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": 0.25892481470692946,
+  "epoch": 0.32365601838366187,
   "eval_steps": 5,
-  "global_step": 2000,
+  "global_step": 2500,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -20423,6 +20423,5106 @@
       "eval_samples_per_second": 3.108,
       "eval_steps_per_second": 0.212,
       "step": 2000
+    },
+    {
+      "epoch": 0.2590542771142829,
+      "grad_norm": 3.7414586362143853,
+      "learning_rate": 1.1561844216386054e-06,
+      "loss": 3.1995,
+      "step": 2001
+    },
+    {
+      "epoch": 0.2591837395216364,
+      "grad_norm": 3.174259407155458,
+      "learning_rate": 1.1560355157520717e-06,
+      "loss": 3.1294,
+      "step": 2002
+    },
+    {
+      "epoch": 0.2593132019289899,
+      "grad_norm": 3.281577783948224,
+      "learning_rate": 1.1558865013907703e-06,
+      "loss": 2.9639,
+      "step": 2003
+    },
+    {
+      "epoch": 0.25944266433634333,
+      "grad_norm": 2.1271887396031803,
+      "learning_rate": 1.1557373785851404e-06,
+      "loss": 2.6738,
+      "step": 2004
+    },
+    {
+      "epoch": 0.2595721267436968,
+      "grad_norm": 2.774438768866771,
+      "learning_rate": 1.1555881473656433e-06,
+      "loss": 3.1494,
+      "step": 2005
+    },
+    {
+      "epoch": 0.2595721267436968,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.681684970855713,
+      "eval_runtime": 14.0066,
+      "eval_samples_per_second": 3.141,
+      "eval_steps_per_second": 0.214,
+      "step": 2005
+    },
+    {
+      "epoch": 0.25970158915105024,
+      "grad_norm": 5.198117608582652,
+      "learning_rate": 1.1554388077627625e-06,
+      "loss": 3.2031,
+      "step": 2006
+    },
+    {
+      "epoch": 0.25983105155840375,
+      "grad_norm": 3.78250264798494,
+      "learning_rate": 1.1552893598070039e-06,
+      "loss": 2.9292,
+      "step": 2007
+    },
+    {
+      "epoch": 0.2599605139657572,
+      "grad_norm": 3.8546913392247575,
+      "learning_rate": 1.155139803528895e-06,
+      "loss": 3.1672,
+      "step": 2008
+    },
+    {
+      "epoch": 0.26008997637311065,
+      "grad_norm": 4.48117237011425,
+      "learning_rate": 1.1549901389589857e-06,
+      "loss": 3.1758,
+      "step": 2009
+    },
+    {
+      "epoch": 0.2602194387804641,
+      "grad_norm": 1.8305553827371683,
+      "learning_rate": 1.1548403661278482e-06,
+      "loss": 2.6951,
+      "step": 2010
+    },
+    {
+      "epoch": 0.2602194387804641,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.679643154144287,
+      "eval_runtime": 13.7347,
+      "eval_samples_per_second": 3.204,
+      "eval_steps_per_second": 0.218,
+      "step": 2010
+    },
+    {
+      "epoch": 0.26034890118781756,
+      "grad_norm": 3.109756075960622,
+      "learning_rate": 1.1546904850660764e-06,
+      "loss": 3.0293,
+      "step": 2011
+    },
+    {
+      "epoch": 0.26047836359517107,
+      "grad_norm": 2.8726703270887923,
+      "learning_rate": 1.154540495804287e-06,
+      "loss": 3.1035,
+      "step": 2012
+    },
+    {
+      "epoch": 0.2606078260025245,
+      "grad_norm": 2.9878880802222754,
+      "learning_rate": 1.1543903983731178e-06,
+      "loss": 2.7725,
+      "step": 2013
+    },
+    {
+      "epoch": 0.260737288409878,
+      "grad_norm": 3.444341083421166,
+      "learning_rate": 1.1542401928032295e-06,
+      "loss": 2.6377,
+      "step": 2014
+    },
+    {
+      "epoch": 0.26086675081723143,
+      "grad_norm": 1.7609766969107403,
+      "learning_rate": 1.1540898791253047e-06,
+      "loss": 2.6497,
+      "step": 2015
+    },
+    {
+      "epoch": 0.26086675081723143,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.6788440942764282,
+      "eval_runtime": 14.0198,
+      "eval_samples_per_second": 3.138,
+      "eval_steps_per_second": 0.214,
+      "step": 2015
+    },
+    {
+      "epoch": 0.26099621322458494,
+      "grad_norm": 5.869701754719928,
+      "learning_rate": 1.153939457370048e-06,
+      "loss": 3.166,
+      "step": 2016
+    },
+    {
+      "epoch": 0.2611256756319384,
+      "grad_norm": 2.2778536124349573,
+      "learning_rate": 1.1537889275681857e-06,
+      "loss": 2.5781,
+      "step": 2017
+    },
+    {
+      "epoch": 0.26125513803929185,
+      "grad_norm": 2.468038446597042,
+      "learning_rate": 1.153638289750467e-06,
+      "loss": 2.6182,
+      "step": 2018
+    },
+    {
+      "epoch": 0.2613846004466453,
+      "grad_norm": 5.395096398524735,
+      "learning_rate": 1.1534875439476629e-06,
+      "loss": 3.1719,
+      "step": 2019
+    },
+    {
+      "epoch": 0.26151406285399875,
+      "grad_norm": 3.3207812998236697,
+      "learning_rate": 1.1533366901905656e-06,
+      "loss": 2.7502,
+      "step": 2020
+    },
+    {
+      "epoch": 0.26151406285399875,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.6794655323028564,
+      "eval_runtime": 13.6532,
+      "eval_samples_per_second": 3.223,
+      "eval_steps_per_second": 0.22,
+      "step": 2020
+    },
+    {
+      "epoch": 0.26164352526135226,
+      "grad_norm": 2.427716489335077,
+      "learning_rate": 1.1531857285099906e-06,
+      "loss": 2.6158,
+      "step": 2021
+    },
+    {
+      "epoch": 0.2617729876687057,
+      "grad_norm": 3.6106438357317048,
+      "learning_rate": 1.1530346589367748e-06,
+      "loss": 3.6465,
+      "step": 2022
+    },
+    {
+      "epoch": 0.26190245007605917,
+      "grad_norm": 3.943559957928543,
+      "learning_rate": 1.152883481501777e-06,
+      "loss": 3.6538,
+      "step": 2023
+    },
+    {
+      "epoch": 0.2620319124834126,
+      "grad_norm": 2.743156932662662,
+      "learning_rate": 1.1527321962358785e-06,
+      "loss": 2.7329,
+      "step": 2024
+    },
+    {
+      "epoch": 0.2621613748907661,
+      "grad_norm": 2.6188952317713157,
+      "learning_rate": 1.152580803169982e-06,
+      "loss": 2.8013,
+      "step": 2025
+    },
+    {
+      "epoch": 0.2621613748907661,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.6809303760528564,
+      "eval_runtime": 14.3475,
+      "eval_samples_per_second": 3.067,
+      "eval_steps_per_second": 0.209,
+      "step": 2025
+    },
+    {
+      "epoch": 0.2622908372981196,
+      "grad_norm": 2.7202565358168758,
+      "learning_rate": 1.1524293023350133e-06,
+      "loss": 2.9446,
+      "step": 2026
+    },
+    {
+      "epoch": 0.26242029970547304,
+      "grad_norm": 2.3315902731412477,
+      "learning_rate": 1.1522776937619189e-06,
+      "loss": 2.9011,
+      "step": 2027
+    },
+    {
+      "epoch": 0.2625497621128265,
+      "grad_norm": 2.3251635944723503,
+      "learning_rate": 1.152125977481668e-06,
+      "loss": 2.6204,
+      "step": 2028
+    },
+    {
+      "epoch": 0.26267922452017994,
+      "grad_norm": 3.3325112007094364,
+      "learning_rate": 1.151974153525252e-06,
+      "loss": 2.8809,
+      "step": 2029
+    },
+    {
+      "epoch": 0.2628086869275334,
+      "grad_norm": 2.199691784327611,
+      "learning_rate": 1.1518222219236839e-06,
+      "loss": 2.6538,
+      "step": 2030
+    },
+    {
+      "epoch": 0.2628086869275334,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.6823952198028564,
+      "eval_runtime": 14.006,
+      "eval_samples_per_second": 3.142,
+      "eval_steps_per_second": 0.214,
+      "step": 2030
+    },
+    {
+      "epoch": 0.2629381493348869,
+      "grad_norm": 3.9473563499032025,
+      "learning_rate": 1.1516701827079986e-06,
+      "loss": 2.6489,
+      "step": 2031
+    },
+    {
+      "epoch": 0.26306761174224036,
+      "grad_norm": 3.383719197891166,
+      "learning_rate": 1.1515180359092534e-06,
+      "loss": 3.0449,
+      "step": 2032
+    },
+    {
+      "epoch": 0.2631970741495938,
+      "grad_norm": 2.9950880493165957,
+      "learning_rate": 1.1513657815585276e-06,
+      "loss": 2.9375,
+      "step": 2033
+    },
+    {
+      "epoch": 0.26332653655694727,
+      "grad_norm": 8.46407490493878,
+      "learning_rate": 1.1512134196869214e-06,
+      "loss": 3.7515,
+      "step": 2034
+    },
+    {
+      "epoch": 0.2634559989643007,
+      "grad_norm": 4.780627101920859,
+      "learning_rate": 1.151060950325559e-06,
+      "loss": 3.4456,
+      "step": 2035
+    },
+    {
+      "epoch": 0.2634559989643007,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.6808860301971436,
+      "eval_runtime": 13.8366,
+      "eval_samples_per_second": 3.18,
+      "eval_steps_per_second": 0.217,
+      "step": 2035
+    },
+    {
+      "epoch": 0.26358546137165423,
+      "grad_norm": 2.79364686641326,
+      "learning_rate": 1.1509083735055843e-06,
+      "loss": 3.0391,
+      "step": 2036
+    },
+    {
+      "epoch": 0.2637149237790077,
+      "grad_norm": 1.87933440252873,
+      "learning_rate": 1.1507556892581648e-06,
+      "loss": 2.8245,
+      "step": 2037
+    },
+    {
+      "epoch": 0.26384438618636114,
+      "grad_norm": 1.8936710995485329,
+      "learning_rate": 1.1506028976144892e-06,
+      "loss": 2.6074,
+      "step": 2038
+    },
+    {
+      "epoch": 0.2639738485937146,
+      "grad_norm": 1.4054084526805806,
+      "learning_rate": 1.1504499986057682e-06,
+      "loss": 2.4592,
+      "step": 2039
+    },
+    {
+      "epoch": 0.26410331100106804,
+      "grad_norm": 2.4430917519941975,
+      "learning_rate": 1.1502969922632345e-06,
+      "loss": 2.929,
+      "step": 2040
+    },
+    {
+      "epoch": 0.26410331100106804,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.6823064088821411,
+      "eval_runtime": 14.0187,
+      "eval_samples_per_second": 3.139,
+      "eval_steps_per_second": 0.214,
+      "step": 2040
+    },
+    {
+      "epoch": 0.26423277340842155,
+      "grad_norm": 7.223900577295312,
+      "learning_rate": 1.1501438786181432e-06,
+      "loss": 3.6663,
+      "step": 2041
+    },
+    {
+      "epoch": 0.264362235815775,
+      "grad_norm": 2.318058792070975,
+      "learning_rate": 1.1499906577017703e-06,
+      "loss": 2.8945,
+      "step": 2042
+    },
+    {
+      "epoch": 0.26449169822312846,
+      "grad_norm": 3.871859060919414,
+      "learning_rate": 1.1498373295454142e-06,
+      "loss": 3.21,
+      "step": 2043
+    },
+    {
+      "epoch": 0.2646211606304819,
+      "grad_norm": 5.526379720800749,
+      "learning_rate": 1.149683894180396e-06,
+      "loss": 2.9705,
+      "step": 2044
+    },
+    {
+      "epoch": 0.26475062303783536,
+      "grad_norm": 2.184094217105361,
+      "learning_rate": 1.1495303516380575e-06,
+      "loss": 2.7295,
+      "step": 2045
+    },
+    {
+      "epoch": 0.26475062303783536,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.6807528734207153,
+      "eval_runtime": 14.1793,
+      "eval_samples_per_second": 3.103,
+      "eval_steps_per_second": 0.212,
+      "step": 2045
+    },
+    {
+      "epoch": 0.2648800854451889,
+      "grad_norm": 2.698760329009547,
+      "learning_rate": 1.1493767019497628e-06,
+      "loss": 2.5781,
+      "step": 2046
+    },
+    {
+      "epoch": 0.2650095478525423,
+      "grad_norm": 2.7530255107060464,
+      "learning_rate": 1.149222945146898e-06,
+      "loss": 3.1382,
+      "step": 2047
+    },
+    {
+      "epoch": 0.2651390102598958,
+      "grad_norm": 2.915233836831477,
+      "learning_rate": 1.1490690812608714e-06,
+      "loss": 2.7622,
+      "step": 2048
+    },
+    {
+      "epoch": 0.26526847266724923,
+      "grad_norm": 2.3199207991923876,
+      "learning_rate": 1.1489151103231125e-06,
+      "loss": 2.675,
+      "step": 2049
+    },
+    {
+      "epoch": 0.2653979350746027,
+      "grad_norm": 7.5563831009740365,
+      "learning_rate": 1.148761032365073e-06,
+      "loss": 3.752,
+      "step": 2050
+    },
+    {
+      "epoch": 0.2653979350746027,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.6849032640457153,
+      "eval_runtime": 14.0846,
+      "eval_samples_per_second": 3.124,
+      "eval_steps_per_second": 0.213,
+      "step": 2050
+    },
+    {
+      "epoch": 0.2655273974819562,
+      "grad_norm": 2.4927119883817572,
+      "learning_rate": 1.1486068474182264e-06,
+      "loss": 2.9487,
+      "step": 2051
+    },
+    {
+      "epoch": 0.26565685988930965,
+      "grad_norm": 7.5591324034845675,
+      "learning_rate": 1.1484525555140682e-06,
+      "loss": 3.5735,
+      "step": 2052
+    },
+    {
+      "epoch": 0.2657863222966631,
+      "grad_norm": 2.0337293189836303,
+      "learning_rate": 1.1482981566841156e-06,
+      "loss": 2.9456,
+      "step": 2053
+    },
+    {
+      "epoch": 0.26591578470401656,
+      "grad_norm": 2.389099062282753,
+      "learning_rate": 1.148143650959908e-06,
+      "loss": 3.1396,
+      "step": 2054
+    },
+    {
+      "epoch": 0.26604524711137,
+      "grad_norm": 6.009646897002942,
+      "learning_rate": 1.1479890383730059e-06,
+      "loss": 3.4199,
+      "step": 2055
+    },
+    {
+      "epoch": 0.26604524711137,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.6865234375,
+      "eval_runtime": 15.7285,
+      "eval_samples_per_second": 2.797,
+      "eval_steps_per_second": 0.191,
+      "step": 2055
+    },
+    {
+      "epoch": 0.2661747095187235,
+      "grad_norm": 2.126713921038673,
+      "learning_rate": 1.147834318954992e-06,
+      "loss": 2.551,
+      "step": 2056
+    },
+    {
+      "epoch": 0.26630417192607697,
+      "grad_norm": 3.324447412588428,
+      "learning_rate": 1.1476794927374713e-06,
+      "loss": 2.8799,
+      "step": 2057
+    },
+    {
+      "epoch": 0.2664336343334304,
+      "grad_norm": 2.3368801952968288,
+      "learning_rate": 1.14752455975207e-06,
+      "loss": 2.7803,
+      "step": 2058
+    },
+    {
+      "epoch": 0.2665630967407839,
+      "grad_norm": 2.357680073144056,
+      "learning_rate": 1.1473695200304362e-06,
+      "loss": 2.6907,
+      "step": 2059
+    },
+    {
+      "epoch": 0.2666925591481374,
+      "grad_norm": 3.046425147921824,
+      "learning_rate": 1.1472143736042398e-06,
+      "loss": 2.9746,
+      "step": 2060
+    },
+    {
+      "epoch": 0.2666925591481374,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.6842595338821411,
+      "eval_runtime": 14.2461,
+      "eval_samples_per_second": 3.089,
+      "eval_steps_per_second": 0.211,
+      "step": 2060
+    },
+    {
+      "epoch": 0.26682202155549084,
+      "grad_norm": 4.973065091786923,
+      "learning_rate": 1.1470591205051731e-06,
+      "loss": 3.3354,
+      "step": 2061
+    },
+    {
+      "epoch": 0.2669514839628443,
+      "grad_norm": 2.013444831983804,
+      "learning_rate": 1.1469037607649494e-06,
+      "loss": 2.6704,
