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": 0.5178496294138589,
+  "epoch": 0.5825808330905913,
   "eval_steps": 5,
-  "global_step": 4000,
+  "global_step": 4500,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -40823,6 +40823,5106 @@
       "eval_samples_per_second": 3.077,
       "eval_steps_per_second": 0.21,
       "step": 4000
+    },
+    {
+      "epoch": 0.5179790918212124,
+      "grad_norm": 2.3046162616095063,
+      "learning_rate": 6.947672681532358e-07,
+      "loss": 2.5112,
+      "step": 4001
+    },
+    {
+      "epoch": 0.5181085542285658,
+      "grad_norm": 1.274892818735378,
+      "learning_rate": 6.944865498056849e-07,
+      "loss": 2.2561,
+      "step": 4002
+    },
+    {
+      "epoch": 0.5182380166359194,
+      "grad_norm": 1.3678346798583385,
+      "learning_rate": 6.942058172640981e-07,
+      "loss": 2.2942,
+      "step": 4003
+    },
+    {
+      "epoch": 0.5183674790432728,
+      "grad_norm": 1.6670188504381802,
+      "learning_rate": 6.9392507058582e-07,
+      "loss": 2.4956,
+      "step": 4004
+    },
+    {
+      "epoch": 0.5184969414506263,
+      "grad_norm": 1.8297421986920928,
+      "learning_rate": 6.936443098281993e-07,
+      "loss": 2.6304,
+      "step": 4005
+    },
+    {
+      "epoch": 0.5184969414506263,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5488947629928589,
+      "eval_runtime": 13.9259,
+      "eval_samples_per_second": 3.16,
+      "eval_steps_per_second": 0.215,
+      "step": 4005
+    },
+    {
+      "epoch": 0.5186264038579798,
+      "grad_norm": 2.281361998524018,
+      "learning_rate": 6.93363535048587e-07,
+      "loss": 2.3857,
+      "step": 4006
+    },
+    {
+      "epoch": 0.5187558662653332,
+      "grad_norm": 1.8496701172040177,
+      "learning_rate": 6.93082746304337e-07,
+      "loss": 2.5203,
+      "step": 4007
+    },
+    {
+      "epoch": 0.5188853286726867,
+      "grad_norm": 2.9370386479711414,
+      "learning_rate": 6.928019436528062e-07,
+      "loss": 2.5742,
+      "step": 4008
+    },
+    {
+      "epoch": 0.5190147910800401,
+      "grad_norm": 1.5975234790838664,
+      "learning_rate": 6.92521127151354e-07,
+      "loss": 2.3896,
+      "step": 4009
+    },
+    {
+      "epoch": 0.5191442534873936,
+      "grad_norm": 1.5829937977805089,
+      "learning_rate": 6.922402968573429e-07,
+      "loss": 2.3433,
+      "step": 4010
+    },
+    {
+      "epoch": 0.5191442534873936,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.545365810394287,
+      "eval_runtime": 13.7942,
+      "eval_samples_per_second": 3.19,
+      "eval_steps_per_second": 0.217,
+      "step": 4010
+    },
+    {
+      "epoch": 0.519273715894747,
+      "grad_norm": 2.259205106524019,
+      "learning_rate": 6.919594528281384e-07,
+      "loss": 2.7288,
+      "step": 4011
+    },
+    {
+      "epoch": 0.5194031783021005,
+      "grad_norm": 1.8968202224773139,
+      "learning_rate": 6.916785951211084e-07,
+      "loss": 2.1353,
+      "step": 4012
+    },
+    {
+      "epoch": 0.519532640709454,
+      "grad_norm": 2.9708380498163365,
+      "learning_rate": 6.913977237936239e-07,
+      "loss": 3.1262,
+      "step": 4013
+    },
+    {
+      "epoch": 0.5196621031168075,
+      "grad_norm": 1.4358213582854178,
+      "learning_rate": 6.911168389030584e-07,
+      "loss": 2.5811,
+      "step": 4014
+    },
+    {
+      "epoch": 0.519791565524161,
+      "grad_norm": 1.5399669742781321,
+      "learning_rate": 6.908359405067885e-07,
+      "loss": 2.3269,
+      "step": 4015
+    },
+    {
+      "epoch": 0.519791565524161,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5490056276321411,
+      "eval_runtime": 14.1061,
+      "eval_samples_per_second": 3.119,
+      "eval_steps_per_second": 0.213,
+      "step": 4015
+    },
+    {
+      "epoch": 0.5199210279315144,
+      "grad_norm": 1.3322671984492085,
+      "learning_rate": 6.905550286621933e-07,
+      "loss": 2.4346,
+      "step": 4016
+    },
+    {
+      "epoch": 0.5200504903388679,
+      "grad_norm": 1.5827936533204134,
+      "learning_rate": 6.902741034266548e-07,
+      "loss": 2.3481,
+      "step": 4017
+    },
+    {
+      "epoch": 0.5201799527462213,
+      "grad_norm": 1.2818233116674882,
+      "learning_rate": 6.899931648575575e-07,
+      "loss": 2.4534,
+      "step": 4018
+    },
+    {
+      "epoch": 0.5203094151535748,
+      "grad_norm": 1.8177028507779016,
+      "learning_rate": 6.89712213012289e-07,
+      "loss": 2.6448,
+      "step": 4019
+    },
+    {
+      "epoch": 0.5204388775609282,
+      "grad_norm": 2.977795867828807,
+      "learning_rate": 6.894312479482394e-07,
+      "loss": 2.3749,
+      "step": 4020
+    },
+    {
+      "epoch": 0.5204388775609282,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5468083620071411,
+      "eval_runtime": 13.2784,
+      "eval_samples_per_second": 3.314,
+      "eval_steps_per_second": 0.226,
+      "step": 4020
+    },
+    {
+      "epoch": 0.5205683399682817,
+      "grad_norm": 3.8761697200983587,
+      "learning_rate": 6.891502697228014e-07,
+      "loss": 2.6353,
+      "step": 4021
+    },
+    {
+      "epoch": 0.5206978023756351,
+      "grad_norm": 2.0277540872485567,
+      "learning_rate": 6.888692783933706e-07,
+      "loss": 2.3862,
+      "step": 4022
+    },
+    {
+      "epoch": 0.5208272647829887,
+      "grad_norm": 2.5307349281963223,
+      "learning_rate": 6.88588274017345e-07,
+      "loss": 2.5737,
+      "step": 4023
+    },
+    {
+      "epoch": 0.5209567271903421,
+      "grad_norm": 2.069943371719611,
+      "learning_rate": 6.883072566521255e-07,
+      "loss": 2.4773,
+      "step": 4024
+    },
+    {
+      "epoch": 0.5210861895976956,
+      "grad_norm": 2.153232460800363,
+      "learning_rate": 6.880262263551159e-07,
+      "loss": 2.4248,
+      "step": 4025
+    },
+    {
+      "epoch": 0.5210861895976956,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5490056276321411,
+      "eval_runtime": 13.8471,
+      "eval_samples_per_second": 3.178,
+      "eval_steps_per_second": 0.217,
+      "step": 4025
+    },
+    {
+      "epoch": 0.521215652005049,
+      "grad_norm": 1.6353442719300793,
+      "learning_rate": 6.877451831837219e-07,
+      "loss": 2.3752,
+      "step": 4026
+    },
+    {
+      "epoch": 0.5213451144124025,
+      "grad_norm": 2.2208258264771548,
+      "learning_rate": 6.874641271953525e-07,
+      "loss": 2.4895,
+      "step": 4027
+    },
+    {
+      "epoch": 0.521474576819756,
+      "grad_norm": 2.213323882831792,
+      "learning_rate": 6.871830584474189e-07,
+      "loss": 2.5103,
+      "step": 4028
+    },
+    {
+      "epoch": 0.5216040392271094,
+      "grad_norm": 1.9378846944831578,
+      "learning_rate": 6.869019769973354e-07,
+      "loss": 2.4146,
+      "step": 4029
+    },
+    {
+      "epoch": 0.5217335016344629,
+      "grad_norm": 1.8529593255096792,
+      "learning_rate": 6.866208829025182e-07,
+      "loss": 2.3682,
+      "step": 4030
+    },
+    {
+      "epoch": 0.5217335016344629,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.544633388519287,
+      "eval_runtime": 13.697,
+      "eval_samples_per_second": 3.212,
+      "eval_steps_per_second": 0.219,
+      "step": 4030
+    },
+    {
+      "epoch": 0.5218629640418163,
+      "grad_norm": 1.6765713081694857,
+      "learning_rate": 6.86339776220387e-07,
+      "loss": 2.269,
+      "step": 4031
+    },
+    {
+      "epoch": 0.5219924264491699,
+      "grad_norm": 2.158720777122161,
+      "learning_rate": 6.860586570083629e-07,
+      "loss": 2.4163,
+      "step": 4032
+    },
+    {
+      "epoch": 0.5221218888565233,
+      "grad_norm": 1.8396302914952403,
+      "learning_rate": 6.857775253238706e-07,
+      "loss": 2.6018,
+      "step": 4033
+    },
+    {
+      "epoch": 0.5222513512638768,
+      "grad_norm": 2.89006099770384,
+      "learning_rate": 6.854963812243369e-07,
+      "loss": 2.7969,
+      "step": 4034
+    },
+    {
+      "epoch": 0.5223808136712302,
+      "grad_norm": 1.2566850857976133,
+      "learning_rate": 6.852152247671911e-07,
+      "loss": 2.2957,
+      "step": 4035
+    },
+    {
+      "epoch": 0.5223808136712302,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5445889234542847,
+      "eval_runtime": 13.8081,
+      "eval_samples_per_second": 3.187,
+      "eval_steps_per_second": 0.217,
+      "step": 4035
+    },
+    {
+      "epoch": 0.5225102760785837,
+      "grad_norm": 1.5335537435467135,
+      "learning_rate": 6.84934056009865e-07,
+      "loss": 2.46,
+      "step": 4036
+    },
+    {
+      "epoch": 0.5226397384859371,
+      "grad_norm": 3.795728024773409,
+      "learning_rate": 6.846528750097933e-07,
+      "loss": 2.5276,
+      "step": 4037
+    },
+    {
+      "epoch": 0.5227692008932906,
+      "grad_norm": 1.3880231422149454,
+      "learning_rate": 6.843716818244129e-07,
+      "loss": 2.48,
+      "step": 4038
+    },
+    {
+      "epoch": 0.522898663300644,
+      "grad_norm": 1.4284678586421233,
+      "learning_rate": 6.840904765111629e-07,
+      "loss": 2.4761,
+      "step": 4039
+    },
+    {
+      "epoch": 0.5230281257079975,
+      "grad_norm": 2.2523568981713273,
+      "learning_rate": 6.838092591274855e-07,
+      "loss": 2.5649,
+      "step": 4040
+    },
+    {
+      "epoch": 0.5230281257079975,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5475407838821411,
+      "eval_runtime": 15.0819,
+      "eval_samples_per_second": 2.917,
+      "eval_steps_per_second": 0.199,
+      "step": 4040
+    },
+    {
+      "epoch": 0.523157588115351,
+      "grad_norm": 1.2029077852565317,
+      "learning_rate": 6.835280297308247e-07,
+      "loss": 2.3252,
+      "step": 4041
+    },
+    {
+      "epoch": 0.5232870505227045,
+      "grad_norm": 1.3756174207563119,
+      "learning_rate": 6.832467883786276e-07,
+      "loss": 2.3965,
+      "step": 4042
+    },
+    {
+      "epoch": 0.523416512930058,
+      "grad_norm": 1.4355098504264527,
+      "learning_rate": 6.829655351283434e-07,
+      "loss": 2.4001,
+      "step": 4043
+    },
+    {
+      "epoch": 0.5235459753374114,
+      "grad_norm": 2.45334991143882,
+      "learning_rate": 6.826842700374238e-07,
+      "loss": 2.6587,
+      "step": 4044
+    },
+    {
+      "epoch": 0.5236754377447649,
+      "grad_norm": 1.531338741425381,
+      "learning_rate": 6.824029931633227e-07,
+      "loss": 2.3848,
+      "step": 4045
+    },
+    {
+      "epoch": 0.5236754377447649,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5446555614471436,
+      "eval_runtime": 14.4656,
+      "eval_samples_per_second": 3.042,
+      "eval_steps_per_second": 0.207,
+      "step": 4045
+    },
+    {
+      "epoch": 0.5238049001521183,
+      "grad_norm": 1.8417709072515376,
+      "learning_rate": 6.821217045634968e-07,
+      "loss": 2.2681,
+      "step": 4046
+    },
+    {
+      "epoch": 0.5239343625594718,
+      "grad_norm": 2.226281162429195,
+      "learning_rate": 6.81840404295405e-07,
+      "loss": 2.2002,
+      "step": 4047
+    },
+    {
+      "epoch": 0.5240638249668252,
+      "grad_norm": 2.2354565806073716,
+      "learning_rate": 6.815590924165085e-07,
+      "loss": 2.6931,
+      "step": 4048
+    },
+    {
+      "epoch": 0.5241932873741787,
+      "grad_norm": 1.9605992545028588,
+      "learning_rate": 6.812777689842708e-07,
+      "loss": 2.3494,
+      "step": 4049
+    },
+    {
+      "epoch": 0.5243227497815321,
+      "grad_norm": 1.6412602427597318,
+      "learning_rate": 6.809964340561583e-07,
+      "loss": 2.4158,
+      "step": 4050
+    },
+    {
+      "epoch": 0.5243227497815321,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.543257236480713,
+      "eval_runtime": 14.0658,
+      "eval_samples_per_second": 3.128,
+      "eval_steps_per_second": 0.213,
+      "step": 4050
+    },
+    {
+      "epoch": 0.5244522121888856,
+      "grad_norm": 1.649246079162323,
+      "learning_rate": 6.80715087689639e-07,
+      "loss": 2.3794,
+      "step": 4051
+    },
+    {
+      "epoch": 0.5245816745962392,
+      "grad_norm": 2.5638443834562086,
+      "learning_rate": 6.804337299421837e-07,
+      "loss": 2.5198,
+      "step": 4052
+    },
+    {
+      "epoch": 0.5247111370035926,
+      "grad_norm": 2.1241400341109533,
+      "learning_rate": 6.801523608712659e-07,
+      "loss": 2.5381,
+      "step": 4053
+    },
+    {
+      "epoch": 0.5248405994109461,
+      "grad_norm": 4.498602466975845,
+      "learning_rate": 6.798709805343601e-07,
+      "loss": 2.5005,
+      "step": 4054
+    },
+    {
+      "epoch": 0.5249700618182995,
+      "grad_norm": 2.781289622954872,
+      "learning_rate": 6.795895889889443e-07,
+      "loss": 2.6262,
+      "step": 4055
+    },
+    {
+      "epoch": 0.5249700618182995,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5418590307235718,
+      "eval_runtime": 15.221,
+      "eval_samples_per_second": 2.891,
+      "eval_steps_per_second": 0.197,
+      "step": 4055
+    },
+    {
+      "epoch": 0.525099524225653,
+      "grad_norm": 3.0928074247533215,
+      "learning_rate": 6.793081862924987e-07,
+      "loss": 2.7434,
+      "step": 4056
+    },
+    {
+      "epoch": 0.5252289866330064,
+      "grad_norm": 2.0552540618606683,
+      "learning_rate": 6.790267725025052e-07,
+      "loss": 2.5503,
+      "step": 4057
+    },
+    {
+      "epoch": 0.5253584490403599,
+      "grad_norm": 1.8352263219916283,
+      "learning_rate": 6.787453476764483e-07,
+      "loss": 2.3716,
+      "step": 4058
+    },
+    {
+      "epoch": 0.5254879114477133,
+      "grad_norm": 1.8612845516455199,
+      "learning_rate": 6.784639118718146e-07,
+      "loss": 2.5055,
+      "step": 4059
+    },
+    {
+      "epoch": 0.5256173738550668,
+      "grad_norm": 2.5825970119011976,
+      "learning_rate": 6.781824651460934e-07,
+      "loss": 2.7236,
+      "step": 4060
+    },
+    {
+      "epoch": 0.5256173738550668,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5419255495071411,
+      "eval_runtime": 14.1009,
+      "eval_samples_per_second": 3.12,
+      "eval_steps_per_second": 0.213,
+      "step": 4060
+    },
+    {
+      "epoch": 0.5257468362624202,
+      "grad_norm": 3.364303685325622,
+      "learning_rate": 6.77901007556776e-07,
+      "loss": 2.4731,
+      "step": 4061
+    },
+    {
+      "epoch": 0.5258762986697738,
+      "grad_norm": 1.4290990338587204,
+      "learning_rate": 6.776195391613552e-07,
+      "loss": 2.2952,
