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": 0.32365601838366187,
+  "epoch": 0.3883872220603942,
   "eval_steps": 5,
-  "global_step": 2500,
+  "global_step": 3000,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -25523,6 +25523,5106 @@
       "eval_samples_per_second": 2.162,
       "eval_steps_per_second": 0.147,
       "step": 2500
+    },
+    {
+      "epoch": 0.3237854807910153,
+      "grad_norm": 0.1286293053646549,
+      "learning_rate": 1.0688849175068642e-06,
+      "loss": 0.7847,
+      "step": 2501
+    },
+    {
+      "epoch": 0.3239149431983688,
+      "grad_norm": 0.10727493358640168,
+      "learning_rate": 1.0686860131240251e-06,
+      "loss": 0.78,
+      "step": 2502
+    },
+    {
+      "epoch": 0.32404440560572223,
+      "grad_norm": 0.2976472329155784,
+      "learning_rate": 1.0684870181093246e-06,
+      "loss": 1.0571,
+      "step": 2503
+    },
+    {
+      "epoch": 0.3241738680130757,
+      "grad_norm": 0.12325642751187146,
+      "learning_rate": 1.0682879325034119e-06,
+      "loss": 0.8109,
+      "step": 2504
+    },
+    {
+      "epoch": 0.3243033304204292,
+      "grad_norm": 0.12365696973617851,
+      "learning_rate": 1.0680887563469537e-06,
+      "loss": 0.7683,
+      "step": 2505
+    },
+    {
+      "epoch": 0.3243033304204292,
+      "eval_PRM Accuracy": 0.7169811320754716,
+      "eval_PRM F1": 0.8351648351648352,
+      "eval_PRM F1 AUC": 0.4578313253012048,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7676767676767676,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6237571239471436,
+      "eval_runtime": 19.9083,
+      "eval_samples_per_second": 2.21,
+      "eval_steps_per_second": 0.151,
+      "step": 2505
+    },
+    {
+      "epoch": 0.32443279282778265,
+      "grad_norm": 0.12209497530594306,
+      "learning_rate": 1.0678894896806357e-06,
+      "loss": 0.7454,
+      "step": 2506
+    },
+    {
+      "epoch": 0.3245622552351361,
+      "grad_norm": 0.14609822840491737,
+      "learning_rate": 1.0676901325451628e-06,
+      "loss": 0.8041,
+      "step": 2507
+    },
+    {
+      "epoch": 0.32469171764248955,
+      "grad_norm": 0.13991505378466332,
+      "learning_rate": 1.0674906849812574e-06,
+      "loss": 0.7755,
+      "step": 2508
+    },
+    {
+      "epoch": 0.324821180049843,
+      "grad_norm": 0.12336661808318582,
+      "learning_rate": 1.0672911470296609e-06,
+      "loss": 0.7726,
+      "step": 2509
+    },
+    {
+      "epoch": 0.3249506424571965,
+      "grad_norm": 0.16165952779730758,
+      "learning_rate": 1.0670915187311326e-06,
+      "loss": 0.7122,
+      "step": 2510
+    },
+    {
+      "epoch": 0.3249506424571965,
+      "eval_PRM Accuracy": 0.7169811320754716,
+      "eval_PRM F1": 0.8351648351648352,
+      "eval_PRM F1 AUC": 0.4578313253012048,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7676767676767676,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6232244372367859,
+      "eval_runtime": 19.912,
+      "eval_samples_per_second": 2.21,
+      "eval_steps_per_second": 0.151,
+      "step": 2510
+    },
+    {
+      "epoch": 0.32508010486454997,
+      "grad_norm": 0.15164086770151672,
+      "learning_rate": 1.0668918001264508e-06,
+      "loss": 0.8653,
+      "step": 2511
+    },
+    {
+      "epoch": 0.3252095672719034,
+      "grad_norm": 0.1445976304509436,
+      "learning_rate": 1.0666919912564119e-06,
+      "loss": 0.7969,
+      "step": 2512
+    },
+    {
+      "epoch": 0.3253390296792569,
+      "grad_norm": 0.20763170818268722,
+      "learning_rate": 1.0664920921618313e-06,
+      "loss": 0.8655,
+      "step": 2513
+    },
+    {
+      "epoch": 0.32546849208661033,
+      "grad_norm": 0.12818776808593144,
+      "learning_rate": 1.0662921028835424e-06,
+      "loss": 0.7578,
+      "step": 2514
+    },
+    {
+      "epoch": 0.32559795449396384,
+      "grad_norm": 0.18409897859995156,
+      "learning_rate": 1.0660920234623968e-06,
+      "loss": 0.7402,
+      "step": 2515
+    },
+    {
+      "epoch": 0.32559795449396384,
+      "eval_PRM Accuracy": 0.7264150943396226,
+      "eval_PRM F1": 0.8415300546448088,
+      "eval_PRM F1 AUC": 0.463855421686747,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.77,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6237571239471436,
+      "eval_runtime": 20.4397,
+      "eval_samples_per_second": 2.153,
+      "eval_steps_per_second": 0.147,
+      "step": 2515
+    },
+    {
+      "epoch": 0.3257274169013173,
+      "grad_norm": 0.2316620601513271,
+      "learning_rate": 1.0658918539392645e-06,
+      "loss": 0.5781,
+      "step": 2516
+    },
+    {
+      "epoch": 0.32585687930867074,
+      "grad_norm": 0.1974934235022006,
+      "learning_rate": 1.065691594355035e-06,
+      "loss": 0.7312,
+      "step": 2517
+    },
+    {
+      "epoch": 0.3259863417160242,
+      "grad_norm": 0.16313638175543904,
+      "learning_rate": 1.065491244750615e-06,
+      "loss": 0.712,
+      "step": 2518
+    },
+    {
+      "epoch": 0.32611580412337765,
+      "grad_norm": 0.11093318196282526,
+      "learning_rate": 1.0652908051669296e-06,
+      "loss": 0.7704,
+      "step": 2519
+    },
+    {
+      "epoch": 0.32624526653073116,
+      "grad_norm": 0.10669963210051349,
+      "learning_rate": 1.065090275644923e-06,
+      "loss": 0.7666,
+      "step": 2520
+    },
+    {
+      "epoch": 0.32624526653073116,
+      "eval_PRM Accuracy": 0.7169811320754716,
+      "eval_PRM F1": 0.8351648351648352,
+      "eval_PRM F1 AUC": 0.4578313253012048,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7676767676767676,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6273082494735718,
+      "eval_runtime": 20.2669,
+      "eval_samples_per_second": 2.171,
+      "eval_steps_per_second": 0.148,
+      "step": 2520
+    },
+    {
+      "epoch": 0.3263747289380846,
+      "grad_norm": 0.11546550226086605,
+      "learning_rate": 1.0648896562255576e-06,
+      "loss": 0.7648,
+      "step": 2521
+    },
+    {
+      "epoch": 0.32650419134543807,
+      "grad_norm": 0.11850389204885863,
+      "learning_rate": 1.0646889469498137e-06,
+      "loss": 0.7339,
+      "step": 2522
+    },
+    {
+      "epoch": 0.3266336537527915,
+      "grad_norm": 0.13210643746083273,
+      "learning_rate": 1.06448814785869e-06,
+      "loss": 0.8257,
+      "step": 2523
+    },
+    {
+      "epoch": 0.326763116160145,
+      "grad_norm": 0.16349367600614878,
+      "learning_rate": 1.0642872589932044e-06,
+      "loss": 0.8529,
+      "step": 2524
+    },
+    {
+      "epoch": 0.3268925785674985,
+      "grad_norm": 0.2049814657162059,
+      "learning_rate": 1.0640862803943922e-06,
+      "loss": 0.6689,
+      "step": 2525
+    },
+    {
+      "epoch": 0.3268925785674985,
+      "eval_PRM Accuracy": 0.7169811320754716,
+      "eval_PRM F1": 0.8351648351648352,
+      "eval_PRM F1 AUC": 0.4578313253012048,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7676767676767676,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6239346861839294,
+      "eval_runtime": 19.9404,
+      "eval_samples_per_second": 2.207,
+      "eval_steps_per_second": 0.15,
+      "step": 2525
+    },
+    {
+      "epoch": 0.32702204097485194,
+      "grad_norm": 0.17167872861434297,
+      "learning_rate": 1.0638852121033071e-06,
+      "loss": 0.7202,
+      "step": 2526
+    },
+    {
+      "epoch": 0.3271515033822054,
+      "grad_norm": 0.12325501954976645,
+      "learning_rate": 1.0636840541610216e-06,
+      "loss": 0.788,
+      "step": 2527
+    },
+    {
+      "epoch": 0.32728096578955884,
+      "grad_norm": 0.10114746938236344,
+      "learning_rate": 1.0634828066086266e-06,
+      "loss": 0.6918,
+      "step": 2528
+    },
+    {
+      "epoch": 0.3274104281969123,
+      "grad_norm": 0.14420876282554568,
+      "learning_rate": 1.0632814694872306e-06,
+      "loss": 0.7489,
+      "step": 2529
+    },
+    {
+      "epoch": 0.3275398906042658,
+      "grad_norm": 0.12981954808787974,
+      "learning_rate": 1.0630800428379609e-06,
+      "loss": 0.7265,
+      "step": 2530
+    },
+    {
+      "epoch": 0.3275398906042658,
+      "eval_PRM Accuracy": 0.7169811320754716,
+      "eval_PRM F1": 0.8351648351648352,
+      "eval_PRM F1 AUC": 0.4578313253012048,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7676767676767676,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6258878111839294,
+      "eval_runtime": 19.9699,
+      "eval_samples_per_second": 2.203,
+      "eval_steps_per_second": 0.15,
+      "step": 2530
+    },
+    {
+      "epoch": 0.32766935301161926,
+      "grad_norm": 0.11207075342553162,
+      "learning_rate": 1.0628785267019631e-06,
+      "loss": 0.7076,
+      "step": 2531
+    },
+    {
+      "epoch": 0.3277988154189727,
+      "grad_norm": 0.17207524813827438,
+      "learning_rate": 1.0626769211204007e-06,
+      "loss": 0.9155,
+      "step": 2532
+    },
+    {
+      "epoch": 0.32792827782632616,
+      "grad_norm": 0.2221344887549533,
+      "learning_rate": 1.062475226134456e-06,
+      "loss": 0.5692,
+      "step": 2533
+    },
+    {
+      "epoch": 0.3280577402336796,
+      "grad_norm": 0.12633453353727667,
+      "learning_rate": 1.062273441785329e-06,
+      "loss": 0.7148,
+      "step": 2534
+    },
+    {
+      "epoch": 0.3281872026410331,
+      "grad_norm": 0.1499964540412509,
+      "learning_rate": 1.0620715681142382e-06,
+      "loss": 0.9048,
+      "step": 2535
+    },
+    {
+      "epoch": 0.3281872026410331,
+      "eval_PRM Accuracy": 0.7169811320754716,
+      "eval_PRM F1": 0.8351648351648352,
+      "eval_PRM F1 AUC": 0.4578313253012048,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7676767676767676,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6265980005264282,
+      "eval_runtime": 20.5061,
+      "eval_samples_per_second": 2.146,
+      "eval_steps_per_second": 0.146,
+      "step": 2535
+    },
+    {
+      "epoch": 0.3283166650483866,
+      "grad_norm": 0.22298546756014595,
+      "learning_rate": 1.0618696051624208e-06,
+      "loss": 1.0857,
+      "step": 2536
+    },
+    {
+      "epoch": 0.32844612745574003,
+      "grad_norm": 0.13535827907204878,
+      "learning_rate": 1.0616675529711317e-06,
+      "loss": 0.7972,
+      "step": 2537
+    },
+    {
+      "epoch": 0.3285755898630935,
+      "grad_norm": 0.15485772805981585,
+      "learning_rate": 1.0614654115816439e-06,
+      "loss": 0.67,
+      "step": 2538
+    },
+    {
+      "epoch": 0.328705052270447,
+      "grad_norm": 0.12569407768417754,
+      "learning_rate": 1.0612631810352491e-06,
+      "loss": 0.8186,
+      "step": 2539
+    },
+    {
+      "epoch": 0.32883451467780045,
+      "grad_norm": 0.12500376543993366,
+      "learning_rate": 1.0610608613732568e-06,
+      "loss": 0.7719,
+      "step": 2540
+    },
+    {
+      "epoch": 0.32883451467780045,
+      "eval_PRM Accuracy": 0.7264150943396226,
+      "eval_PRM F1": 0.8415300546448088,
+      "eval_PRM F1 AUC": 0.463855421686747,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.77,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6248224377632141,
+      "eval_runtime": 20.0515,
+      "eval_samples_per_second": 2.194,
+      "eval_steps_per_second": 0.15,
+      "step": 2540
+    },
+    {
+      "epoch": 0.3289639770851539,
+      "grad_norm": 0.10653074587312804,
+      "learning_rate": 1.060858452636995e-06,
+      "loss": 0.7354,
+      "step": 2541
+    },
+    {
+      "epoch": 0.32909343949250736,
+      "grad_norm": 0.12250437988012987,
+      "learning_rate": 1.06065595486781e-06,
+      "loss": 0.6868,
+      "step": 2542
+    },
+    {
+      "epoch": 0.3292229018998608,
+      "grad_norm": 0.14660267004828095,
+      "learning_rate": 1.0604533681070657e-06,
+      "loss": 0.7292,
+      "step": 2543
+    },
+    {
+      "epoch": 0.3293523643072143,
+      "grad_norm": 0.1408864943970288,
+      "learning_rate": 1.0602506923961447e-06,
+      "loss": 0.8792,
+      "step": 2544
+    },
+    {
+      "epoch": 0.32948182671456777,
+      "grad_norm": 0.13038386632171534,
+      "learning_rate": 1.060047927776448e-06,
+      "loss": 0.7838,
+      "step": 2545
+    },
+    {
+      "epoch": 0.32948182671456777,
+      "eval_PRM Accuracy": 0.7264150943396226,
+      "eval_PRM F1": 0.8415300546448088,
+      "eval_PRM F1 AUC": 0.463855421686747,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.77,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6235795617103577,
+      "eval_runtime": 20.4845,
+      "eval_samples_per_second": 2.148,
+      "eval_steps_per_second": 0.146,
+      "step": 2545
+    },
+    {
+      "epoch": 0.3296112891219212,
+      "grad_norm": 0.1388242810052516,
+      "learning_rate": 1.0598450742893937e-06,
+      "loss": 0.8229,
+      "step": 2546
+    },
+    {
+      "epoch": 0.3297407515292747,
+      "grad_norm": 0.12030531003783426,
+      "learning_rate": 1.0596421319764192e-06,
+      "loss": 0.7744,
+      "step": 2547
+    },
+    {
+      "epoch": 0.32987021393662813,
+      "grad_norm": 0.1236488932397587,
+      "learning_rate": 1.0594391008789791e-06,
+      "loss": 0.8347,
+      "step": 2548
+    },
+    {
+      "epoch": 0.32999967634398164,
+      "grad_norm": 0.12531832342752372,
+      "learning_rate": 1.0592359810385475e-06,
+      "loss": 0.8184,
+      "step": 2549
+    },
+    {
+      "epoch": 0.3301291387513351,
+      "grad_norm": 0.13981178835270341,
+      "learning_rate": 1.059032772496615e-06,
+      "loss": 0.8821,
+      "step": 2550
+    },
+    {
+      "epoch": 0.3301291387513351,
+      "eval_PRM Accuracy": 0.7169811320754716,
+      "eval_PRM F1": 0.8351648351648352,
+      "eval_PRM F1 AUC": 0.4578313253012048,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7676767676767676,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6241121888160706,
+      "eval_runtime": 19.907,
+      "eval_samples_per_second": 2.21,
+      "eval_steps_per_second": 0.151,
+      "step": 2550
+    },
+    {
+      "epoch": 0.33025860115868855,
+      "grad_norm": 0.20600820404196976,
+      "learning_rate": 1.0588294752946912e-06,
+      "loss": 0.6261,
+      "step": 2551
+    },
+    {
+      "epoch": 0.330388063566042,
+      "grad_norm": 0.16700591586790886,
+      "learning_rate": 1.0586260894743037e-06,
+      "loss": 0.7218,
+      "step": 2552
+    },
+    {
+      "epoch": 0.33051752597339545,
+      "grad_norm": 0.14946712208579255,
+      "learning_rate": 1.0584226150769983e-06,
+      "loss": 0.701,
+      "step": 2553
+    },
+    {
+      "epoch": 0.33064698838074896,
+      "grad_norm": 0.15750378536137175,
+      "learning_rate": 1.0582190521443388e-06,
+      "loss": 0.6202,
+      "step": 2554
+    },
+    {
+      "epoch": 0.3307764507881024,
+      "grad_norm": 0.10569303380423906,
+      "learning_rate": 1.0580154007179068e-06,
+      "loss": 0.718,
+      "step": 2555
+    },
+    {
+      "epoch": 0.3307764507881024,
+      "eval_PRM Accuracy": 0.7169811320754716,
+      "eval_PRM F1": 0.8351648351648352,
+      "eval_PRM F1 AUC": 0.4578313253012048,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7676767676767676,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6262428760528564,
+      "eval_runtime": 20.4008,
+      "eval_samples_per_second": 2.157,
+      "eval_steps_per_second": 0.147,
+      "step": 2555
+    },
+    {
+      "epoch": 0.33090591319545587,
+      "grad_norm": 0.26722142187332865,
