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": 0.5151817625656052,
+  "epoch": 0.6439772032070065,
   "eval_steps": 5,
-  "global_step": 2000,
+  "global_step": 2500,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -20824,6 +20824,5206 @@
       "eval_samples_per_second": 2.475,
       "eval_steps_per_second": 0.169,
       "step": 2000
+    },
+    {
+      "epoch": 0.515439353446888,
+      "grad_norm": 0.1576943422442954,
+      "learning_rate": 5.6033701947971805e-05,
+      "loss": 0.3773,
+      "step": 2001
+    },
+    {
+      "epoch": 0.5156969443281708,
+      "grad_norm": 0.1874952534596437,
+      "learning_rate": 5.5989058310079254e-05,
+      "loss": 0.4354,
+      "step": 2002
+    },
+    {
+      "epoch": 0.5159545352094536,
+      "grad_norm": 0.13372880522852698,
+      "learning_rate": 5.594440982755146e-05,
+      "loss": 0.3232,
+      "step": 2003
+    },
+    {
+      "epoch": 0.5162121260907364,
+      "grad_norm": 0.1514057779131614,
+      "learning_rate": 5.589975653650516e-05,
+      "loss": 0.311,
+      "step": 2004
+    },
+    {
+      "epoch": 0.5164697169720192,
+      "grad_norm": 0.20948284439508877,
+      "learning_rate": 5.585509847306112e-05,
+      "loss": 0.4186,
+      "step": 2005
+    },
+    {
+      "epoch": 0.5164697169720192,
+      "eval_PRM Accuracy": 0.7830188679245284,
+      "eval_PRM F1": 0.8606060606060606,
+      "eval_PRM F1 AUC": 0.6885804085908852,
+      "eval_PRM F1 AUC (fixed)": 0.8331587218438974,
+      "eval_PRM F1 Neg": 0.5106382978723404,
+      "eval_PRM NPV": 0.5,
+      "eval_PRM Precision": 0.8658536585365854,
+      "eval_PRM Recall": 0.8554216867469879,
+      "eval_PRM Specificty": 0.5217391304347826,
+      "eval_loss": 0.4187677502632141,
+      "eval_runtime": 17.0341,
+      "eval_samples_per_second": 2.583,
+      "eval_steps_per_second": 0.176,
+      "step": 2005
+    },
+    {
+      "epoch": 0.516727307853302,
+      "grad_norm": 0.15092735579258332,
+      "learning_rate": 5.581043567334383e-05,
+      "loss": 0.394,
+      "step": 2006
+    },
+    {
+      "epoch": 0.5169848987345848,
+      "grad_norm": 0.14426891302448921,
+      "learning_rate": 5.576576817348169e-05,
+      "loss": 0.3616,
+      "step": 2007
+    },
+    {
+      "epoch": 0.5172424896158676,
+      "grad_norm": 0.19486454565642172,
+      "learning_rate": 5.5721096009606885e-05,
+      "loss": 0.3272,
+      "step": 2008
+    },
+    {
+      "epoch": 0.5175000804971503,
+      "grad_norm": 0.19109979412742129,
+      "learning_rate": 5.567641921785537e-05,
+      "loss": 0.4214,
+      "step": 2009
+    },
+    {
+      "epoch": 0.5177576713784332,
+      "grad_norm": 0.15701301563469852,
+      "learning_rate": 5.5631737834366836e-05,
+      "loss": 0.3289,
+      "step": 2010
+    },
+    {
+      "epoch": 0.5177576713784332,
+      "eval_PRM Accuracy": 0.7830188679245284,
+      "eval_PRM F1": 0.8606060606060606,
+      "eval_PRM F1 AUC": 0.6885804085908852,
+      "eval_PRM F1 AUC (fixed)": 0.8389209009952854,
+      "eval_PRM F1 Neg": 0.5106382978723404,
+      "eval_PRM NPV": 0.5,
+      "eval_PRM Precision": 0.8658536585365854,
+      "eval_PRM Recall": 0.8554216867469879,
+      "eval_PRM Specificty": 0.5217391304347826,
+      "eval_loss": 0.4164595305919647,
+      "eval_runtime": 17.4099,
+      "eval_samples_per_second": 2.527,
+      "eval_steps_per_second": 0.172,
+      "step": 2010
+    },
+    {
+      "epoch": 0.5180152622597161,
+      "grad_norm": 0.12605676343069108,
+      "learning_rate": 5.5587051895284695e-05,
+      "loss": 0.3397,
+      "step": 2011
+    },
+    {
+      "epoch": 0.5182728531409988,
+      "grad_norm": 0.1824585228760495,
+      "learning_rate": 5.554236143675605e-05,
+      "loss": 0.3743,
+      "step": 2012
+    },
+    {
+      "epoch": 0.5185304440222817,
+      "grad_norm": 0.20069764307974666,
+      "learning_rate": 5.5497666494931654e-05,
+      "loss": 0.4296,
+      "step": 2013
+    },
+    {
+      "epoch": 0.5187880349035644,
+      "grad_norm": 0.1403249954553184,
+      "learning_rate": 5.5452967105965904e-05,
+      "loss": 0.3772,
+      "step": 2014
+    },
+    {
+      "epoch": 0.5190456257848473,
+      "grad_norm": 0.15452067087473162,
+      "learning_rate": 5.5408263306016736e-05,
+      "loss": 0.4272,
+      "step": 2015
+    },
+    {
+      "epoch": 0.5190456257848473,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.874251497005988,
+      "eval_PRM F1 AUC": 0.7006286013619696,
+      "eval_PRM F1 AUC (fixed)": 0.8470403352540596,
+      "eval_PRM F1 Neg": 0.5333333333333333,
+      "eval_PRM NPV": 0.5454545454545454,
+      "eval_PRM Precision": 0.8690476190476191,
+      "eval_PRM Recall": 0.8795180722891566,
+      "eval_PRM Specificty": 0.5217391304347826,
+      "eval_loss": 0.4115767180919647,
+      "eval_runtime": 17.1613,
+      "eval_samples_per_second": 2.564,
+      "eval_steps_per_second": 0.175,
+      "step": 2015
+    },
+    {
+      "epoch": 0.51930321666613,
+      "grad_norm": 0.17455468940712443,
+      "learning_rate": 5.536355513124574e-05,
+      "loss": 0.3338,
+      "step": 2016
+    },
+    {
+      "epoch": 0.5195608075474129,
+      "grad_norm": 0.2010758514492215,
+      "learning_rate": 5.531884261781797e-05,
+      "loss": 0.4625,
+      "step": 2017
+    },
+    {
+      "epoch": 0.5198183984286956,
+      "grad_norm": 0.16226773402413,
+      "learning_rate": 5.527412580190203e-05,
+      "loss": 0.3552,
+      "step": 2018
+    },
+    {
+      "epoch": 0.5200759893099784,
+      "grad_norm": 0.156114481596148,
+      "learning_rate": 5.5229404719669984e-05,
+      "loss": 0.3447,
+      "step": 2019
+    },
+    {
+      "epoch": 0.5203335801912612,
+      "grad_norm": 0.182267158691678,
+      "learning_rate": 5.518467940729739e-05,
+      "loss": 0.443,
+      "step": 2020
+    },
+    {
+      "epoch": 0.5203335801912612,
+      "eval_PRM Accuracy": 0.8207547169811321,
+      "eval_PRM F1": 0.8862275449101796,
+      "eval_PRM F1 AUC": 0.7283918281822945,
+      "eval_PRM F1 AUC (fixed)": 0.853850183342064,
+      "eval_PRM F1 Neg": 0.5777777777777777,
+      "eval_PRM NPV": 0.5909090909090909,
+      "eval_PRM Precision": 0.8809523809523809,
+      "eval_PRM Recall": 0.891566265060241,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.4057173430919647,
+      "eval_runtime": 17.0622,
+      "eval_samples_per_second": 2.579,
+      "eval_steps_per_second": 0.176,
+      "step": 2020
+    },
+    {
+      "epoch": 0.520591171072544,
+      "grad_norm": 0.1405859308577178,
+      "learning_rate": 5.5139949900963174e-05,
+      "loss": 0.3475,
+      "step": 2021
+    },
+    {
+      "epoch": 0.5208487619538268,
+      "grad_norm": 0.12624673647873447,
+      "learning_rate": 5.509521623684968e-05,
+      "loss": 0.3363,
+      "step": 2022
+    },
+    {
+      "epoch": 0.5211063528351096,
+      "grad_norm": 0.11565588606367397,
+      "learning_rate": 5.5050478451142586e-05,
+      "loss": 0.3585,
+      "step": 2023
+    },
+    {
+      "epoch": 0.5213639437163924,
+      "grad_norm": 0.21176890439543344,
+      "learning_rate": 5.5005736580030963e-05,
+      "loss": 0.3405,
+      "step": 2024
+    },
+    {
+      "epoch": 0.5216215345976752,
+      "grad_norm": 0.1317224767338869,
+      "learning_rate": 5.4960990659707136e-05,
+      "loss": 0.3619,
+      "step": 2025
+    },
+    {
+      "epoch": 0.5216215345976752,
+      "eval_PRM Accuracy": 0.8207547169811321,
+      "eval_PRM F1": 0.8862275449101796,
+      "eval_PRM F1 AUC": 0.7283918281822945,
+      "eval_PRM F1 AUC (fixed)": 0.8501833420639078,
+      "eval_PRM F1 Neg": 0.5777777777777777,
+      "eval_PRM NPV": 0.5909090909090909,
+      "eval_PRM Precision": 0.8809523809523809,
+      "eval_PRM Recall": 0.891566265060241,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.4074041247367859,
+      "eval_runtime": 17.0383,
+      "eval_samples_per_second": 2.582,
+      "eval_steps_per_second": 0.176,
+      "step": 2025
+    },
+    {
+      "epoch": 0.5218791254789581,
+      "grad_norm": 0.11847577493241711,
+      "learning_rate": 5.491624072636674e-05,
+      "loss": 0.297,
+      "step": 2026
+    },
+    {
+      "epoch": 0.5221367163602408,
+      "grad_norm": 0.17280346995987633,
+      "learning_rate": 5.487148681620862e-05,
+      "loss": 0.4082,
+      "step": 2027
+    },
+    {
+      "epoch": 0.5223943072415237,
+      "grad_norm": 0.17084021311446107,
+      "learning_rate": 5.482672896543487e-05,
+      "loss": 0.4101,
+      "step": 2028
+    },
+    {
+      "epoch": 0.5226518981228064,
+      "grad_norm": 0.13621840520116943,
+      "learning_rate": 5.478196721025073e-05,
+      "loss": 0.3993,
+      "step": 2029
+    },
+    {
+      "epoch": 0.5229094890040893,
+      "grad_norm": 0.13117663910236727,
+      "learning_rate": 5.4737201586864675e-05,
+      "loss": 0.3307,
+      "step": 2030
+    },
+    {
+      "epoch": 0.5229094890040893,
+      "eval_PRM Accuracy": 0.7924528301886793,
+      "eval_PRM F1": 0.8658536585365854,
+      "eval_PRM F1 AUC": 0.710319539025668,
+      "eval_PRM F1 AUC (fixed)": 0.8407543216343635,
+      "eval_PRM F1 Neg": 0.5416666666666666,
+      "eval_PRM NPV": 0.52,
+      "eval_PRM Precision": 0.8765432098765432,
+      "eval_PRM Recall": 0.8554216867469879,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.39719459414482117,
+      "eval_runtime": 16.8682,
+      "eval_samples_per_second": 2.608,
+      "eval_steps_per_second": 0.178,
+      "step": 2030
+    },
+    {
+      "epoch": 0.523167079885372,
+      "grad_norm": 0.1383710235225394,
+      "learning_rate": 5.469243213148822e-05,
+      "loss": 0.3639,
+      "step": 2031
+    },
+    {
+      "epoch": 0.5234246707666549,
+      "grad_norm": 0.15304363860482864,
+      "learning_rate": 5.464765888033602e-05,
+      "loss": 0.3882,
+      "step": 2032
+    },
+    {
+      "epoch": 0.5236822616479376,
+      "grad_norm": 0.15212229570341806,
+      "learning_rate": 5.460288186962583e-05,
+      "loss": 0.3878,
+      "step": 2033
+    },
+    {
+      "epoch": 0.5239398525292205,
+      "grad_norm": 0.13839664274067834,
+      "learning_rate": 5.455810113557839e-05,
+      "loss": 0.3708,
+      "step": 2034
+    },
+    {
+      "epoch": 0.5241974434105032,
+      "grad_norm": 0.17316272691860943,
+      "learning_rate": 5.451331671441751e-05,
+      "loss": 0.4628,
+      "step": 2035
+    },
+    {
+      "epoch": 0.5241974434105032,
+      "eval_PRM Accuracy": 0.7830188679245284,
+      "eval_PRM F1": 0.8588957055214724,
+      "eval_PRM F1 AUC": 0.7042954426401258,
+      "eval_PRM F1 AUC (fixed)": 0.8423258250392878,
+      "eval_PRM F1 Neg": 0.5306122448979592,
+      "eval_PRM NPV": 0.5,
+      "eval_PRM Precision": 0.875,
+      "eval_PRM Recall": 0.8433734939759037,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.39151278138160706,
+      "eval_runtime": 16.9123,
+      "eval_samples_per_second": 2.602,
+      "eval_steps_per_second": 0.177,
+      "step": 2035
+    },
+    {
+      "epoch": 0.5244550342917861,
+      "grad_norm": 0.14489451523876312,
+      "learning_rate": 5.446852864236991e-05,
+      "loss": 0.3221,
+      "step": 2036
+    },
+    {
+      "epoch": 0.5247126251730688,
+      "grad_norm": 0.1988766080619215,
+      "learning_rate": 5.442373695566534e-05,
+      "loss": 0.4659,
+      "step": 2037
+    },
+    {
+      "epoch": 0.5249702160543517,
+      "grad_norm": 0.1389776176471604,
+      "learning_rate": 5.437894169053641e-05,
+      "loss": 0.3277,
+      "step": 2038
+    },
+    {
+      "epoch": 0.5252278069356345,
+      "grad_norm": 0.17073375948781336,
+      "learning_rate": 5.4334142883218676e-05,
+      "loss": 0.4109,
+      "step": 2039
+    },
+    {
+      "epoch": 0.5254853978169173,
+      "grad_norm": 0.15091599719834037,
+      "learning_rate": 5.428934056995052e-05,
+      "loss": 0.3377,
+      "step": 2040
+    },
+    {
+      "epoch": 0.5254853978169173,
+      "eval_PRM Accuracy": 0.7924528301886793,
+      "eval_PRM F1": 0.8658536585365854,
+      "eval_PRM F1 AUC": 0.710319539025668,
+      "eval_PRM F1 AUC (fixed)": 0.8386589837611316,
+      "eval_PRM F1 Neg": 0.5416666666666666,
+      "eval_PRM NPV": 0.52,
+      "eval_PRM Precision": 0.8765432098765432,
+      "eval_PRM Recall": 0.8554216867469879,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.4031427502632141,
+      "eval_runtime": 17.5676,
+      "eval_samples_per_second": 2.505,
+      "eval_steps_per_second": 0.171,
+      "step": 2040
+    },
+    {
+      "epoch": 0.5257429886982001,
+      "grad_norm": 0.1726553264575567,
+      "learning_rate": 5.4244534786973214e-05,
+      "loss": 0.3972,
+      "step": 2041
+    },
+    {
+      "epoch": 0.5260005795794829,
+      "grad_norm": 0.12642039827318133,
+      "learning_rate": 5.4199725570530735e-05,
+      "loss": 0.3027,
+      "step": 2042
+    },
+    {
+      "epoch": 0.5262581704607657,
+      "grad_norm": 0.15104989331276478,
+      "learning_rate": 5.415491295686995e-05,
+      "loss": 0.4181,
+      "step": 2043
+    },
+    {
+      "epoch": 0.5265157613420485,
+      "grad_norm": 0.23276158945096306,
+      "learning_rate": 5.411009698224041e-05,
+      "loss": 0.4327,
+      "step": 2044
+    },
+    {
+      "epoch": 0.5267733522233313,
+      "grad_norm": 0.14805946848268564,
+      "learning_rate": 5.406527768289441e-05,
+      "loss": 0.3166,
+      "step": 2045
+    },
+    {
+      "epoch": 0.5267733522233313,
+      "eval_PRM Accuracy": 0.7924528301886793,
+      "eval_PRM F1": 0.8690476190476191,
+      "eval_PRM F1 AUC": 0.678889470927187,
+      "eval_PRM F1 AUC (fixed)": 0.8418019905709795,
+      "eval_PRM F1 Neg": 0.5,
+      "eval_PRM NPV": 0.5238095238095238,
+      "eval_PRM Precision": 0.8588235294117647,
+      "eval_PRM Recall": 0.8795180722891566,
+      "eval_PRM Specificty": 0.4782608695652174,
+      "eval_loss": 0.4216974377632141,
+      "eval_runtime": 17.5049,
+      "eval_samples_per_second": 2.514,
+      "eval_steps_per_second": 0.171,
+      "step": 2045
+    },
+    {
+      "epoch": 0.5270309431046141,
+      "grad_norm": 0.13143827483675186,
+      "learning_rate": 5.402045509508692e-05,
+      "loss": 0.3572,
+      "step": 2046
+    },
+    {
+      "epoch": 0.5272885339858969,
+      "grad_norm": 0.1575077674791195,
+      "learning_rate": 5.3975629255075554e-05,
+      "loss": 0.2465,
+      "step": 2047
+    },
+    {
+      "epoch": 0.5275461248671797,
+      "grad_norm": 0.2740141779719599,
+      "learning_rate": 5.3930800199120616e-05,
+      "loss": 0.4939,
+      "step": 2048
+    },
+    {
+      "epoch": 0.5278037157484625,
+      "grad_norm": 0.18515637996824869,
+      "learning_rate": 5.388596796348494e-05,
+      "loss": 0.4005,
+      "step": 2049
+    },
+    {
+      "epoch": 0.5280613066297453,
+      "grad_norm": 0.17690785782001883,
+      "learning_rate": 5.3841132584433986e-05,
+      "loss": 0.3552,
+      "step": 2050
+    },
+    {
+      "epoch": 0.5280613066297453,
+      "eval_PRM Accuracy": 0.7830188679245284,
+      "eval_PRM F1": 0.8622754491017964,
+      "eval_PRM F1 AUC": 0.6728653745416449,
+      "eval_PRM F1 AUC (fixed)": 0.8404924044002096,
+      "eval_PRM F1 Neg": 0.4888888888888889,
+      "eval_PRM NPV": 0.5,
+      "eval_PRM Precision": 0.8571428571428571,
+      "eval_PRM Recall": 0.8674698795180723,
+      "eval_PRM Specificty": 0.4782608695652174,
+      "eval_loss": 0.4098011255264282,
+      "eval_runtime": 17.0428,
+      "eval_samples_per_second": 2.582,
+      "eval_steps_per_second": 0.176,
+      "step": 2050
+    },
+    {
+      "epoch": 0.5283188975110281,
+      "grad_norm": 0.14716600331952415,
+      "learning_rate": 5.379629409823571e-05,
+      "loss": 0.3991,
+      "step": 2051
+    },
+    {
+      "epoch": 0.5285764883923109,
+      "grad_norm": 0.1572104453485029,
+      "learning_rate": 5.375145254116065e-05,
+      "loss": 0.3503,
+      "step": 2052
+    },
+    {
+      "epoch": 0.5288340792735937,
+      "grad_norm": 0.14391001387350164,
+      "learning_rate": 5.3706607949481736e-05,
+      "loss": 0.3557,
+      "step": 2053
+    },
+    {
+      "epoch": 0.5290916701548766,
+      "grad_norm": 0.19161756424057008,
+      "learning_rate": 5.3661760359474446e-05,
+      "loss": 0.3301,
+      "step": 2054
+    },
+    {
+      "epoch": 0.5293492610361593,
+      "grad_norm": 0.2085656851277453,
+      "learning_rate": 5.361690980741663e-05,
+      "loss": 0.3831,
+      "step": 2055
+    },
+    {
+      "epoch": 0.5293492610361593,
+      "eval_PRM Accuracy": 0.7735849056603774,
+      "eval_PRM F1": 0.85,
+      "eval_PRM F1 AUC": 0.7139863803038241,
+      "eval_PRM F1 AUC (fixed)": 0.8412781561026716,
+      "eval_PRM F1 Neg": 0.5384615384615384,
+      "eval_PRM NPV": 0.4827586206896552,
+      "eval_PRM Precision": 0.8831168831168831,
+      "eval_PRM Recall": 0.8192771084337349,
+      "eval_PRM Specificty": 0.6086956521739131,
+      "eval_loss": 0.4046519994735718,
+      "eval_runtime": 17.5267,
+      "eval_samples_per_second": 2.51,
+      "eval_steps_per_second": 0.171,
+      "step": 2055
+    },
+    {
+      "epoch": 0.5296068519174422,
+      "grad_norm": 0.15938303163199383,
+      "learning_rate": 5.357205632958852e-05,
+      "loss": 0.3305,
+      "step": 2056
+    },
+    {
+      "epoch": 0.5298644427987249,
+      "grad_norm": 0.1691346447226298,
+      "learning_rate": 5.352719996227273e-05,
+      "loss": 0.3886,
+      "step": 2057
+    },
+    {
+      "epoch": 0.5301220336800078,
+      "grad_norm": 0.16082307158085632,
+      "learning_rate": 5.348234074175428e-05,
+      "loss": 0.3472,
+      "step": 2058
+    },
+    {
+      "epoch": 0.5303796245612905,
+      "grad_norm": 0.18415935209890774,
+      "learning_rate": 5.3437478704320375e-05,
+      "loss": 0.3332,
+      "step": 2059
+    },
+    {
+      "epoch": 0.5306372154425734,
+      "grad_norm": 0.16943370945706787,
+      "learning_rate": 5.33926138862606e-05,
+      "loss": 0.4161,
+      "step": 2060
+    },
+    {
+      "epoch": 0.5306372154425734,
+      "eval_PRM Accuracy": 0.7735849056603774,
+      "eval_PRM F1": 0.8571428571428571,
+      "eval_PRM F1 AUC": 0.6511262441068623,
+      "eval_PRM F1 AUC (fixed)": 0.8501833420639078,
+      "eval_PRM F1 Neg": 0.45454545454545453,
+      "eval_PRM NPV": 0.47619047619047616,
+      "eval_PRM Precision": 0.8470588235294118,
+      "eval_PRM Recall": 0.8674698795180723,
+      "eval_PRM Specificty": 0.43478260869565216,
+      "eval_loss": 0.40518465638160706,
+      "eval_runtime": 17.0928,
+      "eval_samples_per_second": 2.574,
+      "eval_steps_per_second": 0.176,
+      "step": 2060
+    },
+    {
+      "epoch": 0.5308948063238561,
+      "grad_norm": 0.14028074670246027,
+      "learning_rate": 5.3347746323866724e-05,
+      "loss": 0.3035,
+      "step": 2061
+    },
+    {
+      "epoch": 0.531152397205139,
+      "grad_norm": 0.32292804580343487,
+      "learning_rate": 5.330287605343279e-05,
+      "loss": 0.4956,
+      "step": 2062
+    },
+    {
+      "epoch": 0.5314099880864217,