+      "step": 4062
+    },
+    {
+      "epoch": 0.5260057610771273,
+      "grad_norm": 2.0174493980427184,
+      "learning_rate": 6.773380600173271e-07,
+      "loss": 2.6475,
+      "step": 4063
+    },
+    {
+      "epoch": 0.5261352234844807,
+      "grad_norm": 1.8965380072530311,
+      "learning_rate": 6.770565701821893e-07,
+      "loss": 2.6648,
+      "step": 4064
+    },
+    {
+      "epoch": 0.5262646858918342,
+      "grad_norm": 2.940568522815605,
+      "learning_rate": 6.76775069713442e-07,
+      "loss": 2.2603,
+      "step": 4065
+    },
+    {
+      "epoch": 0.5262646858918342,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5426357984542847,
+      "eval_runtime": 14.3713,
+      "eval_samples_per_second": 3.062,
+      "eval_steps_per_second": 0.209,
+      "step": 4065
+    },
+    {
+      "epoch": 0.5263941482991876,
+      "grad_norm": 1.276968373334518,
+      "learning_rate": 6.764935586685873e-07,
+      "loss": 2.2014,
+      "step": 4066
+    },
+    {
+      "epoch": 0.5265236107065411,
+      "grad_norm": 2.467185405032055,
+      "learning_rate": 6.762120371051296e-07,
+      "loss": 2.8103,
+      "step": 4067
+    },
+    {
+      "epoch": 0.5266530731138945,
+      "grad_norm": 1.6662267916379692,
+      "learning_rate": 6.75930505080575e-07,
+      "loss": 2.4114,
+      "step": 4068
+    },
+    {
+      "epoch": 0.526782535521248,
+      "grad_norm": 1.369969145436834,
+      "learning_rate": 6.756489626524326e-07,
+      "loss": 2.4111,
+      "step": 4069
+    },
+    {
+      "epoch": 0.5269119979286014,
+      "grad_norm": 1.3334883998849394,
+      "learning_rate": 6.753674098782131e-07,
+      "loss": 2.3872,
+      "step": 4070
+    },
+    {
+      "epoch": 0.5269119979286014,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5459872484207153,
+      "eval_runtime": 14.3671,
+      "eval_samples_per_second": 3.063,
+      "eval_steps_per_second": 0.209,
+      "step": 4070
+    },
+    {
+      "epoch": 0.5270414603359549,
+      "grad_norm": 1.7450215341448398,
+      "learning_rate": 6.750858468154291e-07,
+      "loss": 2.6187,
+      "step": 4071
+    },
+    {
+      "epoch": 0.5271709227433085,
+      "grad_norm": 2.390558197174125,
+      "learning_rate": 6.748042735215957e-07,
+      "loss": 2.5771,
+      "step": 4072
+    },
+    {
+      "epoch": 0.5273003851506619,
+      "grad_norm": 1.2022305620236275,
+      "learning_rate": 6.745226900542298e-07,
+      "loss": 2.2388,
+      "step": 4073
+    },
+    {
+      "epoch": 0.5274298475580154,
+      "grad_norm": 1.8862366979778469,
+      "learning_rate": 6.742410964708509e-07,
+      "loss": 2.6208,
+      "step": 4074
+    },
+    {
+      "epoch": 0.5275593099653688,
+      "grad_norm": 1.8438641976208832,
+      "learning_rate": 6.7395949282898e-07,
+      "loss": 2.5164,
+      "step": 4075
+    },
+    {
+      "epoch": 0.5275593099653688,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5446999073028564,
+      "eval_runtime": 15.033,
+      "eval_samples_per_second": 2.927,
+      "eval_steps_per_second": 0.2,
+      "step": 4075
+    },
+    {
+      "epoch": 0.5276887723727223,
+      "grad_norm": 2.251961850422397,
+      "learning_rate": 6.736778791861403e-07,
+      "loss": 2.428,
+      "step": 4076
+    },
+    {
+      "epoch": 0.5278182347800757,
+      "grad_norm": 1.820521444568679,
+      "learning_rate": 6.733962555998569e-07,
+      "loss": 2.3325,
+      "step": 4077
+    },
+    {
+      "epoch": 0.5279476971874292,
+      "grad_norm": 1.9899972253345364,
+      "learning_rate": 6.731146221276575e-07,
+      "loss": 2.5508,
+      "step": 4078
+    },
+    {
+      "epoch": 0.5280771595947826,
+      "grad_norm": 2.0062209838055205,
+      "learning_rate": 6.728329788270713e-07,
+      "loss": 2.7676,
+      "step": 4079
+    },
+    {
+      "epoch": 0.5282066220021361,
+      "grad_norm": 2.3639440229428352,
+      "learning_rate": 6.725513257556296e-07,
+      "loss": 2.4801,
+      "step": 4080
+    },
+    {
+      "epoch": 0.5282066220021361,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5431462526321411,
+      "eval_runtime": 14.5526,
+      "eval_samples_per_second": 3.024,
+      "eval_steps_per_second": 0.206,
+      "step": 4080
+    },
+    {
+      "epoch": 0.5283360844094896,
+      "grad_norm": 1.8032323195283315,
+      "learning_rate": 6.72269662970866e-07,
+      "loss": 2.3364,
+      "step": 4081
+    },
+    {
+      "epoch": 0.5284655468168431,
+      "grad_norm": 2.0547809072510845,
+      "learning_rate": 6.719879905303152e-07,
+      "loss": 2.533,
+      "step": 4082
+    },
+    {
+      "epoch": 0.5285950092241966,
+      "grad_norm": 2.206939725330552,
+      "learning_rate": 6.717063084915153e-07,
+      "loss": 2.5942,
+      "step": 4083
+    },
+    {
+      "epoch": 0.52872447163155,
+      "grad_norm": 2.092499638160591,
+      "learning_rate": 6.71424616912005e-07,
+      "loss": 2.3096,
+      "step": 4084
+    },
+    {
+      "epoch": 0.5288539340389035,
+      "grad_norm": 1.9624236100112382,
+      "learning_rate": 6.711429158493259e-07,
+      "loss": 2.2903,
+      "step": 4085
+    },
+    {
+      "epoch": 0.5288539340389035,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.543212890625,
+      "eval_runtime": 14.9978,
+      "eval_samples_per_second": 2.934,
+      "eval_steps_per_second": 0.2,
+      "step": 4085
+    },
+    {
+      "epoch": 0.5289833964462569,
+      "grad_norm": 2.652543670973225,
+      "learning_rate": 6.708612053610207e-07,
+      "loss": 2.6018,
+      "step": 4086
+    },
+    {
+      "epoch": 0.5291128588536104,
+      "grad_norm": 1.377147046044787,
+      "learning_rate": 6.705794855046349e-07,
+      "loss": 2.344,
+      "step": 4087
+    },
+    {
+      "epoch": 0.5292423212609638,
+      "grad_norm": 1.6605268731313942,
+      "learning_rate": 6.702977563377153e-07,
+      "loss": 2.2849,
+      "step": 4088
+    },
+    {
+      "epoch": 0.5293717836683173,
+      "grad_norm": 2.027139648045615,
+      "learning_rate": 6.700160179178109e-07,
+      "loss": 2.5793,
+      "step": 4089
+    },
+    {
+      "epoch": 0.5295012460756707,
+      "grad_norm": 1.3878072962026224,
+      "learning_rate": 6.697342703024724e-07,
+      "loss": 2.302,
+      "step": 4090
+    },
+    {
+      "epoch": 0.5295012460756707,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5409489870071411,
+      "eval_runtime": 13.6891,
+      "eval_samples_per_second": 3.214,
+      "eval_steps_per_second": 0.219,
+      "step": 4090
+    },
+    {
+      "epoch": 0.5296307084830243,
+      "grad_norm": 1.3877954933325167,
+      "learning_rate": 6.694525135492523e-07,
+      "loss": 2.4109,
+      "step": 4091
+    },
+    {
+      "epoch": 0.5297601708903777,
+      "grad_norm": 1.3435178934375451,
+      "learning_rate": 6.691707477157055e-07,
+      "loss": 2.2239,
+      "step": 4092
+    },
+    {
+      "epoch": 0.5298896332977312,
+      "grad_norm": 3.237947114871341,
+      "learning_rate": 6.688889728593881e-07,
+      "loss": 2.8003,
+      "step": 4093
+    },
+    {
+      "epoch": 0.5300190957050847,
+      "grad_norm": 1.5673185934234255,
+      "learning_rate": 6.686071890378584e-07,
+      "loss": 2.395,
+      "step": 4094
+    },
+    {
+      "epoch": 0.5301485581124381,
+      "grad_norm": 2.051905149309898,
+      "learning_rate": 6.683253963086764e-07,
+      "loss": 2.5671,
+      "step": 4095
+    },
+    {
+      "epoch": 0.5301485581124381,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5425692796707153,
+      "eval_runtime": 13.9015,
+      "eval_samples_per_second": 3.165,
+      "eval_steps_per_second": 0.216,
+      "step": 4095
+    },
+    {
+      "epoch": 0.5302780205197916,
+      "grad_norm": 1.849445812197663,
+      "learning_rate": 6.680435947294041e-07,
+      "loss": 2.6562,
+      "step": 4096
+    },
+    {
+      "epoch": 0.530407482927145,
+      "grad_norm": 2.6091837391899984,
+      "learning_rate": 6.677617843576052e-07,
+      "loss": 2.7395,
+      "step": 4097
+    },
+    {
+      "epoch": 0.5305369453344985,
+      "grad_norm": 1.795511185540601,
+      "learning_rate": 6.67479965250845e-07,
+      "loss": 2.4978,
+      "step": 4098
+    },
+    {
+      "epoch": 0.5306664077418519,
+      "grad_norm": 2.3018198330380897,
+      "learning_rate": 6.67198137466691e-07,
+      "loss": 2.5415,
+      "step": 4099
+    },
+    {
+      "epoch": 0.5307958701492054,
+      "grad_norm": 1.3505060470833354,
+      "learning_rate": 6.66916301062712e-07,
+      "loss": 2.3435,
+      "step": 4100
+    },
+    {
+      "epoch": 0.5307958701492054,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5418146848678589,
+      "eval_runtime": 14.2788,
+      "eval_samples_per_second": 3.081,
+      "eval_steps_per_second": 0.21,
+      "step": 4100
+    },
+    {
+      "epoch": 0.5309253325565589,
+      "grad_norm": 1.3975667251220192,
+      "learning_rate": 6.666344560964788e-07,
+      "loss": 2.2227,
+      "step": 4101
+    },
+    {
+      "epoch": 0.5310547949639124,
+      "grad_norm": 1.5539573154826345,
+      "learning_rate": 6.663526026255641e-07,
+      "loss": 2.4458,
+      "step": 4102
+    },
+    {
+      "epoch": 0.5311842573712658,
+      "grad_norm": 2.344456631655416,
+      "learning_rate": 6.660707407075423e-07,
+      "loss": 2.5728,
+      "step": 4103
+    },
+    {
+      "epoch": 0.5313137197786193,
+      "grad_norm": 2.124932888646604,
+      "learning_rate": 6.657888703999892e-07,
+      "loss": 2.606,
+      "step": 4104
+    },
+    {
+      "epoch": 0.5314431821859728,
+      "grad_norm": 1.748060373519471,
+      "learning_rate": 6.655069917604822e-07,
+      "loss": 2.4187,
+      "step": 4105
+    },
+    {
+      "epoch": 0.5314431821859728,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5445224046707153,
+      "eval_runtime": 14.5717,
+      "eval_samples_per_second": 3.02,
+      "eval_steps_per_second": 0.206,
+      "step": 4105
+    },
+    {
+      "epoch": 0.5315726445933262,
+      "grad_norm": 1.543397018032056,
+      "learning_rate": 6.652251048466016e-07,
+      "loss": 2.3669,
+      "step": 4106
+    },
+    {
+      "epoch": 0.5317021070006797,
+      "grad_norm": 2.7426462524261597,
+      "learning_rate": 6.649432097159278e-07,
+      "loss": 2.3608,
+      "step": 4107
+    },
+    {
+      "epoch": 0.5318315694080331,
+      "grad_norm": 2.2296753841059096,
+      "learning_rate": 6.646613064260439e-07,
+      "loss": 2.6189,
+      "step": 4108
+    },
+    {
+      "epoch": 0.5319610318153866,
+      "grad_norm": 1.5627983300053554,
+      "learning_rate": 6.64379395034534e-07,
+      "loss": 2.3882,
+      "step": 4109
+    },
+    {
+      "epoch": 0.53209049422274,
+      "grad_norm": 1.994126962465963,
+      "learning_rate": 6.640974755989846e-07,
+      "loss": 2.4468,
+      "step": 4110
+    },
+    {
+      "epoch": 0.53209049422274,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.545942783355713,
+      "eval_runtime": 13.7882,
+      "eval_samples_per_second": 3.191,
+      "eval_steps_per_second": 0.218,
+      "step": 4110
+    },
+    {
+      "epoch": 0.5322199566300936,
+      "grad_norm": 1.3010825712566525,
+      "learning_rate": 6.638155481769836e-07,
+      "loss": 2.4158,
+      "step": 4111
+    },
+    {
+      "epoch": 0.532349419037447,
+      "grad_norm": 2.0143028073878986,
+      "learning_rate": 6.635336128261198e-07,
+      "loss": 2.3218,
+      "step": 4112
+    },
+    {
+      "epoch": 0.5324788814448005,
+      "grad_norm": 2.252579202225458,
+      "learning_rate": 6.632516696039847e-07,
+      "loss": 2.5225,
+      "step": 4113
+    },
+    {
+      "epoch": 0.5326083438521539,
+      "grad_norm": 3.0033230100128896,
+      "learning_rate": 6.629697185681706e-07,
+      "loss": 2.813,
+      "step": 4114
+    },
+    {
+      "epoch": 0.5327378062595074,
+      "grad_norm": 1.572045835248253,
+      "learning_rate": 6.626877597762721e-07,
+      "loss": 2.6624,
+      "step": 4115
+    },
+    {
+      "epoch": 0.5327378062595074,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5383079051971436,
+      "eval_runtime": 13.7374,
+      "eval_samples_per_second": 3.203,
+      "eval_steps_per_second": 0.218,
+      "step": 4115
+    },
+    {
+      "epoch": 0.5328672686668608,
+      "grad_norm": 2.0625186151816677,
+      "learning_rate": 6.624057932858846e-07,
+      "loss": 2.5728,
+      "step": 4116
+    },
+    {
+      "epoch": 0.5329967310742143,
+      "grad_norm": 2.4584282769901447,
+      "learning_rate": 6.621238191546057e-07,
+      "loss": 2.5815,
+      "step": 4117
+    },
+    {
+      "epoch": 0.5331261934815678,
+      "grad_norm": 3.739672032630529,
+      "learning_rate": 6.618418374400341e-07,
+      "loss": 2.9109,
+      "step": 4118
+    },
+    {
+      "epoch": 0.5332556558889212,
+      "grad_norm": 2.22265460600069,
+      "learning_rate": 6.615598481997707e-07,
+      "loss": 2.6689,
+      "step": 4119
+    },
+    {
+      "epoch": 0.5333851182962748,
+      "grad_norm": 1.269188844666255,
+      "learning_rate": 6.612778514914172e-07,
+      "loss": 2.48,
+      "step": 4120
+    },
+    {
+      "epoch": 0.5333851182962748,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5423473119735718,
+      "eval_runtime": 14.4042,
+      "eval_samples_per_second": 3.055,
+      "eval_steps_per_second": 0.208,
+      "step": 4120
+    },
+    {
+      "epoch": 0.5335145807036282,
+      "grad_norm": 1.7695175092101905,
+      "learning_rate": 6.609958473725774e-07,
+      "loss": 2.5378,
+      "step": 4121
+    },
+    {
+      "epoch": 0.5336440431109817,
+      "grad_norm": 2.928718448413823,
+      "learning_rate": 6.607138359008558e-07,
+      "loss": 2.4871,
+      "step": 4122
+    },
+    {
+      "epoch": 0.5337735055183351,
+      "grad_norm": 3.269442266337227,
+      "learning_rate": 6.604318171338595e-07,
+      "loss": 3.1372,
+      "step": 4123
+    },
+    {
+      "epoch": 0.5339029679256886,
+      "grad_norm": 1.1861803220090243,
+      "learning_rate": 6.601497911291964e-07,
+      "loss": 2.2971,
+      "step": 4124
+    },
+    {
+      "epoch": 0.534032430333042,
+      "grad_norm": 1.578764117420061,
+      "learning_rate": 6.598677579444761e-07,
+      "loss": 2.5684,
+      "step": 4125
+    },
+    {
+      "epoch": 0.534032430333042,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5388849973678589,
