diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,8603 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.795204913192394,
+  "eval_steps": 2000,
+  "global_step": 5700,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.003149482303846305,
+      "grad_norm": 3.109375,
+      "learning_rate": 5.000000000000001e-07,
+      "logits/chosen": -0.45210394263267517,
+      "logits/rejected": -0.3446429371833801,
+      "logps/chosen": -213.57180786132812,
+      "logps/rejected": -191.9154052734375,
+      "loss": 0.6941,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": 0.005887002218514681,
+      "rewards/margins": -0.0011509137693792582,
+      "rewards/rejected": 0.00703791668638587,
+      "step": 10
+    },
+    {
+      "epoch": 0.00629896460769261,
+      "grad_norm": 2.640625,
+      "learning_rate": 1.0000000000000002e-06,
+      "logits/chosen": -0.46016186475753784,
+      "logits/rejected": -0.2938859760761261,
+      "logps/chosen": -203.55313110351562,
+      "logps/rejected": -186.91030883789062,
+      "loss": 0.6917,
+      "rewards/accuracies": 0.512499988079071,
+      "rewards/chosen": 0.006860991008579731,
+      "rewards/margins": 0.004207834601402283,
+      "rewards/rejected": 0.002653155941516161,
+      "step": 20
+    },
+    {
+      "epoch": 0.009448446911538916,
+      "grad_norm": 2.703125,
+      "learning_rate": 1.5e-06,
+      "logits/chosen": -0.5505405068397522,
+      "logits/rejected": -0.32200556993484497,
+      "logps/chosen": -206.43130493164062,
+      "logps/rejected": -159.43423461914062,
+      "loss": 0.6978,
+      "rewards/accuracies": 0.4749999940395355,
+      "rewards/chosen": -0.004765630234032869,
+      "rewards/margins": -0.00844950508326292,
+      "rewards/rejected": 0.003683874849230051,
+      "step": 30
+    },
+    {
+      "epoch": 0.01259792921538522,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.0000000000000003e-06,
+      "logits/chosen": -0.5206310749053955,
+      "logits/rejected": -0.44945794343948364,
+      "logps/chosen": -192.1409454345703,
+      "logps/rejected": -195.43496704101562,
+      "loss": 0.6938,
+      "rewards/accuracies": 0.48750001192092896,
+      "rewards/chosen": 0.014343030750751495,
+      "rewards/margins": 0.0002199936716351658,
+      "rewards/rejected": 0.014123037457466125,
+      "step": 40
+    },
+    {
+      "epoch": 0.015747411519231525,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.5e-06,
+      "logits/chosen": -0.5311203002929688,
+      "logits/rejected": -0.4016449451446533,
+      "logps/chosen": -208.75094604492188,
+      "logps/rejected": -167.02542114257812,
+      "loss": 0.693,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.009670769795775414,
+      "rewards/margins": 0.001526160747744143,
+      "rewards/rejected": 0.00814460963010788,
+      "step": 50
+    },
+    {
+      "epoch": 0.018896893823077833,
+      "grad_norm": 2.734375,
+      "learning_rate": 3e-06,
+      "logits/chosen": -0.5485053062438965,
+      "logits/rejected": -0.43868058919906616,
+      "logps/chosen": -196.2797088623047,
+      "logps/rejected": -174.60232543945312,
+      "loss": 0.6959,
+      "rewards/accuracies": 0.42500001192092896,
+      "rewards/chosen": 0.008773775771260262,
+      "rewards/margins": -0.004444376099854708,
+      "rewards/rejected": 0.013218151405453682,
+      "step": 60
+    },
+    {
+      "epoch": 0.022046376126924137,
+      "grad_norm": 3.921875,
+      "learning_rate": 3.5e-06,
+      "logits/chosen": -0.5210541486740112,
+      "logits/rejected": -0.37378597259521484,
+      "logps/chosen": -205.04330444335938,
+      "logps/rejected": -167.24517822265625,
+      "loss": 0.6964,
+      "rewards/accuracies": 0.48750001192092896,
+      "rewards/chosen": 0.006323327776044607,
+      "rewards/margins": -0.00504258181899786,
+      "rewards/rejected": 0.011365910060703754,
+      "step": 70
+    },
+    {
+      "epoch": 0.02519585843077044,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.000000000000001e-06,
+      "logits/chosen": -0.4273603558540344,
+      "logits/rejected": -0.31517690420150757,
+      "logps/chosen": -206.41354370117188,
+      "logps/rejected": -190.51402282714844,
+      "loss": 0.6927,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": 0.012041259557008743,
+      "rewards/margins": 0.001850623870268464,
+      "rewards/rejected": 0.010190634056925774,
+      "step": 80
+    },
+    {
+      "epoch": 0.028345340734616746,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.5e-06,
+      "logits/chosen": -0.47311750054359436,
+      "logits/rejected": -0.3815276026725769,
+      "logps/chosen": -201.52731323242188,
+      "logps/rejected": -173.32937622070312,
+      "loss": 0.6893,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.016733495518565178,
+      "rewards/margins": 0.009016195312142372,
+      "rewards/rejected": 0.007717301603406668,
+      "step": 90
+    },
+    {
+      "epoch": 0.03149482303846305,
+      "grad_norm": 2.953125,
+      "learning_rate": 5e-06,
+      "logits/chosen": -0.44015175104141235,
+      "logits/rejected": -0.29738959670066833,
+      "logps/chosen": -233.11642456054688,
+      "logps/rejected": -192.25820922851562,
+      "loss": 0.6937,
+      "rewards/accuracies": 0.4749999940395355,
+      "rewards/chosen": 0.02164524421095848,
+      "rewards/margins": 0.0006244674441404641,
+      "rewards/rejected": 0.021020779386162758,
+      "step": 100
+    },
+    {
+      "epoch": 0.034644305342309355,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.999994443042687e-06,
+      "logits/chosen": -0.4323144853115082,
+      "logits/rejected": -0.325883150100708,
+      "logps/chosen": -224.80685424804688,
+      "logps/rejected": -206.5443572998047,
+      "loss": 0.697,
+      "rewards/accuracies": 0.44999998807907104,
+      "rewards/chosen": 0.016110900789499283,
+      "rewards/margins": -0.005870751105248928,
+      "rewards/rejected": 0.021981652826070786,
+      "step": 110
+    },
+    {
+      "epoch": 0.037793787646155666,
+      "grad_norm": 4.4375,
+      "learning_rate": 4.999977772195451e-06,
+      "logits/chosen": -0.47046566009521484,
+      "logits/rejected": -0.31014284491539,
+      "logps/chosen": -213.2732696533203,
+      "logps/rejected": -178.27183532714844,
+      "loss": 0.6867,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.034097231924533844,
+      "rewards/margins": 0.01459626667201519,
+      "rewards/rejected": 0.019500967115163803,
+      "step": 120
+    },
+    {
+      "epoch": 0.04094326995000197,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.999949987532405e-06,
+      "logits/chosen": -0.5328488349914551,
+      "logits/rejected": -0.3397727906703949,
+      "logps/chosen": -212.8068084716797,
+      "logps/rejected": -165.73016357421875,
+      "loss": 0.6904,
+      "rewards/accuracies": 0.4625000059604645,
+      "rewards/chosen": 0.03291865810751915,
+      "rewards/margins": 0.0070328498259186745,
+      "rewards/rejected": 0.02588580921292305,
+      "step": 130
+    },
+    {
+      "epoch": 0.044092752253848275,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.9999110891770655e-06,
+      "logits/chosen": -0.5087035894393921,
+      "logits/rejected": -0.37333738803863525,
+      "logps/chosen": -197.49069213867188,
+      "logps/rejected": -164.01104736328125,
+      "loss": 0.6848,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.04153250902891159,
+      "rewards/margins": 0.018097227439284325,
+      "rewards/rejected": 0.023435279726982117,
+      "step": 140
+    },
+    {
+      "epoch": 0.04724223455769458,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.999861077302358e-06,
+      "logits/chosen": -0.5480197668075562,
+      "logits/rejected": -0.39153584837913513,
+      "logps/chosen": -197.01953125,
+      "logps/rejected": -172.3303985595703,
+      "loss": 0.6895,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.04789372906088829,
+      "rewards/margins": 0.008711261674761772,
+      "rewards/rejected": 0.03918246552348137,
+      "step": 150
+    },
+    {
+      "epoch": 0.05039171686154088,
+      "grad_norm": 4.0,
+      "learning_rate": 4.999799952130615e-06,
+      "logits/chosen": -0.5040096044540405,
+      "logits/rejected": -0.39294344186782837,
+      "logps/chosen": -214.4684600830078,
+      "logps/rejected": -185.05201721191406,
+      "loss": 0.6902,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.043327994644641876,
+      "rewards/margins": 0.007260638289153576,
+      "rewards/rejected": 0.03606735169887543,
+      "step": 160
+    },
+    {
+      "epoch": 0.05354119916538719,
+      "grad_norm": 2.328125,
+      "learning_rate": 4.999727713933572e-06,
+      "logits/chosen": -0.5509570837020874,
+      "logits/rejected": -0.37693116068840027,
+      "logps/chosen": -184.16673278808594,
+      "logps/rejected": -162.15756225585938,
+      "loss": 0.6923,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.047002751380205154,
+      "rewards/margins": 0.003170955926179886,
+      "rewards/rejected": 0.04383179172873497,
+      "step": 170
+    },
+    {
+      "epoch": 0.05669068146923349,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.999644363032367e-06,
+      "logits/chosen": -0.5072035193443298,
+      "logits/rejected": -0.37213388085365295,
+      "logps/chosen": -195.76577758789062,
+      "logps/rejected": -164.03848266601562,
+      "loss": 0.6903,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.05497163534164429,
+      "rewards/margins": 0.007314900867640972,
+      "rewards/rejected": 0.04765673726797104,
+      "step": 180
+    },
+    {
+      "epoch": 0.0598401637730798,
+      "grad_norm": 3.734375,
+      "learning_rate": 4.999549899797544e-06,
+      "logits/chosen": -0.45937657356262207,
+      "logits/rejected": -0.31247037649154663,
+      "logps/chosen": -218.7798309326172,
+      "logps/rejected": -182.95201110839844,
+      "loss": 0.6905,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.052850522100925446,
+      "rewards/margins": 0.006988201290369034,
+      "rewards/rejected": 0.04586231708526611,
+      "step": 190
+    },
+    {
+      "epoch": 0.0629896460769261,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.999444324649045e-06,
+      "logits/chosen": -0.5390416979789734,
+      "logits/rejected": -0.35482490062713623,
+      "logps/chosen": -199.8936767578125,
+      "logps/rejected": -165.85488891601562,
+      "loss": 0.6856,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.07901380211114883,
+      "rewards/margins": 0.01698421686887741,
+      "rewards/rejected": 0.062029581516981125,
+      "step": 200
+    },
+    {
+      "epoch": 0.0661391283807724,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.999327638056212e-06,
+      "logits/chosen": -0.4826637804508209,
+      "logits/rejected": -0.3051258623600006,
+      "logps/chosen": -221.1772003173828,
+      "logps/rejected": -173.49583435058594,
+      "loss": 0.6906,
+      "rewards/accuracies": 0.42500001192092896,
+      "rewards/chosen": 0.07877618074417114,
+      "rewards/margins": 0.007990067824721336,
+      "rewards/rejected": 0.07078611105680466,
+      "step": 210
+    },
+    {
+      "epoch": 0.06928861068461871,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.999199840537781e-06,
+      "logits/chosen": -0.4253220558166504,
+      "logits/rejected": -0.25771626830101013,
+      "logps/chosen": -218.5417938232422,
+      "logps/rejected": -180.00387573242188,
+      "loss": 0.6865,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": 0.09241167455911636,
+      "rewards/margins": 0.01505853421986103,
+      "rewards/rejected": 0.07735314220190048,
+      "step": 220
+    },
+    {
+      "epoch": 0.07243809298846503,
+      "grad_norm": 3.0,
+      "learning_rate": 4.9990609326618845e-06,
+      "logits/chosen": -0.468860924243927,
+      "logits/rejected": -0.416407972574234,
+      "logps/chosen": -227.7265625,
+      "logps/rejected": -207.37020874023438,
+      "loss": 0.689,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.10431376844644547,
+      "rewards/margins": 0.01053541712462902,
+      "rewards/rejected": 0.09377835690975189,
+      "step": 230
+    },
+    {
+      "epoch": 0.07558757529231133,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.998910915046048e-06,
+      "logits/chosen": -0.47573143243789673,
+      "logits/rejected": -0.3487725853919983,
+      "logps/chosen": -218.22030639648438,
+      "logps/rejected": -186.4251708984375,
+      "loss": 0.6898,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.11783305555582047,
+      "rewards/margins": 0.009454838000237942,
+      "rewards/rejected": 0.10837821662425995,
+      "step": 240
+    },
+    {
+      "epoch": 0.07873705759615764,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.998749788357184e-06,
+      "logits/chosen": -0.491966187953949,
+      "logits/rejected": -0.42401209473609924,
+      "logps/chosen": -209.5353240966797,
+      "logps/rejected": -191.00352478027344,
+      "loss": 0.6781,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.12841267883777618,
+      "rewards/margins": 0.032561883330345154,
+      "rewards/rejected": 0.09585078805685043,
+      "step": 250
+    },
+    {
+      "epoch": 0.08188653990000394,
+      "grad_norm": 2.578125,
+      "learning_rate": 4.998577553311592e-06,
+      "logits/chosen": -0.5209980010986328,
+      "logits/rejected": -0.3212158679962158,
+      "logps/chosen": -207.7479705810547,
+      "logps/rejected": -174.1891632080078,
+      "loss": 0.681,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.12082584202289581,
+      "rewards/margins": 0.02722765877842903,
+      "rewards/rejected": 0.09359817951917648,
+      "step": 260
+    },
+    {
+      "epoch": 0.08503602220385025,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.998394210674954e-06,
+      "logits/chosen": -0.4800085127353668,
+      "logits/rejected": -0.4280025064945221,
+      "logps/chosen": -193.07815551757812,
+      "logps/rejected": -190.64801025390625,
+      "loss": 0.6958,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.1191275492310524,
+      "rewards/margins": -0.0025443979538977146,
+      "rewards/rejected": 0.12167193740606308,
+      "step": 270
+    },
+    {
+      "epoch": 0.08818550450769655,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.998199761262332e-06,
+      "logits/chosen": -0.4940645694732666,
+      "logits/rejected": -0.3298795521259308,
+      "logps/chosen": -200.9434814453125,
+      "logps/rejected": -179.53814697265625,
+      "loss": 0.6842,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.13147035241127014,
+      "rewards/margins": 0.02113470621407032,
+      "rewards/rejected": 0.11033564805984497,
+      "step": 280
+    },
+    {
+      "epoch": 0.09133498681154285,
+      "grad_norm": 2.375,
+      "learning_rate": 4.997994205938164e-06,
+      "logits/chosen": -0.5327965617179871,
+      "logits/rejected": -0.38405364751815796,
+      "logps/chosen": -214.1449432373047,
+      "logps/rejected": -185.3954620361328,
+      "loss": 0.6731,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.16222958266735077,
+      "rewards/margins": 0.0439109206199646,
+      "rewards/rejected": 0.11831866204738617,
+      "step": 290
+    },
+    {
+      "epoch": 0.09448446911538916,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.997777545616258e-06,
+      "logits/chosen": -0.4295479655265808,
+      "logits/rejected": -0.2999122738838196,
+      "logps/chosen": -206.5615997314453,
+      "logps/rejected": -172.51394653320312,
+      "loss": 0.6781,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.15420152246952057,
+      "rewards/margins": 0.033724717795848846,
+      "rewards/rejected": 0.12047679722309113,
+      "step": 300
+    },
+    {
+      "epoch": 0.09763395141923546,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.9975497812597935e-06,
+      "logits/chosen": -0.47075533866882324,
+      "logits/rejected": -0.38670963048934937,
+      "logps/chosen": -209.8960418701172,
+      "logps/rejected": -182.66079711914062,
+      "loss": 0.6837,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.18156033754348755,
+      "rewards/margins": 0.02306472323834896,
+      "rewards/rejected": 0.15849560499191284,
+      "step": 310
+    },
+    {
+      "epoch": 0.10078343372308177,
+      "grad_norm": 3.21875,
+      "learning_rate": 4.997310913881312e-06,
+      "logits/chosen": -0.4988747239112854,
+      "logits/rejected": -0.363254576921463,
+      "logps/chosen": -202.0031280517578,
+      "logps/rejected": -175.33151245117188,
+      "loss": 0.6791,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": 0.1774149388074875,
+      "rewards/margins": 0.03209572285413742,
+      "rewards/rejected": 0.14531922340393066,
+      "step": 320
+    },
+    {
+      "epoch": 0.10393291602692807,
+      "grad_norm": 2.875,
+      "learning_rate": 4.997060944542713e-06,
+      "logits/chosen": -0.5674048662185669,
+      "logits/rejected": -0.4282529950141907,
+      "logps/chosen": -204.34375,
+      "logps/rejected": -169.42828369140625,
+      "loss": 0.6705,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.18822543323040009,
+      "rewards/margins": 0.0504903681576252,
+      "rewards/rejected": 0.13773508369922638,
+      "step": 330
+    },
+    {
+      "epoch": 0.10708239833077438,
+      "grad_norm": 3.578125,
+      "learning_rate": 4.996799874355253e-06,
+      "logits/chosen": -0.548692524433136,
+      "logits/rejected": -0.42939552664756775,
+      "logps/chosen": -204.91748046875,
+      "logps/rejected": -187.4248046875,
+      "loss": 0.686,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.18558073043823242,
+      "rewards/margins": 0.018990959972143173,
+      "rewards/rejected": 0.16658978164196014,
+      "step": 340
+    },
+    {
+      "epoch": 0.11023188063462068,
+      "grad_norm": 4.09375,
+      "learning_rate": 4.996527704479535e-06,
+      "logits/chosen": -0.43033042550086975,
+      "logits/rejected": -0.3002139627933502,
+      "logps/chosen": -215.57754516601562,
+      "logps/rejected": -193.66397094726562,
+      "loss": 0.6858,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.19084212183952332,
+      "rewards/margins": 0.018678244203329086,
+      "rewards/rejected": 0.17216388881206512,
+      "step": 350
+    },
+    {
+      "epoch": 0.11338136293846698,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.9962444361255095e-06,
+      "logits/chosen": -0.5352093577384949,
+      "logits/rejected": -0.3528750538825989,
+      "logps/chosen": -202.76707458496094,
+      "logps/rejected": -155.83926391601562,
+      "loss": 0.6774,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.20079512894153595,
+      "rewards/margins": 0.03638342395424843,
+      "rewards/rejected": 0.164411723613739,
+      "step": 360
+    },
+    {
+      "epoch": 0.11653084524231329,
+      "grad_norm": 3.640625,
+      "learning_rate": 4.995950070552464e-06,
+      "logits/chosen": -0.5055617094039917,
+      "logits/rejected": -0.3472587764263153,
+      "logps/chosen": -225.160888671875,
+      "logps/rejected": -181.47703552246094,
+      "loss": 0.6912,
+      "rewards/accuracies": 0.512499988079071,
+      "rewards/chosen": 0.21311064064502716,
+      "rewards/margins": 0.00886852853000164,
+      "rewards/rejected": 0.20424208045005798,
+      "step": 370
+    },
+    {
+      "epoch": 0.1196803275461596,
+      "grad_norm": 2.828125,
+      "learning_rate": 4.995644609069021e-06,
+      "logits/chosen": -0.48927387595176697,
+      "logits/rejected": -0.42719903588294983,
+      "logps/chosen": -185.412109375,
+      "logps/rejected": -181.981201171875,
+      "loss": 0.6891,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.1894209086894989,
+      "rewards/margins": 0.012292629107832909,
+      "rewards/rejected": 0.17712828516960144,
+      "step": 380
+    },