+      "learning_rate": 1.0578116608393025e-06,
+      "loss": 0.6171,
+      "step": 2556
+    },
+    {
+      "epoch": 0.3310353756028093,
+      "grad_norm": 0.2208170882345125,
+      "learning_rate": 1.0576078325501437e-06,
+      "loss": 0.9171,
+      "step": 2557
+    },
+    {
+      "epoch": 0.3311648380101628,
+      "grad_norm": 0.15997326260438854,
+      "learning_rate": 1.0574039158920665e-06,
+      "loss": 0.6835,
+      "step": 2558
+    },
+    {
+      "epoch": 0.3312943004175163,
+      "grad_norm": 0.15284642047355682,
+      "learning_rate": 1.0571999109067252e-06,
+      "loss": 0.8569,
+      "step": 2559
+    },
+    {
+      "epoch": 0.33142376282486974,
+      "grad_norm": 0.16872619052578008,
+      "learning_rate": 1.0569958176357916e-06,
+      "loss": 0.6519,
+      "step": 2560
+    },
+    {
+      "epoch": 0.33142376282486974,
+      "eval_PRM Accuracy": 0.7169811320754716,
+      "eval_PRM F1": 0.8351648351648352,
+      "eval_PRM F1 AUC": 0.4578313253012048,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7676767676767676,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6255326867103577,
+      "eval_runtime": 19.4266,
+      "eval_samples_per_second": 2.265,
+      "eval_steps_per_second": 0.154,
+      "step": 2560
+    },
+    {
+      "epoch": 0.3315532252322232,
+      "grad_norm": 0.17243617796642635,
+      "learning_rate": 1.0567916361209562e-06,
+      "loss": 0.7642,
+      "step": 2561
+    },
+    {
+      "epoch": 0.33168268763957665,
+      "grad_norm": 0.12301840863130908,
+      "learning_rate": 1.0565873664039272e-06,
+      "loss": 0.8629,
+      "step": 2562
+    },
+    {
+      "epoch": 0.3318121500469301,
+      "grad_norm": 0.12936303240228877,
+      "learning_rate": 1.0563830085264307e-06,
+      "loss": 0.6534,
+      "step": 2563
+    },
+    {
+      "epoch": 0.3319416124542836,
+      "grad_norm": 0.16195616596317314,
+      "learning_rate": 1.056178562530211e-06,
+      "loss": 0.6738,
+      "step": 2564
+    },
+    {
+      "epoch": 0.33207107486163706,
+      "grad_norm": 0.11355581873748197,
+      "learning_rate": 1.0559740284570301e-06,
+      "loss": 0.7377,
+      "step": 2565
+    },
+    {
+      "epoch": 0.33207107486163706,
+      "eval_PRM Accuracy": 0.7169811320754716,
+      "eval_PRM F1": 0.8351648351648352,
+      "eval_PRM F1 AUC": 0.4578313253012048,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7676767676767676,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6274858117103577,
+      "eval_runtime": 20.9385,
+      "eval_samples_per_second": 2.101,
+      "eval_steps_per_second": 0.143,
+      "step": 2565
+    },
+    {
+      "epoch": 0.3322005372689905,
+      "grad_norm": 0.13996208584183203,
+      "learning_rate": 1.0557694063486685e-06,
+      "loss": 0.8141,
+      "step": 2566
+    },
+    {
+      "epoch": 0.33232999967634397,
+      "grad_norm": 0.13054174299374852,
+      "learning_rate": 1.0555646962469247e-06,
+      "loss": 0.6982,
+      "step": 2567
+    },
+    {
+      "epoch": 0.3324594620836974,
+      "grad_norm": 0.11780819763068695,
+      "learning_rate": 1.0553598981936143e-06,
+      "loss": 0.8246,
+      "step": 2568
+    },
+    {
+      "epoch": 0.33258892449105093,
+      "grad_norm": 0.1192339483173683,
+      "learning_rate": 1.0551550122305714e-06,
+      "loss": 0.6584,
+      "step": 2569
+    },
+    {
+      "epoch": 0.3327183868984044,
+      "grad_norm": 0.15337537785786137,
+      "learning_rate": 1.0549500383996487e-06,
+      "loss": 0.7429,
+      "step": 2570
+    },
+    {
+      "epoch": 0.3327183868984044,
+      "eval_PRM Accuracy": 0.7264150943396226,
+      "eval_PRM F1": 0.8415300546448088,
+      "eval_PRM F1 AUC": 0.463855421686747,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.77,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6248224377632141,
+      "eval_runtime": 19.9542,
+      "eval_samples_per_second": 2.205,
+      "eval_steps_per_second": 0.15,
+      "step": 2570
+    },
+    {
+      "epoch": 0.33284784930575784,
+      "grad_norm": 0.1121123409602909,
+      "learning_rate": 1.054744976742716e-06,
+      "loss": 0.7423,
+      "step": 2571
+    },
+    {
+      "epoch": 0.3329773117131113,
+      "grad_norm": 0.12116299423593384,
+      "learning_rate": 1.0545398273016612e-06,
+      "loss": 0.7389,
+      "step": 2572
+    },
+    {
+      "epoch": 0.33310677412046474,
+      "grad_norm": 0.12222673939487311,
+      "learning_rate": 1.0543345901183903e-06,
+      "loss": 0.7754,
+      "step": 2573
+    },
+    {
+      "epoch": 0.33323623652781825,
+      "grad_norm": 0.20383055813797735,
+      "learning_rate": 1.0541292652348273e-06,
+      "loss": 0.7181,
+      "step": 2574
+    },
+    {
+      "epoch": 0.3333656989351717,
+      "grad_norm": 0.19542916785418335,
+      "learning_rate": 1.0539238526929135e-06,
+      "loss": 0.6307,
+      "step": 2575
+    },
+    {
+      "epoch": 0.3333656989351717,
+      "eval_PRM Accuracy": 0.7264150943396226,
+      "eval_PRM F1": 0.8415300546448088,
+      "eval_PRM F1 AUC": 0.463855421686747,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.77,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6248224377632141,
+      "eval_runtime": 19.911,
+      "eval_samples_per_second": 2.21,
+      "eval_steps_per_second": 0.151,
+      "step": 2575
+    },
+    {
+      "epoch": 0.33349516134252516,
+      "grad_norm": 0.1327357600189503,
+      "learning_rate": 1.0537183525346093e-06,
+      "loss": 0.6935,
+      "step": 2576
+    },
+    {
+      "epoch": 0.3336246237498786,
+      "grad_norm": 0.10650846776750161,
+      "learning_rate": 1.0535127648018915e-06,
+      "loss": 0.7089,
+      "step": 2577
+    },
+    {
+      "epoch": 0.33375408615723207,
+      "grad_norm": 0.16704426988904944,
+      "learning_rate": 1.0533070895367562e-06,
+      "loss": 0.8016,
+      "step": 2578
+    },
+    {
+      "epoch": 0.3338835485645856,
+      "grad_norm": 0.21838144382994687,
+      "learning_rate": 1.0531013267812162e-06,
+      "loss": 0.6487,
+      "step": 2579
+    },
+    {
+      "epoch": 0.33401301097193903,
+      "grad_norm": 0.1271165463063533,
+      "learning_rate": 1.0528954765773032e-06,
+      "loss": 0.6818,
+      "step": 2580
+    },
+    {
+      "epoch": 0.33401301097193903,
+      "eval_PRM Accuracy": 0.7264150943396226,
+      "eval_PRM F1": 0.8415300546448088,
+      "eval_PRM F1 AUC": 0.463855421686747,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.77,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.623046875,
+      "eval_runtime": 20.4295,
+      "eval_samples_per_second": 2.154,
+      "eval_steps_per_second": 0.147,
+      "step": 2580
+    },
+    {
+      "epoch": 0.3341424733792925,
+      "grad_norm": 0.12195052851066047,
+      "learning_rate": 1.052689538967066e-06,
+      "loss": 0.7354,
+      "step": 2581
+    },
+    {
+      "epoch": 0.33427193578664594,
+      "grad_norm": 0.17628339737773374,
+      "learning_rate": 1.0524835139925715e-06,
+      "loss": 0.9652,
+      "step": 2582
+    },
+    {
+      "epoch": 0.33440139819399944,
+      "grad_norm": 0.10918984453422255,
+      "learning_rate": 1.0522774016959048e-06,
+      "loss": 0.7352,
+      "step": 2583
+    },
+    {
+      "epoch": 0.3345308606013529,
+      "grad_norm": 0.14462155888792924,
+      "learning_rate": 1.0520712021191682e-06,
+      "loss": 0.6984,
+      "step": 2584
+    },
+    {
+      "epoch": 0.33466032300870635,
+      "grad_norm": 0.09970203797892369,
+      "learning_rate": 1.0518649153044822e-06,
+      "loss": 0.7046,
+      "step": 2585
+    },
+    {
+      "epoch": 0.33466032300870635,
+      "eval_PRM Accuracy": 0.7264150943396226,
+      "eval_PRM F1": 0.8415300546448088,
+      "eval_PRM F1 AUC": 0.463855421686747,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.77,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6255326867103577,
+      "eval_runtime": 20.0041,
+      "eval_samples_per_second": 2.2,
+      "eval_steps_per_second": 0.15,
+      "step": 2585
+    },
+    {
+      "epoch": 0.3347897854160598,
+      "grad_norm": 0.1486874324039095,
+      "learning_rate": 1.0516585412939851e-06,
+      "loss": 0.8538,
+      "step": 2586
+    },
+    {
+      "epoch": 0.33491924782341326,
+      "grad_norm": 0.20030081755227566,
+      "learning_rate": 1.0514520801298328e-06,
+      "loss": 0.8746,
+      "step": 2587
+    },
+    {
+      "epoch": 0.33504871023076677,
+      "grad_norm": 0.09382219697559281,
+      "learning_rate": 1.0512455318541996e-06,
+      "loss": 0.6876,
+      "step": 2588
+    },
+    {
+      "epoch": 0.3351781726381202,
+      "grad_norm": 0.16940404576028745,
+      "learning_rate": 1.0510388965092767e-06,
+      "loss": 0.7438,
+      "step": 2589
+    },
+    {
+      "epoch": 0.3353076350454737,
+      "grad_norm": 0.19741779144205993,
+      "learning_rate": 1.0508321741372738e-06,
+      "loss": 0.635,
+      "step": 2590
+    },
+    {
+      "epoch": 0.3353076350454737,
+      "eval_PRM Accuracy": 0.7169811320754716,
+      "eval_PRM F1": 0.8351648351648352,
+      "eval_PRM F1 AUC": 0.4578313253012048,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7676767676767676,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6239346861839294,
+      "eval_runtime": 19.9553,
+      "eval_samples_per_second": 2.205,
+      "eval_steps_per_second": 0.15,
+      "step": 2590
+    },
+    {
+      "epoch": 0.3354370974528271,
+      "grad_norm": 0.11687311729562885,
+      "learning_rate": 1.0506253647804183e-06,
+      "loss": 0.8301,
+      "step": 2591
+    },
+    {
+      "epoch": 0.3355665598601806,
+      "grad_norm": 0.12279553823303907,
+      "learning_rate": 1.0504184684809548e-06,
+      "loss": 0.7532,
+      "step": 2592
+    },
+    {
+      "epoch": 0.3356960222675341,
+      "grad_norm": 0.19674474560532282,
+      "learning_rate": 1.0502114852811463e-06,
+      "loss": 0.8757,
+      "step": 2593
+    },
+    {
+      "epoch": 0.33582548467488754,
+      "grad_norm": 0.1186047804146949,
+      "learning_rate": 1.0500044152232734e-06,
+      "loss": 0.7018,
+      "step": 2594
+    },
+    {
+      "epoch": 0.335954947082241,
+      "grad_norm": 0.16272681071442963,
+      "learning_rate": 1.0497972583496341e-06,
+      "loss": 0.881,
+      "step": 2595
+    },
+    {
+      "epoch": 0.335954947082241,
+      "eval_PRM Accuracy": 0.7264150943396226,
+      "eval_PRM F1": 0.8415300546448088,
+      "eval_PRM F1 AUC": 0.463855421686747,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.77,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6265980005264282,
+      "eval_runtime": 20.0495,
+      "eval_samples_per_second": 2.195,
+      "eval_steps_per_second": 0.15,
+      "step": 2595
+    },
+    {
+      "epoch": 0.33608440948959445,
+      "grad_norm": 0.2139245702166506,
+      "learning_rate": 1.0495900147025449e-06,
+      "loss": 0.9753,
+      "step": 2596
+    },
+    {
+      "epoch": 0.3362138718969479,
+      "grad_norm": 0.15071971663123399,
+      "learning_rate": 1.0493826843243388e-06,
+      "loss": 0.6392,
+      "step": 2597
+    },
+    {
+      "epoch": 0.3363433343043014,
+      "grad_norm": 0.13654563750288123,
+      "learning_rate": 1.0491752672573676e-06,
+      "loss": 0.7286,
+      "step": 2598
+    },
+    {
+      "epoch": 0.33647279671165486,
+      "grad_norm": 0.20109159622083803,
+      "learning_rate": 1.0489677635440004e-06,
+      "loss": 0.8362,
+      "step": 2599
+    },
+    {
+      "epoch": 0.3366022591190083,
+      "grad_norm": 0.22332801691996676,
+      "learning_rate": 1.0487601732266242e-06,
+      "loss": 0.7119,
+      "step": 2600
+    },
+    {
+      "epoch": 0.3366022591190083,
+      "eval_PRM Accuracy": 0.7264150943396226,
+      "eval_PRM F1": 0.8415300546448088,
+      "eval_PRM F1 AUC": 0.463855421686747,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.77,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6273082494735718,
+      "eval_runtime": 20.2782,
+      "eval_samples_per_second": 2.17,
+      "eval_steps_per_second": 0.148,
+      "step": 2600
+    },
+    {
+      "epoch": 0.33673172152636177,
+      "grad_norm": 0.1925383087652379,
+      "learning_rate": 1.0485524963476433e-06,
+      "loss": 0.7003,
+      "step": 2601
+    },
+    {
+      "epoch": 0.3368611839337152,
+      "grad_norm": 0.15134354594418042,
+      "learning_rate": 1.04834473294948e-06,
+      "loss": 0.7604,
+      "step": 2602
+    },
+    {
+      "epoch": 0.33699064634106873,
+      "grad_norm": 0.14167714467495984,
+      "learning_rate": 1.0481368830745743e-06,
+      "loss": 0.7626,
+      "step": 2603
+    },
+    {
+      "epoch": 0.3371201087484222,
+      "grad_norm": 0.16759947980786186,
+      "learning_rate": 1.0479289467653836e-06,
+      "loss": 0.7953,
+      "step": 2604
+    },
+    {
+      "epoch": 0.33724957115577564,
+      "grad_norm": 0.1544954716106123,
+      "learning_rate": 1.047720924064383e-06,
+      "loss": 0.8048,
+      "step": 2605
+    },
+    {
+      "epoch": 0.33724957115577564,
+      "eval_PRM Accuracy": 0.7264150943396226,
+      "eval_PRM F1": 0.8415300546448088,
+      "eval_PRM F1 AUC": 0.463855421686747,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.77,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6260653138160706,
+      "eval_runtime": 20.1955,
+      "eval_samples_per_second": 2.179,
+      "eval_steps_per_second": 0.149,
+      "step": 2605
+    },
+    {
+      "epoch": 0.3373790335631291,
+      "grad_norm": 0.1524190436451572,
+      "learning_rate": 1.0475128150140658e-06,
+      "loss": 0.7749,
+      "step": 2606
+    },
+    {
+      "epoch": 0.33750849597048255,
+      "grad_norm": 0.17363936514132744,
+      "learning_rate": 1.0473046196569419e-06,
+      "loss": 0.9148,
+      "step": 2607
+    },
+    {
+      "epoch": 0.33763795837783606,
+      "grad_norm": 0.1937717003106368,
+      "learning_rate": 1.0470963380355398e-06,
+      "loss": 0.6096,
+      "step": 2608
+    },
+    {
+      "epoch": 0.3377674207851895,
+      "grad_norm": 0.14336489022879448,
+      "learning_rate": 1.046887970192405e-06,
+      "loss": 0.6964,
+      "step": 2609
+    },
+    {
+      "epoch": 0.33789688319254296,
+      "grad_norm": 0.20991428917306634,
+      "learning_rate": 1.0466795161701013e-06,
+      "loss": 0.6235,
+      "step": 2610
+    },
+    {
+      "epoch": 0.33789688319254296,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6248224377632141,
+      "eval_runtime": 20.36,
+      "eval_samples_per_second": 2.161,
+      "eval_steps_per_second": 0.147,
+      "step": 2610
+    },
+    {
+      "epoch": 0.3380263455998964,
+      "grad_norm": 0.12879521811422606,
+      "learning_rate": 1.0464709760112092e-06,
+      "loss": 0.8732,
+      "step": 2611
+    },
+    {
+      "epoch": 0.33815580800724987,
+      "grad_norm": 0.16017792598991287,
+      "learning_rate": 1.0462623497583274e-06,
+      "loss": 0.6876,
+      "step": 2612
+    },
+    {
+      "epoch": 0.3382852704146034,
+      "grad_norm": 0.1250461625102018,
+      "learning_rate": 1.046053637454072e-06,
+      "loss": 0.8083,
+      "step": 2613
+    },
+    {
+      "epoch": 0.33841473282195683,
+      "grad_norm": 0.1277852383755882,
+      "learning_rate": 1.045844839141077e-06,
+      "loss": 0.7278,
+      "step": 2614
+    },
+    {