+      "eval_runtime": 14.2809,
+      "eval_samples_per_second": 3.081,
+      "eval_steps_per_second": 0.21,
+      "step": 4125
+    },
+    {
+      "epoch": 0.5341618927403955,
+      "grad_norm": 2.086286422744801,
+      "learning_rate": 6.595857176373094e-07,
+      "loss": 2.4326,
+      "step": 4126
+    },
+    {
+      "epoch": 0.534291355147749,
+      "grad_norm": 2.2371130585898817,
+      "learning_rate": 6.593036702653087e-07,
+      "loss": 2.2461,
+      "step": 4127
+    },
+    {
+      "epoch": 0.5344208175551024,
+      "grad_norm": 1.3533887585013036,
+      "learning_rate": 6.590216158860884e-07,
+      "loss": 2.5505,
+      "step": 4128
+    },
+    {
+      "epoch": 0.5345502799624559,
+      "grad_norm": 3.522146648313255,
+      "learning_rate": 6.587395545572632e-07,
+      "loss": 2.916,
+      "step": 4129
+    },
+    {
+      "epoch": 0.5346797423698094,
+      "grad_norm": 1.7016474924008749,
+      "learning_rate": 6.584574863364501e-07,
+      "loss": 2.2529,
+      "step": 4130
+    },
+    {
+      "epoch": 0.5346797423698094,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5431462526321411,
+      "eval_runtime": 14.761,
+      "eval_samples_per_second": 2.981,
+      "eval_steps_per_second": 0.203,
+      "step": 4130
+    },
+    {
+      "epoch": 0.5348092047771629,
+      "grad_norm": 1.5306119576768233,
+      "learning_rate": 6.581754112812673e-07,
+      "loss": 2.4597,
+      "step": 4131
+    },
+    {
+      "epoch": 0.5349386671845163,
+      "grad_norm": 1.262342395383607,
+      "learning_rate": 6.578933294493344e-07,
+      "loss": 2.2175,
+      "step": 4132
+    },
+    {
+      "epoch": 0.5350681295918698,
+      "grad_norm": 2.7684296116377642,
+      "learning_rate": 6.576112408982721e-07,
+      "loss": 2.4918,
+      "step": 4133
+    },
+    {
+      "epoch": 0.5351975919992232,
+      "grad_norm": 4.262698770451707,
+      "learning_rate": 6.573291456857031e-07,
+      "loss": 2.2244,
+      "step": 4134
+    },
+    {
+      "epoch": 0.5353270544065767,
+      "grad_norm": 1.5831741269025994,
+      "learning_rate": 6.570470438692507e-07,
+      "loss": 2.5874,
+      "step": 4135
+    },
+    {
+      "epoch": 0.5353270544065767,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.540971279144287,
+      "eval_runtime": 14.3312,
+      "eval_samples_per_second": 3.07,
+      "eval_steps_per_second": 0.209,
+      "step": 4135
+    },
+    {
+      "epoch": 0.5354565168139301,
+      "grad_norm": 2.356478319566696,
+      "learning_rate": 6.567649355065399e-07,
+      "loss": 2.522,
+      "step": 4136
+    },
+    {
+      "epoch": 0.5355859792212836,
+      "grad_norm": 2.2150551599205524,
+      "learning_rate": 6.564828206551973e-07,
+      "loss": 2.3005,
+      "step": 4137
+    },
+    {
+      "epoch": 0.535715441628637,
+      "grad_norm": 1.4103500087613847,
+      "learning_rate": 6.562006993728504e-07,
+      "loss": 2.22,
+      "step": 4138
+    },
+    {
+      "epoch": 0.5358449040359905,
+      "grad_norm": 1.6810955401953784,
+      "learning_rate": 6.559185717171283e-07,
+      "loss": 2.436,
+      "step": 4139
+    },
+    {
+      "epoch": 0.5359743664433441,
+      "grad_norm": 1.6689741274499452,
+      "learning_rate": 6.556364377456612e-07,
+      "loss": 2.321,
+      "step": 4140
+    },
+    {
+      "epoch": 0.5359743664433441,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5423916578292847,
+      "eval_runtime": 14.1164,
+      "eval_samples_per_second": 3.117,
+      "eval_steps_per_second": 0.213,
+      "step": 4140
+    },
+    {
+      "epoch": 0.5361038288506975,
+      "grad_norm": 2.5132137293247543,
+      "learning_rate": 6.553542975160805e-07,
+      "loss": 2.4812,
+      "step": 4141
+    },
+    {
+      "epoch": 0.536233291258051,
+      "grad_norm": 1.6537520547321056,
+      "learning_rate": 6.550721510860196e-07,
+      "loss": 2.3928,
+      "step": 4142
+    },
+    {
+      "epoch": 0.5363627536654044,
+      "grad_norm": 4.443183664121167,
+      "learning_rate": 6.547899985131122e-07,
+      "loss": 2.6509,
+      "step": 4143
+    },
+    {
+      "epoch": 0.5364922160727579,
+      "grad_norm": 2.591581579111484,
+      "learning_rate": 6.545078398549938e-07,
+      "loss": 2.4866,
+      "step": 4144
+    },
+    {
+      "epoch": 0.5366216784801113,
+      "grad_norm": 1.972224143849903,
+      "learning_rate": 6.542256751693009e-07,
+      "loss": 2.5684,
+      "step": 4145
+    },
+    {
+      "epoch": 0.5366216784801113,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.540238857269287,
+      "eval_runtime": 14.4653,
+      "eval_samples_per_second": 3.042,
+      "eval_steps_per_second": 0.207,
+      "step": 4145
+    },
+    {
+      "epoch": 0.5367511408874648,
+      "grad_norm": 2.434647424660376,
+      "learning_rate": 6.539435045136716e-07,
+      "loss": 2.4805,
+      "step": 4146
+    },
+    {
+      "epoch": 0.5368806032948182,
+      "grad_norm": 2.7099396153766784,
+      "learning_rate": 6.536613279457449e-07,
+      "loss": 2.2415,
+      "step": 4147
+    },
+    {
+      "epoch": 0.5370100657021717,
+      "grad_norm": 2.5720628154340246,
+      "learning_rate": 6.533791455231608e-07,
+      "loss": 2.1792,
+      "step": 4148
+    },
+    {
+      "epoch": 0.5371395281095251,
+      "grad_norm": 1.5318086603967571,
+      "learning_rate": 6.530969573035613e-07,
+      "loss": 2.3389,
+      "step": 4149
+    },
+    {
+      "epoch": 0.5372689905168787,
+      "grad_norm": 1.6718272624233645,
+      "learning_rate": 6.528147633445885e-07,
+      "loss": 2.4729,
+      "step": 4150
+    },
+    {
+      "epoch": 0.5372689905168787,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5431240797042847,
+      "eval_runtime": 13.9217,
+      "eval_samples_per_second": 3.161,
+      "eval_steps_per_second": 0.215,
+      "step": 4150
+    },
+    {
+      "epoch": 0.5373984529242322,
+      "grad_norm": 1.6339644167646443,
+      "learning_rate": 6.525325637038867e-07,
+      "loss": 2.5388,
+      "step": 4151
+    },
+    {
+      "epoch": 0.5375279153315856,
+      "grad_norm": 1.699395503065127,
+      "learning_rate": 6.522503584391006e-07,
+      "loss": 2.5259,
+      "step": 4152
+    },
+    {
+      "epoch": 0.5376573777389391,
+      "grad_norm": 1.8218315704132828,
+      "learning_rate": 6.519681476078765e-07,
+      "loss": 2.686,
+      "step": 4153
+    },
+    {
+      "epoch": 0.5377868401462925,
+      "grad_norm": 2.4566200313116506,
+      "learning_rate": 6.516859312678615e-07,
+      "loss": 2.3599,
+      "step": 4154
+    },
+    {
+      "epoch": 0.537916302553646,
+      "grad_norm": 1.6803469253255623,
+      "learning_rate": 6.514037094767042e-07,
+      "loss": 2.3262,
+      "step": 4155
+    },
+    {
+      "epoch": 0.537916302553646,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.539595127105713,
+      "eval_runtime": 14.257,
+      "eval_samples_per_second": 3.086,
+      "eval_steps_per_second": 0.21,
+      "step": 4155
+    },
+    {
+      "epoch": 0.5380457649609994,
+      "grad_norm": 1.5664799659933828,
+      "learning_rate": 6.511214822920542e-07,
+      "loss": 2.4006,
+      "step": 4156
+    },
+    {
+      "epoch": 0.5381752273683529,
+      "grad_norm": 2.8993082389735854,
+      "learning_rate": 6.508392497715619e-07,
+      "loss": 2.6189,
+      "step": 4157
+    },
+    {
+      "epoch": 0.5383046897757063,
+      "grad_norm": 1.356183825550859,
+      "learning_rate": 6.50557011972879e-07,
+      "loss": 2.4287,
+      "step": 4158
+    },
+    {
+      "epoch": 0.5384341521830598,
+      "grad_norm": 1.3362634663670692,
+      "learning_rate": 6.502747689536586e-07,
+      "loss": 2.3394,
+      "step": 4159
+    },
+    {
+      "epoch": 0.5385636145904134,
+      "grad_norm": 1.6283922815925258,
+      "learning_rate": 6.49992520771554e-07,
+      "loss": 2.4009,
+      "step": 4160
+    },
+    {
+      "epoch": 0.5385636145904134,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.542436122894287,
+      "eval_runtime": 14.245,
+      "eval_samples_per_second": 3.089,
+      "eval_steps_per_second": 0.211,
+      "step": 4160
+    },
+    {
+      "epoch": 0.5386930769977668,
+      "grad_norm": 1.5901340799808281,
+      "learning_rate": 6.49710267484221e-07,
+      "loss": 2.426,
+      "step": 4161
+    },
+    {
+      "epoch": 0.5388225394051203,
+      "grad_norm": 1.4668920979501585,
+      "learning_rate": 6.494280091493147e-07,
+      "loss": 2.4539,
+      "step": 4162
+    },
+    {
+      "epoch": 0.5389520018124737,
+      "grad_norm": 1.3498471310082636,
+      "learning_rate": 6.491457458244921e-07,
+      "loss": 2.3096,
+      "step": 4163
+    },
+    {
+      "epoch": 0.5390814642198272,
+      "grad_norm": 1.761705768284713,
+      "learning_rate": 6.488634775674118e-07,
+      "loss": 2.3091,
+      "step": 4164
+    },
+    {
+      "epoch": 0.5392109266271806,
+      "grad_norm": 2.8964785889206657,
+      "learning_rate": 6.485812044357326e-07,
+      "loss": 3.2476,
+      "step": 4165
+    },
+    {
+      "epoch": 0.5392109266271806,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.541748046875,
+      "eval_runtime": 13.8564,
+      "eval_samples_per_second": 3.175,
+      "eval_steps_per_second": 0.217,
+      "step": 4165
+    },
+    {
+      "epoch": 0.5393403890345341,
+      "grad_norm": 1.522326332860526,
+      "learning_rate": 6.482989264871144e-07,
+      "loss": 2.4175,
+      "step": 4166
+    },
+    {
+      "epoch": 0.5394698514418875,
+      "grad_norm": 2.5806067591169173,
+      "learning_rate": 6.48016643779218e-07,
+      "loss": 2.2136,
+      "step": 4167
+    },
+    {
+      "epoch": 0.539599313849241,
+      "grad_norm": 1.6801112313729012,
+      "learning_rate": 6.477343563697057e-07,
+      "loss": 2.5496,
+      "step": 4168
+    },
+    {
+      "epoch": 0.5397287762565945,
+      "grad_norm": 1.2116867136132727,
+      "learning_rate": 6.474520643162403e-07,
+      "loss": 2.2166,
+      "step": 4169
+    },
+    {
+      "epoch": 0.539858238663948,
+      "grad_norm": 3.5642090884438304,
+      "learning_rate": 6.471697676764859e-07,
+      "loss": 2.9839,
+      "step": 4170
+    },
+    {
+      "epoch": 0.539858238663948,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.540771484375,
+      "eval_runtime": 13.4534,
+      "eval_samples_per_second": 3.271,
+      "eval_steps_per_second": 0.223,
+      "step": 4170
+    },
+    {
+      "epoch": 0.5399877010713015,
+      "grad_norm": 1.9239674828001385,
+      "learning_rate": 6.46887466508107e-07,
+      "loss": 2.4031,
+      "step": 4171
+    },
+    {
+      "epoch": 0.5401171634786549,
+      "grad_norm": 3.0916557154110262,
+      "learning_rate": 6.466051608687691e-07,
+      "loss": 2.3848,
+      "step": 4172
+    },
+    {
+      "epoch": 0.5402466258860084,
+      "grad_norm": 2.224857846619929,
+      "learning_rate": 6.463228508161396e-07,
+      "loss": 2.8987,
+      "step": 4173
+    },
+    {
+      "epoch": 0.5403760882933618,
+      "grad_norm": 1.4484856016602865,
+      "learning_rate": 6.460405364078853e-07,
+      "loss": 2.3452,
+      "step": 4174
+    },
+    {
+      "epoch": 0.5405055507007153,
+      "grad_norm": 3.5544299907319408,
+      "learning_rate": 6.457582177016753e-07,
+      "loss": 2.6343,
+      "step": 4175
+    },
+    {
+      "epoch": 0.5405055507007153,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5425026416778564,
+      "eval_runtime": 14.1343,
+      "eval_samples_per_second": 3.113,
+      "eval_steps_per_second": 0.212,
+      "step": 4175
+    },
+    {
+      "epoch": 0.5406350131080687,
+      "grad_norm": 2.4112992276312366,
+      "learning_rate": 6.454758947551783e-07,
+      "loss": 2.7356,
+      "step": 4176
+    },
+    {
+      "epoch": 0.5407644755154222,
+      "grad_norm": 1.174249291916097,
+      "learning_rate": 6.451935676260646e-07,
+      "loss": 2.3081,
+      "step": 4177
+    },
+    {
+      "epoch": 0.5408939379227756,
+      "grad_norm": 3.502429903794039,
+      "learning_rate": 6.449112363720059e-07,
+      "loss": 2.5579,
+      "step": 4178
+    },
+    {
+      "epoch": 0.5410234003301292,
+      "grad_norm": 1.996221574938458,
+      "learning_rate": 6.446289010506733e-07,
+      "loss": 2.5537,
+      "step": 4179
+    },
+    {
+      "epoch": 0.5411528627374826,
+      "grad_norm": 1.6937322360129383,
+      "learning_rate": 6.443465617197399e-07,
+      "loss": 2.2275,
+      "step": 4180
+    },
+    {
+      "epoch": 0.5411528627374826,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5429909229278564,
+      "eval_runtime": 14.6676,
+      "eval_samples_per_second": 3.0,
+      "eval_steps_per_second": 0.205,
+      "step": 4180
+    },
+    {
+      "epoch": 0.5412823251448361,
+      "grad_norm": 1.7482707874539745,
+      "learning_rate": 6.440642184368791e-07,
+      "loss": 2.4971,
+      "step": 4181
+    },
+    {
+      "epoch": 0.5414117875521895,
+      "grad_norm": 2.371147087751884,
+      "learning_rate": 6.437818712597652e-07,
+      "loss": 2.2202,
+      "step": 4182
+    },
+    {
+      "epoch": 0.541541249959543,
+      "grad_norm": 1.2245064008136306,
+      "learning_rate": 6.434995202460734e-07,
+      "loss": 2.3198,
+      "step": 4183
+    },
+    {
+      "epoch": 0.5416707123668965,
+      "grad_norm": 1.4041473304260155,
+      "learning_rate": 6.432171654534798e-07,
+      "loss": 2.4062,
+      "step": 4184
+    },
+    {
+      "epoch": 0.5418001747742499,
+      "grad_norm": 1.953007238042008,
+      "learning_rate": 6.429348069396608e-07,
+      "loss": 2.3882,
+      "step": 4185
+    },
+    {
+      "epoch": 0.5418001747742499,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5472744703292847,
+      "eval_runtime": 15.1198,
+      "eval_samples_per_second": 2.91,
+      "eval_steps_per_second": 0.198,
+      "step": 4185
+    },
+    {
+      "epoch": 0.5419296371816034,
+      "grad_norm": 1.6887981284720208,
+      "learning_rate": 6.426524447622939e-07,
+      "loss": 2.2783,
+      "step": 4186
+    },
+    {
+      "epoch": 0.5420590995889568,
+      "grad_norm": 1.4744105133337015,
+      "learning_rate": 6.423700789790574e-07,
+      "loss": 2.4482,
+      "step": 4187
+    },
+    {
+      "epoch": 0.5421885619963103,
+      "grad_norm": 1.0822522146984528,
+      "learning_rate": 6.420877096476302e-07,
+      "loss": 2.1797,
+      "step": 4188
+    },
+    {