+      "step": 2062
+    },
+    {
+      "epoch": 0.26708094637019775,
+      "grad_norm": 5.622216107722647,
+      "learning_rate": 1.1467482944153039e-06,
+      "loss": 3.78,
+      "step": 2063
+    },
+    {
+      "epoch": 0.2672104087775512,
+      "grad_norm": 1.798926339828797,
+      "learning_rate": 1.1465927214879938e-06,
+      "loss": 2.7153,
+      "step": 2064
+    },
+    {
+      "epoch": 0.2673398711849047,
+      "grad_norm": 3.6711464441242017,
+      "learning_rate": 1.1464370420147984e-06,
+      "loss": 3.0742,
+      "step": 2065
+    },
+    {
+      "epoch": 0.2673398711849047,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.687056064605713,
+      "eval_runtime": 14.2454,
+      "eval_samples_per_second": 3.089,
+      "eval_steps_per_second": 0.211,
+      "step": 2065
+    },
+    {
+      "epoch": 0.26746933359225816,
+      "grad_norm": 5.410646675977281,
+      "learning_rate": 1.146281256027518e-06,
+      "loss": 3.6685,
+      "step": 2066
+    },
+    {
+      "epoch": 0.2675987959996116,
+      "grad_norm": 3.9489489867754717,
+      "learning_rate": 1.146125363557975e-06,
+      "loss": 2.9485,
+      "step": 2067
+    },
+    {
+      "epoch": 0.26772825840696507,
+      "grad_norm": 5.146815929244116,
+      "learning_rate": 1.145969364638014e-06,
+      "loss": 3.1311,
+      "step": 2068
+    },
+    {
+      "epoch": 0.2678577208143185,
+      "grad_norm": 6.755278783157041,
+      "learning_rate": 1.1458132592995005e-06,
+      "loss": 3.9727,
+      "step": 2069
+    },
+    {
+      "epoch": 0.26798718322167203,
+      "grad_norm": 5.54033753808382,
+      "learning_rate": 1.1456570475743225e-06,
+      "loss": 3.1401,
+      "step": 2070
+    },
+    {
+      "epoch": 0.26798718322167203,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.6835049390792847,
+      "eval_runtime": 14.487,
+      "eval_samples_per_second": 3.037,
+      "eval_steps_per_second": 0.207,
+      "step": 2070
+    },
+    {
+      "epoch": 0.2681166456290255,
+      "grad_norm": 4.096552958120691,
+      "learning_rate": 1.1455007294943895e-06,
+      "loss": 3.2363,
+      "step": 2071
+    },
+    {
+      "epoch": 0.26824610803637894,
+      "grad_norm": 2.5353750762707965,
+      "learning_rate": 1.1453443050916322e-06,
+      "loss": 2.9187,
+      "step": 2072
+    },
+    {
+      "epoch": 0.2683755704437324,
+      "grad_norm": 2.4643309441062984,
+      "learning_rate": 1.1451877743980036e-06,
+      "loss": 2.8909,
+      "step": 2073
+    },
+    {
+      "epoch": 0.26850503285108585,
+      "grad_norm": 1.348337058733433,
+      "learning_rate": 1.1450311374454786e-06,
+      "loss": 2.4468,
+      "step": 2074
+    },
+    {
+      "epoch": 0.26863449525843935,
+      "grad_norm": 2.512610412236078,
+      "learning_rate": 1.1448743942660532e-06,
+      "loss": 2.7932,
+      "step": 2075
+    },
+    {
+      "epoch": 0.26863449525843935,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.6842595338821411,
+      "eval_runtime": 15.5582,
+      "eval_samples_per_second": 2.828,
+      "eval_steps_per_second": 0.193,
+      "step": 2075
+    },
+    {
+      "epoch": 0.2687639576657928,
+      "grad_norm": 3.4194935079848956,
+      "learning_rate": 1.1447175448917452e-06,
+      "loss": 2.9795,
+      "step": 2076
+    },
+    {
+      "epoch": 0.26889342007314626,
+      "grad_norm": 2.2507691559282668,
+      "learning_rate": 1.1445605893545946e-06,
+      "loss": 3.0302,
+      "step": 2077
+    },
+    {
+      "epoch": 0.2690228824804997,
+      "grad_norm": 5.145316035641122,
+      "learning_rate": 1.1444035276866625e-06,
+      "loss": 3.2922,
+      "step": 2078
+    },
+    {
+      "epoch": 0.26915234488785317,
+      "grad_norm": 2.8097246510674085,
+      "learning_rate": 1.144246359920032e-06,
+      "loss": 3.0798,
+      "step": 2079
+    },
+    {
+      "epoch": 0.2692818072952067,
+      "grad_norm": 2.6726345566288794,
+      "learning_rate": 1.1440890860868082e-06,
+      "loss": 3.1318,
+      "step": 2080
+    },
+    {
+      "epoch": 0.2692818072952067,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.6905406713485718,
+      "eval_runtime": 15.2411,
+      "eval_samples_per_second": 2.887,
+      "eval_steps_per_second": 0.197,
+      "step": 2080
+    },
+    {
+      "epoch": 0.26941126970256013,
+      "grad_norm": 5.951832984679558,
+      "learning_rate": 1.1439317062191168e-06,
+      "loss": 3.1069,
+      "step": 2081
+    },
+    {
+      "epoch": 0.2695407321099136,
+      "grad_norm": 3.2621629390696714,
+      "learning_rate": 1.1437742203491062e-06,
+      "loss": 3.1311,
+      "step": 2082
+    },
+    {
+      "epoch": 0.26967019451726704,
+      "grad_norm": 2.6258824367727622,
+      "learning_rate": 1.1436166285089458e-06,
+      "loss": 3.1406,
+      "step": 2083
+    },
+    {
+      "epoch": 0.2697996569246205,
+      "grad_norm": 2.4642538273932812,
+      "learning_rate": 1.1434589307308275e-06,
+      "loss": 2.8,
+      "step": 2084
+    },
+    {
+      "epoch": 0.269929119331974,
+      "grad_norm": 5.099022838437128,
+      "learning_rate": 1.1433011270469638e-06,
+      "loss": 3.3835,
+      "step": 2085
+    },
+    {
+      "epoch": 0.269929119331974,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.6904518604278564,
+      "eval_runtime": 14.9654,
+      "eval_samples_per_second": 2.94,
+      "eval_steps_per_second": 0.2,
+      "step": 2085
+    },
+    {
+      "epoch": 0.27005858173932745,
+      "grad_norm": 2.6385860321112222,
+      "learning_rate": 1.1431432174895894e-06,
+      "loss": 2.8601,
+      "step": 2086
+    },
+    {
+      "epoch": 0.2701880441466809,
+      "grad_norm": 6.821887298439188,
+      "learning_rate": 1.1429852020909603e-06,
+      "loss": 3.4854,
+      "step": 2087
+    },
+    {
+      "epoch": 0.27031750655403436,
+      "grad_norm": 2.7629744843909694,
+      "learning_rate": 1.142827080883355e-06,
+      "loss": 2.7114,
+      "step": 2088
+    },
+    {
+      "epoch": 0.2704469689613878,
+      "grad_norm": 4.467498101925651,
+      "learning_rate": 1.1426688538990724e-06,
+      "loss": 2.9343,
+      "step": 2089
+    },
+    {
+      "epoch": 0.2705764313687413,
+      "grad_norm": 3.044496230423841,
+      "learning_rate": 1.1425105211704338e-06,
+      "loss": 3.2568,
+      "step": 2090
+    },
+    {
+      "epoch": 0.2705764313687413,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.6920276880264282,
+      "eval_runtime": 15.799,
+      "eval_samples_per_second": 2.785,
+      "eval_steps_per_second": 0.19,
+      "step": 2090
+    },
+    {
+      "epoch": 0.2707058937760948,
+      "grad_norm": 1.7772418360489868,
+      "learning_rate": 1.1423520827297814e-06,
+      "loss": 2.6912,
+      "step": 2091
+    },
+    {
+      "epoch": 0.27083535618344823,
+      "grad_norm": 3.4981336783934514,
+      "learning_rate": 1.1421935386094803e-06,
+      "loss": 3.1421,
+      "step": 2092
+    },
+    {
+      "epoch": 0.2709648185908017,
+      "grad_norm": 2.348128872979748,
+      "learning_rate": 1.1420348888419158e-06,
+      "loss": 2.6309,
+      "step": 2093
+    },
+    {
+      "epoch": 0.27109428099815513,
+      "grad_norm": 1.5857791684155143,
+      "learning_rate": 1.1418761334594954e-06,
+      "loss": 2.5281,
+      "step": 2094
+    },
+    {
+      "epoch": 0.27122374340550864,
+      "grad_norm": 2.908288025149171,
+      "learning_rate": 1.1417172724946482e-06,
+      "loss": 3.4331,
+      "step": 2095
+    },
+    {
+      "epoch": 0.27122374340550864,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.6918723583221436,
+      "eval_runtime": 15.2125,
+      "eval_samples_per_second": 2.892,
+      "eval_steps_per_second": 0.197,
+      "step": 2095
+    },
+    {
+      "epoch": 0.2713532058128621,
+      "grad_norm": 8.048230156977537,
+      "learning_rate": 1.1415583059798247e-06,
+      "loss": 3.7832,
+      "step": 2096
+    },
+    {
+      "epoch": 0.27148266822021555,
+      "grad_norm": 4.618960932081442,
+      "learning_rate": 1.1413992339474968e-06,
+      "loss": 3.5208,
+      "step": 2097
+    },
+    {
+      "epoch": 0.271612130627569,
+      "grad_norm": 5.446955840816096,
+      "learning_rate": 1.1412400564301587e-06,
+      "loss": 3.4502,
+      "step": 2098
+    },
+    {
+      "epoch": 0.27174159303492246,
+      "grad_norm": 2.3579238476305817,
+      "learning_rate": 1.1410807734603251e-06,
+      "loss": 2.8325,
+      "step": 2099
+    },
+    {
+      "epoch": 0.27187105544227597,
+      "grad_norm": 4.365697054676799,
+      "learning_rate": 1.1409213850705334e-06,
+      "loss": 3.1313,
+      "step": 2100
+    },
+    {
+      "epoch": 0.27187105544227597,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.6954900026321411,
+      "eval_runtime": 14.5369,
+      "eval_samples_per_second": 3.027,
+      "eval_steps_per_second": 0.206,
+      "step": 2100
+    },
+    {
+      "epoch": 0.2720005178496294,
+      "grad_norm": 2.371784006098729,
+      "learning_rate": 1.1407618912933414e-06,
+      "loss": 2.8833,
+      "step": 2101
+    },
+    {
+      "epoch": 0.2721299802569829,
+      "grad_norm": 4.746801470373952,
+      "learning_rate": 1.140602292161329e-06,
+      "loss": 3.0278,
+      "step": 2102
+    },
+    {
+      "epoch": 0.2722594426643363,
+      "grad_norm": 2.5329249323059995,
+      "learning_rate": 1.1404425877070977e-06,
+      "loss": 2.7031,
+      "step": 2103
+    },
+    {
+      "epoch": 0.27238890507168984,
+      "grad_norm": 2.871788693388005,
+      "learning_rate": 1.1402827779632702e-06,
+      "loss": 2.7466,
+      "step": 2104
+    },
+    {
+      "epoch": 0.2725183674790433,
+      "grad_norm": 3.1461985929759897,
+      "learning_rate": 1.1401228629624912e-06,
+      "loss": 2.7627,
+      "step": 2105
+    },
+    {
+      "epoch": 0.2725183674790433,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.697509765625,
+      "eval_runtime": 14.5422,
+      "eval_samples_per_second": 3.026,
+      "eval_steps_per_second": 0.206,
+      "step": 2105
+    },
+    {
+      "epoch": 0.27264782988639674,
+      "grad_norm": 2.8704105394729185,
+      "learning_rate": 1.1399628427374262e-06,
+      "loss": 2.8513,
+      "step": 2106
+    },
+    {
+      "epoch": 0.2727772922937502,
+      "grad_norm": 3.501116124828323,
+      "learning_rate": 1.139802717320763e-06,
+      "loss": 2.8237,
+      "step": 2107
+    },
+    {
+      "epoch": 0.27290675470110365,
+      "grad_norm": 7.3612594472307045,
+      "learning_rate": 1.1396424867452099e-06,
+      "loss": 3.647,
+      "step": 2108
+    },
+    {
+      "epoch": 0.27303621710845716,
+      "grad_norm": 3.401919848713317,
+      "learning_rate": 1.139482151043498e-06,
+      "loss": 2.9775,
+      "step": 2109
+    },
+    {
+      "epoch": 0.2731656795158106,
+      "grad_norm": 3.0254158567668004,
+      "learning_rate": 1.139321710248378e-06,
+      "loss": 2.8589,
+      "step": 2110
+    },
+    {
+      "epoch": 0.2731656795158106,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.6967995166778564,
+      "eval_runtime": 15.7066,
+      "eval_samples_per_second": 2.801,
+      "eval_steps_per_second": 0.191,
+      "step": 2110
+    },
+    {
+      "epoch": 0.27329514192316406,
+      "grad_norm": 2.956296196905673,
+      "learning_rate": 1.1391611643926243e-06,
+      "loss": 2.8242,
+      "step": 2111
+    },
+    {
+      "epoch": 0.2734246043305175,
+      "grad_norm": 4.845459647337277,
+      "learning_rate": 1.1390005135090313e-06,
+      "loss": 3.3892,
+      "step": 2112
+    },
+    {
+      "epoch": 0.27355406673787097,
+      "grad_norm": 2.6719008999607325,
+      "learning_rate": 1.1388397576304148e-06,
+      "loss": 3.292,
+      "step": 2113
+    },
+    {
+      "epoch": 0.2736835291452245,
+      "grad_norm": 6.575587963808951,
+      "learning_rate": 1.138678896789613e-06,
+      "loss": 3.0684,
+      "step": 2114
+    },
+    {
+      "epoch": 0.27381299155257793,
+      "grad_norm": 2.112497654626105,
+      "learning_rate": 1.1385179310194844e-06,
+      "loss": 2.6973,
+      "step": 2115
+    },
+    {
+      "epoch": 0.27381299155257793,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.6988192796707153,
+      "eval_runtime": 14.3833,
+      "eval_samples_per_second": 3.059,
+      "eval_steps_per_second": 0.209,
+      "step": 2115
+    },
+    {
+      "epoch": 0.2739424539599314,
+      "grad_norm": 4.61997539323117,
+      "learning_rate": 1.1383568603529101e-06,
+      "loss": 2.9185,
+      "step": 2116
+    },
+    {
+      "epoch": 0.27407191636728484,
+      "grad_norm": 5.3024146243248245,
+      "learning_rate": 1.138195684822792e-06,
+      "loss": 3.2292,
+      "step": 2117
+    },
+    {
+      "epoch": 0.2742013787746383,
+      "grad_norm": 2.9069223441234873,
+      "learning_rate": 1.1380344044620529e-06,
+      "loss": 2.7588,
+      "step": 2118
+    },
+    {
+      "epoch": 0.2743308411819918,
+      "grad_norm": 2.107055960389527,
+      "learning_rate": 1.137873019303638e-06,
+      "loss": 2.6982,
+      "step": 2119
+    },
+    {
+      "epoch": 0.27446030358934526,
+      "grad_norm": 2.2558812451416097,
+      "learning_rate": 1.1377115293805136e-06,
+      "loss": 2.8921,
+      "step": 2120
+    },
+    {
+      "epoch": 0.27446030358934526,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7009055614471436,
+      "eval_runtime": 13.3486,
+      "eval_samples_per_second": 3.296,
+      "eval_steps_per_second": 0.225,
+      "step": 2120
+    },
+    {
+      "epoch": 0.2745897659966987,
+      "grad_norm": 5.964490517844797,
+      "learning_rate": 1.137549934725667e-06,
+      "loss": 3.783,
+      "step": 2121
+    },
+    {
+      "epoch": 0.27471922840405216,
+      "grad_norm": 5.298975138485487,
+      "learning_rate": 1.1373882353721074e-06,
+      "loss": 3.0503,
+      "step": 2122
+    },
+    {
+      "epoch": 0.2748486908114056,
+      "grad_norm": 1.7339039669352063,
+      "learning_rate": 1.1372264313528652e-06,
+      "loss": 2.6182,
+      "step": 2123
+    },
+    {
+      "epoch": 0.2749781532187591,
+      "grad_norm": 2.515719379908353,
+      "learning_rate": 1.137064522700992e-06,
+      "loss": 3.1836,
+      "step": 2124
+    },
+    {
+      "epoch": 0.2751076156261126,
+      "grad_norm": 3.359680611777102,
+      "learning_rate": 1.1369025094495606e-06,
+      "loss": 3.2922,
+      "step": 2125
+    },
+    {
+      "epoch": 0.2751076156261126,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7035911083221436,
