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": 0.4531184257371266,
+  "epoch": 0.5178496294138589,
   "eval_steps": 5,
-  "global_step": 3500,
+  "global_step": 4000,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -35723,6 +35723,5106 @@
       "eval_samples_per_second": 3.072,
       "eval_steps_per_second": 0.209,
       "step": 3500
+    },
+    {
+      "epoch": 0.45324788814448,
+      "grad_norm": 4.037712625281316,
+      "learning_rate": 8.321574729247963e-07,
+      "loss": 3.0259,
+      "step": 3501
+    },
+    {
+      "epoch": 0.45337735055183354,
+      "grad_norm": 2.7086904222357995,
+      "learning_rate": 8.318909429090025e-07,
+      "loss": 2.6284,
+      "step": 3502
+    },
+    {
+      "epoch": 0.453506812959187,
+      "grad_norm": 2.138947189395762,
+      "learning_rate": 8.316243706315399e-07,
+      "loss": 2.6094,
+      "step": 3503
+    },
+    {
+      "epoch": 0.45363627536654044,
+      "grad_norm": 2.052154068345744,
+      "learning_rate": 8.313577561468616e-07,
+      "loss": 2.7842,
+      "step": 3504
+    },
+    {
+      "epoch": 0.4537657377738939,
+      "grad_norm": 1.5251842087983352,
+      "learning_rate": 8.31091099509429e-07,
+      "loss": 2.4153,
+      "step": 3505
+    },
+    {
+      "epoch": 0.4537657377738939,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.589888095855713,
+      "eval_runtime": 13.6086,
+      "eval_samples_per_second": 3.233,
+      "eval_steps_per_second": 0.22,
+      "step": 3505
+    },
+    {
+      "epoch": 0.45389520018124735,
+      "grad_norm": 1.8900255833797497,
+      "learning_rate": 8.30824400773712e-07,
+      "loss": 2.7739,
+      "step": 3506
+    },
+    {
+      "epoch": 0.45402466258860086,
+      "grad_norm": 1.3044646331954206,
+      "learning_rate": 8.305576599941893e-07,
+      "loss": 2.384,
+      "step": 3507
+    },
+    {
+      "epoch": 0.4541541249959543,
+      "grad_norm": 3.646796258599669,
+      "learning_rate": 8.302908772253479e-07,
+      "loss": 2.5906,
+      "step": 3508
+    },
+    {
+      "epoch": 0.45428358740330776,
+      "grad_norm": 2.5083107607117356,
+      "learning_rate": 8.300240525216839e-07,
+      "loss": 2.3652,
+      "step": 3509
+    },
+    {
+      "epoch": 0.4544130498106612,
+      "grad_norm": 1.7266092241583402,
+      "learning_rate": 8.297571859377017e-07,
+      "loss": 2.4023,
+      "step": 3510
+    },
+    {
+      "epoch": 0.4544130498106612,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.592041015625,
+      "eval_runtime": 14.584,
+      "eval_samples_per_second": 3.017,
+      "eval_steps_per_second": 0.206,
+      "step": 3510
+    },
+    {
+      "epoch": 0.45454251221801467,
+      "grad_norm": 1.755901879661211,
+      "learning_rate": 8.294902775279139e-07,
+      "loss": 2.5237,
+      "step": 3511
+    },
+    {
+      "epoch": 0.4546719746253682,
+      "grad_norm": 4.733184420278606,
+      "learning_rate": 8.29223327346842e-07,
+      "loss": 2.9595,
+      "step": 3512
+    },
+    {
+      "epoch": 0.45480143703272163,
+      "grad_norm": 2.0147043517710794,
+      "learning_rate": 8.289563354490161e-07,
+      "loss": 2.5232,
+      "step": 3513
+    },
+    {
+      "epoch": 0.4549308994400751,
+      "grad_norm": 2.9454545173513087,
+      "learning_rate": 8.286893018889747e-07,
+      "loss": 2.6023,
+      "step": 3514
+    },
+    {
+      "epoch": 0.45506036184742854,
+      "grad_norm": 1.9390849326008988,
+      "learning_rate": 8.284222267212648e-07,
+      "loss": 2.3904,
+      "step": 3515
+    },
+    {
+      "epoch": 0.45506036184742854,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5878239870071411,
+      "eval_runtime": 14.4249,
+      "eval_samples_per_second": 3.05,
+      "eval_steps_per_second": 0.208,
+      "step": 3515
+    },
+    {
+      "epoch": 0.455189824254782,
+      "grad_norm": 2.407469520826565,
+      "learning_rate": 8.28155110000442e-07,
+      "loss": 2.4973,
+      "step": 3516
+    },
+    {
+      "epoch": 0.4553192866621355,
+      "grad_norm": 2.9037194799249355,
+      "learning_rate": 8.278879517810701e-07,
+      "loss": 2.7073,
+      "step": 3517
+    },
+    {
+      "epoch": 0.45544874906948896,
+      "grad_norm": 2.0242189250031073,
+      "learning_rate": 8.276207521177215e-07,
+      "loss": 2.5239,
+      "step": 3518
+    },
+    {
+      "epoch": 0.4555782114768424,
+      "grad_norm": 1.620279429623691,
+      "learning_rate": 8.273535110649777e-07,
+      "loss": 2.4004,
+      "step": 3519
+    },
+    {
+      "epoch": 0.45570767388419586,
+      "grad_norm": 1.4933697348122157,
+      "learning_rate": 8.270862286774277e-07,
+      "loss": 2.624,
+      "step": 3520
+    },
+    {
+      "epoch": 0.45570767388419586,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.592729091644287,
+      "eval_runtime": 13.7174,
+      "eval_samples_per_second": 3.208,
+      "eval_steps_per_second": 0.219,
+      "step": 3520
+    },
+    {
+      "epoch": 0.4558371362915493,
+      "grad_norm": 2.2558784021244076,
+      "learning_rate": 8.268189050096692e-07,
+      "loss": 2.3595,
+      "step": 3521
+    },
+    {
+      "epoch": 0.4559665986989028,
+      "grad_norm": 1.8697621736823407,
+      "learning_rate": 8.265515401163088e-07,
+      "loss": 2.5815,
+      "step": 3522
+    },
+    {
+      "epoch": 0.4560960611062563,
+      "grad_norm": 2.6995572062800197,
+      "learning_rate": 8.26284134051961e-07,
+      "loss": 2.9189,
+      "step": 3523
+    },
+    {
+      "epoch": 0.45622552351360973,
+      "grad_norm": 1.6164022513948326,
+      "learning_rate": 8.260166868712492e-07,
+      "loss": 2.126,
+      "step": 3524
+    },
+    {
+      "epoch": 0.4563549859209632,
+      "grad_norm": 1.6266596175980144,
+      "learning_rate": 8.257491986288045e-07,
+      "loss": 2.6719,
+      "step": 3525
+    },
+    {
+      "epoch": 0.4563549859209632,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5919743776321411,
+      "eval_runtime": 13.5262,
+      "eval_samples_per_second": 3.253,
+      "eval_steps_per_second": 0.222,
+      "step": 3525
+    },
+    {
+      "epoch": 0.45648444832831664,
+      "grad_norm": 2.2481635107459734,
+      "learning_rate": 8.254816693792669e-07,
+      "loss": 2.6448,
+      "step": 3526
+    },
+    {
+      "epoch": 0.45661391073567015,
+      "grad_norm": 1.264468241384177,
+      "learning_rate": 8.252140991772849e-07,
+      "loss": 2.4055,
+      "step": 3527
+    },
+    {
+      "epoch": 0.4567433731430236,
+      "grad_norm": 1.3532541377530152,
+      "learning_rate": 8.249464880775148e-07,
+      "loss": 2.3101,
+      "step": 3528
+    },
+    {
+      "epoch": 0.45687283555037705,
+      "grad_norm": 2.644552174923601,
+      "learning_rate": 8.246788361346219e-07,
+      "loss": 2.5945,
+      "step": 3529
+    },
+    {
+      "epoch": 0.4570022979577305,
+      "grad_norm": 1.5353137928902914,
+      "learning_rate": 8.244111434032795e-07,
+      "loss": 2.6028,
+      "step": 3530
+    },
+    {
+      "epoch": 0.4570022979577305,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5955921411514282,
+      "eval_runtime": 14.0368,
+      "eval_samples_per_second": 3.135,
+      "eval_steps_per_second": 0.214,
+      "step": 3530
+    },
+    {
+      "epoch": 0.45713176036508396,
+      "grad_norm": 1.817363881152656,
+      "learning_rate": 8.24143409938169e-07,
+      "loss": 2.5415,
+      "step": 3531
+    },
+    {
+      "epoch": 0.45726122277243747,
+      "grad_norm": 1.6082781305740184,
+      "learning_rate": 8.238756357939807e-07,
+      "loss": 2.4641,
+      "step": 3532
+    },
+    {
+      "epoch": 0.4573906851797909,
+      "grad_norm": 1.235996569190046,
+      "learning_rate": 8.236078210254128e-07,
+      "loss": 2.353,
+      "step": 3533
+    },
+    {
+      "epoch": 0.4575201475871444,
+      "grad_norm": 1.7162396031100193,
+      "learning_rate": 8.233399656871718e-07,
+      "loss": 2.6216,
+      "step": 3534
+    },
+    {
+      "epoch": 0.45764960999449783,
+      "grad_norm": 2.8453265415094204,
+      "learning_rate": 8.230720698339726e-07,
+      "loss": 2.4707,
+      "step": 3535
+    },
+    {
+      "epoch": 0.45764960999449783,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.593461513519287,
+      "eval_runtime": 14.0636,
+      "eval_samples_per_second": 3.129,
+      "eval_steps_per_second": 0.213,
+      "step": 3535
+    },
+    {
+      "epoch": 0.45777907240185134,
+      "grad_norm": 2.378157097105098,
+      "learning_rate": 8.228041335205383e-07,
+      "loss": 2.3474,
+      "step": 3536
+    },
+    {
+      "epoch": 0.4579085348092048,
+      "grad_norm": 3.043788773503862,
+      "learning_rate": 8.225361568016006e-07,
+      "loss": 2.6514,
+      "step": 3537
+    },
+    {
+      "epoch": 0.45803799721655825,
+      "grad_norm": 2.0298226832508086,
+      "learning_rate": 8.22268139731899e-07,
+      "loss": 2.4148,
+      "step": 3538
+    },
+    {
+      "epoch": 0.4581674596239117,
+      "grad_norm": 1.951288476113402,
+      "learning_rate": 8.220000823661814e-07,
+      "loss": 2.2568,
+      "step": 3539
+    },
+    {
+      "epoch": 0.45829692203126515,
+      "grad_norm": 1.7093127850108965,
+      "learning_rate": 8.217319847592039e-07,
+      "loss": 2.4971,
+      "step": 3540
+    },
+    {
+      "epoch": 0.45829692203126515,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5949040651321411,
+      "eval_runtime": 13.3483,
+      "eval_samples_per_second": 3.296,
+      "eval_steps_per_second": 0.225,
+      "step": 3540
+    },
+    {
+      "epoch": 0.45842638443861866,
+      "grad_norm": 2.840784446007513,
+      "learning_rate": 8.21463846965731e-07,
+      "loss": 2.9341,
+      "step": 3541
+    },
+    {
+      "epoch": 0.4585558468459721,
+      "grad_norm": 1.5335644139393236,
+      "learning_rate": 8.211956690405351e-07,
+      "loss": 2.5134,
+      "step": 3542
+    },
+    {
+      "epoch": 0.45868530925332557,
+      "grad_norm": 1.3739263786173637,
+      "learning_rate": 8.209274510383972e-07,
+      "loss": 2.4514,
+      "step": 3543
+    },
+    {
+      "epoch": 0.458814771660679,
+      "grad_norm": 1.2930812102965403,
+      "learning_rate": 8.206591930141061e-07,
+      "loss": 2.3064,
+      "step": 3544
+    },
+    {
+      "epoch": 0.4589442340680325,
+      "grad_norm": 1.9304972048455111,
+      "learning_rate": 8.20390895022459e-07,
+      "loss": 2.6584,
+      "step": 3545
+    },
+    {
+      "epoch": 0.4589442340680325,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5948818922042847,
+      "eval_runtime": 13.8604,
+      "eval_samples_per_second": 3.175,
+      "eval_steps_per_second": 0.216,
+      "step": 3545
+    },
+    {
+      "epoch": 0.459073696475386,
+      "grad_norm": 1.7601059949576592,
+      "learning_rate": 8.201225571182613e-07,
+      "loss": 2.3164,
+      "step": 3546
+    },
+    {
+      "epoch": 0.45920315888273944,
+      "grad_norm": 2.182314595338066,
+      "learning_rate": 8.198541793563262e-07,
+      "loss": 2.6887,
+      "step": 3547
+    },
+    {
+      "epoch": 0.4593326212900929,
+      "grad_norm": 1.9765973941016117,
+      "learning_rate": 8.195857617914756e-07,
+      "loss": 2.3545,
+      "step": 3548
+    },
+    {
+      "epoch": 0.45946208369744634,
+      "grad_norm": 1.808539425278726,
+      "learning_rate": 8.193173044785389e-07,
+      "loss": 2.5796,
+      "step": 3549
+    },
+    {
+      "epoch": 0.4595915461047998,
+      "grad_norm": 2.3527738038771657,
+      "learning_rate": 8.190488074723543e-07,
+      "loss": 2.562,
+      "step": 3550
+    },
+    {
+      "epoch": 0.4595915461047998,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5921741724014282,
+      "eval_runtime": 14.4279,
+      "eval_samples_per_second": 3.05,
+      "eval_steps_per_second": 0.208,
+      "step": 3550
+    },
+    {
+      "epoch": 0.4597210085121533,
+      "grad_norm": 2.538172015425518,
+      "learning_rate": 8.187802708277676e-07,
+      "loss": 2.6057,
+      "step": 3551
+    },
+    {
+      "epoch": 0.45985047091950676,
+      "grad_norm": 1.458300915715859,
+      "learning_rate": 8.185116945996329e-07,
+      "loss": 2.4456,
+      "step": 3552
+    },
+    {
+      "epoch": 0.4599799333268602,
+      "grad_norm": 1.541523273447448,
+      "learning_rate": 8.182430788428119e-07,
+      "loss": 2.3953,
+      "step": 3553
+    },
+    {
+      "epoch": 0.46010939573421367,
+      "grad_norm": 2.1204613407306554,
+      "learning_rate": 8.179744236121753e-07,
+      "loss": 2.5063,
+      "step": 3554
+    },
+    {
+      "epoch": 0.4602388581415671,
+      "grad_norm": 1.857757041972434,
+      "learning_rate": 8.177057289626015e-07,
+      "loss": 2.5691,
+      "step": 3555
+    },
+    {
+      "epoch": 0.4602388581415671,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5920188426971436,
+      "eval_runtime": 14.1103,
+      "eval_samples_per_second": 3.118,
+      "eval_steps_per_second": 0.213,
+      "step": 3555
+    },
+    {
+      "epoch": 0.46036832054892063,
+      "grad_norm": 1.6251962162355167,
+      "learning_rate": 8.174369949489766e-07,
+      "loss": 2.7029,
+      "step": 3556
+    },
+    {
+      "epoch": 0.4604977829562741,
+      "grad_norm": 1.902531341562306,
+      "learning_rate": 8.171682216261947e-07,
+      "loss": 2.6323,
+      "step": 3557
+    },
+    {
+      "epoch": 0.46062724536362754,
+      "grad_norm": 4.6452138828166145,
+      "learning_rate": 8.168994090491586e-07,
+      "loss": 2.6743,
+      "step": 3558
+    },
+    {
+      "epoch": 0.460756707770981,
+      "grad_norm": 3.2342474410885282,
+      "learning_rate": 8.166305572727786e-07,
+      "loss": 2.6072,
+      "step": 3559
+    },
+    {
+      "epoch": 0.46088617017833444,
+      "grad_norm": 1.6112562974224787,
+      "learning_rate": 8.163616663519731e-07,
+      "loss": 2.5098,
+      "step": 3560
+    },
+    {
+      "epoch": 0.46088617017833444,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5934836864471436,
+      "eval_runtime": 14.1883,
+      "eval_samples_per_second": 3.101,
+      "eval_steps_per_second": 0.211,
+      "step": 3560
+    },
+    {
+      "epoch": 0.46101563258568795,
+      "grad_norm": 4.058965869435389,
+      "learning_rate": 8.160927363416687e-07,
+      "loss": 2.6421,
+      "step": 3561
+    },
+    {
+      "epoch": 0.4611450949930414,
+      "grad_norm": 1.9515517112472651,