+    {
+      "epoch": 0.12282980985000591,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.995328053033129e-06,
+      "logits/chosen": -0.4682087302207947,
+      "logits/rejected": -0.25708621740341187,
+      "logps/chosen": -217.36709594726562,
+      "logps/rejected": -174.3680877685547,
+      "loss": 0.6665,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.22537508606910706,
+      "rewards/margins": 0.057967256754636765,
+      "rewards/rejected": 0.1674078404903412,
+      "step": 390
+    },
+    {
+      "epoch": 0.1259792921538522,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.995000403852057e-06,
+      "logits/chosen": -0.5313188433647156,
+      "logits/rejected": -0.39059966802597046,
+      "logps/chosen": -197.9477081298828,
+      "logps/rejected": -174.9998779296875,
+      "loss": 0.6761,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.19948481023311615,
+      "rewards/margins": 0.038825444877147675,
+      "rewards/rejected": 0.16065934300422668,
+      "step": 400
+    },
+    {
+      "epoch": 0.12912877445769852,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.994661662982393e-06,
+      "logits/chosen": -0.533206045627594,
+      "logits/rejected": -0.4300476610660553,
+      "logps/chosen": -192.5688934326172,
+      "logps/rejected": -167.0782928466797,
+      "loss": 0.684,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.19937366247177124,
+      "rewards/margins": 0.022861123085021973,
+      "rewards/rejected": 0.17651252448558807,
+      "step": 410
+    },
+    {
+      "epoch": 0.1322782567615448,
+      "grad_norm": 3.796875,
+      "learning_rate": 4.994311831930032e-06,
+      "logits/chosen": -0.4812285006046295,
+      "logits/rejected": -0.35317444801330566,
+      "logps/chosen": -189.956787109375,
+      "logps/rejected": -159.92453002929688,
+      "loss": 0.6706,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.23123323917388916,
+      "rewards/margins": 0.05047137662768364,
+      "rewards/rejected": 0.18076185882091522,
+      "step": 420
+    },
+    {
+      "epoch": 0.13542773906539113,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.993950912250171e-06,
+      "logits/chosen": -0.4360167384147644,
+      "logits/rejected": -0.3796977400779724,
+      "logps/chosen": -203.8537139892578,
+      "logps/rejected": -175.59356689453125,
+      "loss": 0.6806,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.21009023487567902,
+      "rewards/margins": 0.030164580792188644,
+      "rewards/rejected": 0.17992563545703888,
+      "step": 430
+    },
+    {
+      "epoch": 0.13857722136923742,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.9935789055473e-06,
+      "logits/chosen": -0.4947914183139801,
+      "logits/rejected": -0.36814266443252563,
+      "logps/chosen": -189.90982055664062,
+      "logps/rejected": -170.28982543945312,
+      "loss": 0.6788,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.22383634746074677,
+      "rewards/margins": 0.033211078494787216,
+      "rewards/rejected": 0.19062528014183044,
+      "step": 440
+    },
+    {
+      "epoch": 0.14172670367308374,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.993195813475202e-06,
+      "logits/chosen": -0.5398550629615784,
+      "logits/rejected": -0.3718551993370056,
+      "logps/chosen": -193.36532592773438,
+      "logps/rejected": -169.635986328125,
+      "loss": 0.6801,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.2529357373714447,
+      "rewards/margins": 0.03174880892038345,
+      "rewards/rejected": 0.22118695080280304,
+      "step": 450
+    },
+    {
+      "epoch": 0.14487618597693006,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.992801637736937e-06,
+      "logits/chosen": -0.5186210870742798,
+      "logits/rejected": -0.35956451296806335,
+      "logps/chosen": -189.38392639160156,
+      "logps/rejected": -168.08277893066406,
+      "loss": 0.6769,
+      "rewards/accuracies": 0.4749999940395355,
+      "rewards/chosen": 0.2704620957374573,
+      "rewards/margins": 0.03909246250987053,
+      "rewards/rejected": 0.23136961460113525,
+      "step": 460
+    },
+    {
+      "epoch": 0.14802566828077635,
+      "grad_norm": 3.609375,
+      "learning_rate": 4.992396380084839e-06,
+      "logits/chosen": -0.481611967086792,
+      "logits/rejected": -0.3885895907878876,
+      "logps/chosen": -212.15335083007812,
+      "logps/rejected": -216.1449737548828,
+      "loss": 0.6683,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.29815778136253357,
+      "rewards/margins": 0.05739093944430351,
+      "rewards/rejected": 0.24076685309410095,
+      "step": 470
+    },
+    {
+      "epoch": 0.15117515058462266,
+      "grad_norm": 3.671875,
+      "learning_rate": 4.991980042320507e-06,
+      "logits/chosen": -0.515384316444397,
+      "logits/rejected": -0.3729632794857025,
+      "logps/chosen": -199.055908203125,
+      "logps/rejected": -176.830322265625,
+      "loss": 0.6822,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.27753692865371704,
+      "rewards/margins": 0.03093332052230835,
+      "rewards/rejected": 0.2466035783290863,
+      "step": 480
+    },
+    {
+      "epoch": 0.15432463288846895,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.991552626294799e-06,
+      "logits/chosen": -0.477985680103302,
+      "logits/rejected": -0.3193301260471344,
+      "logps/chosen": -205.2687530517578,
+      "logps/rejected": -178.9701385498047,
+      "loss": 0.6786,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.275654137134552,
+      "rewards/margins": 0.03465648740530014,
+      "rewards/rejected": 0.24099759757518768,
+      "step": 490
+    },
+    {
+      "epoch": 0.15747411519231527,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.991114133907822e-06,
+      "logits/chosen": -0.5235245823860168,
+      "logits/rejected": -0.3859252631664276,
+      "logps/chosen": -212.98876953125,
+      "logps/rejected": -178.56101989746094,
+      "loss": 0.6856,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.2918005585670471,
+      "rewards/margins": 0.020267702639102936,
+      "rewards/rejected": 0.2715328335762024,
+      "step": 500
+    },
+    {
+      "epoch": 0.16062359749616156,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.99066456710892e-06,
+      "logits/chosen": -0.5184639096260071,
+      "logits/rejected": -0.3862496316432953,
+      "logps/chosen": -194.21218872070312,
+      "logps/rejected": -152.38967895507812,
+      "loss": 0.6722,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.28834104537963867,
+      "rewards/margins": 0.05140441656112671,
+      "rewards/rejected": 0.23693661391735077,
+      "step": 510
+    },
+    {
+      "epoch": 0.16377307980000788,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.990203927896674e-06,
+      "logits/chosen": -0.5369669198989868,
+      "logits/rejected": -0.42427778244018555,
+      "logps/chosen": -195.22592163085938,
+      "logps/rejected": -173.0403289794922,
+      "loss": 0.6737,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.282554566860199,
+      "rewards/margins": 0.04619991034269333,
+      "rewards/rejected": 0.23635463416576385,
+      "step": 520
+    },
+    {
+      "epoch": 0.16692256210385417,
+      "grad_norm": 2.609375,
+      "learning_rate": 4.9897322183188855e-06,
+      "logits/chosen": -0.49566903710365295,
+      "logits/rejected": -0.42590633034706116,
+      "logps/chosen": -193.32513427734375,
+      "logps/rejected": -163.94113159179688,
+      "loss": 0.68,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.3135472536087036,
+      "rewards/margins": 0.0341959223151207,
+      "rewards/rejected": 0.2793513536453247,
+      "step": 530
+    },
+    {
+      "epoch": 0.1700720444077005,
+      "grad_norm": 2.609375,
+      "learning_rate": 4.989249440472569e-06,
+      "logits/chosen": -0.4742864668369293,
+      "logits/rejected": -0.361924409866333,
+      "logps/chosen": -206.1411590576172,
+      "logps/rejected": -177.32366943359375,
+      "loss": 0.6616,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.2787955701351166,
+      "rewards/margins": 0.07065565139055252,
+      "rewards/rejected": 0.20813994109630585,
+      "step": 540
+    },
+    {
+      "epoch": 0.17322152671154678,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.988755596503948e-06,
+      "logits/chosen": -0.47193408012390137,
+      "logits/rejected": -0.32936495542526245,
+      "logps/chosen": -212.5970916748047,
+      "logps/rejected": -184.22251892089844,
+      "loss": 0.6785,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.3025556802749634,
+      "rewards/margins": 0.03635026142001152,
+      "rewards/rejected": 0.26620543003082275,
+      "step": 550
+    },
+    {
+      "epoch": 0.1763710090153931,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.988250688608436e-06,
+      "logits/chosen": -0.5082842111587524,
+      "logits/rejected": -0.38267362117767334,
+      "logps/chosen": -196.55685424804688,
+      "logps/rejected": -176.552734375,
+      "loss": 0.6866,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.312252402305603,
+      "rewards/margins": 0.022432830184698105,
+      "rewards/rejected": 0.2898195683956146,
+      "step": 560
+    },
+    {
+      "epoch": 0.1795204913192394,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.9877347190306354e-06,
+      "logits/chosen": -0.44343939423561096,
+      "logits/rejected": -0.38272932171821594,
+      "logps/chosen": -187.27857971191406,
+      "logps/rejected": -173.09884643554688,
+      "loss": 0.7007,
+      "rewards/accuracies": 0.44999998807907104,
+      "rewards/chosen": 0.28175923228263855,
+      "rewards/margins": -0.002676494885236025,
+      "rewards/rejected": 0.2844357490539551,
+      "step": 570
+    },
+    {
+      "epoch": 0.1826699736230857,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.987207690064323e-06,
+      "logits/chosen": -0.47279614210128784,
+      "logits/rejected": -0.31698504090309143,
+      "logps/chosen": -226.852783203125,
+      "logps/rejected": -187.92398071289062,
+      "loss": 0.6668,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.3296729326248169,
+      "rewards/margins": 0.06299933046102524,
+      "rewards/rejected": 0.26667362451553345,
+      "step": 580
+    },
+    {
+      "epoch": 0.185819455926932,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.98666960405244e-06,
+      "logits/chosen": -0.5369702577590942,
+      "logits/rejected": -0.4153470993041992,
+      "logps/chosen": -185.39688110351562,
+      "logps/rejected": -150.71243286132812,
+      "loss": 0.6733,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.2983696460723877,
+      "rewards/margins": 0.04920379817485809,
+      "rewards/rejected": 0.24916581809520721,
+      "step": 590
+    },
+    {
+      "epoch": 0.18896893823077832,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.986120463387084e-06,
+      "logits/chosen": -0.4622046947479248,
+      "logits/rejected": -0.3503243923187256,
+      "logps/chosen": -191.92813110351562,
+      "logps/rejected": -176.70423889160156,
+      "loss": 0.6848,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.309493750333786,
+      "rewards/margins": 0.02766551449894905,
+      "rewards/rejected": 0.28182822465896606,
+      "step": 600
+    },
+    {
+      "epoch": 0.19211842053462463,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.985560270509496e-06,
+      "logits/chosen": -0.5258822441101074,
+      "logits/rejected": -0.3977143168449402,
+      "logps/chosen": -215.99563598632812,
+      "logps/rejected": -186.25816345214844,
+      "loss": 0.6636,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.3478468358516693,
+      "rewards/margins": 0.06890861690044403,
+      "rewards/rejected": 0.2789382338523865,
+      "step": 610
+    },
+    {
+      "epoch": 0.19526790283847092,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.9849890279100495e-06,
+      "logits/chosen": -0.49815624952316284,
+      "logits/rejected": -0.4273989796638489,
+      "logps/chosen": -206.10385131835938,
+      "logps/rejected": -180.3357391357422,
+      "loss": 0.6795,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.3327508568763733,
+      "rewards/margins": 0.03660760074853897,
+      "rewards/rejected": 0.29614323377609253,
+      "step": 620
+    },
+    {
+      "epoch": 0.19841738514231724,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.984406738128241e-06,
+      "logits/chosen": -0.4096647799015045,
+      "logits/rejected": -0.3559405505657196,
+      "logps/chosen": -206.38064575195312,
+      "logps/rejected": -185.5215301513672,
+      "loss": 0.6733,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.3558829426765442,
+      "rewards/margins": 0.049683526158332825,
+      "rewards/rejected": 0.3061993718147278,
+      "step": 630
+    },
+    {
+      "epoch": 0.20156686744616353,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.9838134037526795e-06,
+      "logits/chosen": -0.4695689082145691,
+      "logits/rejected": -0.39185652136802673,
+      "logps/chosen": -194.5779266357422,
+      "logps/rejected": -173.6152801513672,
+      "loss": 0.6702,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.33483070135116577,
+      "rewards/margins": 0.05525987595319748,
+      "rewards/rejected": 0.27957087755203247,
+      "step": 640
+    },
+    {
+      "epoch": 0.20471634975000985,
+      "grad_norm": 2.453125,
+      "learning_rate": 4.983209027421072e-06,
+      "logits/chosen": -0.5027821063995361,
+      "logits/rejected": -0.3183223009109497,
+      "logps/chosen": -209.20291137695312,
+      "logps/rejected": -171.07730102539062,
+      "loss": 0.6453,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.3705558776855469,
+      "rewards/margins": 0.10951970517635345,
+      "rewards/rejected": 0.2610361576080322,
+      "step": 650
+    },
+    {
+      "epoch": 0.20786583205385614,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.982593611820211e-06,
+      "logits/chosen": -0.5004459023475647,
+      "logits/rejected": -0.40845292806625366,
+      "logps/chosen": -183.89920043945312,
+      "logps/rejected": -157.65817260742188,
+      "loss": 0.6759,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.3168241083621979,
+      "rewards/margins": 0.04429393634200096,
+      "rewards/rejected": 0.27253013849258423,
+      "step": 660
+    },
+    {
+      "epoch": 0.21101531435770246,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.981967159685969e-06,
+      "logits/chosen": -0.5235536694526672,
+      "logits/rejected": -0.3312370181083679,
+      "logps/chosen": -212.1456298828125,
+      "logps/rejected": -173.4841766357422,
+      "loss": 0.6682,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.37052249908447266,
+      "rewards/margins": 0.06635678559541702,
+      "rewards/rejected": 0.30416572093963623,
+      "step": 670
+    },
+    {
+      "epoch": 0.21416479666154875,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.98132967380328e-06,
+      "logits/chosen": -0.4822749197483063,
+      "logits/rejected": -0.37933364510536194,
+      "logps/chosen": -184.20840454101562,
+      "logps/rejected": -165.41285705566406,
+      "loss": 0.6753,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.3370289206504822,
+      "rewards/margins": 0.04510311037302017,
+      "rewards/rejected": 0.2919257879257202,
+      "step": 680
+    },
+    {
+      "epoch": 0.21731427896539507,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.980681157006129e-06,
+      "logits/chosen": -0.45780807733535767,
+      "logits/rejected": -0.3043513894081116,
+      "logps/chosen": -217.8817901611328,
+      "logps/rejected": -184.9659423828125,
+      "loss": 0.6575,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.38531795144081116,
+      "rewards/margins": 0.08670790493488312,
+      "rewards/rejected": 0.29861006140708923,
+      "step": 690
+    },
+    {
+      "epoch": 0.22046376126924136,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.9800216121775404e-06,
+      "logits/chosen": -0.4495162069797516,
+      "logits/rejected": -0.3232493996620178,
+      "logps/chosen": -209.5262908935547,
+      "logps/rejected": -175.12301635742188,
+      "loss": 0.6529,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.4109676778316498,
+      "rewards/margins": 0.093905009329319,
+      "rewards/rejected": 0.3170626759529114,
+      "step": 700
+    },
+    {
+      "epoch": 0.22361324357308768,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.979351042249564e-06,
+      "logits/chosen": -0.5166658163070679,
+      "logits/rejected": -0.36778515577316284,
+      "logps/chosen": -202.32473754882812,
+      "logps/rejected": -175.3853759765625,
+      "loss": 0.6596,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.3800323009490967,
+      "rewards/margins": 0.07695108652114868,
+      "rewards/rejected": 0.303081214427948,
+      "step": 710
+    },
+    {
+      "epoch": 0.22676272587693397,
+      "grad_norm": 3.75,
+      "learning_rate": 4.978669450203263e-06,
+      "logits/chosen": -0.5254799127578735,
+      "logits/rejected": -0.3579130470752716,
+      "logps/chosen": -201.89544677734375,
+      "logps/rejected": -170.9855194091797,
+      "loss": 0.6548,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.41179266571998596,
+      "rewards/margins": 0.09266375005245209,
+      "rewards/rejected": 0.31912893056869507,
+      "step": 720
+    },
+    {
+      "epoch": 0.2299122081807803,
+      "grad_norm": 2.5,
+      "learning_rate": 4.977976839068699e-06,
+      "logits/chosen": -0.41213518381118774,
+      "logits/rejected": -0.37237733602523804,
+      "logps/chosen": -193.40603637695312,
+      "logps/rejected": -176.70425415039062,
+      "loss": 0.6763,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.3635534644126892,
+      "rewards/margins": 0.05010410025715828,
+      "rewards/rejected": 0.31344935297966003,
+      "step": 730
+    },
+    {
+      "epoch": 0.23306169048462658,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.977273211924921e-06,
+      "logits/chosen": -0.5329135060310364,
+      "logits/rejected": -0.387803316116333,
+      "logps/chosen": -230.49569702148438,
+      "logps/rejected": -199.40280151367188,
+      "loss": 0.6588,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.4647158682346344,
+      "rewards/margins": 0.08450464904308319,
+      "rewards/rejected": 0.38021120429039,
+      "step": 740
+    },
+    {
+      "epoch": 0.2362111727884729,
+      "grad_norm": 2.875,
+      "learning_rate": 4.97655857189995e-06,
+      "logits/chosen": -0.4675888419151306,
+      "logits/rejected": -0.3698478639125824,
+      "logps/chosen": -200.2988739013672,
+      "logps/rejected": -172.2423858642578,
+      "loss": 0.6704,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.38973450660705566,
+      "rewards/margins": 0.05968126654624939,
+      "rewards/rejected": 0.3300532102584839,
+      "step": 750
+    },
+    {
+      "epoch": 0.2393606550923192,
+      "grad_norm": 3.5,
+      "learning_rate": 4.975832922170765e-06,
+      "logits/chosen": -0.48995572328567505,
+      "logits/rejected": -0.3825104236602783,
+      "logps/chosen": -199.5481414794922,
+      "logps/rejected": -186.9451904296875,
+      "loss": 0.6902,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.40951433777809143,
+      "rewards/margins": 0.019469773396849632,
+      "rewards/rejected": 0.39004451036453247,
+      "step": 760
+    },
+    {
+      "epoch": 0.2425101373961655,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.9750962659632886e-06,
+      "logits/chosen": -0.46572384238243103,
+      "logits/rejected": -0.25734081864356995,
+      "logps/chosen": -213.49441528320312,
+      "logps/rejected": -179.05862426757812,
+      "loss": 0.6589,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.4372657835483551,
+      "rewards/margins": 0.08767595142126083,