+      "epoch": 0.3385441952293103,
+      "grad_norm": 0.16342899037129738,
+      "learning_rate": 1.0456359548619934e-06,
+      "loss": 0.7871,
+      "step": 2615
+    },
+    {
+      "epoch": 0.3385441952293103,
+      "eval_PRM Accuracy": 0.7264150943396226,
+      "eval_PRM F1": 0.8415300546448088,
+      "eval_PRM F1 AUC": 0.463855421686747,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.77,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6257102489471436,
+      "eval_runtime": 19.9022,
+      "eval_samples_per_second": 2.211,
+      "eval_steps_per_second": 0.151,
+      "step": 2615
+    },
+    {
+      "epoch": 0.33867365763666374,
+      "grad_norm": 0.12268385230203797,
+      "learning_rate": 1.04542698465949e-06,
+      "loss": 0.806,
+      "step": 2616
+    },
+    {
+      "epoch": 0.3388031200440172,
+      "grad_norm": 0.12212168645413116,
+      "learning_rate": 1.0452179285762534e-06,
+      "loss": 0.7314,
+      "step": 2617
+    },
+    {
+      "epoch": 0.3389325824513707,
+      "grad_norm": 0.18553815474421625,
+      "learning_rate": 1.0450087866549876e-06,
+      "loss": 0.6898,
+      "step": 2618
+    },
+    {
+      "epoch": 0.33906204485872415,
+      "grad_norm": 0.12153516795325979,
+      "learning_rate": 1.0447995589384136e-06,
+      "loss": 0.7773,
+      "step": 2619
+    },
+    {
+      "epoch": 0.3391915072660776,
+      "grad_norm": 0.20135277789850037,
+      "learning_rate": 1.044590245469271e-06,
+      "loss": 0.8654,
+      "step": 2620
+    },
+    {
+      "epoch": 0.3391915072660776,
+      "eval_PRM Accuracy": 0.7264150943396226,
+      "eval_PRM F1": 0.8415300546448088,
+      "eval_PRM F1 AUC": 0.463855421686747,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.77,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6235795617103577,
+      "eval_runtime": 19.9248,
+      "eval_samples_per_second": 2.208,
+      "eval_steps_per_second": 0.151,
+      "step": 2620
+    },
+    {
+      "epoch": 0.33932096967343106,
+      "grad_norm": 0.1918013471562724,
+      "learning_rate": 1.044380846290316e-06,
+      "loss": 0.834,
+      "step": 2621
+    },
+    {
+      "epoch": 0.33945043208078457,
+      "grad_norm": 0.14007894213389382,
+      "learning_rate": 1.0441713614443226e-06,
+      "loss": 0.7924,
+      "step": 2622
+    },
+    {
+      "epoch": 0.339579894488138,
+      "grad_norm": 0.11536178092358225,
+      "learning_rate": 1.0439617909740826e-06,
+      "loss": 0.8242,
+      "step": 2623
+    },
+    {
+      "epoch": 0.3397093568954915,
+      "grad_norm": 0.19881756739596487,
+      "learning_rate": 1.0437521349224048e-06,
+      "loss": 0.5793,
+      "step": 2624
+    },
+    {
+      "epoch": 0.33983881930284493,
+      "grad_norm": 0.17362701791786683,
+      "learning_rate": 1.0435423933321157e-06,
+      "loss": 0.8368,
+      "step": 2625
+    },
+    {
+      "epoch": 0.33983881930284493,
+      "eval_PRM Accuracy": 0.7264150943396226,
+      "eval_PRM F1": 0.8415300546448088,
+      "eval_PRM F1 AUC": 0.463855421686747,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.77,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6242897510528564,
+      "eval_runtime": 20.2982,
+      "eval_samples_per_second": 2.168,
+      "eval_steps_per_second": 0.148,
+      "step": 2625
+    },
+    {
+      "epoch": 0.3399682817101984,
+      "grad_norm": 0.12759920017620227,
+      "learning_rate": 1.043332566246059e-06,
+      "loss": 0.7639,
+      "step": 2626
+    },
+    {
+      "epoch": 0.3400977441175519,
+      "grad_norm": 0.09105113202688621,
+      "learning_rate": 1.0431226537070967e-06,
+      "loss": 0.6909,
+      "step": 2627
+    },
+    {
+      "epoch": 0.34022720652490535,
+      "grad_norm": 0.12903831865269513,
+      "learning_rate": 1.0429126557581073e-06,
+      "loss": 0.7563,
+      "step": 2628
+    },
+    {
+      "epoch": 0.3403566689322588,
+      "grad_norm": 0.11269634428287907,
+      "learning_rate": 1.0427025724419873e-06,
+      "loss": 0.8113,
+      "step": 2629
+    },
+    {
+      "epoch": 0.34048613133961225,
+      "grad_norm": 0.1334134667935212,
+      "learning_rate": 1.0424924038016505e-06,
+      "loss": 0.7981,
+      "step": 2630
+    },
+    {
+      "epoch": 0.34048613133961225,
+      "eval_PRM Accuracy": 0.7264150943396226,
+      "eval_PRM F1": 0.8415300546448088,
+      "eval_PRM F1 AUC": 0.463855421686747,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.77,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6239346861839294,
+      "eval_runtime": 20.2484,
+      "eval_samples_per_second": 2.173,
+      "eval_steps_per_second": 0.148,
+      "step": 2630
+    },
+    {
+      "epoch": 0.3406155937469657,
+      "grad_norm": 0.14078639195361398,
+      "learning_rate": 1.0422821498800282e-06,
+      "loss": 0.7513,
+      "step": 2631
+    },
+    {
+      "epoch": 0.3407450561543192,
+      "grad_norm": 0.13295704302215192,
+      "learning_rate": 1.0420718107200686e-06,
+      "loss": 0.8292,
+      "step": 2632
+    },
+    {
+      "epoch": 0.34087451856167267,
+      "grad_norm": 0.14475051624638857,
+      "learning_rate": 1.041861386364738e-06,
+      "loss": 0.8287,
+      "step": 2633
+    },
+    {
+      "epoch": 0.3410039809690261,
+      "grad_norm": 0.18888318855888467,
+      "learning_rate": 1.04165087685702e-06,
+      "loss": 0.5947,
+      "step": 2634
+    },
+    {
+      "epoch": 0.3411334433763796,
+      "grad_norm": 0.1705580010733914,
+      "learning_rate": 1.041440282239915e-06,
+      "loss": 0.7481,
+      "step": 2635
+    },
+    {
+      "epoch": 0.3411334433763796,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.62109375,
+      "eval_runtime": 20.8497,
+      "eval_samples_per_second": 2.11,
+      "eval_steps_per_second": 0.144,
+      "step": 2635
+    },
+    {
+      "epoch": 0.34126290578373303,
+      "grad_norm": 0.15600534856508574,
+      "learning_rate": 1.0412296025564417e-06,
+      "loss": 0.7549,
+      "step": 2636
+    },
+    {
+      "epoch": 0.34139236819108654,
+      "grad_norm": 0.21404375217749666,
+      "learning_rate": 1.0410188378496356e-06,
+      "loss": 0.9884,
+      "step": 2637
+    },
+    {
+      "epoch": 0.34152183059844,
+      "grad_norm": 0.11357799376180283,
+      "learning_rate": 1.0408079881625495e-06,
+      "loss": 0.7446,
+      "step": 2638
+    },
+    {
+      "epoch": 0.34165129300579344,
+      "grad_norm": 0.11406328782372947,
+      "learning_rate": 1.0405970535382535e-06,
+      "loss": 0.7598,
+      "step": 2639
+    },
+    {
+      "epoch": 0.3417807554131469,
+      "grad_norm": 0.1143277614097742,
+      "learning_rate": 1.0403860340198358e-06,
+      "loss": 0.6803,
+      "step": 2640
+    },
+    {
+      "epoch": 0.3417807554131469,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6251775622367859,
+      "eval_runtime": 19.9878,
+      "eval_samples_per_second": 2.201,
+      "eval_steps_per_second": 0.15,
+      "step": 2640
+    },
+    {
+      "epoch": 0.34191021782050035,
+      "grad_norm": 0.1325834694575373,
+      "learning_rate": 1.040174929650401e-06,
+      "loss": 0.7645,
+      "step": 2641
+    },
+    {
+      "epoch": 0.34203968022785386,
+      "grad_norm": 0.23137663727013474,
+      "learning_rate": 1.0399637404730717e-06,
+      "loss": 0.6006,
+      "step": 2642
+    },
+    {
+      "epoch": 0.3421691426352073,
+      "grad_norm": 0.09997056987710319,
+      "learning_rate": 1.0397524665309875e-06,
+      "loss": 0.7209,
+      "step": 2643
+    },
+    {
+      "epoch": 0.34229860504256077,
+      "grad_norm": 0.13317083402570234,
+      "learning_rate": 1.0395411078673053e-06,
+      "loss": 0.8613,
+      "step": 2644
+    },
+    {
+      "epoch": 0.3424280674499142,
+      "grad_norm": 0.14026283989799987,
+      "learning_rate": 1.0393296645251996e-06,
+      "loss": 0.7736,
+      "step": 2645
+    },
+    {
+      "epoch": 0.3424280674499142,
+      "eval_PRM Accuracy": 0.7264150943396226,
+      "eval_PRM F1": 0.8415300546448088,
+      "eval_PRM F1 AUC": 0.463855421686747,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.77,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6251775622367859,
+      "eval_runtime": 20.4582,
+      "eval_samples_per_second": 2.151,
+      "eval_steps_per_second": 0.147,
+      "step": 2645
+    },
+    {
+      "epoch": 0.3425575298572677,
+      "grad_norm": 0.14405732501478088,
+      "learning_rate": 1.0391181365478616e-06,
+      "loss": 0.7401,
+      "step": 2646
+    },
+    {
+      "epoch": 0.3426869922646212,
+      "grad_norm": 0.16614393788319257,
+      "learning_rate": 1.0389065239785005e-06,
+      "loss": 0.637,
+      "step": 2647
+    },
+    {
+      "epoch": 0.34281645467197464,
+      "grad_norm": 0.09485729293980032,
+      "learning_rate": 1.0386948268603426e-06,
+      "loss": 0.7268,
+      "step": 2648
+    },
+    {
+      "epoch": 0.3429459170793281,
+      "grad_norm": 0.22875285712804944,
+      "learning_rate": 1.0384830452366306e-06,
+      "loss": 1.0294,
+      "step": 2649
+    },
+    {
+      "epoch": 0.34307537948668154,
+      "grad_norm": 0.23790018258796006,
+      "learning_rate": 1.0382711791506262e-06,
+      "loss": 0.6724,
+      "step": 2650
+    },
+    {
+      "epoch": 0.34307537948668154,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6232244372367859,
+      "eval_runtime": 19.3721,
+      "eval_samples_per_second": 2.271,
+      "eval_steps_per_second": 0.155,
+      "step": 2650
+    },
+    {
+      "epoch": 0.343204841894035,
+      "grad_norm": 0.13175074564086442,
+      "learning_rate": 1.0380592286456066e-06,
+      "loss": 0.6892,
+      "step": 2651
+    },
+    {
+      "epoch": 0.3433343043013885,
+      "grad_norm": 0.14110833892011498,
+      "learning_rate": 1.0378471937648674e-06,
+      "loss": 0.7354,
+      "step": 2652
+    },
+    {
+      "epoch": 0.34346376670874196,
+      "grad_norm": 0.1429568980715398,
+      "learning_rate": 1.0376350745517206e-06,
+      "loss": 0.8157,
+      "step": 2653
+    },
+    {
+      "epoch": 0.3435932291160954,
+      "grad_norm": 0.11968003507531476,
+      "learning_rate": 1.0374228710494962e-06,
+      "loss": 0.7716,
+      "step": 2654
+    },
+    {
+      "epoch": 0.34372269152344886,
+      "grad_norm": 0.1617965113197315,
+      "learning_rate": 1.037210583301541e-06,
+      "loss": 0.7498,
+      "step": 2655
+    },
+    {
+      "epoch": 0.34372269152344886,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6281960010528564,
+      "eval_runtime": 20.052,
+      "eval_samples_per_second": 2.194,
+      "eval_steps_per_second": 0.15,
+      "step": 2655
+    },
+    {
+      "epoch": 0.3438521539308023,
+      "grad_norm": 0.20831998808657423,
+      "learning_rate": 1.036998211351219e-06,
+      "loss": 0.6411,
+      "step": 2656
+    },
+    {
+      "epoch": 0.3439816163381558,
+      "grad_norm": 0.12279500439729611,
+      "learning_rate": 1.0367857552419114e-06,
+      "loss": 0.7384,
+      "step": 2657
+    },
+    {
+      "epoch": 0.3441110787455093,
+      "grad_norm": 0.1709773300643467,
+      "learning_rate": 1.0365732150170168e-06,
+      "loss": 0.9459,
+      "step": 2658
+    },
+    {
+      "epoch": 0.34424054115286273,
+      "grad_norm": 0.2040132352795188,
+      "learning_rate": 1.036360590719951e-06,
+      "loss": 0.6287,
+      "step": 2659
+    },
+    {
+      "epoch": 0.3443700035602162,
+      "grad_norm": 0.12573185782788923,
+      "learning_rate": 1.0361478823941464e-06,
+      "loss": 0.7859,
+      "step": 2660
+    },
+    {
+      "epoch": 0.3443700035602162,
+      "eval_PRM Accuracy": 0.7264150943396226,
+      "eval_PRM F1": 0.8415300546448088,
+      "eval_PRM F1 AUC": 0.463855421686747,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.77,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6228693127632141,
+      "eval_runtime": 20.3388,
+      "eval_samples_per_second": 2.163,
+      "eval_steps_per_second": 0.148,
+      "step": 2660
+    },
+    {
+      "epoch": 0.34449946596756964,
+      "grad_norm": 0.13621373743432413,
+      "learning_rate": 1.0359350900830534e-06,
+      "loss": 0.6208,
+      "step": 2661
+    },
+    {
+      "epoch": 0.34462892837492315,
+      "grad_norm": 0.16741725297647234,
+      "learning_rate": 1.035722213830139e-06,
+      "loss": 0.687,
+      "step": 2662
+    },
+    {
+      "epoch": 0.3447583907822766,
+      "grad_norm": 0.1156327315027558,
+      "learning_rate": 1.0355092536788875e-06,
+      "loss": 0.5619,
+      "step": 2663
+    },
+    {
+      "epoch": 0.34488785318963006,
+      "grad_norm": 0.1212208220445449,
+      "learning_rate": 1.0352962096728001e-06,
+      "loss": 0.798,
+      "step": 2664
+    },
+    {
+      "epoch": 0.3450173155969835,
+      "grad_norm": 0.11820724338405045,
+      "learning_rate": 1.0350830818553957e-06,
+      "loss": 0.6941,
+      "step": 2665
+    },
+    {
+      "epoch": 0.3450173155969835,
+      "eval_PRM Accuracy": 0.7264150943396226,
+      "eval_PRM F1": 0.8415300546448088,
+      "eval_PRM F1 AUC": 0.463855421686747,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.77,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6241121888160706,
+      "eval_runtime": 19.897,
+      "eval_samples_per_second": 2.211,
+      "eval_steps_per_second": 0.151,
+      "step": 2665
+    },
+    {
+      "epoch": 0.345146778004337,
+      "grad_norm": 0.15255628485005296,
+      "learning_rate": 1.0348698702702099e-06,
+      "loss": 0.7039,
+      "step": 2666
+    },
+    {
+      "epoch": 0.34527624041169047,
+      "grad_norm": 0.12916764880859496,
+      "learning_rate": 1.0346565749607952e-06,
+      "loss": 0.7878,
+      "step": 2667
+    },
+    {
+      "epoch": 0.3454057028190439,
+      "grad_norm": 0.10036332260030756,
+      "learning_rate": 1.0344431959707219e-06,
+      "loss": 0.7538,
+      "step": 2668
+    },
+    {
+      "epoch": 0.3455351652263974,
+      "grad_norm": 0.10269489481516424,
+      "learning_rate": 1.0342297333435766e-06,
+      "loss": 0.6985,
+      "step": 2669
+    },
+    {
+      "epoch": 0.34566462763375083,
+      "grad_norm": 0.12987353301378946,
+      "learning_rate": 1.0340161871229637e-06,
+      "loss": 0.8119,
+      "step": 2670
+    },
+    {
+      "epoch": 0.34566462763375083,
+      "eval_PRM Accuracy": 0.7264150943396226,
+      "eval_PRM F1": 0.8415300546448088,
+      "eval_PRM F1 AUC": 0.463855421686747,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.77,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6234019994735718,
+      "eval_runtime": 20.5336,
+      "eval_samples_per_second": 2.143,
+      "eval_steps_per_second": 0.146,
+      "step": 2670
+    },
+    {
+      "epoch": 0.34579409004110434,
+      "grad_norm": 0.11580370508488996,
+      "learning_rate": 1.033802557352504e-06,
+      "loss": 0.7507,
+      "step": 2671
+    },
+    {
+      "epoch": 0.3459235524484578,
+      "grad_norm": 0.14533680127135218,
+      "learning_rate": 1.033588844075836e-06,
+      "loss": 0.7997,
+      "step": 2672
+    },
+    {
+      "epoch": 0.34605301485581125,
+      "grad_norm": 0.20721079088658484,
+      "learning_rate": 1.033375047336615e-06,
+      "loss": 0.6609,
+      "step": 2673
+    },
+    {
+      "epoch": 0.3461824772631647,
+      "grad_norm": 0.24362859962560357,
+      "learning_rate": 1.0331611671785132e-06,
+      "loss": 0.6876,