+      "learning_rate": 8.158237672967993e-07,
+      "loss": 2.4978,
+      "step": 3562
+    },
+    {
+      "epoch": 0.46127455740039486,
+      "grad_norm": 1.692602156188153,
+      "learning_rate": 8.155547592723077e-07,
+      "loss": 2.505,
+      "step": 3563
+    },
+    {
+      "epoch": 0.4614040198077483,
+      "grad_norm": 2.112893395470927,
+      "learning_rate": 8.152857123231442e-07,
+      "loss": 2.7878,
+      "step": 3564
+    },
+    {
+      "epoch": 0.46153348221510176,
+      "grad_norm": 2.1354769230516566,
+      "learning_rate": 8.15016626504267e-07,
+      "loss": 2.7136,
+      "step": 3565
+    },
+    {
+      "epoch": 0.46153348221510176,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5921076536178589,
+      "eval_runtime": 14.1796,
+      "eval_samples_per_second": 3.103,
+      "eval_steps_per_second": 0.212,
+      "step": 3565
+    },
+    {
+      "epoch": 0.4616629446224553,
+      "grad_norm": 2.462216370965521,
+      "learning_rate": 8.147475018706425e-07,
+      "loss": 2.5752,
+      "step": 3566
+    },
+    {
+      "epoch": 0.4617924070298087,
+      "grad_norm": 3.175714550415163,
+      "learning_rate": 8.144783384772445e-07,
+      "loss": 2.8098,
+      "step": 3567
+    },
+    {
+      "epoch": 0.4619218694371622,
+      "grad_norm": 1.820835124439004,
+      "learning_rate": 8.142091363790555e-07,
+      "loss": 2.4871,
+      "step": 3568
+    },
+    {
+      "epoch": 0.46205133184451563,
+      "grad_norm": 2.433789894150037,
+      "learning_rate": 8.139398956310654e-07,
+      "loss": 2.4468,
+      "step": 3569
+    },
+    {
+      "epoch": 0.4621807942518691,
+      "grad_norm": 3.9808301447254513,
+      "learning_rate": 8.136706162882719e-07,
+      "loss": 2.6421,
+      "step": 3570
+    },
+    {
+      "epoch": 0.4621807942518691,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.589888095855713,
+      "eval_runtime": 13.6632,
+      "eval_samples_per_second": 3.22,
+      "eval_steps_per_second": 0.22,
+      "step": 3570
+    },
+    {
+      "epoch": 0.4623102566592226,
+      "grad_norm": 2.3166104811786936,
+      "learning_rate": 8.134012984056807e-07,
+      "loss": 2.6492,
+      "step": 3571
+    },
+    {
+      "epoch": 0.46243971906657605,
+      "grad_norm": 1.5633254614133703,
+      "learning_rate": 8.131319420383055e-07,
+      "loss": 2.5996,
+      "step": 3572
+    },
+    {
+      "epoch": 0.4625691814739295,
+      "grad_norm": 2.5436967709171507,
+      "learning_rate": 8.12862547241168e-07,
+      "loss": 2.6721,
+      "step": 3573
+    },
+    {
+      "epoch": 0.46269864388128296,
+      "grad_norm": 1.336551219749576,
+      "learning_rate": 8.125931140692976e-07,
+      "loss": 2.2827,
+      "step": 3574
+    },
+    {
+      "epoch": 0.4628281062886364,
+      "grad_norm": 3.6220279251908925,
+      "learning_rate": 8.123236425777311e-07,
+      "loss": 2.6694,
+      "step": 3575
+    },
+    {
+      "epoch": 0.4628281062886364,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5941494703292847,
+      "eval_runtime": 14.3569,
+      "eval_samples_per_second": 3.065,
+      "eval_steps_per_second": 0.209,
+      "step": 3575
+    },
+    {
+      "epoch": 0.4629575686959899,
+      "grad_norm": 1.6404182937055733,
+      "learning_rate": 8.120541328215138e-07,
+      "loss": 2.5371,
+      "step": 3576
+    },
+    {
+      "epoch": 0.46308703110334337,
+      "grad_norm": 1.316217332753309,
+      "learning_rate": 8.117845848556985e-07,
+      "loss": 2.5742,
+      "step": 3577
+    },
+    {
+      "epoch": 0.4632164935106968,
+      "grad_norm": 3.248335397277242,
+      "learning_rate": 8.115149987353457e-07,
+      "loss": 2.6111,
+      "step": 3578
+    },
+    {
+      "epoch": 0.4633459559180503,
+      "grad_norm": 2.858708893883924,
+      "learning_rate": 8.112453745155241e-07,
+      "loss": 2.3638,
+      "step": 3579
+    },
+    {
+      "epoch": 0.4634754183254038,
+      "grad_norm": 1.523999816902224,
+      "learning_rate": 8.109757122513095e-07,
+      "loss": 2.5203,
+      "step": 3580
+    },
+    {
+      "epoch": 0.4634754183254038,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5940829515457153,
+      "eval_runtime": 14.5063,
+      "eval_samples_per_second": 3.033,
+      "eval_steps_per_second": 0.207,
+      "step": 3580
+    },
+    {
+      "epoch": 0.46360488073275724,
+      "grad_norm": 2.0406504523822786,
+      "learning_rate": 8.107060119977862e-07,
+      "loss": 2.4436,
+      "step": 3581
+    },
+    {
+      "epoch": 0.4637343431401107,
+      "grad_norm": 2.431364476884996,
+      "learning_rate": 8.104362738100459e-07,
+      "loss": 2.7263,
+      "step": 3582
+    },
+    {
+      "epoch": 0.46386380554746415,
+      "grad_norm": 2.2838751982511174,
+      "learning_rate": 8.101664977431882e-07,
+      "loss": 2.4674,
+      "step": 3583
+    },
+    {
+      "epoch": 0.4639932679548176,
+      "grad_norm": 1.9316995644600121,
+      "learning_rate": 8.098966838523199e-07,
+      "loss": 2.6709,
+      "step": 3584
+    },
+    {
+      "epoch": 0.4641227303621711,
+      "grad_norm": 1.4050379286807781,
+      "learning_rate": 8.096268321925564e-07,
+      "loss": 2.4502,
+      "step": 3585
+    },
+    {
+      "epoch": 0.4641227303621711,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5940829515457153,
+      "eval_runtime": 13.3208,
+      "eval_samples_per_second": 3.303,
+      "eval_steps_per_second": 0.225,
+      "step": 3585
+    },
+    {
+      "epoch": 0.46425219276952456,
+      "grad_norm": 1.66131694065598,
+      "learning_rate": 8.093569428190201e-07,
+      "loss": 2.3375,
+      "step": 3586
+    },
+    {
+      "epoch": 0.464381655176878,
+      "grad_norm": 1.3022443891589117,
+      "learning_rate": 8.090870157868416e-07,
+      "loss": 2.2363,
+      "step": 3587
+    },
+    {
+      "epoch": 0.46451111758423147,
+      "grad_norm": 1.448573845830862,
+      "learning_rate": 8.08817051151159e-07,
+      "loss": 2.4866,
+      "step": 3588
+    },
+    {
+      "epoch": 0.4646405799915849,
+      "grad_norm": 1.4590393426530284,
+      "learning_rate": 8.085470489671177e-07,
+      "loss": 2.4417,
+      "step": 3589
+    },
+    {
+      "epoch": 0.46477004239893843,
+      "grad_norm": 2.42888655972702,
+      "learning_rate": 8.082770092898711e-07,
+      "loss": 2.3853,
+      "step": 3590
+    },
+    {
+      "epoch": 0.46477004239893843,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5933283567428589,
+      "eval_runtime": 13.7539,
+      "eval_samples_per_second": 3.199,
+      "eval_steps_per_second": 0.218,
+      "step": 3590
+    },
+    {
+      "epoch": 0.4648995048062919,
+      "grad_norm": 1.9350854766805095,
+      "learning_rate": 8.080069321745808e-07,
+      "loss": 2.5454,
+      "step": 3591
+    },
+    {
+      "epoch": 0.46502896721364534,
+      "grad_norm": 1.6349943734600694,
+      "learning_rate": 8.077368176764152e-07,
+      "loss": 2.738,
+      "step": 3592
+    },
+    {
+      "epoch": 0.4651584296209988,
+      "grad_norm": 1.8505707847809654,
+      "learning_rate": 8.074666658505505e-07,
+      "loss": 2.5532,
+      "step": 3593
+    },
+    {
+      "epoch": 0.46528789202835225,
+      "grad_norm": 3.3952495839951364,
+      "learning_rate": 8.07196476752171e-07,
+      "loss": 2.7383,
+      "step": 3594
+    },
+    {
+      "epoch": 0.46541735443570575,
+      "grad_norm": 2.8892647458860887,
+      "learning_rate": 8.069262504364682e-07,
+      "loss": 2.4192,
+      "step": 3595
+    },
+    {
+      "epoch": 0.46541735443570575,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5948597192764282,
+      "eval_runtime": 13.6196,
+      "eval_samples_per_second": 3.231,
+      "eval_steps_per_second": 0.22,
+      "step": 3595
+    },
+    {
+      "epoch": 0.4655468168430592,
+      "grad_norm": 2.133267236864094,
+      "learning_rate": 8.066559869586412e-07,
+      "loss": 2.7998,
+      "step": 3596
+    },
+    {
+      "epoch": 0.46567627925041266,
+      "grad_norm": 3.3718440265536542,
+      "learning_rate": 8.063856863738969e-07,
+      "loss": 2.6304,
+      "step": 3597
+    },
+    {
+      "epoch": 0.4658057416577661,
+      "grad_norm": 1.9952924172187647,
+      "learning_rate": 8.061153487374497e-07,
+      "loss": 2.7131,
+      "step": 3598
+    },
+    {
+      "epoch": 0.46593520406511957,
+      "grad_norm": 1.5456239466364916,
+      "learning_rate": 8.058449741045214e-07,
+      "loss": 2.4812,
+      "step": 3599
+    },
+    {
+      "epoch": 0.4660646664724731,
+      "grad_norm": 2.583115023447738,
+      "learning_rate": 8.055745625303415e-07,
+      "loss": 2.4417,
+      "step": 3600
+    },
+    {
+      "epoch": 0.4660646664724731,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5906428098678589,
+      "eval_runtime": 14.186,
+      "eval_samples_per_second": 3.102,
+      "eval_steps_per_second": 0.211,
+      "step": 3600
+    },
+    {
+      "epoch": 0.46619412887982653,
+      "grad_norm": 1.1488507797715302,
+      "learning_rate": 8.053041140701473e-07,
+      "loss": 2.2869,
+      "step": 3601
+    },
+    {
+      "epoch": 0.46632359128718,
+      "grad_norm": 1.7556191640489143,
+      "learning_rate": 8.050336287791832e-07,
+      "loss": 2.3716,
+      "step": 3602
+    },
+    {
+      "epoch": 0.46645305369453344,
+      "grad_norm": 1.8969412618488941,
+      "learning_rate": 8.047631067127012e-07,
+      "loss": 2.5596,
+      "step": 3603
+    },
+    {
+      "epoch": 0.4665825161018869,
+      "grad_norm": 1.5199006891523428,
+      "learning_rate": 8.044925479259609e-07,
+      "loss": 2.4817,
+      "step": 3604
+    },
+    {
+      "epoch": 0.4667119785092404,
+      "grad_norm": 1.9122988107655003,
+      "learning_rate": 8.042219524742298e-07,
+      "loss": 2.4094,
+      "step": 3605
+    },
+    {
+      "epoch": 0.4667119785092404,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.591264247894287,
+      "eval_runtime": 14.6212,
+      "eval_samples_per_second": 3.009,
+      "eval_steps_per_second": 0.205,
+      "step": 3605
+    },
+    {
+      "epoch": 0.46684144091659385,
+      "grad_norm": 1.6565579502051928,
+      "learning_rate": 8.039513204127822e-07,
+      "loss": 2.5667,
+      "step": 3606
+    },
+    {
+      "epoch": 0.4669709033239473,
+      "grad_norm": 1.5887212603423024,
+      "learning_rate": 8.036806517969001e-07,
+      "loss": 2.7041,
+      "step": 3607
+    },
+    {
+      "epoch": 0.46710036573130076,
+      "grad_norm": 1.6585018108929768,
+      "learning_rate": 8.034099466818732e-07,
+      "loss": 2.5796,
+      "step": 3608
+    },
+    {
+      "epoch": 0.4672298281386542,
+      "grad_norm": 1.414804864498699,
+      "learning_rate": 8.031392051229984e-07,
+      "loss": 2.4985,
+      "step": 3609
+    },
+    {
+      "epoch": 0.4673592905460077,
+      "grad_norm": 2.34889392652913,
+      "learning_rate": 8.028684271755806e-07,
+      "loss": 2.2983,
+      "step": 3610
+    },
+    {
+      "epoch": 0.4673592905460077,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5906649827957153,
+      "eval_runtime": 13.5233,
+      "eval_samples_per_second": 3.254,
+      "eval_steps_per_second": 0.222,
+      "step": 3610
+    },
+    {
+      "epoch": 0.4674887529533612,
+      "grad_norm": 1.8952544735961896,
+      "learning_rate": 8.02597612894931e-07,
+      "loss": 2.3274,
+      "step": 3611
+    },
+    {
+      "epoch": 0.46761821536071463,
+      "grad_norm": 2.328226795461395,
+      "learning_rate": 8.023267623363693e-07,
+      "loss": 2.697,
+      "step": 3612
+    },
+    {
+      "epoch": 0.4677476777680681,
+      "grad_norm": 1.575542729139767,
+      "learning_rate": 8.020558755552221e-07,
+      "loss": 2.6398,
+      "step": 3613
+    },
+    {
+      "epoch": 0.46787714017542154,
+      "grad_norm": 2.1777157593498475,
+      "learning_rate": 8.017849526068235e-07,
+      "loss": 2.4082,
+      "step": 3614
+    },
+    {
+      "epoch": 0.46800660258277504,
+      "grad_norm": 2.2741259184806144,
+      "learning_rate": 8.015139935465152e-07,
+      "loss": 2.4932,
+      "step": 3615
+    },
+    {
+      "epoch": 0.46800660258277504,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.591264247894287,
+      "eval_runtime": 13.5983,
+      "eval_samples_per_second": 3.236,
+      "eval_steps_per_second": 0.221,
+      "step": 3615
+    },
+    {
+      "epoch": 0.4681360649901285,
+      "grad_norm": 2.939417811929448,
+      "learning_rate": 8.01242998429646e-07,
+      "loss": 2.3882,
+      "step": 3616
+    },
+    {
+      "epoch": 0.46826552739748195,
+      "grad_norm": 4.701175417229919,
+      "learning_rate": 8.00971967311572e-07,
+      "loss": 2.8022,
+      "step": 3617
+    },
+    {
+      "epoch": 0.4683949898048354,
+      "grad_norm": 2.5364959994065366,
+      "learning_rate": 8.007009002476568e-07,
+      "loss": 2.5098,
+      "step": 3618
+    },
+    {
+      "epoch": 0.4685244522121889,
+      "grad_norm": 3.5895201998707016,
+      "learning_rate": 8.004297972932716e-07,
+      "loss": 2.5964,
+      "step": 3619
+    },
+    {
+      "epoch": 0.46865391461954237,
+      "grad_norm": 1.3531610229149031,
+      "learning_rate": 8.001586585037945e-07,
+      "loss": 2.3726,
+      "step": 3620
+    },
+    {
+      "epoch": 0.46865391461954237,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.594038486480713,
+      "eval_runtime": 13.5648,
+      "eval_samples_per_second": 3.244,
+      "eval_steps_per_second": 0.221,
+      "step": 3620
+    },
+    {
+      "epoch": 0.4687833770268958,
+      "grad_norm": 1.217856566210872,
+      "learning_rate": 7.99887483934611e-07,
+      "loss": 2.4648,
+      "step": 3621
+    },
+    {
+      "epoch": 0.4689128394342493,
+      "grad_norm": 2.4974573885477716,
+      "learning_rate": 7.99616273641114e-07,
+      "loss": 2.5525,
+      "step": 3622
+    },
+    {
+      "epoch": 0.4690423018416027,
+      "grad_norm": 1.5620724118722353,
+      "learning_rate": 7.993450276787041e-07,
+      "loss": 2.2842,
+      "step": 3623
+    },
+    {
+      "epoch": 0.46917176424895624,
+      "grad_norm": 1.8211379853260594,
+      "learning_rate": 7.990737461027882e-07,
+      "loss": 2.6074,
+      "step": 3624
+    },
+    {
+      "epoch": 0.4693012266563097,
+      "grad_norm": 1.8456022956077363,
+      "learning_rate": 7.988024289687815e-07,
+      "loss": 2.3943,
+      "step": 3625
+    },
+    {