+      "eval_runtime": 14.7459,
+      "eval_samples_per_second": 2.984,
+      "eval_steps_per_second": 0.203,
+      "step": 2125
+    },
+    {
+      "epoch": 0.27523707803346603,
+      "grad_norm": 2.5993518130739846,
+      "learning_rate": 1.1367403916316663e-06,
+      "loss": 3.1245,
+      "step": 2126
+    },
+    {
+      "epoch": 0.2753665404408195,
+      "grad_norm": 2.409255174076651,
+      "learning_rate": 1.1365781692804242e-06,
+      "loss": 2.8301,
+      "step": 2127
+    },
+    {
+      "epoch": 0.27549600284817294,
+      "grad_norm": 2.5872875135177296,
+      "learning_rate": 1.1364158424289721e-06,
+      "loss": 2.7595,
+      "step": 2128
+    },
+    {
+      "epoch": 0.27562546525552645,
+      "grad_norm": 3.7293917500577956,
+      "learning_rate": 1.136253411110468e-06,
+      "loss": 2.7498,
+      "step": 2129
+    },
+    {
+      "epoch": 0.2757549276628799,
+      "grad_norm": 2.1062540891706947,
+      "learning_rate": 1.1360908753580921e-06,
+      "loss": 2.8459,
+      "step": 2130
+    },
+    {
+      "epoch": 0.2757549276628799,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.702925205230713,
+      "eval_runtime": 15.6268,
+      "eval_samples_per_second": 2.816,
+      "eval_steps_per_second": 0.192,
+      "step": 2130
+    },
+    {
+      "epoch": 0.27588439007023335,
+      "grad_norm": 3.2745964864625994,
+      "learning_rate": 1.1359282352050458e-06,
+      "loss": 3.1592,
+      "step": 2131
+    },
+    {
+      "epoch": 0.2760138524775868,
+      "grad_norm": 8.782837880746253,
+      "learning_rate": 1.135765490684551e-06,
+      "loss": 3.9993,
+      "step": 2132
+    },
+    {
+      "epoch": 0.27614331488494026,
+      "grad_norm": 2.3521076664452263,
+      "learning_rate": 1.1356026418298525e-06,
+      "loss": 2.7415,
+      "step": 2133
+    },
+    {
+      "epoch": 0.27627277729229377,
+      "grad_norm": 4.203858141927902,
+      "learning_rate": 1.1354396886742144e-06,
+      "loss": 3.0024,
+      "step": 2134
+    },
+    {
+      "epoch": 0.2764022396996472,
+      "grad_norm": 8.90162122217747,
+      "learning_rate": 1.1352766312509242e-06,
+      "loss": 4.02,
+      "step": 2135
+    },
+    {
+      "epoch": 0.2764022396996472,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.71142578125,
+      "eval_runtime": 16.0101,
+      "eval_samples_per_second": 2.748,
+      "eval_steps_per_second": 0.187,
+      "step": 2135
+    },
+    {
+      "epoch": 0.2765317021070007,
+      "grad_norm": 2.3512780688100317,
+      "learning_rate": 1.135113469593289e-06,
+      "loss": 2.8206,
+      "step": 2136
+    },
+    {
+      "epoch": 0.27666116451435413,
+      "grad_norm": 3.009829274006804,
+      "learning_rate": 1.134950203734638e-06,
+      "loss": 2.6372,
+      "step": 2137
+    },
+    {
+      "epoch": 0.2767906269217076,
+      "grad_norm": 3.198183367749438,
+      "learning_rate": 1.134786833708322e-06,
+      "loss": 2.8228,
+      "step": 2138
+    },
+    {
+      "epoch": 0.2769200893290611,
+      "grad_norm": 2.9172347484155075,
+      "learning_rate": 1.134623359547712e-06,
+      "loss": 2.9282,
+      "step": 2139
+    },
+    {
+      "epoch": 0.27704955173641455,
+      "grad_norm": 4.423684413944046,
+      "learning_rate": 1.1344597812862016e-06,
+      "loss": 3.2554,
+      "step": 2140
+    },
+    {
+      "epoch": 0.27704955173641455,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7084072828292847,
+      "eval_runtime": 14.5608,
+      "eval_samples_per_second": 3.022,
+      "eval_steps_per_second": 0.206,
+      "step": 2140
+    },
+    {
+      "epoch": 0.277179014143768,
+      "grad_norm": 3.7885066030285275,
+      "learning_rate": 1.1342960989572041e-06,
+      "loss": 3.0969,
+      "step": 2141
+    },
+    {
+      "epoch": 0.27730847655112145,
+      "grad_norm": 2.239514932877214,
+      "learning_rate": 1.1341323125941559e-06,
+      "loss": 2.9482,
+      "step": 2142
+    },
+    {
+      "epoch": 0.2774379389584749,
+      "grad_norm": 4.071755556531883,
+      "learning_rate": 1.1339684222305127e-06,
+      "loss": 3.0747,
+      "step": 2143
+    },
+    {
+      "epoch": 0.2775674013658284,
+      "grad_norm": 3.924183068294491,
+      "learning_rate": 1.1338044278997535e-06,
+      "loss": 3.0068,
+      "step": 2144
+    },
+    {
+      "epoch": 0.27769686377318187,
+      "grad_norm": 2.4623776182123587,
+      "learning_rate": 1.1336403296353765e-06,
+      "loss": 2.6719,
+      "step": 2145
+    },
+    {
+      "epoch": 0.27769686377318187,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7112038135528564,
+      "eval_runtime": 14.625,
+      "eval_samples_per_second": 3.009,
+      "eval_steps_per_second": 0.205,
+      "step": 2145
+    },
+    {
+      "epoch": 0.2778263261805353,
+      "grad_norm": 5.22692754769894,
+      "learning_rate": 1.1334761274709026e-06,
+      "loss": 4.0322,
+      "step": 2146
+    },
+    {
+      "epoch": 0.2779557885878888,
+      "grad_norm": 1.9911624215537151,
+      "learning_rate": 1.1333118214398733e-06,
+      "loss": 2.6882,
+      "step": 2147
+    },
+    {
+      "epoch": 0.2780852509952423,
+      "grad_norm": 2.030006439493325,
+      "learning_rate": 1.1331474115758517e-06,
+      "loss": 2.6562,
+      "step": 2148
+    },
+    {
+      "epoch": 0.27821471340259574,
+      "grad_norm": 2.71867645698034,
+      "learning_rate": 1.1329828979124212e-06,
+      "loss": 2.7881,
+      "step": 2149
+    },
+    {
+      "epoch": 0.2783441758099492,
+      "grad_norm": 3.6141895546256597,
+      "learning_rate": 1.1328182804831878e-06,
+      "loss": 2.8708,
+      "step": 2150
+    },
+    {
+      "epoch": 0.2783441758099492,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7105823755264282,
+      "eval_runtime": 14.4718,
+      "eval_samples_per_second": 3.04,
+      "eval_steps_per_second": 0.207,
+      "step": 2150
+    },
+    {
+      "epoch": 0.27847363821730264,
+      "grad_norm": 2.63689874732753,
+      "learning_rate": 1.1326535593217774e-06,
+      "loss": 3.0059,
+      "step": 2151
+    },
+    {
+      "epoch": 0.2786031006246561,
+      "grad_norm": 1.7058516968153965,
+      "learning_rate": 1.132488734461838e-06,
+      "loss": 2.3843,
+      "step": 2152
+    },
+    {
+      "epoch": 0.2787325630320096,
+      "grad_norm": 2.893541070293363,
+      "learning_rate": 1.132323805937038e-06,
+      "loss": 2.8291,
+      "step": 2153
+    },
+    {
+      "epoch": 0.27886202543936306,
+      "grad_norm": 5.965881199299705,
+      "learning_rate": 1.132158773781068e-06,
+      "loss": 3.4507,
+      "step": 2154
+    },
+    {
+      "epoch": 0.2789914878467165,
+      "grad_norm": 3.6014909656993015,
+      "learning_rate": 1.1319936380276386e-06,
+      "loss": 3.0967,
+      "step": 2155
+    },
+    {
+      "epoch": 0.2789914878467165,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7100497484207153,
+      "eval_runtime": 18.8346,
+      "eval_samples_per_second": 2.336,
+      "eval_steps_per_second": 0.159,
+      "step": 2155
+    },
+    {
+      "epoch": 0.27912095025406997,
+      "grad_norm": 2.1069149059004206,
+      "learning_rate": 1.1318283987104826e-06,
+      "loss": 2.498,
+      "step": 2156
+    },
+    {
+      "epoch": 0.2792504126614234,
+      "grad_norm": 4.101797173354059,
+      "learning_rate": 1.131663055863353e-06,
+      "loss": 3.4045,
+      "step": 2157
+    },
+    {
+      "epoch": 0.27937987506877693,
+      "grad_norm": 2.113263306558502,
+      "learning_rate": 1.1314976095200244e-06,
+      "loss": 2.7646,
+      "step": 2158
+    },
+    {
+      "epoch": 0.2795093374761304,
+      "grad_norm": 2.9764413417627655,
+      "learning_rate": 1.1313320597142932e-06,
+      "loss": 2.7197,
+      "step": 2159
+    },
+    {
+      "epoch": 0.27963879988348384,
+      "grad_norm": 6.3766328601527045,
+      "learning_rate": 1.1311664064799756e-06,
+      "loss": 3.187,
+      "step": 2160
+    },
+    {
+      "epoch": 0.27963879988348384,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7097834348678589,
+      "eval_runtime": 14.5663,
+      "eval_samples_per_second": 3.021,
+      "eval_steps_per_second": 0.206,
+      "step": 2160
+    },
+    {
+      "epoch": 0.2797682622908373,
+      "grad_norm": 2.0292187713159056,
+      "learning_rate": 1.1310006498509102e-06,
+      "loss": 2.717,
+      "step": 2161
+    },
+    {
+      "epoch": 0.27989772469819074,
+      "grad_norm": 2.5814056537863985,
+      "learning_rate": 1.130834789860956e-06,
+      "loss": 2.8288,
+      "step": 2162
+    },
+    {
+      "epoch": 0.28002718710554425,
+      "grad_norm": 3.984005890631887,
+      "learning_rate": 1.130668826543993e-06,
+      "loss": 3.1814,
+      "step": 2163
+    },
+    {
+      "epoch": 0.2801566495128977,
+      "grad_norm": 2.9525404229419685,
+      "learning_rate": 1.1305027599339228e-06,
+      "loss": 2.8315,
+      "step": 2164
+    },
+    {
+      "epoch": 0.28028611192025116,
+      "grad_norm": 3.0591320774934507,
+      "learning_rate": 1.1303365900646675e-06,
+      "loss": 2.7375,
+      "step": 2165
+    },
+    {
+      "epoch": 0.28028611192025116,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7112038135528564,
+      "eval_runtime": 14.2994,
+      "eval_samples_per_second": 3.077,
+      "eval_steps_per_second": 0.21,
+      "step": 2165
+    },
+    {
+      "epoch": 0.2804155743276046,
+      "grad_norm": 1.5957569858895388,
+      "learning_rate": 1.1301703169701713e-06,
+      "loss": 2.6003,
+      "step": 2166
+    },
+    {
+      "epoch": 0.28054503673495806,
+      "grad_norm": 1.9643281898373082,
+      "learning_rate": 1.1300039406843985e-06,
+      "loss": 2.8228,
+      "step": 2167
+    },
+    {
+      "epoch": 0.2806744991423116,
+      "grad_norm": 3.6598082645507644,
+      "learning_rate": 1.129837461241335e-06,
+      "loss": 2.8616,
+      "step": 2168
+    },
+    {
+      "epoch": 0.280803961549665,
+      "grad_norm": 4.983280124518397,
+      "learning_rate": 1.1296708786749872e-06,
+      "loss": 3.1206,
+      "step": 2169
+    },
+    {
+      "epoch": 0.2809334239570185,
+      "grad_norm": 4.4437933766836615,
+      "learning_rate": 1.129504193019383e-06,
+      "loss": 3.0837,
+      "step": 2170
+    },
+    {
+      "epoch": 0.2809334239570185,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7062321901321411,
+      "eval_runtime": 14.4183,
+      "eval_samples_per_second": 3.052,
+      "eval_steps_per_second": 0.208,
+      "step": 2170
+    },
+    {
+      "epoch": 0.28106288636437193,
+      "grad_norm": 5.137792537324083,
+      "learning_rate": 1.129337404308572e-06,
+      "loss": 2.9954,
+      "step": 2171
+    },
+    {
+      "epoch": 0.2811923487717254,
+      "grad_norm": 3.3956087855775547,
+      "learning_rate": 1.1291705125766235e-06,
+      "loss": 2.9438,
+      "step": 2172
+    },
+    {
+      "epoch": 0.2813218111790789,
+      "grad_norm": 1.766554721470038,
+      "learning_rate": 1.1290035178576288e-06,
+      "loss": 2.8037,
+      "step": 2173
+    },
+    {
+      "epoch": 0.28145127358643235,
+      "grad_norm": 2.044865806947404,
+      "learning_rate": 1.1288364201856998e-06,
+      "loss": 3.0144,
+      "step": 2174
+    },
+    {
+      "epoch": 0.2815807359937858,
+      "grad_norm": 4.498626168572207,
+      "learning_rate": 1.1286692195949696e-06,
+      "loss": 3.3193,
+      "step": 2175
+    },
+    {
+      "epoch": 0.2815807359937858,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7090510129928589,
+      "eval_runtime": 14.6364,
+      "eval_samples_per_second": 3.006,
+      "eval_steps_per_second": 0.205,
+      "step": 2175
+    },
+    {
+      "epoch": 0.28171019840113926,
+      "grad_norm": 2.144516666917743,
+      "learning_rate": 1.1285019161195926e-06,
+      "loss": 2.7476,
+      "step": 2176
+    },
+    {
+      "epoch": 0.2818396608084927,
+      "grad_norm": 8.829021740744638,
+      "learning_rate": 1.1283345097937438e-06,
+      "loss": 3.3428,
+      "step": 2177
+    },
+    {
+      "epoch": 0.2819691232158462,
+      "grad_norm": 1.683585915021965,
+      "learning_rate": 1.128167000651619e-06,
+      "loss": 2.6011,
+      "step": 2178
+    },
+    {
+      "epoch": 0.28209858562319967,
+      "grad_norm": 1.7929468918924358,
+      "learning_rate": 1.1279993887274356e-06,
+      "loss": 2.6406,
+      "step": 2179
+    },
+    {
+      "epoch": 0.2822280480305531,
+      "grad_norm": 2.1813791094384816,
+      "learning_rate": 1.1278316740554319e-06,
+      "loss": 2.9067,
+      "step": 2180
+    },
+    {
+      "epoch": 0.2822280480305531,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7127574682235718,
+      "eval_runtime": 14.0195,
+      "eval_samples_per_second": 3.138,
+      "eval_steps_per_second": 0.214,
+      "step": 2180
+    },
+    {
+      "epoch": 0.2823575104379066,
+      "grad_norm": 4.690397057184647,
+      "learning_rate": 1.1276638566698667e-06,
+      "loss": 3.2859,
+      "step": 2181
+    },
+    {
+      "epoch": 0.28248697284526003,
+      "grad_norm": 5.077977473556037,
+      "learning_rate": 1.1274959366050203e-06,
+      "loss": 3.4092,
+      "step": 2182
+    },
+    {
+      "epoch": 0.28261643525261354,
+      "grad_norm": 4.632403734886668,
+      "learning_rate": 1.1273279138951937e-06,
+      "loss": 2.9756,
+      "step": 2183
+    },
+    {
+      "epoch": 0.282745897659967,
+      "grad_norm": 2.325233264494713,
+      "learning_rate": 1.127159788574709e-06,
+      "loss": 2.9412,
+      "step": 2184
+    },
+    {
+      "epoch": 0.28287536006732045,
+      "grad_norm": 5.290140977108026,
+      "learning_rate": 1.1269915606779091e-06,
+      "loss": 3.4653,
+      "step": 2185
+    },
+    {
+      "epoch": 0.28287536006732045,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.708984375,
+      "eval_runtime": 14.488,
+      "eval_samples_per_second": 3.037,
+      "eval_steps_per_second": 0.207,
+      "step": 2185
+    },
+    {
+      "epoch": 0.2830048224746739,
+      "grad_norm": 1.7586007089312548,
+      "learning_rate": 1.126823230239158e-06,
+      "loss": 2.5349,
+      "step": 2186
+    },
+    {
+      "epoch": 0.28313428488202735,
+      "grad_norm": 2.1762952956245547,
+      "learning_rate": 1.1266547972928407e-06,
+      "loss": 2.7998,
+      "step": 2187
+    },
+    {
+      "epoch": 0.28326374728938086,
+      "grad_norm": 1.8217932197025986,
+      "learning_rate": 1.126486261873363e-06,
+      "loss": 2.9033,
+      "step": 2188
+    },