+      "grad_norm": 0.16444934025334107,
+      "learning_rate": 5.325800311125497e-05,
+      "loss": 0.3578,
+      "step": 2063
+    },
+    {
+      "epoch": 0.5316675789677046,
+      "grad_norm": 0.16473755647300412,
+      "learning_rate": 5.3213127533631666e-05,
+      "loss": 0.3578,
+      "step": 2064
+    },
+    {
+      "epoch": 0.5319251698489873,
+      "grad_norm": 0.15958162419322863,
+      "learning_rate": 5.3168249356863354e-05,
+      "loss": 0.4014,
+      "step": 2065
+    },
+    {
+      "epoch": 0.5319251698489873,
+      "eval_PRM Accuracy": 0.7735849056603774,
+      "eval_PRM F1": 0.8571428571428571,
+      "eval_PRM F1 AUC": 0.6511262441068623,
+      "eval_PRM F1 AUC (fixed)": 0.8543740178103719,
+      "eval_PRM F1 Neg": 0.45454545454545453,
+      "eval_PRM NPV": 0.47619047619047616,
+      "eval_PRM Precision": 0.8470588235294118,
+      "eval_PRM Recall": 0.8674698795180723,
+      "eval_PRM Specificty": 0.43478260869565216,
+      "eval_loss": 0.40163353085517883,
+      "eval_runtime": 17.0134,
+      "eval_samples_per_second": 2.586,
+      "eval_steps_per_second": 0.176,
+      "step": 2065
+    },
+    {
+      "epoch": 0.5321827607302702,
+      "grad_norm": 0.20458510984275777,
+      "learning_rate": 5.3123368617252654e-05,
+      "loss": 0.4312,
+      "step": 2066
+    },
+    {
+      "epoch": 0.5324403516115529,
+      "grad_norm": 0.17054775353840954,
+      "learning_rate": 5.3078485351104226e-05,
+      "loss": 0.4225,
+      "step": 2067
+    },
+    {
+      "epoch": 0.5326979424928358,
+      "grad_norm": 0.16286686594336625,
+      "learning_rate": 5.3033599594724794e-05,
+      "loss": 0.4337,
+      "step": 2068
+    },
+    {
+      "epoch": 0.5329555333741186,
+      "grad_norm": 0.13733316270492532,
+      "learning_rate": 5.298871138442307e-05,
+      "loss": 0.3359,
+      "step": 2069
+    },
+    {
+      "epoch": 0.5332131242554013,
+      "grad_norm": 0.17278877168257112,
+      "learning_rate": 5.294382075650981e-05,
+      "loss": 0.4483,
+      "step": 2070
+    },
+    {
+      "epoch": 0.5332131242554013,
+      "eval_PRM Accuracy": 0.7924528301886793,
+      "eval_PRM F1": 0.8658536585365854,
+      "eval_PRM F1 AUC": 0.710319539025668,
+      "eval_PRM F1 AUC (fixed)": 0.8462545835515977,
+      "eval_PRM F1 Neg": 0.5416666666666666,
+      "eval_PRM NPV": 0.52,
+      "eval_PRM Precision": 0.8765432098765432,
+      "eval_PRM Recall": 0.8554216867469879,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.4013671875,
+      "eval_runtime": 17.0377,
+      "eval_samples_per_second": 2.583,
+      "eval_steps_per_second": 0.176,
+      "step": 2070
+    },
+    {
+      "epoch": 0.5334707151366842,
+      "grad_norm": 0.1303627564951731,
+      "learning_rate": 5.289892774729767e-05,
+      "loss": 0.3618,
+      "step": 2071
+    },
+    {
+      "epoch": 0.533728306017967,
+      "grad_norm": 0.15467466904494062,
+      "learning_rate": 5.285403239310123e-05,
+      "loss": 0.3879,
+      "step": 2072
+    },
+    {
+      "epoch": 0.5339858968992498,
+      "grad_norm": 0.17260775514177185,
+      "learning_rate": 5.280913473023702e-05,
+      "loss": 0.4635,
+      "step": 2073
+    },
+    {
+      "epoch": 0.5342434877805325,
+      "grad_norm": 0.15840090883749958,
+      "learning_rate": 5.276423479502338e-05,
+      "loss": 0.455,
+      "step": 2074
+    },
+    {
+      "epoch": 0.5345010786618154,
+      "grad_norm": 0.14397270680650293,
+      "learning_rate": 5.271933262378054e-05,
+      "loss": 0.3334,
+      "step": 2075
+    },
+    {
+      "epoch": 0.5345010786618154,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.8727272727272727,
+      "eval_PRM F1 AUC": 0.7163436354112102,
+      "eval_PRM F1 AUC (fixed)": 0.8386589837611315,
+      "eval_PRM F1 Neg": 0.5531914893617021,
+      "eval_PRM NPV": 0.5416666666666666,
+      "eval_PRM Precision": 0.8780487804878049,
+      "eval_PRM Recall": 0.8674698795180723,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.4106889069080353,
+      "eval_runtime": 17.5082,
+      "eval_samples_per_second": 2.513,
+      "eval_steps_per_second": 0.171,
+      "step": 2075
+    },
+    {
+      "epoch": 0.5347586695430981,
+      "grad_norm": 0.12572985271208612,
+      "learning_rate": 5.267442825283048e-05,
+      "loss": 0.3947,
+      "step": 2076
+    },
+    {
+      "epoch": 0.535016260424381,
+      "grad_norm": 0.15807576156421438,
+      "learning_rate": 5.2629521718497e-05,
+      "loss": 0.3851,
+      "step": 2077
+    },
+    {
+      "epoch": 0.5352738513056637,
+      "grad_norm": 0.13998319009380997,
+      "learning_rate": 5.258461305710563e-05,
+      "loss": 0.3512,
+      "step": 2078
+    },
+    {
+      "epoch": 0.5355314421869466,
+      "grad_norm": 0.12643201934729273,
+      "learning_rate": 5.253970230498367e-05,
+      "loss": 0.3327,
+      "step": 2079
+    },
+    {
+      "epoch": 0.5357890330682293,
+      "grad_norm": 0.13128534875393535,
+      "learning_rate": 5.249478949846004e-05,
+      "loss": 0.4037,
+      "step": 2080
+    },
+    {
+      "epoch": 0.5357890330682293,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.8771929824561403,
+      "eval_PRM F1 AUC": 0.6691985332634888,
+      "eval_PRM F1 AUC (fixed)": 0.8420639078051335,
+      "eval_PRM F1 Neg": 0.4878048780487805,
+      "eval_PRM NPV": 0.5555555555555556,
+      "eval_PRM Precision": 0.8522727272727273,
+      "eval_PRM Recall": 0.9036144578313253,
+      "eval_PRM Specificty": 0.43478260869565216,
+      "eval_loss": 0.42604759335517883,
+      "eval_runtime": 17.7233,
+      "eval_samples_per_second": 2.483,
+      "eval_steps_per_second": 0.169,
+      "step": 2080
+    },
+    {
+      "epoch": 0.5360466239495122,
+      "grad_norm": 0.16721180525336535,
+      "learning_rate": 5.2449874673865365e-05,
+      "loss": 0.3918,
+      "step": 2081
+    },
+    {
+      "epoch": 0.536304214830795,
+      "grad_norm": 0.12368253443734306,
+      "learning_rate": 5.2404957867531876e-05,
+      "loss": 0.3693,
+      "step": 2082
+    },
+    {
+      "epoch": 0.5365618057120778,
+      "grad_norm": 0.14511735553100816,
+      "learning_rate": 5.236003911579345e-05,
+      "loss": 0.313,
+      "step": 2083
+    },
+    {
+      "epoch": 0.5368193965933606,
+      "grad_norm": 0.14055192834613478,
+      "learning_rate": 5.231511845498548e-05,
+      "loss": 0.349,
+      "step": 2084
+    },
+    {
+      "epoch": 0.5370769874746434,
+      "grad_norm": 0.11673532612131145,
+      "learning_rate": 5.2270195921444954e-05,
+      "loss": 0.3176,
+      "step": 2085
+    },
+    {
+      "epoch": 0.5370769874746434,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.8786127167630058,
+      "eval_PRM F1 AUC": 0.6534834992142482,
+      "eval_PRM F1 AUC (fixed)": 0.8520167627029858,
+      "eval_PRM F1 Neg": 0.46153846153846156,
+      "eval_PRM NPV": 0.5625,
+      "eval_PRM Precision": 0.8444444444444444,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.391304347826087,
+      "eval_loss": 0.4271129369735718,
+      "eval_runtime": 16.8752,
+      "eval_samples_per_second": 2.607,
+      "eval_steps_per_second": 0.178,
+      "step": 2085
+    },
+    {
+      "epoch": 0.5373345783559262,
+      "grad_norm": 0.12488802728660602,
+      "learning_rate": 5.2225271551510345e-05,
+      "loss": 0.3735,
+      "step": 2086
+    },
+    {
+      "epoch": 0.537592169237209,
+      "grad_norm": 0.1577739570133298,
+      "learning_rate": 5.2180345381521626e-05,
+      "loss": 0.3625,
+      "step": 2087
+    },
+    {
+      "epoch": 0.5378497601184918,
+      "grad_norm": 0.17065952863639855,
+      "learning_rate": 5.213541744782022e-05,
+      "loss": 0.3356,
+      "step": 2088
+    },
+    {
+      "epoch": 0.5381073509997746,
+      "grad_norm": 0.20943735287421217,
+      "learning_rate": 5.209048778674898e-05,
+      "loss": 0.4261,
+      "step": 2089
+    },
+    {
+      "epoch": 0.5383649418810574,
+      "grad_norm": 0.281969776040582,
+      "learning_rate": 5.204555643465215e-05,
+      "loss": 0.518,
+      "step": 2090
+    },
+    {
+      "epoch": 0.5383649418810574,
+      "eval_PRM Accuracy": 0.8113207547169812,
+      "eval_PRM F1": 0.8837209302325582,
+      "eval_PRM F1 AUC": 0.6752226296490309,
+      "eval_PRM F1 AUC (fixed)": 0.8548978522786799,
+      "eval_PRM F1 Neg": 0.5,
+      "eval_PRM NPV": 0.5882352941176471,
+      "eval_PRM Precision": 0.8539325842696629,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.43478260869565216,
+      "eval_loss": 0.4005681872367859,
+      "eval_runtime": 16.9796,
+      "eval_samples_per_second": 2.591,
+      "eval_steps_per_second": 0.177,
+      "step": 2090
+    },
+    {
+      "epoch": 0.5386225327623402,
+      "grad_norm": 0.1156186939061372,
+      "learning_rate": 5.200062342787534e-05,
+      "loss": 0.3143,
+      "step": 2091
+    },
+    {
+      "epoch": 0.538880123643623,
+      "grad_norm": 0.15135735675938086,
+      "learning_rate": 5.195568880276552e-05,
+      "loss": 0.4564,
+      "step": 2092
+    },
+    {
+      "epoch": 0.5391377145249058,
+      "grad_norm": 0.16978548643168587,
+      "learning_rate": 5.191075259567093e-05,
+      "loss": 0.4064,
+      "step": 2093
+    },
+    {
+      "epoch": 0.5393953054061886,
+      "grad_norm": 0.21282430664393118,
+      "learning_rate": 5.1865814842941154e-05,
+      "loss": 0.4506,
+      "step": 2094
+    },
+    {
+      "epoch": 0.5396528962874714,
+      "grad_norm": 0.16778315374927444,
+      "learning_rate": 5.1820875580926944e-05,
+      "loss": 0.4113,
+      "step": 2095
+    },
+    {
+      "epoch": 0.5396528962874714,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.8679245283018868,
+      "eval_PRM F1 AUC": 0.7634887375589314,
+      "eval_PRM F1 AUC (fixed)": 0.8470403352540598,
+      "eval_PRM F1 Neg": 0.6037735849056604,
+      "eval_PRM NPV": 0.5333333333333333,
+      "eval_PRM Precision": 0.9078947368421053,
+      "eval_PRM Recall": 0.8313253012048193,
+      "eval_PRM Specificty": 0.6956521739130435,
+      "eval_loss": 0.38858309388160706,
+      "eval_runtime": 16.9531,
+      "eval_samples_per_second": 2.595,
+      "eval_steps_per_second": 0.177,
+      "step": 2095
+    },
+    {
+      "epoch": 0.5399104871687542,
+      "grad_norm": 0.16713448520362614,
+      "learning_rate": 5.1775934845980336e-05,
+      "loss": 0.2613,
+      "step": 2096
+    },
+    {
+      "epoch": 0.5401680780500371,
+      "grad_norm": 0.20617830957458205,
+      "learning_rate": 5.173099267445451e-05,
+      "loss": 0.4076,
+      "step": 2097
+    },
+    {
+      "epoch": 0.5404256689313198,
+      "grad_norm": 0.15584144043550646,
+      "learning_rate": 5.1686049102703883e-05,
+      "loss": 0.3088,
+      "step": 2098
+    },
+    {
+      "epoch": 0.5406832598126027,
+      "grad_norm": 0.13971603999676588,
+      "learning_rate": 5.1641104167083897e-05,
+      "loss": 0.3324,
+      "step": 2099
+    },
+    {
+      "epoch": 0.5409408506938854,
+      "grad_norm": 0.15170984031034837,
+      "learning_rate": 5.1596157903951205e-05,
+      "loss": 0.3263,
+      "step": 2100
+    },
+    {
+      "epoch": 0.5409408506938854,
+      "eval_PRM Accuracy": 0.7924528301886793,
+      "eval_PRM F1": 0.8641975308641975,
+      "eval_PRM F1 AUC": 0.7260345730749084,
+      "eval_PRM F1 AUC (fixed)": 0.8412781561026715,
+      "eval_PRM F1 Neg": 0.56,
+      "eval_PRM NPV": 0.5185185185185185,
+      "eval_PRM Precision": 0.8860759493670886,
+      "eval_PRM Recall": 0.8433734939759037,
+      "eval_PRM Specificty": 0.6086956521739131,
+      "eval_loss": 0.38352271914482117,
+      "eval_runtime": 17.1098,
+      "eval_samples_per_second": 2.572,
+      "eval_steps_per_second": 0.175,
+      "step": 2100
+    },
+    {
+      "epoch": 0.5411984415751683,
+      "grad_norm": 0.1264933633203901,
+      "learning_rate": 5.155121034966346e-05,
+      "loss": 0.3324,
+      "step": 2101
+    },
+    {
+      "epoch": 0.541456032456451,
+      "grad_norm": 0.13737548185601106,
+      "learning_rate": 5.150626154057939e-05,
+      "loss": 0.2788,
+      "step": 2102
+    },
+    {
+      "epoch": 0.5417136233377339,
+      "grad_norm": 0.15823635750885143,
+      "learning_rate": 5.146131151305873e-05,
+      "loss": 0.3731,
+      "step": 2103
+    },
+    {
+      "epoch": 0.5419712142190166,
+      "grad_norm": 0.1706746380591667,
+      "learning_rate": 5.1416360303462206e-05,
+      "loss": 0.3554,
+      "step": 2104
+    },
+    {
+      "epoch": 0.5422288051002995,
+      "grad_norm": 0.11837053908751594,
+      "learning_rate": 5.137140794815148e-05,
+      "loss": 0.3516,
+      "step": 2105
+    },
+    {
+      "epoch": 0.5422288051002995,
+      "eval_PRM Accuracy": 0.7924528301886793,
+      "eval_PRM F1": 0.8690476190476191,
+      "eval_PRM F1 AUC": 0.678889470927187,
+      "eval_PRM F1 AUC (fixed)": 0.8357778941854375,
+      "eval_PRM F1 Neg": 0.5,
+      "eval_PRM NPV": 0.5238095238095238,
+      "eval_PRM Precision": 0.8588235294117647,
+      "eval_PRM Recall": 0.8795180722891566,
+      "eval_PRM Specificty": 0.4782608695652174,
+      "eval_loss": 0.4169921875,
+      "eval_runtime": 16.4525,
+      "eval_samples_per_second": 2.674,
+      "eval_steps_per_second": 0.182,
+      "step": 2105
+    },
+    {
+      "epoch": 0.5424863959815822,
+      "grad_norm": 0.16723448052502865,
+      "learning_rate": 5.1326454483489196e-05,
+      "loss": 0.4145,
+      "step": 2106
+    },
+    {
+      "epoch": 0.5427439868628651,
+      "grad_norm": 0.20134589315950735,
+      "learning_rate": 5.1281499945838816e-05,
+      "loss": 0.4247,
+      "step": 2107
+    },
+    {
+      "epoch": 0.5430015777441478,
+      "grad_norm": 0.20585582310824724,
+      "learning_rate": 5.1236544371564735e-05,
+      "loss": 0.4477,
+      "step": 2108
+    },
+    {
+      "epoch": 0.5432591686254307,
+      "grad_norm": 0.14417732100778552,
+      "learning_rate": 5.119158779703215e-05,
+      "loss": 0.3215,
+      "step": 2109
+    },
+    {
+      "epoch": 0.5435167595067134,
+      "grad_norm": 0.18355234493223777,
+      "learning_rate": 5.11466302586071e-05,
+      "loss": 0.4771,
+      "step": 2110
+    },
+    {
+      "epoch": 0.5435167595067134,
+      "eval_PRM Accuracy": 0.8113207547169812,
+      "eval_PRM F1": 0.8823529411764706,
+      "eval_PRM F1 AUC": 0.6909376636982714,
+      "eval_PRM F1 AUC (fixed)": 0.8404924044002096,
+      "eval_PRM F1 Neg": 0.5238095238095238,
+      "eval_PRM NPV": 0.5789473684210527,
+      "eval_PRM Precision": 0.8620689655172413,
+      "eval_PRM Recall": 0.9036144578313253,
+      "eval_PRM Specificty": 0.4782608695652174,
+      "eval_loss": 0.42018821835517883,
+      "eval_runtime": 17.5956,
+      "eval_samples_per_second": 2.501,
+      "eval_steps_per_second": 0.17,
+      "step": 2110
+    },
+    {
+      "epoch": 0.5437743503879963,
+      "grad_norm": 0.17354688880894878,
+      "learning_rate": 5.110167179265636e-05,
+      "loss": 0.3518,
+      "step": 2111
+    },
+    {
+      "epoch": 0.5440319412692791,
+      "grad_norm": 0.1881262761253409,
+      "learning_rate": 5.1056712435547474e-05,
+      "loss": 0.4283,
+      "step": 2112
+    },
+    {
+      "epoch": 0.5442895321505619,
+      "grad_norm": 0.22049463774032257,
+      "learning_rate": 5.101175222364874e-05,
+      "loss": 0.4017,
+      "step": 2113
+    },
+    {
+      "epoch": 0.5445471230318447,
+      "grad_norm": 0.21867777753475734,
+      "learning_rate": 5.096679119332909e-05,
+      "loss": 0.4022,
+      "step": 2114
+    },
+    {
+      "epoch": 0.5448047139131275,
+      "grad_norm": 0.11658058194358625,
+      "learning_rate": 5.0921829380958174e-05,
+      "loss": 0.311,
+      "step": 2115
+    },
+    {
+      "epoch": 0.5448047139131275,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.8711656441717791,
+      "eval_PRM F1 AUC": 0.7320586694604505,
+      "eval_PRM F1 AUC (fixed)": 0.8394447354635934,
+      "eval_PRM F1 Neg": 0.5714285714285714,
+      "eval_PRM NPV": 0.5384615384615384,
+      "eval_PRM Precision": 0.8875,
+      "eval_PRM Recall": 0.8554216867469879,
+      "eval_PRM Specificty": 0.6086956521739131,
+      "eval_loss": 0.38210228085517883,
+      "eval_runtime": 17.0375,
+      "eval_samples_per_second": 2.583,
+      "eval_steps_per_second": 0.176,
+      "step": 2115
+    },
+    {
+      "epoch": 0.5450623047944103,
+      "grad_norm": 0.15349217186802172,
+      "learning_rate": 5.0876866822906245e-05,
+      "loss": 0.4086,
+      "step": 2116
+    },
+    {
+      "epoch": 0.5453198956756931,
+      "grad_norm": 0.16851131583427995,
+      "learning_rate": 5.0831903555544135e-05,
+      "loss": 0.4019,
+      "step": 2117
+    },
+    {
+      "epoch": 0.5455774865569759,
+      "grad_norm": 0.20822578670286693,
+      "learning_rate": 5.078693961524329e-05,
+      "loss": 0.3894,
+      "step": 2118
+    },
+    {
+      "epoch": 0.5458350774382587,
+      "grad_norm": 0.1843641977188793,
+      "learning_rate": 5.07419750383757e-05,
+      "loss": 0.3926,
+      "step": 2119
+    },
+    {
+      "epoch": 0.5460926683195415,
+      "grad_norm": 0.21264790823720447,
+      "learning_rate": 5.069700986131384e-05,
+      "loss": 0.339,
+      "step": 2120
+    },
+    {
+      "epoch": 0.5460926683195415,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.8695652173913043,
+      "eval_PRM F1 AUC": 0.747773703509691,
+      "eval_PRM F1 AUC (fixed)": 0.8404924044002096,
+      "eval_PRM F1 Neg": 0.5882352941176471,
+      "eval_PRM NPV": 0.5357142857142857,
+      "eval_PRM Precision": 0.8974358974358975,
+      "eval_PRM Recall": 0.8433734939759037,
+      "eval_PRM Specificty": 0.6521739130434783,
+      "eval_loss": 0.3820134997367859,
+      "eval_runtime": 17.4691,
+      "eval_samples_per_second": 2.519,
+      "eval_steps_per_second": 0.172,
+      "step": 2120
+    },
+    {
+      "epoch": 0.5463502592008243,
+      "grad_norm": 0.16334836817993717,
+      "learning_rate": 5.0652044120430707e-05,
+      "loss": 0.3713,
+      "step": 2121
+    },
+    {
+      "epoch": 0.5466078500821071,
+      "grad_norm": 0.1623736795236759,
+      "learning_rate": 5.060707785209972e-05,
+      "loss": 0.3496,
+      "step": 2122
+    },
+    {
+      "epoch": 0.5468654409633898,
+      "grad_norm": 0.1856282482779356,
+      "learning_rate": 5.0562111092694744e-05,
+      "loss": 0.3738,
+      "step": 2123
+    },
+    {
+      "epoch": 0.5471230318446727,
+      "grad_norm": 0.16458214441897714,
+      "learning_rate": 5.051714387859001e-05,
+      "loss": 0.4095,
+      "step": 2124
+    },
+    {
+      "epoch": 0.5473806227259554,
+      "grad_norm": 0.15882753369852362,
+      "learning_rate": 5.0472176246160184e-05,
+      "loss": 0.3166,
+      "step": 2125
+    },
+    {
+      "epoch": 0.5473806227259554,
+      "eval_PRM Accuracy": 0.8207547169811321,
+      "eval_PRM F1": 0.8862275449101796,
+      "eval_PRM F1 AUC": 0.7283918281822945,
+      "eval_PRM F1 AUC (fixed)": 0.8496595075955997,
+      "eval_PRM F1 Neg": 0.5777777777777777,
+      "eval_PRM NPV": 0.5909090909090909,