+      "epoch": 0.4693012266563097,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5940607786178589,
+      "eval_runtime": 14.5194,
+      "eval_samples_per_second": 3.03,
+      "eval_steps_per_second": 0.207,
+      "step": 3625
+    },
+    {
+      "epoch": 0.46943068906366314,
+      "grad_norm": 1.135654349596474,
+      "learning_rate": 7.985310763321057e-07,
+      "loss": 2.2441,
+      "step": 3626
+    },
+    {
+      "epoch": 0.4695601514710166,
+      "grad_norm": 1.8172491529857226,
+      "learning_rate": 7.982596882481905e-07,
+      "loss": 2.5605,
+      "step": 3627
+    },
+    {
+      "epoch": 0.46968961387837005,
+      "grad_norm": 1.547347121707995,
+      "learning_rate": 7.979882647724719e-07,
+      "loss": 2.5198,
+      "step": 3628
+    },
+    {
+      "epoch": 0.46981907628572356,
+      "grad_norm": 2.1268555572499714,
+      "learning_rate": 7.97716805960394e-07,
+      "loss": 2.4683,
+      "step": 3629
+    },
+    {
+      "epoch": 0.469948538693077,
+      "grad_norm": 3.692927730484567,
+      "learning_rate": 7.974453118674077e-07,
+      "loss": 2.6782,
+      "step": 3630
+    },
+    {
+      "epoch": 0.469948538693077,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5940607786178589,
+      "eval_runtime": 14.06,
+      "eval_samples_per_second": 3.129,
+      "eval_steps_per_second": 0.213,
+      "step": 3630
+    },
+    {
+      "epoch": 0.47007800110043046,
+      "grad_norm": 1.486069011000907,
+      "learning_rate": 7.971737825489712e-07,
+      "loss": 2.4634,
+      "step": 3631
+    },
+    {
+      "epoch": 0.4702074635077839,
+      "grad_norm": 4.336146760444173,
+      "learning_rate": 7.969022180605496e-07,
+      "loss": 2.748,
+      "step": 3632
+    },
+    {
+      "epoch": 0.47033692591513737,
+      "grad_norm": 1.6069502772995092,
+      "learning_rate": 7.96630618457616e-07,
+      "loss": 2.4838,
+      "step": 3633
+    },
+    {
+      "epoch": 0.4704663883224909,
+      "grad_norm": 1.5285935016895567,
+      "learning_rate": 7.963589837956495e-07,
+      "loss": 2.2715,
+      "step": 3634
+    },
+    {
+      "epoch": 0.47059585072984433,
+      "grad_norm": 3.724211678255899,
+      "learning_rate": 7.960873141301374e-07,
+      "loss": 2.5835,
+      "step": 3635
+    },
+    {
+      "epoch": 0.47059585072984433,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.591264247894287,
+      "eval_runtime": 14.1703,
+      "eval_samples_per_second": 3.105,
+      "eval_steps_per_second": 0.212,
+      "step": 3635
+    },
+    {
+      "epoch": 0.4707253131371978,
+      "grad_norm": 1.8627103727513723,
+      "learning_rate": 7.958156095165736e-07,
+      "loss": 2.5374,
+      "step": 3636
+    },
+    {
+      "epoch": 0.47085477554455124,
+      "grad_norm": 1.8171282842768852,
+      "learning_rate": 7.955438700104592e-07,
+      "loss": 2.7075,
+      "step": 3637
+    },
+    {
+      "epoch": 0.4709842379519047,
+      "grad_norm": 1.973031343685012,
+      "learning_rate": 7.952720956673026e-07,
+      "loss": 2.4497,
+      "step": 3638
+    },
+    {
+      "epoch": 0.4711137003592582,
+      "grad_norm": 4.783104249776493,
+      "learning_rate": 7.950002865426192e-07,
+      "loss": 3.0967,
+      "step": 3639
+    },
+    {
+      "epoch": 0.47124316276661166,
+      "grad_norm": 2.9507778313227306,
+      "learning_rate": 7.947284426919316e-07,
+      "loss": 2.6985,
+      "step": 3640
+    },
+    {
+      "epoch": 0.47124316276661166,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5920188426971436,
+      "eval_runtime": 14.0808,
+      "eval_samples_per_second": 3.125,
+      "eval_steps_per_second": 0.213,
+      "step": 3640
+    },
+    {
+      "epoch": 0.4713726251739651,
+      "grad_norm": 2.8097235245936583,
+      "learning_rate": 7.944565641707694e-07,
+      "loss": 2.4431,
+      "step": 3641
+    },
+    {
+      "epoch": 0.47150208758131856,
+      "grad_norm": 2.1359216751180576,
+      "learning_rate": 7.941846510346691e-07,
+      "loss": 2.3904,
+      "step": 3642
+    },
+    {
+      "epoch": 0.471631549988672,
+      "grad_norm": 2.3195741311969353,
+      "learning_rate": 7.939127033391747e-07,
+      "loss": 2.5706,
+      "step": 3643
+    },
+    {
+      "epoch": 0.4717610123960255,
+      "grad_norm": 1.4403763591503504,
+      "learning_rate": 7.936407211398368e-07,
+      "loss": 2.3937,
+      "step": 3644
+    },
+    {
+      "epoch": 0.471890474803379,
+      "grad_norm": 1.6442525806178,
+      "learning_rate": 7.933687044922137e-07,
+      "loss": 2.4302,
+      "step": 3645
+    },
+    {
+      "epoch": 0.471890474803379,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.592729091644287,
+      "eval_runtime": 13.7173,
+      "eval_samples_per_second": 3.208,
+      "eval_steps_per_second": 0.219,
+      "step": 3645
+    },
+    {
+      "epoch": 0.47201993721073243,
+      "grad_norm": 2.425007683694663,
+      "learning_rate": 7.930966534518697e-07,
+      "loss": 2.3091,
+      "step": 3646
+    },
+    {
+      "epoch": 0.4721493996180859,
+      "grad_norm": 2.0655610388583048,
+      "learning_rate": 7.928245680743774e-07,
+      "loss": 2.4888,
+      "step": 3647
+    },
+    {
+      "epoch": 0.47227886202543934,
+      "grad_norm": 1.5701088927262847,
+      "learning_rate": 7.925524484153151e-07,
+      "loss": 2.4421,
+      "step": 3648
+    },
+    {
+      "epoch": 0.47240832443279285,
+      "grad_norm": 2.3285068063372854,
+      "learning_rate": 7.922802945302692e-07,
+      "loss": 2.4827,
+      "step": 3649
+    },
+    {
+      "epoch": 0.4725377868401463,
+      "grad_norm": 3.840222524902766,
+      "learning_rate": 7.920081064748326e-07,
+      "loss": 2.573,
+      "step": 3650
+    },
+    {
+      "epoch": 0.4725377868401463,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5927512645721436,
+      "eval_runtime": 13.0948,
+      "eval_samples_per_second": 3.36,
+      "eval_steps_per_second": 0.229,
+      "step": 3650
+    },
+    {
+      "epoch": 0.47266724924749975,
+      "grad_norm": 4.430643802202797,
+      "learning_rate": 7.917358843046049e-07,
+      "loss": 2.8379,
+      "step": 3651
+    },
+    {
+      "epoch": 0.4727967116548532,
+      "grad_norm": 1.7401327511814533,
+      "learning_rate": 7.914636280751936e-07,
+      "loss": 2.3899,
+      "step": 3652
+    },
+    {
+      "epoch": 0.47292617406220666,
+      "grad_norm": 3.4415834060918784,
+      "learning_rate": 7.91191337842212e-07,
+      "loss": 2.4988,
+      "step": 3653
+    },
+    {
+      "epoch": 0.47305563646956017,
+      "grad_norm": 1.608745771038558,
+      "learning_rate": 7.909190136612811e-07,
+      "loss": 2.5613,
+      "step": 3654
+    },
+    {
+      "epoch": 0.4731850988769136,
+      "grad_norm": 3.193515669381869,
+      "learning_rate": 7.906466555880287e-07,
+      "loss": 2.6494,
+      "step": 3655
+    },
+    {
+      "epoch": 0.4731850988769136,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5877796411514282,
+      "eval_runtime": 14.9692,
+      "eval_samples_per_second": 2.939,
+      "eval_steps_per_second": 0.2,
+      "step": 3655
+    },
+    {
+      "epoch": 0.4733145612842671,
+      "grad_norm": 1.9220860961258883,
+      "learning_rate": 7.903742636780895e-07,
+      "loss": 2.4949,
+      "step": 3656
+    },
+    {
+      "epoch": 0.47344402369162053,
+      "grad_norm": 2.0942082472127814,
+      "learning_rate": 7.901018379871047e-07,
+      "loss": 2.3389,
+      "step": 3657
+    },
+    {
+      "epoch": 0.473573486098974,
+      "grad_norm": 3.3252440437042887,
+      "learning_rate": 7.898293785707231e-07,
+      "loss": 2.8555,
+      "step": 3658
+    },
+    {
+      "epoch": 0.4737029485063275,
+      "grad_norm": 1.6969295377177454,
+      "learning_rate": 7.895568854846001e-07,
+      "loss": 2.656,
+      "step": 3659
+    },
+    {
+      "epoch": 0.47383241091368095,
+      "grad_norm": 1.1628809286153323,
+      "learning_rate": 7.89284358784398e-07,
+      "loss": 2.2322,
+      "step": 3660
+    },
+    {
+      "epoch": 0.47383241091368095,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5892223119735718,
+      "eval_runtime": 14.5752,
+      "eval_samples_per_second": 3.019,
+      "eval_steps_per_second": 0.206,
+      "step": 3660
+    },
+    {
+      "epoch": 0.4739618733210344,
+      "grad_norm": 1.4014953449138379,
+      "learning_rate": 7.890117985257853e-07,
+      "loss": 2.4456,
+      "step": 3661
+    },
+    {
+      "epoch": 0.47409133572838785,
+      "grad_norm": 1.7908022122096063,
+      "learning_rate": 7.887392047644385e-07,
+      "loss": 2.7241,
+      "step": 3662
+    },
+    {
+      "epoch": 0.47422079813574136,
+      "grad_norm": 2.45014592304961,
+      "learning_rate": 7.884665775560404e-07,
+      "loss": 2.3882,
+      "step": 3663
+    },
+    {
+      "epoch": 0.4743502605430948,
+      "grad_norm": 1.6672833725737877,
+      "learning_rate": 7.881939169562805e-07,
+      "loss": 2.4238,
+      "step": 3664
+    },
+    {
+      "epoch": 0.47447972295044827,
+      "grad_norm": 4.58009051603523,
+      "learning_rate": 7.879212230208552e-07,
+      "loss": 2.7588,
+      "step": 3665
+    },
+    {
+      "epoch": 0.47447972295044827,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5863147974014282,
+      "eval_runtime": 14.3211,
+      "eval_samples_per_second": 3.072,
+      "eval_steps_per_second": 0.209,
+      "step": 3665
+    },
+    {
+      "epoch": 0.4746091853578017,
+      "grad_norm": 1.2909944526768258,
+      "learning_rate": 7.876484958054678e-07,
+      "loss": 2.4045,
+      "step": 3666
+    },
+    {
+      "epoch": 0.4747386477651552,
+      "grad_norm": 1.9253731831295025,
+      "learning_rate": 7.873757353658286e-07,
+      "loss": 2.437,
+      "step": 3667
+    },
+    {
+      "epoch": 0.4748681101725087,
+      "grad_norm": 1.493828060078587,
+      "learning_rate": 7.871029417576541e-07,
+      "loss": 2.4282,
+      "step": 3668
+    },
+    {
+      "epoch": 0.47499757257986214,
+      "grad_norm": 1.7833120546842636,
+      "learning_rate": 7.868301150366682e-07,
+      "loss": 2.4453,
+      "step": 3669
+    },
+    {
+      "epoch": 0.4751270349872156,
+      "grad_norm": 1.5134766080861666,
+      "learning_rate": 7.865572552586009e-07,
+      "loss": 2.4148,
+      "step": 3670
+    },
+    {
+      "epoch": 0.4751270349872156,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5864036083221436,
+      "eval_runtime": 14.7312,
+      "eval_samples_per_second": 2.987,
+      "eval_steps_per_second": 0.204,
+      "step": 3670
+    },
+    {
+      "epoch": 0.47525649739456904,
+      "grad_norm": 1.6231341720498271,
+      "learning_rate": 7.862843624791896e-07,
+      "loss": 2.5667,
+      "step": 3671
+    },
+    {
+      "epoch": 0.4753859598019225,
+      "grad_norm": 2.340347213473683,
+      "learning_rate": 7.860114367541785e-07,
+      "loss": 2.4146,
+      "step": 3672
+    },
+    {
+      "epoch": 0.475515422209276,
+      "grad_norm": 1.719479769558792,
+      "learning_rate": 7.857384781393177e-07,
+      "loss": 2.4636,
+      "step": 3673
+    },
+    {
+      "epoch": 0.47564488461662946,
+      "grad_norm": 1.2499263146873119,
+      "learning_rate": 7.85465486690365e-07,
+      "loss": 2.2085,
+      "step": 3674
+    },
+    {
+      "epoch": 0.4757743470239829,
+      "grad_norm": 2.618130554292385,
+      "learning_rate": 7.851924624630836e-07,
+      "loss": 2.7778,
+      "step": 3675
+    },
+    {
+      "epoch": 0.4757743470239829,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5892223119735718,
+      "eval_runtime": 13.7603,
+      "eval_samples_per_second": 3.198,
+      "eval_steps_per_second": 0.218,
+      "step": 3675
+    },
+    {
+      "epoch": 0.47590380943133637,
+      "grad_norm": 2.6395850335243884,
+      "learning_rate": 7.84919405513245e-07,
+      "loss": 2.47,
+      "step": 3676
+    },
+    {
+      "epoch": 0.4760332718386898,
+      "grad_norm": 3.095993944547029,
+      "learning_rate": 7.846463158966267e-07,
+      "loss": 2.6088,
+      "step": 3677
+    },
+    {
+      "epoch": 0.47616273424604333,
+      "grad_norm": 1.3857736384241748,
+      "learning_rate": 7.843731936690123e-07,
+      "loss": 2.4314,
+      "step": 3678
+    },
+    {
+      "epoch": 0.4762921966533968,
+      "grad_norm": 1.2359118674120453,
+      "learning_rate": 7.841000388861927e-07,
+      "loss": 2.3379,
+      "step": 3679
+    },
+    {
+      "epoch": 0.47642165906075024,
+      "grad_norm": 1.5659393378335085,
+      "learning_rate": 7.838268516039653e-07,
+      "loss": 2.4626,
+      "step": 3680
+    },
+    {
+      "epoch": 0.47642165906075024,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.584916591644287,
+      "eval_runtime": 14.2106,
+      "eval_samples_per_second": 3.096,
+      "eval_steps_per_second": 0.211,
+      "step": 3680
+    },
+    {
+      "epoch": 0.4765511214681037,
+      "grad_norm": 1.6248855815364758,
+      "learning_rate": 7.835536318781342e-07,
+      "loss": 2.3647,
+      "step": 3681
+    },
+    {
+      "epoch": 0.47668058387545714,
+      "grad_norm": 2.3366036662156886,
+      "learning_rate": 7.832803797645101e-07,
+      "loss": 2.623,
+      "step": 3682
+    },
+    {
+      "epoch": 0.47681004628281065,
+      "grad_norm": 1.9540910015132569,
+      "learning_rate": 7.830070953189105e-07,
+      "loss": 2.2856,
+      "step": 3683
+    },
+    {
+      "epoch": 0.4769395086901641,
+      "grad_norm": 1.7453645913102007,
+      "learning_rate": 7.827337785971586e-07,
+      "loss": 2.4363,
+      "step": 3684
+    },
+    {
+      "epoch": 0.47706897109751756,
+      "grad_norm": 3.870806146670109,
+      "learning_rate": 7.824604296550854e-07,
+      "loss": 2.8291,
+      "step": 3685
+    },
+    {
+      "epoch": 0.47706897109751756,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5828746557235718,
+      "eval_runtime": 14.9179,
+      "eval_samples_per_second": 2.949,
+      "eval_steps_per_second": 0.201,
+      "step": 3685
+    },
+    {
+      "epoch": 0.477198433504871,
+      "grad_norm": 1.824831790233935,
+      "learning_rate": 7.821870485485281e-07,
+      "loss": 2.499,
+      "step": 3686
+    },
+    {
+      "epoch": 0.47732789591222446,
+      "grad_norm": 2.0851171491837874,
+      "learning_rate": 7.819136353333298e-07,
+      "loss": 2.5444,
+      "step": 3687
+    },
+    {
+      "epoch": 0.477457358319578,