+    {
+      "epoch": 0.2833932096967343,
+      "grad_norm": 2.1877432405932264,
+      "learning_rate": 1.1263176240151516e-06,
+      "loss": 2.54,
+      "step": 2189
+    },
+    {
+      "epoch": 0.28352267210408777,
+      "grad_norm": 2.557256721108407,
+      "learning_rate": 1.1261488837526543e-06,
+      "loss": 2.688,
+      "step": 2190
+    },
+    {
+      "epoch": 0.28352267210408777,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7084072828292847,
+      "eval_runtime": 14.6506,
+      "eval_samples_per_second": 3.003,
+      "eval_steps_per_second": 0.205,
+      "step": 2190
+    },
+    {
+      "epoch": 0.2836521345114412,
+      "grad_norm": 2.0520267301152955,
+      "learning_rate": 1.1259800411203395e-06,
+      "loss": 2.6309,
+      "step": 2191
+    },
+    {
+      "epoch": 0.28378159691879473,
+      "grad_norm": 3.058716618496961,
+      "learning_rate": 1.1258110961526973e-06,
+      "loss": 2.6948,
+      "step": 2192
+    },
+    {
+      "epoch": 0.2839110593261482,
+      "grad_norm": 3.378395717359709,
+      "learning_rate": 1.1256420488842373e-06,
+      "loss": 3.1841,
+      "step": 2193
+    },
+    {
+      "epoch": 0.28404052173350164,
+      "grad_norm": 2.875080689847274,
+      "learning_rate": 1.1254728993494914e-06,
+      "loss": 2.925,
+      "step": 2194
+    },
+    {
+      "epoch": 0.2841699841408551,
+      "grad_norm": 2.840536060028901,
+      "learning_rate": 1.1253036475830116e-06,
+      "loss": 2.8147,
+      "step": 2195
+    },
+    {
+      "epoch": 0.2841699841408551,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7090065479278564,
+      "eval_runtime": 13.7959,
+      "eval_samples_per_second": 3.189,
+      "eval_steps_per_second": 0.217,
+      "step": 2195
+    },
+    {
+      "epoch": 0.28429944654820855,
+      "grad_norm": 2.804921306109501,
+      "learning_rate": 1.1251342936193713e-06,
+      "loss": 2.8738,
+      "step": 2196
+    },
+    {
+      "epoch": 0.28442890895556205,
+      "grad_norm": 2.3753824637981005,
+      "learning_rate": 1.124964837493164e-06,
+      "loss": 2.8962,
+      "step": 2197
+    },
+    {
+      "epoch": 0.2845583713629155,
+      "grad_norm": 2.6462816839800647,
+      "learning_rate": 1.124795279239005e-06,
+      "loss": 2.8684,
+      "step": 2198
+    },
+    {
+      "epoch": 0.28468783377026896,
+      "grad_norm": 3.0179004137056396,
+      "learning_rate": 1.12462561889153e-06,
+      "loss": 3.0945,
+      "step": 2199
+    },
+    {
+      "epoch": 0.2848172961776224,
+      "grad_norm": 2.4404753685891007,
+      "learning_rate": 1.1244558564853951e-06,
+      "loss": 2.5583,
+      "step": 2200
+    },
+    {
+      "epoch": 0.2848172961776224,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7096946239471436,
+      "eval_runtime": 14.7399,
+      "eval_samples_per_second": 2.985,
+      "eval_steps_per_second": 0.204,
+      "step": 2200
+    },
+    {
+      "epoch": 0.28494675858497587,
+      "grad_norm": 1.3181340072316032,
+      "learning_rate": 1.1242859920552782e-06,
+      "loss": 2.3582,
+      "step": 2201
+    },
+    {
+      "epoch": 0.2850762209923294,
+      "grad_norm": 2.572755352280697,
+      "learning_rate": 1.1241160256358771e-06,
+      "loss": 2.626,
+      "step": 2202
+    },
+    {
+      "epoch": 0.28520568339968283,
+      "grad_norm": 2.1431148659891854,
+      "learning_rate": 1.1239459572619114e-06,
+      "loss": 2.9226,
+      "step": 2203
+    },
+    {
+      "epoch": 0.2853351458070363,
+      "grad_norm": 1.755824724041719,
+      "learning_rate": 1.1237757869681207e-06,
+      "loss": 2.8125,
+      "step": 2204
+    },
+    {
+      "epoch": 0.28546460821438974,
+      "grad_norm": 2.3796147900861357,
+      "learning_rate": 1.1236055147892655e-06,
+      "loss": 2.9813,
+      "step": 2205
+    },
+    {
+      "epoch": 0.28546460821438974,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7118252515792847,
+      "eval_runtime": 13.9685,
+      "eval_samples_per_second": 3.15,
+      "eval_steps_per_second": 0.215,
+      "step": 2205
+    },
+    {
+      "epoch": 0.2855940706217432,
+      "grad_norm": 2.6189704283882884,
+      "learning_rate": 1.123435140760128e-06,
+      "loss": 3.0723,
+      "step": 2206
+    },
+    {
+      "epoch": 0.2857235330290967,
+      "grad_norm": 2.1538075409259925,
+      "learning_rate": 1.12326466491551e-06,
+      "loss": 2.8909,
+      "step": 2207
+    },
+    {
+      "epoch": 0.28585299543645015,
+      "grad_norm": 5.4951468246839195,
+      "learning_rate": 1.123094087290235e-06,
+      "loss": 3.0295,
+      "step": 2208
+    },
+    {
+      "epoch": 0.2859824578438036,
+      "grad_norm": 3.380280735939302,
+      "learning_rate": 1.1229234079191465e-06,
+      "loss": 2.8865,
+      "step": 2209
+    },
+    {
+      "epoch": 0.28611192025115706,
+      "grad_norm": 4.557573953496405,
+      "learning_rate": 1.1227526268371097e-06,
+      "loss": 2.7681,
+      "step": 2210
+    },
+    {
+      "epoch": 0.28611192025115706,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.705322265625,
+      "eval_runtime": 14.0208,
+      "eval_samples_per_second": 3.138,
+      "eval_steps_per_second": 0.214,
+      "step": 2210
+    },
+    {
+      "epoch": 0.2862413826585105,
+      "grad_norm": 2.252841511868517,
+      "learning_rate": 1.1225817440790099e-06,
+      "loss": 2.7974,
+      "step": 2211
+    },
+    {
+      "epoch": 0.286370845065864,
+      "grad_norm": 1.9598246311021126,
+      "learning_rate": 1.1224107596797533e-06,
+      "loss": 2.5537,
+      "step": 2212
+    },
+    {
+      "epoch": 0.2865003074732175,
+      "grad_norm": 3.0782150864525466,
+      "learning_rate": 1.1222396736742667e-06,
+      "loss": 3.0103,
+      "step": 2213
+    },
+    {
+      "epoch": 0.28662976988057093,
+      "grad_norm": 2.8642008163212114,
+      "learning_rate": 1.1220684860974984e-06,
+      "loss": 2.6191,
+      "step": 2214
+    },
+    {
+      "epoch": 0.2867592322879244,
+      "grad_norm": 1.9046792639731418,
+      "learning_rate": 1.1218971969844168e-06,
+      "loss": 2.5088,
+      "step": 2215
+    },
+    {
+      "epoch": 0.2867592322879244,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.71044921875,
+      "eval_runtime": 14.2214,
+      "eval_samples_per_second": 3.094,
+      "eval_steps_per_second": 0.211,
+      "step": 2215
+    },
+    {
+      "epoch": 0.28688869469527783,
+      "grad_norm": 2.3047614889168737,
+      "learning_rate": 1.1217258063700109e-06,
+      "loss": 2.9648,
+      "step": 2216
+    },
+    {
+      "epoch": 0.28701815710263134,
+      "grad_norm": 2.0683882833066893,
+      "learning_rate": 1.121554314289291e-06,
+      "loss": 2.7445,
+      "step": 2217
+    },
+    {
+      "epoch": 0.2871476195099848,
+      "grad_norm": 1.6373505253723921,
+      "learning_rate": 1.1213827207772876e-06,
+      "loss": 2.7129,
+      "step": 2218
+    },
+    {
+      "epoch": 0.28727708191733825,
+      "grad_norm": 2.2017342222320244,
+      "learning_rate": 1.1212110258690522e-06,
+      "loss": 2.6328,
+      "step": 2219
+    },
+    {
+      "epoch": 0.2874065443246917,
+      "grad_norm": 4.138346815214576,
+      "learning_rate": 1.1210392295996572e-06,
+      "loss": 3.5071,
+      "step": 2220
+    },
+    {
+      "epoch": 0.2874065443246917,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.706099033355713,
+      "eval_runtime": 13.7416,
+      "eval_samples_per_second": 3.202,
+      "eval_steps_per_second": 0.218,
+      "step": 2220
+    },
+    {
+      "epoch": 0.28753600673204516,
+      "grad_norm": 2.0699961736404457,
+      "learning_rate": 1.1208673320041954e-06,
+      "loss": 2.7375,
+      "step": 2221
+    },
+    {
+      "epoch": 0.28766546913939867,
+      "grad_norm": 2.132947347687674,
+      "learning_rate": 1.12069533311778e-06,
+      "loss": 2.7334,
+      "step": 2222
+    },
+    {
+      "epoch": 0.2877949315467521,
+      "grad_norm": 3.079188286394741,
+      "learning_rate": 1.1205232329755455e-06,
+      "loss": 2.8538,
+      "step": 2223
+    },
+    {
+      "epoch": 0.2879243939541056,
+      "grad_norm": 3.7583789646619277,
+      "learning_rate": 1.120351031612647e-06,
+      "loss": 2.7488,
+      "step": 2224
+    },
+    {
+      "epoch": 0.288053856361459,
+      "grad_norm": 3.195703457543387,
+      "learning_rate": 1.1201787290642603e-06,
+      "loss": 2.6689,
+      "step": 2225
+    },
+    {
+      "epoch": 0.288053856361459,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7103382349014282,
+      "eval_runtime": 14.2532,
+      "eval_samples_per_second": 3.087,
+      "eval_steps_per_second": 0.21,
+      "step": 2225
+    },
+    {
+      "epoch": 0.2881833187688125,
+      "grad_norm": 4.278129489842916,
+      "learning_rate": 1.120006325365581e-06,
+      "loss": 2.8003,
+      "step": 2226
+    },
+    {
+      "epoch": 0.288312781176166,
+      "grad_norm": 2.6893671025774553,
+      "learning_rate": 1.1198338205518264e-06,
+      "loss": 2.9006,
+      "step": 2227
+    },
+    {
+      "epoch": 0.28844224358351944,
+      "grad_norm": 1.7955960037653107,
+      "learning_rate": 1.1196612146582341e-06,
+      "loss": 2.6873,
+      "step": 2228
+    },
+    {
+      "epoch": 0.2885717059908729,
+      "grad_norm": 3.5214441972193504,
+      "learning_rate": 1.1194885077200625e-06,
+      "loss": 3.0398,
+      "step": 2229
+    },
+    {
+      "epoch": 0.28870116839822635,
+      "grad_norm": 3.3684742701925687,
+      "learning_rate": 1.1193156997725905e-06,
+      "loss": 2.9468,
+      "step": 2230
+    },
+    {
+      "epoch": 0.28870116839822635,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.705277919769287,
+      "eval_runtime": 13.6533,
+      "eval_samples_per_second": 3.223,
+      "eval_steps_per_second": 0.22,
+      "step": 2230
+    },
+    {
+      "epoch": 0.2888306308055798,
+      "grad_norm": 4.425861298149365,
+      "learning_rate": 1.1191427908511171e-06,
+      "loss": 3.2285,
+      "step": 2231
+    },
+    {
+      "epoch": 0.2889600932129333,
+      "grad_norm": 3.7181983296949146,
+      "learning_rate": 1.1189697809909632e-06,
+      "loss": 2.7056,
+      "step": 2232
+    },
+    {
+      "epoch": 0.28908955562028676,
+      "grad_norm": 7.126669767382223,
+      "learning_rate": 1.1187966702274692e-06,
+      "loss": 3.3049,
+      "step": 2233
+    },
+    {
+      "epoch": 0.2892190180276402,
+      "grad_norm": 5.403212436625998,
+      "learning_rate": 1.1186234585959962e-06,
+      "loss": 2.7183,
+      "step": 2234
+    },
+    {
+      "epoch": 0.28934848043499367,
+      "grad_norm": 2.1687910630315694,
+      "learning_rate": 1.1184501461319268e-06,
+      "loss": 2.7959,
+      "step": 2235
+    },
+    {
+      "epoch": 0.28934848043499367,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7081853151321411,
+      "eval_runtime": 14.6207,
+      "eval_samples_per_second": 3.009,
+      "eval_steps_per_second": 0.205,
+      "step": 2235
+    },
+    {
+      "epoch": 0.2894779428423472,
+      "grad_norm": 5.694975752507845,
+      "learning_rate": 1.1182767328706633e-06,
+      "loss": 3.2422,
+      "step": 2236
+    },
+    {
+      "epoch": 0.28960740524970063,
+      "grad_norm": 3.3886008328371173,
+      "learning_rate": 1.1181032188476288e-06,
+      "loss": 3.0723,
+      "step": 2237
+    },
+    {
+      "epoch": 0.2897368676570541,
+      "grad_norm": 2.798038835489442,
+      "learning_rate": 1.117929604098267e-06,
+      "loss": 2.7288,
+      "step": 2238
+    },
+    {
+      "epoch": 0.28986633006440754,
+      "grad_norm": 2.8702299878043536,
+      "learning_rate": 1.1177558886580423e-06,
+      "loss": 2.9917,
+      "step": 2239
+    },
+    {
+      "epoch": 0.289995792471761,
+      "grad_norm": 4.485289812994543,
+      "learning_rate": 1.1175820725624397e-06,
+      "loss": 2.9668,
+      "step": 2240
+    },
+    {
+      "epoch": 0.289995792471761,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7096058130264282,
+      "eval_runtime": 14.228,
+      "eval_samples_per_second": 3.092,
+      "eval_steps_per_second": 0.211,
+      "step": 2240
+    },
+    {
+      "epoch": 0.2901252548791145,
+      "grad_norm": 3.33508683390436,
+      "learning_rate": 1.1174081558469647e-06,
+      "loss": 2.8782,
+      "step": 2241
+    },
+    {
+      "epoch": 0.29025471728646796,
+      "grad_norm": 2.277414213004642,
+      "learning_rate": 1.117234138547143e-06,
+      "loss": 2.7119,
+      "step": 2242
+    },
+    {
+      "epoch": 0.2903841796938214,
+      "grad_norm": 1.7408559160921249,
+      "learning_rate": 1.1170600206985217e-06,
+      "loss": 2.5103,
+      "step": 2243
+    },
+    {
+      "epoch": 0.29051364210117486,
+      "grad_norm": 1.8986201651011303,
+      "learning_rate": 1.1168858023366672e-06,
+      "loss": 2.5569,
+      "step": 2244
+    },
+    {
+      "epoch": 0.2906431045085283,
+      "grad_norm": 4.292140335695194,
+      "learning_rate": 1.1167114834971676e-06,
+      "loss": 2.958,
+      "step": 2245
+    },
+    {
+      "epoch": 0.2906431045085283,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7058327198028564,
+      "eval_runtime": 14.5886,
+      "eval_samples_per_second": 3.016,
+      "eval_steps_per_second": 0.206,
+      "step": 2245
+    },
+    {
+      "epoch": 0.2907725669158818,
+      "grad_norm": 4.806996848236093,
+      "learning_rate": 1.116537064215631e-06,
+      "loss": 2.9956,
+      "step": 2246
+    },
+    {
+      "epoch": 0.2909020293232353,
+      "grad_norm": 2.9439820261049516,
+      "learning_rate": 1.116362544527686e-06,
+      "loss": 2.9871,
+      "step": 2247
+    },
+    {
+      "epoch": 0.29103149173058873,
+      "grad_norm": 2.579305957742088,
+      "learning_rate": 1.1161879244689818e-06,
+      "loss": 2.7568,
+      "step": 2248
+    },
+    {
+      "epoch": 0.2911609541379422,
+      "grad_norm": 8.68177069753059,
+      "learning_rate": 1.1160132040751882e-06,
+      "loss": 3.366,
+      "step": 2249
+    },
+    {
+      "epoch": 0.29129041654529564,
+      "grad_norm": 2.560707001929491,
+      "learning_rate": 1.1158383833819953e-06,
+      "loss": 2.4294,
+      "step": 2250
+    },
+    {
+      "epoch": 0.29129041654529564,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.713090419769287,
+      "eval_runtime": 13.8762,
+      "eval_samples_per_second": 3.171,
+      "eval_steps_per_second": 0.216,
+      "step": 2250
+    },
+    {
+      "epoch": 0.29141987895264915,
+      "grad_norm": 6.28404934998927,