+      "eval_PRM Precision": 0.8809523809523809,
+      "eval_PRM Recall": 0.891566265060241,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.37890625,
+      "eval_runtime": 17.4158,
+      "eval_samples_per_second": 2.526,
+      "eval_steps_per_second": 0.172,
+      "step": 2125
+    },
+    {
+      "epoch": 0.5476382136072383,
+      "grad_norm": 0.1309588342317957,
+      "learning_rate": 5.042720823178022e-05,
+      "loss": 0.3712,
+      "step": 2126
+    },
+    {
+      "epoch": 0.5478958044885212,
+      "grad_norm": 0.15892977107103531,
+      "learning_rate": 5.0382239871825365e-05,
+      "loss": 0.4008,
+      "step": 2127
+    },
+    {
+      "epoch": 0.5481533953698039,
+      "grad_norm": 0.1375237916698276,
+      "learning_rate": 5.03372712026712e-05,
+      "loss": 0.3384,
+      "step": 2128
+    },
+    {
+      "epoch": 0.5484109862510868,
+      "grad_norm": 0.2633674567019821,
+      "learning_rate": 5.029230226069351e-05,
+      "loss": 0.4293,
+      "step": 2129
+    },
+    {
+      "epoch": 0.5486685771323695,
+      "grad_norm": 0.1312412710653147,
+      "learning_rate": 5.024733308226833e-05,
+      "loss": 0.3817,
+      "step": 2130
+    },
+    {
+      "epoch": 0.5486685771323695,
+      "eval_PRM Accuracy": 0.8301886792452831,
+      "eval_PRM F1": 0.8928571428571429,
+      "eval_PRM F1 AUC": 0.7344159245678366,
+      "eval_PRM F1 AUC (fixed)": 0.8517548454688318,
+      "eval_PRM F1 Neg": 0.5909090909090909,
+      "eval_PRM NPV": 0.6190476190476191,
+      "eval_PRM Precision": 0.8823529411764706,
+      "eval_PRM Recall": 0.9036144578313253,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.3759765625,
+      "eval_runtime": 16.9679,
+      "eval_samples_per_second": 2.593,
+      "eval_steps_per_second": 0.177,
+      "step": 2130
+    },
+    {
+      "epoch": 0.5489261680136523,
+      "grad_norm": 0.1328720900684053,
+      "learning_rate": 5.020236370377186e-05,
+      "loss": 0.3462,
+      "step": 2131
+    },
+    {
+      "epoch": 0.5491837588949351,
+      "grad_norm": 0.1751367425938839,
+      "learning_rate": 5.01573941615805e-05,
+      "loss": 0.4523,
+      "step": 2132
+    },
+    {
+      "epoch": 0.549441349776218,
+      "grad_norm": 0.11848844530055186,
+      "learning_rate": 5.0112424492070716e-05,
+      "loss": 0.3,
+      "step": 2133
+    },
+    {
+      "epoch": 0.5496989406575007,
+      "grad_norm": 0.13458102728384466,
+      "learning_rate": 5.006745473161917e-05,
+      "loss": 0.3874,
+      "step": 2134
+    },
+    {
+      "epoch": 0.5499565315387835,
+      "grad_norm": 0.1346350269991044,
+      "learning_rate": 5.002248491660251e-05,
+      "loss": 0.3357,
+      "step": 2135
+    },
+    {
+      "epoch": 0.5499565315387835,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.8727272727272727,
+      "eval_PRM F1 AUC": 0.7163436354112102,
+      "eval_PRM F1 AUC (fixed)": 0.852016762702986,
+      "eval_PRM F1 Neg": 0.5531914893617021,
+      "eval_PRM NPV": 0.5416666666666666,
+      "eval_PRM Precision": 0.8780487804878049,
+      "eval_PRM Recall": 0.8674698795180723,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.3757990002632141,
+      "eval_runtime": 17.0315,
+      "eval_samples_per_second": 2.583,
+      "eval_steps_per_second": 0.176,
+      "step": 2135
+    },
+    {
+      "epoch": 0.5502141224200663,
+      "grad_norm": 0.1280585930813598,
+      "learning_rate": 4.99775150833975e-05,
+      "loss": 0.3617,
+      "step": 2136
+    },
+    {
+      "epoch": 0.5504717133013491,
+      "grad_norm": 0.23752514813094275,
+      "learning_rate": 4.9932545268380824e-05,
+      "loss": 0.4613,
+      "step": 2137
+    },
+    {
+      "epoch": 0.5507293041826319,
+      "grad_norm": 0.12008766716175923,
+      "learning_rate": 4.988757550792929e-05,
+      "loss": 0.3089,
+      "step": 2138
+    },
+    {
+      "epoch": 0.5509868950639147,
+      "grad_norm": 0.174931204391963,
+      "learning_rate": 4.984260583841953e-05,
+      "loss": 0.4089,
+      "step": 2139
+    },
+    {
+      "epoch": 0.5512444859451976,
+      "grad_norm": 0.16401091970078174,
+      "learning_rate": 4.979763629622815e-05,
+      "loss": 0.3493,
+      "step": 2140
+    },
+    {
+      "epoch": 0.5512444859451976,
+      "eval_PRM Accuracy": 0.7830188679245284,
+      "eval_PRM F1": 0.8571428571428571,
+      "eval_PRM F1 AUC": 0.7200104766893662,
+      "eval_PRM F1 AUC (fixed)": 0.8493975903614458,
+      "eval_PRM F1 Neg": 0.5490196078431373,
+      "eval_PRM NPV": 0.5,
+      "eval_PRM Precision": 0.8846153846153846,
+      "eval_PRM Recall": 0.8313253012048193,
+      "eval_PRM Specificty": 0.6086956521739131,
+      "eval_loss": 0.3702947497367859,
+      "eval_runtime": 17.4167,
+      "eval_samples_per_second": 2.526,
+      "eval_steps_per_second": 0.172,
+      "step": 2140
+    },
+    {
+      "epoch": 0.5515020768264803,
+      "grad_norm": 0.1421853946676774,
+      "learning_rate": 4.975266691773168e-05,
+      "loss": 0.373,
+      "step": 2141
+    },
+    {
+      "epoch": 0.5517596677077632,
+      "grad_norm": 0.16680146122780656,
+      "learning_rate": 4.9707697739306494e-05,
+      "loss": 0.3754,
+      "step": 2142
+    },
+    {
+      "epoch": 0.5520172585890459,
+      "grad_norm": 0.1493812218399558,
+      "learning_rate": 4.9662728797328815e-05,
+      "loss": 0.3307,
+      "step": 2143
+    },
+    {
+      "epoch": 0.5522748494703288,
+      "grad_norm": 0.1417776209584457,
+      "learning_rate": 4.961776012817464e-05,
+      "loss": 0.338,
+      "step": 2144
+    },
+    {
+      "epoch": 0.5525324403516115,
+      "grad_norm": 0.19235181629224005,
+      "learning_rate": 4.9572791768219794e-05,
+      "loss": 0.3649,
+      "step": 2145
+    },
+    {
+      "epoch": 0.5525324403516115,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.8695652173913043,
+      "eval_PRM F1 AUC": 0.747773703509691,
+      "eval_PRM F1 AUC (fixed)": 0.856469355683604,
+      "eval_PRM F1 Neg": 0.5882352941176471,
+      "eval_PRM NPV": 0.5357142857142857,
+      "eval_PRM Precision": 0.8974358974358975,
+      "eval_PRM Recall": 0.8433734939759037,
+      "eval_PRM Specificty": 0.6521739130434783,
+      "eval_loss": 0.3668323755264282,
+      "eval_runtime": 17.471,
+      "eval_samples_per_second": 2.518,
+      "eval_steps_per_second": 0.172,
+      "step": 2145
+    },
+    {
+      "epoch": 0.5527900312328944,
+      "grad_norm": 0.18791856734878726,
+      "learning_rate": 4.9527823753839834e-05,
+      "loss": 0.3634,
+      "step": 2146
+    },
+    {
+      "epoch": 0.5530476221141771,
+      "grad_norm": 0.1584590627816371,
+      "learning_rate": 4.9482856121409996e-05,
+      "loss": 0.343,
+      "step": 2147
+    },
+    {
+      "epoch": 0.55330521299546,
+      "grad_norm": 0.1355920607532572,
+      "learning_rate": 4.943788890730529e-05,
+      "loss": 0.386,
+      "step": 2148
+    },
+    {
+      "epoch": 0.5535628038767427,
+      "grad_norm": 0.14377155856871848,
+      "learning_rate": 4.9392922147900294e-05,
+      "loss": 0.3652,
+      "step": 2149
+    },
+    {
+      "epoch": 0.5538203947580256,
+      "grad_norm": 0.14175723136724785,
+      "learning_rate": 4.9347955879569305e-05,
+      "loss": 0.3691,
+      "step": 2150
+    },
+    {
+      "epoch": 0.5538203947580256,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.8727272727272727,
+      "eval_PRM F1 AUC": 0.7163436354112102,
+      "eval_PRM F1 AUC (fixed)": 0.8585646935568361,
+      "eval_PRM F1 Neg": 0.5531914893617021,
+      "eval_PRM NPV": 0.5416666666666666,
+      "eval_PRM Precision": 0.8780487804878049,
+      "eval_PRM Recall": 0.8674698795180723,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.37109375,
+      "eval_runtime": 16.885,
+      "eval_samples_per_second": 2.606,
+      "eval_steps_per_second": 0.178,
+      "step": 2150
+    },
+    {
+      "epoch": 0.5540779856393083,
+      "grad_norm": 0.23453621043292955,
+      "learning_rate": 4.930299013868615e-05,
+      "loss": 0.5249,
+      "step": 2151
+    },
+    {
+      "epoch": 0.5543355765205912,
+      "grad_norm": 0.1344789117262499,
+      "learning_rate": 4.92580249616243e-05,
+      "loss": 0.3675,
+      "step": 2152
+    },
+    {
+      "epoch": 0.5545931674018739,
+      "grad_norm": 0.15020191341899067,
+      "learning_rate": 4.9213060384756716e-05,
+      "loss": 0.3039,
+      "step": 2153
+    },
+    {
+      "epoch": 0.5548507582831568,
+      "grad_norm": 0.13188450363898974,
+      "learning_rate": 4.916809644445588e-05,
+      "loss": 0.3587,
+      "step": 2154
+    },
+    {
+      "epoch": 0.5551083491644396,
+      "grad_norm": 0.15676676363128247,
+      "learning_rate": 4.912313317709378e-05,
+      "loss": 0.3547,
+      "step": 2155
+    },
+    {
+      "epoch": 0.5551083491644396,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.8727272727272727,
+      "eval_PRM F1 AUC": 0.7163436354112102,
+      "eval_PRM F1 AUC (fixed)": 0.8630172865374542,
+      "eval_PRM F1 Neg": 0.5531914893617021,
+      "eval_PRM NPV": 0.5416666666666666,
+      "eval_PRM Precision": 0.8780487804878049,
+      "eval_PRM Recall": 0.8674698795180723,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.3636363744735718,
+      "eval_runtime": 17.2898,
+      "eval_samples_per_second": 2.545,
+      "eval_steps_per_second": 0.174,
+      "step": 2155
+    },
+    {
+      "epoch": 0.5553659400457224,
+      "grad_norm": 0.12183958225770586,
+      "learning_rate": 4.9078170619041824e-05,
+      "loss": 0.3513,
+      "step": 2156
+    },
+    {
+      "epoch": 0.5556235309270052,
+      "grad_norm": 0.14353723932149423,
+      "learning_rate": 4.9033208806670925e-05,
+      "loss": 0.3542,
+      "step": 2157
+    },
+    {
+      "epoch": 0.555881121808288,
+      "grad_norm": 0.12241634739682451,
+      "learning_rate": 4.898824777635126e-05,
+      "loss": 0.3816,
+      "step": 2158
+    },
+    {
+      "epoch": 0.5561387126895708,
+      "grad_norm": 0.17791820413785403,
+      "learning_rate": 4.894328756445254e-05,
+      "loss": 0.3862,
+      "step": 2159
+    },
+    {
+      "epoch": 0.5563963035708536,
+      "grad_norm": 0.18737974074821792,
+      "learning_rate": 4.8898328207343666e-05,
+      "loss": 0.406,
+      "step": 2160
+    },
+    {
+      "epoch": 0.5563963035708536,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.8727272727272727,
+      "eval_PRM F1 AUC": 0.7163436354112102,
+      "eval_PRM F1 AUC (fixed)": 0.8596123624934521,
+      "eval_PRM F1 Neg": 0.5531914893617021,
+      "eval_PRM NPV": 0.5416666666666666,
+      "eval_PRM Precision": 0.8780487804878049,
+      "eval_PRM Recall": 0.8674698795180723,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.3660333752632141,
+      "eval_runtime": 17.3906,
+      "eval_samples_per_second": 2.53,
+      "eval_steps_per_second": 0.173,
+      "step": 2160
+    },
+    {
+      "epoch": 0.5566538944521364,
+      "grad_norm": 0.22074640164072148,
+      "learning_rate": 4.885336974139292e-05,
+      "loss": 0.5398,
+      "step": 2161
+    },
+    {
+      "epoch": 0.5569114853334192,
+      "grad_norm": 0.16297530028327453,
+      "learning_rate": 4.880841220296786e-05,
+      "loss": 0.415,
+      "step": 2162
+    },
+    {
+      "epoch": 0.557169076214702,
+      "grad_norm": 0.11433109053291302,
+      "learning_rate": 4.876345562843527e-05,
+      "loss": 0.3514,
+      "step": 2163
+    },
+    {
+      "epoch": 0.5574266670959848,
+      "grad_norm": 0.1398536676505433,
+      "learning_rate": 4.8718500054161195e-05,
+      "loss": 0.3723,
+      "step": 2164
+    },
+    {
+      "epoch": 0.5576842579772676,
+      "grad_norm": 0.1478329378108419,
+      "learning_rate": 4.867354551651081e-05,
+      "loss": 0.3927,
+      "step": 2165
+    },
+    {
+      "epoch": 0.5576842579772676,
+      "eval_PRM Accuracy": 0.8207547169811321,
+      "eval_PRM F1": 0.8862275449101796,
+      "eval_PRM F1 AUC": 0.7283918281822945,
+      "eval_PRM F1 AUC (fixed)": 0.8491356731272918,
+      "eval_PRM F1 Neg": 0.5777777777777777,
+      "eval_PRM NPV": 0.5909090909090909,
+      "eval_PRM Precision": 0.8809523809523809,
+      "eval_PRM Recall": 0.891566265060241,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.3755326569080353,
+      "eval_runtime": 17.1232,
+      "eval_samples_per_second": 2.57,
+      "eval_steps_per_second": 0.175,
+      "step": 2165
+    },
+    {
+      "epoch": 0.5579418488585504,
+      "grad_norm": 0.13688230907613783,
+      "learning_rate": 4.862859205184853e-05,
+      "loss": 0.356,
+      "step": 2166
+    },
+    {
+      "epoch": 0.5581994397398332,
+      "grad_norm": 0.14598957880294913,
+      "learning_rate": 4.858363969653781e-05,
+      "loss": 0.3203,
+      "step": 2167
+    },
+    {
+      "epoch": 0.558457030621116,
+      "grad_norm": 0.14978586225686757,
+      "learning_rate": 4.853868848694128e-05,
+      "loss": 0.3873,
+      "step": 2168
+    },
+    {
+      "epoch": 0.5587146215023988,
+      "grad_norm": 0.17697588446454016,
+      "learning_rate": 4.8493738459420625e-05,
+      "loss": 0.3713,
+      "step": 2169
+    },
+    {
+      "epoch": 0.5589722123836817,
+      "grad_norm": 0.18040210280731417,
+      "learning_rate": 4.844878965033655e-05,
+      "loss": 0.3624,
+      "step": 2170
+    },
+    {
+      "epoch": 0.5589722123836817,
+      "eval_PRM Accuracy": 0.8113207547169812,
+      "eval_PRM F1": 0.8780487804878049,
+      "eval_PRM F1 AUC": 0.7380827658459927,
+      "eval_PRM F1 AUC (fixed)": 0.8446830801466736,
+      "eval_PRM F1 Neg": 0.5833333333333334,
+      "eval_PRM NPV": 0.56,
+      "eval_PRM Precision": 0.8888888888888888,
+      "eval_PRM Recall": 0.8674698795180723,
+      "eval_PRM Specificty": 0.6086956521739131,
+      "eval_loss": 0.3761541247367859,
+      "eval_runtime": 17.2164,
+      "eval_samples_per_second": 2.556,
+      "eval_steps_per_second": 0.174,
+      "step": 2170
+    },
+    {
+      "epoch": 0.5592298032649644,
+      "grad_norm": 0.1964276593543136,
+      "learning_rate": 4.84038420960488e-05,
+      "loss": 0.3832,
+      "step": 2171
+    },
+    {
+      "epoch": 0.5594873941462473,
+      "grad_norm": 0.16389978703949507,
+      "learning_rate": 4.83588958329161e-05,
+      "loss": 0.4719,
+      "step": 2172
+    },
+    {
+      "epoch": 0.55974498502753,
+      "grad_norm": 0.12567930008990083,
+      "learning_rate": 4.8313950897296135e-05,
+      "loss": 0.3569,
+      "step": 2173
+    },
+    {
+      "epoch": 0.5600025759088129,
+      "grad_norm": 0.11451491285445857,
+      "learning_rate": 4.8269007325545506e-05,
+      "loss": 0.3477,
+      "step": 2174
+    },
+    {
+      "epoch": 0.5602601667900956,
+      "grad_norm": 0.1342974107810076,
+      "learning_rate": 4.8224065154019676e-05,
+      "loss": 0.3394,
+      "step": 2175
+    },
+    {
+      "epoch": 0.5602601667900956,
+      "eval_PRM Accuracy": 0.8113207547169812,
+      "eval_PRM F1": 0.8765432098765432,
+      "eval_PRM F1 AUC": 0.7537977998952332,
+      "eval_PRM F1 AUC (fixed)": 0.8425877422734416,
+      "eval_PRM F1 Neg": 0.6,
+      "eval_PRM NPV": 0.5555555555555556,
+      "eval_PRM Precision": 0.8987341772151899,
+      "eval_PRM Recall": 0.8554216867469879,
+      "eval_PRM Specificty": 0.6521739130434783,
+      "eval_loss": 0.3746448755264282,
+      "eval_runtime": 17.0633,
+      "eval_samples_per_second": 2.579,
+      "eval_steps_per_second": 0.176,
+      "step": 2175
+    },
+    {
+      "epoch": 0.5605177576713785,
+      "grad_norm": 0.19016684018239385,
+      "learning_rate": 4.817912441907308e-05,
+      "loss": 0.3162,
+      "step": 2176
+    },
+    {
+      "epoch": 0.5607753485526612,
+      "grad_norm": 0.14054461274800253,
+      "learning_rate": 4.813418515705885e-05,
+      "loss": 0.3744,
+      "step": 2177
+    },
+    {
+      "epoch": 0.5610329394339441,
+      "grad_norm": 0.16813177450742112,
+      "learning_rate": 4.808924740432907e-05,
+      "loss": 0.2878,
+      "step": 2178
+    },
+    {
+      "epoch": 0.5612905303152268,
+      "grad_norm": 0.14553774665583524,
+      "learning_rate": 4.804431119723449e-05,
+      "loss": 0.358,
+      "step": 2179
+    },
+    {
+      "epoch": 0.5615481211965097,
+      "grad_norm": 0.25533237074550735,
+      "learning_rate": 4.799937657212467e-05,
+      "loss": 0.5151,
+      "step": 2180
+    },
+    {
+      "epoch": 0.5615481211965097,
+      "eval_PRM Accuracy": 0.8113207547169812,
+      "eval_PRM F1": 0.8795180722891566,
+      "eval_PRM F1 AUC": 0.7223677317967523,
+      "eval_PRM F1 AUC (fixed)": 0.8399685699319015,
+      "eval_PRM F1 Neg": 0.5652173913043478,
+      "eval_PRM NPV": 0.5652173913043478,
+      "eval_PRM Precision": 0.8795180722891566,
+      "eval_PRM Recall": 0.8795180722891566,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.3894708752632141,
+      "eval_runtime": 17.0529,
+      "eval_samples_per_second": 2.58,
+      "eval_steps_per_second": 0.176,
+      "step": 2180
+    },
+    {
+      "epoch": 0.5618057120777924,
+      "grad_norm": 0.1297733900890675,
+      "learning_rate": 4.7954443565347865e-05,
+      "loss": 0.3325,
+      "step": 2181
+    },
+    {
+      "epoch": 0.5620633029590753,
+      "grad_norm": 0.1673331346690515,
+      "learning_rate": 4.7909512213251026e-05,
+      "loss": 0.3791,
+      "step": 2182
+    },
+    {
+      "epoch": 0.5623208938403581,
+      "grad_norm": 0.1343079400856452,
+      "learning_rate": 4.7864582552179796e-05,
+      "loss": 0.2454,
+      "step": 2183
+    },
+    {
+      "epoch": 0.5625784847216408,
+      "grad_norm": 0.2087779801405549,
+      "learning_rate": 4.781965461847838e-05,
+      "loss": 0.3994,
+      "step": 2184
+    },
+    {
+      "epoch": 0.5628360756029237,
+      "grad_norm": 0.141789777382412,
+      "learning_rate": 4.7774728448489666e-05,
+      "loss": 0.4016,
+      "step": 2185
+    },
+    {
+      "epoch": 0.5628360756029237,
+      "eval_PRM Accuracy": 0.7924528301886793,
+      "eval_PRM F1": 0.8690476190476191,
+      "eval_PRM F1 AUC": 0.678889470927187,
+      "eval_PRM F1 AUC (fixed)": 0.8365636458878994,
+      "eval_PRM F1 Neg": 0.5,
+      "eval_PRM NPV": 0.5238095238095238,
+      "eval_PRM Precision": 0.8588235294117647,
+      "eval_PRM Recall": 0.8795180722891566,
+      "eval_PRM Specificty": 0.4782608695652174,
+      "eval_loss": 0.39089134335517883,
+      "eval_runtime": 16.7172,
+      "eval_samples_per_second": 2.632,
+      "eval_steps_per_second": 0.179,
+      "step": 2185
+    },
+    {
+      "epoch": 0.5630936664842064,
+      "grad_norm": 0.1965836602182714,
+      "learning_rate": 4.7729804078555044e-05,
+      "loss": 0.4113,
+      "step": 2186
+    },
+    {
+      "epoch": 0.5633512573654893,
+      "grad_norm": 0.18292467806038407,
+      "learning_rate": 4.768488154501454e-05,
+      "loss": 0.3361,
+      "step": 2187
+    },
+    {
+      "epoch": 0.563608848246772,
+      "grad_norm": 0.16613394402746923,
+      "learning_rate": 4.7639960884206576e-05,
+      "loss": 0.3961,
+      "step": 2188
+    },
+    {
+      "epoch": 0.5638664391280549,