+      "grad_norm": 1.2957353852878712,
+      "learning_rate": 7.816401900653414e-07,
+      "loss": 2.3779,
+      "step": 3688
+    },
+    {
+      "epoch": 0.4775868207269314,
+      "grad_norm": 6.37040799261014,
+      "learning_rate": 7.813667128004187e-07,
+      "loss": 2.7166,
+      "step": 3689
+    },
+    {
+      "epoch": 0.4777162831342849,
+      "grad_norm": 2.27888901263236,
+      "learning_rate": 7.810932035944257e-07,
+      "loss": 2.6482,
+      "step": 3690
+    },
+    {
+      "epoch": 0.4777162831342849,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5834294557571411,
+      "eval_runtime": 14.8773,
+      "eval_samples_per_second": 2.958,
+      "eval_steps_per_second": 0.202,
+      "step": 3690
+    },
+    {
+      "epoch": 0.47784574554163833,
+      "grad_norm": 1.7235541416427815,
+      "learning_rate": 7.808196625032323e-07,
+      "loss": 2.4021,
+      "step": 3691
+    },
+    {
+      "epoch": 0.4779752079489918,
+      "grad_norm": 1.5463941959221092,
+      "learning_rate": 7.805460895827143e-07,
+      "loss": 2.6846,
+      "step": 3692
+    },
+    {
+      "epoch": 0.4781046703563453,
+      "grad_norm": 1.354331992013068,
+      "learning_rate": 7.802724848887545e-07,
+      "loss": 2.5376,
+      "step": 3693
+    },
+    {
+      "epoch": 0.47823413276369875,
+      "grad_norm": 2.122552087688215,
+      "learning_rate": 7.799988484772426e-07,
+      "loss": 2.3826,
+      "step": 3694
+    },
+    {
+      "epoch": 0.4783635951710522,
+      "grad_norm": 2.316364150505932,
+      "learning_rate": 7.797251804040742e-07,
+      "loss": 2.6699,
+      "step": 3695
+    },
+    {
+      "epoch": 0.4783635951710522,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5855823755264282,
+      "eval_runtime": 15.2034,
+      "eval_samples_per_second": 2.894,
+      "eval_steps_per_second": 0.197,
+      "step": 3695
+    },
+    {
+      "epoch": 0.47849305757840566,
+      "grad_norm": 2.3547673697725853,
+      "learning_rate": 7.794514807251518e-07,
+      "loss": 2.4785,
+      "step": 3696
+    },
+    {
+      "epoch": 0.4786225199857591,
+      "grad_norm": 2.080325531391821,
+      "learning_rate": 7.791777494963837e-07,
+      "loss": 2.2725,
+      "step": 3697
+    },
+    {
+      "epoch": 0.4787519823931126,
+      "grad_norm": 1.9976893564121538,
+      "learning_rate": 7.78903986773685e-07,
+      "loss": 2.3774,
+      "step": 3698
+    },
+    {
+      "epoch": 0.47888144480046607,
+      "grad_norm": 1.559551048113188,
+      "learning_rate": 7.786301926129778e-07,
+      "loss": 2.5544,
+      "step": 3699
+    },
+    {
+      "epoch": 0.4790109072078195,
+      "grad_norm": 2.3145842669312824,
+      "learning_rate": 7.783563670701897e-07,
+      "loss": 2.6377,
+      "step": 3700
+    },
+    {
+      "epoch": 0.4790109072078195,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5814098119735718,
+      "eval_runtime": 14.2466,
+      "eval_samples_per_second": 3.088,
+      "eval_steps_per_second": 0.211,
+      "step": 3700
+    },
+    {
+      "epoch": 0.479140369615173,
+      "grad_norm": 1.8498585357715334,
+      "learning_rate": 7.780825102012554e-07,
+      "loss": 2.5061,
+      "step": 3701
+    },
+    {
+      "epoch": 0.47926983202252643,
+      "grad_norm": 2.648286277629586,
+      "learning_rate": 7.778086220621153e-07,
+      "loss": 2.3884,
+      "step": 3702
+    },
+    {
+      "epoch": 0.47939929442987994,
+      "grad_norm": 2.170406607629059,
+      "learning_rate": 7.77534702708717e-07,
+      "loss": 2.2874,
+      "step": 3703
+    },
+    {
+      "epoch": 0.4795287568372334,
+      "grad_norm": 1.2229933943224742,
+      "learning_rate": 7.772607521970139e-07,
+      "loss": 2.4102,
+      "step": 3704
+    },
+    {
+      "epoch": 0.47965821924458685,
+      "grad_norm": 2.0640529545408723,
+      "learning_rate": 7.769867705829662e-07,
+      "loss": 2.4568,
+      "step": 3705
+    },
+    {
+      "epoch": 0.47965821924458685,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.581498622894287,
+      "eval_runtime": 14.3845,
+      "eval_samples_per_second": 3.059,
+      "eval_steps_per_second": 0.209,
+      "step": 3705
+    },
+    {
+      "epoch": 0.4797876816519403,
+      "grad_norm": 2.153704351820282,
+      "learning_rate": 7.767127579225396e-07,
+      "loss": 2.4873,
+      "step": 3706
+    },
+    {
+      "epoch": 0.4799171440592938,
+      "grad_norm": 2.1574806866970895,
+      "learning_rate": 7.764387142717072e-07,
+      "loss": 2.5564,
+      "step": 3707
+    },
+    {
+      "epoch": 0.48004660646664726,
+      "grad_norm": 2.20367580910148,
+      "learning_rate": 7.76164639686448e-07,
+      "loss": 2.5771,
+      "step": 3708
+    },
+    {
+      "epoch": 0.4801760688740007,
+      "grad_norm": 2.9764950702641344,
+      "learning_rate": 7.758905342227471e-07,
+      "loss": 2.4888,
+      "step": 3709
+    },
+    {
+      "epoch": 0.48030553128135417,
+      "grad_norm": 4.206729882439295,
+      "learning_rate": 7.756163979365963e-07,
+      "loss": 3.125,
+      "step": 3710
+    },
+    {
+      "epoch": 0.48030553128135417,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.58349609375,
+      "eval_runtime": 14.6897,
+      "eval_samples_per_second": 2.995,
+      "eval_steps_per_second": 0.204,
+      "step": 3710
+    },
+    {
+      "epoch": 0.4804349936887076,
+      "grad_norm": 1.954075134769091,
+      "learning_rate": 7.753422308839932e-07,
+      "loss": 2.5083,
+      "step": 3711
+    },
+    {
+      "epoch": 0.48056445609606113,
+      "grad_norm": 2.457147486965655,
+      "learning_rate": 7.750680331209423e-07,
+      "loss": 2.7341,
+      "step": 3712
+    },
+    {
+      "epoch": 0.4806939185034146,
+      "grad_norm": 3.3155815697949573,
+      "learning_rate": 7.747938047034539e-07,
+      "loss": 2.4932,
+      "step": 3713
+    },
+    {
+      "epoch": 0.48082338091076804,
+      "grad_norm": 1.3728561597210085,
+      "learning_rate": 7.745195456875446e-07,
+      "loss": 2.2705,
+      "step": 3714
+    },
+    {
+      "epoch": 0.4809528433181215,
+      "grad_norm": 1.4511225081906338,
+      "learning_rate": 7.742452561292375e-07,
+      "loss": 2.4126,
+      "step": 3715
+    },
+    {
+      "epoch": 0.4809528433181215,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5842063426971436,
+      "eval_runtime": 14.4649,
+      "eval_samples_per_second": 3.042,
+      "eval_steps_per_second": 0.207,
+      "step": 3715
+    },
+    {
+      "epoch": 0.48108230572547495,
+      "grad_norm": 1.6307034616060005,
+      "learning_rate": 7.739709360845618e-07,
+      "loss": 2.5022,
+      "step": 3716
+    },
+    {
+      "epoch": 0.48121176813282845,
+      "grad_norm": 1.9373055476282899,
+      "learning_rate": 7.736965856095529e-07,
+      "loss": 2.7395,
+      "step": 3717
+    },
+    {
+      "epoch": 0.4813412305401819,
+      "grad_norm": 1.901467686955576,
+      "learning_rate": 7.734222047602523e-07,
+      "loss": 2.168,
+      "step": 3718
+    },
+    {
+      "epoch": 0.48147069294753536,
+      "grad_norm": 2.0944980091406924,
+      "learning_rate": 7.731477935927081e-07,
+      "loss": 2.418,
+      "step": 3719
+    },
+    {
+      "epoch": 0.4816001553548888,
+      "grad_norm": 3.293475368402424,
+      "learning_rate": 7.728733521629743e-07,
+      "loss": 2.5547,
+      "step": 3720
+    },
+    {
+      "epoch": 0.4816001553548888,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5814541578292847,
+      "eval_runtime": 14.6659,
+      "eval_samples_per_second": 3.0,
+      "eval_steps_per_second": 0.205,
+      "step": 3720
+    },
+    {
+      "epoch": 0.48172961776224227,
+      "grad_norm": 2.36494113794181,
+      "learning_rate": 7.725988805271109e-07,
+      "loss": 2.5623,
+      "step": 3721
+    },
+    {
+      "epoch": 0.4818590801695958,
+      "grad_norm": 1.3858640979437618,
+      "learning_rate": 7.723243787411847e-07,
+      "loss": 2.4854,
+      "step": 3722
+    },
+    {
+      "epoch": 0.48198854257694923,
+      "grad_norm": 1.7821239943084766,
+      "learning_rate": 7.720498468612679e-07,
+      "loss": 2.4141,
+      "step": 3723
+    },
+    {
+      "epoch": 0.4821180049843027,
+      "grad_norm": 1.6248640942210661,
+      "learning_rate": 7.717752849434395e-07,
+      "loss": 2.584,
+      "step": 3724
+    },
+    {
+      "epoch": 0.48224746739165614,
+      "grad_norm": 1.2430576428639852,
+      "learning_rate": 7.715006930437838e-07,
+      "loss": 2.3594,
+      "step": 3725
+    },
+    {
+      "epoch": 0.48224746739165614,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5827858448028564,
+      "eval_runtime": 16.2434,
+      "eval_samples_per_second": 2.709,
+      "eval_steps_per_second": 0.185,
+      "step": 3725
+    },
+    {
+      "epoch": 0.4823769297990096,
+      "grad_norm": 3.7523010693157937,
+      "learning_rate": 7.712260712183922e-07,
+      "loss": 2.4336,
+      "step": 3726
+    },
+    {
+      "epoch": 0.4825063922063631,
+      "grad_norm": 2.7849366469205035,
+      "learning_rate": 7.709514195233618e-07,
+      "loss": 2.354,
+      "step": 3727
+    },
+    {
+      "epoch": 0.48263585461371655,
+      "grad_norm": 1.8079914434319315,
+      "learning_rate": 7.706767380147955e-07,
+      "loss": 2.3965,
+      "step": 3728
+    },
+    {
+      "epoch": 0.48276531702107,
+      "grad_norm": 1.5272232049006351,
+      "learning_rate": 7.70402026748803e-07,
+      "loss": 2.4004,
+      "step": 3729
+    },
+    {
+      "epoch": 0.48289477942842346,
+      "grad_norm": 1.4507466332466517,
+      "learning_rate": 7.701272857814991e-07,
+      "loss": 2.4417,
+      "step": 3730
+    },
+    {
+      "epoch": 0.48289477942842346,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5869362354278564,
+      "eval_runtime": 15.0861,
+      "eval_samples_per_second": 2.917,
+      "eval_steps_per_second": 0.199,
+      "step": 3730
+    },
+    {
+      "epoch": 0.4830242418357769,
+      "grad_norm": 1.5855997926424137,
+      "learning_rate": 7.698525151690056e-07,
+      "loss": 2.3074,
+      "step": 3731
+    },
+    {
+      "epoch": 0.4831537042431304,
+      "grad_norm": 2.276389780672671,
+      "learning_rate": 7.6957771496745e-07,
+      "loss": 2.4673,
+      "step": 3732
+    },
+    {
+      "epoch": 0.4832831666504839,
+      "grad_norm": 2.5975593439310316,
+      "learning_rate": 7.693028852329655e-07,
+      "loss": 2.4802,
+      "step": 3733
+    },
+    {
+      "epoch": 0.48341262905783733,
+      "grad_norm": 5.229809334928725,
+      "learning_rate": 7.690280260216917e-07,
+      "loss": 2.5776,
+      "step": 3734
+    },
+    {
+      "epoch": 0.4835420914651908,
+      "grad_norm": 2.566943958441565,
+      "learning_rate": 7.687531373897743e-07,
+      "loss": 2.5281,
+      "step": 3735
+    },
+    {
+      "epoch": 0.4835420914651908,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5834072828292847,
+      "eval_runtime": 14.059,
+      "eval_samples_per_second": 3.13,
+      "eval_steps_per_second": 0.213,
+      "step": 3735
+    },
+    {
+      "epoch": 0.48367155387254424,
+      "grad_norm": 1.2468401957330484,
+      "learning_rate": 7.68478219393365e-07,
+      "loss": 2.2517,
+      "step": 3736
+    },
+    {
+      "epoch": 0.48380101627989774,
+      "grad_norm": 1.5363808577193312,
+      "learning_rate": 7.682032720886213e-07,
+      "loss": 2.3936,
+      "step": 3737
+    },
+    {
+      "epoch": 0.4839304786872512,
+      "grad_norm": 3.4440986799875177,
+      "learning_rate": 7.679282955317063e-07,
+      "loss": 2.8218,
+      "step": 3738
+    },
+    {
+      "epoch": 0.48405994109460465,
+      "grad_norm": 1.5846203120545803,
+      "learning_rate": 7.676532897787903e-07,
+      "loss": 2.5742,
+      "step": 3739
+    },
+    {
+      "epoch": 0.4841894035019581,
+      "grad_norm": 2.2258609691683886,
+      "learning_rate": 7.673782548860481e-07,
+      "loss": 2.3489,
+      "step": 3740
+    },
+    {
+      "epoch": 0.4841894035019581,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5833629369735718,
+      "eval_runtime": 14.3313,
+      "eval_samples_per_second": 3.07,
+      "eval_steps_per_second": 0.209,
+      "step": 3740
+    },
+    {
+      "epoch": 0.48431886590931156,
+      "grad_norm": 1.8681705513875824,
+      "learning_rate": 7.671031909096619e-07,
+      "loss": 2.4861,
+      "step": 3741
+    },
+    {
+      "epoch": 0.48444832831666507,
+      "grad_norm": 2.754221053995706,
+      "learning_rate": 7.668280979058182e-07,
+      "loss": 2.8,
+      "step": 3742
+    },
+    {
+      "epoch": 0.4845777907240185,
+      "grad_norm": 2.8174720182799904,
+      "learning_rate": 7.665529759307108e-07,
+      "loss": 2.6074,
+      "step": 3743
+    },
+    {
+      "epoch": 0.484707253131372,
+      "grad_norm": 1.624618610155561,
+      "learning_rate": 7.662778250405392e-07,
+      "loss": 2.4326,
+      "step": 3744
+    },
+    {
+      "epoch": 0.4848367155387254,
+      "grad_norm": 1.7181149937346343,
+      "learning_rate": 7.660026452915081e-07,
+      "loss": 2.4961,
+      "step": 3745
+    },
+    {
+      "epoch": 0.4848367155387254,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5826305150985718,
+      "eval_runtime": 13.5985,
+      "eval_samples_per_second": 3.236,
+      "eval_steps_per_second": 0.221,
+      "step": 3745
+    },
+    {
+      "epoch": 0.4849661779460789,
+      "grad_norm": 1.4223643090294513,
+      "learning_rate": 7.657274367398288e-07,
+      "loss": 2.3142,
+      "step": 3746
+    },
+    {
+      "epoch": 0.4850956403534324,
+      "grad_norm": 1.4142145578127674,
+      "learning_rate": 7.65452199441718e-07,
+      "loss": 2.374,
+      "step": 3747
+    },
+    {
+      "epoch": 0.48522510276078584,
+      "grad_norm": 1.387880401134795,
+      "learning_rate": 7.651769334533987e-07,
+      "loss": 2.5703,
+      "step": 3748
+    },
+    {
+      "epoch": 0.4853545651681393,
+      "grad_norm": 1.2781165790519644,
+      "learning_rate": 7.649016388310994e-07,
+      "loss": 2.3257,
+      "step": 3749
+    },
+    {
+      "epoch": 0.48548402757549275,
+      "grad_norm": 2.682387195459793,
+      "learning_rate": 7.646263156310549e-07,
+      "loss": 2.9258,
+      "step": 3750
+    },
+    {
+      "epoch": 0.48548402757549275,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5848056077957153,
+      "eval_runtime": 13.5334,