+      "rewards/rejected": 0.34958982467651367,
+      "step": 770
+    },
+    {
+      "epoch": 0.24565961970001182,
+      "grad_norm": 2.609375,
+      "learning_rate": 4.974348606552377e-06,
+      "logits/chosen": -0.49957194924354553,
+      "logits/rejected": -0.44457465410232544,
+      "logps/chosen": -174.76776123046875,
+      "logps/rejected": -162.83163452148438,
+      "loss": 0.6713,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.37218689918518066,
+      "rewards/margins": 0.05890881270170212,
+      "rewards/rejected": 0.31327807903289795,
+      "step": 780
+    },
+    {
+      "epoch": 0.2488091020038581,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.973589947261797e-06,
+      "logits/chosen": -0.43002453446388245,
+      "logits/rejected": -0.26697424054145813,
+      "logps/chosen": -214.3047637939453,
+      "logps/rejected": -164.3812713623047,
+      "loss": 0.6533,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.4211522042751312,
+      "rewards/margins": 0.09669376909732819,
+      "rewards/rejected": 0.324458509683609,
+      "step": 790
+    },
+    {
+      "epoch": 0.2519585843077044,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.972820291464219e-06,
+      "logits/chosen": -0.5306814312934875,
+      "logits/rejected": -0.3864821195602417,
+      "logps/chosen": -212.0864715576172,
+      "logps/rejected": -191.2584991455078,
+      "loss": 0.6768,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.4268336296081543,
+      "rewards/margins": 0.04500560089945793,
+      "rewards/rejected": 0.3818280100822449,
+      "step": 800
+    },
+    {
+      "epoch": 0.2551080666115507,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.972039642581199e-06,
+      "logits/chosen": -0.44598865509033203,
+      "logits/rejected": -0.3495904505252838,
+      "logps/chosen": -206.1003875732422,
+      "logps/rejected": -180.69091796875,
+      "loss": 0.6704,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.4352218508720398,
+      "rewards/margins": 0.06349250674247742,
+      "rewards/rejected": 0.37172931432724,
+      "step": 810
+    },
+    {
+      "epoch": 0.25825754891539704,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.9712480040831626e-06,
+      "logits/chosen": -0.4507770538330078,
+      "logits/rejected": -0.4034315049648285,
+      "logps/chosen": -201.1985626220703,
+      "logps/rejected": -188.18038940429688,
+      "loss": 0.6691,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.4719354212284088,
+      "rewards/margins": 0.06965798884630203,
+      "rewards/rejected": 0.40227746963500977,
+      "step": 820
+    },
+    {
+      "epoch": 0.26140703121924336,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.9704453794893905e-06,
+      "logits/chosen": -0.4594174921512604,
+      "logits/rejected": -0.3546017110347748,
+      "logps/chosen": -208.15328979492188,
+      "logps/rejected": -175.2351531982422,
+      "loss": 0.6596,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.4524504244327545,
+      "rewards/margins": 0.08146923780441284,
+      "rewards/rejected": 0.3709811568260193,
+      "step": 830
+    },
+    {
+      "epoch": 0.2645565135230896,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.969631772368005e-06,
+      "logits/chosen": -0.44061025977134705,
+      "logits/rejected": -0.3553173840045929,
+      "logps/chosen": -204.01336669921875,
+      "logps/rejected": -194.84194946289062,
+      "loss": 0.683,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.4293643534183502,
+      "rewards/margins": 0.03097158670425415,
+      "rewards/rejected": 0.39839276671409607,
+      "step": 840
+    },
+    {
+      "epoch": 0.26770599582693594,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.968807186335948e-06,
+      "logits/chosen": -0.45513710379600525,
+      "logits/rejected": -0.35907578468322754,
+      "logps/chosen": -187.94784545898438,
+      "logps/rejected": -163.60385131835938,
+      "loss": 0.6449,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.4533543586730957,
+      "rewards/margins": 0.11502598226070404,
+      "rewards/rejected": 0.3383283317089081,
+      "step": 850
+    },
+    {
+      "epoch": 0.27085547813078226,
+      "grad_norm": 2.34375,
+      "learning_rate": 4.9679716250589726e-06,
+      "logits/chosen": -0.53230220079422,
+      "logits/rejected": -0.36705273389816284,
+      "logps/chosen": -214.5038299560547,
+      "logps/rejected": -173.00045776367188,
+      "loss": 0.6439,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.4622381627559662,
+      "rewards/margins": 0.11802558600902557,
+      "rewards/rejected": 0.3442125916481018,
+      "step": 860
+    },
+    {
+      "epoch": 0.2740049604346286,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.96712509225162e-06,
+      "logits/chosen": -0.4010530412197113,
+      "logits/rejected": -0.40078672766685486,
+      "logps/chosen": -189.293212890625,
+      "logps/rejected": -190.86239624023438,
+      "loss": 0.7079,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.3847641050815582,
+      "rewards/margins": -0.01615377888083458,
+      "rewards/rejected": 0.4009179174900055,
+      "step": 870
+    },
+    {
+      "epoch": 0.27715444273847484,
+      "grad_norm": 2.3125,
+      "learning_rate": 4.966267591677209e-06,
+      "logits/chosen": -0.47001272439956665,
+      "logits/rejected": -0.36287882924079895,
+      "logps/chosen": -193.36114501953125,
+      "logps/rejected": -163.00997924804688,
+      "loss": 0.6561,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.412198543548584,
+      "rewards/margins": 0.08650766313076019,
+      "rewards/rejected": 0.3256909251213074,
+      "step": 880
+    },
+    {
+      "epoch": 0.28030392504232116,
+      "grad_norm": 3.40625,
+      "learning_rate": 4.965399127147814e-06,
+      "logits/chosen": -0.5800660848617554,
+      "logits/rejected": -0.4225196838378906,
+      "logps/chosen": -199.9794158935547,
+      "logps/rejected": -175.31942749023438,
+      "loss": 0.6812,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.4509803354740143,
+      "rewards/margins": 0.04268043860793114,
+      "rewards/rejected": 0.40829986333847046,
+      "step": 890
+    },
+    {
+      "epoch": 0.2834534073461675,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.964519702524251e-06,
+      "logits/chosen": -0.48715677857398987,
+      "logits/rejected": -0.3676280677318573,
+      "logps/chosen": -206.2698974609375,
+      "logps/rejected": -180.06561279296875,
+      "loss": 0.6616,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.4541000425815582,
+      "rewards/margins": 0.08144226670265198,
+      "rewards/rejected": 0.37265777587890625,
+      "step": 900
+    },
+    {
+      "epoch": 0.2866028896500138,
+      "grad_norm": 2.75,
+      "learning_rate": 4.9636293217160615e-06,
+      "logits/chosen": -0.4000244736671448,
+      "logits/rejected": -0.37658295035362244,
+      "logps/chosen": -201.72274780273438,
+      "logps/rejected": -197.0117645263672,
+      "loss": 0.708,
+      "rewards/accuracies": 0.44999998807907104,
+      "rewards/chosen": 0.4207178056240082,
+      "rewards/margins": -0.009672733955085278,
+      "rewards/rejected": 0.43039053678512573,
+      "step": 910
+    },
+    {
+      "epoch": 0.2897523719538601,
+      "grad_norm": 2.15625,
+      "learning_rate": 4.96272798868149e-06,
+      "logits/chosen": -0.5054947137832642,
+      "logits/rejected": -0.3827149271965027,
+      "logps/chosen": -193.49334716796875,
+      "logps/rejected": -166.5660858154297,
+      "loss": 0.6608,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.4493057131767273,
+      "rewards/margins": 0.08539381623268127,
+      "rewards/rejected": 0.363911896944046,
+      "step": 920
+    },
+    {
+      "epoch": 0.2929018542577064,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.961815707427473e-06,
+      "logits/chosen": -0.4878782629966736,
+      "logits/rejected": -0.34200000762939453,
+      "logps/chosen": -192.65780639648438,
+      "logps/rejected": -171.5297393798828,
+      "loss": 0.6809,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.43513813614845276,
+      "rewards/margins": 0.04086482524871826,
+      "rewards/rejected": 0.3942733108997345,
+      "step": 930
+    },
+    {
+      "epoch": 0.2960513365615527,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.960892482009617e-06,
+      "logits/chosen": -0.5498200058937073,
+      "logits/rejected": -0.3890989422798157,
+      "logps/chosen": -217.5938262939453,
+      "logps/rejected": -185.64468383789062,
+      "loss": 0.6572,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.4587416648864746,
+      "rewards/margins": 0.08828563988208771,
+      "rewards/rejected": 0.3704560697078705,
+      "step": 940
+    },
+    {
+      "epoch": 0.299200818865399,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.959958316532181e-06,
+      "logits/chosen": -0.4381844103336334,
+      "logits/rejected": -0.39938193559646606,
+      "logps/chosen": -202.2256622314453,
+      "logps/rejected": -184.60812377929688,
+      "loss": 0.6555,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.47266441583633423,
+      "rewards/margins": 0.08841115981340408,
+      "rewards/rejected": 0.38425326347351074,
+      "step": 950
+    },
+    {
+      "epoch": 0.30235030116924533,
+      "grad_norm": 2.875,
+      "learning_rate": 4.959013215148059e-06,
+      "logits/chosen": -0.516729474067688,
+      "logits/rejected": -0.37376007437705994,
+      "logps/chosen": -206.91470336914062,
+      "logps/rejected": -165.06771850585938,
+      "loss": 0.6483,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.48910826444625854,
+      "rewards/margins": 0.11284583806991577,
+      "rewards/rejected": 0.3762624263763428,
+      "step": 960
+    },
+    {
+      "epoch": 0.3054997834730916,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.958057182058763e-06,
+      "logits/chosen": -0.4985506534576416,
+      "logits/rejected": -0.3460918366909027,
+      "logps/chosen": -204.8501434326172,
+      "logps/rejected": -156.81150817871094,
+      "loss": 0.6317,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.5237672924995422,
+      "rewards/margins": 0.14426377415657043,
+      "rewards/rejected": 0.3795034885406494,
+      "step": 970
+    },
+    {
+      "epoch": 0.3086492657769379,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.957090221514399e-06,
+      "logits/chosen": -0.44721898436546326,
+      "logits/rejected": -0.3173277676105499,
+      "logps/chosen": -204.72488403320312,
+      "logps/rejected": -172.4529571533203,
+      "loss": 0.6822,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.43838948011398315,
+      "rewards/margins": 0.04825422167778015,
+      "rewards/rejected": 0.3901353180408478,
+      "step": 980
+    },
+    {
+      "epoch": 0.3117987480807842,
+      "grad_norm": 3.53125,
+      "learning_rate": 4.956112337813655e-06,
+      "logits/chosen": -0.5343712568283081,
+      "logits/rejected": -0.42165470123291016,
+      "logps/chosen": -196.4682159423828,
+      "logps/rejected": -159.38626098632812,
+      "loss": 0.6411,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.49275222420692444,
+      "rewards/margins": 0.12367204576730728,
+      "rewards/rejected": 0.369080126285553,
+      "step": 990
+    },
+    {
+      "epoch": 0.31494823038463055,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.955123535303775e-06,
+      "logits/chosen": -0.5008007884025574,
+      "logits/rejected": -0.32471761107444763,
+      "logps/chosen": -215.82382202148438,
+      "logps/rejected": -170.16177368164062,
+      "loss": 0.6278,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.555659294128418,
+      "rewards/margins": 0.15915410220623016,
+      "rewards/rejected": 0.3965051770210266,
+      "step": 1000
+    },
+    {
+      "epoch": 0.3180977126884768,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.95412381838055e-06,
+      "logits/chosen": -0.45208463072776794,
+      "logits/rejected": -0.3023605942726135,
+      "logps/chosen": -211.84561157226562,
+      "logps/rejected": -175.75242614746094,
+      "loss": 0.6482,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.5254601836204529,
+      "rewards/margins": 0.1151600107550621,
+      "rewards/rejected": 0.41030019521713257,
+      "step": 1010
+    },
+    {
+      "epoch": 0.3212471949923231,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.953113191488284e-06,
+      "logits/chosen": -0.5526930093765259,
+      "logits/rejected": -0.4031393527984619,
+      "logps/chosen": -200.52316284179688,
+      "logps/rejected": -160.85092163085938,
+      "loss": 0.6635,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.47668904066085815,
+      "rewards/margins": 0.08161365985870361,
+      "rewards/rejected": 0.3950754404067993,
+      "step": 1020
+    },
+    {
+      "epoch": 0.32439667729616944,
+      "grad_norm": 2.375,
+      "learning_rate": 4.9520916591197865e-06,
+      "logits/chosen": -0.49668893218040466,
+      "logits/rejected": -0.3756228983402252,
+      "logps/chosen": -201.33413696289062,
+      "logps/rejected": -169.83847045898438,
+      "loss": 0.6651,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.5008379220962524,
+      "rewards/margins": 0.08502224832773209,
+      "rewards/rejected": 0.41581565141677856,
+      "step": 1030
+    },
+    {
+      "epoch": 0.32754615960001576,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.951059225816347e-06,
+      "logits/chosen": -0.4749979078769684,
+      "logits/rejected": -0.30828791856765747,
+      "logps/chosen": -224.30178833007812,
+      "logps/rejected": -178.66122436523438,
+      "loss": 0.6517,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.5890859961509705,
+      "rewards/margins": 0.12044923007488251,
+      "rewards/rejected": 0.46863681077957153,
+      "step": 1040
+    },
+    {
+      "epoch": 0.330695641903862,
+      "grad_norm": 2.203125,
+      "learning_rate": 4.950015896167716e-06,
+      "logits/chosen": -0.47923216223716736,
+      "logits/rejected": -0.28145602345466614,
+      "logps/chosen": -189.4966583251953,
+      "logps/rejected": -161.24319458007812,
+      "loss": 0.6675,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.4908905029296875,
+      "rewards/margins": 0.07855306565761566,
+      "rewards/rejected": 0.41233739256858826,
+      "step": 1050
+    },
+    {
+      "epoch": 0.33384512420770834,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.948961674812083e-06,
+      "logits/chosen": -0.49384012818336487,
+      "logits/rejected": -0.36533522605895996,
+      "logps/chosen": -198.85769653320312,
+      "logps/rejected": -174.79847717285156,
+      "loss": 0.6573,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.5101815462112427,
+      "rewards/margins": 0.08699540048837662,
+      "rewards/rejected": 0.42318612337112427,
+      "step": 1060
+    },
+    {
+      "epoch": 0.33699460651155466,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.9478965664360595e-06,
+      "logits/chosen": -0.5174117088317871,
+      "logits/rejected": -0.3790570795536041,
+      "logps/chosen": -199.65078735351562,
+      "logps/rejected": -159.61801147460938,
+      "loss": 0.6417,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.5138002634048462,
+      "rewards/margins": 0.12827709317207336,
+      "rewards/rejected": 0.38552325963974,
+      "step": 1070
+    },
+    {
+      "epoch": 0.340144088815401,
+      "grad_norm": 2.34375,
+      "learning_rate": 4.946820575774654e-06,
+      "logits/chosen": -0.4343532919883728,
+      "logits/rejected": -0.31658726930618286,
+      "logps/chosen": -199.5583038330078,
+      "logps/rejected": -162.34683227539062,
+      "loss": 0.6508,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.5402163863182068,
+      "rewards/margins": 0.12262473255395889,
+      "rewards/rejected": 0.41759172081947327,
+      "step": 1080
+    },
+    {
+      "epoch": 0.3432935711192473,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.945733707611256e-06,
+      "logits/chosen": -0.42169666290283203,
+      "logits/rejected": -0.27337223291397095,
+      "logps/chosen": -210.4072265625,
+      "logps/rejected": -176.20767211914062,
+      "loss": 0.6267,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.6182594299316406,
+      "rewards/margins": 0.16625744104385376,
+      "rewards/rejected": 0.45200204849243164,
+      "step": 1090
+    },
+    {
+      "epoch": 0.34644305342309356,
+      "grad_norm": 2.234375,
+      "learning_rate": 4.944635966777607e-06,
+      "logits/chosen": -0.5145076513290405,
+      "logits/rejected": -0.3866254985332489,
+      "logps/chosen": -186.8591766357422,
+      "logps/rejected": -147.5457305908203,
+      "loss": 0.6267,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.5492256283760071,
+      "rewards/margins": 0.15818223357200623,
+      "rewards/rejected": 0.39104336500167847,
+      "step": 1100
+    },
+    {
+      "epoch": 0.3495925357269399,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.943527358153787e-06,
+      "logits/chosen": -0.4312458634376526,
+      "logits/rejected": -0.2944316267967224,
+      "logps/chosen": -176.56130981445312,
+      "logps/rejected": -144.69210815429688,
+      "loss": 0.6711,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.5271092057228088,
+      "rewards/margins": 0.074883833527565,
+      "rewards/rejected": 0.45222535729408264,
+      "step": 1110
+    },
+    {
+      "epoch": 0.3527420180307862,
+      "grad_norm": 3.25,
+      "learning_rate": 4.942407886668189e-06,
+      "logits/chosen": -0.4668591022491455,
+      "logits/rejected": -0.4000996947288513,
+      "logps/chosen": -192.39215087890625,
+      "logps/rejected": -181.41018676757812,
+      "loss": 0.6692,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.5816969871520996,
+      "rewards/margins": 0.07639260590076447,
+      "rewards/rejected": 0.5053043365478516,
+      "step": 1120
+    },
+    {
+      "epoch": 0.3558915003346325,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.941277557297497e-06,
+      "logits/chosen": -0.5000630617141724,
+      "logits/rejected": -0.3792869448661804,
+      "logps/chosen": -201.1351776123047,
+      "logps/rejected": -163.50692749023438,
+      "loss": 0.6669,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.5100800395011902,
+      "rewards/margins": 0.07874707877635956,
+      "rewards/rejected": 0.4313329756259918,
+      "step": 1130
+    },
+    {
+      "epoch": 0.3590409826384788,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.940136375066664e-06,
+      "logits/chosen": -0.5313334465026855,
+      "logits/rejected": -0.4073900580406189,
+      "logps/chosen": -195.81436157226562,
+      "logps/rejected": -165.0396728515625,
+      "loss": 0.6488,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.5570641756057739,
+      "rewards/margins": 0.12459216266870499,
+      "rewards/rejected": 0.4324720501899719,
+      "step": 1140
+    },
+    {
+      "epoch": 0.3621904649423251,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.938984345048892e-06,
+      "logits/chosen": -0.47552186250686646,
+      "logits/rejected": -0.3426175117492676,
+      "logps/chosen": -226.5735626220703,
+      "logps/rejected": -185.11488342285156,
+      "loss": 0.6607,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.6292687654495239,
+      "rewards/margins": 0.10300488770008087,
+      "rewards/rejected": 0.5262638926506042,
+      "step": 1150
+    },
+    {
+      "epoch": 0.3653399472461714,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.937821472365606e-06,
+      "logits/chosen": -0.45469918847084045,
+      "logits/rejected": -0.2649703621864319,
+      "logps/chosen": -203.91030883789062,
+      "logps/rejected": -154.42904663085938,
+      "loss": 0.6364,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.614439845085144,