+      "step": 2674
+    },
+    {
+      "epoch": 0.34631193967051815,
+      "grad_norm": 0.1431266209570801,
+      "learning_rate": 1.0329472036452198e-06,
+      "loss": 0.6792,
+      "step": 2675
+    },
+    {
+      "epoch": 0.34631193967051815,
+      "eval_PRM Accuracy": 0.7264150943396226,
+      "eval_PRM F1": 0.8415300546448088,
+      "eval_PRM F1 AUC": 0.463855421686747,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.77,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6265980005264282,
+      "eval_runtime": 20.5805,
+      "eval_samples_per_second": 2.138,
+      "eval_steps_per_second": 0.146,
+      "step": 2675
+    },
+    {
+      "epoch": 0.34644140207787166,
+      "grad_norm": 0.2415249924537078,
+      "learning_rate": 1.0327331567804414e-06,
+      "loss": 0.6059,
+      "step": 2676
+    },
+    {
+      "epoch": 0.3465708644852251,
+      "grad_norm": 0.11246736554411921,
+      "learning_rate": 1.0325190266279013e-06,
+      "loss": 0.7438,
+      "step": 2677
+    },
+    {
+      "epoch": 0.34670032689257857,
+      "grad_norm": 0.23526935635338145,
+      "learning_rate": 1.0323048132313402e-06,
+      "loss": 0.8691,
+      "step": 2678
+    },
+    {
+      "epoch": 0.346829789299932,
+      "grad_norm": 0.12992693319037193,
+      "learning_rate": 1.0320905166345153e-06,
+      "loss": 0.8116,
+      "step": 2679
+    },
+    {
+      "epoch": 0.3469592517072855,
+      "grad_norm": 0.1429023472760469,
+      "learning_rate": 1.0318761368812008e-06,
+      "loss": 0.8479,
+      "step": 2680
+    },
+    {
+      "epoch": 0.3469592517072855,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6237571239471436,
+      "eval_runtime": 19.5882,
+      "eval_samples_per_second": 2.246,
+      "eval_steps_per_second": 0.153,
+      "step": 2680
+    },
+    {
+      "epoch": 0.347088714114639,
+      "grad_norm": 0.1528258583526236,
+      "learning_rate": 1.0316616740151885e-06,
+      "loss": 0.6842,
+      "step": 2681
+    },
+    {
+      "epoch": 0.34721817652199244,
+      "grad_norm": 0.16575898010334725,
+      "learning_rate": 1.0314471280802865e-06,
+      "loss": 0.6744,
+      "step": 2682
+    },
+    {
+      "epoch": 0.3473476389293459,
+      "grad_norm": 0.11636198998472357,
+      "learning_rate": 1.0312324991203202e-06,
+      "loss": 0.6772,
+      "step": 2683
+    },
+    {
+      "epoch": 0.34747710133669935,
+      "grad_norm": 0.11412395454855213,
+      "learning_rate": 1.0310177871791322e-06,
+      "loss": 0.7563,
+      "step": 2684
+    },
+    {
+      "epoch": 0.3476065637440528,
+      "grad_norm": 0.1628665221110542,
+      "learning_rate": 1.0308029923005815e-06,
+      "loss": 0.6613,
+      "step": 2685
+    },
+    {
+      "epoch": 0.3476065637440528,
+      "eval_PRM Accuracy": 0.7264150943396226,
+      "eval_PRM F1": 0.8415300546448088,
+      "eval_PRM F1 AUC": 0.463855421686747,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.77,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6242897510528564,
+      "eval_runtime": 20.3174,
+      "eval_samples_per_second": 2.166,
+      "eval_steps_per_second": 0.148,
+      "step": 2685
+    },
+    {
+      "epoch": 0.3477360261514063,
+      "grad_norm": 0.32074752419295516,
+      "learning_rate": 1.0305881145285443e-06,
+      "loss": 0.9639,
+      "step": 2686
+    },
+    {
+      "epoch": 0.34786548855875976,
+      "grad_norm": 0.16925582037262216,
+      "learning_rate": 1.030373153906914e-06,
+      "loss": 0.8098,
+      "step": 2687
+    },
+    {
+      "epoch": 0.3479949509661132,
+      "grad_norm": 0.20625428743496452,
+      "learning_rate": 1.0301581104796002e-06,
+      "loss": 0.769,
+      "step": 2688
+    },
+    {
+      "epoch": 0.34812441337346667,
+      "grad_norm": 0.14538778546930728,
+      "learning_rate": 1.0299429842905303e-06,
+      "loss": 0.6874,
+      "step": 2689
+    },
+    {
+      "epoch": 0.3482538757808201,
+      "grad_norm": 0.1918668567156437,
+      "learning_rate": 1.0297277753836482e-06,
+      "loss": 0.9476,
+      "step": 2690
+    },
+    {
+      "epoch": 0.3482538757808201,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6214488744735718,
+      "eval_runtime": 20.1661,
+      "eval_samples_per_second": 2.182,
+      "eval_steps_per_second": 0.149,
+      "step": 2690
+    },
+    {
+      "epoch": 0.34838333818817363,
+      "grad_norm": 0.14259114546137175,
+      "learning_rate": 1.0295124838029143e-06,
+      "loss": 0.7056,
+      "step": 2691
+    },
+    {
+      "epoch": 0.3485128005955271,
+      "grad_norm": 0.17073909321017092,
+      "learning_rate": 1.0292971095923065e-06,
+      "loss": 0.7964,
+      "step": 2692
+    },
+    {
+      "epoch": 0.34864226300288054,
+      "grad_norm": 0.1826840704302061,
+      "learning_rate": 1.0290816527958194e-06,
+      "loss": 0.7531,
+      "step": 2693
+    },
+    {
+      "epoch": 0.348771725410234,
+      "grad_norm": 0.11257348838214089,
+      "learning_rate": 1.0288661134574643e-06,
+      "loss": 0.7407,
+      "step": 2694
+    },
+    {
+      "epoch": 0.34890118781758744,
+      "grad_norm": 0.11395645096633086,
+      "learning_rate": 1.0286504916212697e-06,
+      "loss": 0.738,
+      "step": 2695
+    },
+    {
+      "epoch": 0.34890118781758744,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6232244372367859,
+      "eval_runtime": 20.2687,
+      "eval_samples_per_second": 2.171,
+      "eval_steps_per_second": 0.148,
+      "step": 2695
+    },
+    {
+      "epoch": 0.34903065022494095,
+      "grad_norm": 0.1817868819971395,
+      "learning_rate": 1.0284347873312804e-06,
+      "loss": 0.6821,
+      "step": 2696
+    },
+    {
+      "epoch": 0.3491601126322944,
+      "grad_norm": 0.11870942962323508,
+      "learning_rate": 1.0282190006315584e-06,
+      "loss": 0.6689,
+      "step": 2697
+    },
+    {
+      "epoch": 0.34928957503964786,
+      "grad_norm": 0.19407363199818237,
+      "learning_rate": 1.0280031315661828e-06,
+      "loss": 0.858,
+      "step": 2698
+    },
+    {
+      "epoch": 0.3494190374470013,
+      "grad_norm": 0.10666355644435606,
+      "learning_rate": 1.0277871801792493e-06,
+      "loss": 0.7075,
+      "step": 2699
+    },
+    {
+      "epoch": 0.34954849985435477,
+      "grad_norm": 0.109787623204738,
+      "learning_rate": 1.0275711465148699e-06,
+      "loss": 0.762,
+      "step": 2700
+    },
+    {
+      "epoch": 0.34954849985435477,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6244673132896423,
+      "eval_runtime": 20.7565,
+      "eval_samples_per_second": 2.12,
+      "eval_steps_per_second": 0.145,
+      "step": 2700
+    },
+    {
+      "epoch": 0.3496779622617083,
+      "grad_norm": 0.16854566995053066,
+      "learning_rate": 1.027355030617174e-06,
+      "loss": 0.6534,
+      "step": 2701
+    },
+    {
+      "epoch": 0.34980742466906173,
+      "grad_norm": 0.18780546546749913,
+      "learning_rate": 1.027138832530308e-06,
+      "loss": 0.6724,
+      "step": 2702
+    },
+    {
+      "epoch": 0.3499368870764152,
+      "grad_norm": 0.14995509366403234,
+      "learning_rate": 1.0269225522984342e-06,
+      "loss": 0.9214,
+      "step": 2703
+    },
+    {
+      "epoch": 0.35006634948376864,
+      "grad_norm": 0.12674617365309715,
+      "learning_rate": 1.0267061899657327e-06,
+      "loss": 0.738,
+      "step": 2704
+    },
+    {
+      "epoch": 0.3501958118911221,
+      "grad_norm": 0.12061226051122682,
+      "learning_rate": 1.0264897455763996e-06,
+      "loss": 0.8054,
+      "step": 2705
+    },
+    {
+      "epoch": 0.3501958118911221,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6246448755264282,
+      "eval_runtime": 20.7527,
+      "eval_samples_per_second": 2.12,
+      "eval_steps_per_second": 0.145,
+      "step": 2705
+    },
+    {
+      "epoch": 0.3503252742984756,
+      "grad_norm": 0.12638993342007726,
+      "learning_rate": 1.026273219174648e-06,
+      "loss": 0.7728,
+      "step": 2706
+    },
+    {
+      "epoch": 0.35045473670582905,
+      "grad_norm": 0.1318676914925145,
+      "learning_rate": 1.0260566108047081e-06,
+      "loss": 0.6862,
+      "step": 2707
+    },
+    {
+      "epoch": 0.3505841991131825,
+      "grad_norm": 0.11870047900732393,
+      "learning_rate": 1.0258399205108263e-06,
+      "loss": 0.6953,
+      "step": 2708
+    },
+    {
+      "epoch": 0.35071366152053596,
+      "grad_norm": 0.15554761835008327,
+      "learning_rate": 1.0256231483372662e-06,
+      "loss": 0.6822,
+      "step": 2709
+    },
+    {
+      "epoch": 0.35084312392788947,
+      "grad_norm": 0.1283686676932774,
+      "learning_rate": 1.0254062943283078e-06,
+      "loss": 0.7934,
+      "step": 2710
+    },
+    {
+      "epoch": 0.35084312392788947,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.625,
+      "eval_runtime": 20.9696,
+      "eval_samples_per_second": 2.098,
+      "eval_steps_per_second": 0.143,
+      "step": 2710
+    },
+    {
+      "epoch": 0.3509725863352429,
+      "grad_norm": 0.1387498407129159,
+      "learning_rate": 1.0251893585282477e-06,
+      "loss": 0.7303,
+      "step": 2711
+    },
+    {
+      "epoch": 0.3511020487425964,
+      "grad_norm": 0.14614957357735758,
+      "learning_rate": 1.0249723409814e-06,
+      "loss": 0.7035,
+      "step": 2712
+    },
+    {
+      "epoch": 0.3512315111499498,
+      "grad_norm": 0.12361410582688825,
+      "learning_rate": 1.0247552417320942e-06,
+      "loss": 0.6467,
+      "step": 2713
+    },
+    {
+      "epoch": 0.3513609735573033,
+      "grad_norm": 0.1366073718909575,
+      "learning_rate": 1.024538060824678e-06,
+      "loss": 0.7207,
+      "step": 2714
+    },
+    {
+      "epoch": 0.3514904359646568,
+      "grad_norm": 0.12849696107517175,
+      "learning_rate": 1.0243207983035143e-06,
+      "loss": 0.7145,
+      "step": 2715
+    },
+    {
+      "epoch": 0.3514904359646568,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6260653138160706,
+      "eval_runtime": 20.1479,
+      "eval_samples_per_second": 2.184,
+      "eval_steps_per_second": 0.149,
+      "step": 2715
+    },
+    {
+      "epoch": 0.35161989837201024,
+      "grad_norm": 0.21569020090123464,
+      "learning_rate": 1.0241034542129838e-06,
+      "loss": 0.8474,
+      "step": 2716
+    },
+    {
+      "epoch": 0.3517493607793637,
+      "grad_norm": 0.1985221261226252,
+      "learning_rate": 1.0238860285974833e-06,
+      "loss": 0.6548,
+      "step": 2717
+    },
+    {
+      "epoch": 0.35187882318671715,
+      "grad_norm": 0.12307991479943844,
+      "learning_rate": 1.0236685215014264e-06,
+      "loss": 0.8794,
+      "step": 2718
+    },
+    {
+      "epoch": 0.3520082855940706,
+      "grad_norm": 0.1299775778247633,
+      "learning_rate": 1.0234509329692432e-06,
+      "loss": 0.6559,
+      "step": 2719
+    },
+    {
+      "epoch": 0.3521377480014241,
+      "grad_norm": 0.19516247011483848,
+      "learning_rate": 1.0232332630453811e-06,
+      "loss": 0.6024,
+      "step": 2720
+    },
+    {
+      "epoch": 0.3521377480014241,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6257102489471436,
+      "eval_runtime": 20.0734,
+      "eval_samples_per_second": 2.192,
+      "eval_steps_per_second": 0.149,
+      "step": 2720
+    },
+    {
+      "epoch": 0.35226721040877756,
+      "grad_norm": 0.1347558549548527,
+      "learning_rate": 1.023015511774303e-06,
+      "loss": 0.8176,
+      "step": 2721
+    },
+    {
+      "epoch": 0.352396672816131,
+      "grad_norm": 0.13407150459822206,
+      "learning_rate": 1.022797679200489e-06,
+      "loss": 0.6418,
+      "step": 2722
+    },
+    {
+      "epoch": 0.35252613522348447,
+      "grad_norm": 0.15077833826565806,
+      "learning_rate": 1.0225797653684364e-06,
+      "loss": 0.8223,
+      "step": 2723
+    },
+    {
+      "epoch": 0.3526555976308379,
+      "grad_norm": 0.18608418341572158,
+      "learning_rate": 1.022361770322658e-06,
+      "loss": 0.8953,
+      "step": 2724
+    },
+    {
+      "epoch": 0.35278506003819143,
+      "grad_norm": 0.11848103082757616,
+      "learning_rate": 1.0221436941076835e-06,
+      "loss": 0.8021,
+      "step": 2725
+    },
+    {
+      "epoch": 0.35278506003819143,
+      "eval_PRM Accuracy": 0.7264150943396226,
+      "eval_PRM F1": 0.8415300546448088,
+      "eval_PRM F1 AUC": 0.463855421686747,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.77,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6251775622367859,
+      "eval_runtime": 20.1668,
+      "eval_samples_per_second": 2.182,
+      "eval_steps_per_second": 0.149,
+      "step": 2725
+    },
+    {
+      "epoch": 0.3529145224455449,
+      "grad_norm": 0.2116902047789001,
+      "learning_rate": 1.0219255367680603e-06,
+      "loss": 0.897,
+      "step": 2726
+    },
+    {
+      "epoch": 0.35304398485289834,
+      "grad_norm": 0.16139614503438354,
+      "learning_rate": 1.0217072983483507e-06,
+      "loss": 0.6772,
+      "step": 2727
+    },
+    {
+      "epoch": 0.3531734472602518,
+      "grad_norm": 0.22100733821146543,
+      "learning_rate": 1.0214889788931343e-06,
+      "loss": 0.963,
+      "step": 2728
+    },
+    {
+      "epoch": 0.35330290966760525,
+      "grad_norm": 0.12001734517579127,
+      "learning_rate": 1.0212705784470077e-06,
+      "loss": 0.7664,
+      "step": 2729
+    },
+    {
+      "epoch": 0.35343237207495876,
+      "grad_norm": 0.14850484282638185,
+      "learning_rate": 1.0210520970545835e-06,
+      "loss": 0.7637,
+      "step": 2730
+    },
+    {
+      "epoch": 0.35343237207495876,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6253551244735718,
+      "eval_runtime": 20.3843,
+      "eval_samples_per_second": 2.159,
+      "eval_steps_per_second": 0.147,
+      "step": 2730
+    },
+    {
+      "epoch": 0.3535618344823122,
+      "grad_norm": 0.11724829907098497,
+      "learning_rate": 1.0208335347604907e-06,
+      "loss": 0.6985,
+      "step": 2731
+    },
+    {
+      "epoch": 0.35369129688966566,
+      "grad_norm": 0.11813228026793304,
+      "learning_rate": 1.0206148916093753e-06,
+      "loss": 0.7053,
+      "step": 2732
+    },
+    {
+      "epoch": 0.3538207592970191,
+      "grad_norm": 0.11701838798036542,
+      "learning_rate": 1.0203961676458994e-06,
+      "loss": 0.7621,
+      "step": 2733
+    },
+    {
+      "epoch": 0.35395022170437257,
+      "grad_norm": 0.12572205898941735,
+      "learning_rate": 1.020177362914742e-06,
+      "loss": 0.8169,
+      "step": 2734
+    },
+    {
+      "epoch": 0.3540796841117261,
+      "grad_norm": 0.16882291578642253,
+      "learning_rate": 1.0199584774605985e-06,
+      "loss": 0.7242,
+      "step": 2735
+    },
+    {
+      "epoch": 0.3540796841117261,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6232244372367859,
+      "eval_runtime": 20.8221,
+      "eval_samples_per_second": 2.113,
+      "eval_steps_per_second": 0.144,
+      "step": 2735
+    },
+    {
+      "epoch": 0.35420914651907953,
+      "grad_norm": 0.11653044732796512,
+      "learning_rate": 1.0197395113281804e-06,
+      "loss": 0.7321,
+      "step": 2736
+    },
+    {
+      "epoch": 0.354338608926433,
+      "grad_norm": 0.11571717859652658,
+      "learning_rate": 1.0195204645622163e-06,
+      "loss": 0.7371,
+      "step": 2737
+    },
+    {
+      "epoch": 0.35446807133378644,
+      "grad_norm": 0.11754856623706685,