+      "eval_samples_per_second": 3.251,
+      "eval_steps_per_second": 0.222,
+      "step": 3750
+    },
+    {
+      "epoch": 0.48561348998284626,
+      "grad_norm": 2.115571873693488,
+      "learning_rate": 7.643509639095053e-07,
+      "loss": 2.3838,
+      "step": 3751
+    },
+    {
+      "epoch": 0.4857429523901997,
+      "grad_norm": 1.645931359818129,
+      "learning_rate": 7.640755837226965e-07,
+      "loss": 2.5327,
+      "step": 3752
+    },
+    {
+      "epoch": 0.48587241479755316,
+      "grad_norm": 1.3347804292284098,
+      "learning_rate": 7.638001751268812e-07,
+      "loss": 2.3938,
+      "step": 3753
+    },
+    {
+      "epoch": 0.4860018772049066,
+      "grad_norm": 2.9185762888402813,
+      "learning_rate": 7.635247381783165e-07,
+      "loss": 2.8037,
+      "step": 3754
+    },
+    {
+      "epoch": 0.48613133961226007,
+      "grad_norm": 1.715877741532721,
+      "learning_rate": 7.632492729332664e-07,
+      "loss": 2.5388,
+      "step": 3755
+    },
+    {
+      "epoch": 0.48613133961226007,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5834739208221436,
+      "eval_runtime": 14.5946,
+      "eval_samples_per_second": 3.015,
+      "eval_steps_per_second": 0.206,
+      "step": 3755
+    },
+    {
+      "epoch": 0.4862608020196136,
+      "grad_norm": 2.406502812620039,
+      "learning_rate": 7.629737794480001e-07,
+      "loss": 2.5679,
+      "step": 3756
+    },
+    {
+      "epoch": 0.48639026442696703,
+      "grad_norm": 2.749982137582843,
+      "learning_rate": 7.626982577787925e-07,
+      "loss": 2.4883,
+      "step": 3757
+    },
+    {
+      "epoch": 0.4865197268343205,
+      "grad_norm": 2.599471880256545,
+      "learning_rate": 7.624227079819251e-07,
+      "loss": 2.4792,
+      "step": 3758
+    },
+    {
+      "epoch": 0.48664918924167394,
+      "grad_norm": 2.2316005291715233,
+      "learning_rate": 7.62147130113684e-07,
+      "loss": 2.2654,
+      "step": 3759
+    },
+    {
+      "epoch": 0.4867786516490274,
+      "grad_norm": 1.61230855683663,
+      "learning_rate": 7.618715242303616e-07,
+      "loss": 2.479,
+      "step": 3760
+    },
+    {
+      "epoch": 0.4867786516490274,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5819424390792847,
+      "eval_runtime": 14.186,
+      "eval_samples_per_second": 3.102,
+      "eval_steps_per_second": 0.211,
+      "step": 3760
+    },
+    {
+      "epoch": 0.4869081140563809,
+      "grad_norm": 1.5997071523205781,
+      "learning_rate": 7.615958903882562e-07,
+      "loss": 2.6201,
+      "step": 3761
+    },
+    {
+      "epoch": 0.48703757646373436,
+      "grad_norm": 1.263696374332714,
+      "learning_rate": 7.613202286436716e-07,
+      "loss": 2.4778,
+      "step": 3762
+    },
+    {
+      "epoch": 0.4871670388710878,
+      "grad_norm": 2.0725752991641766,
+      "learning_rate": 7.610445390529172e-07,
+      "loss": 2.3997,
+      "step": 3763
+    },
+    {
+      "epoch": 0.48729650127844126,
+      "grad_norm": 2.061199888559606,
+      "learning_rate": 7.607688216723083e-07,
+      "loss": 2.511,
+      "step": 3764
+    },
+    {
+      "epoch": 0.4874259636857947,
+      "grad_norm": 2.1263173244821743,
+      "learning_rate": 7.604930765581655e-07,
+      "loss": 2.6099,
+      "step": 3765
+    },
+    {
+      "epoch": 0.4874259636857947,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5812321901321411,
+      "eval_runtime": 14.4445,
+      "eval_samples_per_second": 3.046,
+      "eval_steps_per_second": 0.208,
+      "step": 3765
+    },
+    {
+      "epoch": 0.4875554260931482,
+      "grad_norm": 2.603026123162556,
+      "learning_rate": 7.602173037668155e-07,
+      "loss": 2.6643,
+      "step": 3766
+    },
+    {
+      "epoch": 0.4876848885005017,
+      "grad_norm": 1.821807698417415,
+      "learning_rate": 7.599415033545906e-07,
+      "loss": 2.2778,
+      "step": 3767
+    },
+    {
+      "epoch": 0.48781435090785513,
+      "grad_norm": 2.176638610369739,
+      "learning_rate": 7.596656753778284e-07,
+      "loss": 2.3701,
+      "step": 3768
+    },
+    {
+      "epoch": 0.4879438133152086,
+      "grad_norm": 2.748369441027363,
+      "learning_rate": 7.593898198928725e-07,
+      "loss": 2.3882,
+      "step": 3769
+    },
+    {
+      "epoch": 0.48807327572256204,
+      "grad_norm": 2.673306715070197,
+      "learning_rate": 7.591139369560717e-07,
+      "loss": 2.5186,
+      "step": 3770
+    },
+    {
+      "epoch": 0.48807327572256204,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5784579515457153,
+      "eval_runtime": 14.2177,
+      "eval_samples_per_second": 3.095,
+      "eval_steps_per_second": 0.211,
+      "step": 3770
+    },
+    {
+      "epoch": 0.48820273812991555,
+      "grad_norm": 1.9444332531518242,
+      "learning_rate": 7.588380266237811e-07,
+      "loss": 2.4504,
+      "step": 3771
+    },
+    {
+      "epoch": 0.488332200537269,
+      "grad_norm": 2.10486202104313,
+      "learning_rate": 7.585620889523609e-07,
+      "loss": 2.5176,
+      "step": 3772
+    },
+    {
+      "epoch": 0.48846166294462245,
+      "grad_norm": 3.366771178935237,
+      "learning_rate": 7.582861239981767e-07,
+      "loss": 3.1654,
+      "step": 3773
+    },
+    {
+      "epoch": 0.4885911253519759,
+      "grad_norm": 2.1460931983728173,
+      "learning_rate": 7.580101318176001e-07,
+      "loss": 2.4222,
+      "step": 3774
+    },
+    {
+      "epoch": 0.48872058775932936,
+      "grad_norm": 1.9874755259185424,
+      "learning_rate": 7.577341124670082e-07,
+      "loss": 2.5044,
+      "step": 3775
+    },
+    {
+      "epoch": 0.48872058775932936,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.58056640625,
+      "eval_runtime": 14.8579,
+      "eval_samples_per_second": 2.961,
+      "eval_steps_per_second": 0.202,
+      "step": 3775
+    },
+    {
+      "epoch": 0.48885005016668287,
+      "grad_norm": 1.417694617421724,
+      "learning_rate": 7.574580660027833e-07,
+      "loss": 2.5955,
+      "step": 3776
+    },
+    {
+      "epoch": 0.4889795125740363,
+      "grad_norm": 1.5890110186592572,
+      "learning_rate": 7.571819924813138e-07,
+      "loss": 2.4797,
+      "step": 3777
+    },
+    {
+      "epoch": 0.4891089749813898,
+      "grad_norm": 1.7896499862697623,
+      "learning_rate": 7.56905891958993e-07,
+      "loss": 2.6299,
+      "step": 3778
+    },
+    {
+      "epoch": 0.48923843738874323,
+      "grad_norm": 1.4127749632758106,
+      "learning_rate": 7.566297644922202e-07,
+      "loss": 2.3528,
+      "step": 3779
+    },
+    {
+      "epoch": 0.4893678997960967,
+      "grad_norm": 1.506628794055902,
+      "learning_rate": 7.563536101373999e-07,
+      "loss": 2.3928,
+      "step": 3780
+    },
+    {
+      "epoch": 0.4893678997960967,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5820090770721436,
+      "eval_runtime": 15.0042,
+      "eval_samples_per_second": 2.933,
+      "eval_steps_per_second": 0.2,
+      "step": 3780
+    },
+    {
+      "epoch": 0.4894973622034502,
+      "grad_norm": 1.835602741807069,
+      "learning_rate": 7.560774289509423e-07,
+      "loss": 2.4072,
+      "step": 3781
+    },
+    {
+      "epoch": 0.48962682461080365,
+      "grad_norm": 1.872658240403593,
+      "learning_rate": 7.55801220989263e-07,
+      "loss": 2.5349,
+      "step": 3782
+    },
+    {
+      "epoch": 0.4897562870181571,
+      "grad_norm": 2.059157933072759,
+      "learning_rate": 7.555249863087833e-07,
+      "loss": 2.9045,
+      "step": 3783
+    },
+    {
+      "epoch": 0.48988574942551055,
+      "grad_norm": 4.654972077338492,
+      "learning_rate": 7.552487249659293e-07,
+      "loss": 2.7256,
+      "step": 3784
+    },
+    {
+      "epoch": 0.490015211832864,
+      "grad_norm": 4.282968235896998,
+      "learning_rate": 7.549724370171332e-07,
+      "loss": 3.1448,
+      "step": 3785
+    },
+    {
+      "epoch": 0.490015211832864,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.582075595855713,
+      "eval_runtime": 15.2266,
+      "eval_samples_per_second": 2.89,
+      "eval_steps_per_second": 0.197,
+      "step": 3785
+    },
+    {
+      "epoch": 0.4901446742402175,
+      "grad_norm": 1.7328692945690685,
+      "learning_rate": 7.546961225188326e-07,
+      "loss": 2.5132,
+      "step": 3786
+    },
+    {
+      "epoch": 0.49027413664757097,
+      "grad_norm": 1.9279686658187183,
+      "learning_rate": 7.544197815274701e-07,
+      "loss": 2.3301,
+      "step": 3787
+    },
+    {
+      "epoch": 0.4904035990549244,
+      "grad_norm": 1.8035244152791998,
+      "learning_rate": 7.541434140994939e-07,
+      "loss": 2.4041,
+      "step": 3788
+    },
+    {
+      "epoch": 0.4905330614622779,
+      "grad_norm": 1.4612428600763918,
+      "learning_rate": 7.538670202913578e-07,
+      "loss": 2.2925,
+      "step": 3789
+    },
+    {
+      "epoch": 0.49066252386963133,
+      "grad_norm": 2.40106368432215,
+      "learning_rate": 7.535906001595208e-07,
+      "loss": 2.3408,
+      "step": 3790
+    },
+    {
+      "epoch": 0.49066252386963133,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5819646120071411,
+      "eval_runtime": 13.9747,
+      "eval_samples_per_second": 3.149,
+      "eval_steps_per_second": 0.215,
+      "step": 3790
+    },
+    {
+      "epoch": 0.49079198627698484,
+      "grad_norm": 1.4759548614521,
+      "learning_rate": 7.533141537604474e-07,
+      "loss": 2.4529,
+      "step": 3791
+    },
+    {
+      "epoch": 0.4909214486843383,
+      "grad_norm": 1.806778938780607,
+      "learning_rate": 7.530376811506073e-07,
+      "loss": 2.3044,
+      "step": 3792
+    },
+    {
+      "epoch": 0.49105091109169174,
+      "grad_norm": 2.4843736152069993,
+      "learning_rate": 7.527611823864753e-07,
+      "loss": 2.6257,
+      "step": 3793
+    },
+    {
+      "epoch": 0.4911803734990452,
+      "grad_norm": 1.3744583028057233,
+      "learning_rate": 7.524846575245326e-07,
+      "loss": 2.2251,
+      "step": 3794
+    },
+    {
+      "epoch": 0.4913098359063987,
+      "grad_norm": 2.4159379021213367,
+      "learning_rate": 7.522081066212644e-07,
+      "loss": 2.8132,
+      "step": 3795
+    },
+    {
+      "epoch": 0.4913098359063987,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5798561573028564,
+      "eval_runtime": 15.1108,
+      "eval_samples_per_second": 2.912,
+      "eval_steps_per_second": 0.199,
+      "step": 3795
+    },
+    {
+      "epoch": 0.49143929831375216,
+      "grad_norm": 1.9351467813129752,
+      "learning_rate": 7.519315297331619e-07,
+      "loss": 2.4399,
+      "step": 3796
+    },
+    {
+      "epoch": 0.4915687607211056,
+      "grad_norm": 1.7565517773032409,
+      "learning_rate": 7.51654926916722e-07,
+      "loss": 2.4822,
+      "step": 3797
+    },
+    {
+      "epoch": 0.49169822312845907,
+      "grad_norm": 1.8909031824520097,
+      "learning_rate": 7.513782982284458e-07,
+      "loss": 2.5505,
+      "step": 3798
+    },
+    {
+      "epoch": 0.4918276855358125,
+      "grad_norm": 2.0266424653585844,
+      "learning_rate": 7.511016437248407e-07,
+      "loss": 2.2991,
+      "step": 3799
+    },
+    {
+      "epoch": 0.49195714794316603,
+      "grad_norm": 1.9321306233451563,
+      "learning_rate": 7.508249634624191e-07,
+      "loss": 2.5488,
+      "step": 3800
+    },
+    {
+      "epoch": 0.49195714794316603,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5777698755264282,
+      "eval_runtime": 15.4091,
+      "eval_samples_per_second": 2.855,
+      "eval_steps_per_second": 0.195,
+      "step": 3800
+    },
+    {
+      "epoch": 0.4920866103505195,
+      "grad_norm": 2.249859434112645,
+      "learning_rate": 7.505482574976981e-07,
+      "loss": 2.4429,
+      "step": 3801
+    },
+    {
+      "epoch": 0.49221607275787294,
+      "grad_norm": 1.9578910076494158,
+      "learning_rate": 7.502715258872009e-07,
+      "loss": 2.5437,
+      "step": 3802
+    },
+    {
+      "epoch": 0.4923455351652264,
+      "grad_norm": 1.6559026230322365,
+      "learning_rate": 7.499947686874553e-07,
+      "loss": 2.4465,
+      "step": 3803
+    },
+    {
+      "epoch": 0.49247499757257984,
+      "grad_norm": 1.971016537092765,
+      "learning_rate": 7.497179859549948e-07,
+      "loss": 2.4014,
+      "step": 3804
+    },
+    {
+      "epoch": 0.49260445997993335,
+      "grad_norm": 2.1029887299355785,
+      "learning_rate": 7.494411777463578e-07,
+      "loss": 2.5227,
+      "step": 3805
+    },
+    {
+      "epoch": 0.49260445997993335,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5769264698028564,
+      "eval_runtime": 14.0806,
+      "eval_samples_per_second": 3.125,
+      "eval_steps_per_second": 0.213,
+      "step": 3805
+    },
+    {
+      "epoch": 0.4927339223872868,
+      "grad_norm": 1.6023708569528559,
+      "learning_rate": 7.491643441180876e-07,
+      "loss": 2.5186,
+      "step": 3806
+    },
+    {
+      "epoch": 0.49286338479464026,
+      "grad_norm": 1.8521373085127388,
+      "learning_rate": 7.488874851267336e-07,
+      "loss": 2.781,
+      "step": 3807
+    },
+    {
+      "epoch": 0.4929928472019937,
+      "grad_norm": 1.7576352957359789,
+      "learning_rate": 7.486106008288499e-07,
+      "loss": 2.4624,
+      "step": 3808
+    },
+    {
+      "epoch": 0.49312230960934716,
+      "grad_norm": 1.819612450542925,
+      "learning_rate": 7.483336912809952e-07,
+      "loss": 2.3779,
+      "step": 3809
+    },
+    {
+      "epoch": 0.4932517720167007,
+      "grad_norm": 1.5098069548233064,
+      "learning_rate": 7.480567565397344e-07,
+      "loss": 2.5537,
+      "step": 3810
+    },
+    {
+      "epoch": 0.4932517720167007,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5764381885528564,
+      "eval_runtime": 14.0195,
+      "eval_samples_per_second": 3.138,
+      "eval_steps_per_second": 0.214,
+      "step": 3810
+    },
+    {
+      "epoch": 0.4933812344240541,
+      "grad_norm": 1.419732364950187,
+      "learning_rate": 7.477797966616366e-07,
+      "loss": 2.4734,
+      "step": 3811
+    },
+    {
+      "epoch": 0.4935106968314076,
+      "grad_norm": 1.3622137753538932,
+      "learning_rate": 7.475028117032769e-07,
+      "loss": 2.4253,
+      "step": 3812
+    },
+    {
+      "epoch": 0.49364015923876103,
+      "grad_norm": 2.1862426362924943,
+      "learning_rate": 7.472258017212348e-07,
+      "loss": 2.4448,
+      "step": 3813
+    },