+      "rewards/margins": 0.14579293131828308,
+      "rewards/rejected": 0.46864691376686096,
+      "step": 1160
+    },
+    {
+      "epoch": 0.36848942955001773,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.9366477621864325e-06,
+      "logits/chosen": -0.45428067445755005,
+      "logits/rejected": -0.38431745767593384,
+      "logps/chosen": -193.93759155273438,
+      "logps/rejected": -166.46226501464844,
+      "loss": 0.6829,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.5578614473342896,
+      "rewards/margins": 0.04849938303232193,
+      "rewards/rejected": 0.5093621015548706,
+      "step": 1170
+    },
+    {
+      "epoch": 0.371638911853864,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.935463219729178e-06,
+      "logits/chosen": -0.43662381172180176,
+      "logits/rejected": -0.3796136975288391,
+      "logps/chosen": -208.486083984375,
+      "logps/rejected": -186.07241821289062,
+      "loss": 0.6391,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.5909144878387451,
+      "rewards/margins": 0.12893392145633698,
+      "rewards/rejected": 0.46198058128356934,
+      "step": 1180
+    },
+    {
+      "epoch": 0.3747883941577103,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.934267850259802e-06,
+      "logits/chosen": -0.5133141875267029,
+      "logits/rejected": -0.36373409628868103,
+      "logps/chosen": -194.3529815673828,
+      "logps/rejected": -159.5540771484375,
+      "loss": 0.6617,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.5327507853507996,
+      "rewards/margins": 0.08738649636507034,
+      "rewards/rejected": 0.4453642964363098,
+      "step": 1190
+    },
+    {
+      "epoch": 0.37793787646155663,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.933061659092401e-06,
+      "logits/chosen": -0.45688313245773315,
+      "logits/rejected": -0.3101183772087097,
+      "logps/chosen": -197.06399536132812,
+      "logps/rejected": -178.8423309326172,
+      "loss": 0.6921,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.5263506174087524,
+      "rewards/margins": 0.027934294193983078,
+      "rewards/rejected": 0.4984162747859955,
+      "step": 1200
+    },
+    {
+      "epoch": 0.38108735876540295,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.931844651589176e-06,
+      "logits/chosen": -0.3927622437477112,
+      "logits/rejected": -0.3346686065196991,
+      "logps/chosen": -192.22146606445312,
+      "logps/rejected": -177.29522705078125,
+      "loss": 0.6793,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.5813957452774048,
+      "rewards/margins": 0.05229301005601883,
+      "rewards/rejected": 0.5291028022766113,
+      "step": 1210
+    },
+    {
+      "epoch": 0.38423684106924927,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.930616833160414e-06,
+      "logits/chosen": -0.503675639629364,
+      "logits/rejected": -0.3028663694858551,
+      "logps/chosen": -209.7771759033203,
+      "logps/rejected": -166.1737060546875,
+      "loss": 0.6183,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.6333845853805542,
+      "rewards/margins": 0.17845922708511353,
+      "rewards/rejected": 0.45492544770240784,
+      "step": 1220
+    },
+    {
+      "epoch": 0.38738632337309553,
+      "grad_norm": 2.359375,
+      "learning_rate": 4.929378209264464e-06,
+      "logits/chosen": -0.4029599130153656,
+      "logits/rejected": -0.2958356738090515,
+      "logps/chosen": -201.47354125976562,
+      "logps/rejected": -175.35316467285156,
+      "loss": 0.6546,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.6188668012619019,
+      "rewards/margins": 0.10686089843511581,
+      "rewards/rejected": 0.5120059251785278,
+      "step": 1230
+    },
+    {
+      "epoch": 0.39053580567694185,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.9281287854077075e-06,
+      "logits/chosen": -0.5093222856521606,
+      "logits/rejected": -0.3469759523868561,
+      "logps/chosen": -203.4248809814453,
+      "logps/rejected": -168.985595703125,
+      "loss": 0.6675,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.5747894048690796,
+      "rewards/margins": 0.08418375253677368,
+      "rewards/rejected": 0.4906056821346283,
+      "step": 1240
+    },
+    {
+      "epoch": 0.39368528798078817,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.926868567144543e-06,
+      "logits/chosen": -0.43734902143478394,
+      "logits/rejected": -0.27301496267318726,
+      "logps/chosen": -217.2354736328125,
+      "logps/rejected": -179.50477600097656,
+      "loss": 0.6439,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.602939784526825,
+      "rewards/margins": 0.13366705179214478,
+      "rewards/rejected": 0.46927279233932495,
+      "step": 1250
+    },
+    {
+      "epoch": 0.3968347702846345,
+      "grad_norm": 2.015625,
+      "learning_rate": 4.9255975600773506e-06,
+      "logits/chosen": -0.46365243196487427,
+      "logits/rejected": -0.369037002325058,
+      "logps/chosen": -193.8307647705078,
+      "logps/rejected": -167.16226196289062,
+      "loss": 0.6528,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.5885978937149048,
+      "rewards/margins": 0.11199425160884857,
+      "rewards/rejected": 0.4766036570072174,
+      "step": 1260
+    },
+    {
+      "epoch": 0.39998425258848075,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.92431576985648e-06,
+      "logits/chosen": -0.5200837850570679,
+      "logits/rejected": -0.3553524315357208,
+      "logps/chosen": -207.01754760742188,
+      "logps/rejected": -175.571533203125,
+      "loss": 0.6289,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.6721639633178711,
+      "rewards/margins": 0.16469912230968475,
+      "rewards/rejected": 0.5074647665023804,
+      "step": 1270
+    },
+    {
+      "epoch": 0.40313373489232707,
+      "grad_norm": 2.828125,
+      "learning_rate": 4.9230232021802116e-06,
+      "logits/chosen": -0.4531930088996887,
+      "logits/rejected": -0.33995673060417175,
+      "logps/chosen": -182.60243225097656,
+      "logps/rejected": -167.54733276367188,
+      "loss": 0.6632,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.5690507292747498,
+      "rewards/margins": 0.08255408704280853,
+      "rewards/rejected": 0.4864966869354248,
+      "step": 1280
+    },
+    {
+      "epoch": 0.4062832171961734,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.921719862794741e-06,
+      "logits/chosen": -0.5172183513641357,
+      "logits/rejected": -0.43691587448120117,
+      "logps/chosen": -198.16488647460938,
+      "logps/rejected": -176.88308715820312,
+      "loss": 0.6577,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.5772544741630554,
+      "rewards/margins": 0.09580488502979279,
+      "rewards/rejected": 0.4814496636390686,
+      "step": 1290
+    },
+    {
+      "epoch": 0.4094326995000197,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.920405757494147e-06,
+      "logits/chosen": -0.5242056846618652,
+      "logits/rejected": -0.41546088457107544,
+      "logps/chosen": -205.5856475830078,
+      "logps/rejected": -165.4510040283203,
+      "loss": 0.6403,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.6132642030715942,
+      "rewards/margins": 0.13079342246055603,
+      "rewards/rejected": 0.4824707508087158,
+      "step": 1300
+    },
+    {
+      "epoch": 0.41258218180386597,
+      "grad_norm": 3.78125,
+      "learning_rate": 4.919080892120375e-06,
+      "logits/chosen": -0.4625665247440338,
+      "logits/rejected": -0.36404842138290405,
+      "logps/chosen": -186.5456085205078,
+      "logps/rejected": -172.53269958496094,
+      "loss": 0.6757,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.5738548040390015,
+      "rewards/margins": 0.07151280343532562,
+      "rewards/rejected": 0.5023420453071594,
+      "step": 1310
+    },
+    {
+      "epoch": 0.4157316641077123,
+      "grad_norm": 2.21875,
+      "learning_rate": 4.917745272563198e-06,
+      "logits/chosen": -0.514171302318573,
+      "logits/rejected": -0.4131143093109131,
+      "logps/chosen": -190.60104370117188,
+      "logps/rejected": -163.16876220703125,
+      "loss": 0.6377,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.644982635974884,
+      "rewards/margins": 0.149316668510437,
+      "rewards/rejected": 0.49566593766212463,
+      "step": 1320
+    },
+    {
+      "epoch": 0.4188811464115586,
+      "grad_norm": 3.25,
+      "learning_rate": 4.916398904760202e-06,
+      "logits/chosen": -0.407987117767334,
+      "logits/rejected": -0.3450384736061096,
+      "logps/chosen": -204.9993133544922,
+      "logps/rejected": -177.96536254882812,
+      "loss": 0.6839,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.6174831390380859,
+      "rewards/margins": 0.06308461725711823,
+      "rewards/rejected": 0.5543986558914185,
+      "step": 1330
+    },
+    {
+      "epoch": 0.4220306287154049,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.915041794696755e-06,
+      "logits/chosen": -0.4585428833961487,
+      "logits/rejected": -0.32048022747039795,
+      "logps/chosen": -205.7061004638672,
+      "logps/rejected": -188.31765747070312,
+      "loss": 0.6532,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6447404623031616,
+      "rewards/margins": 0.11661551147699356,
+      "rewards/rejected": 0.5281249284744263,
+      "step": 1340
+    },
+    {
+      "epoch": 0.4251801110192512,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.913673948405977e-06,
+      "logits/chosen": -0.425508975982666,
+      "logits/rejected": -0.3008427619934082,
+      "logps/chosen": -188.8653564453125,
+      "logps/rejected": -154.46444702148438,
+      "loss": 0.6551,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.5882565379142761,
+      "rewards/margins": 0.11206309497356415,
+      "rewards/rejected": 0.4761934280395508,
+      "step": 1350
+    },
+    {
+      "epoch": 0.4283295933230975,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.91229537196872e-06,
+      "logits/chosen": -0.466022253036499,
+      "logits/rejected": -0.40916579961776733,
+      "logps/chosen": -188.9302978515625,
+      "logps/rejected": -176.31539916992188,
+      "loss": 0.6762,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.5576030015945435,
+      "rewards/margins": 0.059089891612529755,
+      "rewards/rejected": 0.4985131323337555,
+      "step": 1360
+    },
+    {
+      "epoch": 0.4314790756269438,
+      "grad_norm": 2.203125,
+      "learning_rate": 4.910906071513536e-06,
+      "logits/chosen": -0.468529611825943,
+      "logits/rejected": -0.3725103735923767,
+      "logps/chosen": -189.84571838378906,
+      "logps/rejected": -157.89552307128906,
+      "loss": 0.6576,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.5729373693466187,
+      "rewards/margins": 0.10841979831457138,
+      "rewards/rejected": 0.46451759338378906,
+      "step": 1370
+    },
+    {
+      "epoch": 0.43462855793079014,
+      "grad_norm": 2.578125,
+      "learning_rate": 4.9095060532166515e-06,
+      "logits/chosen": -0.4988088607788086,
+      "logits/rejected": -0.3795704245567322,
+      "logps/chosen": -187.31626892089844,
+      "logps/rejected": -166.83358764648438,
+      "loss": 0.6576,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.5874306559562683,
+      "rewards/margins": 0.10258360207080841,
+      "rewards/rejected": 0.4848470687866211,
+      "step": 1380
+    },
+    {
+      "epoch": 0.43777804023463646,
+      "grad_norm": 2.875,
+      "learning_rate": 4.90809532330194e-06,
+      "logits/chosen": -0.44124871492385864,
+      "logits/rejected": -0.33731183409690857,
+      "logps/chosen": -193.7071990966797,
+      "logps/rejected": -182.87266540527344,
+      "loss": 0.6483,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.6297098398208618,
+      "rewards/margins": 0.11017205566167831,
+      "rewards/rejected": 0.5195378065109253,
+      "step": 1390
+    },
+    {
+      "epoch": 0.4409275225384827,
+      "grad_norm": 2.75,
+      "learning_rate": 4.906673888040895e-06,
+      "logits/chosen": -0.470440149307251,
+      "logits/rejected": -0.40894460678100586,
+      "logps/chosen": -201.374755859375,
+      "logps/rejected": -162.78060913085938,
+      "loss": 0.6548,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.6208814382553101,
+      "rewards/margins": 0.12041078507900238,
+      "rewards/rejected": 0.5004706382751465,
+      "step": 1400
+    },
+    {
+      "epoch": 0.44407700484232904,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.905241753752599e-06,
+      "logits/chosen": -0.45281878113746643,
+      "logits/rejected": -0.3816523253917694,
+      "logps/chosen": -187.10189819335938,
+      "logps/rejected": -180.3949737548828,
+      "loss": 0.6594,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6287654638290405,
+      "rewards/margins": 0.09312457591295242,
+      "rewards/rejected": 0.5356410145759583,
+      "step": 1410
+    },
+    {
+      "epoch": 0.44722648714617536,
+      "grad_norm": 2.390625,
+      "learning_rate": 4.903798926803701e-06,
+      "logits/chosen": -0.5540148615837097,
+      "logits/rejected": -0.31469932198524475,
+      "logps/chosen": -195.7555694580078,
+      "logps/rejected": -160.02955627441406,
+      "loss": 0.6663,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.5776636004447937,
+      "rewards/margins": 0.08535424619913101,
+      "rewards/rejected": 0.4923093318939209,
+      "step": 1420
+    },
+    {
+      "epoch": 0.4503759694500217,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.902345413608382e-06,
+      "logits/chosen": -0.5317307710647583,
+      "logits/rejected": -0.4287118911743164,
+      "logps/chosen": -212.67172241210938,
+      "logps/rejected": -184.7850799560547,
+      "loss": 0.6522,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.6378628015518188,
+      "rewards/margins": 0.11098313331604004,
+      "rewards/rejected": 0.5268796682357788,
+      "step": 1430
+    },
+    {
+      "epoch": 0.45352545175386794,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.900881220628332e-06,
+      "logits/chosen": -0.4813712537288666,
+      "logits/rejected": -0.39985647797584534,
+      "logps/chosen": -198.0911407470703,
+      "logps/rejected": -183.64688110351562,
+      "loss": 0.6738,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.603533148765564,
+      "rewards/margins": 0.07522164285182953,
+      "rewards/rejected": 0.528311550617218,
+      "step": 1440
+    },
+    {
+      "epoch": 0.45667493405771425,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.899406354372716e-06,
+      "logits/chosen": -0.48485565185546875,
+      "logits/rejected": -0.335989385843277,
+      "logps/chosen": -213.813720703125,
+      "logps/rejected": -169.41334533691406,
+      "loss": 0.6284,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.667735755443573,
+      "rewards/margins": 0.18240274488925934,
+      "rewards/rejected": 0.48533302545547485,
+      "step": 1450
+    },
+    {
+      "epoch": 0.4598244163615606,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.897920821398149e-06,
+      "logits/chosen": -0.36127400398254395,
+      "logits/rejected": -0.33091285824775696,
+      "logps/chosen": -198.51376342773438,
+      "logps/rejected": -179.6190643310547,
+      "loss": 0.6766,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.6157578825950623,
+      "rewards/margins": 0.06863512098789215,
+      "rewards/rejected": 0.5471227765083313,
+      "step": 1460
+    },
+    {
+      "epoch": 0.4629738986654069,
+      "grad_norm": 2.34375,
+      "learning_rate": 4.896424628308666e-06,
+      "logits/chosen": -0.49601975083351135,
+      "logits/rejected": -0.3733476996421814,
+      "logps/chosen": -201.9625701904297,
+      "logps/rejected": -163.45664978027344,
+      "loss": 0.646,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.6690108180046082,
+      "rewards/margins": 0.130323126912117,
+      "rewards/rejected": 0.5386877059936523,
+      "step": 1470
+    },
+    {
+      "epoch": 0.46612338096925315,
+      "grad_norm": 2.625,
+      "learning_rate": 4.894917781755693e-06,
+      "logits/chosen": -0.40988796949386597,
+      "logits/rejected": -0.2908838391304016,
+      "logps/chosen": -174.8596954345703,
+      "logps/rejected": -152.51849365234375,
+      "loss": 0.6666,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.5752810835838318,
+      "rewards/margins": 0.07995191216468811,
+      "rewards/rejected": 0.4953291416168213,
+      "step": 1480
+    },
+    {
+      "epoch": 0.46927286327309947,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.893400288438013e-06,
+      "logits/chosen": -0.4711819291114807,
+      "logits/rejected": -0.34161874651908875,
+      "logps/chosen": -210.2778778076172,
+      "logps/rejected": -188.28152465820312,
+      "loss": 0.652,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.6171534061431885,
+      "rewards/margins": 0.11316442489624023,
+      "rewards/rejected": 0.5039889216423035,
+      "step": 1490
+    },
+    {
+      "epoch": 0.4724223455769458,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.891872155101746e-06,
+      "logits/chosen": -0.45833712816238403,
+      "logits/rejected": -0.3544694781303406,
+      "logps/chosen": -205.4825439453125,
+      "logps/rejected": -174.62255859375,
+      "loss": 0.6504,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.6301782131195068,
+      "rewards/margins": 0.11798008531332016,
+      "rewards/rejected": 0.5121980905532837,
+      "step": 1500
+    },
+    {
+      "epoch": 0.4755718278807921,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.890333388540306e-06,
+      "logits/chosen": -0.3727184236049652,
+      "logits/rejected": -0.2826997637748718,
+      "logps/chosen": -219.3084716796875,
+      "logps/rejected": -186.0581512451172,
+      "loss": 0.6574,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6568127870559692,
+      "rewards/margins": 0.10541248321533203,
+      "rewards/rejected": 0.551400363445282,
+      "step": 1510
+    },
+    {
+      "epoch": 0.4787213101846384,
+      "grad_norm": 4.21875,
+      "learning_rate": 4.888783995594383e-06,
+      "logits/chosen": -0.5030876398086548,
+      "logits/rejected": -0.41888195276260376,
+      "logps/chosen": -210.5947723388672,
+      "logps/rejected": -189.0829620361328,
+      "loss": 0.6691,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.6370301842689514,
+      "rewards/margins": 0.07603044807910919,
+      "rewards/rejected": 0.5609997510910034,
+      "step": 1520
+    },
+    {
+      "epoch": 0.4818707924884847,
+      "grad_norm": 2.828125,
+      "learning_rate": 4.887223983151905e-06,
+      "logits/chosen": -0.45775994658470154,
+      "logits/rejected": -0.3528696894645691,
+      "logps/chosen": -205.0533447265625,
+      "logps/rejected": -169.550048828125,
+      "loss": 0.6478,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.6256412863731384,
+      "rewards/margins": 0.12434478104114532,
+      "rewards/rejected": 0.5012965202331543,
+      "step": 1530
+    },
+    {
+      "epoch": 0.485020274792331,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.88565335814801e-06,
+      "logits/chosen": -0.4910427927970886,
+      "logits/rejected": -0.3841659128665924,
+      "logps/chosen": -224.6226348876953,
+      "logps/rejected": -191.9573516845703,
+      "loss": 0.6349,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.6829046010971069,
+      "rewards/margins": 0.15164795517921448,
+      "rewards/rejected": 0.5312565565109253,
+      "step": 1540
+    },
+    {
+      "epoch": 0.4881697570961773,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.884072127565015e-06,
+      "logits/chosen": -0.5355531573295593,
+      "logits/rejected": -0.4093754291534424,
+      "logps/chosen": -190.81552124023438,
+      "logps/rejected": -157.70872497558594,