+      "learning_rate": 1.0193013372074507e-06,
+      "loss": 0.7966,
+      "step": 2738
+    },
+    {
+      "epoch": 0.3545975337411399,
+      "grad_norm": 0.13469327786209867,
+      "learning_rate": 1.0190821293086448e-06,
+      "loss": 0.7305,
+      "step": 2739
+    },
+    {
+      "epoch": 0.3547269961484934,
+      "grad_norm": 0.12976069946690014,
+      "learning_rate": 1.018862840910576e-06,
+      "loss": 0.6887,
+      "step": 2740
+    },
+    {
+      "epoch": 0.3547269961484934,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6246448755264282,
+      "eval_runtime": 19.6303,
+      "eval_samples_per_second": 2.241,
+      "eval_steps_per_second": 0.153,
+      "step": 2740
+    },
+    {
+      "epoch": 0.35485645855584685,
+      "grad_norm": 0.12786321279416296,
+      "learning_rate": 1.0186434720580391e-06,
+      "loss": 0.7698,
+      "step": 2741
+    },
+    {
+      "epoch": 0.3549859209632003,
+      "grad_norm": 0.10220789842983692,
+      "learning_rate": 1.0184240227958442e-06,
+      "loss": 0.7167,
+      "step": 2742
+    },
+    {
+      "epoch": 0.35511538337055376,
+      "grad_norm": 0.21902866455938091,
+      "learning_rate": 1.018204493168818e-06,
+      "loss": 0.666,
+      "step": 2743
+    },
+    {
+      "epoch": 0.3552448457779072,
+      "grad_norm": 0.16010945358101447,
+      "learning_rate": 1.0179848832218044e-06,
+      "loss": 0.6352,
+      "step": 2744
+    },
+    {
+      "epoch": 0.3553743081852607,
+      "grad_norm": 0.11510753384554984,
+      "learning_rate": 1.0177651929996632e-06,
+      "loss": 0.8212,
+      "step": 2745
+    },
+    {
+      "epoch": 0.3553743081852607,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6258878111839294,
+      "eval_runtime": 20.5043,
+      "eval_samples_per_second": 2.146,
+      "eval_steps_per_second": 0.146,
+      "step": 2745
+    },
+    {
+      "epoch": 0.3555037705926142,
+      "grad_norm": 0.1734162631080907,
+      "learning_rate": 1.01754542254727e-06,
+      "loss": 0.6626,
+      "step": 2746
+    },
+    {
+      "epoch": 0.35563323299996763,
+      "grad_norm": 0.10734341116006511,
+      "learning_rate": 1.0173255719095178e-06,
+      "loss": 0.7329,
+      "step": 2747
+    },
+    {
+      "epoch": 0.3557626954073211,
+      "grad_norm": 0.1597415965179653,
+      "learning_rate": 1.0171056411313155e-06,
+      "loss": 0.8628,
+      "step": 2748
+    },
+    {
+      "epoch": 0.35589215781467454,
+      "grad_norm": 0.1402622554858559,
+      "learning_rate": 1.0168856302575884e-06,
+      "loss": 0.7981,
+      "step": 2749
+    },
+    {
+      "epoch": 0.35602162022202805,
+      "grad_norm": 0.17858580163135207,
+      "learning_rate": 1.016665539333278e-06,
+      "loss": 0.891,
+      "step": 2750
+    },
+    {
+      "epoch": 0.35602162022202805,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6242897510528564,
+      "eval_runtime": 20.5857,
+      "eval_samples_per_second": 2.137,
+      "eval_steps_per_second": 0.146,
+      "step": 2750
+    },
+    {
+      "epoch": 0.3561510826293815,
+      "grad_norm": 0.11870390244309614,
+      "learning_rate": 1.0164453684033429e-06,
+      "loss": 0.7383,
+      "step": 2751
+    },
+    {
+      "epoch": 0.35628054503673495,
+      "grad_norm": 0.13655355148023363,
+      "learning_rate": 1.0162251175127565e-06,
+      "loss": 0.6854,
+      "step": 2752
+    },
+    {
+      "epoch": 0.3564100074440884,
+      "grad_norm": 0.11107294337546103,
+      "learning_rate": 1.0160047867065107e-06,
+      "loss": 0.7694,
+      "step": 2753
+    },
+    {
+      "epoch": 0.3565394698514419,
+      "grad_norm": 0.1132502104250624,
+      "learning_rate": 1.0157843760296114e-06,
+      "loss": 0.7491,
+      "step": 2754
+    },
+    {
+      "epoch": 0.35666893225879537,
+      "grad_norm": 0.11393725242873398,
+      "learning_rate": 1.0155638855270824e-06,
+      "loss": 0.7587,
+      "step": 2755
+    },
+    {
+      "epoch": 0.35666893225879537,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6248224377632141,
+      "eval_runtime": 20.6429,
+      "eval_samples_per_second": 2.131,
+      "eval_steps_per_second": 0.145,
+      "step": 2755
+    },
+    {
+      "epoch": 0.3567983946661488,
+      "grad_norm": 0.131522192438505,
+      "learning_rate": 1.0153433152439639e-06,
+      "loss": 0.6835,
+      "step": 2756
+    },
+    {
+      "epoch": 0.3569278570735023,
+      "grad_norm": 0.12180491173852623,
+      "learning_rate": 1.015122665225311e-06,
+      "loss": 0.7074,
+      "step": 2757
+    },
+    {
+      "epoch": 0.35705731948085573,
+      "grad_norm": 0.20500319445625226,
+      "learning_rate": 1.0149019355161964e-06,
+      "loss": 0.6676,
+      "step": 2758
+    },
+    {
+      "epoch": 0.35718678188820924,
+      "grad_norm": 0.12044705076411193,
+      "learning_rate": 1.0146811261617085e-06,
+      "loss": 0.7926,
+      "step": 2759
+    },
+    {
+      "epoch": 0.3573162442955627,
+      "grad_norm": 0.10619269377339655,
+      "learning_rate": 1.0144602372069522e-06,
+      "loss": 0.7463,
+      "step": 2760
+    },
+    {
+      "epoch": 0.3573162442955627,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6260653138160706,
+      "eval_runtime": 20.7202,
+      "eval_samples_per_second": 2.124,
+      "eval_steps_per_second": 0.145,
+      "step": 2760
+    },
+    {
+      "epoch": 0.35744570670291614,
+      "grad_norm": 0.11891466555305893,
+      "learning_rate": 1.0142392686970485e-06,
+      "loss": 0.7551,
+      "step": 2761
+    },
+    {
+      "epoch": 0.3575751691102696,
+      "grad_norm": 0.13609856749765495,
+      "learning_rate": 1.0140182206771342e-06,
+      "loss": 0.6885,
+      "step": 2762
+    },
+    {
+      "epoch": 0.35770463151762305,
+      "grad_norm": 0.11425428381603756,
+      "learning_rate": 1.0137970931923638e-06,
+      "loss": 0.7473,
+      "step": 2763
+    },
+    {
+      "epoch": 0.35783409392497656,
+      "grad_norm": 0.1582178144541826,
+      "learning_rate": 1.0135758862879064e-06,
+      "loss": 0.9011,
+      "step": 2764
+    },
+    {
+      "epoch": 0.35796355633233,
+      "grad_norm": 0.1120917242428175,
+      "learning_rate": 1.013354600008948e-06,
+      "loss": 0.7391,
+      "step": 2765
+    },
+    {
+      "epoch": 0.35796355633233,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6246448755264282,
+      "eval_runtime": 20.3676,
+      "eval_samples_per_second": 2.16,
+      "eval_steps_per_second": 0.147,
+      "step": 2765
+    },
+    {
+      "epoch": 0.35809301873968347,
+      "grad_norm": 0.15703359705665684,
+      "learning_rate": 1.0131332344006912e-06,
+      "loss": 0.7189,
+      "step": 2766
+    },
+    {
+      "epoch": 0.3582224811470369,
+      "grad_norm": 0.10909456100818825,
+      "learning_rate": 1.012911789508354e-06,
+      "loss": 0.7361,
+      "step": 2767
+    },
+    {
+      "epoch": 0.3583519435543904,
+      "grad_norm": 0.11376200325607701,
+      "learning_rate": 1.0126902653771713e-06,
+      "loss": 0.6575,
+      "step": 2768
+    },
+    {
+      "epoch": 0.3584814059617439,
+      "grad_norm": 0.1262968396491346,
+      "learning_rate": 1.0124686620523935e-06,
+      "loss": 0.8145,
+      "step": 2769
+    },
+    {
+      "epoch": 0.35861086836909734,
+      "grad_norm": 0.13926513650013556,
+      "learning_rate": 1.0122469795792882e-06,
+      "loss": 0.8274,
+      "step": 2770
+    },
+    {
+      "epoch": 0.35861086836909734,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.625,
+      "eval_runtime": 20.6053,
+      "eval_samples_per_second": 2.135,
+      "eval_steps_per_second": 0.146,
+      "step": 2770
+    },
+    {
+      "epoch": 0.3587403307764508,
+      "grad_norm": 0.15627731270756506,
+      "learning_rate": 1.012025218003138e-06,
+      "loss": 0.6954,
+      "step": 2771
+    },
+    {
+      "epoch": 0.35886979318380424,
+      "grad_norm": 0.12242087056096711,
+      "learning_rate": 1.0118033773692423e-06,
+      "loss": 0.7094,
+      "step": 2772
+    },
+    {
+      "epoch": 0.3589992555911577,
+      "grad_norm": 0.1264656957752253,
+      "learning_rate": 1.011581457722917e-06,
+      "loss": 0.7654,
+      "step": 2773
+    },
+    {
+      "epoch": 0.3591287179985112,
+      "grad_norm": 0.17999341009987485,
+      "learning_rate": 1.011359459109493e-06,
+      "loss": 0.5994,
+      "step": 2774
+    },
+    {
+      "epoch": 0.35925818040586466,
+      "grad_norm": 0.10936517829904863,
+      "learning_rate": 1.0111373815743181e-06,
+      "loss": 0.7556,
+      "step": 2775
+    },
+    {
+      "epoch": 0.35925818040586466,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6253551244735718,
+      "eval_runtime": 20.2195,
+      "eval_samples_per_second": 2.176,
+      "eval_steps_per_second": 0.148,
+      "step": 2775
+    },
+    {
+      "epoch": 0.3593876428132181,
+      "grad_norm": 0.17755414915472456,
+      "learning_rate": 1.010915225162757e-06,
+      "loss": 0.6494,
+      "step": 2776
+    },
+    {
+      "epoch": 0.35951710522057156,
+      "grad_norm": 0.1611005102659837,
+      "learning_rate": 1.0106929899201884e-06,
+      "loss": 0.7393,
+      "step": 2777
+    },
+    {
+      "epoch": 0.359646567627925,
+      "grad_norm": 0.18229186403387432,
+      "learning_rate": 1.0104706758920093e-06,
+      "loss": 0.6657,
+      "step": 2778
+    },
+    {
+      "epoch": 0.3597760300352785,
+      "grad_norm": 0.1896330364951413,
+      "learning_rate": 1.0102482831236314e-06,
+      "loss": 0.6361,
+      "step": 2779
+    },
+    {
+      "epoch": 0.359905492442632,
+      "grad_norm": 0.12466767141565187,
+      "learning_rate": 1.010025811660483e-06,
+      "loss": 0.8573,
+      "step": 2780
+    },
+    {
+      "epoch": 0.359905492442632,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.625,
+      "eval_runtime": 20.283,
+      "eval_samples_per_second": 2.169,
+      "eval_steps_per_second": 0.148,
+      "step": 2780
+    },
+    {
+      "epoch": 0.36003495484998543,
+      "grad_norm": 0.14535080032320413,
+      "learning_rate": 1.0098032615480086e-06,
+      "loss": 0.8849,
+      "step": 2781
+    },
+    {
+      "epoch": 0.3601644172573389,
+      "grad_norm": 0.1999147027160217,
+      "learning_rate": 1.0095806328316683e-06,
+      "loss": 0.6954,
+      "step": 2782
+    },
+    {
+      "epoch": 0.36029387966469234,
+      "grad_norm": 0.15927990805136685,
+      "learning_rate": 1.0093579255569387e-06,
+      "loss": 0.6552,
+      "step": 2783
+    },
+    {
+      "epoch": 0.36042334207204585,
+      "grad_norm": 0.15835118981907673,
+      "learning_rate": 1.0091351397693123e-06,
+      "loss": 0.6371,
+      "step": 2784
+    },
+    {
+      "epoch": 0.3605528044793993,
+      "grad_norm": 0.10386258069867864,
+      "learning_rate": 1.0089122755142974e-06,
+      "loss": 0.7109,
+      "step": 2785
+    },
+    {
+      "epoch": 0.3605528044793993,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6257102489471436,
+      "eval_runtime": 20.8722,
+      "eval_samples_per_second": 2.108,
+      "eval_steps_per_second": 0.144,
+      "step": 2785
+    },
+    {
+      "epoch": 0.36068226688675276,
+      "grad_norm": 0.17030186734538136,
+      "learning_rate": 1.0086893328374187e-06,
+      "loss": 0.7106,
+      "step": 2786
+    },
+    {
+      "epoch": 0.3608117292941062,
+      "grad_norm": 0.1550846234054508,
+      "learning_rate": 1.0084663117842168e-06,
+      "loss": 0.6976,
+      "step": 2787
+    },
+    {
+      "epoch": 0.36094119170145966,
+      "grad_norm": 0.2264992995774288,
+      "learning_rate": 1.0082432124002483e-06,
+      "loss": 0.8905,
+      "step": 2788
+    },
+    {
+      "epoch": 0.36107065410881317,
+      "grad_norm": 0.2007499073979965,
+      "learning_rate": 1.0080200347310853e-06,
+      "loss": 0.6783,
+      "step": 2789
+    },
+    {
+      "epoch": 0.3612001165161666,
+      "grad_norm": 0.12898495169101912,
+      "learning_rate": 1.0077967788223172e-06,
+      "loss": 0.7843,
+      "step": 2790
+    },
+    {
+      "epoch": 0.3612001165161666,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6251775622367859,
+      "eval_runtime": 20.7158,
+      "eval_samples_per_second": 2.124,
+      "eval_steps_per_second": 0.145,
+      "step": 2790
+    },
+    {
+      "epoch": 0.3613295789235201,
+      "grad_norm": 0.1550602620277481,
+      "learning_rate": 1.0075734447195478e-06,
+      "loss": 0.6628,
+      "step": 2791
+    },
+    {
+      "epoch": 0.36145904133087353,
+      "grad_norm": 0.12766404479282328,
+      "learning_rate": 1.0073500324683983e-06,
+      "loss": 0.6885,
+      "step": 2792
+    },
+    {
+      "epoch": 0.361588503738227,
+      "grad_norm": 0.12716156577780283,
+      "learning_rate": 1.0071265421145042e-06,
+      "loss": 0.696,
+      "step": 2793
+    },
+    {
+      "epoch": 0.3617179661455805,
+      "grad_norm": 0.10859967223212445,
+      "learning_rate": 1.0069029737035188e-06,
+      "loss": 0.7709,
+      "step": 2794
+    },
+    {
+      "epoch": 0.36184742855293395,
+      "grad_norm": 0.14587594387496694,
+      "learning_rate": 1.0066793272811104e-06,
+      "loss": 0.6575,
+      "step": 2795
+    },
+    {
+      "epoch": 0.36184742855293395,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6225141882896423,
+      "eval_runtime": 20.7019,
+      "eval_samples_per_second": 2.125,
+      "eval_steps_per_second": 0.145,
+      "step": 2795
+    },
+    {
+      "epoch": 0.3619768909602874,
+      "grad_norm": 0.1745755996814566,
+      "learning_rate": 1.0064556028929628e-06,
+      "loss": 0.8647,
+      "step": 2796
+    },
+    {
+      "epoch": 0.36210635336764085,
+      "grad_norm": 0.21949138484570782,
+      "learning_rate": 1.0062318005847769e-06,
+      "loss": 0.624,
+      "step": 2797
+    },
+    {
+      "epoch": 0.36223581577499436,
+      "grad_norm": 0.14046554710327563,
+      "learning_rate": 1.0060079204022681e-06,
+      "loss": 0.6584,
+      "step": 2798
+    },
+    {
+      "epoch": 0.3623652781823478,
+      "grad_norm": 0.2817723441517102,
+      "learning_rate": 1.0057839623911692e-06,
+      "loss": 0.925,
+      "step": 2799
+    },
+    {
+      "epoch": 0.36249474058970127,
+      "grad_norm": 0.11721244501914892,
+      "learning_rate": 1.0055599265972277e-06,
+      "loss": 0.5902,
+      "step": 2800
+    },
+    {
+      "epoch": 0.36249474058970127,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6253551244735718,
+      "eval_runtime": 20.7911,
+      "eval_samples_per_second": 2.116,
+      "eval_steps_per_second": 0.144,
+      "step": 2800
+    },
+    {
+      "epoch": 0.3626242029970547,
+      "grad_norm": 0.22128517155194466,
+      "learning_rate": 1.0053358130662076e-06,
+      "loss": 0.6342,
+      "step": 2801
+    },
+    {
+      "epoch": 0.3627536654044082,
+      "grad_norm": 0.17332877937342372,
+      "learning_rate": 1.0051116218438886e-06,
+      "loss": 0.5484,
+      "step": 2802
+    },
+    {
+      "epoch": 0.3628831278117617,
+      "grad_norm": 0.18114344254945527,
+      "learning_rate": 1.0048873529760663e-06,
+      "loss": 0.6136,
+      "step": 2803
+    },
+    {
+      "epoch": 0.36301259021911514,
+      "grad_norm": 0.13865993771006485,
+      "learning_rate": 1.0046630065085524e-06,
+      "loss": 0.795,
+      "step": 2804
+    },
+    {
+      "epoch": 0.3631420526264686,