+    {
+      "epoch": 0.4937696216461145,
+      "grad_norm": 1.903007143071898,
+      "learning_rate": 7.469487667720954e-07,
+      "loss": 2.6011,
+      "step": 3814
+    },
+    {
+      "epoch": 0.493899084053468,
+      "grad_norm": 1.379961011257503,
+      "learning_rate": 7.466717069124485e-07,
+      "loss": 2.4663,
+      "step": 3815
+    },
+    {
+      "epoch": 0.493899084053468,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5742853879928589,
+      "eval_runtime": 14.1682,
+      "eval_samples_per_second": 3.106,
+      "eval_steps_per_second": 0.212,
+      "step": 3815
+    },
+    {
+      "epoch": 0.49402854646082145,
+      "grad_norm": 4.395612935373734,
+      "learning_rate": 7.463946221988894e-07,
+      "loss": 2.6821,
+      "step": 3816
+    },
+    {
+      "epoch": 0.4941580088681749,
+      "grad_norm": 1.9107067356600425,
+      "learning_rate": 7.46117512688018e-07,
+      "loss": 2.3472,
+      "step": 3817
+    },
+    {
+      "epoch": 0.49428747127552836,
+      "grad_norm": 1.5604783790335544,
+      "learning_rate": 7.458403784364397e-07,
+      "loss": 2.5479,
+      "step": 3818
+    },
+    {
+      "epoch": 0.4944169336828818,
+      "grad_norm": 1.9559556506019429,
+      "learning_rate": 7.45563219500765e-07,
+      "loss": 2.4883,
+      "step": 3819
+    },
+    {
+      "epoch": 0.4945463960902353,
+      "grad_norm": 3.671482138816925,
+      "learning_rate": 7.452860359376087e-07,
+      "loss": 2.6936,
+      "step": 3820
+    },
+    {
+      "epoch": 0.4945463960902353,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5728648900985718,
+      "eval_runtime": 14.8591,
+      "eval_samples_per_second": 2.961,
+      "eval_steps_per_second": 0.202,
+      "step": 3820
+    },
+    {
+      "epoch": 0.49467585849758877,
+      "grad_norm": 1.394284610016648,
+      "learning_rate": 7.450088278035916e-07,
+      "loss": 2.5481,
+      "step": 3821
+    },
+    {
+      "epoch": 0.4948053209049422,
+      "grad_norm": 1.7266522832586098,
+      "learning_rate": 7.447315951553391e-07,
+      "loss": 2.3232,
+      "step": 3822
+    },
+    {
+      "epoch": 0.4949347833122957,
+      "grad_norm": 2.038626792647284,
+      "learning_rate": 7.444543380494813e-07,
+      "loss": 2.8459,
+      "step": 3823
+    },
+    {
+      "epoch": 0.49506424571964913,
+      "grad_norm": 2.8266099786324745,
+      "learning_rate": 7.441770565426537e-07,
+      "loss": 2.9866,
+      "step": 3824
+    },
+    {
+      "epoch": 0.49519370812700264,
+      "grad_norm": 2.0805122319699954,
+      "learning_rate": 7.438997506914968e-07,
+      "loss": 2.5251,
+      "step": 3825
+    },
+    {
+      "epoch": 0.49519370812700264,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5750178098678589,
+      "eval_runtime": 13.5697,
+      "eval_samples_per_second": 3.243,
+      "eval_steps_per_second": 0.221,
+      "step": 3825
+    },
+    {
+      "epoch": 0.4953231705343561,
+      "grad_norm": 2.0092854279822165,
+      "learning_rate": 7.436224205526557e-07,
+      "loss": 2.5103,
+      "step": 3826
+    },
+    {
+      "epoch": 0.49545263294170955,
+      "grad_norm": 2.6995394306328153,
+      "learning_rate": 7.433450661827811e-07,
+      "loss": 3.4343,
+      "step": 3827
+    },
+    {
+      "epoch": 0.495582095349063,
+      "grad_norm": 2.0084445623864977,
+      "learning_rate": 7.430676876385278e-07,
+      "loss": 2.4236,
+      "step": 3828
+    },
+    {
+      "epoch": 0.49571155775641645,
+      "grad_norm": 1.8805177996097957,
+      "learning_rate": 7.42790284976556e-07,
+      "loss": 2.429,
+      "step": 3829
+    },
+    {
+      "epoch": 0.49584102016376996,
+      "grad_norm": 1.8158045047655909,
+      "learning_rate": 7.425128582535313e-07,
+      "loss": 2.6643,
+      "step": 3830
+    },
+    {
+      "epoch": 0.49584102016376996,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5728648900985718,
+      "eval_runtime": 13.3939,
+      "eval_samples_per_second": 3.285,
+      "eval_steps_per_second": 0.224,
+      "step": 3830
+    },
+    {
+      "epoch": 0.4959704825711234,
+      "grad_norm": 1.811621330528476,
+      "learning_rate": 7.422354075261232e-07,
+      "loss": 2.2358,
+      "step": 3831
+    },
+    {
+      "epoch": 0.49609994497847687,
+      "grad_norm": 2.313425687473408,
+      "learning_rate": 7.419579328510069e-07,
+      "loss": 2.4932,
+      "step": 3832
+    },
+    {
+      "epoch": 0.4962294073858303,
+      "grad_norm": 2.6069791484869964,
+      "learning_rate": 7.416804342848623e-07,
+      "loss": 2.7866,
+      "step": 3833
+    },
+    {
+      "epoch": 0.4963588697931838,
+      "grad_norm": 1.5991832552492549,
+      "learning_rate": 7.414029118843738e-07,
+      "loss": 2.4744,
+      "step": 3834
+    },
+    {
+      "epoch": 0.4964883322005373,
+      "grad_norm": 1.8881290900592334,
+      "learning_rate": 7.411253657062313e-07,
+      "loss": 2.345,
+      "step": 3835
+    },
+    {
+      "epoch": 0.4964883322005373,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5743075609207153,
+      "eval_runtime": 14.4064,
+      "eval_samples_per_second": 3.054,
+      "eval_steps_per_second": 0.208,
+      "step": 3835
+    },
+    {
+      "epoch": 0.49661779460789074,
+      "grad_norm": 3.204383757763302,
+      "learning_rate": 7.408477958071291e-07,
+      "loss": 2.4419,
+      "step": 3836
+    },
+    {
+      "epoch": 0.4967472570152442,
+      "grad_norm": 1.3868120509253685,
+      "learning_rate": 7.405702022437666e-07,
+      "loss": 2.4214,
+      "step": 3837
+    },
+    {
+      "epoch": 0.49687671942259765,
+      "grad_norm": 1.3155182370097174,
+      "learning_rate": 7.402925850728475e-07,
+      "loss": 2.4441,
+      "step": 3838
+    },
+    {
+      "epoch": 0.49700618182995115,
+      "grad_norm": 1.8934473387512372,
+      "learning_rate": 7.400149443510814e-07,
+      "loss": 2.4619,
+      "step": 3839
+    },
+    {
+      "epoch": 0.4971356442373046,
+      "grad_norm": 1.3701282877287062,
+      "learning_rate": 7.397372801351814e-07,
+      "loss": 2.3213,
+      "step": 3840
+    },
+    {
+      "epoch": 0.4971356442373046,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5686701536178589,
+      "eval_runtime": 14.2887,
+      "eval_samples_per_second": 3.079,
+      "eval_steps_per_second": 0.21,
+      "step": 3840
+    },
+    {
+      "epoch": 0.49726510664465806,
+      "grad_norm": 2.381051717378634,
+      "learning_rate": 7.394595924818665e-07,
+      "loss": 2.6194,
+      "step": 3841
+    },
+    {
+      "epoch": 0.4973945690520115,
+      "grad_norm": 3.912663294756281,
+      "learning_rate": 7.391818814478595e-07,
+      "loss": 2.627,
+      "step": 3842
+    },
+    {
+      "epoch": 0.49752403145936497,
+      "grad_norm": 1.8801207395755364,
+      "learning_rate": 7.38904147089889e-07,
+      "loss": 2.5083,
+      "step": 3843
+    },
+    {
+      "epoch": 0.4976534938667185,
+      "grad_norm": 1.8499822206012733,
+      "learning_rate": 7.38626389464688e-07,
+      "loss": 2.484,
+      "step": 3844
+    },
+    {
+      "epoch": 0.49778295627407193,
+      "grad_norm": 1.9300649295209942,
+      "learning_rate": 7.383486086289936e-07,
+      "loss": 2.6553,
+      "step": 3845
+    },
+    {
+      "epoch": 0.49778295627407193,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5678932666778564,
+      "eval_runtime": 14.9147,
+      "eval_samples_per_second": 2.95,
+      "eval_steps_per_second": 0.201,
+      "step": 3845
+    },
+    {
+      "epoch": 0.4979124186814254,
+      "grad_norm": 1.4945354598537437,
+      "learning_rate": 7.380708046395485e-07,
+      "loss": 2.3804,
+      "step": 3846
+    },
+    {
+      "epoch": 0.49804188108877884,
+      "grad_norm": 1.287166983856708,
+      "learning_rate": 7.377929775530997e-07,
+      "loss": 2.3335,
+      "step": 3847
+    },
+    {
+      "epoch": 0.4981713434961323,
+      "grad_norm": 2.041089072972178,
+      "learning_rate": 7.375151274263989e-07,
+      "loss": 2.4495,
+      "step": 3848
+    },
+    {
+      "epoch": 0.4983008059034858,
+      "grad_norm": 1.734460772955245,
+      "learning_rate": 7.37237254316203e-07,
+      "loss": 2.5708,
+      "step": 3849
+    },
+    {
+      "epoch": 0.49843026831083925,
+      "grad_norm": 2.155783042283747,
+      "learning_rate": 7.369593582792727e-07,
+      "loss": 2.552,
+      "step": 3850
+    },
+    {
+      "epoch": 0.49843026831083925,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5707341432571411,
+      "eval_runtime": 13.9557,
+      "eval_samples_per_second": 3.153,
+      "eval_steps_per_second": 0.215,
+      "step": 3850
+    },
+    {
+      "epoch": 0.4985597307181927,
+      "grad_norm": 1.8270992707659444,
+      "learning_rate": 7.366814393723742e-07,
+      "loss": 2.5442,
+      "step": 3851
+    },
+    {
+      "epoch": 0.49868919312554616,
+      "grad_norm": 1.526756141936261,
+      "learning_rate": 7.364034976522783e-07,
+      "loss": 2.5999,
+      "step": 3852
+    },
+    {
+      "epoch": 0.4988186555328996,
+      "grad_norm": 2.1889065723581616,
+      "learning_rate": 7.361255331757597e-07,
+      "loss": 2.689,
+      "step": 3853
+    },
+    {
+      "epoch": 0.4989481179402531,
+      "grad_norm": 1.4077063254995932,
+      "learning_rate": 7.358475459995985e-07,
+      "loss": 2.5684,
+      "step": 3854
+    },
+    {
+      "epoch": 0.4990775803476066,
+      "grad_norm": 2.28805826148618,
+      "learning_rate": 7.355695361805794e-07,
+      "loss": 2.3989,
+      "step": 3855
+    },
+    {
+      "epoch": 0.4990775803476066,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5650302171707153,
+      "eval_runtime": 14.6487,
+      "eval_samples_per_second": 3.004,
+      "eval_steps_per_second": 0.205,
+      "step": 3855
+    },
+    {
+      "epoch": 0.49920704275496003,
+      "grad_norm": 1.5251554579287894,
+      "learning_rate": 7.352915037754912e-07,
+      "loss": 2.5068,
+      "step": 3856
+    },
+    {
+      "epoch": 0.4993365051623135,
+      "grad_norm": 2.1011453941041363,
+      "learning_rate": 7.35013448841128e-07,
+      "loss": 2.8315,
+      "step": 3857
+    },
+    {
+      "epoch": 0.49946596756966694,
+      "grad_norm": 1.1925082996351501,
+      "learning_rate": 7.347353714342881e-07,
+      "loss": 2.272,
+      "step": 3858
+    },
+    {
+      "epoch": 0.49959542997702044,
+      "grad_norm": 1.7177547031808236,
+      "learning_rate": 7.344572716117742e-07,
+      "loss": 2.6719,
+      "step": 3859
+    },
+    {
+      "epoch": 0.4997248923843739,
+      "grad_norm": 2.021125481308124,
+      "learning_rate": 7.34179149430394e-07,
+      "loss": 2.4043,
+      "step": 3860
+    },
+    {
+      "epoch": 0.4997248923843739,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5636540651321411,
+      "eval_runtime": 14.5011,
+      "eval_samples_per_second": 3.034,
+      "eval_steps_per_second": 0.207,
+      "step": 3860
+    },
+    {
+      "epoch": 0.49985435479172735,
+      "grad_norm": 1.3750962233632043,
+      "learning_rate": 7.339010049469596e-07,
+      "loss": 2.2122,
+      "step": 3861
+    },
+    {
+      "epoch": 0.4999838171990808,
+      "grad_norm": 1.5203316871122106,
+      "learning_rate": 7.336228382182875e-07,
+      "loss": 2.4861,
+      "step": 3862
+    },
+    {
+      "epoch": 0.5001132796064343,
+      "grad_norm": 2.0263422902254677,
+      "learning_rate": 7.333446493011991e-07,
+      "loss": 2.418,
+      "step": 3863
+    },
+    {
+      "epoch": 0.5002427420137877,
+      "grad_norm": 2.407502207446351,
+      "learning_rate": 7.330664382525201e-07,
+      "loss": 2.5923,
+      "step": 3864
+    },
+    {
+      "epoch": 0.5003722044211412,
+      "grad_norm": 2.111981918023866,
+      "learning_rate": 7.327882051290805e-07,
+      "loss": 2.9321,
+      "step": 3865
+    },
+    {
+      "epoch": 0.5003722044211412,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5658291578292847,
+      "eval_runtime": 13.6826,
+      "eval_samples_per_second": 3.216,
+      "eval_steps_per_second": 0.219,
+      "step": 3865
+    },
+    {
+      "epoch": 0.5005016668284947,
+      "grad_norm": 2.0362133554203052,
+      "learning_rate": 7.325099499877154e-07,
+      "loss": 2.4763,
+      "step": 3866
+    },
+    {
+      "epoch": 0.5006311292358482,
+      "grad_norm": 1.3532069968107463,
+      "learning_rate": 7.32231672885264e-07,
+      "loss": 2.4005,
+      "step": 3867
+    },
+    {
+      "epoch": 0.5007605916432016,
+      "grad_norm": 1.8315827762511798,
+      "learning_rate": 7.319533738785696e-07,
+      "loss": 2.5701,
+      "step": 3868
+    },
+    {
+      "epoch": 0.5008900540505551,
+      "grad_norm": 1.9170065064572093,
+      "learning_rate": 7.316750530244809e-07,
+      "loss": 2.4468,
+      "step": 3869
+    },
+    {
+      "epoch": 0.5010195164579085,
+      "grad_norm": 2.0260918662704515,
+      "learning_rate": 7.313967103798504e-07,
+      "loss": 2.6094,
+      "step": 3870
+    },
+    {
+      "epoch": 0.5010195164579085,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5665394067764282,
+      "eval_runtime": 13.5102,
+      "eval_samples_per_second": 3.257,
+      "eval_steps_per_second": 0.222,
+      "step": 3870
+    },
+    {
+      "epoch": 0.501148978865262,
+      "grad_norm": 1.5321029807546402,
+      "learning_rate": 7.311183460015353e-07,
+      "loss": 2.4431,
+      "step": 3871
+    },
+    {
+      "epoch": 0.5012784412726154,
+      "grad_norm": 1.7593700539247659,
+      "learning_rate": 7.308399599463973e-07,
+      "loss": 2.3701,
+      "step": 3872
+    },
+    {
+      "epoch": 0.5014079036799689,
+      "grad_norm": 1.882865198824058,
+      "learning_rate": 7.305615522713018e-07,
+      "loss": 2.4648,
+      "step": 3873
+    },
+    {
+      "epoch": 0.5015373660873224,
+      "grad_norm": 1.4862211139501802,
+      "learning_rate": 7.302831230331197e-07,
+      "loss": 2.4026,
+      "step": 3874
+    },
+    {
+      "epoch": 0.5016668284946758,
+      "grad_norm": 2.947035413963958,
+      "learning_rate": 7.300046722887259e-07,
+      "loss": 2.6794,
+      "step": 3875
+    },
+    {
+      "epoch": 0.5016668284946758,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5650745630264282,
+      "eval_runtime": 14.482,
+      "eval_samples_per_second": 3.038,
+      "eval_steps_per_second": 0.207,
+      "step": 3875
+    },
+    {
+      "epoch": 0.5017962909020294,