+      "grad_norm": 0.18376328117834048,
+      "learning_rate": 4.759504213246813e-05,
+      "loss": 0.3649,
+      "step": 2189
+    },
+    {
+      "epoch": 0.5641240300093376,
+      "grad_norm": 0.15517133244551723,
+      "learning_rate": 4.755012532613465e-05,
+      "loss": 0.3249,
+      "step": 2190
+    },
+    {
+      "epoch": 0.5641240300093376,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.8727272727272727,
+      "eval_PRM F1 AUC": 0.7163436354112102,
+      "eval_PRM F1 AUC (fixed)": 0.8381351492928234,
+      "eval_PRM F1 Neg": 0.5531914893617021,
+      "eval_PRM NPV": 0.5416666666666666,
+      "eval_PRM Precision": 0.8780487804878049,
+      "eval_PRM Recall": 0.8674698795180723,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.3722478747367859,
+      "eval_runtime": 16.9506,
+      "eval_samples_per_second": 2.596,
+      "eval_steps_per_second": 0.177,
+      "step": 2190
+    },
+    {
+      "epoch": 0.5643816208906205,
+      "grad_norm": 0.12600302601372262,
+      "learning_rate": 4.7505210501539966e-05,
+      "loss": 0.366,
+      "step": 2191
+    },
+    {
+      "epoch": 0.5646392117719032,
+      "grad_norm": 0.16316320475752782,
+      "learning_rate": 4.746029769501634e-05,
+      "loss": 0.3683,
+      "step": 2192
+    },
+    {
+      "epoch": 0.5648968026531861,
+      "grad_norm": 0.18499636302691863,
+      "learning_rate": 4.741538694289437e-05,
+      "loss": 0.4064,
+      "step": 2193
+    },
+    {
+      "epoch": 0.5651543935344688,
+      "grad_norm": 0.1303058154252544,
+      "learning_rate": 4.7370478281503014e-05,
+      "loss": 0.3632,
+      "step": 2194
+    },
+    {
+      "epoch": 0.5654119844157517,
+      "grad_norm": 0.15355381606296517,
+      "learning_rate": 4.7325571747169545e-05,
+      "loss": 0.3509,
+      "step": 2195
+    },
+    {
+      "epoch": 0.5654119844157517,
+      "eval_PRM Accuracy": 0.7830188679245284,
+      "eval_PRM F1": 0.8588957055214724,
+      "eval_PRM F1 AUC": 0.7042954426401258,
+      "eval_PRM F1 AUC (fixed)": 0.8410162388685175,
+      "eval_PRM F1 Neg": 0.5306122448979592,
+      "eval_PRM NPV": 0.5,
+      "eval_PRM Precision": 0.875,
+      "eval_PRM Recall": 0.8433734939759037,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.3628373444080353,
+      "eval_runtime": 17.0342,
+      "eval_samples_per_second": 2.583,
+      "eval_steps_per_second": 0.176,
+      "step": 2195
+    },
+    {
+      "epoch": 0.5656695752970344,
+      "grad_norm": 0.1760926218321264,
+      "learning_rate": 4.728066737621947e-05,
+      "loss": 0.4315,
+      "step": 2196
+    },
+    {
+      "epoch": 0.5659271661783173,
+      "grad_norm": 0.21462043388412655,
+      "learning_rate": 4.723576520497663e-05,
+      "loss": 0.4928,
+      "step": 2197
+    },
+    {
+      "epoch": 0.5661847570596001,
+      "grad_norm": 0.12322984749559669,
+      "learning_rate": 4.719086526976298e-05,
+      "loss": 0.2569,
+      "step": 2198
+    },
+    {
+      "epoch": 0.5664423479408829,
+      "grad_norm": 0.12928031124907874,
+      "learning_rate": 4.714596760689878e-05,
+      "loss": 0.3709,
+      "step": 2199
+    },
+    {
+      "epoch": 0.5666999388221657,
+      "grad_norm": 0.14365617703509492,
+      "learning_rate": 4.710107225270236e-05,
+      "loss": 0.3403,
+      "step": 2200
+    },
+    {
+      "epoch": 0.5666999388221657,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.8727272727272727,
+      "eval_PRM F1 AUC": 0.7163436354112102,
+      "eval_PRM F1 AUC (fixed)": 0.8538501833420639,
+      "eval_PRM F1 Neg": 0.5531914893617021,
+      "eval_PRM NPV": 0.5416666666666666,
+      "eval_PRM Precision": 0.8780487804878049,
+      "eval_PRM Recall": 0.8674698795180723,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.35342684388160706,
+      "eval_runtime": 16.8425,
+      "eval_samples_per_second": 2.612,
+      "eval_steps_per_second": 0.178,
+      "step": 2200
+    },
+    {
+      "epoch": 0.5669575297034485,
+      "grad_norm": 0.12062531516149663,
+      "learning_rate": 4.70561792434902e-05,
+      "loss": 0.3082,
+      "step": 2201
+    },
+    {
+      "epoch": 0.5672151205847313,
+      "grad_norm": 0.1526087638521343,
+      "learning_rate": 4.7011288615576934e-05,
+      "loss": 0.3406,
+      "step": 2202
+    },
+    {
+      "epoch": 0.5674727114660141,
+      "grad_norm": 0.15644291328617355,
+      "learning_rate": 4.6966400405275224e-05,
+      "loss": 0.3636,
+      "step": 2203
+    },
+    {
+      "epoch": 0.5677303023472969,
+      "grad_norm": 0.13913944417182417,
+      "learning_rate": 4.6921514648895786e-05,
+      "loss": 0.3085,
+      "step": 2204
+    },
+    {
+      "epoch": 0.5679878932285797,
+      "grad_norm": 0.16084278441444716,
+      "learning_rate": 4.687663138274735e-05,
+      "loss": 0.3669,
+      "step": 2205
+    },
+    {
+      "epoch": 0.5679878932285797,
+      "eval_PRM Accuracy": 0.8113207547169812,
+      "eval_PRM F1": 0.8809523809523809,
+      "eval_PRM F1 AUC": 0.7066526977475118,
+      "eval_PRM F1 AUC (fixed)": 0.8627553693033002,
+      "eval_PRM F1 Neg": 0.5454545454545454,
+      "eval_PRM NPV": 0.5714285714285714,
+      "eval_PRM Precision": 0.8705882352941177,
+      "eval_PRM Recall": 0.891566265060241,
+      "eval_PRM Specificty": 0.5217391304347826,
+      "eval_loss": 0.3511185944080353,
+      "eval_runtime": 17.1619,
+      "eval_samples_per_second": 2.564,
+      "eval_steps_per_second": 0.175,
+      "step": 2205
+    },
+    {
+      "epoch": 0.5682454841098625,
+      "grad_norm": 0.20549808343963985,
+      "learning_rate": 4.683175064313666e-05,
+      "loss": 0.4485,
+      "step": 2206
+    },
+    {
+      "epoch": 0.5685030749911453,
+      "grad_norm": 0.12326408605710353,
+      "learning_rate": 4.6786872466368346e-05,
+      "loss": 0.3371,
+      "step": 2207
+    },
+    {
+      "epoch": 0.5687606658724281,
+      "grad_norm": 0.1786740596929532,
+      "learning_rate": 4.6741996888745034e-05,
+      "loss": 0.3124,
+      "step": 2208
+    },
+    {
+      "epoch": 0.5690182567537109,
+      "grad_norm": 0.15079983176315676,
+      "learning_rate": 4.6697123946567227e-05,
+      "loss": 0.3673,
+      "step": 2209
+    },
+    {
+      "epoch": 0.5692758476349937,
+      "grad_norm": 0.19944306111265092,
+      "learning_rate": 4.665225367613328e-05,
+      "loss": 0.4476,
+      "step": 2210
+    },
+    {
+      "epoch": 0.5692758476349937,
+      "eval_PRM Accuracy": 0.8113207547169812,
+      "eval_PRM F1": 0.8809523809523809,
+      "eval_PRM F1 AUC": 0.7066526977475118,
+      "eval_PRM F1 AUC (fixed)": 0.8656364588789942,
+      "eval_PRM F1 Neg": 0.5454545454545454,
+      "eval_PRM NPV": 0.5714285714285714,
+      "eval_PRM Precision": 0.8705882352941177,
+      "eval_PRM Recall": 0.891566265060241,
+      "eval_PRM Specificty": 0.5217391304347826,
+      "eval_loss": 0.3511185944080353,
+      "eval_runtime": 16.9474,
+      "eval_samples_per_second": 2.596,
+      "eval_steps_per_second": 0.177,
+      "step": 2210
+    },
+    {
+      "epoch": 0.5695334385162765,
+      "grad_norm": 0.14572671723711672,
+      "learning_rate": 4.6607386113739413e-05,
+      "loss": 0.3723,
+      "step": 2211
+    },
+    {
+      "epoch": 0.5697910293975593,
+      "grad_norm": 0.15106906527292013,
+      "learning_rate": 4.656252129567963e-05,
+      "loss": 0.3103,
+      "step": 2212
+    },
+    {
+      "epoch": 0.5700486202788422,
+      "grad_norm": 0.2634210600639788,
+      "learning_rate": 4.6517659258245734e-05,
+      "loss": 0.5098,
+      "step": 2213
+    },
+    {
+      "epoch": 0.5703062111601249,
+      "grad_norm": 0.11451975560835995,
+      "learning_rate": 4.647280003772728e-05,
+      "loss": 0.3236,
+      "step": 2214
+    },
+    {
+      "epoch": 0.5705638020414078,
+      "grad_norm": 0.1586227318953602,
+      "learning_rate": 4.6427943670411495e-05,
+      "loss": 0.3795,
+      "step": 2215
+    },
+    {
+      "epoch": 0.5705638020414078,
+      "eval_PRM Accuracy": 0.8301886792452831,
+      "eval_PRM F1": 0.8902439024390244,
+      "eval_PRM F1 AUC": 0.7658459926663174,
+      "eval_PRM F1 AUC (fixed)": 0.8643268727082243,
+      "eval_PRM F1 Neg": 0.625,
+      "eval_PRM NPV": 0.6,
+      "eval_PRM Precision": 0.9012345679012346,
+      "eval_PRM Recall": 0.8795180722891566,
+      "eval_PRM Specificty": 0.6521739130434783,
+      "eval_loss": 0.34108665585517883,
+      "eval_runtime": 17.7693,
+      "eval_samples_per_second": 2.476,
+      "eval_steps_per_second": 0.169,
+      "step": 2215
+    },
+    {
+      "epoch": 0.5708213929226905,
+      "grad_norm": 0.15676782722268975,
+      "learning_rate": 4.63830901925834e-05,
+      "loss": 0.3931,
+      "step": 2216
+    },
+    {
+      "epoch": 0.5710789838039734,
+      "grad_norm": 0.13454389321114535,
+      "learning_rate": 4.633823964052555e-05,
+      "loss": 0.369,
+      "step": 2217
+    },
+    {
+      "epoch": 0.5713365746852561,
+      "grad_norm": 0.12218625857130747,
+      "learning_rate": 4.629339205051827e-05,
+      "loss": 0.3256,
+      "step": 2218
+    },
+    {
+      "epoch": 0.571594165566539,
+      "grad_norm": 0.1375886521697247,
+      "learning_rate": 4.6248547458839365e-05,
+      "loss": 0.3486,
+      "step": 2219
+    },
+    {
+      "epoch": 0.5718517564478217,
+      "grad_norm": 0.1461004670609193,
+      "learning_rate": 4.62037059017643e-05,
+      "loss": 0.3915,
+      "step": 2220
+    },
+    {
+      "epoch": 0.5718517564478217,
+      "eval_PRM Accuracy": 0.8207547169811321,
+      "eval_PRM F1": 0.8834355828220859,
+      "eval_PRM F1 AUC": 0.7598218962807753,
+      "eval_PRM F1 AUC (fixed)": 0.859350445259298,
+      "eval_PRM F1 Neg": 0.6122448979591837,
+      "eval_PRM NPV": 0.5769230769230769,
+      "eval_PRM Precision": 0.9,
+      "eval_PRM Recall": 0.8674698795180723,
+      "eval_PRM Specificty": 0.6521739130434783,
+      "eval_loss": 0.34765625,
+      "eval_runtime": 16.9494,
+      "eval_samples_per_second": 2.596,
+      "eval_steps_per_second": 0.177,
+      "step": 2220
+    },
+    {
+      "epoch": 0.5721093473291046,
+      "grad_norm": 0.13185660987919753,
+      "learning_rate": 4.615886741556604e-05,
+      "loss": 0.3213,
+      "step": 2221
+    },
+    {
+      "epoch": 0.5723669382103873,
+      "grad_norm": 0.16499953154104235,
+      "learning_rate": 4.611403203651507e-05,
+      "loss": 0.3698,
+      "step": 2222
+    },
+    {
+      "epoch": 0.5726245290916702,
+      "grad_norm": 0.1837604920686579,
+      "learning_rate": 4.60691998008794e-05,
+      "loss": 0.3873,
+      "step": 2223
+    },
+    {
+      "epoch": 0.5728821199729529,
+      "grad_norm": 0.142755388671106,
+      "learning_rate": 4.602437074492445e-05,
+      "loss": 0.4132,
+      "step": 2224
+    },
+    {
+      "epoch": 0.5731397108542358,
+      "grad_norm": 0.15381425681621197,
+      "learning_rate": 4.59795449049131e-05,
+      "loss": 0.339,
+      "step": 2225
+    },
+    {
+      "epoch": 0.5731397108542358,
+      "eval_PRM Accuracy": 0.8113207547169812,
+      "eval_PRM F1": 0.8795180722891566,
+      "eval_PRM F1 AUC": 0.7223677317967523,
+      "eval_PRM F1 AUC (fixed)": 0.8548978522786799,
+      "eval_PRM F1 Neg": 0.5652173913043478,
+      "eval_PRM NPV": 0.5652173913043478,
+      "eval_PRM Precision": 0.8795180722891566,
+      "eval_PRM Recall": 0.8795180722891566,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.3608842194080353,
+      "eval_runtime": 17.1193,
+      "eval_samples_per_second": 2.57,
+      "eval_steps_per_second": 0.175,
+      "step": 2225
+    },
+    {
+      "epoch": 0.5733973017355186,
+      "grad_norm": 0.13265249259935646,
+      "learning_rate": 4.593472231710559e-05,
+      "loss": 0.296,
+      "step": 2226
+    },
+    {
+      "epoch": 0.5736548926168014,
+      "grad_norm": 0.10971577687365057,
+      "learning_rate": 4.58899030177596e-05,
+      "loss": 0.2991,
+      "step": 2227
+    },
+    {
+      "epoch": 0.5739124834980842,
+      "grad_norm": 0.1346092443842548,
+      "learning_rate": 4.584508704313006e-05,
+      "loss": 0.3516,
+      "step": 2228
+    },
+    {
+      "epoch": 0.574170074379367,
+      "grad_norm": 0.28546397698404885,
+      "learning_rate": 4.580027442946928e-05,
+      "loss": 0.5096,
+      "step": 2229
+    },
+    {
+      "epoch": 0.5744276652606498,
+      "grad_norm": 0.2153285699974356,
+      "learning_rate": 4.575546521302681e-05,
+      "loss": 0.4035,
+      "step": 2230
+    },
+    {
+      "epoch": 0.5744276652606498,
+      "eval_PRM Accuracy": 0.8301886792452831,
+      "eval_PRM F1": 0.8928571428571429,
+      "eval_PRM F1 AUC": 0.7344159245678366,
+      "eval_PRM F1 AUC (fixed)": 0.855945521215296,
+      "eval_PRM F1 Neg": 0.5909090909090909,
+      "eval_PRM NPV": 0.6190476190476191,
+      "eval_PRM Precision": 0.8823529411764706,
+      "eval_PRM Recall": 0.9036144578313253,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.376953125,
+      "eval_runtime": 16.9774,
+      "eval_samples_per_second": 2.592,
+      "eval_steps_per_second": 0.177,
+      "step": 2230
+    },
+    {
+      "epoch": 0.5746852561419326,
+      "grad_norm": 0.15833971603216263,
+      "learning_rate": 4.571065943004947e-05,
+      "loss": 0.403,
+      "step": 2231
+    },
+    {
+      "epoch": 0.5749428470232154,
+      "grad_norm": 0.16444949348925209,
+      "learning_rate": 4.566585711678133e-05,
+      "loss": 0.3791,
+      "step": 2232
+    },
+    {
+      "epoch": 0.5752004379044982,
+      "grad_norm": 0.1459790553316448,
+      "learning_rate": 4.56210583094636e-05,
+      "loss": 0.3827,
+      "step": 2233
+    },
+    {
+      "epoch": 0.575458028785781,
+      "grad_norm": 0.13441591004776918,
+      "learning_rate": 4.557626304433467e-05,
+      "loss": 0.4107,
+      "step": 2234
+    },
+    {
+      "epoch": 0.5757156196670637,
+      "grad_norm": 0.15489351648664135,
+      "learning_rate": 4.5531471357630114e-05,
+      "loss": 0.3618,
+      "step": 2235
+    },
+    {
+      "epoch": 0.5757156196670637,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.8711656441717791,
+      "eval_PRM F1 AUC": 0.7320586694604505,
+      "eval_PRM F1 AUC (fixed)": 0.8530644316396019,
+      "eval_PRM F1 Neg": 0.5714285714285714,
+      "eval_PRM NPV": 0.5384615384615384,
+      "eval_PRM Precision": 0.8875,
+      "eval_PRM Recall": 0.8554216867469879,
+      "eval_PRM Specificty": 0.6086956521739131,
+      "eval_loss": 0.37428978085517883,
+      "eval_runtime": 17.4467,
+      "eval_samples_per_second": 2.522,
+      "eval_steps_per_second": 0.172,
+      "step": 2235
+    },
+    {
+      "epoch": 0.5759732105483466,
+      "grad_norm": 0.1277990860423348,
+      "learning_rate": 4.54866832855825e-05,
+      "loss": 0.4014,
+      "step": 2236
+    },
+    {
+      "epoch": 0.5762308014296293,
+      "grad_norm": 0.15228208517801683,
+      "learning_rate": 4.544189886442162e-05,
+      "loss": 0.372,
+      "step": 2237
+    },
+    {
+      "epoch": 0.5764883923109122,
+      "grad_norm": 0.1254931002378707,
+      "learning_rate": 4.5397118130374175e-05,
+      "loss": 0.3067,
+      "step": 2238
+    },
+    {
+      "epoch": 0.5767459831921949,
+      "grad_norm": 0.15408585564454477,
+      "learning_rate": 4.535234111966399e-05,
+      "loss": 0.4115,
+      "step": 2239
+    },
+    {
+      "epoch": 0.5770035740734778,
+      "grad_norm": 0.15631776075479115,
+      "learning_rate": 4.5307567868511795e-05,
+      "loss": 0.4342,
+      "step": 2240
+    },
+    {
+      "epoch": 0.5770035740734778,
+      "eval_PRM Accuracy": 0.7924528301886793,
+      "eval_PRM F1": 0.8641975308641975,
+      "eval_PRM F1 AUC": 0.7260345730749084,
+      "eval_PRM F1 AUC (fixed)": 0.8473022524882137,
+      "eval_PRM F1 Neg": 0.56,
+      "eval_PRM NPV": 0.5185185185185185,
+      "eval_PRM Precision": 0.8860759493670886,
+      "eval_PRM Recall": 0.8433734939759037,
+      "eval_PRM Specificty": 0.6086956521739131,
+      "eval_loss": 0.3833451569080353,
+      "eval_runtime": 17.0431,
+      "eval_samples_per_second": 2.582,
+      "eval_steps_per_second": 0.176,
+      "step": 2240
+    },
+    {
+      "epoch": 0.5772611649547607,
+      "grad_norm": 0.1204536283126352,
+      "learning_rate": 4.5262798413135344e-05,
+      "loss": 0.3439,
+      "step": 2241
+    },
+    {
+      "epoch": 0.5775187558360434,
+      "grad_norm": 0.11277574795480862,
+      "learning_rate": 4.5218032789749275e-05,
+      "loss": 0.2972,
+      "step": 2242
+    },
+    {
+      "epoch": 0.5777763467173263,
+      "grad_norm": 0.13040420288598253,
+      "learning_rate": 4.517327103456515e-05,
+      "loss": 0.3522,
+      "step": 2243
+    },
+    {
+      "epoch": 0.578033937598609,
+      "grad_norm": 0.12038259999115875,
+      "learning_rate": 4.5128513183791386e-05,
+      "loss": 0.298,
+      "step": 2244
+    },
+    {
+      "epoch": 0.5782915284798918,
+      "grad_norm": 0.12879970384020212,
+      "learning_rate": 4.5083759273633267e-05,
+      "loss": 0.3444,
+      "step": 2245
+    },
+    {
+      "epoch": 0.5782915284798918,
+      "eval_PRM Accuracy": 0.8113207547169812,
+      "eval_PRM F1": 0.8780487804878049,
+      "eval_PRM F1 AUC": 0.7380827658459927,
+      "eval_PRM F1 AUC (fixed)": 0.8473022524882137,
+      "eval_PRM F1 Neg": 0.5833333333333334,
+      "eval_PRM NPV": 0.56,
+      "eval_PRM Precision": 0.8888888888888888,
+      "eval_PRM Recall": 0.8674698795180723,
+      "eval_PRM Specificty": 0.6086956521739131,
+      "eval_loss": 0.38387784361839294,
+      "eval_runtime": 16.4767,
+      "eval_samples_per_second": 2.67,
+      "eval_steps_per_second": 0.182,
+      "step": 2245
+    },
+    {
+      "epoch": 0.5785491193611746,
+      "grad_norm": 0.1312248062851336,
+      "learning_rate": 4.503900934029287e-05,
+      "loss": 0.3658,
+      "step": 2246
+    },
+    {
+      "epoch": 0.5788067102424574,
+      "grad_norm": 0.1264918196677566,
+      "learning_rate": 4.4994263419969055e-05,
+      "loss": 0.3519,
+      "step": 2247
+    },
+    {
+      "epoch": 0.5790643011237402,
+      "grad_norm": 0.11470048536136435,
+      "learning_rate": 4.4949521548857426e-05,
+      "loss": 0.349,
+      "step": 2248
+    },
+    {
+      "epoch": 0.579321892005023,
+      "grad_norm": 0.15467134731951687,
+      "learning_rate": 4.490478376315035e-05,
+      "loss": 0.3333,
+      "step": 2249
+    },
+    {
+      "epoch": 0.5795794828863058,
+      "grad_norm": 0.1514767223007565,
+      "learning_rate": 4.486005009903684e-05,
+      "loss": 0.3993,
+      "step": 2250
+    },
+    {
+      "epoch": 0.5795794828863058,
+      "eval_PRM Accuracy": 0.8207547169811321,
+      "eval_PRM F1": 0.8862275449101796,
+      "eval_PRM F1 AUC": 0.7283918281822945,
+      "eval_PRM F1 AUC (fixed)": 0.855945521215296,
+      "eval_PRM F1 Neg": 0.5777777777777777,
+      "eval_PRM NPV": 0.5909090909090909,
+      "eval_PRM Precision": 0.8809523809523809,
+      "eval_PRM Recall": 0.891566265060241,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.3881392180919647,