+      "learning_rate": 1.1156634624251135e-06,
+      "loss": 3.0762,
+      "step": 2251
+    },
+    {
+      "epoch": 0.2915493413600026,
+      "grad_norm": 4.00146180175229,
+      "learning_rate": 1.1154884412402747e-06,
+      "loss": 3.4414,
+      "step": 2252
+    },
+    {
+      "epoch": 0.29167880376735605,
+      "grad_norm": 2.1571934650948927,
+      "learning_rate": 1.1153133198632296e-06,
+      "loss": 2.8652,
+      "step": 2253
+    },
+    {
+      "epoch": 0.2918082661747095,
+      "grad_norm": 2.6638967026898395,
+      "learning_rate": 1.1151380983297508e-06,
+      "loss": 2.5972,
+      "step": 2254
+    },
+    {
+      "epoch": 0.29193772858206296,
+      "grad_norm": 4.952341860227406,
+      "learning_rate": 1.114962776675631e-06,
+      "loss": 3.4187,
+      "step": 2255
+    },
+    {
+      "epoch": 0.29193772858206296,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.713578701019287,
+      "eval_runtime": 14.5247,
+      "eval_samples_per_second": 3.029,
+      "eval_steps_per_second": 0.207,
+      "step": 2255
+    },
+    {
+      "epoch": 0.29206719098941647,
+      "grad_norm": 3.129056514406557,
+      "learning_rate": 1.1147873549366827e-06,
+      "loss": 2.8794,
+      "step": 2256
+    },
+    {
+      "epoch": 0.2921966533967699,
+      "grad_norm": 2.1370040412469504,
+      "learning_rate": 1.1146118331487398e-06,
+      "loss": 2.637,
+      "step": 2257
+    },
+    {
+      "epoch": 0.2923261158041234,
+      "grad_norm": 2.457779404919745,
+      "learning_rate": 1.1144362113476558e-06,
+      "loss": 2.5955,
+      "step": 2258
+    },
+    {
+      "epoch": 0.29245557821147683,
+      "grad_norm": 4.973510975174887,
+      "learning_rate": 1.1142604895693055e-06,
+      "loss": 3.0322,
+      "step": 2259
+    },
+    {
+      "epoch": 0.2925850406188303,
+      "grad_norm": 1.8751211881924508,
+      "learning_rate": 1.114084667849583e-06,
+      "loss": 2.533,
+      "step": 2260
+    },
+    {
+      "epoch": 0.2925850406188303,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7122913599014282,
+      "eval_runtime": 15.8396,
+      "eval_samples_per_second": 2.778,
+      "eval_steps_per_second": 0.189,
+      "step": 2260
+    },
+    {
+      "epoch": 0.2927145030261838,
+      "grad_norm": 2.0114337638364734,
+      "learning_rate": 1.1139087462244038e-06,
+      "loss": 2.6738,
+      "step": 2261
+    },
+    {
+      "epoch": 0.29284396543353725,
+      "grad_norm": 5.29225363324954,
+      "learning_rate": 1.113732724729704e-06,
+      "loss": 2.9424,
+      "step": 2262
+    },
+    {
+      "epoch": 0.2929734278408907,
+      "grad_norm": 1.9882829970955298,
+      "learning_rate": 1.1135566034014384e-06,
+      "loss": 2.9143,
+      "step": 2263
+    },
+    {
+      "epoch": 0.29310289024824415,
+      "grad_norm": 2.542988863606374,
+      "learning_rate": 1.1133803822755843e-06,
+      "loss": 2.6362,
+      "step": 2264
+    },
+    {
+      "epoch": 0.2932323526555976,
+      "grad_norm": 2.3777462071011293,
+      "learning_rate": 1.1132040613881378e-06,
+      "loss": 2.7874,
+      "step": 2265
+    },
+    {
+      "epoch": 0.2932323526555976,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7107599973678589,
+      "eval_runtime": 14.3715,
+      "eval_samples_per_second": 3.062,
+      "eval_steps_per_second": 0.209,
+      "step": 2265
+    },
+    {
+      "epoch": 0.2933618150629511,
+      "grad_norm": 1.618431288280825,
+      "learning_rate": 1.1130276407751167e-06,
+      "loss": 2.3726,
+      "step": 2266
+    },
+    {
+      "epoch": 0.29349127747030457,
+      "grad_norm": 1.7050746325844315,
+      "learning_rate": 1.1128511204725579e-06,
+      "loss": 2.5781,
+      "step": 2267
+    },
+    {
+      "epoch": 0.293620739877658,
+      "grad_norm": 1.5178288970628233,
+      "learning_rate": 1.1126745005165194e-06,
+      "loss": 2.4829,
+      "step": 2268
+    },
+    {
+      "epoch": 0.2937502022850115,
+      "grad_norm": 4.677534963476821,
+      "learning_rate": 1.1124977809430794e-06,
+      "loss": 2.4578,
+      "step": 2269
+    },
+    {
+      "epoch": 0.29387966469236493,
+      "grad_norm": 2.0305203104429923,
+      "learning_rate": 1.1123209617883368e-06,
+      "loss": 2.5623,
+      "step": 2270
+    },
+    {
+      "epoch": 0.29387966469236493,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7080078125,
+      "eval_runtime": 14.8327,
+      "eval_samples_per_second": 2.966,
+      "eval_steps_per_second": 0.202,
+      "step": 2270
+    },
+    {
+      "epoch": 0.29400912709971844,
+      "grad_norm": 3.9270302038057214,
+      "learning_rate": 1.11214404308841e-06,
+      "loss": 2.7964,
+      "step": 2271
+    },
+    {
+      "epoch": 0.2941385895070719,
+      "grad_norm": 2.462702196807948,
+      "learning_rate": 1.1119670248794384e-06,
+      "loss": 2.502,
+      "step": 2272
+    },
+    {
+      "epoch": 0.29426805191442534,
+      "grad_norm": 4.34227279463719,
+      "learning_rate": 1.1117899071975819e-06,
+      "loss": 2.9658,
+      "step": 2273
+    },
+    {
+      "epoch": 0.2943975143217788,
+      "grad_norm": 4.6700635772428125,
+      "learning_rate": 1.1116126900790197e-06,
+      "loss": 2.9985,
+      "step": 2274
+    },
+    {
+      "epoch": 0.29452697672913225,
+      "grad_norm": 10.694848827573995,
+      "learning_rate": 1.1114353735599525e-06,
+      "loss": 3.7954,
+      "step": 2275
+    },
+    {
+      "epoch": 0.29452697672913225,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7079411745071411,
+      "eval_runtime": 13.7701,
+      "eval_samples_per_second": 3.195,
+      "eval_steps_per_second": 0.218,
+      "step": 2275
+    },
+    {
+      "epoch": 0.29465643913648576,
+      "grad_norm": 6.199938665661265,
+      "learning_rate": 1.111257957676601e-06,
+      "loss": 3.0825,
+      "step": 2276
+    },
+    {
+      "epoch": 0.2947859015438392,
+      "grad_norm": 2.3000652490116313,
+      "learning_rate": 1.111080442465205e-06,
+      "loss": 2.6416,
+      "step": 2277
+    },
+    {
+      "epoch": 0.29491536395119267,
+      "grad_norm": 1.608724350549297,
+      "learning_rate": 1.1109028279620266e-06,
+      "loss": 2.7773,
+      "step": 2278
+    },
+    {
+      "epoch": 0.2950448263585461,
+      "grad_norm": 3.5004184752607883,
+      "learning_rate": 1.1107251142033468e-06,
+      "loss": 2.8203,
+      "step": 2279
+    },
+    {
+      "epoch": 0.29517428876589963,
+      "grad_norm": 6.078583931008407,
+      "learning_rate": 1.110547301225467e-06,
+      "loss": 3.416,
+      "step": 2280
+    },
+    {
+      "epoch": 0.29517428876589963,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.713667392730713,
+      "eval_runtime": 14.9524,
+      "eval_samples_per_second": 2.943,
+      "eval_steps_per_second": 0.201,
+      "step": 2280
+    },
+    {
+      "epoch": 0.2953037511732531,
+      "grad_norm": 2.5075926605183962,
+      "learning_rate": 1.1103693890647097e-06,
+      "loss": 2.6797,
+      "step": 2281
+    },
+    {
+      "epoch": 0.29543321358060654,
+      "grad_norm": 2.725853349756007,
+      "learning_rate": 1.1101913777574164e-06,
+      "loss": 2.7324,
+      "step": 2282
+    },
+    {
+      "epoch": 0.29556267598796,
+      "grad_norm": 2.8218512901821917,
+      "learning_rate": 1.1100132673399498e-06,
+      "loss": 3.1111,
+      "step": 2283
+    },
+    {
+      "epoch": 0.29569213839531344,
+      "grad_norm": 5.088712382362241,
+      "learning_rate": 1.1098350578486927e-06,
+      "loss": 2.8916,
+      "step": 2284
+    },
+    {
+      "epoch": 0.29582160080266695,
+      "grad_norm": 5.494502510637724,
+      "learning_rate": 1.1096567493200477e-06,
+      "loss": 3.0034,
+      "step": 2285
+    },
+    {
+      "epoch": 0.29582160080266695,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7105602025985718,
+      "eval_runtime": 13.5554,
+      "eval_samples_per_second": 3.246,
+      "eval_steps_per_second": 0.221,
+      "step": 2285
+    },
+    {
+      "epoch": 0.2959510632100204,
+      "grad_norm": 3.1714050045330002,
+      "learning_rate": 1.1094783417904385e-06,
+      "loss": 3.0225,
+      "step": 2286
+    },
+    {
+      "epoch": 0.29608052561737386,
+      "grad_norm": 2.990935943544762,
+      "learning_rate": 1.1092998352963078e-06,
+      "loss": 2.9847,
+      "step": 2287
+    },
+    {
+      "epoch": 0.2962099880247273,
+      "grad_norm": 3.1417296172973757,
+      "learning_rate": 1.1091212298741195e-06,
+      "loss": 2.8726,
+      "step": 2288
+    },
+    {
+      "epoch": 0.29633945043208076,
+      "grad_norm": 1.759831550793735,
+      "learning_rate": 1.108942525560357e-06,
+      "loss": 2.6489,
+      "step": 2289
+    },
+    {
+      "epoch": 0.2964689128394343,
+      "grad_norm": 2.8246089030720825,
+      "learning_rate": 1.1087637223915249e-06,
+      "loss": 2.6543,
+      "step": 2290
+    },
+    {
+      "epoch": 0.2964689128394343,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7157536745071411,
+      "eval_runtime": 15.0643,
+      "eval_samples_per_second": 2.921,
+      "eval_steps_per_second": 0.199,
+      "step": 2290
+    },
+    {
+      "epoch": 0.2965983752467877,
+      "grad_norm": 4.684436599881485,
+      "learning_rate": 1.1085848204041468e-06,
+      "loss": 2.7144,
+      "step": 2291
+    },
+    {
+      "epoch": 0.2967278376541412,
+      "grad_norm": 5.325197371726781,
+      "learning_rate": 1.1084058196347676e-06,
+      "loss": 3.3655,
+      "step": 2292
+    },
+    {
+      "epoch": 0.29685730006149463,
+      "grad_norm": 3.5037919949092777,
+      "learning_rate": 1.1082267201199515e-06,
+      "loss": 2.814,
+      "step": 2293
+    },
+    {
+      "epoch": 0.2969867624688481,
+      "grad_norm": 3.2014843591185596,
+      "learning_rate": 1.1080475218962832e-06,
+      "loss": 2.9329,
+      "step": 2294
+    },
+    {
+      "epoch": 0.2971162248762016,
+      "grad_norm": 7.271137073874824,
+      "learning_rate": 1.1078682250003677e-06,
+      "loss": 3.5168,
+      "step": 2295
+    },
+    {
+      "epoch": 0.2971162248762016,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7149547338485718,
+      "eval_runtime": 14.2664,
+      "eval_samples_per_second": 3.084,
+      "eval_steps_per_second": 0.21,
+      "step": 2295
+    },
+    {
+      "epoch": 0.29724568728355505,
+      "grad_norm": 4.901814231369322,
+      "learning_rate": 1.1076888294688298e-06,
+      "loss": 3.1787,
+      "step": 2296
+    },
+    {
+      "epoch": 0.2973751496909085,
+      "grad_norm": 2.219257235597538,
+      "learning_rate": 1.1075093353383148e-06,
+      "loss": 2.8198,
+      "step": 2297
+    },
+    {
+      "epoch": 0.29750461209826196,
+      "grad_norm": 1.357931993492483,
+      "learning_rate": 1.1073297426454883e-06,
+      "loss": 2.3682,
+      "step": 2298
+    },
+    {
+      "epoch": 0.2976340745056154,
+      "grad_norm": 5.267575887649589,
+      "learning_rate": 1.1071500514270354e-06,
+      "loss": 3.2883,
+      "step": 2299
+    },
+    {
+      "epoch": 0.2977635369129689,
+      "grad_norm": 3.817558876719962,
+      "learning_rate": 1.1069702617196616e-06,
+      "loss": 3.0457,
+      "step": 2300
+    },
+    {
+      "epoch": 0.2977635369129689,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7156871557235718,
+      "eval_runtime": 13.8965,
+      "eval_samples_per_second": 3.166,
+      "eval_steps_per_second": 0.216,
+      "step": 2300
+    },
+    {
+      "epoch": 0.29789299932032237,
+      "grad_norm": 2.2686835609527787,
+      "learning_rate": 1.1067903735600928e-06,
+      "loss": 3.019,
+      "step": 2301
+    },
+    {
+      "epoch": 0.2980224617276758,
+      "grad_norm": 8.127838148779066,
+      "learning_rate": 1.1066103869850749e-06,
+      "loss": 3.4963,
+      "step": 2302
+    },
+    {
+      "epoch": 0.2981519241350293,
+      "grad_norm": 7.454411254477891,
+      "learning_rate": 1.1064303020313736e-06,
+      "loss": 3.6494,
+      "step": 2303
+    },
+    {
+      "epoch": 0.29828138654238273,
+      "grad_norm": 3.620502155925507,
+      "learning_rate": 1.1062501187357747e-06,
+      "loss": 2.8923,
+      "step": 2304
+    },
+    {
+      "epoch": 0.29841084894973624,
+      "grad_norm": 7.703118869295989,
+      "learning_rate": 1.1060698371350849e-06,
+      "loss": 3.0095,
+      "step": 2305
+    },
+    {
+      "epoch": 0.29841084894973624,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.720414638519287,
+      "eval_runtime": 13.8658,
+      "eval_samples_per_second": 3.173,
+      "eval_steps_per_second": 0.216,
+      "step": 2305
+    },
+    {
+      "epoch": 0.2985403113570897,
+      "grad_norm": 5.672908820019964,
+      "learning_rate": 1.10588945726613e-06,
+      "loss": 3.0894,
+      "step": 2306
+    },
+    {
+      "epoch": 0.29866977376444315,
+      "grad_norm": 1.7377480439703714,
+      "learning_rate": 1.105708979165756e-06,
+      "loss": 2.5754,
+      "step": 2307
+    },
+    {
+      "epoch": 0.2987992361717966,
+      "grad_norm": 2.755588309548811,
+      "learning_rate": 1.1055284028708297e-06,
+      "loss": 2.4216,
+      "step": 2308
+    },
+    {
+      "epoch": 0.29892869857915005,
+      "grad_norm": 4.3430421940125346,
+      "learning_rate": 1.1053477284182373e-06,
+      "loss": 2.8965,
+      "step": 2309
+    },
+    {
+      "epoch": 0.29905816098650356,
+      "grad_norm": 2.9417969979345906,
+      "learning_rate": 1.105166955844885e-06,
+      "loss": 2.8525,
+      "step": 2310
+    },
+    {
+      "epoch": 0.29905816098650356,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.719682216644287,
+      "eval_runtime": 14.4237,
+      "eval_samples_per_second": 3.051,
+      "eval_steps_per_second": 0.208,
+      "step": 2310
+    },
+    {
+      "epoch": 0.299187623393857,
+      "grad_norm": 5.464198809873469,
+      "learning_rate": 1.1049860851876996e-06,
+      "loss": 3.3276,
+      "step": 2311
+    },
+    {
+      "epoch": 0.29931708580121047,
+      "grad_norm": 4.174417484442815,
+      "learning_rate": 1.1048051164836271e-06,
+      "loss": 3.0596,
+      "step": 2312
+    },
+    {
+      "epoch": 0.2994465482085639,
+      "grad_norm": 2.583972012195666,
+      "learning_rate": 1.1046240497696345e-06,
+      "loss": 2.8989,
+      "step": 2313
+    },
+    {
+      "epoch": 0.2995760106159174,
+      "grad_norm": 4.851229023177933,
+      "learning_rate": 1.1044428850827083e-06,
+      "loss": 3.2666,
+      "step": 2314
+    },
+    {
+      "epoch": 0.2997054730232709,