+      "grad_norm": 2.41367090513958,
+      "learning_rate": 7.297262000949993e-07,
+      "loss": 2.4866,
+      "step": 3876
+    },
+    {
+      "epoch": 0.5019257533093828,
+      "grad_norm": 1.9178427941864862,
+      "learning_rate": 7.294477065088237e-07,
+      "loss": 2.5181,
+      "step": 3877
+    },
+    {
+      "epoch": 0.5020552157167363,
+      "grad_norm": 2.326835907076696,
+      "learning_rate": 7.291691915870869e-07,
+      "loss": 2.3979,
+      "step": 3878
+    },
+    {
+      "epoch": 0.5021846781240897,
+      "grad_norm": 3.4223439682426218,
+      "learning_rate": 7.288906553866815e-07,
+      "loss": 2.4858,
+      "step": 3879
+    },
+    {
+      "epoch": 0.5023141405314432,
+      "grad_norm": 2.908525290796561,
+      "learning_rate": 7.286120979645038e-07,
+      "loss": 2.4551,
+      "step": 3880
+    },
+    {
+      "epoch": 0.5023141405314432,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.56005859375,
+      "eval_runtime": 14.2513,
+      "eval_samples_per_second": 3.087,
+      "eval_steps_per_second": 0.211,
+      "step": 3880
+    },
+    {
+      "epoch": 0.5024436029387966,
+      "grad_norm": 3.0164933433794934,
+      "learning_rate": 7.28333519377455e-07,
+      "loss": 2.4829,
+      "step": 3881
+    },
+    {
+      "epoch": 0.5025730653461501,
+      "grad_norm": 2.490369015888935,
+      "learning_rate": 7.280549196824405e-07,
+      "loss": 2.5938,
+      "step": 3882
+    },
+    {
+      "epoch": 0.5027025277535035,
+      "grad_norm": 1.8304420042134857,
+      "learning_rate": 7.277762989363696e-07,
+      "loss": 2.417,
+      "step": 3883
+    },
+    {
+      "epoch": 0.502831990160857,
+      "grad_norm": 4.388618020094771,
+      "learning_rate": 7.274976571961566e-07,
+      "loss": 2.3291,
+      "step": 3884
+    },
+    {
+      "epoch": 0.5029614525682105,
+      "grad_norm": 4.2393510992134456,
+      "learning_rate": 7.272189945187196e-07,
+      "loss": 2.6987,
+      "step": 3885
+    },
+    {
+      "epoch": 0.5029614525682105,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5587490797042847,
+      "eval_runtime": 13.4433,
+      "eval_samples_per_second": 3.273,
+      "eval_steps_per_second": 0.223,
+      "step": 3885
+    },
+    {
+      "epoch": 0.503090914975564,
+      "grad_norm": 1.3787967172423385,
+      "learning_rate": 7.269403109609811e-07,
+      "loss": 2.3911,
+      "step": 3886
+    },
+    {
+      "epoch": 0.5032203773829175,
+      "grad_norm": 2.333546024328004,
+      "learning_rate": 7.266616065798679e-07,
+      "loss": 2.7061,
+      "step": 3887
+    },
+    {
+      "epoch": 0.5033498397902709,
+      "grad_norm": 3.205586913066561,
+      "learning_rate": 7.263828814323107e-07,
+      "loss": 2.228,
+      "step": 3888
+    },
+    {
+      "epoch": 0.5034793021976244,
+      "grad_norm": 2.4211409160020594,
+      "learning_rate": 7.261041355752453e-07,
+      "loss": 2.8258,
+      "step": 3889
+    },
+    {
+      "epoch": 0.5036087646049778,
+      "grad_norm": 3.4646879677254367,
+      "learning_rate": 7.258253690656109e-07,
+      "loss": 3.0598,
+      "step": 3890
+    },
+    {
+      "epoch": 0.5036087646049778,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5587269067764282,
+      "eval_runtime": 13.2947,
+      "eval_samples_per_second": 3.31,
+      "eval_steps_per_second": 0.226,
+      "step": 3890
+    },
+    {
+      "epoch": 0.5037382270123313,
+      "grad_norm": 2.6102513763882227,
+      "learning_rate": 7.255465819603512e-07,
+      "loss": 2.6096,
+      "step": 3891
+    },
+    {
+      "epoch": 0.5038676894196847,
+      "grad_norm": 1.4244695399444645,
+      "learning_rate": 7.25267774316414e-07,
+      "loss": 2.51,
+      "step": 3892
+    },
+    {
+      "epoch": 0.5039971518270382,
+      "grad_norm": 1.7718620456748764,
+      "learning_rate": 7.249889461907517e-07,
+      "loss": 2.3584,
+      "step": 3893
+    },
+    {
+      "epoch": 0.5041266142343916,
+      "grad_norm": 1.8930299446891008,
+      "learning_rate": 7.247100976403204e-07,
+      "loss": 2.6296,
+      "step": 3894
+    },
+    {
+      "epoch": 0.5042560766417451,
+      "grad_norm": 1.7490787792980549,
+      "learning_rate": 7.244312287220806e-07,
+      "loss": 2.4651,
+      "step": 3895
+    },
+    {
+      "epoch": 0.5042560766417451,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5616122484207153,
+      "eval_runtime": 13.649,
+      "eval_samples_per_second": 3.224,
+      "eval_steps_per_second": 0.22,
+      "step": 3895
+    },
+    {
+      "epoch": 0.5043855390490987,
+      "grad_norm": 1.9619250480295667,
+      "learning_rate": 7.241523394929971e-07,
+      "loss": 2.4622,
+      "step": 3896
+    },
+    {
+      "epoch": 0.5045150014564521,
+      "grad_norm": 1.8031755554424054,
+      "learning_rate": 7.238734300100383e-07,
+      "loss": 2.4492,
+      "step": 3897
+    },
+    {
+      "epoch": 0.5046444638638056,
+      "grad_norm": 3.7399693913344443,
+      "learning_rate": 7.235945003301776e-07,
+      "loss": 2.6377,
+      "step": 3898
+    },
+    {
+      "epoch": 0.504773926271159,
+      "grad_norm": 3.5956869862531398,
+      "learning_rate": 7.233155505103916e-07,
+      "loss": 2.8918,
+      "step": 3899
+    },
+    {
+      "epoch": 0.5049033886785125,
+      "grad_norm": 2.3889617369703333,
+      "learning_rate": 7.230365806076618e-07,
+      "loss": 2.5273,
+      "step": 3900
+    },
+    {
+      "epoch": 0.5049033886785125,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.560990810394287,
+      "eval_runtime": 15.2679,
+      "eval_samples_per_second": 2.882,
+      "eval_steps_per_second": 0.196,
+      "step": 3900
+    },
+    {
+      "epoch": 0.5050328510858659,
+      "grad_norm": 1.4238818349690794,
+      "learning_rate": 7.227575906789731e-07,
+      "loss": 2.6147,
+      "step": 3901
+    },
+    {
+      "epoch": 0.5051623134932194,
+      "grad_norm": 1.6769802198278922,
+      "learning_rate": 7.224785807813152e-07,
+      "loss": 2.4956,
+      "step": 3902
+    },
+    {
+      "epoch": 0.5052917759005728,
+      "grad_norm": 1.7444086808256676,
+      "learning_rate": 7.221995509716815e-07,
+      "loss": 2.5999,
+      "step": 3903
+    },
+    {
+      "epoch": 0.5054212383079263,
+      "grad_norm": 1.4715754415043543,
+      "learning_rate": 7.219205013070692e-07,
+      "loss": 2.3098,
+      "step": 3904
+    },
+    {
+      "epoch": 0.5055507007152799,
+      "grad_norm": 1.7070369466948605,
+      "learning_rate": 7.216414318444799e-07,
+      "loss": 2.5869,
+      "step": 3905
+    },
+    {
+      "epoch": 0.5055507007152799,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.559525966644287,
+      "eval_runtime": 14.4639,
+      "eval_samples_per_second": 3.042,
+      "eval_steps_per_second": 0.207,
+      "step": 3905
+    },
+    {
+      "epoch": 0.5056801631226333,
+      "grad_norm": 2.655850881190418,
+      "learning_rate": 7.213623426409194e-07,
+      "loss": 2.6255,
+      "step": 3906
+    },
+    {
+      "epoch": 0.5058096255299868,
+      "grad_norm": 3.762788410686041,
+      "learning_rate": 7.210832337533973e-07,
+      "loss": 2.5059,
+      "step": 3907
+    },
+    {
+      "epoch": 0.5059390879373402,
+      "grad_norm": 1.948873950767315,
+      "learning_rate": 7.208041052389273e-07,
+      "loss": 2.6294,
+      "step": 3908
+    },
+    {
+      "epoch": 0.5060685503446937,
+      "grad_norm": 1.5547507570351142,
+      "learning_rate": 7.205249571545268e-07,
+      "loss": 2.5928,
+      "step": 3909
+    },
+    {
+      "epoch": 0.5061980127520471,
+      "grad_norm": 3.789349689907264,
+      "learning_rate": 7.202457895572176e-07,
+      "loss": 2.895,
+      "step": 3910
+    },
+    {
+      "epoch": 0.5061980127520471,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5601474046707153,
+      "eval_runtime": 14.714,
+      "eval_samples_per_second": 2.99,
+      "eval_steps_per_second": 0.204,
+      "step": 3910
+    },
+    {
+      "epoch": 0.5063274751594006,
+      "grad_norm": 2.6536148555537125,
+      "learning_rate": 7.199666025040253e-07,
+      "loss": 2.9846,
+      "step": 3911
+    },
+    {
+      "epoch": 0.506456937566754,
+      "grad_norm": 1.193291118307913,
+      "learning_rate": 7.196873960519795e-07,
+      "loss": 2.2935,
+      "step": 3912
+    },
+    {
+      "epoch": 0.5065863999741075,
+      "grad_norm": 1.9758267587284766,
+      "learning_rate": 7.194081702581141e-07,
+      "loss": 2.604,
+      "step": 3913
+    },
+    {
+      "epoch": 0.5067158623814609,
+      "grad_norm": 1.3407589899149672,
+      "learning_rate": 7.191289251794663e-07,
+      "loss": 2.3889,
+      "step": 3914
+    },
+    {
+      "epoch": 0.5068453247888145,
+      "grad_norm": 1.6273721120582303,
+      "learning_rate": 7.188496608730773e-07,
+      "loss": 2.5203,
+      "step": 3915
+    },
+    {
+      "epoch": 0.5068453247888145,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.562943935394287,
+      "eval_runtime": 14.343,
+      "eval_samples_per_second": 3.068,
+      "eval_steps_per_second": 0.209,
+      "step": 3915
+    },
+    {
+      "epoch": 0.506974787196168,
+      "grad_norm": 2.3531567114958154,
+      "learning_rate": 7.185703773959932e-07,
+      "loss": 2.4966,
+      "step": 3916
+    },
+    {
+      "epoch": 0.5071042496035214,
+      "grad_norm": 1.3115626162294656,
+      "learning_rate": 7.182910748052627e-07,
+      "loss": 2.3882,
+      "step": 3917
+    },
+    {
+      "epoch": 0.5072337120108749,
+      "grad_norm": 1.4346453004413782,
+      "learning_rate": 7.180117531579394e-07,
+      "loss": 2.3844,
+      "step": 3918
+    },
+    {
+      "epoch": 0.5073631744182283,
+      "grad_norm": 2.0330923507423715,
+      "learning_rate": 7.177324125110801e-07,
+      "loss": 2.386,
+      "step": 3919
+    },
+    {
+      "epoch": 0.5074926368255818,
+      "grad_norm": 2.489217714225753,
+      "learning_rate": 7.174530529217459e-07,
+      "loss": 2.5239,
+      "step": 3920
+    },
+    {
+      "epoch": 0.5074926368255818,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5594815015792847,
+      "eval_runtime": 13.2785,
+      "eval_samples_per_second": 3.314,
+      "eval_steps_per_second": 0.226,
+      "step": 3920
+    },
+    {
+      "epoch": 0.5076220992329352,
+      "grad_norm": 2.594991739503318,
+      "learning_rate": 7.171736744470019e-07,
+      "loss": 2.5317,
+      "step": 3921
+    },
+    {
+      "epoch": 0.5077515616402887,
+      "grad_norm": 1.5725199053060537,
+      "learning_rate": 7.168942771439167e-07,
+      "loss": 2.4084,
+      "step": 3922
+    },
+    {
+      "epoch": 0.5078810240476421,
+      "grad_norm": 1.881095004664746,
+      "learning_rate": 7.166148610695625e-07,
+      "loss": 2.4331,
+      "step": 3923
+    },
+    {
+      "epoch": 0.5080104864549956,
+      "grad_norm": 2.4266930175903747,
+      "learning_rate": 7.16335426281016e-07,
+      "loss": 2.5654,
+      "step": 3924
+    },
+    {
+      "epoch": 0.5081399488623491,
+      "grad_norm": 1.796259741508993,
+      "learning_rate": 7.160559728353576e-07,
+      "loss": 2.4663,
+      "step": 3925
+    },
+    {
+      "epoch": 0.5081399488623491,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5594815015792847,
+      "eval_runtime": 14.1639,
+      "eval_samples_per_second": 3.106,
+      "eval_steps_per_second": 0.212,
+      "step": 3925
+    },
+    {
+      "epoch": 0.5082694112697026,
+      "grad_norm": 1.4423450777449243,
+      "learning_rate": 7.157765007896709e-07,
+      "loss": 2.3232,
+      "step": 3926
+    },
+    {
+      "epoch": 0.508398873677056,
+      "grad_norm": 2.194804788908348,
+      "learning_rate": 7.154970102010441e-07,
+      "loss": 2.4922,
+      "step": 3927
+    },
+    {
+      "epoch": 0.5085283360844095,
+      "grad_norm": 2.2222308648950824,
+      "learning_rate": 7.152175011265688e-07,
+      "loss": 2.4585,
+      "step": 3928
+    },
+    {
+      "epoch": 0.508657798491763,
+      "grad_norm": 1.524060422558183,
+      "learning_rate": 7.1493797362334e-07,
+      "loss": 2.5251,
+      "step": 3929
+    },
+    {
+      "epoch": 0.5087872608991164,
+      "grad_norm": 2.2265936815451086,
+      "learning_rate": 7.146584277484574e-07,
+      "loss": 2.3047,
+      "step": 3930
+    },
+    {
+      "epoch": 0.5087872608991164,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5573508739471436,
+      "eval_runtime": 14.3982,
+      "eval_samples_per_second": 3.056,
+      "eval_steps_per_second": 0.208,
+      "step": 3930
+    },
+    {
+      "epoch": 0.5089167233064699,
+      "grad_norm": 1.6994007040797923,
+      "learning_rate": 7.143788635590235e-07,
+      "loss": 2.5957,
+      "step": 3931
+    },
+    {
+      "epoch": 0.5090461857138233,
+      "grad_norm": 1.5348308774429156,
+      "learning_rate": 7.140992811121449e-07,
+      "loss": 2.4495,
+      "step": 3932
+    },
+    {
+      "epoch": 0.5091756481211768,
+      "grad_norm": 2.0577768073031377,
+      "learning_rate": 7.138196804649325e-07,
+      "loss": 2.7041,
+      "step": 3933
+    },
+    {
+      "epoch": 0.5093051105285302,
+      "grad_norm": 2.1932914550121456,
+      "learning_rate": 7.135400616744998e-07,
+      "loss": 2.5571,
+      "step": 3934
+    },
+    {
+      "epoch": 0.5094345729358838,
+      "grad_norm": 1.7350578899840934,
+      "learning_rate": 7.132604247979651e-07,
+      "loss": 2.5276,
+      "step": 3935
+    },
+    {
+      "epoch": 0.5094345729358838,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5574396848678589,
+      "eval_runtime": 13.5905,
+      "eval_samples_per_second": 3.238,
+      "eval_steps_per_second": 0.221,
+      "step": 3935
+    },
+    {
+      "epoch": 0.5095640353432372,
+      "grad_norm": 1.3493533949797523,
+      "learning_rate": 7.129807698924496e-07,
+      "loss": 2.3362,
+      "step": 3936
+    },
+    {
+      "epoch": 0.5096934977505907,
+      "grad_norm": 1.497058971274792,
+      "learning_rate": 7.127010970150784e-07,
+      "loss": 2.262,
+      "step": 3937
+    },
+    {
+      "epoch": 0.5098229601579441,
+      "grad_norm": 2.2848055261385163,
+      "learning_rate": 7.124214062229806e-07,
+      "loss": 2.6392,
+      "step": 3938
+    },
+    {
+      "epoch": 0.5099524225652976,
+      "grad_norm": 1.5324710314121792,
+      "learning_rate": 7.121416975732889e-07,
+      "loss": 2.5166,