+      "loss": 0.6334,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.6082229018211365,
+      "rewards/margins": 0.15607663989067078,
+      "rewards/rejected": 0.4521462321281433,
+      "step": 1550
+    },
+    {
+      "epoch": 0.49131923940002364,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.882480298432384e-06,
+      "logits/chosen": -0.4936140179634094,
+      "logits/rejected": -0.3383339047431946,
+      "logps/chosen": -200.6589813232422,
+      "logps/rejected": -171.00064086914062,
+      "loss": 0.6331,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.6500834226608276,
+      "rewards/margins": 0.15458612143993378,
+      "rewards/rejected": 0.49549728631973267,
+      "step": 1560
+    },
+    {
+      "epoch": 0.4944687217038699,
+      "grad_norm": 3.578125,
+      "learning_rate": 4.8808778778266985e-06,
+      "logits/chosen": -0.4825199246406555,
+      "logits/rejected": -0.3065822124481201,
+      "logps/chosen": -222.0049285888672,
+      "logps/rejected": -180.4205780029297,
+      "loss": 0.6483,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.7011119723320007,
+      "rewards/margins": 0.13294753432273865,
+      "rewards/rejected": 0.5681644678115845,
+      "step": 1570
+    },
+    {
+      "epoch": 0.4976182040077162,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.879264872871625e-06,
+      "logits/chosen": -0.47797495126724243,
+      "logits/rejected": -0.3094409704208374,
+      "logps/chosen": -211.0563201904297,
+      "logps/rejected": -168.85647583007812,
+      "loss": 0.6436,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.65972501039505,
+      "rewards/margins": 0.13198061287403107,
+      "rewards/rejected": 0.5277442932128906,
+      "step": 1580
+    },
+    {
+      "epoch": 0.5007676863115625,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.8776412907378845e-06,
+      "logits/chosen": -0.4951102137565613,
+      "logits/rejected": -0.4047602117061615,
+      "logps/chosen": -200.99998474121094,
+      "logps/rejected": -183.0837860107422,
+      "loss": 0.6603,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.6028062105178833,
+      "rewards/margins": 0.10012507438659668,
+      "rewards/rejected": 0.5026811361312866,
+      "step": 1590
+    },
+    {
+      "epoch": 0.5039171686154088,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.876007138643216e-06,
+      "logits/chosen": -0.47414493560791016,
+      "logits/rejected": -0.40259408950805664,
+      "logps/chosen": -184.4724884033203,
+      "logps/rejected": -163.58192443847656,
+      "loss": 0.6457,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.6546593904495239,
+      "rewards/margins": 0.1281435191631317,
+      "rewards/rejected": 0.5265159010887146,
+      "step": 1600
+    },
+    {
+      "epoch": 0.5070666509192552,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.874362423852352e-06,
+      "logits/chosen": -0.5160477161407471,
+      "logits/rejected": -0.43918365240097046,
+      "logps/chosen": -185.5178985595703,
+      "logps/rejected": -165.6835479736328,
+      "loss": 0.6668,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.6197860836982727,
+      "rewards/margins": 0.08066993951797485,
+      "rewards/rejected": 0.5391160845756531,
+      "step": 1610
+    },
+    {
+      "epoch": 0.5102161332231014,
+      "grad_norm": 3.40625,
+      "learning_rate": 4.872707153676979e-06,
+      "logits/chosen": -0.466305673122406,
+      "logits/rejected": -0.3604885935783386,
+      "logps/chosen": -213.26834106445312,
+      "logps/rejected": -181.22439575195312,
+      "loss": 0.6655,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.6425263285636902,
+      "rewards/margins": 0.08319854736328125,
+      "rewards/rejected": 0.5593277812004089,
+      "step": 1620
+    },
+    {
+      "epoch": 0.5133656155269477,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.871041335475712e-06,
+      "logits/chosen": -0.5054647922515869,
+      "logits/rejected": -0.4364451467990875,
+      "logps/chosen": -186.38011169433594,
+      "logps/rejected": -157.52462768554688,
+      "loss": 0.674,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.5790429711341858,
+      "rewards/margins": 0.07125049829483032,
+      "rewards/rejected": 0.5077924728393555,
+      "step": 1630
+    },
+    {
+      "epoch": 0.5165150978307941,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.869364976654052e-06,
+      "logits/chosen": -0.5144768953323364,
+      "logits/rejected": -0.4074745774269104,
+      "logps/chosen": -209.39431762695312,
+      "logps/rejected": -187.3798065185547,
+      "loss": 0.6849,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.655154824256897,
+      "rewards/margins": 0.0526747927069664,
+      "rewards/rejected": 0.6024800539016724,
+      "step": 1640
+    },
+    {
+      "epoch": 0.5196645801346403,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.867678084664365e-06,
+      "logits/chosen": -0.4594438672065735,
+      "logits/rejected": -0.3518297076225281,
+      "logps/chosen": -200.11427307128906,
+      "logps/rejected": -160.406494140625,
+      "loss": 0.6205,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.6651702523231506,
+      "rewards/margins": 0.18755197525024414,
+      "rewards/rejected": 0.4776183068752289,
+      "step": 1650
+    },
+    {
+      "epoch": 0.5228140624384867,
+      "grad_norm": 2.515625,
+      "learning_rate": 4.865980667005839e-06,
+      "logits/chosen": -0.5308347940444946,
+      "logits/rejected": -0.38161686062812805,
+      "logps/chosen": -228.52413940429688,
+      "logps/rejected": -180.50119018554688,
+      "loss": 0.6231,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.7223028540611267,
+      "rewards/margins": 0.1841433346271515,
+      "rewards/rejected": 0.5381595492362976,
+      "step": 1660
+    },
+    {
+      "epoch": 0.525963544742333,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.864272731224457e-06,
+      "logits/chosen": -0.4550582468509674,
+      "logits/rejected": -0.36829763650894165,
+      "logps/chosen": -199.5106964111328,
+      "logps/rejected": -177.867919921875,
+      "loss": 0.6481,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.7055026888847351,
+      "rewards/margins": 0.13270851969718933,
+      "rewards/rejected": 0.5727940797805786,
+      "step": 1670
+    },
+    {
+      "epoch": 0.5291130270461792,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.862554284912961e-06,
+      "logits/chosen": -0.5029420256614685,
+      "logits/rejected": -0.37519973516464233,
+      "logps/chosen": -193.17857360839844,
+      "logps/rejected": -151.67193603515625,
+      "loss": 0.6445,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.5841437578201294,
+      "rewards/margins": 0.12721844017505646,
+      "rewards/rejected": 0.4569253921508789,
+      "step": 1680
+    },
+    {
+      "epoch": 0.5322625093500256,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.860825335710815e-06,
+      "logits/chosen": -0.4503244459629059,
+      "logits/rejected": -0.33839210867881775,
+      "logps/chosen": -221.6410675048828,
+      "logps/rejected": -179.1236114501953,
+      "loss": 0.6475,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.7080546617507935,
+      "rewards/margins": 0.12341801822185516,
+      "rewards/rejected": 0.5846366286277771,
+      "step": 1690
+    },
+    {
+      "epoch": 0.5354119916538719,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.8590858913041775e-06,
+      "logits/chosen": -0.5011571645736694,
+      "logits/rejected": -0.34209686517715454,
+      "logps/chosen": -194.4629364013672,
+      "logps/rejected": -162.49404907226562,
+      "loss": 0.6226,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.6779108643531799,
+      "rewards/margins": 0.17823724448680878,
+      "rewards/rejected": 0.49967360496520996,
+      "step": 1700
+    },
+    {
+      "epoch": 0.5385614739577183,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.857335959425864e-06,
+      "logits/chosen": -0.47282689809799194,
+      "logits/rejected": -0.387722373008728,
+      "logps/chosen": -188.14022827148438,
+      "logps/rejected": -163.66702270507812,
+      "loss": 0.6431,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.6209579110145569,
+      "rewards/margins": 0.12763172388076782,
+      "rewards/rejected": 0.49332618713378906,
+      "step": 1710
+    },
+    {
+      "epoch": 0.5417109562615645,
+      "grad_norm": 3.125,
+      "learning_rate": 4.85557554785531e-06,
+      "logits/chosen": -0.5164914727210999,
+      "logits/rejected": -0.4268696904182434,
+      "logps/chosen": -194.6099090576172,
+      "logps/rejected": -171.3253631591797,
+      "loss": 0.6742,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.6653293371200562,
+      "rewards/margins": 0.08088545501232147,
+      "rewards/rejected": 0.5844438672065735,
+      "step": 1720
+    },
+    {
+      "epoch": 0.5448604385654108,
+      "grad_norm": 2.390625,
+      "learning_rate": 4.853804664418543e-06,
+      "logits/chosen": -0.5048812627792358,
+      "logits/rejected": -0.36949628591537476,
+      "logps/chosen": -186.7891845703125,
+      "logps/rejected": -177.64492797851562,
+      "loss": 0.6929,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.6204543113708496,
+      "rewards/margins": 0.03395826369524002,
+      "rewards/rejected": 0.5864960551261902,
+      "step": 1730
+    },
+    {
+      "epoch": 0.5480099208692571,
+      "grad_norm": 2.75,
+      "learning_rate": 4.85202331698814e-06,
+      "logits/chosen": -0.5082224011421204,
+      "logits/rejected": -0.3899040222167969,
+      "logps/chosen": -189.59127807617188,
+      "logps/rejected": -176.00732421875,
+      "loss": 0.6808,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.5799592733383179,
+      "rewards/margins": 0.04768219217658043,
+      "rewards/rejected": 0.5322771072387695,
+      "step": 1740
+    },
+    {
+      "epoch": 0.5511594031731034,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.8502315134832e-06,
+      "logits/chosen": -0.45675116777420044,
+      "logits/rejected": -0.35309940576553345,
+      "logps/chosen": -182.9220733642578,
+      "logps/rejected": -161.3640899658203,
+      "loss": 0.6708,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.605731189250946,
+      "rewards/margins": 0.07132869213819504,
+      "rewards/rejected": 0.5344024896621704,
+      "step": 1750
+    },
+    {
+      "epoch": 0.5543088854769497,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.848429261869303e-06,
+      "logits/chosen": -0.47061842679977417,
+      "logits/rejected": -0.32226455211639404,
+      "logps/chosen": -206.2819366455078,
+      "logps/rejected": -163.6657257080078,
+      "loss": 0.627,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.7078389525413513,
+      "rewards/margins": 0.1701158881187439,
+      "rewards/rejected": 0.5377230644226074,
+      "step": 1760
+    },
+    {
+      "epoch": 0.557458367780796,
+      "grad_norm": 2.5,
+      "learning_rate": 4.8466165701584766e-06,
+      "logits/chosen": -0.46921786665916443,
+      "logits/rejected": -0.33290085196495056,
+      "logps/chosen": -185.49661254882812,
+      "logps/rejected": -155.57052612304688,
+      "loss": 0.6234,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.6777461767196655,
+      "rewards/margins": 0.18197762966156006,
+      "rewards/rejected": 0.49576863646507263,
+      "step": 1770
+    },
+    {
+      "epoch": 0.5606078500846423,
+      "grad_norm": 3.53125,
+      "learning_rate": 4.844793446409162e-06,
+      "logits/chosen": -0.4977359175682068,
+      "logits/rejected": -0.3592739999294281,
+      "logps/chosen": -234.7196807861328,
+      "logps/rejected": -196.9207763671875,
+      "loss": 0.6667,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.725782573223114,
+      "rewards/margins": 0.08639247715473175,
+      "rewards/rejected": 0.6393901109695435,
+      "step": 1780
+    },
+    {
+      "epoch": 0.5637573323884887,
+      "grad_norm": 3.703125,
+      "learning_rate": 4.842959898726175e-06,
+      "logits/chosen": -0.42303353548049927,
+      "logits/rejected": -0.34682708978652954,
+      "logps/chosen": -233.24038696289062,
+      "logps/rejected": -207.0811309814453,
+      "loss": 0.6478,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.75257807970047,
+      "rewards/margins": 0.12484292685985565,
+      "rewards/rejected": 0.6277351975440979,
+      "step": 1790
+    },
+    {
+      "epoch": 0.566906814692335,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.8411159352606735e-06,
+      "logits/chosen": -0.49121102690696716,
+      "logits/rejected": -0.3855198621749878,
+      "logps/chosen": -201.4905548095703,
+      "logps/rejected": -184.22596740722656,
+      "loss": 0.6408,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.6799174547195435,
+      "rewards/margins": 0.1396731436252594,
+      "rewards/rejected": 0.5402444005012512,
+      "step": 1800
+    },
+    {
+      "epoch": 0.5700562969961812,
+      "grad_norm": 3.21875,
+      "learning_rate": 4.839261564210118e-06,
+      "logits/chosen": -0.4372677206993103,
+      "logits/rejected": -0.3174059987068176,
+      "logps/chosen": -182.1688995361328,
+      "logps/rejected": -167.50186157226562,
+      "loss": 0.6455,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.5665633678436279,
+      "rewards/margins": 0.12508925795555115,
+      "rewards/rejected": 0.441474050283432,
+      "step": 1810
+    },
+    {
+      "epoch": 0.5732057793000276,
+      "grad_norm": 2.453125,
+      "learning_rate": 4.837396793818237e-06,
+      "logits/chosen": -0.5189486742019653,
+      "logits/rejected": -0.4490106701850891,
+      "logps/chosen": -169.6758575439453,
+      "logps/rejected": -160.53372192382812,
+      "loss": 0.6999,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.5530275106430054,
+      "rewards/margins": 0.015836771577596664,
+      "rewards/rejected": 0.5371907353401184,
+      "step": 1820
+    },
+    {
+      "epoch": 0.5763552616038738,
+      "grad_norm": 2.625,
+      "learning_rate": 4.83552163237499e-06,
+      "logits/chosen": -0.44089236855506897,
+      "logits/rejected": -0.31842055916786194,
+      "logps/chosen": -189.9591064453125,
+      "logps/rejected": -159.47850036621094,
+      "loss": 0.6333,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.6711987257003784,
+      "rewards/margins": 0.16579605638980865,
+      "rewards/rejected": 0.505402684211731,
+      "step": 1830
+    },
+    {
+      "epoch": 0.5795047439077202,
+      "grad_norm": 2.390625,
+      "learning_rate": 4.8336360882165315e-06,
+      "logits/chosen": -0.4447326064109802,
+      "logits/rejected": -0.3511047661304474,
+      "logps/chosen": -189.64212036132812,
+      "logps/rejected": -163.2635955810547,
+      "loss": 0.6612,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.6063627600669861,
+      "rewards/margins": 0.10564740747213364,
+      "rewards/rejected": 0.5007153749465942,
+      "step": 1840
+    },
+    {
+      "epoch": 0.5826542262115665,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.831740169725172e-06,
+      "logits/chosen": -0.37459006905555725,
+      "logits/rejected": -0.27248674631118774,
+      "logps/chosen": -191.74412536621094,
+      "logps/rejected": -170.22665405273438,
+      "loss": 0.6765,
+      "rewards/accuracies": 0.512499988079071,
+      "rewards/chosen": 0.6292241811752319,
+      "rewards/margins": 0.06548759341239929,
+      "rewards/rejected": 0.5637365579605103,
+      "step": 1850
+    },
+    {
+      "epoch": 0.5858037085154127,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.829833885329341e-06,
+      "logits/chosen": -0.5164798498153687,
+      "logits/rejected": -0.3869698643684387,
+      "logps/chosen": -195.4289093017578,
+      "logps/rejected": -161.2754364013672,
+      "loss": 0.6763,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.5856376886367798,
+      "rewards/margins": 0.06769655644893646,
+      "rewards/rejected": 0.5179411172866821,
+      "step": 1860
+    },
+    {
+      "epoch": 0.5889531908192591,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.827917243503552e-06,
+      "logits/chosen": -0.4844232499599457,
+      "logits/rejected": -0.38613173365592957,
+      "logps/chosen": -212.82217407226562,
+      "logps/rejected": -174.99215698242188,
+      "loss": 0.6457,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.6664649844169617,
+      "rewards/margins": 0.13369064033031464,
+      "rewards/rejected": 0.5327743291854858,
+      "step": 1870
+    },
+    {
+      "epoch": 0.5921026731231054,
+      "grad_norm": 3.6875,
+      "learning_rate": 4.825990252768362e-06,
+      "logits/chosen": -0.4278056025505066,
+      "logits/rejected": -0.3540639579296112,
+      "logps/chosen": -191.68344116210938,
+      "logps/rejected": -180.2418670654297,
+      "loss": 0.6856,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.6736811399459839,
+      "rewards/margins": 0.05576536804437637,
+      "rewards/rejected": 0.6179158091545105,
+      "step": 1880
+    },
+    {
+      "epoch": 0.5952521554269516,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.824052921690337e-06,
+      "logits/chosen": -0.5202184915542603,
+      "logits/rejected": -0.38284236192703247,
+      "logps/chosen": -211.123046875,
+      "logps/rejected": -174.93304443359375,
+      "loss": 0.6467,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.7286884188652039,
+      "rewards/margins": 0.13865350186824799,
+      "rewards/rejected": 0.5900349020957947,
+      "step": 1890
+    },
+    {
+      "epoch": 0.598401637730798,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.822105258882007e-06,
+      "logits/chosen": -0.553580105304718,
+      "logits/rejected": -0.40661874413490295,
+      "logps/chosen": -231.7763214111328,
+      "logps/rejected": -190.30320739746094,
+      "loss": 0.6755,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.7072332501411438,
+      "rewards/margins": 0.06532811373472214,
+      "rewards/rejected": 0.6419050693511963,
+      "step": 1900
+    },
+    {
+      "epoch": 0.6015511200346443,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.8201472730018386e-06,
+      "logits/chosen": -0.4518907964229584,
+      "logits/rejected": -0.3795849680900574,
+      "logps/chosen": -205.5015411376953,
+      "logps/rejected": -180.77366638183594,
+      "loss": 0.6682,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.6349667310714722,
+      "rewards/margins": 0.07740394026041031,
+      "rewards/rejected": 0.5575627684593201,
+      "step": 1910
+    },
+    {
+      "epoch": 0.6047006023384907,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.818178972754184e-06,
+      "logits/chosen": -0.43323999643325806,
+      "logits/rejected": -0.4050220549106598,
+      "logps/chosen": -190.7349090576172,
+      "logps/rejected": -176.9745330810547,
+      "loss": 0.6589,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.6163308024406433,
+      "rewards/margins": 0.09825930744409561,
+      "rewards/rejected": 0.5180714726448059,
+      "step": 1920
+    },
+    {
+      "epoch": 0.6078500846423369,
+      "grad_norm": 2.390625,
+      "learning_rate": 4.816200366889252e-06,
+      "logits/chosen": -0.48036471009254456,
+      "logits/rejected": -0.34876304864883423,
+      "logps/chosen": -202.17532348632812,
+      "logps/rejected": -171.40567016601562,
+      "loss": 0.6655,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.6224693059921265,
+      "rewards/margins": 0.08968141674995422,
+      "rewards/rejected": 0.5327879190444946,
+      "step": 1930
+    },
+    {
+      "epoch": 0.6109995669461832,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.8142114642030665e-06,
+      "logits/chosen": -0.5331605672836304,
+      "logits/rejected": -0.35749322175979614,
+      "logps/chosen": -200.44296264648438,
+      "logps/rejected": -155.3129119873047,