+      "grad_norm": 0.19357737146631498,
+      "learning_rate": 1.0044385824871738e-06,
+      "loss": 0.8451,
+      "step": 2805
+    },
+    {
+      "epoch": 0.3631420526264686,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6246448755264282,
+      "eval_runtime": 20.4191,
+      "eval_samples_per_second": 2.155,
+      "eval_steps_per_second": 0.147,
+      "step": 2805
+    },
+    {
+      "epoch": 0.36327151503382205,
+      "grad_norm": 0.11869806826167473,
+      "learning_rate": 1.004214080957774e-06,
+      "loss": 0.6781,
+      "step": 2806
+    },
+    {
+      "epoch": 0.3634009774411755,
+      "grad_norm": 0.15712017771014644,
+      "learning_rate": 1.0039895019662115e-06,
+      "loss": 0.8055,
+      "step": 2807
+    },
+    {
+      "epoch": 0.363530439848529,
+      "grad_norm": 0.1886070857501483,
+      "learning_rate": 1.0037648455583619e-06,
+      "loss": 0.7979,
+      "step": 2808
+    },
+    {
+      "epoch": 0.36365990225588246,
+      "grad_norm": 0.09699037192668848,
+      "learning_rate": 1.0035401117801152e-06,
+      "loss": 0.6787,
+      "step": 2809
+    },
+    {
+      "epoch": 0.3637893646632359,
+      "grad_norm": 0.1301472360259722,
+      "learning_rate": 1.003315300677378e-06,
+      "loss": 0.8563,
+      "step": 2810
+    },
+    {
+      "epoch": 0.3637893646632359,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6251775622367859,
+      "eval_runtime": 20.73,
+      "eval_samples_per_second": 2.123,
+      "eval_steps_per_second": 0.145,
+      "step": 2810
+    },
+    {
+      "epoch": 0.36391882707058937,
+      "grad_norm": 0.17629520010489655,
+      "learning_rate": 1.0030904122960725e-06,
+      "loss": 0.6196,
+      "step": 2811
+    },
+    {
+      "epoch": 0.3640482894779428,
+      "grad_norm": 0.17110750514810258,
+      "learning_rate": 1.002865446682137e-06,
+      "loss": 0.8717,
+      "step": 2812
+    },
+    {
+      "epoch": 0.36417775188529633,
+      "grad_norm": 0.11661881615758461,
+      "learning_rate": 1.0026404038815246e-06,
+      "loss": 0.6324,
+      "step": 2813
+    },
+    {
+      "epoch": 0.3643072142926498,
+      "grad_norm": 0.17321263941495776,
+      "learning_rate": 1.0024152839402055e-06,
+      "loss": 0.6805,
+      "step": 2814
+    },
+    {
+      "epoch": 0.36443667670000324,
+      "grad_norm": 0.18243577324490526,
+      "learning_rate": 1.0021900869041648e-06,
+      "loss": 0.9711,
+      "step": 2815
+    },
+    {
+      "epoch": 0.36443667670000324,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6253551244735718,
+      "eval_runtime": 20.174,
+      "eval_samples_per_second": 2.181,
+      "eval_steps_per_second": 0.149,
+      "step": 2815
+    },
+    {
+      "epoch": 0.3645661391073567,
+      "grad_norm": 0.35969707544601937,
+      "learning_rate": 1.0019648128194035e-06,
+      "loss": 1.0547,
+      "step": 2816
+    },
+    {
+      "epoch": 0.36469560151471014,
+      "grad_norm": 0.15667076978530703,
+      "learning_rate": 1.0017394617319387e-06,
+      "loss": 0.8734,
+      "step": 2817
+    },
+    {
+      "epoch": 0.36482506392206365,
+      "grad_norm": 0.1583859620711222,
+      "learning_rate": 1.0015140336878024e-06,
+      "loss": 0.6794,
+      "step": 2818
+    },
+    {
+      "epoch": 0.3649545263294171,
+      "grad_norm": 0.13531924142054677,
+      "learning_rate": 1.0012885287330432e-06,
+      "loss": 0.8154,
+      "step": 2819
+    },
+    {
+      "epoch": 0.36508398873677056,
+      "grad_norm": 0.23001703415401786,
+      "learning_rate": 1.001062946913725e-06,
+      "loss": 0.9592,
+      "step": 2820
+    },
+    {
+      "epoch": 0.36508398873677056,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6255326867103577,
+      "eval_runtime": 20.3471,
+      "eval_samples_per_second": 2.162,
+      "eval_steps_per_second": 0.147,
+      "step": 2820
+    },
+    {
+      "epoch": 0.365213451144124,
+      "grad_norm": 0.10967483737984751,
+      "learning_rate": 1.0008372882759275e-06,
+      "loss": 0.6429,
+      "step": 2821
+    },
+    {
+      "epoch": 0.36534291355147747,
+      "grad_norm": 0.18289014759832528,
+      "learning_rate": 1.000611552865746e-06,
+      "loss": 0.7673,
+      "step": 2822
+    },
+    {
+      "epoch": 0.365472375958831,
+      "grad_norm": 0.11244747425236269,
+      "learning_rate": 1.0003857407292913e-06,
+      "loss": 0.7706,
+      "step": 2823
+    },
+    {
+      "epoch": 0.36560183836618443,
+      "grad_norm": 0.10748853614266381,
+      "learning_rate": 1.0001598519126906e-06,
+      "loss": 0.7705,
+      "step": 2824
+    },
+    {
+      "epoch": 0.3657313007735379,
+      "grad_norm": 0.18569021687321619,
+      "learning_rate": 9.999338864620857e-07,
+      "loss": 0.6119,
+      "step": 2825
+    },
+    {
+      "epoch": 0.3657313007735379,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.623046875,
+      "eval_runtime": 20.6616,
+      "eval_samples_per_second": 2.13,
+      "eval_steps_per_second": 0.145,
+      "step": 2825
+    },
+    {
+      "epoch": 0.36586076318089134,
+      "grad_norm": 0.1404997605463703,
+      "learning_rate": 9.997078444236353e-07,
+      "loss": 0.8663,
+      "step": 2826
+    },
+    {
+      "epoch": 0.3659902255882448,
+      "grad_norm": 0.12635455672583473,
+      "learning_rate": 9.994817258435125e-07,
+      "loss": 0.745,
+      "step": 2827
+    },
+    {
+      "epoch": 0.3661196879955983,
+      "grad_norm": 0.19570866842944618,
+      "learning_rate": 9.992555307679069e-07,
+      "loss": 0.6388,
+      "step": 2828
+    },
+    {
+      "epoch": 0.36624915040295175,
+      "grad_norm": 0.13974682735196295,
+      "learning_rate": 9.990292592430233e-07,
+      "loss": 0.7522,
+      "step": 2829
+    },
+    {
+      "epoch": 0.3663786128103052,
+      "grad_norm": 0.12582437913045555,
+      "learning_rate": 9.98802911315082e-07,
+      "loss": 0.7726,
+      "step": 2830
+    },
+    {
+      "epoch": 0.3663786128103052,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6207386255264282,
+      "eval_runtime": 19.9816,
+      "eval_samples_per_second": 2.202,
+      "eval_steps_per_second": 0.15,
+      "step": 2830
+    },
+    {
+      "epoch": 0.36650807521765866,
+      "grad_norm": 0.12879098767911193,
+      "learning_rate": 9.9857648703032e-07,
+      "loss": 0.7976,
+      "step": 2831
+    },
+    {
+      "epoch": 0.3666375376250121,
+      "grad_norm": 0.1113684299180072,
+      "learning_rate": 9.98349986434988e-07,
+      "loss": 0.6083,
+      "step": 2832
+    },
+    {
+      "epoch": 0.3667670000323656,
+      "grad_norm": 0.20367847289459148,
+      "learning_rate": 9.981234095753538e-07,
+      "loss": 0.741,
+      "step": 2833
+    },
+    {
+      "epoch": 0.3668964624397191,
+      "grad_norm": 0.16565516350602624,
+      "learning_rate": 9.978967564977006e-07,
+      "loss": 0.6956,
+      "step": 2834
+    },
+    {
+      "epoch": 0.3670259248470725,
+      "grad_norm": 0.149223061260409,
+      "learning_rate": 9.976700272483263e-07,
+      "loss": 0.8771,
+      "step": 2835
+    },
+    {
+      "epoch": 0.3670259248470725,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6255326867103577,
+      "eval_runtime": 19.825,
+      "eval_samples_per_second": 2.219,
+      "eval_steps_per_second": 0.151,
+      "step": 2835
+    },
+    {
+      "epoch": 0.367155387254426,
+      "grad_norm": 0.1368947890017426,
+      "learning_rate": 9.974432218735455e-07,
+      "loss": 0.8422,
+      "step": 2836
+    },
+    {
+      "epoch": 0.36728484966177943,
+      "grad_norm": 0.18992851341581415,
+      "learning_rate": 9.972163404196876e-07,
+      "loss": 0.8458,
+      "step": 2837
+    },
+    {
+      "epoch": 0.36741431206913294,
+      "grad_norm": 0.1020817477929157,
+      "learning_rate": 9.969893829330976e-07,
+      "loss": 0.7202,
+      "step": 2838
+    },
+    {
+      "epoch": 0.3675437744764864,
+      "grad_norm": 0.22575054944290548,
+      "learning_rate": 9.96762349460136e-07,
+      "loss": 0.693,
+      "step": 2839
+    },
+    {
+      "epoch": 0.36767323688383985,
+      "grad_norm": 0.149959951158413,
+      "learning_rate": 9.965352400471796e-07,
+      "loss": 0.7114,
+      "step": 2840
+    },
+    {
+      "epoch": 0.36767323688383985,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6242897510528564,
+      "eval_runtime": 20.3143,
+      "eval_samples_per_second": 2.166,
+      "eval_steps_per_second": 0.148,
+      "step": 2840
+    },
+    {
+      "epoch": 0.3678026992911933,
+      "grad_norm": 0.2256009121617485,
+      "learning_rate": 9.963080547406194e-07,
+      "loss": 0.8301,
+      "step": 2841
+    },
+    {
+      "epoch": 0.3679321616985468,
+      "grad_norm": 0.10723478803848362,
+      "learning_rate": 9.960807935868634e-07,
+      "loss": 0.7053,
+      "step": 2842
+    },
+    {
+      "epoch": 0.36806162410590026,
+      "grad_norm": 0.130704032270869,
+      "learning_rate": 9.958534566323336e-07,
+      "loss": 0.5854,
+      "step": 2843
+    },
+    {
+      "epoch": 0.3681910865132537,
+      "grad_norm": 0.13733285112940952,
+      "learning_rate": 9.956260439234683e-07,
+      "loss": 0.7458,
+      "step": 2844
+    },
+    {
+      "epoch": 0.36832054892060717,
+      "grad_norm": 0.10798055070264732,
+      "learning_rate": 9.953985555067216e-07,
+      "loss": 0.6639,
+      "step": 2845
+    },
+    {
+      "epoch": 0.36832054892060717,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6218039989471436,
+      "eval_runtime": 19.2582,
+      "eval_samples_per_second": 2.285,
+      "eval_steps_per_second": 0.156,
+      "step": 2845
+    },
+    {
+      "epoch": 0.3684500113279606,
+      "grad_norm": 0.17381986305454794,
+      "learning_rate": 9.95170991428562e-07,
+      "loss": 0.6963,
+      "step": 2846
+    },
+    {
+      "epoch": 0.36857947373531413,
+      "grad_norm": 0.11041443323013606,
+      "learning_rate": 9.949433517354745e-07,
+      "loss": 0.8016,
+      "step": 2847
+    },
+    {
+      "epoch": 0.3687089361426676,
+      "grad_norm": 0.1342537235632201,
+      "learning_rate": 9.947156364739591e-07,
+      "loss": 0.7283,
+      "step": 2848
+    },
+    {
+      "epoch": 0.36883839855002104,
+      "grad_norm": 0.11340855921890433,
+      "learning_rate": 9.94487845690531e-07,
+      "loss": 0.8123,
+      "step": 2849
+    },
+    {
+      "epoch": 0.3689678609573745,
+      "grad_norm": 0.1148836825914116,
+      "learning_rate": 9.942599794317212e-07,
+      "loss": 0.7489,
+      "step": 2850
+    },
+    {
+      "epoch": 0.3689678609573745,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6242897510528564,
+      "eval_runtime": 20.0357,
+      "eval_samples_per_second": 2.196,
+      "eval_steps_per_second": 0.15,
+      "step": 2850
+    },
+    {
+      "epoch": 0.36909732336472795,
+      "grad_norm": 0.11598163033572025,
+      "learning_rate": 9.940320377440763e-07,
+      "loss": 0.7452,
+      "step": 2851
+    },
+    {
+      "epoch": 0.36922678577208146,
+      "grad_norm": 0.18576733544059618,
+      "learning_rate": 9.938040206741574e-07,
+      "loss": 0.9213,
+      "step": 2852
+    },
+    {
+      "epoch": 0.3693562481794349,
+      "grad_norm": 0.1541508092868443,
+      "learning_rate": 9.93575928268542e-07,
+      "loss": 0.7845,
+      "step": 2853
+    },
+    {
+      "epoch": 0.36948571058678836,
+      "grad_norm": 0.13106887946643705,
+      "learning_rate": 9.933477605738228e-07,
+      "loss": 0.8201,
+      "step": 2854
+    },
+    {
+      "epoch": 0.3696151729941418,
+      "grad_norm": 0.12708264906921904,
+      "learning_rate": 9.931195176366073e-07,
+      "loss": 0.7166,
+      "step": 2855
+    },
+    {
+      "epoch": 0.3696151729941418,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6214488744735718,
+      "eval_runtime": 20.9235,
+      "eval_samples_per_second": 2.103,
+      "eval_steps_per_second": 0.143,
+      "step": 2855
+    },
+    {
+      "epoch": 0.36974463540149527,
+      "grad_norm": 0.20759925584505856,
+      "learning_rate": 9.92891199503519e-07,
+      "loss": 0.9233,
+      "step": 2856
+    },
+    {
+      "epoch": 0.3698740978088488,
+      "grad_norm": 0.1182791972346759,
+      "learning_rate": 9.926628062211963e-07,
+      "loss": 0.783,
+      "step": 2857
+    },
+    {
+      "epoch": 0.37000356021620223,
+      "grad_norm": 0.19592505470800373,
+      "learning_rate": 9.924343378362934e-07,
+      "loss": 0.6206,
+      "step": 2858
+    },
+    {
+      "epoch": 0.3701330226235557,
+      "grad_norm": 0.09279704123529552,
+      "learning_rate": 9.92205794395479e-07,
+      "loss": 0.676,
+      "step": 2859
+    },
+    {
+      "epoch": 0.37026248503090914,
+      "grad_norm": 0.13837394202458592,
+      "learning_rate": 9.919771759454386e-07,
+      "loss": 0.7532,
+      "step": 2860
+    },
+    {
+      "epoch": 0.37026248503090914,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6228693127632141,
+      "eval_runtime": 20.4147,
+      "eval_samples_per_second": 2.155,
+      "eval_steps_per_second": 0.147,
+      "step": 2860
+    },
+    {
+      "epoch": 0.3703919474382626,
+      "grad_norm": 0.2749197806110179,
+      "learning_rate": 9.917484825328717e-07,
+      "loss": 0.9917,
+      "step": 2861
+    },
+    {
+      "epoch": 0.3705214098456161,
+      "grad_norm": 0.1316747837115949,
+      "learning_rate": 9.915197142044933e-07,
+      "loss": 0.7505,
+      "step": 2862
+    },
+    {
+      "epoch": 0.37065087225296955,
+      "grad_norm": 0.08914873041230541,
+      "learning_rate": 9.912908710070344e-07,
+      "loss": 0.6862,
+      "step": 2863
+    },
+    {
+      "epoch": 0.370780334660323,
+      "grad_norm": 0.15896317627144219,
+      "learning_rate": 9.910619529872408e-07,
+      "loss": 0.8815,
+      "step": 2864
+    },
+    {
+      "epoch": 0.37090979706767646,
+      "grad_norm": 0.13847414093318494,
+      "learning_rate": 9.908329601918735e-07,
+      "loss": 0.7295,
+      "step": 2865
+    },
+    {
+      "epoch": 0.37090979706767646,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6244673132896423,
+      "eval_runtime": 20.6997,
+      "eval_samples_per_second": 2.126,
+      "eval_steps_per_second": 0.145,
+      "step": 2865
+    },
+    {
+      "epoch": 0.3710392594750299,
+      "grad_norm": 0.133110081768871,
+      "learning_rate": 9.90603892667709e-07,
+      "loss": 0.7982,
+      "step": 2866
+    },
+    {
+      "epoch": 0.3711687218823834,
+      "grad_norm": 0.1386340023273188,
+      "learning_rate": 9.903747504615391e-07,
+      "loss": 0.8813,
+      "step": 2867
+    },
+    {
+      "epoch": 0.3712981842897369,
+      "grad_norm": 0.12937511884512193,
+      "learning_rate": 9.901455336201704e-07,
+      "loss": 0.7798,
+      "step": 2868
+    },
+    {
+      "epoch": 0.37142764669709033,
+      "grad_norm": 0.10752945915865027,
+      "learning_rate": 9.899162421904254e-07,
+      "loss": 0.7404,
+      "step": 2869
+    },
+    {
+      "epoch": 0.3715571091044438,
+      "grad_norm": 0.19242984857130088,
+      "learning_rate": 9.896868762191415e-07,
+      "loss": 0.6317,
+      "step": 2870
+    },
+    {
+      "epoch": 0.3715571091044438,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.619140625,
+      "eval_runtime": 21.217,
+      "eval_samples_per_second": 2.074,
+      "eval_steps_per_second": 0.141,
+      "step": 2870
+    },
+    {