+      "step": 3939
+    },
+    {
+      "epoch": 0.5100818849726511,
+      "grad_norm": 2.2047869311356485,
+      "learning_rate": 7.118619711231389e-07,
+      "loss": 2.6123,
+      "step": 3940
+    },
+    {
+      "epoch": 0.5100818849726511,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5545543432235718,
+      "eval_runtime": 14.0838,
+      "eval_samples_per_second": 3.124,
+      "eval_steps_per_second": 0.213,
+      "step": 3940
+    },
+    {
+      "epoch": 0.5102113473800045,
+      "grad_norm": 1.6734007618058986,
+      "learning_rate": 7.115822269296708e-07,
+      "loss": 2.4174,
+      "step": 3941
+    },
+    {
+      "epoch": 0.510340809787358,
+      "grad_norm": 1.9927410656670073,
+      "learning_rate": 7.113024650500278e-07,
+      "loss": 2.4111,
+      "step": 3942
+    },
+    {
+      "epoch": 0.5104702721947114,
+      "grad_norm": 3.482649138463418,
+      "learning_rate": 7.110226855413571e-07,
+      "loss": 2.5056,
+      "step": 3943
+    },
+    {
+      "epoch": 0.510599734602065,
+      "grad_norm": 1.7149359690220838,
+      "learning_rate": 7.107428884608095e-07,
+      "loss": 2.4897,
+      "step": 3944
+    },
+    {
+      "epoch": 0.5107291970094184,
+      "grad_norm": 1.2134008456724723,
+      "learning_rate": 7.104630738655387e-07,
+      "loss": 2.3611,
+      "step": 3945
+    },
+    {
+      "epoch": 0.5107291970094184,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.556596279144287,
+      "eval_runtime": 14.6474,
+      "eval_samples_per_second": 3.004,
+      "eval_steps_per_second": 0.205,
+      "step": 3945
+    },
+    {
+      "epoch": 0.5108586594167719,
+      "grad_norm": 2.0132074726290883,
+      "learning_rate": 7.101832418127029e-07,
+      "loss": 2.4846,
+      "step": 3946
+    },
+    {
+      "epoch": 0.5109881218241253,
+      "grad_norm": 1.5541086630709071,
+      "learning_rate": 7.099033923594634e-07,
+      "loss": 2.3997,
+      "step": 3947
+    },
+    {
+      "epoch": 0.5111175842314788,
+      "grad_norm": 1.3299252453677401,
+      "learning_rate": 7.09623525562985e-07,
+      "loss": 2.3718,
+      "step": 3948
+    },
+    {
+      "epoch": 0.5112470466388322,
+      "grad_norm": 2.5221613480053224,
+      "learning_rate": 7.093436414804363e-07,
+      "loss": 2.4031,
+      "step": 3949
+    },
+    {
+      "epoch": 0.5113765090461857,
+      "grad_norm": 1.4096505860276634,
+      "learning_rate": 7.090637401689894e-07,
+      "loss": 2.3093,
+      "step": 3950
+    },
+    {
+      "epoch": 0.5113765090461857,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5552645921707153,
+      "eval_runtime": 13.9075,
+      "eval_samples_per_second": 3.164,
+      "eval_steps_per_second": 0.216,
+      "step": 3950
+    },
+    {
+      "epoch": 0.5115059714535392,
+      "grad_norm": 2.066049643975334,
+      "learning_rate": 7.087838216858194e-07,
+      "loss": 2.4053,
+      "step": 3951
+    },
+    {
+      "epoch": 0.5116354338608926,
+      "grad_norm": 1.7286364982543136,
+      "learning_rate": 7.085038860881058e-07,
+      "loss": 2.4419,
+      "step": 3952
+    },
+    {
+      "epoch": 0.5117648962682461,
+      "grad_norm": 1.4051318477919965,
+      "learning_rate": 7.082239334330309e-07,
+      "loss": 2.396,
+      "step": 3953
+    },
+    {
+      "epoch": 0.5118943586755996,
+      "grad_norm": 1.6146455394077133,
+      "learning_rate": 7.079439637777808e-07,
+      "loss": 2.5259,
+      "step": 3954
+    },
+    {
+      "epoch": 0.5120238210829531,
+      "grad_norm": 3.449450569319739,
+      "learning_rate": 7.076639771795447e-07,
+      "loss": 2.5874,
+      "step": 3955
+    },
+    {
+      "epoch": 0.5120238210829531,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5573508739471436,
+      "eval_runtime": 15.0765,
+      "eval_samples_per_second": 2.918,
+      "eval_steps_per_second": 0.199,
+      "step": 3955
+    },
+    {
+      "epoch": 0.5121532834903065,
+      "grad_norm": 2.060301125091945,
+      "learning_rate": 7.073839736955159e-07,
+      "loss": 2.7292,
+      "step": 3956
+    },
+    {
+      "epoch": 0.51228274589766,
+      "grad_norm": 2.9594007014426653,
+      "learning_rate": 7.071039533828905e-07,
+      "loss": 2.5938,
+      "step": 3957
+    },
+    {
+      "epoch": 0.5124122083050134,
+      "grad_norm": 1.8220953475582509,
+      "learning_rate": 7.068239162988687e-07,
+      "loss": 2.478,
+      "step": 3958
+    },
+    {
+      "epoch": 0.5125416707123669,
+      "grad_norm": 2.5455585823633564,
+      "learning_rate": 7.065438625006533e-07,
+      "loss": 2.5444,
+      "step": 3959
+    },
+    {
+      "epoch": 0.5126711331197203,
+      "grad_norm": 2.134688878493036,
+      "learning_rate": 7.06263792045451e-07,
+      "loss": 2.4739,
+      "step": 3960
+    },
+    {
+      "epoch": 0.5126711331197203,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5510475635528564,
+      "eval_runtime": 14.6108,
+      "eval_samples_per_second": 3.011,
+      "eval_steps_per_second": 0.205,
+      "step": 3960
+    },
+    {
+      "epoch": 0.5128005955270738,
+      "grad_norm": 4.072923458133599,
+      "learning_rate": 7.059837049904723e-07,
+      "loss": 2.8989,
+      "step": 3961
+    },
+    {
+      "epoch": 0.5129300579344273,
+      "grad_norm": 1.5879376346767393,
+      "learning_rate": 7.057036013929304e-07,
+      "loss": 2.5625,
+      "step": 3962
+    },
+    {
+      "epoch": 0.5130595203417807,
+      "grad_norm": 1.853374831338135,
+      "learning_rate": 7.05423481310042e-07,
+      "loss": 2.3704,
+      "step": 3963
+    },
+    {
+      "epoch": 0.5131889827491343,
+      "grad_norm": 3.11869981616625,
+      "learning_rate": 7.051433447990275e-07,
+      "loss": 2.665,
+      "step": 3964
+    },
+    {
+      "epoch": 0.5133184451564877,
+      "grad_norm": 1.9902384840853993,
+      "learning_rate": 7.048631919171103e-07,
+      "loss": 2.4248,
+      "step": 3965
+    },
+    {
+      "epoch": 0.5133184451564877,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.55517578125,
+      "eval_runtime": 13.9622,
+      "eval_samples_per_second": 3.151,
+      "eval_steps_per_second": 0.215,
+      "step": 3965
+    },
+    {
+      "epoch": 0.5134479075638412,
+      "grad_norm": 1.4680747921815727,
+      "learning_rate": 7.045830227215175e-07,
+      "loss": 2.2361,
+      "step": 3966
+    },
+    {
+      "epoch": 0.5135773699711946,
+      "grad_norm": 1.8960457945657165,
+      "learning_rate": 7.043028372694794e-07,
+      "loss": 2.5029,
+      "step": 3967
+    },
+    {
+      "epoch": 0.5137068323785481,
+      "grad_norm": 1.7457386696546948,
+      "learning_rate": 7.04022635618229e-07,
+      "loss": 2.4343,
+      "step": 3968
+    },
+    {
+      "epoch": 0.5138362947859015,
+      "grad_norm": 1.7655426299646468,
+      "learning_rate": 7.037424178250038e-07,
+      "loss": 2.4026,
+      "step": 3969
+    },
+    {
+      "epoch": 0.513965757193255,
+      "grad_norm": 1.485243791585457,
+      "learning_rate": 7.034621839470436e-07,
+      "loss": 2.4216,
+      "step": 3970
+    },
+    {
+      "epoch": 0.513965757193255,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.552290439605713,
+      "eval_runtime": 13.9204,
+      "eval_samples_per_second": 3.161,
+      "eval_steps_per_second": 0.216,
+      "step": 3970
+    },
+    {
+      "epoch": 0.5140952196006084,
+      "grad_norm": 3.732439611569006,
+      "learning_rate": 7.031819340415922e-07,
+      "loss": 2.4043,
+      "step": 3971
+    },
+    {
+      "epoch": 0.5142246820079619,
+      "grad_norm": 3.1296279316006124,
+      "learning_rate": 7.029016681658959e-07,
+      "loss": 2.5725,
+      "step": 3972
+    },
+    {
+      "epoch": 0.5143541444153154,
+      "grad_norm": 1.833534723667105,
+      "learning_rate": 7.026213863772048e-07,
+      "loss": 2.7883,
+      "step": 3973
+    },
+    {
+      "epoch": 0.5144836068226689,
+      "grad_norm": 1.7938191067710845,
+      "learning_rate": 7.023410887327723e-07,
+      "loss": 2.6987,
+      "step": 3974
+    },
+    {
+      "epoch": 0.5146130692300224,
+      "grad_norm": 1.924824007484302,
+      "learning_rate": 7.02060775289855e-07,
+      "loss": 2.5598,
+      "step": 3975
+    },
+    {
+      "epoch": 0.5146130692300224,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5531116724014282,
+      "eval_runtime": 14.4951,
+      "eval_samples_per_second": 3.036,
+      "eval_steps_per_second": 0.207,
+      "step": 3975
+    },
+    {
+      "epoch": 0.5147425316373758,
+      "grad_norm": 1.832494485582077,
+      "learning_rate": 7.017804461057122e-07,
+      "loss": 2.7869,
+      "step": 3976
+    },
+    {
+      "epoch": 0.5148719940447293,
+      "grad_norm": 3.2238741576907795,
+      "learning_rate": 7.01500101237607e-07,
+      "loss": 2.5723,
+      "step": 3977
+    },
+    {
+      "epoch": 0.5150014564520827,
+      "grad_norm": 2.179860322047722,
+      "learning_rate": 7.012197407428056e-07,
+      "loss": 2.5591,
+      "step": 3978
+    },
+    {
+      "epoch": 0.5151309188594362,
+      "grad_norm": 2.1896499209017257,
+      "learning_rate": 7.009393646785773e-07,
+      "loss": 2.4724,
+      "step": 3979
+    },
+    {
+      "epoch": 0.5152603812667896,
+      "grad_norm": 1.3999990607693729,
+      "learning_rate": 7.006589731021946e-07,
+      "loss": 2.3999,
+      "step": 3980
+    },
+    {
+      "epoch": 0.5152603812667896,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5510032176971436,
+      "eval_runtime": 14.3346,
+      "eval_samples_per_second": 3.069,
+      "eval_steps_per_second": 0.209,
+      "step": 3980
+    },
+    {
+      "epoch": 0.5153898436741431,
+      "grad_norm": 2.923499299527811,
+      "learning_rate": 7.003785660709329e-07,
+      "loss": 2.4344,
+      "step": 3981
+    },
+    {
+      "epoch": 0.5155193060814965,
+      "grad_norm": 2.7561421328259437,
+      "learning_rate": 7.000981436420713e-07,
+      "loss": 2.582,
+      "step": 3982
+    },
+    {
+      "epoch": 0.51564876848885,
+      "grad_norm": 1.590944618509092,
+      "learning_rate": 6.998177058728917e-07,
+      "loss": 2.3236,
+      "step": 3983
+    },
+    {
+      "epoch": 0.5157782308962036,
+      "grad_norm": 1.4916695803307185,
+      "learning_rate": 6.995372528206793e-07,
+      "loss": 2.3311,
+      "step": 3984
+    },
+    {
+      "epoch": 0.515907693303557,
+      "grad_norm": 3.491227262486324,
+      "learning_rate": 6.99256784542722e-07,
+      "loss": 2.7056,
+      "step": 3985
+    },
+    {
+      "epoch": 0.515907693303557,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.550337314605713,
+      "eval_runtime": 14.7473,
+      "eval_samples_per_second": 2.984,
+      "eval_steps_per_second": 0.203,
+      "step": 3985
+    },
+    {
+      "epoch": 0.5160371557109105,
+      "grad_norm": 1.812358430861342,
+      "learning_rate": 6.989763010963116e-07,
+      "loss": 2.6284,
+      "step": 3986
+    },
+    {
+      "epoch": 0.5161666181182639,
+      "grad_norm": 3.0750105845363764,
+      "learning_rate": 6.98695802538742e-07,
+      "loss": 2.5698,
+      "step": 3987
+    },
+    {
+      "epoch": 0.5162960805256174,
+      "grad_norm": 1.6985289106885613,
+      "learning_rate": 6.984152889273111e-07,
+      "loss": 2.4294,
+      "step": 3988
+    },
+    {
+      "epoch": 0.5164255429329708,
+      "grad_norm": 1.7012894884849699,
+      "learning_rate": 6.981347603193194e-07,
+      "loss": 2.5469,
+      "step": 3989
+    },
+    {
+      "epoch": 0.5165550053403243,
+      "grad_norm": 2.6308187205415035,
+      "learning_rate": 6.978542167720705e-07,
+      "loss": 2.7463,
+      "step": 3990
+    },
+    {
+      "epoch": 0.5165550053403243,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5511142015457153,
+      "eval_runtime": 14.3667,
+      "eval_samples_per_second": 3.063,
+      "eval_steps_per_second": 0.209,
+      "step": 3990
+    },
+    {
+      "epoch": 0.5166844677476777,
+      "grad_norm": 1.2373409490242695,
+      "learning_rate": 6.97573658342871e-07,
+      "loss": 2.2629,
+      "step": 3991
+    },
+    {
+      "epoch": 0.5168139301550312,
+      "grad_norm": 2.1358560904142605,
+      "learning_rate": 6.97293085089031e-07,
+      "loss": 2.4246,
+      "step": 3992
+    },
+    {
+      "epoch": 0.5169433925623848,
+      "grad_norm": 1.7000094902465779,
+      "learning_rate": 6.97012497067863e-07,
+      "loss": 2.4238,
+      "step": 3993
+    },
+    {
+      "epoch": 0.5170728549697382,
+      "grad_norm": 2.7223489131476413,
+      "learning_rate": 6.967318943366828e-07,
+      "loss": 2.5874,
+      "step": 3994
+    },
+    {
+      "epoch": 0.5172023173770917,
+      "grad_norm": 5.732676217185138,
+      "learning_rate": 6.964512769528094e-07,
+      "loss": 3.3306,
+      "step": 3995
+    },
+    {
+      "epoch": 0.5172023173770917,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.548828125,
+      "eval_runtime": 14.9967,
+      "eval_samples_per_second": 2.934,
+      "eval_steps_per_second": 0.2,
+      "step": 3995
+    },
+    {
+      "epoch": 0.5173317797844451,
+      "grad_norm": 1.7775936567587456,
+      "learning_rate": 6.96170644973564e-07,
+      "loss": 2.6416,
+      "step": 3996
+    },
+    {
+      "epoch": 0.5174612421917986,
+      "grad_norm": 1.7078130269268268,
+      "learning_rate": 6.958899984562724e-07,
+      "loss": 2.4714,
+      "step": 3997
+    },
+    {
+      "epoch": 0.517590704599152,
+      "grad_norm": 1.9191300921989958,
+      "learning_rate": 6.956093374582614e-07,
+      "loss": 2.3521,
+      "step": 3998
+    },
+    {
+      "epoch": 0.5177201670065055,
+      "grad_norm": 1.5271119778326883,
+      "learning_rate": 6.95328662036862e-07,
+      "loss": 2.3293,
+      "step": 3999
+    },
+    {
+      "epoch": 0.5178496294138589,
+      "grad_norm": 1.501818407490641,
+      "learning_rate": 6.95047972249408e-07,
+      "loss": 2.5022,
+      "step": 4000
+    },
+    {
+      "epoch": 0.5178496294138589,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5495827198028564,
+      "eval_runtime": 14.299,
+      "eval_samples_per_second": 3.077,
+      "eval_steps_per_second": 0.21,
+      "step": 4000
     }
   ],
   "logging_steps": 1,
@@ -35742,7 +40842,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 2941030172688384.0,
+  "total_flos": 3361768083062784.0,
   "train_batch_size": 4,
   "trial_name": null,
   "trial_params": null