+      "loss": 0.6277,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.686188817024231,
+      "rewards/margins": 0.17992666363716125,
+      "rewards/rejected": 0.5062621831893921,
+      "step": 1940
+    },
+    {
+      "epoch": 0.6141490492500296,
+      "grad_norm": 3.65625,
+      "learning_rate": 4.812212273537426e-06,
+      "logits/chosen": -0.5703829526901245,
+      "logits/rejected": -0.3986208736896515,
+      "logps/chosen": -207.4144287109375,
+      "logps/rejected": -166.04586791992188,
+      "loss": 0.635,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.6602171659469604,
+      "rewards/margins": 0.1550987958908081,
+      "rewards/rejected": 0.5051184296607971,
+      "step": 1950
+    },
+    {
+      "epoch": 0.6172985315538758,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.810202803779862e-06,
+      "logits/chosen": -0.49680987000465393,
+      "logits/rejected": -0.3343147039413452,
+      "logps/chosen": -194.55484008789062,
+      "logps/rejected": -171.62588500976562,
+      "loss": 0.6414,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6576918959617615,
+      "rewards/margins": 0.14642241597175598,
+      "rewards/rejected": 0.5112695693969727,
+      "step": 1960
+    },
+    {
+      "epoch": 0.6204480138577222,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.808183063863606e-06,
+      "logits/chosen": -0.5439049005508423,
+      "logits/rejected": -0.4092562794685364,
+      "logps/chosen": -223.9301300048828,
+      "logps/rejected": -191.12759399414062,
+      "loss": 0.6458,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.7495763897895813,
+      "rewards/margins": 0.13388705253601074,
+      "rewards/rejected": 0.6156893372535706,
+      "step": 1970
+    },
+    {
+      "epoch": 0.6235974961615685,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.806153062767544e-06,
+      "logits/chosen": -0.4681766629219055,
+      "logits/rejected": -0.3462735116481781,
+      "logps/chosen": -192.7504425048828,
+      "logps/rejected": -170.0450897216797,
+      "loss": 0.6479,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.6624137163162231,
+      "rewards/margins": 0.12909933924674988,
+      "rewards/rejected": 0.5333144068717957,
+      "step": 1980
+    },
+    {
+      "epoch": 0.6267469784654147,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.804112809516181e-06,
+      "logits/chosen": -0.5108489394187927,
+      "logits/rejected": -0.3038786053657532,
+      "logps/chosen": -207.41226196289062,
+      "logps/rejected": -163.39027404785156,
+      "loss": 0.6386,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.6961920261383057,
+      "rewards/margins": 0.14503511786460876,
+      "rewards/rejected": 0.5511568188667297,
+      "step": 1990
+    },
+    {
+      "epoch": 0.6298964607692611,
+      "grad_norm": 2.171875,
+      "learning_rate": 4.802062313179595e-06,
+      "logits/chosen": -0.4810725748538971,
+      "logits/rejected": -0.40924152731895447,
+      "logps/chosen": -187.09451293945312,
+      "logps/rejected": -162.91554260253906,
+      "loss": 0.661,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.5915666818618774,
+      "rewards/margins": 0.10626377910375595,
+      "rewards/rejected": 0.4853029251098633,
+      "step": 2000
+    },
+    {
+      "epoch": 0.6298964607692611,
+      "eval_logits/chosen": -0.6020957231521606,
+      "eval_logits/rejected": -0.4771311283111572,
+      "eval_logps/chosen": -243.46115112304688,
+      "eval_logps/rejected": -222.5963897705078,
+      "eval_loss": 0.6673439145088196,
+      "eval_rewards/accuracies": 0.5848915576934814,
+      "eval_rewards/chosen": 0.7640087604522705,
+      "eval_rewards/margins": 0.09086808562278748,
+      "eval_rewards/rejected": 0.6731407046318054,
+      "eval_runtime": 3657.8705,
+      "eval_samples_per_second": 0.366,
+      "eval_steps_per_second": 0.366,
+      "step": 2000
+    },
+    {
+      "epoch": 0.6330459430731074,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.800001582873405e-06,
+      "logits/chosen": -0.42106738686561584,
+      "logits/rejected": -0.35292476415634155,
+      "logps/chosen": -202.25643920898438,
+      "logps/rejected": -198.75399780273438,
+      "loss": 0.6953,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.6455667614936829,
+      "rewards/margins": 0.026390869170427322,
+      "rewards/rejected": 0.6191757917404175,
+      "step": 2010
+    },
+    {
+      "epoch": 0.6361954253769536,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.797930627758721e-06,
+      "logits/chosen": -0.3941357731819153,
+      "logits/rejected": -0.3840845823287964,
+      "logps/chosen": -193.46145629882812,
+      "logps/rejected": -182.74398803710938,
+      "loss": 0.6844,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.6705250144004822,
+      "rewards/margins": 0.047097403556108475,
+      "rewards/rejected": 0.6234275698661804,
+      "step": 2020
+    },
+    {
+      "epoch": 0.6393449076808,
+      "grad_norm": 2.1875,
+      "learning_rate": 4.795849457042112e-06,
+      "logits/chosen": -0.5017488598823547,
+      "logits/rejected": -0.40243926644325256,
+      "logps/chosen": -190.13934326171875,
+      "logps/rejected": -166.44187927246094,
+      "loss": 0.6766,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.6177536845207214,
+      "rewards/margins": 0.07091078162193298,
+      "rewards/rejected": 0.5468429327011108,
+      "step": 2030
+    },
+    {
+      "epoch": 0.6424943899846463,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.793758079975559e-06,
+      "logits/chosen": -0.47853583097457886,
+      "logits/rejected": -0.3796117901802063,
+      "logps/chosen": -183.77645874023438,
+      "logps/rejected": -150.51734924316406,
+      "loss": 0.6283,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.6325706243515015,
+      "rewards/margins": 0.16535897552967072,
+      "rewards/rejected": 0.46721166372299194,
+      "step": 2040
+    },
+    {
+      "epoch": 0.6456438722884926,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.791656505856416e-06,
+      "logits/chosen": -0.49306803941726685,
+      "logits/rejected": -0.37095504999160767,
+      "logps/chosen": -193.6649932861328,
+      "logps/rejected": -169.91397094726562,
+      "loss": 0.6596,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.6382758617401123,
+      "rewards/margins": 0.09585778415203094,
+      "rewards/rejected": 0.5424180030822754,
+      "step": 2050
+    },
+    {
+      "epoch": 0.6487933545923389,
+      "grad_norm": 4.375,
+      "learning_rate": 4.789544744027369e-06,
+      "logits/chosen": -0.456062376499176,
+      "logits/rejected": -0.326261430978775,
+      "logps/chosen": -202.9215087890625,
+      "logps/rejected": -158.33755493164062,
+      "loss": 0.6307,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6835106015205383,
+      "rewards/margins": 0.16993892192840576,
+      "rewards/rejected": 0.5135716199874878,
+      "step": 2060
+    },
+    {
+      "epoch": 0.6519428368961852,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.787422803876394e-06,
+      "logits/chosen": -0.4271954596042633,
+      "logits/rejected": -0.32812589406967163,
+      "logps/chosen": -217.31643676757812,
+      "logps/rejected": -189.05111694335938,
+      "loss": 0.654,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.7037121057510376,
+      "rewards/margins": 0.1129276379942894,
+      "rewards/rejected": 0.59078449010849,
+      "step": 2070
+    },
+    {
+      "epoch": 0.6550923192000315,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.785290694836719e-06,
+      "logits/chosen": -0.4718199670314789,
+      "logits/rejected": -0.29942384362220764,
+      "logps/chosen": -194.9972686767578,
+      "logps/rejected": -162.9912872314453,
+      "loss": 0.6402,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.7091890573501587,
+      "rewards/margins": 0.13689157366752625,
+      "rewards/rejected": 0.5722974538803101,
+      "step": 2080
+    },
+    {
+      "epoch": 0.6582418015038778,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.783148426386771e-06,
+      "logits/chosen": -0.4268282353878021,
+      "logits/rejected": -0.2675458788871765,
+      "logps/chosen": -199.43051147460938,
+      "logps/rejected": -162.5355987548828,
+      "loss": 0.6125,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.665716826915741,
+      "rewards/margins": 0.20410867035388947,
+      "rewards/rejected": 0.4616081714630127,
+      "step": 2090
+    },
+    {
+      "epoch": 0.661391283807724,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.7809960080501464e-06,
+      "logits/chosen": -0.4748079776763916,
+      "logits/rejected": -0.4310288429260254,
+      "logps/chosen": -201.29617309570312,
+      "logps/rejected": -184.75961303710938,
+      "loss": 0.6435,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.7009934186935425,
+      "rewards/margins": 0.13489434123039246,
+      "rewards/rejected": 0.5660991668701172,
+      "step": 2100
+    },
+    {
+      "epoch": 0.6645407661115704,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.778833449395563e-06,
+      "logits/chosen": -0.4732258915901184,
+      "logits/rejected": -0.3779997229576111,
+      "logps/chosen": -211.35055541992188,
+      "logps/rejected": -177.94529724121094,
+      "loss": 0.6271,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.6552049517631531,
+      "rewards/margins": 0.1688978374004364,
+      "rewards/rejected": 0.48630720376968384,
+      "step": 2110
+    },
+    {
+      "epoch": 0.6676902484154167,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.77666076003682e-06,
+      "logits/chosen": -0.5109846591949463,
+      "logits/rejected": -0.4116583466529846,
+      "logps/chosen": -201.0147705078125,
+      "logps/rejected": -173.7718048095703,
+      "loss": 0.6458,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.6840208172798157,
+      "rewards/margins": 0.1273798793554306,
+      "rewards/rejected": 0.5566409230232239,
+      "step": 2120
+    },
+    {
+      "epoch": 0.6708397307192631,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.774477949632747e-06,
+      "logits/chosen": -0.4809556007385254,
+      "logits/rejected": -0.37704578042030334,
+      "logps/chosen": -215.3496551513672,
+      "logps/rejected": -185.5915985107422,
+      "loss": 0.6738,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.6900166273117065,
+      "rewards/margins": 0.08503197878599167,
+      "rewards/rejected": 0.604984700679779,
+      "step": 2130
+    },
+    {
+      "epoch": 0.6739892130231093,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.772285027887174e-06,
+      "logits/chosen": -0.4865281581878662,
+      "logits/rejected": -0.3789558410644531,
+      "logps/chosen": -212.9844512939453,
+      "logps/rejected": -181.038818359375,
+      "loss": 0.6399,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6945775747299194,
+      "rewards/margins": 0.15178106725215912,
+      "rewards/rejected": 0.5427964329719543,
+      "step": 2140
+    },
+    {
+      "epoch": 0.6771386953269556,
+      "grad_norm": 2.28125,
+      "learning_rate": 4.770082004548878e-06,
+      "logits/chosen": -0.5607801675796509,
+      "logits/rejected": -0.4424077868461609,
+      "logps/chosen": -197.34896850585938,
+      "logps/rejected": -168.6945343017578,
+      "loss": 0.6788,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.6424630880355835,
+      "rewards/margins": 0.06348511576652527,
+      "rewards/rejected": 0.5789780020713806,
+      "step": 2150
+    },
+    {
+      "epoch": 0.680288177630802,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.767868889411545e-06,
+      "logits/chosen": -0.37069040536880493,
+      "logits/rejected": -0.29210925102233887,
+      "logps/chosen": -189.8524169921875,
+      "logps/rejected": -173.9405517578125,
+      "loss": 0.6838,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": 0.648335337638855,
+      "rewards/margins": 0.061441101133823395,
+      "rewards/rejected": 0.586894154548645,
+      "step": 2160
+    },
+    {
+      "epoch": 0.6834376599346482,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.765645692313724e-06,
+      "logits/chosen": -0.4717496931552887,
+      "logits/rejected": -0.38448435068130493,
+      "logps/chosen": -197.7617645263672,
+      "logps/rejected": -176.4130401611328,
+      "loss": 0.6377,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.6932880282402039,
+      "rewards/margins": 0.14973357319831848,
+      "rewards/rejected": 0.543554425239563,
+      "step": 2170
+    },
+    {
+      "epoch": 0.6865871422384946,
+      "grad_norm": 3.265625,
+      "learning_rate": 4.763412423138784e-06,
+      "logits/chosen": -0.46070584654808044,
+      "logits/rejected": -0.37532466650009155,
+      "logps/chosen": -211.4508514404297,
+      "logps/rejected": -192.87716674804688,
+      "loss": 0.6801,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.6881066560745239,
+      "rewards/margins": 0.05569648742675781,
+      "rewards/rejected": 0.6324101686477661,
+      "step": 2180
+    },
+    {
+      "epoch": 0.6897366245423409,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.761169091814869e-06,
+      "logits/chosen": -0.44056111574172974,
+      "logits/rejected": -0.3600524961948395,
+      "logps/chosen": -212.5589599609375,
+      "logps/rejected": -194.01473999023438,
+      "loss": 0.66,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.7078835368156433,
+      "rewards/margins": 0.10578237473964691,
+      "rewards/rejected": 0.60210120677948,
+      "step": 2190
+    },
+    {
+      "epoch": 0.6928861068461871,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.758915708314858e-06,
+      "logits/chosen": -0.528985857963562,
+      "logits/rejected": -0.3610820770263672,
+      "logps/chosen": -214.4971923828125,
+      "logps/rejected": -172.77456665039062,
+      "loss": 0.6279,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.7492777705192566,
+      "rewards/margins": 0.17471325397491455,
+      "rewards/rejected": 0.5745645761489868,
+      "step": 2200
+    },
+    {
+      "epoch": 0.6960355891500335,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.756652282656314e-06,
+      "logits/chosen": -0.46792277693748474,
+      "logits/rejected": -0.3924694061279297,
+      "logps/chosen": -189.278076171875,
+      "logps/rejected": -168.0936737060547,
+      "loss": 0.6298,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.6673378944396973,
+      "rewards/margins": 0.18005801737308502,
+      "rewards/rejected": 0.48727989196777344,
+      "step": 2210
+    },
+    {
+      "epoch": 0.6991850714538798,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.754378824901447e-06,
+      "logits/chosen": -0.42372870445251465,
+      "logits/rejected": -0.3045392334461212,
+      "logps/chosen": -217.94894409179688,
+      "logps/rejected": -181.64312744140625,
+      "loss": 0.6352,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.7333195209503174,
+      "rewards/margins": 0.15828455984592438,
+      "rewards/rejected": 0.5750349760055542,
+      "step": 2220
+    },
+    {
+      "epoch": 0.702334553757726,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.752095345157062e-06,
+      "logits/chosen": -0.5427747368812561,
+      "logits/rejected": -0.3744064271450043,
+      "logps/chosen": -196.52786254882812,
+      "logps/rejected": -179.39456176757812,
+      "loss": 0.6659,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.7010904550552368,
+      "rewards/margins": 0.10356839001178741,
+      "rewards/rejected": 0.5975220203399658,
+      "step": 2230
+    },
+    {
+      "epoch": 0.7054840360615724,
+      "grad_norm": 2.875,
+      "learning_rate": 4.7498018535745175e-06,
+      "logits/chosen": -0.45889267325401306,
+      "logits/rejected": -0.3741144835948944,
+      "logps/chosen": -194.2584228515625,
+      "logps/rejected": -169.21490478515625,
+      "loss": 0.6514,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.6648105382919312,
+      "rewards/margins": 0.12509003281593323,
+      "rewards/rejected": 0.5397205352783203,
+      "step": 2240
+    },
+    {
+      "epoch": 0.7086335183654187,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.747498360349681e-06,
+      "logits/chosen": -0.4748617708683014,
+      "logits/rejected": -0.3516277074813843,
+      "logps/chosen": -219.1868438720703,
+      "logps/rejected": -182.24143981933594,
+      "loss": 0.6369,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.7704683542251587,
+      "rewards/margins": 0.16334478557109833,
+      "rewards/rejected": 0.6071235537528992,
+      "step": 2250
+    },
+    {
+      "epoch": 0.711783000669265,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.745184875722887e-06,
+      "logits/chosen": -0.5461896657943726,
+      "logits/rejected": -0.4441626965999603,
+      "logps/chosen": -177.56777954101562,
+      "logps/rejected": -164.29269409179688,
+      "loss": 0.6576,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.63064044713974,
+      "rewards/margins": 0.11562252044677734,
+      "rewards/rejected": 0.5150178670883179,
+      "step": 2260
+    },
+    {
+      "epoch": 0.7149324829731113,
+      "grad_norm": 2.625,
+      "learning_rate": 4.7428614099788804e-06,
+      "logits/chosen": -0.4357683062553406,
+      "logits/rejected": -0.40783005952835083,
+      "logps/chosen": -180.0326690673828,
+      "logps/rejected": -161.7758331298828,
+      "loss": 0.6844,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.5830633640289307,
+      "rewards/margins": 0.04567595571279526,
+      "rewards/rejected": 0.5373873710632324,
+      "step": 2270
+    },
+    {
+      "epoch": 0.7180819652769576,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.740527973446782e-06,
+      "logits/chosen": -0.4672257900238037,
+      "logits/rejected": -0.38863813877105713,
+      "logps/chosen": -172.70326232910156,
+      "logps/rejected": -155.18206787109375,
+      "loss": 0.6836,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.5959118008613586,
+      "rewards/margins": 0.06164885684847832,
+      "rewards/rejected": 0.5342629551887512,
+      "step": 2280
+    },
+    {
+      "epoch": 0.7212314475808039,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.738184576500038e-06,
+      "logits/chosen": -0.4562186300754547,
+      "logits/rejected": -0.3595152199268341,
+      "logps/chosen": -202.3937530517578,
+      "logps/rejected": -184.3506622314453,
+      "loss": 0.6476,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.6859248280525208,
+      "rewards/margins": 0.13335470855236053,
+      "rewards/rejected": 0.5525700449943542,
+      "step": 2290
+    },
+    {
+      "epoch": 0.7243809298846502,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.735831229556374e-06,
+      "logits/chosen": -0.5141324996948242,
+      "logits/rejected": -0.4093483090400696,
+      "logps/chosen": -219.1161651611328,
+      "logps/rejected": -183.11634826660156,
+      "loss": 0.6545,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.7390071153640747,
+      "rewards/margins": 0.13933688402175903,
+      "rewards/rejected": 0.5996701717376709,
+      "step": 2300
+    },
+    {
+      "epoch": 0.7275304121884966,
+      "grad_norm": 3.75,
+      "learning_rate": 4.733467943077747e-06,
+      "logits/chosen": -0.43999338150024414,
+      "logits/rejected": -0.3529604375362396,
+      "logps/chosen": -179.51895141601562,
+      "logps/rejected": -177.3287811279297,
+      "loss": 0.7027,
+      "rewards/accuracies": 0.4749999940395355,
+      "rewards/chosen": 0.5721325278282166,
+      "rewards/margins": 0.017483506351709366,
+      "rewards/rejected": 0.5546489953994751,
+      "step": 2310
+    },
+    {
+      "epoch": 0.7306798944923428,
+      "grad_norm": 3.71875,
+      "learning_rate": 4.731094727570305e-06,
+      "logits/chosen": -0.39424124360084534,