+      "epoch": 0.5423180244036638,
+      "grad_norm": 2.441575103806913,
+      "learning_rate": 6.418053368256921e-07,
+      "loss": 2.8159,
+      "step": 4189
+    },
+    {
+      "epoch": 0.5424474868110173,
+      "grad_norm": 1.4886098520316235,
+      "learning_rate": 6.41522960570923e-07,
+      "loss": 2.3335,
+      "step": 4190
+    },
+    {
+      "epoch": 0.5424474868110173,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5444557666778564,
+      "eval_runtime": 13.9678,
+      "eval_samples_per_second": 3.15,
+      "eval_steps_per_second": 0.215,
+      "step": 4190
+    },
+    {
+      "epoch": 0.5425769492183707,
+      "grad_norm": 1.9406476922778426,
+      "learning_rate": 6.412405809410044e-07,
+      "loss": 2.626,
+      "step": 4191
+    },
+    {
+      "epoch": 0.5427064116257242,
+      "grad_norm": 2.3393791085247666,
+      "learning_rate": 6.409581979936182e-07,
+      "loss": 2.5432,
+      "step": 4192
+    },
+    {
+      "epoch": 0.5428358740330776,
+      "grad_norm": 2.2694496386550984,
+      "learning_rate": 6.406758117864465e-07,
+      "loss": 2.2891,
+      "step": 4193
+    },
+    {
+      "epoch": 0.5429653364404311,
+      "grad_norm": 1.6616169648274188,
+      "learning_rate": 6.403934223771723e-07,
+      "loss": 2.1733,
+      "step": 4194
+    },
+    {
+      "epoch": 0.5430947988477846,
+      "grad_norm": 1.7415433677808254,
+      "learning_rate": 6.401110298234797e-07,
+      "loss": 2.4353,
+      "step": 4195
+    },
+    {
+      "epoch": 0.5430947988477846,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.546586513519287,
+      "eval_runtime": 14.4774,
+      "eval_samples_per_second": 3.039,
+      "eval_steps_per_second": 0.207,
+      "step": 4195
+    },
+    {
+      "epoch": 0.543224261255138,
+      "grad_norm": 2.1996264029862456,
+      "learning_rate": 6.39828634183053e-07,
+      "loss": 2.5642,
+      "step": 4196
+    },
+    {
+      "epoch": 0.5433537236624915,
+      "grad_norm": 2.8892846477758334,
+      "learning_rate": 6.395462355135771e-07,
+      "loss": 2.3782,
+      "step": 4197
+    },
+    {
+      "epoch": 0.5434831860698449,
+      "grad_norm": 2.087094379000283,
+      "learning_rate": 6.39263833872738e-07,
+      "loss": 2.4329,
+      "step": 4198
+    },
+    {
+      "epoch": 0.5436126484771985,
+      "grad_norm": 1.3318643034196034,
+      "learning_rate": 6.389814293182217e-07,
+      "loss": 2.4294,
+      "step": 4199
+    },
+    {
+      "epoch": 0.5437421108845519,
+      "grad_norm": 1.1322866633949251,
+      "learning_rate": 6.386990219077149e-07,
+      "loss": 2.1982,
+      "step": 4200
+    },
+    {
+      "epoch": 0.5437421108845519,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5423694849014282,
+      "eval_runtime": 14.232,
+      "eval_samples_per_second": 3.092,
+      "eval_steps_per_second": 0.211,
+      "step": 4200
+    },
+    {
+      "epoch": 0.5438715732919054,
+      "grad_norm": 2.5152851648724677,
+      "learning_rate": 6.384166116989052e-07,
+      "loss": 2.447,
+      "step": 4201
+    },
+    {
+      "epoch": 0.5440010356992588,
+      "grad_norm": 2.620407039986505,
+      "learning_rate": 6.381341987494806e-07,
+      "loss": 2.3965,
+      "step": 4202
+    },
+    {
+      "epoch": 0.5441304981066123,
+      "grad_norm": 2.748626575249578,
+      "learning_rate": 6.378517831171298e-07,
+      "loss": 2.5415,
+      "step": 4203
+    },
+    {
+      "epoch": 0.5442599605139657,
+      "grad_norm": 4.251181066911142,
+      "learning_rate": 6.375693648595417e-07,
+      "loss": 2.6841,
+      "step": 4204
+    },
+    {
+      "epoch": 0.5443894229213192,
+      "grad_norm": 1.189665140244015,
+      "learning_rate": 6.37286944034406e-07,
+      "loss": 2.3726,
+      "step": 4205
+    },
+    {
+      "epoch": 0.5443894229213192,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5437678098678589,
+      "eval_runtime": 14.3562,
+      "eval_samples_per_second": 3.065,
+      "eval_steps_per_second": 0.209,
+      "step": 4205
+    },
+    {
+      "epoch": 0.5445188853286727,
+      "grad_norm": 2.076593322638028,
+      "learning_rate": 6.370045206994133e-07,
+      "loss": 2.4871,
+      "step": 4206
+    },
+    {
+      "epoch": 0.5446483477360261,
+      "grad_norm": 1.4388034624808688,
+      "learning_rate": 6.367220949122536e-07,
+      "loss": 2.3867,
+      "step": 4207
+    },
+    {
+      "epoch": 0.5447778101433797,
+      "grad_norm": 2.493220957586294,
+      "learning_rate": 6.364396667306186e-07,
+      "loss": 2.6929,
+      "step": 4208
+    },
+    {
+      "epoch": 0.5449072725507331,
+      "grad_norm": 3.426691707669274,
+      "learning_rate": 6.361572362121997e-07,
+      "loss": 2.447,
+      "step": 4209
+    },
+    {
+      "epoch": 0.5450367349580866,
+      "grad_norm": 1.4371328325093933,
+      "learning_rate": 6.35874803414689e-07,
+      "loss": 2.3025,
+      "step": 4210
+    },
+    {
+      "epoch": 0.5450367349580866,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5452326536178589,
+      "eval_runtime": 14.5668,
+      "eval_samples_per_second": 3.021,
+      "eval_steps_per_second": 0.206,
+      "step": 4210
+    },
+    {
+      "epoch": 0.54516619736544,
+      "grad_norm": 1.6803092972037372,
+      "learning_rate": 6.355923683957796e-07,
+      "loss": 2.4741,
+      "step": 4211
+    },
+    {
+      "epoch": 0.5452956597727935,
+      "grad_norm": 1.5433213481873256,
+      "learning_rate": 6.353099312131641e-07,
+      "loss": 2.47,
+      "step": 4212
+    },
+    {
+      "epoch": 0.5454251221801469,
+      "grad_norm": 1.6174433267701191,
+      "learning_rate": 6.35027491924536e-07,
+      "loss": 2.5542,
+      "step": 4213
+    },
+    {
+      "epoch": 0.5455545845875004,
+      "grad_norm": 2.8481688644180476,
+      "learning_rate": 6.347450505875898e-07,
+      "loss": 2.5293,
+      "step": 4214
+    },
+    {
+      "epoch": 0.5456840469948538,
+      "grad_norm": 1.408684476527369,
+      "learning_rate": 6.344626072600192e-07,
+      "loss": 2.2739,
+      "step": 4215
+    },
+    {
+      "epoch": 0.5456840469948538,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5416592359542847,
+      "eval_runtime": 14.4188,
+      "eval_samples_per_second": 3.052,
+      "eval_steps_per_second": 0.208,
+      "step": 4215
+    },
+    {
+      "epoch": 0.5458135094022073,
+      "grad_norm": 1.4355391005547333,
+      "learning_rate": 6.341801619995192e-07,
+      "loss": 2.1799,
+      "step": 4216
+    },
+    {
+      "epoch": 0.5459429718095608,
+      "grad_norm": 1.6207047158663324,
+      "learning_rate": 6.338977148637854e-07,
+      "loss": 2.459,
+      "step": 4217
+    },
+    {
+      "epoch": 0.5460724342169143,
+      "grad_norm": 1.9303170020412486,
+      "learning_rate": 6.336152659105127e-07,
+      "loss": 2.5396,
+      "step": 4218
+    },
+    {
+      "epoch": 0.5462018966242678,
+      "grad_norm": 2.3186850222738093,
+      "learning_rate": 6.333328151973974e-07,
+      "loss": 2.3923,
+      "step": 4219
+    },
+    {
+      "epoch": 0.5463313590316212,
+      "grad_norm": 2.0628762619456857,
+      "learning_rate": 6.330503627821358e-07,
+      "loss": 2.7161,
+      "step": 4220
+    },
+    {
+      "epoch": 0.5463313590316212,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5359108448028564,
+      "eval_runtime": 13.5612,
+      "eval_samples_per_second": 3.245,
+      "eval_steps_per_second": 0.221,
+      "step": 4220
+    },
+    {
+      "epoch": 0.5464608214389747,
+      "grad_norm": 1.7857025300073248,
+      "learning_rate": 6.327679087224244e-07,
+      "loss": 2.5642,
+      "step": 4221
+    },
+    {
+      "epoch": 0.5465902838463281,
+      "grad_norm": 1.8159878219151797,
+      "learning_rate": 6.324854530759598e-07,
+      "loss": 2.5762,
+      "step": 4222
+    },
+    {
+      "epoch": 0.5467197462536816,
+      "grad_norm": 1.0916356142439672,
+      "learning_rate": 6.322029959004403e-07,
+      "loss": 2.1165,
+      "step": 4223
+    },
+    {
+      "epoch": 0.546849208661035,
+      "grad_norm": 1.6807476770558634,
+      "learning_rate": 6.319205372535625e-07,
+      "loss": 2.4902,
+      "step": 4224
+    },
+    {
+      "epoch": 0.5469786710683885,
+      "grad_norm": 1.296506550576479,
+      "learning_rate": 6.316380771930249e-07,
+      "loss": 2.2305,
+      "step": 4225
+    },
+    {
+      "epoch": 0.5469786710683885,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5417258739471436,
+      "eval_runtime": 14.1958,
+      "eval_samples_per_second": 3.1,
+      "eval_steps_per_second": 0.211,
+      "step": 4225
+    },
+    {
+      "epoch": 0.5471081334757419,
+      "grad_norm": 4.201644742027217,
+      "learning_rate": 6.313556157765252e-07,
+      "loss": 2.5933,
+      "step": 4226
+    },
+    {
+      "epoch": 0.5472375958830954,
+      "grad_norm": 2.0314146327915976,
+      "learning_rate": 6.310731530617624e-07,
+      "loss": 2.5894,
+      "step": 4227
+    },
+    {
+      "epoch": 0.547367058290449,
+      "grad_norm": 1.8474975378034588,
+      "learning_rate": 6.307906891064349e-07,
+      "loss": 2.3545,
+      "step": 4228
+    },
+    {
+      "epoch": 0.5474965206978024,
+      "grad_norm": 1.6345361699498737,
+      "learning_rate": 6.305082239682418e-07,
+      "loss": 2.5654,
+      "step": 4229
+    },
+    {
+      "epoch": 0.5476259831051559,
+      "grad_norm": 2.105173565976692,
+      "learning_rate": 6.302257577048824e-07,
+      "loss": 2.5134,
+      "step": 4230
+    },
+    {
+      "epoch": 0.5476259831051559,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5423251390457153,
+      "eval_runtime": 14.2269,
+      "eval_samples_per_second": 3.093,
+      "eval_steps_per_second": 0.211,
+      "step": 4230
+    },
+    {
+      "epoch": 0.5477554455125093,
+      "grad_norm": 1.48930641554959,
+      "learning_rate": 6.299432903740559e-07,
+      "loss": 2.4819,
+      "step": 4231
+    },
+    {
+      "epoch": 0.5478849079198628,
+      "grad_norm": 1.2684982608570394,
+      "learning_rate": 6.296608220334623e-07,
+      "loss": 2.3135,
+      "step": 4232
+    },
+    {
+      "epoch": 0.5480143703272162,
+      "grad_norm": 1.4004164751858577,
+      "learning_rate": 6.293783527408011e-07,
+      "loss": 2.3877,
+      "step": 4233
+    },
+    {
+      "epoch": 0.5481438327345697,
+      "grad_norm": 2.467489378204922,
+      "learning_rate": 6.290958825537729e-07,
+      "loss": 2.5466,
+      "step": 4234
+    },
+    {
+      "epoch": 0.5482732951419231,
+      "grad_norm": 2.37339525455699,
+      "learning_rate": 6.288134115300776e-07,
+      "loss": 2.5435,
+      "step": 4235
+    },
+    {
+      "epoch": 0.5482732951419231,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.541703701019287,
+      "eval_runtime": 14.5334,
+      "eval_samples_per_second": 3.028,
+      "eval_steps_per_second": 0.206,
+      "step": 4235
+    },
+    {
+      "epoch": 0.5484027575492766,
+      "grad_norm": 1.5220531485817708,
+      "learning_rate": 6.285309397274156e-07,
+      "loss": 2.3633,
+      "step": 4236
+    },
+    {
+      "epoch": 0.54853221995663,
+      "grad_norm": 1.8425814639014737,
+      "learning_rate": 6.282484672034875e-07,
+      "loss": 2.4744,
+      "step": 4237
+    },
+    {
+      "epoch": 0.5486616823639836,
+      "grad_norm": 1.8795602805475373,
+      "learning_rate": 6.279659940159944e-07,
+      "loss": 2.3457,
+      "step": 4238
+    },
+    {
+      "epoch": 0.5487911447713371,
+      "grad_norm": 3.2850836472195053,
+      "learning_rate": 6.276835202226368e-07,
+      "loss": 2.6023,
+      "step": 4239
+    },
+    {
+      "epoch": 0.5489206071786905,
+      "grad_norm": 1.9113612809426985,
+      "learning_rate": 6.274010458811155e-07,
+      "loss": 2.3223,
+      "step": 4240
+    },
+    {
+      "epoch": 0.5489206071786905,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.543257236480713,
+      "eval_runtime": 14.1931,
+      "eval_samples_per_second": 3.1,
+      "eval_steps_per_second": 0.211,
+      "step": 4240
+    },
+    {
+      "epoch": 0.549050069586044,
+      "grad_norm": 1.3494014291103476,
+      "learning_rate": 6.271185710491322e-07,
+      "loss": 2.405,
+      "step": 4241
+    },
+    {
+      "epoch": 0.5491795319933974,
+      "grad_norm": 1.619297404680498,
+      "learning_rate": 6.268360957843879e-07,
+      "loss": 2.3254,
+      "step": 4242
+    },
+    {
+      "epoch": 0.5493089944007509,
+      "grad_norm": 2.5865667349566017,
+      "learning_rate": 6.265536201445837e-07,
+      "loss": 2.3953,
+      "step": 4243
+    },
+    {
+      "epoch": 0.5494384568081043,
+      "grad_norm": 1.891944822875691,
+      "learning_rate": 6.262711441874212e-07,
+      "loss": 2.4565,
+      "step": 4244
+    },
+    {
+      "epoch": 0.5495679192154578,
+      "grad_norm": 1.3004415238380485,
+      "learning_rate": 6.259886679706015e-07,
+      "loss": 2.3877,
+      "step": 4245
+    },
+    {
+      "epoch": 0.5495679192154578,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.545210361480713,
+      "eval_runtime": 15.0134,
+      "eval_samples_per_second": 2.931,
+      "eval_steps_per_second": 0.2,
+      "step": 4245
+    },
+    {
+      "epoch": 0.5496973816228112,
+      "grad_norm": 1.776819784881242,
+      "learning_rate": 6.257061915518264e-07,
+      "loss": 2.302,
+      "step": 4246
+    },
+    {
+      "epoch": 0.5498268440301648,
+      "grad_norm": 2.280689985617081,
+      "learning_rate": 6.254237149887975e-07,
+      "loss": 2.4583,
+      "step": 4247
+    },
+    {
+      "epoch": 0.5499563064375182,
+      "grad_norm": 2.7203691094558065,
+      "learning_rate": 6.251412383392162e-07,
+      "loss": 2.5894,
+      "step": 4248
+    },
+    {
+      "epoch": 0.5500857688448717,
+      "grad_norm": 2.969062585147217,
+      "learning_rate": 6.248587616607841e-07,
+      "loss": 2.4214,
+      "step": 4249
+    },
+    {
+      "epoch": 0.5502152312522252,
+      "grad_norm": 1.296427052441223,
+      "learning_rate": 6.245762850112027e-07,
+      "loss": 2.1907,
+      "step": 4250
+    },
+    {
+      "epoch": 0.5502152312522252,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5416370630264282,
+      "eval_runtime": 13.8872,
+      "eval_samples_per_second": 3.168,
+      "eval_steps_per_second": 0.216,
+      "step": 4250
+    },
+    {
+      "epoch": 0.5503446936595786,
+      "grad_norm": 1.402864582983678,