+      "grad_norm": 4.0543917094532,
+      "learning_rate": 1.104261622459855e-06,
+      "loss": 3.0627,
+      "step": 2315
+    },
+    {
+      "epoch": 0.2997054730232709,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.723921298980713,
+      "eval_runtime": 14.797,
+      "eval_samples_per_second": 2.974,
+      "eval_steps_per_second": 0.203,
+      "step": 2315
+    },
+    {
+      "epoch": 0.29983493543062434,
+      "grad_norm": 4.332318062558142,
+      "learning_rate": 1.104080261938101e-06,
+      "loss": 2.9961,
+      "step": 2316
+    },
+    {
+      "epoch": 0.2999643978379778,
+      "grad_norm": 2.3053382259492112,
+      "learning_rate": 1.103898803554493e-06,
+      "loss": 2.739,
+      "step": 2317
+    },
+    {
+      "epoch": 0.30009386024533125,
+      "grad_norm": 5.479569282279513,
+      "learning_rate": 1.1037172473460976e-06,
+      "loss": 2.949,
+      "step": 2318
+    },
+    {
+      "epoch": 0.30022332265268475,
+      "grad_norm": 4.024732523908445,
+      "learning_rate": 1.1035355933500012e-06,
+      "loss": 2.7861,
+      "step": 2319
+    },
+    {
+      "epoch": 0.3003527850600382,
+      "grad_norm": 2.3241983284303758,
+      "learning_rate": 1.1033538416033105e-06,
+      "loss": 2.7703,
+      "step": 2320
+    },
+    {
+      "epoch": 0.3003527850600382,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7218571901321411,
+      "eval_runtime": 14.0487,
+      "eval_samples_per_second": 3.132,
+      "eval_steps_per_second": 0.214,
+      "step": 2320
+    },
+    {
+      "epoch": 0.30048224746739166,
+      "grad_norm": 2.8018145010136335,
+      "learning_rate": 1.1031719921431517e-06,
+      "loss": 2.9968,
+      "step": 2321
+    },
+    {
+      "epoch": 0.3006117098747451,
+      "grad_norm": 2.7994496703703327,
+      "learning_rate": 1.1029900450066715e-06,
+      "loss": 2.4685,
+      "step": 2322
+    },
+    {
+      "epoch": 0.30074117228209857,
+      "grad_norm": 2.359341875876882,
+      "learning_rate": 1.1028080002310361e-06,
+      "loss": 2.4744,
+      "step": 2323
+    },
+    {
+      "epoch": 0.3008706346894521,
+      "grad_norm": 3.3150985146941876,
+      "learning_rate": 1.1026258578534322e-06,
+      "loss": 2.825,
+      "step": 2324
+    },
+    {
+      "epoch": 0.30100009709680553,
+      "grad_norm": 2.311207815437649,
+      "learning_rate": 1.1024436179110654e-06,
+      "loss": 2.8203,
+      "step": 2325
+    },
+    {
+      "epoch": 0.30100009709680553,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.726806640625,
+      "eval_runtime": 14.0322,
+      "eval_samples_per_second": 3.136,
+      "eval_steps_per_second": 0.214,
+      "step": 2325
+    },
+    {
+      "epoch": 0.301129559504159,
+      "grad_norm": 2.7065193663620923,
+      "learning_rate": 1.1022612804411625e-06,
+      "loss": 2.9934,
+      "step": 2326
+    },
+    {
+      "epoch": 0.30125902191151244,
+      "grad_norm": 6.4200675593452265,
+      "learning_rate": 1.1020788454809695e-06,
+      "loss": 3.2385,
+      "step": 2327
+    },
+    {
+      "epoch": 0.3013884843188659,
+      "grad_norm": 3.918861280770376,
+      "learning_rate": 1.1018963130677523e-06,
+      "loss": 2.6145,
+      "step": 2328
+    },
+    {
+      "epoch": 0.3015179467262194,
+      "grad_norm": 2.73809457857486,
+      "learning_rate": 1.1017136832387967e-06,
+      "loss": 2.6553,
+      "step": 2329
+    },
+    {
+      "epoch": 0.30164740913357285,
+      "grad_norm": 1.8389206163537233,
+      "learning_rate": 1.101530956031409e-06,
+      "loss": 2.6775,
+      "step": 2330
+    },
+    {
+      "epoch": 0.30164740913357285,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7268732786178589,
+      "eval_runtime": 15.0474,
+      "eval_samples_per_second": 2.924,
+      "eval_steps_per_second": 0.199,
+      "step": 2330
+    },
+    {
+      "epoch": 0.3017768715409263,
+      "grad_norm": 2.3396551197934907,
+      "learning_rate": 1.1013481314829146e-06,
+      "loss": 2.7395,
+      "step": 2331
+    },
+    {
+      "epoch": 0.30190633394827976,
+      "grad_norm": 2.202878194051444,
+      "learning_rate": 1.1011652096306593e-06,
+      "loss": 2.4819,
+      "step": 2332
+    },
+    {
+      "epoch": 0.3020357963556332,
+      "grad_norm": 2.1879117401839703,
+      "learning_rate": 1.1009821905120087e-06,
+      "loss": 2.7915,
+      "step": 2333
+    },
+    {
+      "epoch": 0.3021652587629867,
+      "grad_norm": 2.233376883365686,
+      "learning_rate": 1.1007990741643476e-06,
+      "loss": 2.6038,
+      "step": 2334
+    },
+    {
+      "epoch": 0.3022947211703402,
+      "grad_norm": 4.103359338062117,
+      "learning_rate": 1.1006158606250816e-06,
+      "loss": 3.2329,
+      "step": 2335
+    },
+    {
+      "epoch": 0.3022947211703402,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7268732786178589,
+      "eval_runtime": 14.1474,
+      "eval_samples_per_second": 3.11,
+      "eval_steps_per_second": 0.212,
+      "step": 2335
+    },
+    {
+      "epoch": 0.30242418357769363,
+      "grad_norm": 3.1153083607544505,
+      "learning_rate": 1.100432549931636e-06,
+      "loss": 2.5474,
+      "step": 2336
+    },
+    {
+      "epoch": 0.3025536459850471,
+      "grad_norm": 2.2457332628017412,
+      "learning_rate": 1.1002491421214552e-06,
+      "loss": 2.6084,
+      "step": 2337
+    },
+    {
+      "epoch": 0.30268310839240054,
+      "grad_norm": 3.770406804964287,
+      "learning_rate": 1.1000656372320045e-06,
+      "loss": 2.7412,
+      "step": 2338
+    },
+    {
+      "epoch": 0.30281257079975404,
+      "grad_norm": 3.290597648196265,
+      "learning_rate": 1.0998820353007683e-06,
+      "loss": 3.0625,
+      "step": 2339
+    },
+    {
+      "epoch": 0.3029420332071075,
+      "grad_norm": 3.3024932641343496,
+      "learning_rate": 1.099698336365251e-06,
+      "loss": 2.8193,
+      "step": 2340
+    },
+    {
+      "epoch": 0.3029420332071075,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7282493114471436,
+      "eval_runtime": 14.838,
+      "eval_samples_per_second": 2.965,
+      "eval_steps_per_second": 0.202,
+      "step": 2340
+    },
+    {
+      "epoch": 0.30307149561446095,
+      "grad_norm": 2.673703705595916,
+      "learning_rate": 1.0995145404629764e-06,
+      "loss": 2.813,
+      "step": 2341
+    },
+    {
+      "epoch": 0.3032009580218144,
+      "grad_norm": 1.9614944555464808,
+      "learning_rate": 1.0993306476314892e-06,
+      "loss": 2.8359,
+      "step": 2342
+    },
+    {
+      "epoch": 0.30333042042916786,
+      "grad_norm": 4.077729405270378,
+      "learning_rate": 1.099146657908353e-06,
+      "loss": 2.5928,
+      "step": 2343
+    },
+    {
+      "epoch": 0.30345988283652137,
+      "grad_norm": 2.163937911996922,
+      "learning_rate": 1.0989625713311514e-06,
+      "loss": 2.6296,
+      "step": 2344
+    },
+    {
+      "epoch": 0.3035893452438748,
+      "grad_norm": 1.7445607961316325,
+      "learning_rate": 1.0987783879374878e-06,
+      "loss": 2.6089,
+      "step": 2345
+    },
+    {
+      "epoch": 0.3035893452438748,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7274724245071411,
+      "eval_runtime": 14.3576,
+      "eval_samples_per_second": 3.065,
+      "eval_steps_per_second": 0.209,
+      "step": 2345
+    },
+    {
+      "epoch": 0.3037188076512283,
+      "grad_norm": 3.306514351501887,
+      "learning_rate": 1.0985941077649854e-06,
+      "loss": 2.8125,
+      "step": 2346
+    },
+    {
+      "epoch": 0.3038482700585817,
+      "grad_norm": 1.5654794429116161,
+      "learning_rate": 1.098409730851287e-06,
+      "loss": 2.5405,
+      "step": 2347
+    },
+    {
+      "epoch": 0.3039777324659352,
+      "grad_norm": 3.028447916983368,
+      "learning_rate": 1.098225257234056e-06,
+      "loss": 2.9534,
+      "step": 2348
+    },
+    {
+      "epoch": 0.3041071948732887,
+      "grad_norm": 2.555158984805008,
+      "learning_rate": 1.0980406869509737e-06,
+      "loss": 2.5488,
+      "step": 2349
+    },
+    {
+      "epoch": 0.30423665728064214,
+      "grad_norm": 2.1549844505405695,
+      "learning_rate": 1.0978560200397434e-06,
+      "loss": 2.7473,
+      "step": 2350
+    },
+    {
+      "epoch": 0.30423665728064214,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.728227138519287,
+      "eval_runtime": 14.7034,
+      "eval_samples_per_second": 2.993,
+      "eval_steps_per_second": 0.204,
+      "step": 2350
+    },
+    {
+      "epoch": 0.3043661196879956,
+      "grad_norm": 4.344356650267298,
+      "learning_rate": 1.0976712565380868e-06,
+      "loss": 2.7275,
+      "step": 2351
+    },
+    {
+      "epoch": 0.30449558209534905,
+      "grad_norm": 1.9759526783379688,
+      "learning_rate": 1.0974863964837452e-06,
+      "loss": 2.7452,
+      "step": 2352
+    },
+    {
+      "epoch": 0.3046250445027025,
+      "grad_norm": 2.918339179192121,
+      "learning_rate": 1.0973014399144804e-06,
+      "loss": 2.8164,
+      "step": 2353
+    },
+    {
+      "epoch": 0.304754506910056,
+      "grad_norm": 1.8913136195210352,
+      "learning_rate": 1.0971163868680732e-06,
+      "loss": 2.567,
+      "step": 2354
+    },
+    {
+      "epoch": 0.30488396931740946,
+      "grad_norm": 3.5293196141146024,
+      "learning_rate": 1.0969312373823246e-06,
+      "loss": 2.5986,
+      "step": 2355
+    },
+    {
+      "epoch": 0.30488396931740946,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7251864671707153,
+      "eval_runtime": 13.9786,
+      "eval_samples_per_second": 3.148,
+      "eval_steps_per_second": 0.215,
+      "step": 2355
+    },
+    {
+      "epoch": 0.3050134317247629,
+      "grad_norm": 1.3062140494467265,
+      "learning_rate": 1.0967459914950555e-06,
+      "loss": 2.3357,
+      "step": 2356
+    },
+    {
+      "epoch": 0.30514289413211637,
+      "grad_norm": 2.7197663773225043,
+      "learning_rate": 1.0965606492441055e-06,
+      "loss": 3.1416,
+      "step": 2357
+    },
+    {
+      "epoch": 0.3052723565394698,
+      "grad_norm": 3.2655455323575144,
+      "learning_rate": 1.0963752106673348e-06,
+      "loss": 2.874,
+      "step": 2358
+    },
+    {
+      "epoch": 0.30540181894682333,
+      "grad_norm": 3.49802099192436,
+      "learning_rate": 1.0961896758026231e-06,
+      "loss": 2.9258,
+      "step": 2359
+    },
+    {
+      "epoch": 0.3055312813541768,
+      "grad_norm": 1.9628910595002096,
+      "learning_rate": 1.0960040446878691e-06,
+      "loss": 2.7124,
+      "step": 2360
+    },
+    {
+      "epoch": 0.3055312813541768,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7253639698028564,
+      "eval_runtime": 14.1944,
+      "eval_samples_per_second": 3.1,
+      "eval_steps_per_second": 0.211,
+      "step": 2360
+    },
+    {
+      "epoch": 0.30566074376153024,
+      "grad_norm": 2.4587835627703334,
+      "learning_rate": 1.0958183173609927e-06,
+      "loss": 2.7305,
+      "step": 2361
+    },
+    {
+      "epoch": 0.3057902061688837,
+      "grad_norm": 1.6832969713945503,
+      "learning_rate": 1.0956324938599317e-06,
+      "loss": 2.4036,
+      "step": 2362
+    },
+    {
+      "epoch": 0.3059196685762372,
+      "grad_norm": 4.404962521109919,
+      "learning_rate": 1.0954465742226446e-06,
+      "loss": 2.76,
+      "step": 2363
+    },
+    {
+      "epoch": 0.30604913098359066,
+      "grad_norm": 2.374721501109396,
+      "learning_rate": 1.0952605584871092e-06,
+      "loss": 2.5376,
+      "step": 2364
+    },
+    {
+      "epoch": 0.3061785933909441,
+      "grad_norm": 3.664580468008269,
+      "learning_rate": 1.095074446691323e-06,
+      "loss": 2.5071,
+      "step": 2365
+    },
+    {
+      "epoch": 0.3061785933909441,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.723832607269287,
+      "eval_runtime": 14.6734,
+      "eval_samples_per_second": 2.999,
+      "eval_steps_per_second": 0.204,
+      "step": 2365
+    },
+    {
+      "epoch": 0.30630805579829756,
+      "grad_norm": 2.9028991689676436,
+      "learning_rate": 1.0948882388733028e-06,
+      "loss": 2.6248,
+      "step": 2366
+    },
+    {
+      "epoch": 0.306437518205651,
+      "grad_norm": 1.7629813529037595,
+      "learning_rate": 1.094701935071086e-06,
+      "loss": 2.5582,
+      "step": 2367
+    },
+    {
+      "epoch": 0.3065669806130045,
+      "grad_norm": 2.092818393140363,
+      "learning_rate": 1.0945155353227284e-06,
+      "loss": 2.6292,
+      "step": 2368
+    },
+    {
+      "epoch": 0.306696443020358,
+      "grad_norm": 6.977125321100761,
+      "learning_rate": 1.094329039666306e-06,
+      "loss": 2.5737,
+      "step": 2369
+    },
+    {
+      "epoch": 0.30682590542771143,
+      "grad_norm": 8.652030826571032,
+      "learning_rate": 1.0941424481399145e-06,
+      "loss": 3.5762,
+      "step": 2370
+    },
+    {
+      "epoch": 0.30682590542771143,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7288485765457153,
+      "eval_runtime": 15.0853,
+      "eval_samples_per_second": 2.917,
+      "eval_steps_per_second": 0.199,
+      "step": 2370
+    },
+    {
+      "epoch": 0.3069553678350649,
+      "grad_norm": 2.345543682802869,
+      "learning_rate": 1.0939557607816687e-06,
+      "loss": 2.759,
+      "step": 2371
+    },
+    {
+      "epoch": 0.30708483024241834,
+      "grad_norm": 3.5004531029562096,
+      "learning_rate": 1.0937689776297036e-06,
+      "loss": 2.8242,
+      "step": 2372
+    },
+    {
+      "epoch": 0.30721429264977185,
+      "grad_norm": 3.0975953965554592,
+      "learning_rate": 1.0935820987221734e-06,
+      "loss": 2.8235,
+      "step": 2373
+    },
+    {
+      "epoch": 0.3073437550571253,
+      "grad_norm": 3.548805030095774,
+      "learning_rate": 1.0933951240972517e-06,
+      "loss": 3.1929,
+      "step": 2374
+    },
+    {
+      "epoch": 0.30747321746447875,
+      "grad_norm": 2.3444064758342638,
+      "learning_rate": 1.0932080537931319e-06,
+      "loss": 2.7817,
+      "step": 2375
+    },
+    {
+      "epoch": 0.30747321746447875,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.728759765625,
+      "eval_runtime": 14.9551,
+      "eval_samples_per_second": 2.942,
+      "eval_steps_per_second": 0.201,
+      "step": 2375
+    },
+    {
+      "epoch": 0.3076026798718322,
+      "grad_norm": 1.9652062446980842,
+      "learning_rate": 1.093020887848027e-06,
+      "loss": 2.5242,
+      "step": 2376
+    },
+    {
+      "epoch": 0.30773214227918566,
+      "grad_norm": 4.292845713711578,
+      "learning_rate": 1.0928336263001694e-06,