+      "eval_runtime": 17.398,
+      "eval_samples_per_second": 2.529,
+      "eval_steps_per_second": 0.172,
+      "step": 2250
+    },
+    {
+      "epoch": 0.5798370737675886,
+      "grad_norm": 0.20915160984468872,
+      "learning_rate": 4.481532059270262e-05,
+      "loss": 0.3845,
+      "step": 2251
+    },
+    {
+      "epoch": 0.5800946646488714,
+      "grad_norm": 0.16023912248270797,
+      "learning_rate": 4.477059528033001e-05,
+      "loss": 0.4117,
+      "step": 2252
+    },
+    {
+      "epoch": 0.5803522555301542,
+      "grad_norm": 0.18700292286454162,
+      "learning_rate": 4.472587419809798e-05,
+      "loss": 0.4213,
+      "step": 2253
+    },
+    {
+      "epoch": 0.580609846411437,
+      "grad_norm": 0.12058361241555736,
+      "learning_rate": 4.468115738218206e-05,
+      "loss": 0.4035,
+      "step": 2254
+    },
+    {
+      "epoch": 0.5808674372927198,
+      "grad_norm": 0.15736977413293526,
+      "learning_rate": 4.463644486875427e-05,
+      "loss": 0.4465,
+      "step": 2255
+    },
+    {
+      "epoch": 0.5808674372927198,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.8711656441717791,
+      "eval_PRM F1 AUC": 0.7320586694604505,
+      "eval_PRM F1 AUC (fixed)": 0.8533263488737559,
+      "eval_PRM F1 Neg": 0.5714285714285714,
+      "eval_PRM NPV": 0.5384615384615384,
+      "eval_PRM Precision": 0.8875,
+      "eval_PRM Recall": 0.8554216867469879,
+      "eval_PRM Specificty": 0.6086956521739131,
+      "eval_loss": 0.37180396914482117,
+      "eval_runtime": 16.8676,
+      "eval_samples_per_second": 2.609,
+      "eval_steps_per_second": 0.178,
+      "step": 2255
+    },
+    {
+      "epoch": 0.5811250281740027,
+      "grad_norm": 0.14231846289545302,
+      "learning_rate": 4.459173669398328e-05,
+      "loss": 0.3966,
+      "step": 2256
+    },
+    {
+      "epoch": 0.5813826190552854,
+      "grad_norm": 0.15200014494583405,
+      "learning_rate": 4.454703289403411e-05,
+      "loss": 0.4429,
+      "step": 2257
+    },
+    {
+      "epoch": 0.5816402099365683,
+      "grad_norm": 0.1484057973180248,
+      "learning_rate": 4.450233350506836e-05,
+      "loss": 0.3177,
+      "step": 2258
+    },
+    {
+      "epoch": 0.581897800817851,
+      "grad_norm": 0.1454146292852184,
+      "learning_rate": 4.4457638563243954e-05,
+      "loss": 0.3001,
+      "step": 2259
+    },
+    {
+      "epoch": 0.5821553916991339,
+      "grad_norm": 0.18622364318324686,
+      "learning_rate": 4.4412948104715316e-05,
+      "loss": 0.3707,
+      "step": 2260
+    },
+    {
+      "epoch": 0.5821553916991339,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.8695652173913043,
+      "eval_PRM F1 AUC": 0.747773703509691,
+      "eval_PRM F1 AUC (fixed)": 0.8467784180199056,
+      "eval_PRM F1 Neg": 0.5882352941176471,
+      "eval_PRM NPV": 0.5357142857142857,
+      "eval_PRM Precision": 0.8974358974358975,
+      "eval_PRM Recall": 0.8433734939759037,
+      "eval_PRM Specificty": 0.6521739130434783,
+      "eval_loss": 0.37411221861839294,
+      "eval_runtime": 17.0686,
+      "eval_samples_per_second": 2.578,
+      "eval_steps_per_second": 0.176,
+      "step": 2260
+    },
+    {
+      "epoch": 0.5824129825804166,
+      "grad_norm": 0.18289434281705721,
+      "learning_rate": 4.436826216563318e-05,
+      "loss": 0.3948,
+      "step": 2261
+    },
+    {
+      "epoch": 0.5826705734616995,
+      "grad_norm": 0.12791602046198763,
+      "learning_rate": 4.432358078214464e-05,
+      "loss": 0.308,
+      "step": 2262
+    },
+    {
+      "epoch": 0.5829281643429822,
+      "grad_norm": 0.14830241151439877,
+      "learning_rate": 4.4278903990393127e-05,
+      "loss": 0.2993,
+      "step": 2263
+    },
+    {
+      "epoch": 0.5831857552242651,
+      "grad_norm": 0.15756705571132135,
+      "learning_rate": 4.423423182651831e-05,
+      "loss": 0.4263,
+      "step": 2264
+    },
+    {
+      "epoch": 0.5834433461055478,
+      "grad_norm": 0.14695797512698278,
+      "learning_rate": 4.418956432665618e-05,
+      "loss": 0.4075,
+      "step": 2265
+    },
+    {
+      "epoch": 0.5834433461055478,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.8727272727272727,
+      "eval_PRM F1 AUC": 0.7163436354112102,
+      "eval_PRM F1 AUC (fixed)": 0.8470403352540599,
+      "eval_PRM F1 Neg": 0.5531914893617021,
+      "eval_PRM NPV": 0.5416666666666666,
+      "eval_PRM Precision": 0.8780487804878049,
+      "eval_PRM Recall": 0.8674698795180723,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.3751775622367859,
+      "eval_runtime": 17.512,
+      "eval_samples_per_second": 2.513,
+      "eval_steps_per_second": 0.171,
+      "step": 2265
+    },
+    {
+      "epoch": 0.5837009369868307,
+      "grad_norm": 0.13259319305805958,
+      "learning_rate": 4.4144901526938886e-05,
+      "loss": 0.398,
+      "step": 2266
+    },
+    {
+      "epoch": 0.5839585278681134,
+      "grad_norm": 0.1593042136574566,
+      "learning_rate": 4.410024346349484e-05,
+      "loss": 0.4099,
+      "step": 2267
+    },
+    {
+      "epoch": 0.5842161187493963,
+      "grad_norm": 0.1451339019173708,
+      "learning_rate": 4.405559017244857e-05,
+      "loss": 0.3785,
+      "step": 2268
+    },
+    {
+      "epoch": 0.5844737096306791,
+      "grad_norm": 0.2515281631502885,
+      "learning_rate": 4.401094168992075e-05,
+      "loss": 0.4094,
+      "step": 2269
+    },
+    {
+      "epoch": 0.5847313005119619,
+      "grad_norm": 0.19495752680724165,
+      "learning_rate": 4.396629805202821e-05,
+      "loss": 0.3853,
+      "step": 2270
+    },
+    {
+      "epoch": 0.5847313005119619,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.8727272727272727,
+      "eval_PRM F1 AUC": 0.7163436354112102,
+      "eval_PRM F1 AUC (fixed)": 0.8446830801466737,
+      "eval_PRM F1 Neg": 0.5531914893617021,
+      "eval_PRM NPV": 0.5416666666666666,
+      "eval_PRM Precision": 0.8780487804878049,
+      "eval_PRM Recall": 0.8674698795180723,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.3761541247367859,
+      "eval_runtime": 17.172,
+      "eval_samples_per_second": 2.562,
+      "eval_steps_per_second": 0.175,
+      "step": 2270
+    },
+    {
+      "epoch": 0.5849888913932447,
+      "grad_norm": 0.15725677876385077,
+      "learning_rate": 4.392165929488381e-05,
+      "loss": 0.3737,
+      "step": 2271
+    },
+    {
+      "epoch": 0.5852464822745275,
+      "grad_norm": 0.13637584450534468,
+      "learning_rate": 4.387702545459649e-05,
+      "loss": 0.3327,
+      "step": 2272
+    },
+    {
+      "epoch": 0.5855040731558103,
+      "grad_norm": 0.1561340421962783,
+      "learning_rate": 4.383239656727119e-05,
+      "loss": 0.4173,
+      "step": 2273
+    },
+    {
+      "epoch": 0.5857616640370931,
+      "grad_norm": 0.2638794545999978,
+      "learning_rate": 4.378777266900889e-05,
+      "loss": 0.5171,
+      "step": 2274
+    },
+    {
+      "epoch": 0.5860192549183759,
+      "grad_norm": 0.12291816196261775,
+      "learning_rate": 4.374315379590649e-05,
+      "loss": 0.3587,
+      "step": 2275
+    },
+    {
+      "epoch": 0.5860192549183759,
+      "eval_PRM Accuracy": 0.7924528301886793,
+      "eval_PRM F1": 0.8641975308641975,
+      "eval_PRM F1 AUC": 0.7260345730749084,
+      "eval_PRM F1 AUC (fixed)": 0.8368255631220533,
+      "eval_PRM F1 Neg": 0.56,
+      "eval_PRM NPV": 0.5185185185185185,
+      "eval_PRM Precision": 0.8860759493670886,
+      "eval_PRM Recall": 0.8433734939759037,
+      "eval_PRM Specificty": 0.6086956521739131,
+      "eval_loss": 0.37801846861839294,
+      "eval_runtime": 17.4753,
+      "eval_samples_per_second": 2.518,
+      "eval_steps_per_second": 0.172,
+      "step": 2275
+    },
+    {
+      "epoch": 0.5862768457996587,
+      "grad_norm": 0.1803958721122535,
+      "learning_rate": 4.36985399840568e-05,
+      "loss": 0.4139,
+      "step": 2276
+    },
+    {
+      "epoch": 0.5865344366809415,
+      "grad_norm": 0.19998996472480873,
+      "learning_rate": 4.365393126954863e-05,
+      "loss": 0.3019,
+      "step": 2277
+    },
+    {
+      "epoch": 0.5867920275622243,
+      "grad_norm": 0.14917447809816536,
+      "learning_rate": 4.3609327688466546e-05,
+      "loss": 0.3539,
+      "step": 2278
+    },
+    {
+      "epoch": 0.5870496184435071,
+      "grad_norm": 0.154365871065513,
+      "learning_rate": 4.356472927689109e-05,
+      "loss": 0.3742,
+      "step": 2279
+    },
+    {
+      "epoch": 0.5873072093247899,
+      "grad_norm": 0.12829951247327792,
+      "learning_rate": 4.352013607089852e-05,
+      "loss": 0.365,
+      "step": 2280
+    },
+    {
+      "epoch": 0.5873072093247899,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.8711656441717791,
+      "eval_PRM F1 AUC": 0.7320586694604505,
+      "eval_PRM F1 AUC (fixed)": 0.8302776322682034,
+      "eval_PRM F1 Neg": 0.5714285714285714,
+      "eval_PRM NPV": 0.5384615384615384,
+      "eval_PRM Precision": 0.8875,
+      "eval_PRM Recall": 0.8554216867469879,
+      "eval_PRM Specificty": 0.6086956521739131,
+      "eval_loss": 0.3888494372367859,
+      "eval_runtime": 17.2207,
+      "eval_samples_per_second": 2.555,
+      "eval_steps_per_second": 0.174,
+      "step": 2280
+    },
+    {
+      "epoch": 0.5875648002060727,
+      "grad_norm": 0.16134000007480917,
+      "learning_rate": 4.347554810656093e-05,
+      "loss": 0.4404,
+      "step": 2281
+    },
+    {
+      "epoch": 0.5878223910873555,
+      "grad_norm": 0.13756861935496567,
+      "learning_rate": 4.343096541994617e-05,
+      "loss": 0.3328,
+      "step": 2282
+    },
+    {
+      "epoch": 0.5880799819686383,
+      "grad_norm": 0.11227877119448444,
+      "learning_rate": 4.3386388047117796e-05,
+      "loss": 0.2859,
+      "step": 2283
+    },
+    {
+      "epoch": 0.5883375728499212,
+      "grad_norm": 0.1295742460879775,
+      "learning_rate": 4.334181602413511e-05,
+      "loss": 0.3618,
+      "step": 2284
+    },
+    {
+      "epoch": 0.5885951637312039,
+      "grad_norm": 0.13980724148963705,
+      "learning_rate": 4.329724938705304e-05,
+      "loss": 0.3653,
+      "step": 2285
+    },
+    {
+      "epoch": 0.5885951637312039,
+      "eval_PRM Accuracy": 0.7830188679245284,
+      "eval_PRM F1": 0.8622754491017964,
+      "eval_PRM F1 AUC": 0.6728653745416449,
+      "eval_PRM F1 AUC (fixed)": 0.8300157150340493,
+      "eval_PRM F1 Neg": 0.4888888888888889,
+      "eval_PRM NPV": 0.5,
+      "eval_PRM Precision": 0.8571428571428571,
+      "eval_PRM Recall": 0.8674698795180723,
+      "eval_PRM Specificty": 0.4782608695652174,
+      "eval_loss": 0.4008345305919647,
+      "eval_runtime": 17.3421,
+      "eval_samples_per_second": 2.537,
+      "eval_steps_per_second": 0.173,
+      "step": 2285
+    },
+    {
+      "epoch": 0.5888527546124868,
+      "grad_norm": 0.13831366656318222,
+      "learning_rate": 4.32526881719222e-05,
+      "loss": 0.3624,
+      "step": 2286
+    },
+    {
+      "epoch": 0.5891103454937695,
+      "grad_norm": 0.13796271646368452,
+      "learning_rate": 4.320813241478876e-05,
+      "loss": 0.3287,
+      "step": 2287
+    },
+    {
+      "epoch": 0.5893679363750524,
+      "grad_norm": 0.20813717625802478,
+      "learning_rate": 4.3163582151694573e-05,
+      "loss": 0.3713,
+      "step": 2288
+    },
+    {
+      "epoch": 0.5896255272563351,
+      "grad_norm": 0.12250415747063537,
+      "learning_rate": 4.3119037418676933e-05,
+      "loss": 0.2924,
+      "step": 2289
+    },
+    {
+      "epoch": 0.589883118137618,
+      "grad_norm": 0.19821969688382862,
+      "learning_rate": 4.3074498251768716e-05,
+      "loss": 0.4017,
+      "step": 2290
+    },
+    {
+      "epoch": 0.589883118137618,
+      "eval_PRM Accuracy": 0.8113207547169812,
+      "eval_PRM F1": 0.8823529411764706,
+      "eval_PRM F1 AUC": 0.6909376636982714,
+      "eval_PRM F1 AUC (fixed)": 0.8433734939759038,
+      "eval_PRM F1 Neg": 0.5238095238095238,
+      "eval_PRM NPV": 0.5789473684210527,
+      "eval_PRM Precision": 0.8620689655172413,
+      "eval_PRM Recall": 0.9036144578313253,
+      "eval_PRM Specificty": 0.4782608695652174,
+      "eval_loss": 0.4139293432235718,
+      "eval_runtime": 17.5685,
+      "eval_samples_per_second": 2.504,
+      "eval_steps_per_second": 0.171,
+      "step": 2290
+    },
+    {
+      "epoch": 0.5901407090189007,
+      "grad_norm": 0.11948246377924135,
+      "learning_rate": 4.302996468699833e-05,
+      "loss": 0.2829,
+      "step": 2291
+    },
+    {
+      "epoch": 0.5903982999001836,
+      "grad_norm": 0.2492267578473308,
+      "learning_rate": 4.298543676038958e-05,
+      "loss": 0.4189,
+      "step": 2292
+    },
+    {
+      "epoch": 0.5906558907814663,
+      "grad_norm": 0.19005796072967893,
+      "learning_rate": 4.2940914507961775e-05,
+      "loss": 0.3656,
+      "step": 2293
+    },
+    {
+      "epoch": 0.5909134816627492,
+      "grad_norm": 0.24832687920935803,
+      "learning_rate": 4.2896397965729575e-05,
+      "loss": 0.3959,
+      "step": 2294
+    },
+    {
+      "epoch": 0.5911710725440319,
+      "grad_norm": 0.2203340009224552,
+      "learning_rate": 4.285188716970307e-05,
+      "loss": 0.4418,
+      "step": 2295
+    },
+    {
+      "epoch": 0.5911710725440319,
+      "eval_PRM Accuracy": 0.7924528301886793,
+      "eval_PRM F1": 0.8658536585365854,
+      "eval_PRM F1 AUC": 0.710319539025668,
+      "eval_PRM F1 AUC (fixed)": 0.8397066526977477,
+      "eval_PRM F1 Neg": 0.5416666666666666,
+      "eval_PRM NPV": 0.52,
+      "eval_PRM Precision": 0.8765432098765432,
+      "eval_PRM Recall": 0.8554216867469879,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.3896484375,
+      "eval_runtime": 17.6247,
+      "eval_samples_per_second": 2.496,
+      "eval_steps_per_second": 0.17,
+      "step": 2295
+    },
+    {
+      "epoch": 0.5914286634253147,
+      "grad_norm": 0.13878427099407956,
+      "learning_rate": 4.280738215588769e-05,
+      "loss": 0.3719,
+      "step": 2296
+    },
+    {
+      "epoch": 0.5916862543065975,
+      "grad_norm": 0.15998587752502577,
+      "learning_rate": 4.276288296028413e-05,
+      "loss": 0.3582,
+      "step": 2297
+    },
+    {
+      "epoch": 0.5919438451878803,
+      "grad_norm": 0.16923710750646784,
+      "learning_rate": 4.27183896188885e-05,
+      "loss": 0.3764,
+      "step": 2298
+    },
+    {
+      "epoch": 0.5922014360691632,
+      "grad_norm": 0.13725800110456088,
+      "learning_rate": 4.267390216769205e-05,
+      "loss": 0.3212,
+      "step": 2299
+    },
+    {
+      "epoch": 0.5924590269504459,
+      "grad_norm": 0.1443056880246243,
+      "learning_rate": 4.262942064268134e-05,
+      "loss": 0.3848,
+      "step": 2300
+    },
+    {
+      "epoch": 0.5924590269504459,
+      "eval_PRM Accuracy": 0.7924528301886793,
+      "eval_PRM F1": 0.8625,
+      "eval_PRM F1 AUC": 0.7417496071241488,
+      "eval_PRM F1 AUC (fixed)": 0.8397066526977476,
+      "eval_PRM F1 Neg": 0.5769230769230769,
+      "eval_PRM NPV": 0.5172413793103449,
+      "eval_PRM Precision": 0.8961038961038961,
+      "eval_PRM Recall": 0.8313253012048193,
+      "eval_PRM Specificty": 0.6521739130434783,
+      "eval_loss": 0.3832564055919647,
+      "eval_runtime": 17.5859,
+      "eval_samples_per_second": 2.502,
+      "eval_steps_per_second": 0.171,
+      "step": 2300
+    },
+    {
+      "epoch": 0.5927166178317288,
+      "grad_norm": 0.1925039231194253,
+      "learning_rate": 4.2584945079838096e-05,
+      "loss": 0.3163,
+      "step": 2301
+    },
+    {
+      "epoch": 0.5929742087130115,
+      "grad_norm": 0.21111701745013034,
+      "learning_rate": 4.2540475515139264e-05,
+      "loss": 0.4348,
+      "step": 2302
+    },
+    {
+      "epoch": 0.5932317995942944,
+      "grad_norm": 0.16305082349405703,
+      "learning_rate": 4.249601198455691e-05,
+      "loss": 0.3986,
+      "step": 2303
+    },
+    {
+      "epoch": 0.5934893904755771,
+      "grad_norm": 0.13293669947815345,
+      "learning_rate": 4.245155452405819e-05,
+      "loss": 0.2639,
+      "step": 2304
+    },
+    {
+      "epoch": 0.59374698135686,
+      "grad_norm": 0.18048699718115427,
+      "learning_rate": 4.2407103169605436e-05,
+      "loss": 0.4084,
+      "step": 2305
+    },
+    {
+      "epoch": 0.59374698135686,
+      "eval_PRM Accuracy": 0.7924528301886793,
+      "eval_PRM F1": 0.8625,
+      "eval_PRM F1 AUC": 0.7417496071241488,
+      "eval_PRM F1 AUC (fixed)": 0.8441592456783658,
+      "eval_PRM F1 Neg": 0.5769230769230769,
+      "eval_PRM NPV": 0.5172413793103449,
+      "eval_PRM Precision": 0.8961038961038961,
+      "eval_PRM Recall": 0.8313253012048193,
+      "eval_PRM Specificty": 0.6521739130434783,
+      "eval_loss": 0.37784090638160706,
+      "eval_runtime": 16.8174,
+      "eval_samples_per_second": 2.616,
+      "eval_steps_per_second": 0.178,
+      "step": 2305
+    },
+    {
+      "epoch": 0.5940045722381427,
+      "grad_norm": 0.1798571551854095,
+      "learning_rate": 4.2362657957155934e-05,
+      "loss": 0.4594,
+      "step": 2306
+    },
+    {
+      "epoch": 0.5942621631194256,
+      "grad_norm": 0.14420653302408004,
+      "learning_rate": 4.23182189226621e-05,
+      "loss": 0.2714,
+      "step": 2307
+    },
+    {
+      "epoch": 0.5945197540007083,
+      "grad_norm": 0.1505100604261495,
+      "learning_rate": 4.2273786102071275e-05,
+      "loss": 0.3587,
+      "step": 2308
+    },
+    {
+      "epoch": 0.5947773448819912,
+      "grad_norm": 0.14331384765206512,
+      "learning_rate": 4.2229359531325804e-05,
+      "loss": 0.402,
+      "step": 2309
+    },
+    {
+      "epoch": 0.5950349357632739,
+      "grad_norm": 0.14206105415515183,
+      "learning_rate": 4.2184939246363005e-05,
+      "loss": 0.4072,
+      "step": 2310
+    },
+    {
+      "epoch": 0.5950349357632739,
+      "eval_PRM Accuracy": 0.7924528301886793,
+      "eval_PRM F1": 0.8658536585365854,
+      "eval_PRM F1 AUC": 0.710319539025668,
+      "eval_PRM F1 AUC (fixed)": 0.8528025144054479,
+      "eval_PRM F1 Neg": 0.5416666666666666,
+      "eval_PRM NPV": 0.52,
+      "eval_PRM Precision": 0.8765432098765432,
+      "eval_PRM Recall": 0.8554216867469879,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.3757990002632141,
+      "eval_runtime": 17.7482,
+      "eval_samples_per_second": 2.479,
+      "eval_steps_per_second": 0.169,
+      "step": 2310
+    },
+    {
+      "epoch": 0.5952925266445568,
+      "grad_norm": 0.12844389785248492,
+      "learning_rate": 4.214052528311506e-05,
+      "loss": 0.3422,
+      "step": 2311
+    },
+    {
+      "epoch": 0.5955501175258396,
+      "grad_norm": 0.1417563698914759,
+      "learning_rate": 4.209611767750908e-05,
+      "loss": 0.3675,
+      "step": 2312
+    },
+    {
+      "epoch": 0.5958077084071224,
+      "grad_norm": 0.18466106503914595,
+      "learning_rate": 4.2051716465466986e-05,
+      "loss": 0.4099,
+      "step": 2313
+    },
+    {
+      "epoch": 0.5960652992884052,
+      "grad_norm": 0.21904032491039754,
+      "learning_rate": 4.20073216829056e-05,