+      "learning_rate": 6.242938084481737e-07,
+      "loss": 2.3792,
+      "step": 4251
+    },
+    {
+      "epoch": 0.5504741560669321,
+      "grad_norm": 1.3751387286957826,
+      "learning_rate": 6.240113320293987e-07,
+      "loss": 2.3184,
+      "step": 4252
+    },
+    {
+      "epoch": 0.5506036184742855,
+      "grad_norm": 2.1869111357667874,
+      "learning_rate": 6.237288558125792e-07,
+      "loss": 2.3779,
+      "step": 4253
+    },
+    {
+      "epoch": 0.550733080881639,
+      "grad_norm": 1.5283618607460008,
+      "learning_rate": 6.234463798554164e-07,
+      "loss": 2.1719,
+      "step": 4254
+    },
+    {
+      "epoch": 0.5508625432889924,
+      "grad_norm": 1.4464525499124974,
+      "learning_rate": 6.231639042156124e-07,
+      "loss": 2.2969,
+      "step": 4255
+    },
+    {
+      "epoch": 0.5508625432889924,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5423916578292847,
+      "eval_runtime": 14.0399,
+      "eval_samples_per_second": 3.134,
+      "eval_steps_per_second": 0.214,
+      "step": 4255
+    },
+    {
+      "epoch": 0.5509920056963459,
+      "grad_norm": 1.5550531398759564,
+      "learning_rate": 6.22881428950868e-07,
+      "loss": 2.3955,
+      "step": 4256
+    },
+    {
+      "epoch": 0.5511214681036994,
+      "grad_norm": 2.0309621929565633,
+      "learning_rate": 6.225989541188847e-07,
+      "loss": 2.4414,
+      "step": 4257
+    },
+    {
+      "epoch": 0.5512509305110529,
+      "grad_norm": 1.5501826796899372,
+      "learning_rate": 6.223164797773634e-07,
+      "loss": 2.3711,
+      "step": 4258
+    },
+    {
+      "epoch": 0.5513803929184063,
+      "grad_norm": 2.692452107080833,
+      "learning_rate": 6.220340059840057e-07,
+      "loss": 2.4541,
+      "step": 4259
+    },
+    {
+      "epoch": 0.5515098553257598,
+      "grad_norm": 1.963313779702523,
+      "learning_rate": 6.217515327965128e-07,
+      "loss": 2.6089,
+      "step": 4260
+    },
+    {
+      "epoch": 0.5515098553257598,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5409489870071411,
+      "eval_runtime": 14.1201,
+      "eval_samples_per_second": 3.116,
+      "eval_steps_per_second": 0.212,
+      "step": 4260
+    },
+    {
+      "epoch": 0.5516393177331133,
+      "grad_norm": 1.050193208919441,
+      "learning_rate": 6.214690602725848e-07,
+      "loss": 2.2222,
+      "step": 4261
+    },
+    {
+      "epoch": 0.5517687801404667,
+      "grad_norm": 1.3499979398460218,
+      "learning_rate": 6.211865884699227e-07,
+      "loss": 2.3208,
+      "step": 4262
+    },
+    {
+      "epoch": 0.5518982425478202,
+      "grad_norm": 1.5321107954241715,
+      "learning_rate": 6.209041174462273e-07,
+      "loss": 2.3252,
+      "step": 4263
+    },
+    {
+      "epoch": 0.5520277049551736,
+      "grad_norm": 2.990441835117077,
+      "learning_rate": 6.206216472591989e-07,
+      "loss": 2.4578,
+      "step": 4264
+    },
+    {
+      "epoch": 0.5521571673625271,
+      "grad_norm": 1.559652379821359,
+      "learning_rate": 6.203391779665378e-07,
+      "loss": 2.3167,
+      "step": 4265
+    },
+    {
+      "epoch": 0.5521571673625271,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5409934520721436,
+      "eval_runtime": 13.6904,
+      "eval_samples_per_second": 3.214,
+      "eval_steps_per_second": 0.219,
+      "step": 4265
+    },
+    {
+      "epoch": 0.5522866297698805,
+      "grad_norm": 1.8119123071096452,
+      "learning_rate": 6.200567096259442e-07,
+      "loss": 2.5085,
+      "step": 4266
+    },
+    {
+      "epoch": 0.5524160921772341,
+      "grad_norm": 1.5331208358071444,
+      "learning_rate": 6.197742422951179e-07,
+      "loss": 2.3789,
+      "step": 4267
+    },
+    {
+      "epoch": 0.5525455545845875,
+      "grad_norm": 3.4507193190543246,
+      "learning_rate": 6.194917760317583e-07,
+      "loss": 2.6619,
+      "step": 4268
+    },
+    {
+      "epoch": 0.552675016991941,
+      "grad_norm": 1.7432181431325713,
+      "learning_rate": 6.192093108935652e-07,
+      "loss": 2.5391,
+      "step": 4269
+    },
+    {
+      "epoch": 0.5528044793992944,
+      "grad_norm": 1.0047514710300003,
+      "learning_rate": 6.189268469382377e-07,
+      "loss": 2.1633,
+      "step": 4270
+    },
+    {
+      "epoch": 0.5528044793992944,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5408824682235718,
+      "eval_runtime": 14.2773,
+      "eval_samples_per_second": 3.082,
+      "eval_steps_per_second": 0.21,
+      "step": 4270
+    },
+    {
+      "epoch": 0.5529339418066479,
+      "grad_norm": 2.847447302301485,
+      "learning_rate": 6.18644384223475e-07,
+      "loss": 2.4878,
+      "step": 4271
+    },
+    {
+      "epoch": 0.5530634042140014,
+      "grad_norm": 3.564824162049398,
+      "learning_rate": 6.183619228069754e-07,
+      "loss": 2.5781,
+      "step": 4272
+    },
+    {
+      "epoch": 0.5531928666213548,
+      "grad_norm": 2.190422739362663,
+      "learning_rate": 6.180794627464376e-07,
+      "loss": 2.5781,
+      "step": 4273
+    },
+    {
+      "epoch": 0.5533223290287083,
+      "grad_norm": 3.335455848514011,
+      "learning_rate": 6.1779700409956e-07,
+      "loss": 2.6113,
+      "step": 4274
+    },
+    {
+      "epoch": 0.5534517914360617,
+      "grad_norm": 3.240592517370985,
+      "learning_rate": 6.175145469240402e-07,
+      "loss": 2.437,
+      "step": 4275
+    },
+    {
+      "epoch": 0.5534517914360617,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5424582958221436,
+      "eval_runtime": 14.1941,
+      "eval_samples_per_second": 3.1,
+      "eval_steps_per_second": 0.211,
+      "step": 4275
+    },
+    {
+      "epoch": 0.5535812538434152,
+      "grad_norm": 1.1368239219851894,
+      "learning_rate": 6.172320912775759e-07,
+      "loss": 2.3291,
+      "step": 4276
+    },
+    {
+      "epoch": 0.5537107162507687,
+      "grad_norm": 1.6462575268764186,
+      "learning_rate": 6.169496372178645e-07,
+      "loss": 2.5496,
+      "step": 4277
+    },
+    {
+      "epoch": 0.5538401786581222,
+      "grad_norm": 2.7383191619532736,
+      "learning_rate": 6.166671848026027e-07,
+      "loss": 2.5112,
+      "step": 4278
+    },
+    {
+      "epoch": 0.5539696410654756,
+      "grad_norm": 1.3810044211872197,
+      "learning_rate": 6.163847340894875e-07,
+      "loss": 2.3259,
+      "step": 4279
+    },
+    {
+      "epoch": 0.5540991034728291,
+      "grad_norm": 1.3782291577102648,
+      "learning_rate": 6.161022851362149e-07,
+      "loss": 2.3877,
+      "step": 4280
+    },
+    {
+      "epoch": 0.5540991034728291,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5381081104278564,
+      "eval_runtime": 13.7889,
+      "eval_samples_per_second": 3.191,
+      "eval_steps_per_second": 0.218,
+      "step": 4280
+    },
+    {
+      "epoch": 0.5542285658801825,
+      "grad_norm": 1.6345559226615984,
+      "learning_rate": 6.158198380004807e-07,
+      "loss": 2.3025,
+      "step": 4281
+    },
+    {
+      "epoch": 0.554358028287536,
+      "grad_norm": 2.134404514212594,
+      "learning_rate": 6.155373927399811e-07,
+      "loss": 2.6165,
+      "step": 4282
+    },
+    {
+      "epoch": 0.5544874906948895,
+      "grad_norm": 1.8754157751704479,
+      "learning_rate": 6.152549494124105e-07,
+      "loss": 2.4297,
+      "step": 4283
+    },
+    {
+      "epoch": 0.5546169531022429,
+      "grad_norm": 1.4589409114717964,
+      "learning_rate": 6.149725080754641e-07,
+      "loss": 2.293,
+      "step": 4284
+    },
+    {
+      "epoch": 0.5547464155095964,
+      "grad_norm": 2.3647478201961722,
+      "learning_rate": 6.146900687868361e-07,
+      "loss": 2.3967,
+      "step": 4285
+    },
+    {
+      "epoch": 0.5547464155095964,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5380192995071411,
+      "eval_runtime": 14.1434,
+      "eval_samples_per_second": 3.111,
+      "eval_steps_per_second": 0.212,
+      "step": 4285
+    },
+    {
+      "epoch": 0.5548758779169498,
+      "grad_norm": 1.8037358080536374,
+      "learning_rate": 6.144076316042206e-07,
+      "loss": 2.5027,
+      "step": 4286
+    },
+    {
+      "epoch": 0.5550053403243034,
+      "grad_norm": 2.010237073853137,
+      "learning_rate": 6.14125196585311e-07,
+      "loss": 2.3926,
+      "step": 4287
+    },
+    {
+      "epoch": 0.5551348027316568,
+      "grad_norm": 2.028880778931612,
+      "learning_rate": 6.138427637878007e-07,
+      "loss": 2.2804,
+      "step": 4288
+    },
+    {
+      "epoch": 0.5552642651390103,
+      "grad_norm": 1.6381121289506992,
+      "learning_rate": 6.135603332693816e-07,
+      "loss": 2.3596,
+      "step": 4289
+    },
+    {
+      "epoch": 0.5553937275463637,
+      "grad_norm": 1.155760356910515,
+      "learning_rate": 6.132779050877465e-07,
+      "loss": 2.2751,
+      "step": 4290
+    },
+    {
+      "epoch": 0.5553937275463637,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5388849973678589,
+      "eval_runtime": 14.3499,
+      "eval_samples_per_second": 3.066,
+      "eval_steps_per_second": 0.209,
+      "step": 4290
+    },
+    {
+      "epoch": 0.5555231899537172,
+      "grad_norm": 1.4974920409708508,
+      "learning_rate": 6.129954793005868e-07,
+      "loss": 2.4294,
+      "step": 4291
+    },
+    {
+      "epoch": 0.5556526523610706,
+      "grad_norm": 1.6028331554743795,
+      "learning_rate": 6.127130559655938e-07,
+      "loss": 2.47,
+      "step": 4292
+    },
+    {
+      "epoch": 0.5557821147684241,
+      "grad_norm": 2.846220988905065,
+      "learning_rate": 6.124306351404583e-07,
+      "loss": 2.5627,
+      "step": 4293
+    },
+    {
+      "epoch": 0.5559115771757775,
+      "grad_norm": 1.5526617947905548,
+      "learning_rate": 6.121482168828704e-07,
+      "loss": 2.2983,
+      "step": 4294
+    },
+    {
+      "epoch": 0.556041039583131,
+      "grad_norm": 2.3601307970385452,
+      "learning_rate": 6.118658012505195e-07,
+      "loss": 2.4534,
+      "step": 4295
+    },
+    {
+      "epoch": 0.556041039583131,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5347567796707153,
+      "eval_runtime": 14.4575,
+      "eval_samples_per_second": 3.043,
+      "eval_steps_per_second": 0.208,
+      "step": 4295
+    },
+    {
+      "epoch": 0.5561705019904846,
+      "grad_norm": 1.0591689115265042,
+      "learning_rate": 6.11583388301095e-07,
+      "loss": 2.1467,
+      "step": 4296
+    },
+    {
+      "epoch": 0.556299964397838,
+      "grad_norm": 3.093537640550959,
+      "learning_rate": 6.113009780922853e-07,
+      "loss": 2.8789,
+      "step": 4297
+    },
+    {
+      "epoch": 0.5564294268051915,
+      "grad_norm": 1.4199783428734807,
+      "learning_rate": 6.110185706817785e-07,
+      "loss": 2.5833,
+      "step": 4298
+    },
+    {
+      "epoch": 0.5565588892125449,
+      "grad_norm": 1.7854911614667743,
+      "learning_rate": 6.107361661272622e-07,
+      "loss": 2.6311,
+      "step": 4299
+    },
+    {
+      "epoch": 0.5566883516198984,
+      "grad_norm": 1.692967792531868,
+      "learning_rate": 6.10453764486423e-07,
+      "loss": 2.4094,
+      "step": 4300
+    },
+    {
+      "epoch": 0.5566883516198984,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.535888671875,
+      "eval_runtime": 13.742,
+      "eval_samples_per_second": 3.202,
+      "eval_steps_per_second": 0.218,
+      "step": 4300
+    },
+    {
+      "epoch": 0.5568178140272518,
+      "grad_norm": 1.621198340454262,
+      "learning_rate": 6.101713658169471e-07,
+      "loss": 2.5679,
+      "step": 4301
+    },
+    {
+      "epoch": 0.5569472764346053,
+      "grad_norm": 1.7649924775661157,
+      "learning_rate": 6.098889701765204e-07,
+      "loss": 2.3979,
+      "step": 4302
+    },
+    {
+      "epoch": 0.5570767388419587,
+      "grad_norm": 1.9243137756792954,
+      "learning_rate": 6.096065776228279e-07,
+      "loss": 2.2397,
+      "step": 4303
+    },
+    {
+      "epoch": 0.5572062012493122,
+      "grad_norm": 1.1067917529171238,
+      "learning_rate": 6.093241882135536e-07,
+      "loss": 2.2673,
+      "step": 4304
+    },
+    {
+      "epoch": 0.5573356636566656,
+      "grad_norm": 1.8037687154397128,
+      "learning_rate": 6.090418020063821e-07,
+      "loss": 2.3108,
+      "step": 4305
+    },
+    {
+      "epoch": 0.5573356636566656,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5361106395721436,
+      "eval_runtime": 13.5547,
+      "eval_samples_per_second": 3.246,
+      "eval_steps_per_second": 0.221,
+      "step": 4305
+    },
+    {
+      "epoch": 0.5574651260640192,
+      "grad_norm": 1.6941644609536064,
+      "learning_rate": 6.087594190589957e-07,
+      "loss": 2.4482,
+      "step": 4306
+    },
+    {
+      "epoch": 0.5575945884713727,
+      "grad_norm": 1.6380269526534845,
+      "learning_rate": 6.084770394290772e-07,
+      "loss": 2.3022,
+      "step": 4307
+    },
+    {
+      "epoch": 0.5577240508787261,
+      "grad_norm": 2.0084433010521,
+      "learning_rate": 6.081946631743081e-07,
+      "loss": 2.6982,
+      "step": 4308
+    },
+    {
+      "epoch": 0.5578535132860796,
+      "grad_norm": 2.01592731315132,
+      "learning_rate": 6.079122903523699e-07,
+      "loss": 2.4358,
+      "step": 4309
+    },
+    {
+      "epoch": 0.557982975693433,
+      "grad_norm": 1.0106757896882663,
+      "learning_rate": 6.076299210209429e-07,
+      "loss": 2.167,
+      "step": 4310
+    },
+    {
+      "epoch": 0.557982975693433,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5338467359542847,
+      "eval_runtime": 13.6083,
+      "eval_samples_per_second": 3.233,
+      "eval_steps_per_second": 0.22,
+      "step": 4310
+    },
+    {
+      "epoch": 0.5581124381007865,
+      "grad_norm": 1.280734552829846,
+      "learning_rate": 6.073475552377065e-07,
+      "loss": 2.4146,
+      "step": 4311
+    },
+    {
+      "epoch": 0.5582419005081399,
+      "grad_norm": 1.670267584144987,
+      "learning_rate": 6.070651930603394e-07,
+      "loss": 2.2913,
+      "step": 4312
+    },
+    {
+      "epoch": 0.5583713629154934,
+      "grad_norm": 1.7265189822582097,
+      "learning_rate": 6.067828345465203e-07,
+      "loss": 2.3843,
+      "step": 4313
+    },
+    {
+      "epoch": 0.5585008253228468,
+      "grad_norm": 2.002063738470711,
+      "learning_rate": 6.065004797539266e-07,
+      "loss": 2.3066,
+      "step": 4314
+    },
+    {
+      "epoch": 0.5586302877302003,
+      "grad_norm": 1.9168009317211723,