+      "loss": 2.8331,
+      "step": 2377
+    },
+    {
+      "epoch": 0.30786160468653917,
+      "grad_norm": 2.1080984858369374,
+      "learning_rate": 1.0926462691878113e-06,
+      "loss": 2.7549,
+      "step": 2378
+    },
+    {
+      "epoch": 0.3079910670938926,
+      "grad_norm": 2.9578317131351026,
+      "learning_rate": 1.0924588165492236e-06,
+      "loss": 2.9641,
+      "step": 2379
+    },
+    {
+      "epoch": 0.3081205295012461,
+      "grad_norm": 4.041543951211056,
+      "learning_rate": 1.0922712684226977e-06,
+      "loss": 3.1533,
+      "step": 2380
+    },
+    {
+      "epoch": 0.3081205295012461,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7217463254928589,
+      "eval_runtime": 14.1731,
+      "eval_samples_per_second": 3.104,
+      "eval_steps_per_second": 0.212,
+      "step": 2380
+    },
+    {
+      "epoch": 0.30824999190859953,
+      "grad_norm": 3.7349426704360096,
+      "learning_rate": 1.0920836248465444e-06,
+      "loss": 2.8679,
+      "step": 2381
+    },
+    {
+      "epoch": 0.308379454315953,
+      "grad_norm": 2.3819826233324326,
+      "learning_rate": 1.091895885859093e-06,
+      "loss": 2.7231,
+      "step": 2382
+    },
+    {
+      "epoch": 0.3085089167233065,
+      "grad_norm": 1.5817564064631358,
+      "learning_rate": 1.0917080514986936e-06,
+      "loss": 2.5522,
+      "step": 2383
+    },
+    {
+      "epoch": 0.30863837913065995,
+      "grad_norm": 3.1587163866305343,
+      "learning_rate": 1.091520121803715e-06,
+      "loss": 3.1121,
+      "step": 2384
+    },
+    {
+      "epoch": 0.3087678415380134,
+      "grad_norm": 4.826990616632971,
+      "learning_rate": 1.0913320968125454e-06,
+      "loss": 2.856,
+      "step": 2385
+    },
+    {
+      "epoch": 0.3087678415380134,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7281161546707153,
+      "eval_runtime": 13.8847,
+      "eval_samples_per_second": 3.169,
+      "eval_steps_per_second": 0.216,
+      "step": 2385
+    },
+    {
+      "epoch": 0.30889730394536685,
+      "grad_norm": 2.2901727560733858,
+      "learning_rate": 1.0911439765635927e-06,
+      "loss": 2.6177,
+      "step": 2386
+    },
+    {
+      "epoch": 0.3090267663527203,
+      "grad_norm": 2.7665876493584816,
+      "learning_rate": 1.090955761095285e-06,
+      "loss": 2.4453,
+      "step": 2387
+    },
+    {
+      "epoch": 0.3091562287600738,
+      "grad_norm": 2.071003556745559,
+      "learning_rate": 1.0907674504460684e-06,
+      "loss": 2.666,
+      "step": 2388
+    },
+    {
+      "epoch": 0.30928569116742727,
+      "grad_norm": 5.602577345978535,
+      "learning_rate": 1.0905790446544094e-06,
+      "loss": 2.7539,
+      "step": 2389
+    },
+    {
+      "epoch": 0.3094151535747807,
+      "grad_norm": 2.509865921768195,
+      "learning_rate": 1.0903905437587934e-06,
+      "loss": 2.9897,
+      "step": 2390
+    },
+    {
+      "epoch": 0.3094151535747807,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.723100185394287,
+      "eval_runtime": 14.0773,
+      "eval_samples_per_second": 3.126,
+      "eval_steps_per_second": 0.213,
+      "step": 2390
+    },
+    {
+      "epoch": 0.3095446159821342,
+      "grad_norm": 4.721741217295041,
+      "learning_rate": 1.090201947797726e-06,
+      "loss": 2.9995,
+      "step": 2391
+    },
+    {
+      "epoch": 0.30967407838948763,
+      "grad_norm": 3.1456676592206936,
+      "learning_rate": 1.0900132568097315e-06,
+      "loss": 2.9624,
+      "step": 2392
+    },
+    {
+      "epoch": 0.30980354079684114,
+      "grad_norm": 1.4610708989058652,
+      "learning_rate": 1.089824470833354e-06,
+      "loss": 2.561,
+      "step": 2393
+    },
+    {
+      "epoch": 0.3099330032041946,
+      "grad_norm": 2.6181156043233225,
+      "learning_rate": 1.0896355899071568e-06,
+      "loss": 2.5471,
+      "step": 2394
+    },
+    {
+      "epoch": 0.31006246561154804,
+      "grad_norm": 5.9134217414205485,
+      "learning_rate": 1.0894466140697225e-06,
+      "loss": 2.7651,
+      "step": 2395
+    },
+    {
+      "epoch": 0.31006246561154804,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7242985963821411,
+      "eval_runtime": 15.146,
+      "eval_samples_per_second": 2.905,
+      "eval_steps_per_second": 0.198,
+      "step": 2395
+    },
+    {
+      "epoch": 0.3101919280189015,
+      "grad_norm": 8.412361559878502,
+      "learning_rate": 1.0892575433596533e-06,
+      "loss": 3.1011,
+      "step": 2396
+    },
+    {
+      "epoch": 0.31032139042625495,
+      "grad_norm": 5.419321444894031,
+      "learning_rate": 1.0890683778155708e-06,
+      "loss": 2.8979,
+      "step": 2397
+    },
+    {
+      "epoch": 0.31045085283360846,
+      "grad_norm": 8.117119304665449,
+      "learning_rate": 1.0888791174761162e-06,
+      "loss": 3.2632,
+      "step": 2398
+    },
+    {
+      "epoch": 0.3105803152409619,
+      "grad_norm": 5.778830542819061,
+      "learning_rate": 1.088689762379949e-06,
+      "loss": 3.231,
+      "step": 2399
+    },
+    {
+      "epoch": 0.31070977764831537,
+      "grad_norm": 3.503842564948514,
+      "learning_rate": 1.0885003125657497e-06,
+      "loss": 2.8774,
+      "step": 2400
+    },
+    {
+      "epoch": 0.31070977764831537,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7264958620071411,
+      "eval_runtime": 14.483,
+      "eval_samples_per_second": 3.038,
+      "eval_steps_per_second": 0.207,
+      "step": 2400
+    },
+    {
+      "epoch": 0.3108392400556688,
+      "grad_norm": 2.609445130844578,
+      "learning_rate": 1.0883107680722167e-06,
+      "loss": 2.793,
+      "step": 2401
+    },
+    {
+      "epoch": 0.3109687024630223,
+      "grad_norm": 3.201048399294415,
+      "learning_rate": 1.0881211289380682e-06,
+      "loss": 2.7988,
+      "step": 2402
+    },
+    {
+      "epoch": 0.3110981648703758,
+      "grad_norm": 2.9433106975568752,
+      "learning_rate": 1.0879313952020424e-06,
+      "loss": 2.7368,
+      "step": 2403
+    },
+    {
+      "epoch": 0.31122762727772924,
+      "grad_norm": 3.287459843621182,
+      "learning_rate": 1.087741566902896e-06,
+      "loss": 2.4546,
+      "step": 2404
+    },
+    {
+      "epoch": 0.3113570896850827,
+      "grad_norm": 2.545160737961092,
+      "learning_rate": 1.0875516440794048e-06,
+      "loss": 2.5977,
+      "step": 2405
+    },
+    {
+      "epoch": 0.3113570896850827,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7250310182571411,
+      "eval_runtime": 14.8953,
+      "eval_samples_per_second": 2.954,
+      "eval_steps_per_second": 0.201,
+      "step": 2405
+    },
+    {
+      "epoch": 0.31148655209243614,
+      "grad_norm": 4.7298583669312215,
+      "learning_rate": 1.0873616267703651e-06,
+      "loss": 2.7993,
+      "step": 2406
+    },
+    {
+      "epoch": 0.31161601449978965,
+      "grad_norm": 1.6697275418840263,
+      "learning_rate": 1.0871715150145915e-06,
+      "loss": 2.6411,
+      "step": 2407
+    },
+    {
+      "epoch": 0.3117454769071431,
+      "grad_norm": 4.145696219088108,
+      "learning_rate": 1.086981308850918e-06,
+      "loss": 3.0166,
+      "step": 2408
+    },
+    {
+      "epoch": 0.31187493931449656,
+      "grad_norm": 4.1341049890440935,
+      "learning_rate": 1.0867910083181984e-06,
+      "loss": 3.1152,
+      "step": 2409
+    },
+    {
+      "epoch": 0.31200440172185,
+      "grad_norm": 1.7365756726224395,
+      "learning_rate": 1.0866006134553053e-06,
+      "loss": 2.5522,
+      "step": 2410
+    },
+    {
+      "epoch": 0.31200440172185,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7258079051971436,
+      "eval_runtime": 14.1523,
+      "eval_samples_per_second": 3.109,
+      "eval_steps_per_second": 0.212,
+      "step": 2410
+    },
+    {
+      "epoch": 0.31213386412920346,
+      "grad_norm": 3.8991441737204475,
+      "learning_rate": 1.0864101243011306e-06,
+      "loss": 2.5142,
+      "step": 2411
+    },
+    {
+      "epoch": 0.312263326536557,
+      "grad_norm": 3.2662779745530073,
+      "learning_rate": 1.0862195408945856e-06,
+      "loss": 2.8677,
+      "step": 2412
+    },
+    {
+      "epoch": 0.3123927889439104,
+      "grad_norm": 4.277262769897312,
+      "learning_rate": 1.0860288632746007e-06,
+      "loss": 2.8499,
+      "step": 2413
+    },
+    {
+      "epoch": 0.3125222513512639,
+      "grad_norm": 5.01450381901903,
+      "learning_rate": 1.085838091480126e-06,
+      "loss": 2.8115,
+      "step": 2414
+    },
+    {
+      "epoch": 0.31265171375861733,
+      "grad_norm": 5.140632361962117,
+      "learning_rate": 1.0856472255501307e-06,
+      "loss": 2.7261,
+      "step": 2415
+    },
+    {
+      "epoch": 0.31265171375861733,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7264958620071411,
+      "eval_runtime": 15.1146,
+      "eval_samples_per_second": 2.911,
+      "eval_steps_per_second": 0.198,
+      "step": 2415
+    },
+    {
+      "epoch": 0.3127811761659708,
+      "grad_norm": 1.5762636180296328,
+      "learning_rate": 1.0854562655236022e-06,
+      "loss": 2.6777,
+      "step": 2416
+    },
+    {
+      "epoch": 0.3129106385733243,
+      "grad_norm": 2.811376902037237,
+      "learning_rate": 1.0852652114395488e-06,
+      "loss": 2.6733,
+      "step": 2417
+    },
+    {
+      "epoch": 0.31304010098067775,
+      "grad_norm": 2.764862726756628,
+      "learning_rate": 1.0850740633369965e-06,
+      "loss": 2.3457,
+      "step": 2418
+    },
+    {
+      "epoch": 0.3131695633880312,
+      "grad_norm": 6.20786143692676,
+      "learning_rate": 1.0848828212549917e-06,
+      "loss": 3.0454,
+      "step": 2419
+    },
+    {
+      "epoch": 0.31329902579538466,
+      "grad_norm": 2.931228851104307,
+      "learning_rate": 1.0846914852325994e-06,
+      "loss": 2.9658,
+      "step": 2420
+    },
+    {
+      "epoch": 0.31329902579538466,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.730668544769287,
+      "eval_runtime": 15.5422,
+      "eval_samples_per_second": 2.831,
+      "eval_steps_per_second": 0.193,
+      "step": 2420
+    },
+    {
+      "epoch": 0.3134284882027381,
+      "grad_norm": 4.246217231741376,
+      "learning_rate": 1.0845000553089032e-06,
+      "loss": 2.8145,
+      "step": 2421
+    },
+    {
+      "epoch": 0.3135579506100916,
+      "grad_norm": 3.7715143370034263,
+      "learning_rate": 1.0843085315230076e-06,
+      "loss": 3.0273,
+      "step": 2422
+    },
+    {
+      "epoch": 0.31368741301744507,
+      "grad_norm": 1.8352538755017078,
+      "learning_rate": 1.0841169139140345e-06,
+      "loss": 2.696,
+      "step": 2423
+    },
+    {
+      "epoch": 0.3138168754247985,
+      "grad_norm": 3.951555566734122,
+      "learning_rate": 1.0839252025211259e-06,
+      "loss": 2.5566,
+      "step": 2424
+    },
+    {
+      "epoch": 0.313946337832152,
+      "grad_norm": 5.255888607410781,
+      "learning_rate": 1.0837333973834423e-06,
+      "loss": 2.8184,
+      "step": 2425
+    },
+    {
+      "epoch": 0.313946337832152,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7328213453292847,
+      "eval_runtime": 13.8188,
+      "eval_samples_per_second": 3.184,
+      "eval_steps_per_second": 0.217,
+      "step": 2425
+    },
+    {
+      "epoch": 0.31407580023950543,
+      "grad_norm": 4.27767564354849,
+      "learning_rate": 1.0835414985401646e-06,
+      "loss": 2.6543,
+      "step": 2426
+    },
+    {
+      "epoch": 0.31420526264685894,
+      "grad_norm": 2.0119119746602645,
+      "learning_rate": 1.0833495060304916e-06,
+      "loss": 2.5239,
+      "step": 2427
+    },
+    {
+      "epoch": 0.3143347250542124,
+      "grad_norm": 2.0564351305921904,
+      "learning_rate": 1.0831574198936414e-06,
+      "loss": 2.7466,
+      "step": 2428
+    },
+    {
+      "epoch": 0.31446418746156585,
+      "grad_norm": 3.328244710080324,
+      "learning_rate": 1.082965240168852e-06,
+      "loss": 2.8818,
+      "step": 2429
+    },
+    {
+      "epoch": 0.3145936498689193,
+      "grad_norm": 5.202876420862465,
+      "learning_rate": 1.08277296689538e-06,
+      "loss": 2.9172,
+      "step": 2430
+    },
+    {
+      "epoch": 0.3145936498689193,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7305575609207153,
+      "eval_runtime": 14.3207,
+      "eval_samples_per_second": 3.072,
+      "eval_steps_per_second": 0.209,
+      "step": 2430
+    },
+    {
+      "epoch": 0.31472311227627275,
+      "grad_norm": 2.3954730238284814,
+      "learning_rate": 1.0825806001125003e-06,
+      "loss": 2.6597,
+      "step": 2431
+    },
+    {
+      "epoch": 0.31485257468362626,
+      "grad_norm": 2.45988008419959,
+      "learning_rate": 1.0823881398595085e-06,
+      "loss": 2.7405,
+      "step": 2432
+    },
+    {
+      "epoch": 0.3149820370909797,
+      "grad_norm": 1.784575661810735,
+      "learning_rate": 1.0821955861757185e-06,
+      "loss": 2.6685,
+      "step": 2433
+    },
+    {
+      "epoch": 0.31511149949833317,
+      "grad_norm": 2.581102977328374,
+      "learning_rate": 1.082002939100463e-06,
+      "loss": 2.7251,
+      "step": 2434
+    },
+    {
+      "epoch": 0.3152409619056866,
+      "grad_norm": 3.8718914848734465,
+      "learning_rate": 1.0818101986730942e-06,
+      "loss": 2.7039,
+      "step": 2435
+    },
+    {
+      "epoch": 0.3152409619056866,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7306907176971436,
+      "eval_runtime": 14.7789,
+      "eval_samples_per_second": 2.977,
+      "eval_steps_per_second": 0.203,
+      "step": 2435
+    },
+    {
+      "epoch": 0.3153704243130401,
+      "grad_norm": 2.655205853014172,
+      "learning_rate": 1.0816173649329831e-06,
+      "loss": 2.6807,
+      "step": 2436
+    },
+    {
+      "epoch": 0.3154998867203936,
+      "grad_norm": 2.6546314761607563,
+      "learning_rate": 1.0814244379195199e-06,
+      "loss": 2.7903,
+      "step": 2437
+    },
+    {
+      "epoch": 0.31562934912774704,
+      "grad_norm": 2.115226926621325,
+      "learning_rate": 1.0812314176721141e-06,
+      "loss": 2.7529,
+      "step": 2438
+    },
+    {
+      "epoch": 0.3157588115351005,
+      "grad_norm": 2.36549653962922,
+      "learning_rate": 1.0810383042301938e-06,
+      "loss": 2.4944,
+      "step": 2439
+    },
+    {
+      "epoch": 0.31588827394245395,
+      "grad_norm": 3.4032755258886045,
+      "learning_rate": 1.0808450976332064e-06,
+      "loss": 2.947,
+      "step": 2440
+    },
+    {
+      "epoch": 0.31588827394245395,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7308460474014282,