+      "loss": 0.4417,
+      "step": 2314
+    },
+    {
+      "epoch": 0.596322890169688,
+      "grad_norm": 0.14913382778989,
+      "learning_rate": 4.196293336573648e-05,
+      "loss": 0.3804,
+      "step": 2315
+    },
+    {
+      "epoch": 0.596322890169688,
+      "eval_PRM Accuracy": 0.8207547169811321,
+      "eval_PRM F1": 0.8862275449101796,
+      "eval_PRM F1 AUC": 0.7283918281822945,
+      "eval_PRM F1 AUC (fixed)": 0.8517548454688318,
+      "eval_PRM F1 Neg": 0.5777777777777777,
+      "eval_PRM NPV": 0.5909090909090909,
+      "eval_PRM Precision": 0.8809523809523809,
+      "eval_PRM Recall": 0.891566265060241,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.3802379369735718,
+      "eval_runtime": 16.9919,
+      "eval_samples_per_second": 2.589,
+      "eval_steps_per_second": 0.177,
+      "step": 2315
+    },
+    {
+      "epoch": 0.5965804810509708,
+      "grad_norm": 0.13716170546344314,
+      "learning_rate": 4.1918551549865934e-05,
+      "loss": 0.3671,
+      "step": 2316
+    },
+    {
+      "epoch": 0.5968380719322536,
+      "grad_norm": 0.13506852473226214,
+      "learning_rate": 4.187417627119511e-05,
+      "loss": 0.3434,
+      "step": 2317
+    },
+    {
+      "epoch": 0.5970956628135364,
+      "grad_norm": 0.1313861150309788,
+      "learning_rate": 4.182980756561976e-05,
+      "loss": 0.36,
+      "step": 2318
+    },
+    {
+      "epoch": 0.5973532536948192,
+      "grad_norm": 0.12254271130288111,
+      "learning_rate": 4.178544546903042e-05,
+      "loss": 0.3909,
+      "step": 2319
+    },
+    {
+      "epoch": 0.597610844576102,
+      "grad_norm": 0.14060596925147895,
+      "learning_rate": 4.1741090017312166e-05,
+      "loss": 0.3432,
+      "step": 2320
+    },
+    {
+      "epoch": 0.597610844576102,
+      "eval_PRM Accuracy": 0.839622641509434,
+      "eval_PRM F1": 0.896969696969697,
+      "eval_PRM F1 AUC": 0.7718700890518595,
+      "eval_PRM F1 AUC (fixed)": 0.8444211629125197,
+      "eval_PRM F1 Neg": 0.6382978723404256,
+      "eval_PRM NPV": 0.625,
+      "eval_PRM Precision": 0.9024390243902439,
+      "eval_PRM Recall": 0.891566265060241,
+      "eval_PRM Specificty": 0.6521739130434783,
+      "eval_loss": 0.38503196835517883,
+      "eval_runtime": 17.1014,
+      "eval_samples_per_second": 2.573,
+      "eval_steps_per_second": 0.175,
+      "step": 2320
+    },
+    {
+      "epoch": 0.5978684354573848,
+      "grad_norm": 0.168004846727644,
+      "learning_rate": 4.169674124634481e-05,
+      "loss": 0.4644,
+      "step": 2321
+    },
+    {
+      "epoch": 0.5981260263386676,
+      "grad_norm": 0.1415225286134437,
+      "learning_rate": 4.165239919200269e-05,
+      "loss": 0.2826,
+      "step": 2322
+    },
+    {
+      "epoch": 0.5983836172199504,
+      "grad_norm": 0.13137494742688535,
+      "learning_rate": 4.1608063890154715e-05,
+      "loss": 0.4054,
+      "step": 2323
+    },
+    {
+      "epoch": 0.5986412081012332,
+      "grad_norm": 0.1470378613652157,
+      "learning_rate": 4.156373537666437e-05,
+      "loss": 0.3757,
+      "step": 2324
+    },
+    {
+      "epoch": 0.598898798982516,
+      "grad_norm": 0.1269355336304008,
+      "learning_rate": 4.1519413687389615e-05,
+      "loss": 0.3243,
+      "step": 2325
+    },
+    {
+      "epoch": 0.598898798982516,
+      "eval_PRM Accuracy": 0.8301886792452831,
+      "eval_PRM F1": 0.8941176470588236,
+      "eval_PRM F1 AUC": 0.7187008905185961,
+      "eval_PRM F1 AUC (fixed)": 0.8452069146149818,
+      "eval_PRM F1 Neg": 0.5714285714285714,
+      "eval_PRM NPV": 0.631578947368421,
+      "eval_PRM Precision": 0.8735632183908046,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.5217391304347826,
+      "eval_loss": 0.3919566869735718,
+      "eval_runtime": 17.288,
+      "eval_samples_per_second": 2.545,
+      "eval_steps_per_second": 0.174,
+      "step": 2325
+    },
+    {
+      "epoch": 0.5991563898637988,
+      "grad_norm": 0.14258608967689515,
+      "learning_rate": 4.147509885818292e-05,
+      "loss": 0.3029,
+      "step": 2326
+    },
+    {
+      "epoch": 0.5994139807450817,
+      "grad_norm": 0.13317579369354718,
+      "learning_rate": 4.143079092489114e-05,
+      "loss": 0.3345,
+      "step": 2327
+    },
+    {
+      "epoch": 0.5996715716263644,
+      "grad_norm": 0.18059828585327012,
+      "learning_rate": 4.138648992335566e-05,
+      "loss": 0.4281,
+      "step": 2328
+    },
+    {
+      "epoch": 0.5999291625076473,
+      "grad_norm": 0.13015801332157306,
+      "learning_rate": 4.1342195889412166e-05,
+      "loss": 0.4118,
+      "step": 2329
+    },
+    {
+      "epoch": 0.60018675338893,
+      "grad_norm": 0.21798644743809673,
+      "learning_rate": 4.129790885889072e-05,
+      "loss": 0.4471,
+      "step": 2330
+    },
+    {
+      "epoch": 0.60018675338893,
+      "eval_PRM Accuracy": 0.8207547169811321,
+      "eval_PRM F1": 0.8888888888888888,
+      "eval_PRM F1 AUC": 0.6969617600838135,
+      "eval_PRM F1 AUC (fixed)": 0.8410162388685175,
+      "eval_PRM F1 Neg": 0.5365853658536586,
+      "eval_PRM NPV": 0.6111111111111112,
+      "eval_PRM Precision": 0.8636363636363636,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.4782608695652174,
+      "eval_loss": 0.4052290618419647,
+      "eval_runtime": 17.4006,
+      "eval_samples_per_second": 2.529,
+      "eval_steps_per_second": 0.172,
+      "step": 2330
+    },
+    {
+      "epoch": 0.6004443442702129,
+      "grad_norm": 0.13610508936668073,
+      "learning_rate": 4.125362886761577e-05,
+      "loss": 0.3657,
+      "step": 2331
+    },
+    {
+      "epoch": 0.6007019351514956,
+      "grad_norm": 0.15504773900939664,
+      "learning_rate": 4.1209355951406e-05,
+      "loss": 0.4261,
+      "step": 2332
+    },
+    {
+      "epoch": 0.6009595260327785,
+      "grad_norm": 0.1446863802292206,
+      "learning_rate": 4.1165090146074446e-05,
+      "loss": 0.3632,
+      "step": 2333
+    },
+    {
+      "epoch": 0.6012171169140612,
+      "grad_norm": 0.13063788776631363,
+      "learning_rate": 4.112083148742833e-05,
+      "loss": 0.3635,
+      "step": 2334
+    },
+    {
+      "epoch": 0.6014747077953441,
+      "grad_norm": 0.12052295245282314,
+      "learning_rate": 4.107658001126913e-05,
+      "loss": 0.3628,
+      "step": 2335
+    },
+    {
+      "epoch": 0.6014747077953441,
+      "eval_PRM Accuracy": 0.8113207547169812,
+      "eval_PRM F1": 0.8809523809523809,
+      "eval_PRM F1 AUC": 0.7066526977475118,
+      "eval_PRM F1 AUC (fixed)": 0.8305395495023573,
+      "eval_PRM F1 Neg": 0.5454545454545454,
+      "eval_PRM NPV": 0.5714285714285714,
+      "eval_PRM Precision": 0.8705882352941177,
+      "eval_PRM Recall": 0.891566265060241,
+      "eval_PRM Specificty": 0.5217391304347826,
+      "eval_loss": 0.4041193127632141,
+      "eval_runtime": 17.7381,
+      "eval_samples_per_second": 2.481,
+      "eval_steps_per_second": 0.169,
+      "step": 2335
+    },
+    {
+      "epoch": 0.6017322986766268,
+      "grad_norm": 0.14273888256518527,
+      "learning_rate": 4.1032335753392505e-05,
+      "loss": 0.4155,
+      "step": 2336
+    },
+    {
+      "epoch": 0.6019898895579097,
+      "grad_norm": 0.19249131742072517,
+      "learning_rate": 4.098809874958822e-05,
+      "loss": 0.4896,
+      "step": 2337
+    },
+    {
+      "epoch": 0.6022474804391924,
+      "grad_norm": 0.1203162215241167,
+      "learning_rate": 4.09438690356403e-05,
+      "loss": 0.4073,
+      "step": 2338
+    },
+    {
+      "epoch": 0.6025050713204753,
+      "grad_norm": 0.1726093816833048,
+      "learning_rate": 4.089964664732674e-05,
+      "loss": 0.4277,
+      "step": 2339
+    },
+    {
+      "epoch": 0.602762662201758,
+      "grad_norm": 0.1295230914383838,
+      "learning_rate": 4.085543162041969e-05,
+      "loss": 0.3623,
+      "step": 2340
+    },
+    {
+      "epoch": 0.602762662201758,
+      "eval_PRM Accuracy": 0.7924528301886793,
+      "eval_PRM F1": 0.8641975308641975,
+      "eval_PRM F1 AUC": 0.7260345730749084,
+      "eval_PRM F1 AUC (fixed)": 0.8250392875851231,
+      "eval_PRM F1 Neg": 0.56,
+      "eval_PRM NPV": 0.5185185185185185,
+      "eval_PRM Precision": 0.8860759493670886,
+      "eval_PRM Recall": 0.8433734939759037,
+      "eval_PRM Specificty": 0.6086956521739131,
+      "eval_loss": 0.40065696835517883,
+      "eval_runtime": 17.0919,
+      "eval_samples_per_second": 2.574,
+      "eval_steps_per_second": 0.176,
+      "step": 2340
+    },
+    {
+      "epoch": 0.6030202530830409,
+      "grad_norm": 0.173569420588442,
+      "learning_rate": 4.0811223990685306e-05,
+      "loss": 0.38,
+      "step": 2341
+    },
+    {
+      "epoch": 0.6032778439643237,
+      "grad_norm": 0.13684583572111672,
+      "learning_rate": 4.0767023793883785e-05,
+      "loss": 0.3606,
+      "step": 2342
+    },
+    {
+      "epoch": 0.6035354348456065,
+      "grad_norm": 0.14005902767896153,
+      "learning_rate": 4.0722831065769304e-05,
+      "loss": 0.3749,
+      "step": 2343
+    },
+    {
+      "epoch": 0.6037930257268893,
+      "grad_norm": 0.12740103105685202,
+      "learning_rate": 4.067864584208997e-05,
+      "loss": 0.3653,
+      "step": 2344
+    },
+    {
+      "epoch": 0.604050616608172,
+      "grad_norm": 0.15660047142322808,
+      "learning_rate": 4.063446815858788e-05,
+      "loss": 0.3316,
+      "step": 2345
+    },
+    {
+      "epoch": 0.604050616608172,
+      "eval_PRM Accuracy": 0.7735849056603774,
+      "eval_PRM F1": 0.8536585365853658,
+      "eval_PRM F1 AUC": 0.6825563122053432,
+      "eval_PRM F1 AUC (fixed)": 0.821634363541121,
+      "eval_PRM F1 Neg": 0.5,
+      "eval_PRM NPV": 0.48,
+      "eval_PRM Precision": 0.8641975308641975,
+      "eval_PRM Recall": 0.8433734939759037,
+      "eval_PRM Specificty": 0.5217391304347826,
+      "eval_loss": 0.40731534361839294,
+      "eval_runtime": 17.0906,
+      "eval_samples_per_second": 2.575,
+      "eval_steps_per_second": 0.176,
+      "step": 2345
+    },
+    {
+      "epoch": 0.6043082074894549,
+      "grad_norm": 0.1520180262232338,
+      "learning_rate": 4.0590298050998964e-05,
+      "loss": 0.3688,
+      "step": 2346
+    },
+    {
+      "epoch": 0.6045657983707377,
+      "grad_norm": 0.1483852238797389,
+      "learning_rate": 4.05461355550531e-05,
+      "loss": 0.3624,
+      "step": 2347
+    },
+    {
+      "epoch": 0.6048233892520205,
+      "grad_norm": 0.1601064996885839,
+      "learning_rate": 4.05019807064739e-05,
+      "loss": 0.456,
+      "step": 2348
+    },
+    {
+      "epoch": 0.6050809801333032,
+      "grad_norm": 0.1775581775355606,
+      "learning_rate": 4.045783354097893e-05,
+      "loss": 0.3666,
+      "step": 2349
+    },
+    {
+      "epoch": 0.6053385710145861,
+      "grad_norm": 0.1391482916141042,
+      "learning_rate": 4.041369409427941e-05,
+      "loss": 0.3885,
+      "step": 2350
+    },
+    {
+      "epoch": 0.6053385710145861,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.8757396449704142,
+      "eval_PRM F1 AUC": 0.6849135673127291,
+      "eval_PRM F1 AUC (fixed)": 0.8268727082242011,
+      "eval_PRM F1 Neg": 0.5116279069767442,
+      "eval_PRM NPV": 0.55,
+      "eval_PRM Precision": 0.8604651162790697,
+      "eval_PRM Recall": 0.891566265060241,
+      "eval_PRM Specificty": 0.4782608695652174,
+      "eval_loss": 0.42001065611839294,
+      "eval_runtime": 17.0055,
+      "eval_samples_per_second": 2.587,
+      "eval_steps_per_second": 0.176,
+      "step": 2350
+    },
+    {
+      "epoch": 0.6055961618958688,
+      "grad_norm": 0.16933116300541795,
+      "learning_rate": 4.0369562402080384e-05,
+      "loss": 0.4159,
+      "step": 2351
+    },
+    {
+      "epoch": 0.6058537527771517,
+      "grad_norm": 0.16754478600512965,
+      "learning_rate": 4.032543850008063e-05,
+      "loss": 0.3919,
+      "step": 2352
+    },
+    {
+      "epoch": 0.6061113436584344,
+      "grad_norm": 0.19447554248718874,
+      "learning_rate": 4.028132242397258e-05,
+      "loss": 0.4139,
+      "step": 2353
+    },
+    {
+      "epoch": 0.6063689345397173,
+      "grad_norm": 0.13432879039547974,
+      "learning_rate": 4.0237214209442394e-05,
+      "loss": 0.365,
+      "step": 2354
+    },
+    {
+      "epoch": 0.6066265254210002,
+      "grad_norm": 0.15885519743295454,
+      "learning_rate": 4.0193113892169804e-05,
+      "loss": 0.3813,
+      "step": 2355
+    },
+    {
+      "epoch": 0.6066265254210002,
+      "eval_PRM Accuracy": 0.7830188679245284,
+      "eval_PRM F1": 0.8622754491017964,
+      "eval_PRM F1 AUC": 0.6728653745416449,
+      "eval_PRM F1 AUC (fixed)": 0.8218962807752749,
+      "eval_PRM F1 Neg": 0.4888888888888889,
+      "eval_PRM NPV": 0.5,
+      "eval_PRM Precision": 0.8571428571428571,
+      "eval_PRM Recall": 0.8674698795180723,
+      "eval_PRM Specificty": 0.4782608695652174,
+      "eval_loss": 0.4205433130264282,
+      "eval_runtime": 17.5093,
+      "eval_samples_per_second": 2.513,
+      "eval_steps_per_second": 0.171,
+      "step": 2355
+    },
+    {
+      "epoch": 0.6068841163022829,
+      "grad_norm": 0.16671735189095732,
+      "learning_rate": 4.0149021507828224e-05,
+      "loss": 0.4095,
+      "step": 2356
+    },
+    {
+      "epoch": 0.6071417071835657,
+      "grad_norm": 0.17378346582878013,
+      "learning_rate": 4.01049370920846e-05,
+      "loss": 0.3777,
+      "step": 2357
+    },
+    {
+      "epoch": 0.6073992980648485,
+      "grad_norm": 0.12070700699832426,
+      "learning_rate": 4.006086068059943e-05,
+      "loss": 0.3213,
+      "step": 2358
+    },
+    {
+      "epoch": 0.6076568889461313,
+      "grad_norm": 0.16891870723656224,
+      "learning_rate": 4.00167923090268e-05,
+      "loss": 0.3578,
+      "step": 2359
+    },
+    {
+      "epoch": 0.6079144798274141,
+      "grad_norm": 0.1308638140610988,
+      "learning_rate": 3.9972732013014206e-05,
+      "loss": 0.3198,
+      "step": 2360
+    },
+    {
+      "epoch": 0.6079144798274141,
+      "eval_PRM Accuracy": 0.7735849056603774,
+      "eval_PRM F1": 0.8536585365853658,
+      "eval_PRM F1 AUC": 0.6825563122053432,
+      "eval_PRM F1 AUC (fixed)": 0.820848611838659,
+      "eval_PRM F1 Neg": 0.5,
+      "eval_PRM NPV": 0.48,
+      "eval_PRM Precision": 0.8641975308641975,
+      "eval_PRM Recall": 0.8433734939759037,
+      "eval_PRM Specificty": 0.5217391304347826,
+      "eval_loss": 0.4169921875,
+      "eval_runtime": 17.5358,
+      "eval_samples_per_second": 2.509,
+      "eval_steps_per_second": 0.171,
+      "step": 2360
+    },
+    {
+      "epoch": 0.6081720707086969,
+      "grad_norm": 0.1317729555391042,
+      "learning_rate": 3.992867982820268e-05,
+      "loss": 0.3508,
+      "step": 2361
+    },
+    {
+      "epoch": 0.6084296615899797,
+      "grad_norm": 0.1767411732431218,
+      "learning_rate": 3.988463579022666e-05,
+      "loss": 0.4106,
+      "step": 2362
+    },
+    {
+      "epoch": 0.6086872524712625,
+      "grad_norm": 0.18155428167496843,
+      "learning_rate": 3.984059993471399e-05,
+      "loss": 0.4019,
+      "step": 2363
+    },
+    {
+      "epoch": 0.6089448433525453,
+      "grad_norm": 0.19324560638494176,
+      "learning_rate": 3.9796572297285926e-05,
+      "loss": 0.51,
+      "step": 2364
+    },
+    {
+      "epoch": 0.6092024342338281,
+      "grad_norm": 0.17087083249532514,
+      "learning_rate": 3.9752552913557036e-05,
+      "loss": 0.367,
+      "step": 2365
+    },
+    {
+      "epoch": 0.6092024342338281,
+      "eval_PRM Accuracy": 0.7735849056603774,
+      "eval_PRM F1": 0.8536585365853658,
+      "eval_PRM F1 AUC": 0.6825563122053432,
+      "eval_PRM F1 AUC (fixed)": 0.817967522262965,
+      "eval_PRM F1 Neg": 0.5,
+      "eval_PRM NPV": 0.48,
+      "eval_PRM Precision": 0.8641975308641975,
+      "eval_PRM Recall": 0.8433734939759037,
+      "eval_PRM Specificty": 0.5217391304347826,
+      "eval_loss": 0.41432884335517883,
+      "eval_runtime": 17.0782,
+      "eval_samples_per_second": 2.576,
+      "eval_steps_per_second": 0.176,
+      "step": 2365
+    },
+    {
+      "epoch": 0.6094600251151109,
+      "grad_norm": 0.15022774199433786,
+      "learning_rate": 3.970854181913526e-05,
+      "loss": 0.4312,
+      "step": 2366
+    },
+    {
+      "epoch": 0.6097176159963937,
+      "grad_norm": 0.14147350650677862,
+      "learning_rate": 3.966453904962176e-05,
+      "loss": 0.2885,
+      "step": 2367
+    },
+    {
+      "epoch": 0.6099752068776765,
+      "grad_norm": 0.1900434964583597,
+      "learning_rate": 3.962054464061106e-05,
+      "loss": 0.4551,
+      "step": 2368
+    },
+    {
+      "epoch": 0.6102327977589593,
+      "grad_norm": 0.14243337413848095,
+      "learning_rate": 3.9576558627690826e-05,
+      "loss": 0.3873,
+      "step": 2369
+    },
+    {
+      "epoch": 0.6104903886402422,
+      "grad_norm": 0.14661969644213976,
+      "learning_rate": 3.9532581046442e-05,
+      "loss": 0.2973,
+      "step": 2370
+    },
+    {
+      "epoch": 0.6104903886402422,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.874251497005988,
+      "eval_PRM F1 AUC": 0.7006286013619696,
+      "eval_PRM F1 AUC (fixed)": 0.8260869565217391,
+      "eval_PRM F1 Neg": 0.5333333333333333,
+      "eval_PRM NPV": 0.5454545454545454,
+      "eval_PRM Precision": 0.8690476190476191,
+      "eval_PRM Recall": 0.8795180722891566,
+      "eval_PRM Specificty": 0.5217391304347826,
+      "eval_loss": 0.40731534361839294,
+      "eval_runtime": 17.0664,
+      "eval_samples_per_second": 2.578,
+      "eval_steps_per_second": 0.176,
+      "step": 2370
+    },
+    {
+      "epoch": 0.6107479795215249,
+      "grad_norm": 0.11402039346775396,
+      "learning_rate": 3.9488611932438666e-05,
+      "loss": 0.3182,
+      "step": 2371
+    },
+    {
+      "epoch": 0.6110055704028078,
+      "grad_norm": 0.15915651498180694,
+      "learning_rate": 3.944465132124807e-05,
+      "loss": 0.3995,
+      "step": 2372
+    },
+    {
+      "epoch": 0.6112631612840905,
+      "grad_norm": 0.13158790616500857,
+      "learning_rate": 3.94006992484306e-05,
+      "loss": 0.2991,
+      "step": 2373
+    },
+    {
+      "epoch": 0.6115207521653734,
+      "grad_norm": 0.11553152695883866,
+      "learning_rate": 3.93567557495397e-05,
+      "loss": 0.3199,
+      "step": 2374
+    },
+    {
+      "epoch": 0.6117783430466561,
+      "grad_norm": 0.20815237796381744,
+      "learning_rate": 3.931282086012191e-05,
+      "loss": 0.4048,
+      "step": 2375
+    },
+    {
+      "epoch": 0.6117783430466561,
+      "eval_PRM Accuracy": 0.8018867924528302,
+      "eval_PRM F1": 0.8757396449704142,
+      "eval_PRM F1 AUC": 0.6849135673127291,
+      "eval_PRM F1 AUC (fixed)": 0.820062860136197,
+      "eval_PRM F1 Neg": 0.5116279069767442,
+      "eval_PRM NPV": 0.55,
+      "eval_PRM Precision": 0.8604651162790697,
+      "eval_PRM Recall": 0.891566265060241,
+      "eval_PRM Specificty": 0.4782608695652174,
+      "eval_loss": 0.4189896881580353,
+      "eval_runtime": 17.2565,
+      "eval_samples_per_second": 2.55,
+      "eval_steps_per_second": 0.174,