+      "learning_rate": 6.062181287402349e-07,
+      "loss": 2.5767,
+      "step": 4315
+    },
+    {
+      "epoch": 0.5586302877302003,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5359553098678589,
+      "eval_runtime": 14.0475,
+      "eval_samples_per_second": 3.132,
+      "eval_steps_per_second": 0.214,
+      "step": 4315
+    },
+    {
+      "epoch": 0.5587597501375539,
+      "grad_norm": 1.403384796059386,
+      "learning_rate": 6.059357815631211e-07,
+      "loss": 2.207,
+      "step": 4316
+    },
+    {
+      "epoch": 0.5588892125449073,
+      "grad_norm": 4.448100370827739,
+      "learning_rate": 6.056534382802604e-07,
+      "loss": 2.75,
+      "step": 4317
+    },
+    {
+      "epoch": 0.5590186749522608,
+      "grad_norm": 2.121133381597139,
+      "learning_rate": 6.053710989493268e-07,
+      "loss": 2.4905,
+      "step": 4318
+    },
+    {
+      "epoch": 0.5591481373596142,
+      "grad_norm": 6.075048779560015,
+      "learning_rate": 6.050887636279943e-07,
+      "loss": 3.207,
+      "step": 4319
+    },
+    {
+      "epoch": 0.5592775997669677,
+      "grad_norm": 1.8484663206337315,
+      "learning_rate": 6.048064323739353e-07,
+      "loss": 2.4114,
+      "step": 4320
+    },
+    {
+      "epoch": 0.5592775997669677,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.533891201019287,
+      "eval_runtime": 14.5707,
+      "eval_samples_per_second": 3.02,
+      "eval_steps_per_second": 0.206,
+      "step": 4320
+    },
+    {
+      "epoch": 0.5594070621743211,
+      "grad_norm": 4.268453327901289,
+      "learning_rate": 6.04524105244822e-07,
+      "loss": 2.7529,
+      "step": 4321
+    },
+    {
+      "epoch": 0.5595365245816746,
+      "grad_norm": 2.1370555224338714,
+      "learning_rate": 6.04241782298325e-07,
+      "loss": 2.3604,
+      "step": 4322
+    },
+    {
+      "epoch": 0.559665986989028,
+      "grad_norm": 2.2960287423991486,
+      "learning_rate": 6.039594635921149e-07,
+      "loss": 2.3677,
+      "step": 4323
+    },
+    {
+      "epoch": 0.5597954493963815,
+      "grad_norm": 2.8704269197703756,
+      "learning_rate": 6.036771491838607e-07,
+      "loss": 2.4443,
+      "step": 4324
+    },
+    {
+      "epoch": 0.5599249118037349,
+      "grad_norm": 1.4825200247912658,
+      "learning_rate": 6.03394839131231e-07,
+      "loss": 2.5254,
+      "step": 4325
+    },
+    {
+      "epoch": 0.5599249118037349,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5359996557235718,
+      "eval_runtime": 14.266,
+      "eval_samples_per_second": 3.084,
+      "eval_steps_per_second": 0.21,
+      "step": 4325
+    },
+    {
+      "epoch": 0.5600543742110885,
+      "grad_norm": 1.442340105331669,
+      "learning_rate": 6.031125334918933e-07,
+      "loss": 2.2185,
+      "step": 4326
+    },
+    {
+      "epoch": 0.560183836618442,
+      "grad_norm": 1.2449860375339152,
+      "learning_rate": 6.028302323235144e-07,
+      "loss": 2.3076,
+      "step": 4327
+    },
+    {
+      "epoch": 0.5603132990257954,
+      "grad_norm": 1.7380514771513373,
+      "learning_rate": 6.025479356837598e-07,
+      "loss": 2.1694,
+      "step": 4328
+    },
+    {
+      "epoch": 0.5604427614331489,
+      "grad_norm": 2.657681475032726,
+      "learning_rate": 6.022656436302944e-07,
+      "loss": 2.8411,
+      "step": 4329
+    },
+    {
+      "epoch": 0.5605722238405023,
+      "grad_norm": 2.0844999473675343,
+      "learning_rate": 6.019833562207821e-07,
+      "loss": 2.3096,
+      "step": 4330
+    },
+    {
+      "epoch": 0.5605722238405023,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5352894067764282,
+      "eval_runtime": 14.2294,
+      "eval_samples_per_second": 3.092,
+      "eval_steps_per_second": 0.211,
+      "step": 4330
+    },
+    {
+      "epoch": 0.5607016862478558,
+      "grad_norm": 1.5067865702213148,
+      "learning_rate": 6.017010735128857e-07,
+      "loss": 2.4738,
+      "step": 4331
+    },
+    {
+      "epoch": 0.5608311486552092,
+      "grad_norm": 2.1401722891417294,
+      "learning_rate": 6.014187955642677e-07,
+      "loss": 2.3574,
+      "step": 4332
+    },
+    {
+      "epoch": 0.5609606110625627,
+      "grad_norm": 1.7115771031567406,
+      "learning_rate": 6.011365224325884e-07,
+      "loss": 2.4009,
+      "step": 4333
+    },
+    {
+      "epoch": 0.5610900734699161,
+      "grad_norm": 1.6361342118722764,
+      "learning_rate": 6.008542541755081e-07,
+      "loss": 2.3208,
+      "step": 4334
+    },
+    {
+      "epoch": 0.5612195358772697,
+      "grad_norm": 1.356496895199791,
+      "learning_rate": 6.005719908506856e-07,
+      "loss": 2.4309,
+      "step": 4335
+    },
+    {
+      "epoch": 0.5612195358772697,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5339133739471436,
+      "eval_runtime": 15.2556,
+      "eval_samples_per_second": 2.884,
+      "eval_steps_per_second": 0.197,
+      "step": 4335
+    },
+    {
+      "epoch": 0.5613489982846231,
+      "grad_norm": 1.221308370546581,
+      "learning_rate": 6.002897325157793e-07,
+      "loss": 2.1519,
+      "step": 4336
+    },
+    {
+      "epoch": 0.5614784606919766,
+      "grad_norm": 2.675358818356185,
+      "learning_rate": 6.00007479228446e-07,
+      "loss": 2.7185,
+      "step": 4337
+    },
+    {
+      "epoch": 0.56160792309933,
+      "grad_norm": 1.8337408764339245,
+      "learning_rate": 5.997252310463418e-07,
+      "loss": 2.47,
+      "step": 4338
+    },
+    {
+      "epoch": 0.5617373855066835,
+      "grad_norm": 1.7740760470384846,
+      "learning_rate": 5.994429880271211e-07,
+      "loss": 2.559,
+      "step": 4339
+    },
+    {
+      "epoch": 0.561866847914037,
+      "grad_norm": 1.927484913934167,
+      "learning_rate": 5.991607502284383e-07,
+      "loss": 2.5396,
+      "step": 4340
+    },
+    {
+      "epoch": 0.561866847914037,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.532470703125,
+      "eval_runtime": 13.7042,
+      "eval_samples_per_second": 3.211,
+      "eval_steps_per_second": 0.219,
+      "step": 4340
+    },
+    {
+      "epoch": 0.5619963103213904,
+      "grad_norm": 1.7099284959469516,
+      "learning_rate": 5.988785177079459e-07,
+      "loss": 2.457,
+      "step": 4341
+    },
+    {
+      "epoch": 0.5621257727287439,
+      "grad_norm": 4.647874236406694,
+      "learning_rate": 5.985962905232959e-07,
+      "loss": 2.416,
+      "step": 4342
+    },
+    {
+      "epoch": 0.5622552351360973,
+      "grad_norm": 1.9542066969976484,
+      "learning_rate": 5.983140687321386e-07,
+      "loss": 2.5364,
+      "step": 4343
+    },
+    {
+      "epoch": 0.5623846975434508,
+      "grad_norm": 1.3377028975600906,
+      "learning_rate": 5.980318523921238e-07,
+      "loss": 2.3984,
+      "step": 4344
+    },
+    {
+      "epoch": 0.5625141599508043,
+      "grad_norm": 2.450650803461354,
+      "learning_rate": 5.977496415608997e-07,
+      "loss": 2.5984,
+      "step": 4345
+    },
+    {
+      "epoch": 0.5625141599508043,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5274325609207153,
+      "eval_runtime": 14.2484,
+      "eval_samples_per_second": 3.088,
+      "eval_steps_per_second": 0.211,
+      "step": 4345
+    },
+    {
+      "epoch": 0.5626436223581578,
+      "grad_norm": 2.247019952296008,
+      "learning_rate": 5.974674362961135e-07,
+      "loss": 2.6516,
+      "step": 4346
+    },
+    {
+      "epoch": 0.5627730847655112,
+      "grad_norm": 2.3293819167551955,
+      "learning_rate": 5.971852366554116e-07,
+      "loss": 2.8132,
+      "step": 4347
+    },
+    {
+      "epoch": 0.5629025471728647,
+      "grad_norm": 2.2004301921375125,
+      "learning_rate": 5.969030426964389e-07,
+      "loss": 2.4553,
+      "step": 4348
+    },
+    {
+      "epoch": 0.5630320095802182,
+      "grad_norm": 1.5640593573696209,
+      "learning_rate": 5.966208544768394e-07,
+      "loss": 2.5265,
+      "step": 4349
+    },
+    {
+      "epoch": 0.5631614719875716,
+      "grad_norm": 2.3508723234105955,
+      "learning_rate": 5.963386720542554e-07,
+      "loss": 2.3914,
+      "step": 4350
+    },
+    {
+      "epoch": 0.5631614719875716,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5288752317428589,
+      "eval_runtime": 15.089,
+      "eval_samples_per_second": 2.916,
+      "eval_steps_per_second": 0.199,
+      "step": 4350
+    },
+    {
+      "epoch": 0.5632909343949251,
+      "grad_norm": 1.719696224374733,
+      "learning_rate": 5.960564954863287e-07,
+      "loss": 2.4253,
+      "step": 4351
+    },
+    {
+      "epoch": 0.5634203968022785,
+      "grad_norm": 1.6928187089090314,
+      "learning_rate": 5.957743248306992e-07,
+      "loss": 2.3608,
+      "step": 4352
+    },
+    {
+      "epoch": 0.563549859209632,
+      "grad_norm": 1.2109778019038435,
+      "learning_rate": 5.954921601450063e-07,
+      "loss": 2.3892,
+      "step": 4353
+    },
+    {
+      "epoch": 0.5636793216169854,
+      "grad_norm": 1.5762652315588639,
+      "learning_rate": 5.952100014868879e-07,
+      "loss": 2.5181,
+      "step": 4354
+    },
+    {
+      "epoch": 0.563808784024339,
+      "grad_norm": 1.3981577179119586,
+      "learning_rate": 5.949278489139808e-07,
+      "loss": 2.3237,
+      "step": 4355
+    },
+    {
+      "epoch": 0.563808784024339,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.531782627105713,
+      "eval_runtime": 15.3357,
+      "eval_samples_per_second": 2.869,
+      "eval_steps_per_second": 0.196,
+      "step": 4355
+    },
+    {
+      "epoch": 0.5639382464316924,
+      "grad_norm": 1.4754293718042735,
+      "learning_rate": 5.946457024839196e-07,
+      "loss": 2.3621,
+      "step": 4356
+    },
+    {
+      "epoch": 0.5640677088390459,
+      "grad_norm": 1.712249368352622,
+      "learning_rate": 5.94363562254339e-07,
+      "loss": 2.3892,
+      "step": 4357
+    },
+    {
+      "epoch": 0.5641971712463993,
+      "grad_norm": 2.6254923200702622,
+      "learning_rate": 5.940814282828719e-07,
+      "loss": 2.3833,
+      "step": 4358
+    },
+    {
+      "epoch": 0.5643266336537528,
+      "grad_norm": 3.885289785657173,
+      "learning_rate": 5.937993006271497e-07,
+      "loss": 2.6719,
+      "step": 4359
+    },
+    {
+      "epoch": 0.5644560960611062,
+      "grad_norm": 2.363808303837714,
+      "learning_rate": 5.93517179344803e-07,
+      "loss": 2.5852,
+      "step": 4360
+    },
+    {
+      "epoch": 0.5644560960611062,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.531538486480713,
+      "eval_runtime": 14.0524,
+      "eval_samples_per_second": 3.131,
+      "eval_steps_per_second": 0.213,
+      "step": 4360
+    },
+    {
+      "epoch": 0.5645855584684597,
+      "grad_norm": 1.602069822782967,
+      "learning_rate": 5.932350644934602e-07,
+      "loss": 2.2119,
+      "step": 4361
+    },
+    {
+      "epoch": 0.5647150208758132,
+      "grad_norm": 1.8281623782924643,
+      "learning_rate": 5.929529561307496e-07,
+      "loss": 2.6133,
+      "step": 4362
+    },
+    {
+      "epoch": 0.5648444832831666,
+      "grad_norm": 2.0200877176521717,
+      "learning_rate": 5.926708543142971e-07,
+      "loss": 2.5155,
+      "step": 4363
+    },
+    {
+      "epoch": 0.5649739456905201,
+      "grad_norm": 1.8284045627951315,
+      "learning_rate": 5.923887591017278e-07,
+      "loss": 2.6443,
+      "step": 4364
+    },
+    {
+      "epoch": 0.5651034080978736,
+      "grad_norm": 2.687058918958252,
+      "learning_rate": 5.921066705506657e-07,
+      "loss": 2.2771,
+      "step": 4365
+    },
+    {
+      "epoch": 0.5651034080978736,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5296519994735718,
+      "eval_runtime": 15.087,
+      "eval_samples_per_second": 2.916,
+      "eval_steps_per_second": 0.199,
+      "step": 4365
+    },
+    {
+      "epoch": 0.5652328705052271,
+      "grad_norm": 1.8236095391076637,
+      "learning_rate": 5.918245887187328e-07,
+      "loss": 2.2871,
+      "step": 4366
+    },
+    {
+      "epoch": 0.5653623329125805,
+      "grad_norm": 1.5896062479065716,
+      "learning_rate": 5.915425136635501e-07,
+      "loss": 2.192,
+      "step": 4367
+    },
+    {
+      "epoch": 0.565491795319934,
+      "grad_norm": 1.6472960425831566,
+      "learning_rate": 5.912604454427371e-07,
+      "loss": 2.5454,
+      "step": 4368
+    },
+    {
+      "epoch": 0.5656212577272874,
+      "grad_norm": 2.9543109508071046,
+      "learning_rate": 5.909783841139119e-07,
+      "loss": 2.6353,
+      "step": 4369
+    },
+    {
+      "epoch": 0.5657507201346409,
+      "grad_norm": 1.4553244028788497,
+      "learning_rate": 5.906963297346913e-07,
+      "loss": 2.3591,
+      "step": 4370
+    },
+    {
+      "epoch": 0.5657507201346409,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5302734375,
+      "eval_runtime": 14.4201,
+      "eval_samples_per_second": 3.051,
+      "eval_steps_per_second": 0.208,
+      "step": 4370
+    },
+    {
+      "epoch": 0.5658801825419943,
+      "grad_norm": 2.1224827481378656,
+      "learning_rate": 5.90414282362691e-07,
+      "loss": 2.2778,
+      "step": 4371
+    },
+    {
+      "epoch": 0.5660096449493478,
+      "grad_norm": 1.182766858602216,
+      "learning_rate": 5.901322420555242e-07,
+      "loss": 2.2878,
+      "step": 4372
+    },
+    {
+      "epoch": 0.5661391073567013,
+      "grad_norm": 2.4107673011574464,
+      "learning_rate": 5.898502088708038e-07,
+      "loss": 2.7124,
+      "step": 4373
+    },
+    {
+      "epoch": 0.5662685697640547,
+      "grad_norm": 1.819894293678706,
+      "learning_rate": 5.895681828661407e-07,
+      "loss": 2.4404,
+      "step": 4374
+    },
+    {
+      "epoch": 0.5663980321714083,
+      "grad_norm": 2.159685988426527,
+      "learning_rate": 5.892861640991443e-07,
+      "loss": 2.4075,
+      "step": 4375
+    },
+    {
+      "epoch": 0.5663980321714083,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.527388095855713,
+      "eval_runtime": 14.5617,
+      "eval_samples_per_second": 3.022,
+      "eval_steps_per_second": 0.206,
+      "step": 4375
+    },
+    {
+      "epoch": 0.5665274945787617,
+      "grad_norm": 2.043764536141545,
+      "learning_rate": 5.890041526274227e-07,
+      "loss": 2.4575,
+      "step": 4376
+    },
+    {
+      "epoch": 0.5666569569861152,
+      "grad_norm": 1.5799764033172574,
+      "learning_rate": 5.887221485085831e-07,
+      "loss": 2.3267,