+      "logits/rejected": -0.32468560338020325,
+      "logps/chosen": -201.04409790039062,
+      "logps/rejected": -186.39602661132812,
+      "loss": 0.6709,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.615720808506012,
+      "rewards/margins": 0.08357678353786469,
+      "rewards/rejected": 0.5321440100669861,
+      "step": 2320
+    },
+    {
+      "epoch": 0.7338293767961891,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.7287115935843335e-06,
+      "logits/chosen": -0.4321005940437317,
+      "logits/rejected": -0.3664671778678894,
+      "logps/chosen": -183.7452392578125,
+      "logps/rejected": -170.99435424804688,
+      "loss": 0.6807,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.6428938508033752,
+      "rewards/margins": 0.06662772595882416,
+      "rewards/rejected": 0.5762661695480347,
+      "step": 2330
+    },
+    {
+      "epoch": 0.7369788591000355,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.72631855171421e-06,
+      "logits/chosen": -0.407909095287323,
+      "logits/rejected": -0.3485127389431,
+      "logps/chosen": -189.66488647460938,
+      "logps/rejected": -168.92410278320312,
+      "loss": 0.6538,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6725813150405884,
+      "rewards/margins": 0.11997097730636597,
+      "rewards/rejected": 0.5526103377342224,
+      "step": 2340
+    },
+    {
+      "epoch": 0.7401283414038817,
+      "grad_norm": 3.765625,
+      "learning_rate": 4.72391561259836e-06,
+      "logits/chosen": -0.4830864369869232,
+      "logits/rejected": -0.39991340041160583,
+      "logps/chosen": -196.81080627441406,
+      "logps/rejected": -173.09869384765625,
+      "loss": 0.6644,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.6581705808639526,
+      "rewards/margins": 0.10404298454523087,
+      "rewards/rejected": 0.55412757396698,
+      "step": 2350
+    },
+    {
+      "epoch": 0.743277823707728,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.721502786919209e-06,
+      "logits/chosen": -0.5260214805603027,
+      "logits/rejected": -0.45939525961875916,
+      "logps/chosen": -191.50491333007812,
+      "logps/rejected": -162.9628143310547,
+      "loss": 0.6571,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6285362243652344,
+      "rewards/margins": 0.10894250869750977,
+      "rewards/rejected": 0.5195937156677246,
+      "step": 2360
+    },
+    {
+      "epoch": 0.7464273060115744,
+      "grad_norm": 2.390625,
+      "learning_rate": 4.719080085403131e-06,
+      "logits/chosen": -0.4764222204685211,
+      "logits/rejected": -0.37835368514060974,
+      "logps/chosen": -207.5322723388672,
+      "logps/rejected": -173.14564514160156,
+      "loss": 0.6403,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.7107085585594177,
+      "rewards/margins": 0.15078091621398926,
+      "rewards/rejected": 0.5599276423454285,
+      "step": 2370
+    },
+    {
+      "epoch": 0.7495767883154206,
+      "grad_norm": 3.703125,
+      "learning_rate": 4.716647518820406e-06,
+      "logits/chosen": -0.51595538854599,
+      "logits/rejected": -0.3755626380443573,
+      "logps/chosen": -191.32260131835938,
+      "logps/rejected": -156.11282348632812,
+      "loss": 0.676,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.6388931274414062,
+      "rewards/margins": 0.06408585608005524,
+      "rewards/rejected": 0.5748072862625122,
+      "step": 2380
+    },
+    {
+      "epoch": 0.752726270619267,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.714205097985169e-06,
+      "logits/chosen": -0.4605388045310974,
+      "logits/rejected": -0.31849172711372375,
+      "logps/chosen": -183.9641876220703,
+      "logps/rejected": -168.69387817382812,
+      "loss": 0.6416,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.7079667448997498,
+      "rewards/margins": 0.15685173869132996,
+      "rewards/rejected": 0.5511150360107422,
+      "step": 2390
+    },
+    {
+      "epoch": 0.7558757529231133,
+      "grad_norm": 3.5625,
+      "learning_rate": 4.711752833755362e-06,
+      "logits/chosen": -0.4923486113548279,
+      "logits/rejected": -0.36627882719039917,
+      "logps/chosen": -211.5367431640625,
+      "logps/rejected": -170.20559692382812,
+      "loss": 0.6441,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.7119361162185669,
+      "rewards/margins": 0.14048628509044647,
+      "rewards/rejected": 0.571449875831604,
+      "step": 2400
+    },
+    {
+      "epoch": 0.7590252352269595,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.7092907370326876e-06,
+      "logits/chosen": -0.4083434045314789,
+      "logits/rejected": -0.3242022693157196,
+      "logps/chosen": -195.62118530273438,
+      "logps/rejected": -170.5220489501953,
+      "loss": 0.6861,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.6857455968856812,
+      "rewards/margins": 0.0565875768661499,
+      "rewards/rejected": 0.6291579604148865,
+      "step": 2410
+    },
+    {
+      "epoch": 0.7621747175308059,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.706818818762558e-06,
+      "logits/chosen": -0.5046082735061646,
+      "logits/rejected": -0.35835257172584534,
+      "logps/chosen": -212.45217895507812,
+      "logps/rejected": -180.62155151367188,
+      "loss": 0.6265,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.7465167045593262,
+      "rewards/margins": 0.1813477873802185,
+      "rewards/rejected": 0.5651688575744629,
+      "step": 2420
+    },
+    {
+      "epoch": 0.7653241998346522,
+      "grad_norm": 2.453125,
+      "learning_rate": 4.7043370899340505e-06,
+      "logits/chosen": -0.4568845331668854,
+      "logits/rejected": -0.4346873164176941,
+      "logps/chosen": -202.90444946289062,
+      "logps/rejected": -197.64659118652344,
+      "loss": 0.6894,
+      "rewards/accuracies": 0.4749999940395355,
+      "rewards/chosen": 0.6897009015083313,
+      "rewards/margins": 0.04634975641965866,
+      "rewards/rejected": 0.6433511972427368,
+      "step": 2430
+    },
+    {
+      "epoch": 0.7684736821384985,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.701845561579853e-06,
+      "logits/chosen": -0.4123512804508209,
+      "logits/rejected": -0.3234286606311798,
+      "logps/chosen": -199.5214385986328,
+      "logps/rejected": -182.79100036621094,
+      "loss": 0.6436,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6898704767227173,
+      "rewards/margins": 0.13404139876365662,
+      "rewards/rejected": 0.5558291077613831,
+      "step": 2440
+    },
+    {
+      "epoch": 0.7716231644423448,
+      "grad_norm": 2.390625,
+      "learning_rate": 4.6993442447762185e-06,
+      "logits/chosen": -0.4709036350250244,
+      "logits/rejected": -0.3212384283542633,
+      "logps/chosen": -214.52877807617188,
+      "logps/rejected": -172.2939910888672,
+      "loss": 0.6349,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.7454115152359009,
+      "rewards/margins": 0.16362550854682922,
+      "rewards/rejected": 0.5817859768867493,
+      "step": 2450
+    },
+    {
+      "epoch": 0.7747726467461911,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.696833150642916e-06,
+      "logits/chosen": -0.4889756143093109,
+      "logits/rejected": -0.3762696385383606,
+      "logps/chosen": -200.27764892578125,
+      "logps/rejected": -167.56094360351562,
+      "loss": 0.646,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6864475011825562,
+      "rewards/margins": 0.13881604373455048,
+      "rewards/rejected": 0.5476315021514893,
+      "step": 2460
+    },
+    {
+      "epoch": 0.7779221290500374,
+      "grad_norm": 2.359375,
+      "learning_rate": 4.694312290343178e-06,
+      "logits/chosen": -0.5132132768630981,
+      "logits/rejected": -0.38872334361076355,
+      "logps/chosen": -200.6832733154297,
+      "logps/rejected": -171.80210876464844,
+      "loss": 0.6265,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.709212601184845,
+      "rewards/margins": 0.17675474286079407,
+      "rewards/rejected": 0.5324578881263733,
+      "step": 2470
+    },
+    {
+      "epoch": 0.7810716113538837,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.691781675083658e-06,
+      "logits/chosen": -0.543319582939148,
+      "logits/rejected": -0.4530103802680969,
+      "logps/chosen": -193.91519165039062,
+      "logps/rejected": -169.05667114257812,
+      "loss": 0.6525,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.6816273927688599,
+      "rewards/margins": 0.11835716664791107,
+      "rewards/rejected": 0.5632702112197876,
+      "step": 2480
+    },
+    {
+      "epoch": 0.78422109365773,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.689241316114373e-06,
+      "logits/chosen": -0.42033663392066956,
+      "logits/rejected": -0.35506805777549744,
+      "logps/chosen": -197.7043914794922,
+      "logps/rejected": -181.24746704101562,
+      "loss": 0.6528,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.761360764503479,
+      "rewards/margins": 0.1297047734260559,
+      "rewards/rejected": 0.6316559910774231,
+      "step": 2490
+    },
+    {
+      "epoch": 0.7873705759615763,
+      "grad_norm": 2.828125,
+      "learning_rate": 4.686691224728652e-06,
+      "logits/chosen": -0.5081063508987427,
+      "logits/rejected": -0.3558959364891052,
+      "logps/chosen": -215.3639373779297,
+      "logps/rejected": -175.36788940429688,
+      "loss": 0.6667,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.7302805781364441,
+      "rewards/margins": 0.10294453799724579,
+      "rewards/rejected": 0.6273361444473267,
+      "step": 2500
+    },
+    {
+      "epoch": 0.7905200582654226,
+      "grad_norm": 3.0,
+      "learning_rate": 4.684131412263098e-06,
+      "logits/chosen": -0.35183241963386536,
+      "logits/rejected": -0.3214506208896637,
+      "logps/chosen": -177.6040496826172,
+      "logps/rejected": -171.63784790039062,
+      "loss": 0.6771,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.6120871901512146,
+      "rewards/margins": 0.06991194188594818,
+      "rewards/rejected": 0.5421752333641052,
+      "step": 2510
+    },
+    {
+      "epoch": 0.793669540569269,
+      "grad_norm": 2.28125,
+      "learning_rate": 4.681561890097525e-06,
+      "logits/chosen": -0.4272761344909668,
+      "logits/rejected": -0.35165831446647644,
+      "logps/chosen": -195.2444305419922,
+      "logps/rejected": -184.37887573242188,
+      "loss": 0.6776,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.68397456407547,
+      "rewards/margins": 0.0739150270819664,
+      "rewards/rejected": 0.6100595593452454,
+      "step": 2520
+    },
+    {
+      "epoch": 0.7968190228731152,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.678982669654912e-06,
+      "logits/chosen": -0.423833429813385,
+      "logits/rejected": -0.37776055932044983,
+      "logps/chosen": -202.28225708007812,
+      "logps/rejected": -188.2001190185547,
+      "loss": 0.6442,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.6994912624359131,
+      "rewards/margins": 0.13222193717956543,
+      "rewards/rejected": 0.5672692656517029,
+      "step": 2530
+    },
+    {
+      "epoch": 0.7999685051769615,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.676393762401354e-06,
+      "logits/chosen": -0.4404567778110504,
+      "logits/rejected": -0.32071155309677124,
+      "logps/chosen": -197.72959899902344,
+      "logps/rejected": -161.87881469726562,
+      "loss": 0.6561,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.6411703824996948,
+      "rewards/margins": 0.11388187110424042,
+      "rewards/rejected": 0.527288556098938,
+      "step": 2540
+    },
+    {
+      "epoch": 0.8031179874808079,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.673795179846008e-06,
+      "logits/chosen": -0.4287866950035095,
+      "logits/rejected": -0.33532360196113586,
+      "logps/chosen": -207.0975341796875,
+      "logps/rejected": -175.67420959472656,
+      "loss": 0.6759,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.6876527667045593,
+      "rewards/margins": 0.06763847172260284,
+      "rewards/rejected": 0.6200142502784729,
+      "step": 2550
+    },
+    {
+      "epoch": 0.8062674697846541,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.671186933541044e-06,
+      "logits/chosen": -0.4286137521266937,
+      "logits/rejected": -0.37905916571617126,
+      "logps/chosen": -191.0441131591797,
+      "logps/rejected": -179.77825927734375,
+      "loss": 0.695,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.6264538168907166,
+      "rewards/margins": 0.020585432648658752,
+      "rewards/rejected": 0.6058684587478638,
+      "step": 2560
+    },
+    {
+      "epoch": 0.8094169520885005,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.668569035081594e-06,
+      "logits/chosen": -0.4553300440311432,
+      "logits/rejected": -0.4031829833984375,
+      "logps/chosen": -182.02127075195312,
+      "logps/rejected": -171.6331787109375,
+      "loss": 0.6973,
+      "rewards/accuracies": 0.4625000059604645,
+      "rewards/chosen": 0.6412879228591919,
+      "rewards/margins": 0.04937012866139412,
+      "rewards/rejected": 0.5919178128242493,
+      "step": 2570
+    },
+    {
+      "epoch": 0.8125664343923468,
+      "grad_norm": 2.828125,
+      "learning_rate": 4.665941496105697e-06,
+      "logits/chosen": -0.5085742473602295,
+      "logits/rejected": -0.3790132701396942,
+      "logps/chosen": -187.68478393554688,
+      "logps/rejected": -153.02972412109375,
+      "loss": 0.6575,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.6509996056556702,
+      "rewards/margins": 0.11632029712200165,
+      "rewards/rejected": 0.5346792936325073,
+      "step": 2580
+    },
+    {
+      "epoch": 0.815715916696193,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.663304328294251e-06,
+      "logits/chosen": -0.4928790032863617,
+      "logits/rejected": -0.31242626905441284,
+      "logps/chosen": -204.3889617919922,
+      "logps/rejected": -167.4778594970703,
+      "loss": 0.645,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.6917664408683777,
+      "rewards/margins": 0.1491299867630005,
+      "rewards/rejected": 0.5426364541053772,
+      "step": 2590
+    },
+    {
+      "epoch": 0.8188653990000394,
+      "grad_norm": 3.609375,
+      "learning_rate": 4.660657543370958e-06,
+      "logits/chosen": -0.4446497857570648,
+      "logits/rejected": -0.35457903146743774,
+      "logps/chosen": -181.77471923828125,
+      "logps/rejected": -157.31527709960938,
+      "loss": 0.6608,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.6617356538772583,
+      "rewards/margins": 0.09582933783531189,
+      "rewards/rejected": 0.5659063458442688,
+      "step": 2600
+    },
+    {
+      "epoch": 0.8220148813038857,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.658001153102276e-06,
+      "logits/chosen": -0.44886890053749084,
+      "logits/rejected": -0.43147921562194824,
+      "logps/chosen": -186.99864196777344,
+      "logps/rejected": -188.51278686523438,
+      "loss": 0.6707,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.6480501890182495,
+      "rewards/margins": 0.08241890370845795,
+      "rewards/rejected": 0.5656312108039856,
+      "step": 2610
+    },
+    {
+      "epoch": 0.8251643636077319,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.655335169297363e-06,
+      "logits/chosen": -0.4991195797920227,
+      "logits/rejected": -0.3592751622200012,
+      "logps/chosen": -185.76898193359375,
+      "logps/rejected": -163.68087768554688,
+      "loss": 0.6537,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6216106414794922,
+      "rewards/margins": 0.11423293501138687,
+      "rewards/rejected": 0.5073777437210083,
+      "step": 2620
+    },
+    {
+      "epoch": 0.8283138459115783,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.652659603808024e-06,
+      "logits/chosen": -0.46877604722976685,
+      "logits/rejected": -0.3875640332698822,
+      "logps/chosen": -194.6293182373047,
+      "logps/rejected": -162.6995391845703,
+      "loss": 0.6556,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.6514769792556763,
+      "rewards/margins": 0.10939432680606842,
+      "rewards/rejected": 0.542082667350769,
+      "step": 2630
+    },
+    {
+      "epoch": 0.8314633282154246,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.6499744685286626e-06,
+      "logits/chosen": -0.4762224555015564,
+      "logits/rejected": -0.3948310911655426,
+      "logps/chosen": -176.2687530517578,
+      "logps/rejected": -162.93235778808594,
+      "loss": 0.678,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.6167873740196228,
+      "rewards/margins": 0.05972397327423096,
+      "rewards/rejected": 0.5570634007453918,
+      "step": 2640
+    },
+    {
+      "epoch": 0.8346128105192709,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.6472797753962255e-06,
+      "logits/chosen": -0.46065980195999146,
+      "logits/rejected": -0.4272928237915039,
+      "logps/chosen": -203.83956909179688,
+      "logps/rejected": -183.5655059814453,
+      "loss": 0.6521,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.7077460289001465,
+      "rewards/margins": 0.13481923937797546,
+      "rewards/rejected": 0.5729268193244934,
+      "step": 2650
+    },
+    {
+      "epoch": 0.8377622928231172,
+      "grad_norm": 2.046875,
+      "learning_rate": 4.6445755363901465e-06,
+      "logits/chosen": -0.4757087826728821,
+      "logits/rejected": -0.38740482926368713,
+      "logps/chosen": -201.98423767089844,
+      "logps/rejected": -168.28964233398438,
+      "loss": 0.662,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.6773720979690552,
+      "rewards/margins": 0.10439164936542511,
+      "rewards/rejected": 0.5729804039001465,
+      "step": 2660
+    },
+    {
+      "epoch": 0.8409117751269635,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.641861763532299e-06,
+      "logits/chosen": -0.4650692939758301,
+      "logits/rejected": -0.3750719130039215,
+      "logps/chosen": -202.7363739013672,
+      "logps/rejected": -168.4525909423828,
+      "loss": 0.6652,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.6195183992385864,
+      "rewards/margins": 0.08996371924877167,
+      "rewards/rejected": 0.529554545879364,
+      "step": 2670
+    },
+    {
+      "epoch": 0.8440612574308098,
+      "grad_norm": 3.0,
+      "learning_rate": 4.639138468886939e-06,
+      "logits/chosen": -0.49985313415527344,
+      "logits/rejected": -0.4105320870876312,
+      "logps/chosen": -200.1018524169922,
+      "logps/rejected": -182.53408813476562,
+      "loss": 0.6707,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.68363356590271,
+      "rewards/margins": 0.08553630113601685,
+      "rewards/rejected": 0.5980972051620483,
+      "step": 2680
+    },
+    {
+      "epoch": 0.8472107397346561,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.636405664560652e-06,
+      "logits/chosen": -0.5170512199401855,
+      "logits/rejected": -0.4552704691886902,
+      "logps/chosen": -205.1244354248047,
+      "logps/rejected": -176.17617797851562,
+      "loss": 0.6613,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.7613251805305481,
+      "rewards/margins": 0.1176910400390625,
+      "rewards/rejected": 0.6436341404914856,
+      "step": 2690
+    },
+    {
+      "epoch": 0.8503602220385024,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.6336633627023e-06,
+      "logits/chosen": -0.5023754835128784,
+      "logits/rejected": -0.3591773808002472,
+      "logps/chosen": -198.65286254882812,
+      "logps/rejected": -173.5179443359375,
+      "loss": 0.6609,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.6720755696296692,
+      "rewards/margins": 0.10540957748889923,
+      "rewards/rejected": 0.5666660070419312,
+      "step": 2700
+    },
+    {
+      "epoch": 0.8535097043423487,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.630911575502967e-06,