+      "eval_runtime": 14.0494,
+      "eval_samples_per_second": 3.132,
+      "eval_steps_per_second": 0.214,
+      "step": 2440
+    },
+    {
+      "epoch": 0.3160177363498074,
+      "grad_norm": 3.160928109932279,
+      "learning_rate": 1.0806517979206183e-06,
+      "loss": 2.6792,
+      "step": 2441
+    },
+    {
+      "epoch": 0.3161471987571609,
+      "grad_norm": 3.9320513444446266,
+      "learning_rate": 1.0804584051319146e-06,
+      "loss": 2.8491,
+      "step": 2442
+    },
+    {
+      "epoch": 0.31627666116451436,
+      "grad_norm": 3.0947554246258244,
+      "learning_rate": 1.0802649193066003e-06,
+      "loss": 2.7412,
+      "step": 2443
+    },
+    {
+      "epoch": 0.3164061235718678,
+      "grad_norm": 3.229832254575071,
+      "learning_rate": 1.0800713404841982e-06,
+      "loss": 2.5986,
+      "step": 2444
+    },
+    {
+      "epoch": 0.31653558597922127,
+      "grad_norm": 3.2882254814348744,
+      "learning_rate": 1.079877668704251e-06,
+      "loss": 2.7722,
+      "step": 2445
+    },
+    {
+      "epoch": 0.31653558597922127,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7264736890792847,
+      "eval_runtime": 14.1039,
+      "eval_samples_per_second": 3.12,
+      "eval_steps_per_second": 0.213,
+      "step": 2445
+    },
+    {
+      "epoch": 0.3166650483865747,
+      "grad_norm": 8.512871943307267,
+      "learning_rate": 1.07968390400632e-06,
+      "loss": 2.9233,
+      "step": 2446
+    },
+    {
+      "epoch": 0.31679451079392823,
+      "grad_norm": 1.8824532694724163,
+      "learning_rate": 1.0794900464299856e-06,
+      "loss": 2.5317,
+      "step": 2447
+    },
+    {
+      "epoch": 0.3169239732012817,
+      "grad_norm": 2.31123638509395,
+      "learning_rate": 1.0792960960148472e-06,
+      "loss": 2.7749,
+      "step": 2448
+    },
+    {
+      "epoch": 0.31705343560863514,
+      "grad_norm": 1.9713933039519702,
+      "learning_rate": 1.0791020528005228e-06,
+      "loss": 2.7087,
+      "step": 2449
+    },
+    {
+      "epoch": 0.3171828980159886,
+      "grad_norm": 6.951210945555584,
+      "learning_rate": 1.07890791682665e-06,
+      "loss": 2.9065,
+      "step": 2450
+    },
+    {
+      "epoch": 0.3171828980159886,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.725053310394287,
+      "eval_runtime": 14.3067,
+      "eval_samples_per_second": 3.075,
+      "eval_steps_per_second": 0.21,
+      "step": 2450
+    },
+    {
+      "epoch": 0.3173123604233421,
+      "grad_norm": 1.5610967085634204,
+      "learning_rate": 1.078713688132885e-06,
+      "loss": 2.6248,
+      "step": 2451
+    },
+    {
+      "epoch": 0.31744182283069555,
+      "grad_norm": 2.6455580630439464,
+      "learning_rate": 1.0785193667589029e-06,
+      "loss": 2.7341,
+      "step": 2452
+    },
+    {
+      "epoch": 0.317571285238049,
+      "grad_norm": 1.6280969815206268,
+      "learning_rate": 1.0783249527443973e-06,
+      "loss": 2.5791,
+      "step": 2453
+    },
+    {
+      "epoch": 0.31770074764540246,
+      "grad_norm": 4.436350882045652,
+      "learning_rate": 1.0781304461290821e-06,
+      "loss": 2.603,
+      "step": 2454
+    },
+    {
+      "epoch": 0.3178302100527559,
+      "grad_norm": 3.1764506320770693,
+      "learning_rate": 1.0779358469526886e-06,
+      "loss": 2.8035,
+      "step": 2455
+    },
+    {
+      "epoch": 0.3178302100527559,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.722367763519287,
+      "eval_runtime": 13.7875,
+      "eval_samples_per_second": 3.191,
+      "eval_steps_per_second": 0.218,
+      "step": 2455
+    },
+    {
+      "epoch": 0.3179596724601094,
+      "grad_norm": 2.553147022982123,
+      "learning_rate": 1.0777411552549675e-06,
+      "loss": 2.7024,
+      "step": 2456
+    },
+    {
+      "epoch": 0.3180891348674629,
+      "grad_norm": 4.184804910513455,
+      "learning_rate": 1.077546371075689e-06,
+      "loss": 3.0112,
+      "step": 2457
+    },
+    {
+      "epoch": 0.31821859727481633,
+      "grad_norm": 2.8942752790324002,
+      "learning_rate": 1.0773514944546414e-06,
+      "loss": 3.0034,
+      "step": 2458
+    },
+    {
+      "epoch": 0.3183480596821698,
+      "grad_norm": 2.7530188522035974,
+      "learning_rate": 1.0771565254316327e-06,
+      "loss": 2.8999,
+      "step": 2459
+    },
+    {
+      "epoch": 0.31847752208952324,
+      "grad_norm": 2.6219202515606526,
+      "learning_rate": 1.0769614640464883e-06,
+      "loss": 2.8696,
+      "step": 2460
+    },
+    {
+      "epoch": 0.31847752208952324,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7179287672042847,
+      "eval_runtime": 14.4399,
+      "eval_samples_per_second": 3.047,
+      "eval_steps_per_second": 0.208,
+      "step": 2460
+    },
+    {
+      "epoch": 0.31860698449687674,
+      "grad_norm": 2.84148124067926,
+      "learning_rate": 1.0767663103390546e-06,
+      "loss": 2.7959,
+      "step": 2461
+    },
+    {
+      "epoch": 0.3187364469042302,
+      "grad_norm": 2.756660157319629,
+      "learning_rate": 1.0765710643491947e-06,
+      "loss": 3.083,
+      "step": 2462
+    },
+    {
+      "epoch": 0.31886590931158365,
+      "grad_norm": 4.201171957519662,
+      "learning_rate": 1.0763757261167923e-06,
+      "loss": 2.6294,
+      "step": 2463
+    },
+    {
+      "epoch": 0.3189953717189371,
+      "grad_norm": 4.321268425030079,
+      "learning_rate": 1.0761802956817486e-06,
+      "loss": 2.6978,
+      "step": 2464
+    },
+    {
+      "epoch": 0.31912483412629056,
+      "grad_norm": 1.7481726935604223,
+      "learning_rate": 1.0759847730839847e-06,
+      "loss": 2.865,
+      "step": 2465
+    },
+    {
+      "epoch": 0.31912483412629056,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7188388109207153,
+      "eval_runtime": 16.101,
+      "eval_samples_per_second": 2.733,
+      "eval_steps_per_second": 0.186,
+      "step": 2465
+    },
+    {
+      "epoch": 0.31925429653364407,
+      "grad_norm": 2.9924227571415556,
+      "learning_rate": 1.0757891583634398e-06,
+      "loss": 2.6526,
+      "step": 2466
+    },
+    {
+      "epoch": 0.3193837589409975,
+      "grad_norm": 2.5318284906218587,
+      "learning_rate": 1.0755934515600721e-06,
+      "loss": 2.667,
+      "step": 2467
+    },
+    {
+      "epoch": 0.319513221348351,
+      "grad_norm": 3.454384497407765,
+      "learning_rate": 1.075397652713859e-06,
+      "loss": 3.1702,
+      "step": 2468
+    },
+    {
+      "epoch": 0.3196426837557044,
+      "grad_norm": 3.5162367439322177,
+      "learning_rate": 1.0752017618647959e-06,
+      "loss": 2.6421,
+      "step": 2469
+    },
+    {
+      "epoch": 0.3197721461630579,
+      "grad_norm": 3.327768218711653,
+      "learning_rate": 1.075005779052898e-06,
+      "loss": 2.9227,
+      "step": 2470
+    },
+    {
+      "epoch": 0.3197721461630579,
+      "eval_PRM Accuracy": 0.22641509433962265,
+      "eval_PRM F1": 0.023809523809523808,
+      "eval_PRM F1 AUC": 0.5060240963855421,
+      "eval_PRM F1 Neg": 0.359375,
+      "eval_PRM NPV": 0.21904761904761905,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.012048192771084338,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7175514698028564,
+      "eval_runtime": 15.8133,
+      "eval_samples_per_second": 2.782,
+      "eval_steps_per_second": 0.19,
+      "step": 2470
+    },
+    {
+      "epoch": 0.3199016085704114,
+      "grad_norm": 2.0077234479844415,
+      "learning_rate": 1.0748097043181984e-06,
+      "loss": 2.7979,
+      "step": 2471
+    },
+    {
+      "epoch": 0.32003107097776484,
+      "grad_norm": 2.907360607367345,
+      "learning_rate": 1.0746135377007496e-06,
+      "loss": 2.728,
+      "step": 2472
+    },
+    {
+      "epoch": 0.3201605333851183,
+      "grad_norm": 6.426058586786103,
+      "learning_rate": 1.0744172792406225e-06,
+      "loss": 2.6758,
+      "step": 2473
+    },
+    {
+      "epoch": 0.32028999579247175,
+      "grad_norm": 1.2931018024081558,
+      "learning_rate": 1.0742209289779069e-06,
+      "loss": 2.449,
+      "step": 2474
+    },
+    {
+      "epoch": 0.3204194581998252,
+      "grad_norm": 3.1610311723919224,
+      "learning_rate": 1.074024486952711e-06,
+      "loss": 2.7855,
+      "step": 2475
+    },
+    {
+      "epoch": 0.3204194581998252,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.72021484375,
+      "eval_runtime": 15.1815,
+      "eval_samples_per_second": 2.898,
+      "eval_steps_per_second": 0.198,
+      "step": 2475
+    },
+    {
+      "epoch": 0.3205489206071787,
+      "grad_norm": 2.648317789620508,
+      "learning_rate": 1.0738279532051625e-06,
+      "loss": 2.8459,
+      "step": 2476
+    },
+    {
+      "epoch": 0.32067838301453216,
+      "grad_norm": 1.5009025415287334,
+      "learning_rate": 1.073631327775407e-06,
+      "loss": 2.439,
+      "step": 2477
+    },
+    {
+      "epoch": 0.3208078454218856,
+      "grad_norm": 1.6609496433491577,
+      "learning_rate": 1.0734346107036097e-06,
+      "loss": 2.4409,
+      "step": 2478
+    },
+    {
+      "epoch": 0.32093730782923907,
+      "grad_norm": 2.290237376431311,
+      "learning_rate": 1.073237802029954e-06,
+      "loss": 2.7837,
+      "step": 2479
+    },
+    {
+      "epoch": 0.3210667702365925,
+      "grad_norm": 3.1314784364313724,
+      "learning_rate": 1.0730409017946414e-06,
+      "loss": 2.6189,
+      "step": 2480
+    },
+    {
+      "epoch": 0.3210667702365925,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.716020107269287,
+      "eval_runtime": 14.5865,
+      "eval_samples_per_second": 3.016,
+      "eval_steps_per_second": 0.206,
+      "step": 2480
+    },
+    {
+      "epoch": 0.32119623264394603,
+      "grad_norm": 2.480958092689093,
+      "learning_rate": 1.0728439100378935e-06,
+      "loss": 2.6064,
+      "step": 2481
+    },
+    {
+      "epoch": 0.3213256950512995,
+      "grad_norm": 1.485956326766334,
+      "learning_rate": 1.0726468267999495e-06,
+      "loss": 2.5059,
+      "step": 2482
+    },
+    {
+      "epoch": 0.32145515745865294,
+      "grad_norm": 3.9416117523117613,
+      "learning_rate": 1.072449652121068e-06,
+      "loss": 3.1322,
+      "step": 2483
+    },
+    {
+      "epoch": 0.3215846198660064,
+      "grad_norm": 2.118533380022602,
+      "learning_rate": 1.0722523860415251e-06,
+      "loss": 2.7395,
+      "step": 2484
+    },
+    {
+      "epoch": 0.32171408227335985,
+      "grad_norm": 7.2211385103012224,
+      "learning_rate": 1.0720550286016174e-06,
+      "loss": 3.2327,
+      "step": 2485
+    },
+    {
+      "epoch": 0.32171408227335985,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.714555263519287,
+      "eval_runtime": 14.4227,
+      "eval_samples_per_second": 3.051,
+      "eval_steps_per_second": 0.208,
+      "step": 2485
+    },
+    {
+      "epoch": 0.32184354468071336,
+      "grad_norm": 2.9886870997262838,
+      "learning_rate": 1.0718575798416585e-06,
+      "loss": 2.5952,
+      "step": 2486
+    },
+    {
+      "epoch": 0.3219730070880668,
+      "grad_norm": 6.80834374440038,
+      "learning_rate": 1.0716600398019818e-06,
+      "loss": 2.417,
+      "step": 2487
+    },
+    {
+      "epoch": 0.32210246949542026,
+      "grad_norm": 6.216963909338738,
+      "learning_rate": 1.0714624085229383e-06,
+      "loss": 2.6121,
+      "step": 2488
+    },
+    {
+      "epoch": 0.3222319319027737,
+      "grad_norm": 3.567677258886948,
+      "learning_rate": 1.0712646860448985e-06,
+      "loss": 3.1226,
+      "step": 2489
+    },
+    {
+      "epoch": 0.32236139431012717,
+      "grad_norm": 4.148947984312593,
+      "learning_rate": 1.0710668724082516e-06,
+      "loss": 2.6794,
+      "step": 2490
+    },
+    {
+      "epoch": 0.32236139431012717,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7180397510528564,
+      "eval_runtime": 14.2145,
+      "eval_samples_per_second": 3.095,
+      "eval_steps_per_second": 0.211,
+      "step": 2490
+    },
+    {
+      "epoch": 0.3224908567174807,
+      "grad_norm": 2.409469581957139,
+      "learning_rate": 1.0708689676534044e-06,
+      "loss": 2.7512,
+      "step": 2491
+    },
+    {
+      "epoch": 0.32262031912483413,
+      "grad_norm": 1.989219831788488,
+      "learning_rate": 1.070670971820783e-06,
+      "loss": 2.6553,
+      "step": 2492
+    },
+    {
+      "epoch": 0.3227497815321876,
+      "grad_norm": 1.2342416306836181,
+      "learning_rate": 1.0704728849508324e-06,
+      "loss": 2.2351,
+      "step": 2493
+    },
+    {
+      "epoch": 0.32287924393954104,
+      "grad_norm": 2.6907820017073503,
+      "learning_rate": 1.070274707084016e-06,
+      "loss": 2.7109,
+      "step": 2494
+    },
+    {
+      "epoch": 0.32300870634689455,
+      "grad_norm": 3.650577051923038,
+      "learning_rate": 1.0700764382608148e-06,
+      "loss": 2.9163,
+      "step": 2495
+    },
+    {
+      "epoch": 0.32300870634689455,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7165749073028564,
+      "eval_runtime": 14.7022,
+      "eval_samples_per_second": 2.993,
+      "eval_steps_per_second": 0.204,
+      "step": 2495
+    },
+    {
+      "epoch": 0.323138168754248,
+      "grad_norm": 6.1785950075176235,
+      "learning_rate": 1.0698780785217302e-06,
+      "loss": 3.3462,
+      "step": 2496
+    },
+    {
+      "epoch": 0.32326763116160145,
+      "grad_norm": 2.5343607882108694,
+      "learning_rate": 1.0696796279072807e-06,
+      "loss": 2.4346,
+      "step": 2497
+    },
+    {
+      "epoch": 0.3233970935689549,
+      "grad_norm": 1.846750528713725,
+      "learning_rate": 1.0694810864580041e-06,
+      "loss": 2.501,
+      "step": 2498
+    },
+    {
+      "epoch": 0.32352655597630836,
+      "grad_norm": 2.0810508054385037,
+      "learning_rate": 1.0692824542144562e-06,
+      "loss": 2.564,
+      "step": 2499
+    },
+    {
+      "epoch": 0.32365601838366187,
+      "grad_norm": 2.5691675830070513,
+      "learning_rate": 1.0690837312172117e-06,
+      "loss": 2.5383,
+      "step": 2500
+    },
+    {
+      "epoch": 0.32365601838366187,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.7138671875,
+      "eval_runtime": 14.264,
+      "eval_samples_per_second": 3.085,
+      "eval_steps_per_second": 0.21,
+      "step": 2500
     }
   ],
   "logging_steps": 1,
@@ -20442,7 +25542,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 1679157807448064.0,
+  "total_flos": 2098934106324992.0,
   "train_batch_size": 4,
   "trial_name": null,
   "trial_params": null