+      "step": 4377
+    },
+    {
+      "epoch": 0.5667864193934686,
+      "grad_norm": 1.3463578493932928,
+      "learning_rate": 5.884401518002294e-07,
+      "loss": 2.4819,
+      "step": 4378
+    },
+    {
+      "epoch": 0.5669158818008221,
+      "grad_norm": 2.0349278503485153,
+      "learning_rate": 5.88158162559966e-07,
+      "loss": 2.4309,
+      "step": 4379
+    },
+    {
+      "epoch": 0.5670453442081755,
+      "grad_norm": 2.770127127905956,
+      "learning_rate": 5.878761808453945e-07,
+      "loss": 2.248,
+      "step": 4380
+    },
+    {
+      "epoch": 0.5670453442081755,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5281871557235718,
+      "eval_runtime": 14.9465,
+      "eval_samples_per_second": 2.944,
+      "eval_steps_per_second": 0.201,
+      "step": 4380
+    },
+    {
+      "epoch": 0.567174806615529,
+      "grad_norm": 2.134112231434973,
+      "learning_rate": 5.875942067141155e-07,
+      "loss": 2.6089,
+      "step": 4381
+    },
+    {
+      "epoch": 0.5673042690228824,
+      "grad_norm": 2.994367714922088,
+      "learning_rate": 5.87312240223728e-07,
+      "loss": 2.4468,
+      "step": 4382
+    },
+    {
+      "epoch": 0.5674337314302359,
+      "grad_norm": 1.7663851518555604,
+      "learning_rate": 5.870302814318297e-07,
+      "loss": 2.394,
+      "step": 4383
+    },
+    {
+      "epoch": 0.5675631938375895,
+      "grad_norm": 1.4659378815165225,
+      "learning_rate": 5.867483303960156e-07,
+      "loss": 2.4839,
+      "step": 4384
+    },
+    {
+      "epoch": 0.5676926562449429,
+      "grad_norm": 1.1933853110709893,
+      "learning_rate": 5.864663871738804e-07,
+      "loss": 2.3306,
+      "step": 4385
+    },
+    {
+      "epoch": 0.5676926562449429,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5296963453292847,
+      "eval_runtime": 14.9068,
+      "eval_samples_per_second": 2.952,
+      "eval_steps_per_second": 0.201,
+      "step": 4385
+    },
+    {
+      "epoch": 0.5678221186522964,
+      "grad_norm": 2.4091167840293055,
+      "learning_rate": 5.861844518230166e-07,
+      "loss": 2.7019,
+      "step": 4386
+    },
+    {
+      "epoch": 0.5679515810596498,
+      "grad_norm": 1.988042526869207,
+      "learning_rate": 5.859025244010154e-07,
+      "loss": 2.3662,
+      "step": 4387
+    },
+    {
+      "epoch": 0.5680810434670033,
+      "grad_norm": 1.8566397518006972,
+      "learning_rate": 5.856206049654662e-07,
+      "loss": 2.4619,
+      "step": 4388
+    },
+    {
+      "epoch": 0.5682105058743567,
+      "grad_norm": 1.4740486551990213,
+      "learning_rate": 5.853386935739565e-07,
+      "loss": 2.446,
+      "step": 4389
+    },
+    {
+      "epoch": 0.5683399682817102,
+      "grad_norm": 2.8905062659500493,
+      "learning_rate": 5.850567902840725e-07,
+      "loss": 2.3464,
+      "step": 4390
+    },
+    {
+      "epoch": 0.5683399682817102,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5274103879928589,
+      "eval_runtime": 13.3875,
+      "eval_samples_per_second": 3.287,
+      "eval_steps_per_second": 0.224,
+      "step": 4390
+    },
+    {
+      "epoch": 0.5684694306890636,
+      "grad_norm": 1.4397807669253861,
+      "learning_rate": 5.847748951533986e-07,
+      "loss": 2.4487,
+      "step": 4391
+    },
+    {
+      "epoch": 0.5685988930964171,
+      "grad_norm": 1.5826540808394478,
+      "learning_rate": 5.844930082395179e-07,
+      "loss": 2.4158,
+      "step": 4392
+    },
+    {
+      "epoch": 0.5687283555037705,
+      "grad_norm": 1.4787467223629225,
+      "learning_rate": 5.842111296000111e-07,
+      "loss": 2.5039,
+      "step": 4393
+    },
+    {
+      "epoch": 0.5688578179111241,
+      "grad_norm": 2.338872559020097,
+      "learning_rate": 5.83929259292458e-07,
+      "loss": 2.543,
+      "step": 4394
+    },
+    {
+      "epoch": 0.5689872803184776,
+      "grad_norm": 2.0668034917106537,
+      "learning_rate": 5.836473973744361e-07,
+      "loss": 2.387,
+      "step": 4395
+    },
+    {
+      "epoch": 0.5689872803184776,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5296076536178589,
+      "eval_runtime": 14.3069,
+      "eval_samples_per_second": 3.075,
+      "eval_steps_per_second": 0.21,
+      "step": 4395
+    },
+    {
+      "epoch": 0.569116742725831,
+      "grad_norm": 1.400165456144604,
+      "learning_rate": 5.833655439035214e-07,
+      "loss": 2.3931,
+      "step": 4396
+    },
+    {
+      "epoch": 0.5692462051331845,
+      "grad_norm": 1.4042769598962068,
+      "learning_rate": 5.830836989372884e-07,
+      "loss": 2.2837,
+      "step": 4397
+    },
+    {
+      "epoch": 0.5693756675405379,
+      "grad_norm": 1.9484711577465819,
+      "learning_rate": 5.828018625333092e-07,
+      "loss": 2.3738,
+      "step": 4398
+    },
+    {
+      "epoch": 0.5695051299478914,
+      "grad_norm": 1.9466189831270702,
+      "learning_rate": 5.825200347491553e-07,
+      "loss": 2.4841,
+      "step": 4399
+    },
+    {
+      "epoch": 0.5696345923552448,
+      "grad_norm": 1.791127433568799,
+      "learning_rate": 5.822382156423951e-07,
+      "loss": 2.4585,
+      "step": 4400
+    },
+    {
+      "epoch": 0.5696345923552448,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.525346279144287,
+      "eval_runtime": 14.0008,
+      "eval_samples_per_second": 3.143,
+      "eval_steps_per_second": 0.214,
+      "step": 4400
+    },
+    {
+      "epoch": 0.5697640547625983,
+      "grad_norm": 2.8891716286120426,
+      "learning_rate": 5.81956405270596e-07,
+      "loss": 2.4109,
+      "step": 4401
+    },
+    {
+      "epoch": 0.5698935171699517,
+      "grad_norm": 2.7550617339644146,
+      "learning_rate": 5.816746036913238e-07,
+      "loss": 2.4573,
+      "step": 4402
+    },
+    {
+      "epoch": 0.5700229795773052,
+      "grad_norm": 2.59713592337112,
+      "learning_rate": 5.813928109621417e-07,
+      "loss": 2.5371,
+      "step": 4403
+    },
+    {
+      "epoch": 0.5701524419846588,
+      "grad_norm": 2.5926111759252564,
+      "learning_rate": 5.811110271406119e-07,
+      "loss": 2.3833,
+      "step": 4404
+    },
+    {
+      "epoch": 0.5702819043920122,
+      "grad_norm": 2.022572978459419,
+      "learning_rate": 5.808292522842948e-07,
+      "loss": 2.658,
+      "step": 4405
+    },
+    {
+      "epoch": 0.5702819043920122,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.528120517730713,
+      "eval_runtime": 13.6321,
+      "eval_samples_per_second": 3.228,
+      "eval_steps_per_second": 0.22,
+      "step": 4405
+    },
+    {
+      "epoch": 0.5704113667993657,
+      "grad_norm": 3.3056888083335267,
+      "learning_rate": 5.805474864507479e-07,
+      "loss": 2.6514,
+      "step": 4406
+    },
+    {
+      "epoch": 0.5705408292067191,
+      "grad_norm": 1.7753400137407758,
+      "learning_rate": 5.802657296975278e-07,
+      "loss": 2.4954,
+      "step": 4407
+    },
+    {
+      "epoch": 0.5706702916140726,
+      "grad_norm": 1.8646872930363578,
+      "learning_rate": 5.799839820821892e-07,
+      "loss": 2.4954,
+      "step": 4408
+    },
+    {
+      "epoch": 0.570799754021426,
+      "grad_norm": 2.3099731334676683,
+      "learning_rate": 5.797022436622847e-07,
+      "loss": 2.5325,
+      "step": 4409
+    },
+    {
+      "epoch": 0.5709292164287795,
+      "grad_norm": 1.3895810608022325,
+      "learning_rate": 5.794205144953654e-07,
+      "loss": 2.4773,
+      "step": 4410
+    },
+    {
+      "epoch": 0.5709292164287795,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5260564088821411,
+      "eval_runtime": 13.7311,
+      "eval_samples_per_second": 3.204,
+      "eval_steps_per_second": 0.218,
+      "step": 4410
+    },
+    {
+      "epoch": 0.5710586788361329,
+      "grad_norm": 1.3705900301873855,
+      "learning_rate": 5.791387946389794e-07,
+      "loss": 2.2119,
+      "step": 4411
+    },
+    {
+      "epoch": 0.5711881412434864,
+      "grad_norm": 1.9473085089700983,
+      "learning_rate": 5.788570841506744e-07,
+      "loss": 2.7031,
+      "step": 4412
+    },
+    {
+      "epoch": 0.5713176036508398,
+      "grad_norm": 1.354616648337542,
+      "learning_rate": 5.785753830879951e-07,
+      "loss": 2.2866,
+      "step": 4413
+    },
+    {
+      "epoch": 0.5714470660581934,
+      "grad_norm": 1.9862347815882009,
+      "learning_rate": 5.782936915084849e-07,
+      "loss": 2.4004,
+      "step": 4414
+    },
+    {
+      "epoch": 0.5715765284655469,
+      "grad_norm": 1.129499225406658,
+      "learning_rate": 5.780120094696848e-07,
+      "loss": 2.3333,
+      "step": 4415
+    },
+    {
+      "epoch": 0.5715765284655469,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5262340307235718,
+      "eval_runtime": 14.5752,
+      "eval_samples_per_second": 3.019,
+      "eval_steps_per_second": 0.206,
+      "step": 4415
+    },
+    {
+      "epoch": 0.5717059908729003,
+      "grad_norm": 1.6202474907493667,
+      "learning_rate": 5.777303370291345e-07,
+      "loss": 2.7019,
+      "step": 4416
+    },
+    {
+      "epoch": 0.5718354532802538,
+      "grad_norm": 1.9191105811353815,
+      "learning_rate": 5.774486742443706e-07,
+      "loss": 2.4966,
+      "step": 4417
+    },
+    {
+      "epoch": 0.5719649156876072,
+      "grad_norm": 2.037320423421432,
+      "learning_rate": 5.771670211729289e-07,
+      "loss": 2.4553,
+      "step": 4418
+    },
+    {
+      "epoch": 0.5720943780949607,
+      "grad_norm": 2.510736120238449,
+      "learning_rate": 5.768853778723426e-07,
+      "loss": 2.5156,
+      "step": 4419
+    },
+    {
+      "epoch": 0.5722238405023141,
+      "grad_norm": 1.4703227582133935,
+      "learning_rate": 5.766037444001432e-07,
+      "loss": 2.4041,
+      "step": 4420
+    },
+    {
+      "epoch": 0.5722238405023141,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5289639234542847,
+      "eval_runtime": 14.9362,
+      "eval_samples_per_second": 2.946,
+      "eval_steps_per_second": 0.201,
+      "step": 4420
+    },
+    {
+      "epoch": 0.5723533029096676,
+      "grad_norm": 1.8304570511644205,
+      "learning_rate": 5.763221208138599e-07,
+      "loss": 2.3875,
+      "step": 4421
+    },
+    {
+      "epoch": 0.572482765317021,
+      "grad_norm": 2.473405558929656,
+      "learning_rate": 5.760405071710203e-07,
+      "loss": 2.4817,
+      "step": 4422
+    },
+    {
+      "epoch": 0.5726122277243746,
+      "grad_norm": 2.281016263263221,
+      "learning_rate": 5.757589035291493e-07,
+      "loss": 2.4146,
+      "step": 4423
+    },
+    {
+      "epoch": 0.572741690131728,
+      "grad_norm": 1.8188954735236613,
+      "learning_rate": 5.754773099457703e-07,
+      "loss": 2.4629,
+      "step": 4424
+    },
+    {
+      "epoch": 0.5728711525390815,
+      "grad_norm": 1.5946402814098786,
+      "learning_rate": 5.751957264784045e-07,
+      "loss": 2.4077,
+      "step": 4425
+    },
+    {
+      "epoch": 0.5728711525390815,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5260342359542847,
+      "eval_runtime": 14.7625,
+      "eval_samples_per_second": 2.981,
+      "eval_steps_per_second": 0.203,
+      "step": 4425
+    },
+    {
+      "epoch": 0.573000614946435,
+      "grad_norm": 1.4585366481901132,
+      "learning_rate": 5.74914153184571e-07,
+      "loss": 2.2961,
+      "step": 4426
+    },
+    {
+      "epoch": 0.5731300773537884,
+      "grad_norm": 2.3032195190792484,
+      "learning_rate": 5.746325901217873e-07,
+      "loss": 2.4172,
+      "step": 4427
+    },
+    {
+      "epoch": 0.5732595397611419,
+      "grad_norm": 1.2597632216703234,
+      "learning_rate": 5.743510373475676e-07,
+      "loss": 2.4263,
+      "step": 4428
+    },
+    {
+      "epoch": 0.5733890021684953,
+      "grad_norm": 1.5173838874780994,
+      "learning_rate": 5.740694949194252e-07,
+      "loss": 2.4333,
+      "step": 4429
+    },
+    {
+      "epoch": 0.5735184645758488,
+      "grad_norm": 2.401684593596505,
+      "learning_rate": 5.737879628948706e-07,
+      "loss": 2.3127,
+      "step": 4430
+    },
+    {
+      "epoch": 0.5735184645758488,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5231267213821411,
+      "eval_runtime": 13.8635,
+      "eval_samples_per_second": 3.174,
+      "eval_steps_per_second": 0.216,
+      "step": 4430
+    },
+    {
+      "epoch": 0.5736479269832022,
+      "grad_norm": 1.4510733736592982,
+      "learning_rate": 5.735064413314127e-07,
+      "loss": 2.311,
+      "step": 4431
+    },
+    {
+      "epoch": 0.5737773893905557,
+      "grad_norm": 1.6857797273584392,
+      "learning_rate": 5.73224930286558e-07,
+      "loss": 2.6533,
+      "step": 4432
+    },
+    {
+      "epoch": 0.5739068517979092,
+      "grad_norm": 1.3458299598883972,
+      "learning_rate": 5.729434298178109e-07,
+      "loss": 2.2466,
+      "step": 4433
+    },
+    {
+      "epoch": 0.5740363142052627,
+      "grad_norm": 2.2297329226935325,
+      "learning_rate": 5.72661939982673e-07,
+      "loss": 2.4878,
+      "step": 4434
+    },
+    {
+      "epoch": 0.5741657766126161,
+      "grad_norm": 2.213536585372666,
+      "learning_rate": 5.723804608386449e-07,
+      "loss": 2.4495,
+      "step": 4435
+    },
+    {
+      "epoch": 0.5741657766126161,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.525346279144287,
+      "eval_runtime": 15.4433,
+      "eval_samples_per_second": 2.849,
+      "eval_steps_per_second": 0.194,
+      "step": 4435
+    },
+    {
+      "epoch": 0.5742952390199696,
+      "grad_norm": 1.7665518674641991,
+      "learning_rate": 5.720989924432242e-07,
+      "loss": 2.4937,
+      "step": 4436
+    },
+    {
+      "epoch": 0.574424701427323,
+      "grad_norm": 1.9313632127953595,
+      "learning_rate": 5.718175348539066e-07,
+      "loss": 2.6377,
+      "step": 4437
+    },
+    {
+      "epoch": 0.5745541638346765,
+      "grad_norm": 1.4411096246688953,
+      "learning_rate": 5.715360881281855e-07,
+      "loss": 2.4055,
+      "step": 4438
+    },
+    {
+      "epoch": 0.57468362624203,
+      "grad_norm": 2.4802929117215347,
+      "learning_rate": 5.712546523235519e-07,
+      "loss": 2.2656,
+      "step": 4439
+    },
+    {
+      "epoch": 0.5748130886493834,
+      "grad_norm": 2.0575709791767127,
+      "learning_rate": 5.70973227497495e-07,
+      "loss": 2.4155,
+      "step": 4440
+    },
+    {
+      "epoch": 0.5748130886493834,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5282315015792847,