+      "step": 2375
+    },
+    {
+      "epoch": 0.612035933927939,
+      "grad_norm": 0.11404295721713555,
+      "learning_rate": 3.9268894615716804e-05,
+      "loss": 0.2881,
+      "step": 2376
+    },
+    {
+      "epoch": 0.6122935248092217,
+      "grad_norm": 0.1531556969722233,
+      "learning_rate": 3.9224977051856904e-05,
+      "loss": 0.446,
+      "step": 2377
+    },
+    {
+      "epoch": 0.6125511156905046,
+      "grad_norm": 0.13607225651561186,
+      "learning_rate": 3.918106820406782e-05,
+      "loss": 0.2904,
+      "step": 2378
+    },
+    {
+      "epoch": 0.6128087065717873,
+      "grad_norm": 0.1830553290422174,
+      "learning_rate": 3.9137168107867994e-05,
+      "loss": 0.4065,
+      "step": 2379
+    },
+    {
+      "epoch": 0.6130662974530702,
+      "grad_norm": 0.1357494360307605,
+      "learning_rate": 3.909327679876888e-05,
+      "loss": 0.3653,
+      "step": 2380
+    },
+    {
+      "epoch": 0.6130662974530702,
+      "eval_PRM Accuracy": 0.8113207547169812,
+      "eval_PRM F1": 0.8823529411764706,
+      "eval_PRM F1 AUC": 0.6909376636982714,
+      "eval_PRM F1 AUC (fixed)": 0.821634363541121,
+      "eval_PRM F1 Neg": 0.5238095238095238,
+      "eval_PRM NPV": 0.5789473684210527,
+      "eval_PRM Precision": 0.8620689655172413,
+      "eval_PRM Recall": 0.9036144578313253,
+      "eval_PRM Specificty": 0.4782608695652174,
+      "eval_loss": 0.41317471861839294,
+      "eval_runtime": 16.8684,
+      "eval_samples_per_second": 2.608,
+      "eval_steps_per_second": 0.178,
+      "step": 2380
+    },
+    {
+      "epoch": 0.6133238883343529,
+      "grad_norm": 0.19432163887414694,
+      "learning_rate": 3.9049394312274765e-05,
+      "loss": 0.3545,
+      "step": 2381
+    },
+    {
+      "epoch": 0.6135814792156358,
+      "grad_norm": 0.143143247958846,
+      "learning_rate": 3.9005520683882854e-05,
+      "loss": 0.3856,
+      "step": 2382
+    },
+    {
+      "epoch": 0.6138390700969185,
+      "grad_norm": 0.13596461157727238,
+      "learning_rate": 3.896165594908313e-05,
+      "loss": 0.3579,
+      "step": 2383
+    },
+    {
+      "epoch": 0.6140966609782014,
+      "grad_norm": 0.18731344597012275,
+      "learning_rate": 3.8917800143358404e-05,
+      "loss": 0.4522,
+      "step": 2384
+    },
+    {
+      "epoch": 0.6143542518594842,
+      "grad_norm": 0.19021340220262192,
+      "learning_rate": 3.887395330218429e-05,
+      "loss": 0.3945,
+      "step": 2385
+    },
+    {
+      "epoch": 0.6143542518594842,
+      "eval_PRM Accuracy": 0.8207547169811321,
+      "eval_PRM F1": 0.8848484848484849,
+      "eval_PRM F1 AUC": 0.7441068622315348,
+      "eval_PRM F1 AUC (fixed)": 0.8205866946045051,
+      "eval_PRM F1 Neg": 0.5957446808510638,
+      "eval_PRM NPV": 0.5833333333333334,
+      "eval_PRM Precision": 0.8902439024390244,
+      "eval_PRM Recall": 0.8795180722891566,
+      "eval_PRM Specificty": 0.6086956521739131,
+      "eval_loss": 0.3953302502632141,
+      "eval_runtime": 17.6129,
+      "eval_samples_per_second": 2.498,
+      "eval_steps_per_second": 0.17,
+      "step": 2385
+    },
+    {
+      "epoch": 0.614611842740767,
+      "grad_norm": 0.1611457927631445,
+      "learning_rate": 3.883011546102909e-05,
+      "loss": 0.3255,
+      "step": 2386
+    },
+    {
+      "epoch": 0.6148694336220498,
+      "grad_norm": 0.17976558652728208,
+      "learning_rate": 3.878628665535392e-05,
+      "loss": 0.3647,
+      "step": 2387
+    },
+    {
+      "epoch": 0.6151270245033326,
+      "grad_norm": 0.18633645755451828,
+      "learning_rate": 3.874246692061246e-05,
+      "loss": 0.3515,
+      "step": 2388
+    },
+    {
+      "epoch": 0.6153846153846154,
+      "grad_norm": 0.1476256519035803,
+      "learning_rate": 3.869865629225119e-05,
+      "loss": 0.3342,
+      "step": 2389
+    },
+    {
+      "epoch": 0.6156422062658982,
+      "grad_norm": 0.13511375655045682,
+      "learning_rate": 3.865485480570912e-05,
+      "loss": 0.3907,
+      "step": 2390
+    },
+    {
+      "epoch": 0.6156422062658982,
+      "eval_PRM Accuracy": 0.8113207547169812,
+      "eval_PRM F1": 0.8809523809523809,
+      "eval_PRM F1 AUC": 0.7066526977475118,
+      "eval_PRM F1 AUC (fixed)": 0.8190151911995809,
+      "eval_PRM F1 Neg": 0.5454545454545454,
+      "eval_PRM NPV": 0.5714285714285714,
+      "eval_PRM Precision": 0.8705882352941177,
+      "eval_PRM Recall": 0.891566265060241,
+      "eval_PRM Specificty": 0.5217391304347826,
+      "eval_loss": 0.40305396914482117,
+      "eval_runtime": 16.8874,
+      "eval_samples_per_second": 2.605,
+      "eval_steps_per_second": 0.178,
+      "step": 2390
+    },
+    {
+      "epoch": 0.615899797147181,
+      "grad_norm": 0.1454825346465254,
+      "learning_rate": 3.861106249641789e-05,
+      "loss": 0.3343,
+      "step": 2391
+    },
+    {
+      "epoch": 0.6161573880284638,
+      "grad_norm": 0.15286212775344593,
+      "learning_rate": 3.856727939980176e-05,
+      "loss": 0.4346,
+      "step": 2392
+    },
+    {
+      "epoch": 0.6164149789097466,
+      "grad_norm": 0.13970371113977845,
+      "learning_rate": 3.852350555127748e-05,
+      "loss": 0.3177,
+      "step": 2393
+    },
+    {
+      "epoch": 0.6166725697910294,
+      "grad_norm": 0.19705685890823973,
+      "learning_rate": 3.8479740986254365e-05,
+      "loss": 0.356,
+      "step": 2394
+    },
+    {
+      "epoch": 0.6169301606723122,
+      "grad_norm": 0.2535529334088201,
+      "learning_rate": 3.843598574013418e-05,
+      "loss": 0.4189,
+      "step": 2395
+    },
+    {
+      "epoch": 0.6169301606723122,
+      "eval_PRM Accuracy": 0.8207547169811321,
+      "eval_PRM F1": 0.8888888888888888,
+      "eval_PRM F1 AUC": 0.6969617600838135,
+      "eval_PRM F1 AUC (fixed)": 0.8276584599266632,
+      "eval_PRM F1 Neg": 0.5365853658536586,
+      "eval_PRM NPV": 0.6111111111111112,
+      "eval_PRM Precision": 0.8636363636363636,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.4782608695652174,
+      "eval_loss": 0.412109375,
+      "eval_runtime": 17.4744,
+      "eval_samples_per_second": 2.518,
+      "eval_steps_per_second": 0.172,
+      "step": 2395
+    },
+    {
+      "epoch": 0.617187751553595,
+      "grad_norm": 0.18794186191451187,
+      "learning_rate": 3.839223984831119e-05,
+      "loss": 0.4021,
+      "step": 2396
+    },
+    {
+      "epoch": 0.6174453424348778,
+      "grad_norm": 0.14441165243612405,
+      "learning_rate": 3.8348503346172074e-05,
+      "loss": 0.3452,
+      "step": 2397
+    },
+    {
+      "epoch": 0.6177029333161607,
+      "grad_norm": 0.23925161516194815,
+      "learning_rate": 3.830477626909589e-05,
+      "loss": 0.4133,
+      "step": 2398
+    },
+    {
+      "epoch": 0.6179605241974434,
+      "grad_norm": 0.17828211742232414,
+      "learning_rate": 3.826105865245414e-05,
+      "loss": 0.3388,
+      "step": 2399
+    },
+    {
+      "epoch": 0.6182181150787263,
+      "grad_norm": 0.13630425549357833,
+      "learning_rate": 3.82173505316106e-05,
+      "loss": 0.3799,
+      "step": 2400
+    },
+    {
+      "epoch": 0.6182181150787263,
+      "eval_PRM Accuracy": 0.8207547169811321,
+      "eval_PRM F1": 0.8901734104046243,
+      "eval_PRM F1 AUC": 0.6812467260345731,
+      "eval_PRM F1 AUC (fixed)": 0.8389209009952855,
+      "eval_PRM F1 Neg": 0.5128205128205128,
+      "eval_PRM NPV": 0.625,
+      "eval_PRM Precision": 0.8555555555555555,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.43478260869565216,
+      "eval_loss": 0.4106889069080353,
+      "eval_runtime": 16.5579,
+      "eval_samples_per_second": 2.657,
+      "eval_steps_per_second": 0.181,
+      "step": 2400
+    },
+    {
+      "epoch": 0.618475705960009,
+      "grad_norm": 0.13869335502670038,
+      "learning_rate": 3.81736519419214e-05,
+      "loss": 0.346,
+      "step": 2401
+    },
+    {
+      "epoch": 0.6187332968412919,
+      "grad_norm": 0.21401245288632476,
+      "learning_rate": 3.812996291873496e-05,
+      "loss": 0.4715,
+      "step": 2402
+    },
+    {
+      "epoch": 0.6189908877225746,
+      "grad_norm": 0.1699494837457687,
+      "learning_rate": 3.808628349739198e-05,
+      "loss": 0.3264,
+      "step": 2403
+    },
+    {
+      "epoch": 0.6192484786038575,
+      "grad_norm": 0.12204448338944414,
+      "learning_rate": 3.8042613713225346e-05,
+      "loss": 0.3131,
+      "step": 2404
+    },
+    {
+      "epoch": 0.6195060694851402,
+      "grad_norm": 0.1295816329473272,
+      "learning_rate": 3.7998953601560175e-05,
+      "loss": 0.3065,
+      "step": 2405
+    },
+    {
+      "epoch": 0.6195060694851402,
+      "eval_PRM Accuracy": 0.8207547169811321,
+      "eval_PRM F1": 0.8888888888888888,
+      "eval_PRM F1 AUC": 0.6969617600838135,
+      "eval_PRM F1 AUC (fixed)": 0.8381351492928236,
+      "eval_PRM F1 Neg": 0.5365853658536586,
+      "eval_PRM NPV": 0.6111111111111112,
+      "eval_PRM Precision": 0.8636363636363636,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.4782608695652174,
+      "eval_loss": 0.3964399993419647,
+      "eval_runtime": 17.0733,
+      "eval_samples_per_second": 2.577,
+      "eval_steps_per_second": 0.176,
+      "step": 2405
+    },
+    {
+      "epoch": 0.619763660366423,
+      "grad_norm": 0.12782589842749065,
+      "learning_rate": 3.7955303197713774e-05,
+      "loss": 0.2942,
+      "step": 2406
+    },
+    {
+      "epoch": 0.6200212512477058,
+      "grad_norm": 0.22347031911550472,
+      "learning_rate": 3.791166253699555e-05,
+      "loss": 0.3996,
+      "step": 2407
+    },
+    {
+      "epoch": 0.6202788421289887,
+      "grad_norm": 0.15465719210193768,
+      "learning_rate": 3.7868031654707105e-05,
+      "loss": 0.3739,
+      "step": 2408
+    },
+    {
+      "epoch": 0.6205364330102714,
+      "grad_norm": 0.15084946354307485,
+      "learning_rate": 3.782441058614203e-05,
+      "loss": 0.3447,
+      "step": 2409
+    },
+    {
+      "epoch": 0.6207940238915542,
+      "grad_norm": 0.14473470469746796,
+      "learning_rate": 3.778079936658611e-05,
+      "loss": 0.3595,
+      "step": 2410
+    },
+    {
+      "epoch": 0.6207940238915542,
+      "eval_PRM Accuracy": 0.839622641509434,
+      "eval_PRM F1": 0.8982035928143712,
+      "eval_PRM F1 AUC": 0.7561550550026193,
+      "eval_PRM F1 AUC (fixed)": 0.8273965426925091,
+      "eval_PRM F1 Neg": 0.6222222222222222,
+      "eval_PRM NPV": 0.6363636363636364,
+      "eval_PRM Precision": 0.8928571428571429,
+      "eval_PRM Recall": 0.9036144578313253,
+      "eval_PRM Specificty": 0.6086956521739131,
+      "eval_loss": 0.3949751555919647,
+      "eval_runtime": 17.4771,
+      "eval_samples_per_second": 2.518,
+      "eval_steps_per_second": 0.172,
+      "step": 2410
+    },
+    {
+      "epoch": 0.621051614772837,
+      "grad_norm": 0.19841184109055246,
+      "learning_rate": 3.773719803131702e-05,
+      "loss": 0.2876,
+      "step": 2411
+    },
+    {
+      "epoch": 0.6213092056541198,
+      "grad_norm": 0.18140941867294844,
+      "learning_rate": 3.769360661560453e-05,
+      "loss": 0.3564,
+      "step": 2412
+    },
+    {
+      "epoch": 0.6215667965354027,
+      "grad_norm": 0.14336788233772046,
+      "learning_rate": 3.765002515471035e-05,
+      "loss": 0.38,
+      "step": 2413
+    },
+    {
+      "epoch": 0.6218243874166854,
+      "grad_norm": 0.16433034958269993,
+      "learning_rate": 3.760645368388817e-05,
+      "loss": 0.4335,
+      "step": 2414
+    },
+    {
+      "epoch": 0.6220819782979683,
+      "grad_norm": 0.130249698037053,
+      "learning_rate": 3.7562892238383576e-05,
+      "loss": 0.3519,
+      "step": 2415
+    },
+    {
+      "epoch": 0.6220819782979683,
+      "eval_PRM Accuracy": 0.8490566037735849,
+      "eval_PRM F1": 0.9036144578313253,
+      "eval_PRM F1 AUC": 0.7778941854374019,
+      "eval_PRM F1 AUC (fixed)": 0.8349921424829754,
+      "eval_PRM F1 Neg": 0.6521739130434783,
+      "eval_PRM NPV": 0.6521739130434783,
+      "eval_PRM Precision": 0.9036144578313253,
+      "eval_PRM Recall": 0.9036144578313253,
+      "eval_PRM Specificty": 0.6521739130434783,
+      "eval_loss": 0.38991478085517883,
+      "eval_runtime": 16.4684,
+      "eval_samples_per_second": 2.672,
+      "eval_steps_per_second": 0.182,
+      "step": 2415
+    },
+    {
+      "epoch": 0.622339569179251,
+      "grad_norm": 0.1691572313466575,
+      "learning_rate": 3.7519340853434044e-05,
+      "loss": 0.385,
+      "step": 2416
+    },
+    {
+      "epoch": 0.6225971600605339,
+      "grad_norm": 0.3209029787796321,
+      "learning_rate": 3.747579956426888e-05,
+      "loss": 0.483,
+      "step": 2417
+    },
+    {
+      "epoch": 0.6228547509418166,
+      "grad_norm": 0.14144139932379493,
+      "learning_rate": 3.7432268406109326e-05,
+      "loss": 0.3275,
+      "step": 2418
+    },
+    {
+      "epoch": 0.6231123418230995,
+      "grad_norm": 0.17592595116774362,
+      "learning_rate": 3.73887474141683e-05,
+      "loss": 0.355,
+      "step": 2419
+    },
+    {
+      "epoch": 0.6233699327043822,
+      "grad_norm": 0.1516450321416382,
+      "learning_rate": 3.734523662365059e-05,
+      "loss": 0.417,
+      "step": 2420
+    },
+    {
+      "epoch": 0.6233699327043822,
+      "eval_PRM Accuracy": 0.8490566037735849,
+      "eval_PRM F1": 0.9036144578313253,
+      "eval_PRM F1 AUC": 0.7778941854374019,
+      "eval_PRM F1 AUC (fixed)": 0.8376113148245155,
+      "eval_PRM F1 Neg": 0.6521739130434783,
+      "eval_PRM NPV": 0.6521739130434783,
+      "eval_PRM Precision": 0.9036144578313253,
+      "eval_PRM Recall": 0.9036144578313253,
+      "eval_PRM Specificty": 0.6521739130434783,
+      "eval_loss": 0.37863990664482117,
+      "eval_runtime": 17.5193,
+      "eval_samples_per_second": 2.512,
+      "eval_steps_per_second": 0.171,
+      "step": 2420
+    },
+    {
+      "epoch": 0.6236275235856651,
+      "grad_norm": 0.19136966587799933,
+      "learning_rate": 3.730173606975268e-05,
+      "loss": 0.4759,
+      "step": 2421
+    },
+    {
+      "epoch": 0.6238851144669478,
+      "grad_norm": 0.16603655661634564,
+      "learning_rate": 3.725824578766281e-05,
+      "loss": 0.355,
+      "step": 2422
+    },
+    {
+      "epoch": 0.6241427053482307,
+      "grad_norm": 0.2053708825957282,
+      "learning_rate": 3.721476581256089e-05,
+      "loss": 0.3731,
+      "step": 2423
+    },
+    {
+      "epoch": 0.6244002962295134,
+      "grad_norm": 0.18248176347253492,
+      "learning_rate": 3.7171296179618476e-05,
+      "loss": 0.34,
+      "step": 2424
+    },
+    {
+      "epoch": 0.6246578871107963,
+      "grad_norm": 0.12733024912347352,
+      "learning_rate": 3.712783692399881e-05,
+      "loss": 0.3072,
+      "step": 2425
+    },
+    {
+      "epoch": 0.6246578871107963,
+      "eval_PRM Accuracy": 0.839622641509434,
+      "eval_PRM F1": 0.896969696969697,
+      "eval_PRM F1 AUC": 0.7718700890518595,
+      "eval_PRM F1 AUC (fixed)": 0.8404924044002094,
+      "eval_PRM F1 Neg": 0.6382978723404256,
+      "eval_PRM NPV": 0.625,
+      "eval_PRM Precision": 0.9024390243902439,
+      "eval_PRM Recall": 0.891566265060241,
+      "eval_PRM Specificty": 0.6521739130434783,
+      "eval_loss": 0.3765092194080353,
+      "eval_runtime": 17.0684,
+      "eval_samples_per_second": 2.578,
+      "eval_steps_per_second": 0.176,
+      "step": 2425
+    },
+    {
+      "epoch": 0.624915477992079,
+      "grad_norm": 0.1621235773141948,
+      "learning_rate": 3.708438808085668e-05,
+      "loss": 0.2946,
+      "step": 2426
+    },
+    {
+      "epoch": 0.6251730688733619,
+      "grad_norm": 0.15820404950724934,
+      "learning_rate": 3.704094968533852e-05,
+      "loss": 0.3633,
+      "step": 2427
+    },
+    {
+      "epoch": 0.6254306597546447,
+      "grad_norm": 0.15525345620753025,
+      "learning_rate": 3.69975217725822e-05,
+      "loss": 0.3775,
+      "step": 2428
+    },
+    {
+      "epoch": 0.6256882506359275,
+      "grad_norm": 0.13987459511490624,
+      "learning_rate": 3.695410437771727e-05,
+      "loss": 0.313,
+      "step": 2429
+    },
+    {
+      "epoch": 0.6259458415172103,
+      "grad_norm": 0.16656774734893298,
+      "learning_rate": 3.691069753586462e-05,
+      "loss": 0.3671,
+      "step": 2430
+    },
+    {
+      "epoch": 0.6259458415172103,
+      "eval_PRM Accuracy": 0.839622641509434,
+      "eval_PRM F1": 0.8994082840236687,
+      "eval_PRM F1 AUC": 0.7404400209533788,
+      "eval_PRM F1 AUC (fixed)": 0.8459926663174437,
+      "eval_PRM F1 Neg": 0.6046511627906976,
+      "eval_PRM NPV": 0.65,
+      "eval_PRM Precision": 0.8837209302325582,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.3775745630264282,
+      "eval_runtime": 17.0016,
+      "eval_samples_per_second": 2.588,
+      "eval_steps_per_second": 0.176,
+      "step": 2430
+    },
+    {
+      "epoch": 0.6262034323984931,
+      "grad_norm": 0.12236490588156322,
+      "learning_rate": 3.686730128213668e-05,
+      "loss": 0.2818,
+      "step": 2431
+    },
+    {
+      "epoch": 0.6264610232797759,
+      "grad_norm": 0.1527398652229691,
+      "learning_rate": 3.682391565163732e-05,
+      "loss": 0.3281,
+      "step": 2432
+    },
+    {
+      "epoch": 0.6267186141610587,
+      "grad_norm": 0.15293359843564117,
+      "learning_rate": 3.6780540679461784e-05,
+      "loss": 0.3706,
+      "step": 2433
+    },
+    {
+      "epoch": 0.6269762050423415,
+      "grad_norm": 0.16229893172953397,
+      "learning_rate": 3.673717640069672e-05,
+      "loss": 0.3909,
+      "step": 2434
+    },
+    {
+      "epoch": 0.6272337959236243,
+      "grad_norm": 0.14385377128288312,
+      "learning_rate": 3.669382285042011e-05,
+      "loss": 0.2975,
+      "step": 2435
+    },
+    {
+      "epoch": 0.6272337959236243,
+      "eval_PRM Accuracy": 0.839622641509434,
+      "eval_PRM F1": 0.8994082840236687,
+      "eval_PRM F1 AUC": 0.7404400209533788,
+      "eval_PRM F1 AUC (fixed)": 0.8483499214248297,
+      "eval_PRM F1 Neg": 0.6046511627906976,
+      "eval_PRM NPV": 0.65,
+      "eval_PRM Precision": 0.8837209302325582,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.3722034692764282,
+      "eval_runtime": 16.9658,
+      "eval_samples_per_second": 2.593,
+      "eval_steps_per_second": 0.177,
+      "step": 2435
+    },
+    {
+      "epoch": 0.6274913868049071,
+      "grad_norm": 0.14195669348417514,
+      "learning_rate": 3.6650480063701265e-05,
+      "loss": 0.3913,
+      "step": 2436
+    },
+    {
+      "epoch": 0.6277489776861899,
+      "grad_norm": 0.1711043094421518,
+      "learning_rate": 3.66071480756008e-05,
+      "loss": 0.4659,
+      "step": 2437
+    },
+    {
+      "epoch": 0.6280065685674727,
+      "grad_norm": 0.15588465134278923,
+      "learning_rate": 3.6563826921170544e-05,
+      "loss": 0.3509,
+      "step": 2438
+    },
+    {
+      "epoch": 0.6282641594487555,
+      "grad_norm": 0.18307559032814893,
+      "learning_rate": 3.652051663545366e-05,
+      "loss": 0.4066,
+      "step": 2439
+    },
+    {
+      "epoch": 0.6285217503300383,
+      "grad_norm": 0.16824615853320427,