+      "epoch": 0.37168657151179724,
+      "grad_norm": 0.16927344739673844,
+      "learning_rate": 9.894574357531708e-07,
+      "loss": 0.6477,
+      "step": 2871
+    },
+    {
+      "epoch": 0.37181603391915075,
+      "grad_norm": 0.17839618971087978,
+      "learning_rate": 9.892279208393822e-07,
+      "loss": 0.8459,
+      "step": 2872
+    },
+    {
+      "epoch": 0.3719454963265042,
+      "grad_norm": 0.16120500834577478,
+      "learning_rate": 9.889983315246579e-07,
+      "loss": 0.7805,
+      "step": 2873
+    },
+    {
+      "epoch": 0.37207495873385765,
+      "grad_norm": 0.12044896850124229,
+      "learning_rate": 9.887686678558966e-07,
+      "loss": 0.7936,
+      "step": 2874
+    },
+    {
+      "epoch": 0.3722044211412111,
+      "grad_norm": 0.196596106119202,
+      "learning_rate": 9.885389298800116e-07,
+      "loss": 0.8745,
+      "step": 2875
+    },
+    {
+      "epoch": 0.3722044211412111,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6209161877632141,
+      "eval_runtime": 20.5102,
+      "eval_samples_per_second": 2.145,
+      "eval_steps_per_second": 0.146,
+      "step": 2875
+    },
+    {
+      "epoch": 0.37233388354856456,
+      "grad_norm": 0.10856748403868338,
+      "learning_rate": 9.883091176439317e-07,
+      "loss": 0.754,
+      "step": 2876
+    },
+    {
+      "epoch": 0.37246334595591807,
+      "grad_norm": 0.10659709455841847,
+      "learning_rate": 9.880792311946002e-07,
+      "loss": 0.7786,
+      "step": 2877
+    },
+    {
+      "epoch": 0.3725928083632715,
+      "grad_norm": 0.10193075910208234,
+      "learning_rate": 9.878492705789768e-07,
+      "loss": 0.703,
+      "step": 2878
+    },
+    {
+      "epoch": 0.372722270770625,
+      "grad_norm": 0.27018595012761365,
+      "learning_rate": 9.876192358440352e-07,
+      "loss": 0.8665,
+      "step": 2879
+    },
+    {
+      "epoch": 0.37285173317797843,
+      "grad_norm": 0.2192297338242662,
+      "learning_rate": 9.873891270367647e-07,
+      "loss": 0.8121,
+      "step": 2880
+    },
+    {
+      "epoch": 0.37285173317797843,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.62109375,
+      "eval_runtime": 19.8857,
+      "eval_samples_per_second": 2.213,
+      "eval_steps_per_second": 0.151,
+      "step": 2880
+    },
+    {
+      "epoch": 0.3729811955853319,
+      "grad_norm": 0.10687442096021815,
+      "learning_rate": 9.871589442041697e-07,
+      "loss": 0.6902,
+      "step": 2881
+    },
+    {
+      "epoch": 0.3731106579926854,
+      "grad_norm": 0.1190306321432561,
+      "learning_rate": 9.869286873932696e-07,
+      "loss": 0.8167,
+      "step": 2882
+    },
+    {
+      "epoch": 0.37324012040003884,
+      "grad_norm": 0.11254228096620711,
+      "learning_rate": 9.866983566510994e-07,
+      "loss": 0.6722,
+      "step": 2883
+    },
+    {
+      "epoch": 0.3733695828073923,
+      "grad_norm": 0.20083797984382643,
+      "learning_rate": 9.864679520247084e-07,
+      "loss": 0.8613,
+      "step": 2884
+    },
+    {
+      "epoch": 0.37349904521474575,
+      "grad_norm": 0.13950813833990097,
+      "learning_rate": 9.862374735611615e-07,
+      "loss": 0.6549,
+      "step": 2885
+    },
+    {
+      "epoch": 0.37349904521474575,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6202059388160706,
+      "eval_runtime": 20.0603,
+      "eval_samples_per_second": 2.193,
+      "eval_steps_per_second": 0.15,
+      "step": 2885
+    },
+    {
+      "epoch": 0.37362850762209926,
+      "grad_norm": 0.10845310820567529,
+      "learning_rate": 9.860069213075388e-07,
+      "loss": 0.7228,
+      "step": 2886
+    },
+    {
+      "epoch": 0.3737579700294527,
+      "grad_norm": 0.29280655580377535,
+      "learning_rate": 9.85776295310935e-07,
+      "loss": 1.0416,
+      "step": 2887
+    },
+    {
+      "epoch": 0.37388743243680617,
+      "grad_norm": 0.1732519069824405,
+      "learning_rate": 9.855455956184603e-07,
+      "loss": 0.9031,
+      "step": 2888
+    },
+    {
+      "epoch": 0.3740168948441596,
+      "grad_norm": 0.1442694528254996,
+      "learning_rate": 9.853148222772401e-07,
+      "loss": 0.6379,
+      "step": 2889
+    },
+    {
+      "epoch": 0.3741463572515131,
+      "grad_norm": 0.210322397461802,
+      "learning_rate": 9.85083975334414e-07,
+      "loss": 0.5677,
+      "step": 2890
+    },
+    {
+      "epoch": 0.3741463572515131,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6193181872367859,
+      "eval_runtime": 20.1002,
+      "eval_samples_per_second": 2.189,
+      "eval_steps_per_second": 0.149,
+      "step": 2890
+    },
+    {
+      "epoch": 0.3742758196588666,
+      "grad_norm": 0.13601654045992892,
+      "learning_rate": 9.848530548371373e-07,
+      "loss": 0.6717,
+      "step": 2891
+    },
+    {
+      "epoch": 0.37440528206622004,
+      "grad_norm": 0.161632015639682,
+      "learning_rate": 9.846220608325803e-07,
+      "loss": 0.8059,
+      "step": 2892
+    },
+    {
+      "epoch": 0.3745347444735735,
+      "grad_norm": 0.14065969069575485,
+      "learning_rate": 9.843909933679281e-07,
+      "loss": 0.6122,
+      "step": 2893
+    },
+    {
+      "epoch": 0.37466420688092694,
+      "grad_norm": 0.15436419168454768,
+      "learning_rate": 9.84159852490381e-07,
+      "loss": 0.7066,
+      "step": 2894
+    },
+    {
+      "epoch": 0.3747936692882804,
+      "grad_norm": 0.10746396080468937,
+      "learning_rate": 9.839286382471544e-07,
+      "loss": 0.676,
+      "step": 2895
+    },
+    {
+      "epoch": 0.3747936692882804,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6221590638160706,
+      "eval_runtime": 20.844,
+      "eval_samples_per_second": 2.111,
+      "eval_steps_per_second": 0.144,
+      "step": 2895
+    },
+    {
+      "epoch": 0.3749231316956339,
+      "grad_norm": 0.1424428802203209,
+      "learning_rate": 9.83697350685478e-07,
+      "loss": 0.6363,
+      "step": 2896
+    },
+    {
+      "epoch": 0.37505259410298736,
+      "grad_norm": 0.17816174444259716,
+      "learning_rate": 9.834659898525973e-07,
+      "loss": 0.7845,
+      "step": 2897
+    },
+    {
+      "epoch": 0.3751820565103408,
+      "grad_norm": 0.14746849829737738,
+      "learning_rate": 9.832345557957727e-07,
+      "loss": 0.7157,
+      "step": 2898
+    },
+    {
+      "epoch": 0.37531151891769426,
+      "grad_norm": 0.12054749979829008,
+      "learning_rate": 9.830030485622787e-07,
+      "loss": 0.6608,
+      "step": 2899
+    },
+    {
+      "epoch": 0.3754409813250477,
+      "grad_norm": 0.12373977935547659,
+      "learning_rate": 9.827714681994055e-07,
+      "loss": 0.7675,
+      "step": 2900
+    },
+    {
+      "epoch": 0.3754409813250477,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6198508739471436,
+      "eval_runtime": 20.7095,
+      "eval_samples_per_second": 2.125,
+      "eval_steps_per_second": 0.145,
+      "step": 2900
+    },
+    {
+      "epoch": 0.3755704437324012,
+      "grad_norm": 0.1025071281951921,
+      "learning_rate": 9.825398147544585e-07,
+      "loss": 0.686,
+      "step": 2901
+    },
+    {
+      "epoch": 0.3756999061397547,
+      "grad_norm": 0.13855938622203118,
+      "learning_rate": 9.823080882747574e-07,
+      "loss": 0.8522,
+      "step": 2902
+    },
+    {
+      "epoch": 0.37582936854710813,
+      "grad_norm": 0.2007044350243728,
+      "learning_rate": 9.820762888076369e-07,
+      "loss": 0.8264,
+      "step": 2903
+    },
+    {
+      "epoch": 0.3759588309544616,
+      "grad_norm": 0.14910523321869876,
+      "learning_rate": 9.818444164004465e-07,
+      "loss": 0.8308,
+      "step": 2904
+    },
+    {
+      "epoch": 0.37608829336181504,
+      "grad_norm": 0.14577532600215753,
+      "learning_rate": 9.816124711005514e-07,
+      "loss": 0.7068,
+      "step": 2905
+    },
+    {
+      "epoch": 0.37608829336181504,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.619140625,
+      "eval_runtime": 20.3865,
+      "eval_samples_per_second": 2.158,
+      "eval_steps_per_second": 0.147,
+      "step": 2905
+    },
+    {
+      "epoch": 0.37621775576916855,
+      "grad_norm": 0.10747425253265232,
+      "learning_rate": 9.81380452955331e-07,
+      "loss": 0.7368,
+      "step": 2906
+    },
+    {
+      "epoch": 0.376347218176522,
+      "grad_norm": 0.16717792766842757,
+      "learning_rate": 9.811483620121796e-07,
+      "loss": 0.6138,
+      "step": 2907
+    },
+    {
+      "epoch": 0.37647668058387546,
+      "grad_norm": 0.1391080390369382,
+      "learning_rate": 9.80916198318506e-07,
+      "loss": 0.7,
+      "step": 2908
+    },
+    {
+      "epoch": 0.3766061429912289,
+      "grad_norm": 0.17051142695814003,
+      "learning_rate": 9.806839619217353e-07,
+      "loss": 0.6503,
+      "step": 2909
+    },
+    {
+      "epoch": 0.37673560539858236,
+      "grad_norm": 0.13019302956333167,
+      "learning_rate": 9.804516528693056e-07,
+      "loss": 0.8029,
+      "step": 2910
+    },
+    {
+      "epoch": 0.37673560539858236,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6207386255264282,
+      "eval_runtime": 20.7218,
+      "eval_samples_per_second": 2.123,
+      "eval_steps_per_second": 0.145,
+      "step": 2910
+    },
+    {
+      "epoch": 0.37686506780593587,
+      "grad_norm": 0.18289266796873432,
+      "learning_rate": 9.802192712086713e-07,
+      "loss": 0.8928,
+      "step": 2911
+    },
+    {
+      "epoch": 0.3769945302132893,
+      "grad_norm": 0.16085694768119557,
+      "learning_rate": 9.799868169873011e-07,
+      "loss": 0.9163,
+      "step": 2912
+    },
+    {
+      "epoch": 0.3771239926206428,
+      "grad_norm": 0.13196289845834508,
+      "learning_rate": 9.797542902526779e-07,
+      "loss": 0.6727,
+      "step": 2913
+    },
+    {
+      "epoch": 0.37725345502799623,
+      "grad_norm": 0.18268665637029438,
+      "learning_rate": 9.795216910523006e-07,
+      "loss": 0.6888,
+      "step": 2914
+    },
+    {
+      "epoch": 0.3773829174353497,
+      "grad_norm": 0.11628970806080995,
+      "learning_rate": 9.79289019433682e-07,
+      "loss": 0.7185,
+      "step": 2915
+    },
+    {
+      "epoch": 0.3773829174353497,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6207386255264282,
+      "eval_runtime": 20.739,
+      "eval_samples_per_second": 2.122,
+      "eval_steps_per_second": 0.145,
+      "step": 2915
+    },
+    {
+      "epoch": 0.3775123798427032,
+      "grad_norm": 0.14413704069575264,
+      "learning_rate": 9.7905627544435e-07,
+      "loss": 0.839,
+      "step": 2916
+    },
+    {
+      "epoch": 0.37764184225005665,
+      "grad_norm": 0.1621074243863637,
+      "learning_rate": 9.788234591318476e-07,
+      "loss": 0.8204,
+      "step": 2917
+    },
+    {
+      "epoch": 0.3777713046574101,
+      "grad_norm": 0.1433322417043379,
+      "learning_rate": 9.785905705437318e-07,
+      "loss": 0.8381,
+      "step": 2918
+    },
+    {
+      "epoch": 0.37790076706476355,
+      "grad_norm": 0.16274607016418605,
+      "learning_rate": 9.783576097275751e-07,
+      "loss": 0.6924,
+      "step": 2919
+    },
+    {
+      "epoch": 0.378030229472117,
+      "grad_norm": 0.12163896570491606,
+      "learning_rate": 9.781245767309644e-07,
+      "loss": 0.7842,
+      "step": 2920
+    },
+    {
+      "epoch": 0.378030229472117,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6219815611839294,
+      "eval_runtime": 19.7696,
+      "eval_samples_per_second": 2.226,
+      "eval_steps_per_second": 0.152,
+      "step": 2920
+    },
+    {
+      "epoch": 0.3781596918794705,
+      "grad_norm": 0.1711905994356605,
+      "learning_rate": 9.778914716015012e-07,
+      "loss": 0.6746,
+      "step": 2921
+    },
+    {
+      "epoch": 0.37828915428682397,
+      "grad_norm": 0.16140728647998293,
+      "learning_rate": 9.776582943868023e-07,
+      "loss": 0.6947,
+      "step": 2922
+    },
+    {
+      "epoch": 0.3784186166941774,
+      "grad_norm": 0.12306153651013194,
+      "learning_rate": 9.774250451344987e-07,
+      "loss": 0.7122,
+      "step": 2923
+    },
+    {
+      "epoch": 0.3785480791015309,
+      "grad_norm": 0.1033864125430614,
+      "learning_rate": 9.771917238922362e-07,
+      "loss": 0.7742,
+      "step": 2924
+    },
+    {
+      "epoch": 0.37867754150888433,
+      "grad_norm": 0.14437274575784365,
+      "learning_rate": 9.769583307076757e-07,
+      "loss": 0.8065,
+      "step": 2925
+    },
+    {
+      "epoch": 0.37867754150888433,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.619140625,
+      "eval_runtime": 20.1743,
+      "eval_samples_per_second": 2.181,
+      "eval_steps_per_second": 0.149,
+      "step": 2925
+    },
+    {
+      "epoch": 0.37880700391623784,
+      "grad_norm": 0.12768225459881963,
+      "learning_rate": 9.767248656284923e-07,
+      "loss": 0.6436,
+      "step": 2926
+    },
+    {
+      "epoch": 0.3789364663235913,
+      "grad_norm": 0.19376344204495813,
+      "learning_rate": 9.764913287023756e-07,
+      "loss": 0.6767,
+      "step": 2927
+    },
+    {
+      "epoch": 0.37906592873094475,
+      "grad_norm": 0.17283468523541107,
+      "learning_rate": 9.762577199770308e-07,
+      "loss": 0.8026,
+      "step": 2928
+    },
+    {
+      "epoch": 0.3791953911382982,
+      "grad_norm": 0.12603899729750628,
+      "learning_rate": 9.76024039500177e-07,
+      "loss": 0.8136,
+      "step": 2929
+    },
+    {
+      "epoch": 0.3793248535456517,
+      "grad_norm": 0.11609354010733448,
+      "learning_rate": 9.757902873195483e-07,
+      "loss": 0.6344,
+      "step": 2930
+    },
+    {
+      "epoch": 0.3793248535456517,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6194957494735718,
+      "eval_runtime": 19.8324,
+      "eval_samples_per_second": 2.219,
+      "eval_steps_per_second": 0.151,
+      "step": 2930
+    },
+    {
+      "epoch": 0.37945431595300516,
+      "grad_norm": 0.20049524647476305,
+      "learning_rate": 9.755564634828932e-07,
+      "loss": 0.6322,
+      "step": 2931
+    },
+    {
+      "epoch": 0.3795837783603586,
+      "grad_norm": 0.12885743062912688,
+      "learning_rate": 9.75322568037975e-07,
+      "loss": 0.8063,
+      "step": 2932
+    },
+    {
+      "epoch": 0.37971324076771207,
+      "grad_norm": 0.12177216476930756,
+      "learning_rate": 9.750886010325714e-07,
+      "loss": 0.6625,
+      "step": 2933
+    },
+    {
+      "epoch": 0.3798427031750655,
+      "grad_norm": 0.3506098334993272,
+      "learning_rate": 9.748545625144754e-07,
+      "loss": 0.9276,
+      "step": 2934
+    },
+    {
+      "epoch": 0.37997216558241903,
+      "grad_norm": 0.12137965121930647,
+      "learning_rate": 9.746204525314933e-07,
+      "loss": 0.7565,
+      "step": 2935
+    },
+    {
+      "epoch": 0.37997216558241903,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6189630627632141,
+      "eval_runtime": 19.9211,
+      "eval_samples_per_second": 2.209,
+      "eval_steps_per_second": 0.151,
+      "step": 2935
+    },
+    {
+      "epoch": 0.3801016279897725,
+      "grad_norm": 0.1699646772733835,
+      "learning_rate": 9.743862711314474e-07,
+      "loss": 0.8575,
+      "step": 2936
+    },
+    {
+      "epoch": 0.38023109039712594,
+      "grad_norm": 0.12775046294973874,
+      "learning_rate": 9.741520183621743e-07,
+      "loss": 0.8253,
+      "step": 2937
+    },
+    {
+      "epoch": 0.3803605528044794,