+      "eval_runtime": 15.0436,
+      "eval_samples_per_second": 2.925,
+      "eval_steps_per_second": 0.199,
+      "step": 4440
+    },
+    {
+      "epoch": 0.5749425510567369,
+      "grad_norm": 1.6075160428351194,
+      "learning_rate": 5.706918137075016e-07,
+      "loss": 2.3403,
+      "step": 4441
+    },
+    {
+      "epoch": 0.5750720134640903,
+      "grad_norm": 2.0210791928496072,
+      "learning_rate": 5.704104110110558e-07,
+      "loss": 2.5894,
+      "step": 4442
+    },
+    {
+      "epoch": 0.5752014758714439,
+      "grad_norm": 3.581780719538034,
+      "learning_rate": 5.701290194656401e-07,
+      "loss": 2.5481,
+      "step": 4443
+    },
+    {
+      "epoch": 0.5753309382787973,
+      "grad_norm": 2.3132010101035094,
+      "learning_rate": 5.698476391287346e-07,
+      "loss": 2.5286,
+      "step": 4444
+    },
+    {
+      "epoch": 0.5754604006861508,
+      "grad_norm": 2.1634873054041095,
+      "learning_rate": 5.695662700578164e-07,
+      "loss": 2.7505,
+      "step": 4445
+    },
+    {
+      "epoch": 0.5754604006861508,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5262340307235718,
+      "eval_runtime": 15.2581,
+      "eval_samples_per_second": 2.884,
+      "eval_steps_per_second": 0.197,
+      "step": 4445
+    },
+    {
+      "epoch": 0.5755898630935042,
+      "grad_norm": 4.710934020276544,
+      "learning_rate": 5.692849123103612e-07,
+      "loss": 2.8469,
+      "step": 4446
+    },
+    {
+      "epoch": 0.5757193255008577,
+      "grad_norm": 2.5049038293057797,
+      "learning_rate": 5.69003565943842e-07,
+      "loss": 2.1226,
+      "step": 4447
+    },
+    {
+      "epoch": 0.5758487879082111,
+      "grad_norm": 1.85147163157558,
+      "learning_rate": 5.687222310157292e-07,
+      "loss": 2.5762,
+      "step": 4448
+    },
+    {
+      "epoch": 0.5759782503155646,
+      "grad_norm": 2.6613966355536616,
+      "learning_rate": 5.68440907583492e-07,
+      "loss": 2.2188,
+      "step": 4449
+    },
+    {
+      "epoch": 0.576107712722918,
+      "grad_norm": 3.426319030938563,
+      "learning_rate": 5.681595957045953e-07,
+      "loss": 2.5869,
+      "step": 4450
+    },
+    {
+      "epoch": 0.576107712722918,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5259898900985718,
+      "eval_runtime": 14.7706,
+      "eval_samples_per_second": 2.979,
+      "eval_steps_per_second": 0.203,
+      "step": 4450
+    },
+    {
+      "epoch": 0.5762371751302715,
+      "grad_norm": 1.383261421992041,
+      "learning_rate": 5.678782954365035e-07,
+      "loss": 2.3679,
+      "step": 4451
+    },
+    {
+      "epoch": 0.576366637537625,
+      "grad_norm": 2.1947644694985136,
+      "learning_rate": 5.675970068366775e-07,
+      "loss": 2.2419,
+      "step": 4452
+    },
+    {
+      "epoch": 0.5764960999449785,
+      "grad_norm": 3.7721305425759724,
+      "learning_rate": 5.673157299625763e-07,
+      "loss": 2.3655,
+      "step": 4453
+    },
+    {
+      "epoch": 0.576625562352332,
+      "grad_norm": 1.2867292570156783,
+      "learning_rate": 5.670344648716567e-07,
+      "loss": 2.2977,
+      "step": 4454
+    },
+    {
+      "epoch": 0.5767550247596854,
+      "grad_norm": 1.5538208293258153,
+      "learning_rate": 5.667532116213727e-07,
+      "loss": 2.4429,
+      "step": 4455
+    },
+    {
+      "epoch": 0.5767550247596854,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.52880859375,
+      "eval_runtime": 14.1241,
+      "eval_samples_per_second": 3.115,
+      "eval_steps_per_second": 0.212,
+      "step": 4455
+    },
+    {
+      "epoch": 0.5768844871670389,
+      "grad_norm": 1.965334129077527,
+      "learning_rate": 5.664719702691755e-07,
+      "loss": 2.604,
+      "step": 4456
+    },
+    {
+      "epoch": 0.5770139495743923,
+      "grad_norm": 4.026632870232991,
+      "learning_rate": 5.661907408725148e-07,
+      "loss": 2.5515,
+      "step": 4457
+    },
+    {
+      "epoch": 0.5771434119817458,
+      "grad_norm": 4.676562484974604,
+      "learning_rate": 5.659095234888372e-07,
+      "loss": 2.624,
+      "step": 4458
+    },
+    {
+      "epoch": 0.5772728743890992,
+      "grad_norm": 1.730909330499002,
+      "learning_rate": 5.656283181755872e-07,
+      "loss": 2.3916,
+      "step": 4459
+    },
+    {
+      "epoch": 0.5774023367964527,
+      "grad_norm": 1.5817781744530113,
+      "learning_rate": 5.653471249902069e-07,
+      "loss": 2.4617,
+      "step": 4460
+    },
+    {
+      "epoch": 0.5774023367964527,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5274769067764282,
+      "eval_runtime": 14.4784,
+      "eval_samples_per_second": 3.039,
+      "eval_steps_per_second": 0.207,
+      "step": 4460
+    },
+    {
+      "epoch": 0.5775317992038062,
+      "grad_norm": 2.818841732056183,
+      "learning_rate": 5.65065943990135e-07,
+      "loss": 2.3599,
+      "step": 4461
+    },
+    {
+      "epoch": 0.5776612616111596,
+      "grad_norm": 1.0506262480272102,
+      "learning_rate": 5.64784775232809e-07,
+      "loss": 2.2012,
+      "step": 4462
+    },
+    {
+      "epoch": 0.5777907240185132,
+      "grad_norm": 1.5922862410622813,
+      "learning_rate": 5.645036187756633e-07,
+      "loss": 2.4409,
+      "step": 4463
+    },
+    {
+      "epoch": 0.5779201864258666,
+      "grad_norm": 1.4695994767055949,
+      "learning_rate": 5.642224746761296e-07,
+      "loss": 2.3076,
+      "step": 4464
+    },
+    {
+      "epoch": 0.5780496488332201,
+      "grad_norm": 1.6604665725323235,
+      "learning_rate": 5.639413429916372e-07,
+      "loss": 2.4216,
+      "step": 4465
+    },
+    {
+      "epoch": 0.5780496488332201,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.527388095855713,
+      "eval_runtime": 14.5962,
+      "eval_samples_per_second": 3.014,
+      "eval_steps_per_second": 0.206,
+      "step": 4465
+    },
+    {
+      "epoch": 0.5781791112405735,
+      "grad_norm": 1.9544969857678574,
+      "learning_rate": 5.636602237796134e-07,
+      "loss": 2.3052,
+      "step": 4466
+    },
+    {
+      "epoch": 0.578308573647927,
+      "grad_norm": 1.3974319686009802,
+      "learning_rate": 5.633791170974819e-07,
+      "loss": 2.457,
+      "step": 4467
+    },
+    {
+      "epoch": 0.5784380360552804,
+      "grad_norm": 1.7721137061460703,
+      "learning_rate": 5.630980230026649e-07,
+      "loss": 2.1443,
+      "step": 4468
+    },
+    {
+      "epoch": 0.5785674984626339,
+      "grad_norm": 1.5041703755625444,
+      "learning_rate": 5.628169415525812e-07,
+      "loss": 2.4646,
+      "step": 4469
+    },
+    {
+      "epoch": 0.5786969608699873,
+      "grad_norm": 1.8893228776421431,
+      "learning_rate": 5.625358728046478e-07,
+      "loss": 2.531,
+      "step": 4470
+    },
+    {
+      "epoch": 0.5786969608699873,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.529541015625,
+      "eval_runtime": 14.0562,
+      "eval_samples_per_second": 3.13,
+      "eval_steps_per_second": 0.213,
+      "step": 4470
+    },
+    {
+      "epoch": 0.5788264232773408,
+      "grad_norm": 2.5719945207698816,
+      "learning_rate": 5.622548168162781e-07,
+      "loss": 2.6311,
+      "step": 4471
+    },
+    {
+      "epoch": 0.5789558856846944,
+      "grad_norm": 1.9246790601651034,
+      "learning_rate": 5.619737736448844e-07,
+      "loss": 2.6304,
+      "step": 4472
+    },
+    {
+      "epoch": 0.5790853480920478,
+      "grad_norm": 2.0237890836361982,
+      "learning_rate": 5.616927433478747e-07,
+      "loss": 2.2788,
+      "step": 4473
+    },
+    {
+      "epoch": 0.5792148104994013,
+      "grad_norm": 2.9359188910380736,
+      "learning_rate": 5.614117259826553e-07,
+      "loss": 2.4771,
+      "step": 4474
+    },
+    {
+      "epoch": 0.5793442729067547,
+      "grad_norm": 1.1558513778999286,
+      "learning_rate": 5.611307216066296e-07,
+      "loss": 2.3115,
+      "step": 4475
+    },
+    {
+      "epoch": 0.5793442729067547,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5287864208221436,
+      "eval_runtime": 14.0072,
+      "eval_samples_per_second": 3.141,
+      "eval_steps_per_second": 0.214,
+      "step": 4475
+    },
+    {
+      "epoch": 0.5794737353141082,
+      "grad_norm": 1.8909683015214567,
+      "learning_rate": 5.608497302771987e-07,
+      "loss": 2.1995,
+      "step": 4476
+    },
+    {
+      "epoch": 0.5796031977214616,
+      "grad_norm": 1.4889476465015383,
+      "learning_rate": 5.605687520517609e-07,
+      "loss": 2.3591,
+      "step": 4477
+    },
+    {
+      "epoch": 0.5797326601288151,
+      "grad_norm": 1.4010784570748078,
+      "learning_rate": 5.602877869877113e-07,
+      "loss": 2.4226,
+      "step": 4478
+    },
+    {
+      "epoch": 0.5798621225361685,
+      "grad_norm": 3.120190715671266,
+      "learning_rate": 5.600068351424426e-07,
+      "loss": 2.4722,
+      "step": 4479
+    },
+    {
+      "epoch": 0.579991584943522,
+      "grad_norm": 1.350775285953974,
+      "learning_rate": 5.597258965733453e-07,
+      "loss": 2.4858,
+      "step": 4480
+    },
+    {
+      "epoch": 0.579991584943522,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5295631885528564,
+      "eval_runtime": 14.354,
+      "eval_samples_per_second": 3.065,
+      "eval_steps_per_second": 0.209,
+      "step": 4480
+    },
+    {
+      "epoch": 0.5801210473508754,
+      "grad_norm": 1.0767364308174234,
+      "learning_rate": 5.594449713378068e-07,
+      "loss": 2.3174,
+      "step": 4481
+    },
+    {
+      "epoch": 0.580250509758229,
+      "grad_norm": 1.4328183638098897,
+      "learning_rate": 5.591640594932116e-07,
+      "loss": 2.394,
+      "step": 4482
+    },
+    {
+      "epoch": 0.5803799721655825,
+      "grad_norm": 1.6070932440065626,
+      "learning_rate": 5.588831610969417e-07,
+      "loss": 2.5144,
+      "step": 4483
+    },
+    {
+      "epoch": 0.5805094345729359,
+      "grad_norm": 4.211060137167303,
+      "learning_rate": 5.586022762063762e-07,
+      "loss": 2.9375,
+      "step": 4484
+    },
+    {
+      "epoch": 0.5806388969802894,
+      "grad_norm": 2.2107982240612287,
+      "learning_rate": 5.583214048788917e-07,
+      "loss": 2.262,
+      "step": 4485
+    },
+    {
+      "epoch": 0.5806388969802894,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.528076171875,
+      "eval_runtime": 14.5999,
+      "eval_samples_per_second": 3.014,
+      "eval_steps_per_second": 0.205,
+      "step": 4485
+    },
+    {
+      "epoch": 0.5807683593876428,
+      "grad_norm": 1.4940383019804222,
+      "learning_rate": 5.580405471718617e-07,
+      "loss": 2.3945,
+      "step": 4486
+    },
+    {
+      "epoch": 0.5808978217949963,
+      "grad_norm": 3.9055127690181153,
+      "learning_rate": 5.577597031426572e-07,
+      "loss": 2.5847,
+      "step": 4487
+    },
+    {
+      "epoch": 0.5810272842023497,
+      "grad_norm": 1.5190296165711294,
+      "learning_rate": 5.574788728486463e-07,
+      "loss": 2.4238,
+      "step": 4488
+    },
+    {
+      "epoch": 0.5811567466097032,
+      "grad_norm": 2.553099489896722,
+      "learning_rate": 5.571980563471941e-07,
+      "loss": 2.5259,
+      "step": 4489
+    },
+    {
+      "epoch": 0.5812862090170566,
+      "grad_norm": 1.2215964268693216,
+      "learning_rate": 5.569172536956632e-07,
+      "loss": 2.3818,
+      "step": 4490
+    },
+    {
+      "epoch": 0.5812862090170566,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5287864208221436,
+      "eval_runtime": 15.8192,
+      "eval_samples_per_second": 2.781,
+      "eval_steps_per_second": 0.19,
+      "step": 4490
+    },
+    {
+      "epoch": 0.5814156714244101,
+      "grad_norm": 1.3329980872126876,
+      "learning_rate": 5.566364649514132e-07,
+      "loss": 2.4604,
+      "step": 4491
+    },
+    {
+      "epoch": 0.5815451338317636,
+      "grad_norm": 4.380748109242493,
+      "learning_rate": 5.563556901718008e-07,
+      "loss": 2.7351,
+      "step": 4492
+    },
+    {
+      "epoch": 0.5816745962391171,
+      "grad_norm": 1.4112095905544813,
+      "learning_rate": 5.560749294141801e-07,
+      "loss": 2.3372,
+      "step": 4493
+    },
+    {
+      "epoch": 0.5818040586464706,
+      "grad_norm": 1.8576856419271832,
+      "learning_rate": 5.557941827359023e-07,
+      "loss": 2.5603,
+      "step": 4494
+    },
+    {
+      "epoch": 0.581933521053824,
+      "grad_norm": 1.1999207118768667,
+      "learning_rate": 5.555134501943153e-07,
+      "loss": 2.4482,
+      "step": 4495
+    },
+    {
+      "epoch": 0.581933521053824,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.5308948755264282,
+      "eval_runtime": 14.637,
+      "eval_samples_per_second": 3.006,
+      "eval_steps_per_second": 0.205,
+      "step": 4495
+    },
+    {
+      "epoch": 0.5820629834611775,
+      "grad_norm": 1.9533276199200365,
+      "learning_rate": 5.552327318467645e-07,
+      "loss": 2.3313,
+      "step": 4496
+    },
+    {
+      "epoch": 0.5821924458685309,
+      "grad_norm": 1.617461968051022,
+      "learning_rate": 5.549520277505922e-07,
+      "loss": 2.2388,
+      "step": 4497
+    },
+    {
+      "epoch": 0.5823219082758844,
+      "grad_norm": 1.7036046199161534,
+      "learning_rate": 5.54671337963138e-07,
+      "loss": 2.5518,
+      "step": 4498
+    },
+    {
+      "epoch": 0.5824513706832378,
+      "grad_norm": 1.766900389112874,
+      "learning_rate": 5.543906625417389e-07,
+      "loss": 2.3081,
+      "step": 4499
+    },
+    {
+      "epoch": 0.5825808330905913,
+      "grad_norm": 2.142857262001564,
+      "learning_rate": 5.54110001543728e-07,
+      "loss": 2.4951,
+      "step": 4500
+    },
+    {
+      "epoch": 0.5825808330905913,
+      "eval_PRM Accuracy": 0.2358490566037736,
+      "eval_PRM F1": 0.047058823529411764,
+      "eval_PRM F1 AUC": 0.5120481927710844,
+      "eval_PRM F1 Neg": 0.36220472440944884,
+      "eval_PRM NPV": 0.22115384615384615,
+      "eval_PRM Precision": 1.0,
+      "eval_PRM Recall": 0.024096385542168676,
+      "eval_PRM Specificty": 1.0,
+      "eval_loss": 1.52880859375,
+      "eval_runtime": 14.8016,
+      "eval_samples_per_second": 2.973,
+      "eval_steps_per_second": 0.203,
+      "step": 4500
     }
   ],
   "logging_steps": 1,
@@ -40842,7 +45942,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 3361768083062784.0,
+  "total_flos": 3779727029731328.0,
   "train_batch_size": 4,
   "trial_name": null,
   "trial_params": null