+      "learning_rate": 3.64772172534844e-05,
+      "loss": 0.2902,
+      "step": 2440
+    },
+    {
+      "epoch": 0.6285217503300383,
+      "eval_PRM Accuracy": 0.839622641509434,
+      "eval_PRM F1": 0.8994082840236687,
+      "eval_PRM F1 AUC": 0.7404400209533788,
+      "eval_PRM F1 AUC (fixed)": 0.8504452592980618,
+      "eval_PRM F1 Neg": 0.6046511627906976,
+      "eval_PRM NPV": 0.65,
+      "eval_PRM Precision": 0.8837209302325582,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.3701171875,
+      "eval_runtime": 17.5261,
+      "eval_samples_per_second": 2.511,
+      "eval_steps_per_second": 0.171,
+      "step": 2440
+    },
+    {
+      "epoch": 0.6287793412113212,
+      "grad_norm": 0.15941201118061468,
+      "learning_rate": 3.643392881028829e-05,
+      "loss": 0.384,
+      "step": 2441
+    },
+    {
+      "epoch": 0.6290369320926039,
+      "grad_norm": 0.1481415780534842,
+      "learning_rate": 3.639065134088196e-05,
+      "loss": 0.3438,
+      "step": 2442
+    },
+    {
+      "epoch": 0.6292945229738868,
+      "grad_norm": 0.13971340614909494,
+      "learning_rate": 3.63473848802732e-05,
+      "loss": 0.3288,
+      "step": 2443
+    },
+    {
+      "epoch": 0.6295521138551695,
+      "grad_norm": 0.19468720322757044,
+      "learning_rate": 3.630412946346084e-05,
+      "loss": 0.4081,
+      "step": 2444
+    },
+    {
+      "epoch": 0.6298097047364524,
+      "grad_norm": 0.14912771442970982,
+      "learning_rate": 3.6260885125434815e-05,
+      "loss": 0.3622,
+      "step": 2445
+    },
+    {
+      "epoch": 0.6298097047364524,
+      "eval_PRM Accuracy": 0.8301886792452831,
+      "eval_PRM F1": 0.8941176470588236,
+      "eval_PRM F1 AUC": 0.7187008905185961,
+      "eval_PRM F1 AUC (fixed)": 0.8493975903614458,
+      "eval_PRM F1 Neg": 0.5714285714285714,
+      "eval_PRM NPV": 0.631578947368421,
+      "eval_PRM Precision": 0.8735632183908046,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.5217391304347826,
+      "eval_loss": 0.3699840307235718,
+      "eval_runtime": 16.9876,
+      "eval_samples_per_second": 2.59,
+      "eval_steps_per_second": 0.177,
+      "step": 2445
+    },
+    {
+      "epoch": 0.6300672956177351,
+      "grad_norm": 0.19585779447466803,
+      "learning_rate": 3.621765190117612e-05,
+      "loss": 0.4418,
+      "step": 2446
+    },
+    {
+      "epoch": 0.630324886499018,
+      "grad_norm": 0.15398334378027992,
+      "learning_rate": 3.6174429825656685e-05,
+      "loss": 0.2998,
+      "step": 2447
+    },
+    {
+      "epoch": 0.6305824773803007,
+      "grad_norm": 0.18073986802135503,
+      "learning_rate": 3.613121893383952e-05,
+      "loss": 0.4453,
+      "step": 2448
+    },
+    {
+      "epoch": 0.6308400682615836,
+      "grad_norm": 0.21775943023545297,
+      "learning_rate": 3.6088019260678465e-05,
+      "loss": 0.4535,
+      "step": 2449
+    },
+    {
+      "epoch": 0.6310976591428663,
+      "grad_norm": 0.19220132536823256,
+      "learning_rate": 3.6044830841118434e-05,
+      "loss": 0.285,
+      "step": 2450
+    },
+    {
+      "epoch": 0.6310976591428663,
+      "eval_PRM Accuracy": 0.839622641509434,
+      "eval_PRM F1": 0.8994082840236687,
+      "eval_PRM F1 AUC": 0.7404400209533788,
+      "eval_PRM F1 AUC (fixed)": 0.855683603981142,
+      "eval_PRM F1 Neg": 0.6046511627906976,
+      "eval_PRM NPV": 0.65,
+      "eval_PRM Precision": 0.8837209302325582,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.3622603118419647,
+      "eval_runtime": 17.0728,
+      "eval_samples_per_second": 2.577,
+      "eval_steps_per_second": 0.176,
+      "step": 2450
+    },
+    {
+      "epoch": 0.6313552500241492,
+      "grad_norm": 0.12850460632582075,
+      "learning_rate": 3.600165371009511e-05,
+      "loss": 0.2786,
+      "step": 2451
+    },
+    {
+      "epoch": 0.6316128409054319,
+      "grad_norm": 0.17613917177631971,
+      "learning_rate": 3.59584879025351e-05,
+      "loss": 0.4074,
+      "step": 2452
+    },
+    {
+      "epoch": 0.6318704317867148,
+      "grad_norm": 0.14661266019960154,
+      "learning_rate": 3.591533345335587e-05,
+      "loss": 0.3353,
+      "step": 2453
+    },
+    {
+      "epoch": 0.6321280226679975,
+      "grad_norm": 0.13134303722339521,
+      "learning_rate": 3.587219039746564e-05,
+      "loss": 0.3593,
+      "step": 2454
+    },
+    {
+      "epoch": 0.6323856135492804,
+      "grad_norm": 0.18462412448019225,
+      "learning_rate": 3.5829058769763475e-05,
+      "loss": 0.3275,
+      "step": 2455
+    },
+    {
+      "epoch": 0.6323856135492804,
+      "eval_PRM Accuracy": 0.8301886792452831,
+      "eval_PRM F1": 0.8941176470588236,
+      "eval_PRM F1 AUC": 0.7187008905185961,
+      "eval_PRM F1 AUC (fixed)": 0.858040859088528,
+      "eval_PRM F1 Neg": 0.5714285714285714,
+      "eval_PRM NPV": 0.631578947368421,
+      "eval_PRM Precision": 0.8735632183908046,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.5217391304347826,
+      "eval_loss": 0.36581143736839294,
+      "eval_runtime": 16.937,
+      "eval_samples_per_second": 2.598,
+      "eval_steps_per_second": 0.177,
+      "step": 2455
+    },
+    {
+      "epoch": 0.6326432044305632,
+      "grad_norm": 0.17257292064812652,
+      "learning_rate": 3.5785938605139145e-05,
+      "loss": 0.3864,
+      "step": 2456
+    },
+    {
+      "epoch": 0.632900795311846,
+      "grad_norm": 0.1569296924716655,
+      "learning_rate": 3.57428299384732e-05,
+      "loss": 0.4272,
+      "step": 2457
+    },
+    {
+      "epoch": 0.6331583861931288,
+      "grad_norm": 0.13782728055888355,
+      "learning_rate": 3.569973280463684e-05,
+      "loss": 0.333,
+      "step": 2458
+    },
+    {
+      "epoch": 0.6334159770744116,
+      "grad_norm": 0.2277672956516817,
+      "learning_rate": 3.565664723849195e-05,
+      "loss": 0.439,
+      "step": 2459
+    },
+    {
+      "epoch": 0.6336735679556944,
+      "grad_norm": 0.16675031645959432,
+      "learning_rate": 3.561357327489108e-05,
+      "loss": 0.2939,
+      "step": 2460
+    },
+    {
+      "epoch": 0.6336735679556944,
+      "eval_PRM Accuracy": 0.8301886792452831,
+      "eval_PRM F1": 0.8941176470588236,
+      "eval_PRM F1 AUC": 0.7187008905185961,
+      "eval_PRM F1 AUC (fixed)": 0.8546359350445258,
+      "eval_PRM F1 Neg": 0.5714285714285714,
+      "eval_PRM NPV": 0.631578947368421,
+      "eval_PRM Precision": 0.8735632183908046,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.5217391304347826,
+      "eval_loss": 0.3665216565132141,
+      "eval_runtime": 17.0327,
+      "eval_samples_per_second": 2.583,
+      "eval_steps_per_second": 0.176,
+      "step": 2460
+    },
+    {
+      "epoch": 0.6339311588369771,
+      "grad_norm": 0.12459223290713027,
+      "learning_rate": 3.557051094867735e-05,
+      "loss": 0.2948,
+      "step": 2461
+    },
+    {
+      "epoch": 0.63418874971826,
+      "grad_norm": 0.1423469842293259,
+      "learning_rate": 3.552746029468452e-05,
+      "loss": 0.3972,
+      "step": 2462
+    },
+    {
+      "epoch": 0.6344463405995427,
+      "grad_norm": 0.1843782731483223,
+      "learning_rate": 3.5484421347736874e-05,
+      "loss": 0.3366,
+      "step": 2463
+    },
+    {
+      "epoch": 0.6347039314808256,
+      "grad_norm": 0.15991191460921128,
+      "learning_rate": 3.5441394142649245e-05,
+      "loss": 0.3603,
+      "step": 2464
+    },
+    {
+      "epoch": 0.6349615223621083,
+      "grad_norm": 0.16679660401059462,
+      "learning_rate": 3.5398378714226946e-05,
+      "loss": 0.3868,
+      "step": 2465
+    },
+    {
+      "epoch": 0.6349615223621083,
+      "eval_PRM Accuracy": 0.839622641509434,
+      "eval_PRM F1": 0.9005847953216374,
+      "eval_PRM F1 AUC": 0.7247249869041383,
+      "eval_PRM F1 AUC (fixed)": 0.8588266107909901,
+      "eval_PRM F1 Neg": 0.5853658536585366,
+      "eval_PRM NPV": 0.6666666666666666,
+      "eval_PRM Precision": 0.875,
+      "eval_PRM Recall": 0.927710843373494,
+      "eval_PRM Specificty": 0.5217391304347826,
+      "eval_loss": 0.36843040585517883,
+      "eval_runtime": 16.8694,
+      "eval_samples_per_second": 2.608,
+      "eval_steps_per_second": 0.178,
+      "step": 2465
+    },
+    {
+      "epoch": 0.6352191132433912,
+      "grad_norm": 0.18915726926721044,
+      "learning_rate": 3.535537509726575e-05,
+      "loss": 0.3844,
+      "step": 2466
+    },
+    {
+      "epoch": 0.6354767041246739,
+      "grad_norm": 0.1617886697675632,
+      "learning_rate": 3.531238332655197e-05,
+      "loss": 0.294,
+      "step": 2467
+    },
+    {
+      "epoch": 0.6357342950059568,
+      "grad_norm": 0.1838773212150472,
+      "learning_rate": 3.5269403436862175e-05,
+      "loss": 0.3225,
+      "step": 2468
+    },
+    {
+      "epoch": 0.6359918858872395,
+      "grad_norm": 0.20907038325245178,
+      "learning_rate": 3.5226435462963504e-05,
+      "loss": 0.314,
+      "step": 2469
+    },
+    {
+      "epoch": 0.6362494767685224,
+      "grad_norm": 0.24405684396483357,
+      "learning_rate": 3.518347943961331e-05,
+      "loss": 0.4034,
+      "step": 2470
+    },
+    {
+      "epoch": 0.6362494767685224,
+      "eval_PRM Accuracy": 0.839622641509434,
+      "eval_PRM F1": 0.8994082840236687,
+      "eval_PRM F1 AUC": 0.7404400209533788,
+      "eval_PRM F1 AUC (fixed)": 0.858302776322682,
+      "eval_PRM F1 Neg": 0.6046511627906976,
+      "eval_PRM NPV": 0.65,
+      "eval_PRM Precision": 0.8837209302325582,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.36319246888160706,
+      "eval_runtime": 17.5333,
+      "eval_samples_per_second": 2.51,
+      "eval_steps_per_second": 0.171,
+      "step": 2470
+    },
+    {
+      "epoch": 0.6365070676498052,
+      "grad_norm": 0.18978304088865044,
+      "learning_rate": 3.5140535401559336e-05,
+      "loss": 0.4053,
+      "step": 2471
+    },
+    {
+      "epoch": 0.636764658531088,
+      "grad_norm": 0.12154912178567832,
+      "learning_rate": 3.509760338353967e-05,
+      "loss": 0.3001,
+      "step": 2472
+    },
+    {
+      "epoch": 0.6370222494123708,
+      "grad_norm": 0.14639788033094533,
+      "learning_rate": 3.5054683420282605e-05,
+      "loss": 0.3305,
+      "step": 2473
+    },
+    {
+      "epoch": 0.6372798402936536,
+      "grad_norm": 0.17426631388467265,
+      "learning_rate": 3.5011775546506736e-05,
+      "loss": 0.4327,
+      "step": 2474
+    },
+    {
+      "epoch": 0.6375374311749364,
+      "grad_norm": 0.18550190630188318,
+      "learning_rate": 3.496887979692084e-05,
+      "loss": 0.4209,
+      "step": 2475
+    },
+    {
+      "epoch": 0.6375374311749364,
+      "eval_PRM Accuracy": 0.8301886792452831,
+      "eval_PRM F1": 0.891566265060241,
+      "eval_PRM F1 AUC": 0.7501309586170771,
+      "eval_PRM F1 AUC (fixed)": 0.8501833420639079,
+      "eval_PRM F1 Neg": 0.6086956521739131,
+      "eval_PRM NPV": 0.6086956521739131,
+      "eval_PRM Precision": 0.891566265060241,
+      "eval_PRM Recall": 0.891566265060241,
+      "eval_PRM Specificty": 0.6086956521739131,
+      "eval_loss": 0.3565784692764282,
+      "eval_runtime": 16.9846,
+      "eval_samples_per_second": 2.591,
+      "eval_steps_per_second": 0.177,
+      "step": 2475
+    },
+    {
+      "epoch": 0.6377950220562192,
+      "grad_norm": 0.19721703746705196,
+      "learning_rate": 3.492599620622392e-05,
+      "loss": 0.3477,
+      "step": 2476
+    },
+    {
+      "epoch": 0.638052612937502,
+      "grad_norm": 0.1553088976127993,
+      "learning_rate": 3.488312480910514e-05,
+      "loss": 0.2938,
+      "step": 2477
+    },
+    {
+      "epoch": 0.6383102038187848,
+      "grad_norm": 0.1360575047069072,
+      "learning_rate": 3.484026564024373e-05,
+      "loss": 0.3407,
+      "step": 2478
+    },
+    {
+      "epoch": 0.6385677947000676,
+      "grad_norm": 0.21361931804260326,
+      "learning_rate": 3.479741873430917e-05,
+      "loss": 0.4115,
+      "step": 2479
+    },
+    {
+      "epoch": 0.6388253855813504,
+      "grad_norm": 0.15468757231204278,
+      "learning_rate": 3.4754584125960864e-05,
+      "loss": 0.3112,
+      "step": 2480
+    },
+    {
+      "epoch": 0.6388253855813504,
+      "eval_PRM Accuracy": 0.8301886792452831,
+      "eval_PRM F1": 0.8928571428571429,
+      "eval_PRM F1 AUC": 0.7344159245678366,
+      "eval_PRM F1 AUC (fixed)": 0.8467784180199057,
+      "eval_PRM F1 Neg": 0.5909090909090909,
+      "eval_PRM NPV": 0.6190476190476191,
+      "eval_PRM Precision": 0.8823529411764706,
+      "eval_PRM Recall": 0.9036144578313253,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.36305931210517883,
+      "eval_runtime": 17.1046,
+      "eval_samples_per_second": 2.572,
+      "eval_steps_per_second": 0.175,
+      "step": 2480
+    },
+    {
+      "epoch": 0.6390829764626332,
+      "grad_norm": 0.1495308777069929,
+      "learning_rate": 3.471176184984839e-05,
+      "loss": 0.3673,
+      "step": 2481
+    },
+    {
+      "epoch": 0.639340567343916,
+      "grad_norm": 0.14968331179907596,
+      "learning_rate": 3.466895194061128e-05,
+      "loss": 0.3066,
+      "step": 2482
+    },
+    {
+      "epoch": 0.6395981582251988,
+      "grad_norm": 0.22761623239442078,
+      "learning_rate": 3.462615443287909e-05,
+      "loss": 0.3957,
+      "step": 2483
+    },
+    {
+      "epoch": 0.6398557491064817,
+      "grad_norm": 0.14780407170317722,
+      "learning_rate": 3.458336936127134e-05,
+      "loss": 0.3565,
+      "step": 2484
+    },
+    {
+      "epoch": 0.6401133399877644,
+      "grad_norm": 0.16277120545258716,
+      "learning_rate": 3.4540596760397484e-05,
+      "loss": 0.3883,
+      "step": 2485
+    },
+    {
+      "epoch": 0.6401133399877644,
+      "eval_PRM Accuracy": 0.839622641509434,
+      "eval_PRM F1": 0.8994082840236687,
+      "eval_PRM F1 AUC": 0.7404400209533788,
+      "eval_PRM F1 AUC (fixed)": 0.8452069146149817,
+      "eval_PRM F1 Neg": 0.6046511627906976,
+      "eval_PRM NPV": 0.65,
+      "eval_PRM Precision": 0.8837209302325582,
+      "eval_PRM Recall": 0.9156626506024096,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.37033912539482117,
+      "eval_runtime": 16.8255,
+      "eval_samples_per_second": 2.615,
+      "eval_steps_per_second": 0.178,
+      "step": 2485
+    },
+    {
+      "epoch": 0.6403709308690473,
+      "grad_norm": 0.16935247016016258,
+      "learning_rate": 3.449783666485691e-05,
+      "loss": 0.4021,
+      "step": 2486
+    },
+    {
+      "epoch": 0.64062852175033,
+      "grad_norm": 0.14878556083839845,
+      "learning_rate": 3.4455089109238835e-05,
+      "loss": 0.3479,
+      "step": 2487
+    },
+    {
+      "epoch": 0.6408861126316129,
+      "grad_norm": 0.22207223552251215,
+      "learning_rate": 3.4412354128122424e-05,
+      "loss": 0.4185,
+      "step": 2488
+    },
+    {
+      "epoch": 0.6411437035128956,
+      "grad_norm": 0.12561016373143336,
+      "learning_rate": 3.436963175607656e-05,
+      "loss": 0.3105,
+      "step": 2489
+    },
+    {
+      "epoch": 0.6414012943941785,
+      "grad_norm": 0.1504741636653908,
+      "learning_rate": 3.4326922027660036e-05,
+      "loss": 0.3924,
+      "step": 2490
+    },
+    {
+      "epoch": 0.6414012943941785,
+      "eval_PRM Accuracy": 0.8207547169811321,
+      "eval_PRM F1": 0.8862275449101796,
+      "eval_PRM F1 AUC": 0.7283918281822945,
+      "eval_PRM F1 AUC (fixed)": 0.8431115767417496,
+      "eval_PRM F1 Neg": 0.5777777777777777,
+      "eval_PRM NPV": 0.5909090909090909,
+      "eval_PRM Precision": 0.8809523809523809,
+      "eval_PRM Recall": 0.891566265060241,
+      "eval_PRM Specificty": 0.5652173913043478,
+      "eval_loss": 0.3628373444080353,
+      "eval_runtime": 17.2019,
+      "eval_samples_per_second": 2.558,
+      "eval_steps_per_second": 0.174,
+      "step": 2490
+    },
+    {
+      "epoch": 0.6416588852754612,
+      "grad_norm": 0.12600048006648018,
+      "learning_rate": 3.428422497742133e-05,
+      "loss": 0.3471,
+      "step": 2491
+    },
+    {
+      "epoch": 0.6419164761567441,
+      "grad_norm": 0.14653266088375858,
+      "learning_rate": 3.42415406398987e-05,
+      "loss": 0.3411,
+      "step": 2492
+    },
+    {
+      "epoch": 0.6421740670380268,
+      "grad_norm": 0.14251994876680993,
+      "learning_rate": 3.419886904962014e-05,
+      "loss": 0.3597,
+      "step": 2493
+    },
+    {
+      "epoch": 0.6424316579193097,
+      "grad_norm": 0.16747605739930146,
+      "learning_rate": 3.4156210241103284e-05,
+      "loss": 0.3475,
+      "step": 2494
+    },
+    {
+      "epoch": 0.6426892488005924,
+      "grad_norm": 0.13704030627396535,
+      "learning_rate": 3.411356424885549e-05,
+      "loss": 0.3667,
+      "step": 2495
+    },
+    {
+      "epoch": 0.6426892488005924,
+      "eval_PRM Accuracy": 0.8301886792452831,
+      "eval_PRM F1": 0.8902439024390244,
+      "eval_PRM F1 AUC": 0.7658459926663174,
+      "eval_PRM F1 AUC (fixed)": 0.8365636458878994,
+      "eval_PRM F1 Neg": 0.625,
+      "eval_PRM NPV": 0.6,
+      "eval_PRM Precision": 0.9012345679012346,
+      "eval_PRM Recall": 0.8795180722891566,
+      "eval_PRM Specificty": 0.6521739130434783,
+      "eval_loss": 0.36026278138160706,
+      "eval_runtime": 16.9453,
+      "eval_samples_per_second": 2.597,
+      "eval_steps_per_second": 0.177,
+      "step": 2495
+    },
+    {
+      "epoch": 0.6429468396818753,
+      "grad_norm": 0.3215333205805997,
+      "learning_rate": 3.4070931107373675e-05,
+      "loss": 0.3883,
+      "step": 2496
+    },
+    {
+      "epoch": 0.643204430563158,
+      "grad_norm": 0.14689143727399998,
+      "learning_rate": 3.402831085114444e-05,
+      "loss": 0.3536,
+      "step": 2497
+    },
+    {
+      "epoch": 0.6434620214444409,
+      "grad_norm": 0.22675565527247427,
+      "learning_rate": 3.398570351464391e-05,
+      "loss": 0.5071,
+      "step": 2498
+    },
+    {
+      "epoch": 0.6437196123257237,
+      "grad_norm": 0.1408923021693267,
+      "learning_rate": 3.394310913233774e-05,
+      "loss": 0.303,
+      "step": 2499
+    },
+    {
+      "epoch": 0.6439772032070065,
+      "grad_norm": 0.15506363825547645,
+      "learning_rate": 3.390052773868116e-05,
+      "loss": 0.3007,
+      "step": 2500
+    },
+    {
+      "epoch": 0.6439772032070065,
+      "eval_PRM Accuracy": 0.8301886792452831,
+      "eval_PRM F1": 0.8902439024390244,
+      "eval_PRM F1 AUC": 0.7658459926663174,
+      "eval_PRM F1 AUC (fixed)": 0.8365636458878996,
+      "eval_PRM F1 Neg": 0.625,
+      "eval_PRM NPV": 0.6,
+      "eval_PRM Precision": 0.9012345679012346,
+      "eval_PRM Recall": 0.8795180722891566,
+      "eval_PRM Specificty": 0.6521739130434783,
+      "eval_loss": 0.3553355932235718,
+      "eval_runtime": 16.5546,
+      "eval_samples_per_second": 2.658,
+      "eval_steps_per_second": 0.181,
+      "step": 2500
     }
   ],
   "logging_steps": 1,
@@ -20843,7 +26043,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 4365188093968384.0,
+  "total_flos": 5453105592795136.0,
   "train_batch_size": 4,
   "trial_name": null,
   "trial_params": null