+      "grad_norm": 0.13359175217216981,
+      "learning_rate": 9.739176942715242e-07,
+      "loss": 0.6843,
+      "step": 2938
+    },
+    {
+      "epoch": 0.38049001521183284,
+      "grad_norm": 0.1341214894303281,
+      "learning_rate": 9.736832989073627e-07,
+      "loss": 0.6556,
+      "step": 2939
+    },
+    {
+      "epoch": 0.38061947761918635,
+      "grad_norm": 0.11736389042064316,
+      "learning_rate": 9.7344883231757e-07,
+      "loss": 0.791,
+      "step": 2940
+    },
+    {
+      "epoch": 0.38061947761918635,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6212713122367859,
+      "eval_runtime": 20.0101,
+      "eval_samples_per_second": 2.199,
+      "eval_steps_per_second": 0.15,
+      "step": 2940
+    },
+    {
+      "epoch": 0.3807489400265398,
+      "grad_norm": 0.11781251235836517,
+      "learning_rate": 9.732142945500406e-07,
+      "loss": 0.7035,
+      "step": 2941
+    },
+    {
+      "epoch": 0.38087840243389326,
+      "grad_norm": 0.13578805227585336,
+      "learning_rate": 9.729796856526835e-07,
+      "loss": 0.7827,
+      "step": 2942
+    },
+    {
+      "epoch": 0.3810078648412467,
+      "grad_norm": 0.12253230081529547,
+      "learning_rate": 9.727450056734224e-07,
+      "loss": 0.8142,
+      "step": 2943
+    },
+    {
+      "epoch": 0.38113732724860017,
+      "grad_norm": 0.12978567074701472,
+      "learning_rate": 9.725102546601954e-07,
+      "loss": 0.6324,
+      "step": 2944
+    },
+    {
+      "epoch": 0.3812667896559537,
+      "grad_norm": 0.15150022400599913,
+      "learning_rate": 9.72275432660955e-07,
+      "loss": 0.8481,
+      "step": 2945
+    },
+    {
+      "epoch": 0.3812667896559537,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6187855005264282,
+      "eval_runtime": 20.1174,
+      "eval_samples_per_second": 2.187,
+      "eval_steps_per_second": 0.149,
+      "step": 2945
+    },
+    {
+      "epoch": 0.38139625206330713,
+      "grad_norm": 0.12496904279896774,
+      "learning_rate": 9.720405397236691e-07,
+      "loss": 0.7777,
+      "step": 2946
+    },
+    {
+      "epoch": 0.3815257144706606,
+      "grad_norm": 0.17337355995235101,
+      "learning_rate": 9.718055758963181e-07,
+      "loss": 0.8457,
+      "step": 2947
+    },
+    {
+      "epoch": 0.38165517687801404,
+      "grad_norm": 0.29201564674071445,
+      "learning_rate": 9.71570541226899e-07,
+      "loss": 0.9115,
+      "step": 2948
+    },
+    {
+      "epoch": 0.3817846392853675,
+      "grad_norm": 0.11403922625743408,
+      "learning_rate": 9.713354357634226e-07,
+      "loss": 0.7421,
+      "step": 2949
+    },
+    {
+      "epoch": 0.381914101692721,
+      "grad_norm": 0.17551327863611849,
+      "learning_rate": 9.71100259553913e-07,
+      "loss": 0.719,
+      "step": 2950
+    },
+    {
+      "epoch": 0.381914101692721,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6175426244735718,
+      "eval_runtime": 20.7676,
+      "eval_samples_per_second": 2.119,
+      "eval_steps_per_second": 0.144,
+      "step": 2950
+    },
+    {
+      "epoch": 0.38204356410007445,
+      "grad_norm": 0.2683226372989131,
+      "learning_rate": 9.708650126464107e-07,
+      "loss": 0.8911,
+      "step": 2951
+    },
+    {
+      "epoch": 0.3821730265074279,
+      "grad_norm": 0.12698481816769971,
+      "learning_rate": 9.706296950889689e-07,
+      "loss": 0.802,
+      "step": 2952
+    },
+    {
+      "epoch": 0.38230248891478136,
+      "grad_norm": 0.14449187930380666,
+      "learning_rate": 9.703943069296563e-07,
+      "loss": 0.8191,
+      "step": 2953
+    },
+    {
+      "epoch": 0.3824319513221348,
+      "grad_norm": 0.17288285607499515,
+      "learning_rate": 9.701588482165559e-07,
+      "loss": 0.9005,
+      "step": 2954
+    },
+    {
+      "epoch": 0.3825614137294883,
+      "grad_norm": 0.11128933102164836,
+      "learning_rate": 9.699233189977645e-07,
+      "loss": 0.7004,
+      "step": 2955
+    },
+    {
+      "epoch": 0.3825614137294883,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6205610632896423,
+      "eval_runtime": 20.4147,
+      "eval_samples_per_second": 2.155,
+      "eval_steps_per_second": 0.147,
+      "step": 2955
+    },
+    {
+      "epoch": 0.3826908761368418,
+      "grad_norm": 0.14522726841511438,
+      "learning_rate": 9.696877193213938e-07,
+      "loss": 0.8127,
+      "step": 2956
+    },
+    {
+      "epoch": 0.3828203385441952,
+      "grad_norm": 0.10892132776729214,
+      "learning_rate": 9.694520492355702e-07,
+      "loss": 0.7451,
+      "step": 2957
+    },
+    {
+      "epoch": 0.3829498009515487,
+      "grad_norm": 0.2910502845523496,
+      "learning_rate": 9.692163087884336e-07,
+      "loss": 0.8813,
+      "step": 2958
+    },
+    {
+      "epoch": 0.38307926335890213,
+      "grad_norm": 0.1327059346049371,
+      "learning_rate": 9.689804980281392e-07,
+      "loss": 0.6716,
+      "step": 2959
+    },
+    {
+      "epoch": 0.38320872576625564,
+      "grad_norm": 0.112394389186232,
+      "learning_rate": 9.687446170028559e-07,
+      "loss": 0.7485,
+      "step": 2960
+    },
+    {
+      "epoch": 0.38320872576625564,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6184303760528564,
+      "eval_runtime": 20.5626,
+      "eval_samples_per_second": 2.14,
+      "eval_steps_per_second": 0.146,
+      "step": 2960
+    },
+    {
+      "epoch": 0.3833381881736091,
+      "grad_norm": 0.11608336240880529,
+      "learning_rate": 9.685086657607668e-07,
+      "loss": 0.7906,
+      "step": 2961
+    },
+    {
+      "epoch": 0.38346765058096255,
+      "grad_norm": 0.1069424901883626,
+      "learning_rate": 9.682726443500707e-07,
+      "loss": 0.6702,
+      "step": 2962
+    },
+    {
+      "epoch": 0.383597112988316,
+      "grad_norm": 0.12422148674825287,
+      "learning_rate": 9.680365528189787e-07,
+      "loss": 0.7229,
+      "step": 2963
+    },
+    {
+      "epoch": 0.38372657539566946,
+      "grad_norm": 0.11725133943594489,
+      "learning_rate": 9.678003912157178e-07,
+      "loss": 0.6699,
+      "step": 2964
+    },
+    {
+      "epoch": 0.38385603780302296,
+      "grad_norm": 0.16245212612765417,
+      "learning_rate": 9.67564159588529e-07,
+      "loss": 0.8082,
+      "step": 2965
+    },
+    {
+      "epoch": 0.38385603780302296,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6159446239471436,
+      "eval_runtime": 20.4146,
+      "eval_samples_per_second": 2.155,
+      "eval_steps_per_second": 0.147,
+      "step": 2965
+    },
+    {
+      "epoch": 0.3839855002103764,
+      "grad_norm": 0.2723255227454983,
+      "learning_rate": 9.67327857985667e-07,
+      "loss": 0.99,
+      "step": 2966
+    },
+    {
+      "epoch": 0.38411496261772987,
+      "grad_norm": 0.11293641649025954,
+      "learning_rate": 9.670914864554014e-07,
+      "loss": 0.6996,
+      "step": 2967
+    },
+    {
+      "epoch": 0.3842444250250833,
+      "grad_norm": 0.127342427222284,
+      "learning_rate": 9.668550450460156e-07,
+      "loss": 0.804,
+      "step": 2968
+    },
+    {
+      "epoch": 0.38437388743243683,
+      "grad_norm": 0.12020174994253749,
+      "learning_rate": 9.666185338058083e-07,
+      "loss": 0.7474,
+      "step": 2969
+    },
+    {
+      "epoch": 0.3845033498397903,
+      "grad_norm": 0.15999556900609463,
+      "learning_rate": 9.663819527830906e-07,
+      "loss": 0.6677,
+      "step": 2970
+    },
+    {
+      "epoch": 0.3845033498397903,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6166548132896423,
+      "eval_runtime": 20.3236,
+      "eval_samples_per_second": 2.165,
+      "eval_steps_per_second": 0.148,
+      "step": 2970
+    },
+    {
+      "epoch": 0.38463281224714374,
+      "grad_norm": 0.16644370180520565,
+      "learning_rate": 9.661453020261898e-07,
+      "loss": 0.7926,
+      "step": 2971
+    },
+    {
+      "epoch": 0.3847622746544972,
+      "grad_norm": 0.13131534994832633,
+      "learning_rate": 9.659085815834463e-07,
+      "loss": 0.7896,
+      "step": 2972
+    },
+    {
+      "epoch": 0.38489173706185065,
+      "grad_norm": 0.10275317655902884,
+      "learning_rate": 9.656717915032153e-07,
+      "loss": 0.6989,
+      "step": 2973
+    },
+    {
+      "epoch": 0.38502119946920416,
+      "grad_norm": 0.1332665948345099,
+      "learning_rate": 9.654349318338655e-07,
+      "loss": 0.7637,
+      "step": 2974
+    },
+    {
+      "epoch": 0.3851506618765576,
+      "grad_norm": 0.1414660732726309,
+      "learning_rate": 9.651980026237805e-07,
+      "loss": 0.7838,
+      "step": 2975
+    },
+    {
+      "epoch": 0.3851506618765576,
+      "eval_PRM Accuracy": 0.7452830188679245,
+      "eval_PRM F1": 0.8540540540540541,
+      "eval_PRM F1 AUC": 0.4759036144578313,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7745098039215687,
+      "eval_PRM Recall": 0.9518072289156626,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6166548132896423,
+      "eval_runtime": 19.8971,
+      "eval_samples_per_second": 2.211,
+      "eval_steps_per_second": 0.151,
+      "step": 2975
+    },
+    {
+      "epoch": 0.38528012428391106,
+      "grad_norm": 0.10030760074649209,
+      "learning_rate": 9.649610039213584e-07,
+      "loss": 0.6753,
+      "step": 2976
+    },
+    {
+      "epoch": 0.3854095866912645,
+      "grad_norm": 0.16798701373016434,
+      "learning_rate": 9.647239357750101e-07,
+      "loss": 0.9323,
+      "step": 2977
+    },
+    {
+      "epoch": 0.38553904909861797,
+      "grad_norm": 0.13719937524078818,
+      "learning_rate": 9.644867982331623e-07,
+      "loss": 0.6757,
+      "step": 2978
+    },
+    {
+      "epoch": 0.3856685115059715,
+      "grad_norm": 0.3258988939345767,
+      "learning_rate": 9.642495913442548e-07,
+      "loss": 1.0037,
+      "step": 2979
+    },
+    {
+      "epoch": 0.38579797391332493,
+      "grad_norm": 0.2500398927720127,
+      "learning_rate": 9.64012315156742e-07,
+      "loss": 0.6485,
+      "step": 2980
+    },
+    {
+      "epoch": 0.38579797391332493,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6166548132896423,
+      "eval_runtime": 20.3246,
+      "eval_samples_per_second": 2.165,
+      "eval_steps_per_second": 0.148,
+      "step": 2980
+    },
+    {
+      "epoch": 0.3859274363206784,
+      "grad_norm": 0.14195969638249414,
+      "learning_rate": 9.637749697190924e-07,
+      "loss": 0.847,
+      "step": 2981
+    },
+    {
+      "epoch": 0.38605689872803184,
+      "grad_norm": 0.19284062836922308,
+      "learning_rate": 9.635375550797885e-07,
+      "loss": 0.626,
+      "step": 2982
+    },
+    {
+      "epoch": 0.3861863611353853,
+      "grad_norm": 0.14693518026601537,
+      "learning_rate": 9.633000712873272e-07,
+      "loss": 0.8036,
+      "step": 2983
+    },
+    {
+      "epoch": 0.3863158235427388,
+      "grad_norm": 0.1276245523638328,
+      "learning_rate": 9.630625183902193e-07,
+      "loss": 0.8201,
+      "step": 2984
+    },
+    {
+      "epoch": 0.38644528595009225,
+      "grad_norm": 0.12996353056823864,
+      "learning_rate": 9.628248964369898e-07,
+      "loss": 0.6917,
+      "step": 2985
+    },
+    {
+      "epoch": 0.38644528595009225,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6184303760528564,
+      "eval_runtime": 19.8595,
+      "eval_samples_per_second": 2.216,
+      "eval_steps_per_second": 0.151,
+      "step": 2985
+    },
+    {
+      "epoch": 0.3865747483574457,
+      "grad_norm": 0.14461082624521332,
+      "learning_rate": 9.625872054761777e-07,
+      "loss": 0.7872,
+      "step": 2986
+    },
+    {
+      "epoch": 0.38670421076479916,
+      "grad_norm": 0.10981781767572543,
+      "learning_rate": 9.623494455563365e-07,
+      "loss": 0.7173,
+      "step": 2987
+    },
+    {
+      "epoch": 0.3868336731721526,
+      "grad_norm": 0.18090521427014442,
+      "learning_rate": 9.62111616726033e-07,
+      "loss": 0.6464,
+      "step": 2988
+    },
+    {
+      "epoch": 0.3869631355795061,
+      "grad_norm": 0.15986378981868013,
+      "learning_rate": 9.61873719033849e-07,
+      "loss": 0.8995,
+      "step": 2989
+    },
+    {
+      "epoch": 0.3870925979868596,
+      "grad_norm": 0.11538584014845356,
+      "learning_rate": 9.6163575252838e-07,
+      "loss": 0.7437,
+      "step": 2990
+    },
+    {
+      "epoch": 0.3870925979868596,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6162996888160706,
+      "eval_runtime": 20.4872,
+      "eval_samples_per_second": 2.148,
+      "eval_steps_per_second": 0.146,
+      "step": 2990
+    },
+    {
+      "epoch": 0.38722206039421303,
+      "grad_norm": 0.1886340654569701,
+      "learning_rate": 9.61397717258235e-07,
+      "loss": 0.6417,
+      "step": 2991
+    },
+    {
+      "epoch": 0.3873515228015665,
+      "grad_norm": 0.12209644550522541,
+      "learning_rate": 9.611596132720377e-07,
+      "loss": 0.7263,
+      "step": 2992
+    },
+    {
+      "epoch": 0.38748098520891994,
+      "grad_norm": 0.13047979406009533,
+      "learning_rate": 9.609214406184258e-07,
+      "loss": 0.6543,
+      "step": 2993
+    },
+    {
+      "epoch": 0.38761044761627345,
+      "grad_norm": 0.22390987067116666,
+      "learning_rate": 9.60683199346051e-07,
+      "loss": 0.9119,
+      "step": 2994
+    },
+    {
+      "epoch": 0.3877399100236269,
+      "grad_norm": 0.20221461348124425,
+      "learning_rate": 9.604448895035787e-07,
+      "loss": 0.9195,
+      "step": 2995
+    },
+    {
+      "epoch": 0.3877399100236269,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6148792505264282,
+      "eval_runtime": 20.7237,
+      "eval_samples_per_second": 2.123,
+      "eval_steps_per_second": 0.145,
+      "step": 2995
+    },
+    {
+      "epoch": 0.38786937243098035,
+      "grad_norm": 0.18992182302509658,
+      "learning_rate": 9.602065111396886e-07,
+      "loss": 0.6474,
+      "step": 2996
+    },
+    {
+      "epoch": 0.3879988348383338,
+      "grad_norm": 0.13798151412899176,
+      "learning_rate": 9.59968064303074e-07,
+      "loss": 0.7939,
+      "step": 2997
+    },
+    {
+      "epoch": 0.38812829724568726,
+      "grad_norm": 0.14072398182924187,
+      "learning_rate": 9.59729549042443e-07,
+      "loss": 0.7848,
+      "step": 2998
+    },
+    {
+      "epoch": 0.38825775965304077,
+      "grad_norm": 0.08599506231353685,
+      "learning_rate": 9.594909654065168e-07,
+      "loss": 0.6457,
+      "step": 2999
+    },
+    {
+      "epoch": 0.3883872220603942,
+      "grad_norm": 0.11391958686440753,
+      "learning_rate": 9.592523134440312e-07,
+      "loss": 0.6804,
+      "step": 3000
+    },
+    {
+      "epoch": 0.3883872220603942,
+      "eval_PRM Accuracy": 0.7358490566037735,
+      "eval_PRM F1": 0.8478260869565217,
+      "eval_PRM F1 AUC": 0.46987951807228917,
+      "eval_PRM F1 Neg": 0.0,
+      "eval_PRM NPV": 0.0,
+      "eval_PRM Precision": 0.7722772277227723,
+      "eval_PRM Recall": 0.9397590361445783,
+      "eval_PRM Specificty": 0.0,
+      "eval_loss": 0.6141690611839294,
+      "eval_runtime": 20.919,
+      "eval_samples_per_second": 2.103,
+      "eval_steps_per_second": 0.143,
+      "step": 3000
     }
   ],
   "logging_steps": 1,
@@ -25542,7 +30642,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 2098934106324992.0,
+  "total_flos": 2518198355329024.0,
   "train_batch_size": 4,
   "trial_name": null,
   "trial_params": null