+      "logits/chosen": -0.4533337652683258,
+      "logits/rejected": -0.3400491774082184,
+      "logps/chosen": -190.54476928710938,
+      "logps/rejected": -167.88975524902344,
+      "loss": 0.6551,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.662422776222229,
+      "rewards/margins": 0.11403118073940277,
+      "rewards/rejected": 0.5483915209770203,
+      "step": 2710
+    },
+    {
+      "epoch": 0.856659186646195,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.628150315195902e-06,
+      "logits/chosen": -0.44884181022644043,
+      "logits/rejected": -0.3520192801952362,
+      "logps/chosen": -192.61509704589844,
+      "logps/rejected": -169.34854125976562,
+      "loss": 0.6865,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.6516313552856445,
+      "rewards/margins": 0.05064528435468674,
+      "rewards/rejected": 0.6009860634803772,
+      "step": 2720
+    },
+    {
+      "epoch": 0.8598086689500414,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.625379594056472e-06,
+      "logits/chosen": -0.43172770738601685,
+      "logits/rejected": -0.3771997094154358,
+      "logps/chosen": -181.5345916748047,
+      "logps/rejected": -166.65841674804688,
+      "loss": 0.6848,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.6033766865730286,
+      "rewards/margins": 0.052743762731552124,
+      "rewards/rejected": 0.5506329536437988,
+      "step": 2730
+    },
+    {
+      "epoch": 0.8629581512538876,
+      "grad_norm": 2.25,
+      "learning_rate": 4.6225994244020984e-06,
+      "logits/chosen": -0.4917459487915039,
+      "logits/rejected": -0.37200021743774414,
+      "logps/chosen": -209.8813018798828,
+      "logps/rejected": -180.4649658203125,
+      "loss": 0.6435,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.7169169187545776,
+      "rewards/margins": 0.15931591391563416,
+      "rewards/rejected": 0.5576010346412659,
+      "step": 2740
+    },
+    {
+      "epoch": 0.8661076335577339,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.61980981859221e-06,
+      "logits/chosen": -0.5261596441268921,
+      "logits/rejected": -0.4045354425907135,
+      "logps/chosen": -197.52328491210938,
+      "logps/rejected": -175.49851989746094,
+      "loss": 0.6605,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.6951828598976135,
+      "rewards/margins": 0.11784076690673828,
+      "rewards/rejected": 0.5773420929908752,
+      "step": 2750
+    },
+    {
+      "epoch": 0.8692571158615803,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.6170107890281826e-06,
+      "logits/chosen": -0.4892495274543762,
+      "logits/rejected": -0.3943815231323242,
+      "logps/chosen": -214.80078125,
+      "logps/rejected": -192.01844787597656,
+      "loss": 0.6471,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.7895753979682922,
+      "rewards/margins": 0.12512117624282837,
+      "rewards/rejected": 0.6644541025161743,
+      "step": 2760
+    },
+    {
+      "epoch": 0.8724065981654265,
+      "grad_norm": 3.5625,
+      "learning_rate": 4.614202348153285e-06,
+      "logits/chosen": -0.506980299949646,
+      "logits/rejected": -0.37280508875846863,
+      "logps/chosen": -206.6781005859375,
+      "logps/rejected": -187.54193115234375,
+      "loss": 0.6726,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.6904038190841675,
+      "rewards/margins": 0.0823880210518837,
+      "rewards/rejected": 0.6080158352851868,
+      "step": 2770
+    },
+    {
+      "epoch": 0.8755560804692729,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.611384508452629e-06,
+      "logits/chosen": -0.47033706307411194,
+      "logits/rejected": -0.3309301435947418,
+      "logps/chosen": -221.22421264648438,
+      "logps/rejected": -190.77188110351562,
+      "loss": 0.6516,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.759658932685852,
+      "rewards/margins": 0.12068722397089005,
+      "rewards/rejected": 0.638971745967865,
+      "step": 2780
+    },
+    {
+      "epoch": 0.8787055627731192,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.608557282453104e-06,
+      "logits/chosen": -0.5095947980880737,
+      "logits/rejected": -0.45932531356811523,
+      "logps/chosen": -189.14663696289062,
+      "logps/rejected": -171.67837524414062,
+      "loss": 0.6528,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.6966843008995056,
+      "rewards/margins": 0.13209667801856995,
+      "rewards/rejected": 0.5645877122879028,
+      "step": 2790
+    },
+    {
+      "epoch": 0.8818550450769654,
+      "grad_norm": 2.390625,
+      "learning_rate": 4.605720682723331e-06,
+      "logits/chosen": -0.46834999322891235,
+      "logits/rejected": -0.3549017608165741,
+      "logps/chosen": -210.1326446533203,
+      "logps/rejected": -179.3318634033203,
+      "loss": 0.6451,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.761681318283081,
+      "rewards/margins": 0.15264348685741425,
+      "rewards/rejected": 0.6090378761291504,
+      "step": 2800
+    },
+    {
+      "epoch": 0.8850045273808118,
+      "grad_norm": 2.140625,
+      "learning_rate": 4.602874721873599e-06,
+      "logits/chosen": -0.5375205874443054,
+      "logits/rejected": -0.34866297245025635,
+      "logps/chosen": -188.57321166992188,
+      "logps/rejected": -155.67984008789062,
+      "loss": 0.6321,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.6381895542144775,
+      "rewards/margins": 0.16713955998420715,
+      "rewards/rejected": 0.47105008363723755,
+      "step": 2810
+    },
+    {
+      "epoch": 0.8881540096846581,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.600019412555816e-06,
+      "logits/chosen": -0.4357992112636566,
+      "logits/rejected": -0.28150540590286255,
+      "logps/chosen": -199.49620056152344,
+      "logps/rejected": -153.14419555664062,
+      "loss": 0.6476,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.6509748101234436,
+      "rewards/margins": 0.14667873084545135,
+      "rewards/rejected": 0.5042960047721863,
+      "step": 2820
+    },
+    {
+      "epoch": 0.8913034919885043,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.597154767463448e-06,
+      "logits/chosen": -0.4881827235221863,
+      "logits/rejected": -0.3464208245277405,
+      "logps/chosen": -201.01864624023438,
+      "logps/rejected": -183.83456420898438,
+      "loss": 0.6627,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.6923896074295044,
+      "rewards/margins": 0.09767551720142365,
+      "rewards/rejected": 0.5947140455245972,
+      "step": 2830
+    },
+    {
+      "epoch": 0.8944529742923507,
+      "grad_norm": 2.203125,
+      "learning_rate": 4.594280799331461e-06,
+      "logits/chosen": -0.516327977180481,
+      "logits/rejected": -0.4000956416130066,
+      "logps/chosen": -177.13763427734375,
+      "logps/rejected": -154.5478515625,
+      "loss": 0.6463,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.6087929010391235,
+      "rewards/margins": 0.1329253911972046,
+      "rewards/rejected": 0.47586750984191895,
+      "step": 2840
+    },
+    {
+      "epoch": 0.897602456596197,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.591397520936271e-06,
+      "logits/chosen": -0.4523008465766907,
+      "logits/rejected": -0.32660526037216187,
+      "logps/chosen": -190.58114624023438,
+      "logps/rejected": -168.5622100830078,
+      "loss": 0.6439,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.6767908930778503,
+      "rewards/margins": 0.1442452073097229,
+      "rewards/rejected": 0.5325456857681274,
+      "step": 2850
+    },
+    {
+      "epoch": 0.9007519389000433,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.588504945095684e-06,
+      "logits/chosen": -0.4655417799949646,
+      "logits/rejected": -0.3397763669490814,
+      "logps/chosen": -184.3746337890625,
+      "logps/rejected": -158.1368408203125,
+      "loss": 0.6466,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6702816486358643,
+      "rewards/margins": 0.1449626088142395,
+      "rewards/rejected": 0.5253190398216248,
+      "step": 2860
+    },
+    {
+      "epoch": 0.9039014212038896,
+      "grad_norm": 2.75,
+      "learning_rate": 4.585603084668833e-06,
+      "logits/chosen": -0.48204731941223145,
+      "logits/rejected": -0.3869457542896271,
+      "logps/chosen": -211.1409149169922,
+      "logps/rejected": -177.21664428710938,
+      "loss": 0.6278,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.693871796131134,
+      "rewards/margins": 0.17502956092357635,
+      "rewards/rejected": 0.5188421607017517,
+      "step": 2870
+    },
+    {
+      "epoch": 0.9070509035077359,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.582691952556131e-06,
+      "logits/chosen": -0.516753077507019,
+      "logits/rejected": -0.3798757791519165,
+      "logps/chosen": -185.08062744140625,
+      "logps/rejected": -157.95521545410156,
+      "loss": 0.6113,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.6912366151809692,
+      "rewards/margins": 0.21181932091712952,
+      "rewards/rejected": 0.4794173240661621,
+      "step": 2880
+    },
+    {
+      "epoch": 0.9102003858115822,
+      "grad_norm": 2.375,
+      "learning_rate": 4.579771561699208e-06,
+      "logits/chosen": -0.5301073789596558,
+      "logits/rejected": -0.37922126054763794,
+      "logps/chosen": -200.29745483398438,
+      "logps/rejected": -163.3575439453125,
+      "loss": 0.6527,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.655502200126648,
+      "rewards/margins": 0.1373116374015808,
+      "rewards/rejected": 0.5181905627250671,
+      "step": 2890
+    },
+    {
+      "epoch": 0.9133498681154285,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.576841925080853e-06,
+      "logits/chosen": -0.3888477683067322,
+      "logits/rejected": -0.32840999960899353,
+      "logps/chosen": -185.23165893554688,
+      "logps/rejected": -163.85369873046875,
+      "loss": 0.6592,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.6471211314201355,
+      "rewards/margins": 0.11747239530086517,
+      "rewards/rejected": 0.5296487808227539,
+      "step": 2900
+    },
+    {
+      "epoch": 0.9164993504192749,
+      "grad_norm": 2.234375,
+      "learning_rate": 4.5739030557249595e-06,
+      "logits/chosen": -0.5069789886474609,
+      "logits/rejected": -0.3604618310928345,
+      "logps/chosen": -196.76856994628906,
+      "logps/rejected": -165.517822265625,
+      "loss": 0.6391,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.6826483607292175,
+      "rewards/margins": 0.1527707278728485,
+      "rewards/rejected": 0.5298776030540466,
+      "step": 2910
+    },
+    {
+      "epoch": 0.9196488327231211,
+      "grad_norm": 2.375,
+      "learning_rate": 4.570954966696464e-06,
+      "logits/chosen": -0.43340712785720825,
+      "logits/rejected": -0.3375625014305115,
+      "logps/chosen": -195.20223999023438,
+      "logps/rejected": -172.97694396972656,
+      "loss": 0.6641,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.6797399520874023,
+      "rewards/margins": 0.09967435896396637,
+      "rewards/rejected": 0.5800655484199524,
+      "step": 2920
+    },
+    {
+      "epoch": 0.9227983150269674,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.56799767110129e-06,
+      "logits/chosen": -0.4303937554359436,
+      "logits/rejected": -0.38990846276283264,
+      "logps/chosen": -192.70545959472656,
+      "logps/rejected": -176.0408172607422,
+      "loss": 0.6566,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6560736894607544,
+      "rewards/margins": 0.10354320704936981,
+      "rewards/rejected": 0.5525304079055786,
+      "step": 2930
+    },
+    {
+      "epoch": 0.9259477973308138,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.565031182086291e-06,
+      "logits/chosen": -0.50087571144104,
+      "logits/rejected": -0.3773635923862457,
+      "logps/chosen": -194.03750610351562,
+      "logps/rejected": -171.54971313476562,
+      "loss": 0.6346,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.7052168846130371,
+      "rewards/margins": 0.16069751977920532,
+      "rewards/rejected": 0.544519305229187,
+      "step": 2940
+    },
+    {
+      "epoch": 0.92909727963466,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.562055512839189e-06,
+      "logits/chosen": -0.46178698539733887,
+      "logits/rejected": -0.37201589345932007,
+      "logps/chosen": -205.4782257080078,
+      "logps/rejected": -171.72808837890625,
+      "loss": 0.6217,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.7175599336624146,
+      "rewards/margins": 0.19239577651023865,
+      "rewards/rejected": 0.5251641273498535,
+      "step": 2950
+    },
+    {
+      "epoch": 0.9322467619385063,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.559070676588516e-06,
+      "logits/chosen": -0.4568074643611908,
+      "logits/rejected": -0.37853866815567017,
+      "logps/chosen": -214.366455078125,
+      "logps/rejected": -184.67774963378906,
+      "loss": 0.6514,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.7769818305969238,
+      "rewards/margins": 0.1274883896112442,
+      "rewards/rejected": 0.6494934558868408,
+      "step": 2960
+    },
+    {
+      "epoch": 0.9353962442423527,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.55607668660356e-06,
+      "logits/chosen": -0.48704949021339417,
+      "logits/rejected": -0.4139239192008972,
+      "logps/chosen": -193.26736450195312,
+      "logps/rejected": -175.77593994140625,
+      "loss": 0.6544,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.6955569982528687,
+      "rewards/margins": 0.11222921311855316,
+      "rewards/rejected": 0.5833277702331543,
+      "step": 2970
+    },
+    {
+      "epoch": 0.9385457265461989,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.5530735561943e-06,
+      "logits/chosen": -0.4977632462978363,
+      "logits/rejected": -0.3273804187774658,
+      "logps/chosen": -193.0499725341797,
+      "logps/rejected": -158.284423828125,
+      "loss": 0.6462,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.6608245968818665,
+      "rewards/margins": 0.13379625976085663,
+      "rewards/rejected": 0.5270283222198486,
+      "step": 2980
+    },
+    {
+      "epoch": 0.9416952088500453,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.55006129871135e-06,
+      "logits/chosen": -0.5545817017555237,
+      "logits/rejected": -0.43782949447631836,
+      "logps/chosen": -191.93934631347656,
+      "logps/rejected": -167.23953247070312,
+      "loss": 0.6625,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.6882120370864868,
+      "rewards/margins": 0.1209985762834549,
+      "rewards/rejected": 0.5672134160995483,
+      "step": 2990
+    },
+    {
+      "epoch": 0.9448446911538916,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.547039927545899e-06,
+      "logits/chosen": -0.4707905650138855,
+      "logits/rejected": -0.3295817971229553,
+      "logps/chosen": -186.739501953125,
+      "logps/rejected": -162.34422302246094,
+      "loss": 0.6588,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.579537034034729,
+      "rewards/margins": 0.0972566232085228,
+      "rewards/rejected": 0.48228034377098083,
+      "step": 3000
+    },
+    {
+      "epoch": 0.9479941734577378,
+      "grad_norm": 2.578125,
+      "learning_rate": 4.544009456129651e-06,
+      "logits/chosen": -0.40460291504859924,
+      "logits/rejected": -0.36005955934524536,
+      "logps/chosen": -191.5205841064453,
+      "logps/rejected": -164.66702270507812,
+      "loss": 0.6927,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.615665078163147,
+      "rewards/margins": 0.033144764602184296,
+      "rewards/rejected": 0.5825203061103821,
+      "step": 3010
+    },
+    {
+      "epoch": 0.9511436557615842,
+      "grad_norm": 3.125,
+      "learning_rate": 4.540969897934767e-06,
+      "logits/chosen": -0.4697691798210144,
+      "logits/rejected": -0.3581236004829407,
+      "logps/chosen": -205.1107940673828,
+      "logps/rejected": -161.09967041015625,
+      "loss": 0.6513,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.6768245697021484,
+      "rewards/margins": 0.13728924095630646,
+      "rewards/rejected": 0.5395353436470032,
+      "step": 3020
+    },
+    {
+      "epoch": 0.9542931380654305,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.537921266473802e-06,
+      "logits/chosen": -0.4509238302707672,
+      "logits/rejected": -0.41182202100753784,
+      "logps/chosen": -198.67684936523438,
+      "logps/rejected": -188.21649169921875,
+      "loss": 0.7044,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.7039377093315125,
+      "rewards/margins": 0.03385835513472557,
+      "rewards/rejected": 0.6700793504714966,
+      "step": 3030
+    },
+    {
+      "epoch": 0.9574426203692769,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.53486357529965e-06,
+      "logits/chosen": -0.48828214406967163,
+      "logits/rejected": -0.3500337302684784,
+      "logps/chosen": -187.39865112304688,
+      "logps/rejected": -156.00173950195312,
+      "loss": 0.64,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.6695148944854736,
+      "rewards/margins": 0.1509798765182495,
+      "rewards/rejected": 0.5185350179672241,
+      "step": 3040
+    },
+    {
+      "epoch": 0.9605921026731231,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.531796838005477e-06,
+      "logits/chosen": -0.5069050192832947,
+      "logits/rejected": -0.3991912305355072,
+      "logps/chosen": -197.77066040039062,
+      "logps/rejected": -163.296875,
+      "loss": 0.6187,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.7524086236953735,
+      "rewards/margins": 0.1966829001903534,
+      "rewards/rejected": 0.5557257533073425,
+      "step": 3050
+    },
+    {
+      "epoch": 0.9637415849769694,
+      "grad_norm": 2.5,
+      "learning_rate": 4.5287210682246655e-06,
+      "logits/chosen": -0.42384999990463257,
+      "logits/rejected": -0.3603217601776123,
+      "logps/chosen": -183.27471923828125,
+      "logps/rejected": -170.30142211914062,
+      "loss": 0.6832,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.5444117784500122,
+      "rewards/margins": 0.05068939924240112,
+      "rewards/rejected": 0.49372243881225586,
+      "step": 3060
+    },
+    {
+      "epoch": 0.9668910672808158,
+      "grad_norm": 6.25,
+      "learning_rate": 4.525636279630752e-06,
+      "logits/chosen": -0.4649208188056946,
+      "logits/rejected": -0.3279884159564972,
+      "logps/chosen": -204.52191162109375,
+      "logps/rejected": -171.454833984375,
+      "loss": 0.6164,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.7678558826446533,
+      "rewards/margins": 0.21700289845466614,
+      "rewards/rejected": 0.5508529543876648,
+      "step": 3070
+    },
+    {
+      "epoch": 0.970040549584662,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.522542485937369e-06,
+      "logits/chosen": -0.48648127913475037,
+      "logits/rejected": -0.3391122817993164,
+      "logps/chosen": -191.50942993164062,
+      "logps/rejected": -159.47317504882812,
+      "loss": 0.6609,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.6665128469467163,
+      "rewards/margins": 0.10518161207437515,
+      "rewards/rejected": 0.5613312721252441,
+      "step": 3080
+    },
+    {
+      "epoch": 0.9731900318885083,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.519439700898179e-06,
+      "logits/chosen": -0.5188948512077332,
+      "logits/rejected": -0.4177464544773102,
+      "logps/chosen": -205.6922607421875,
+      "logps/rejected": -177.05718994140625,
+      "loss": 0.6589,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.7047790288925171,
+      "rewards/margins": 0.11938399076461792,
+      "rewards/rejected": 0.5853949785232544,
+      "step": 3090
+    },
+    {
+      "epoch": 0.9763395141923547,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.516327938306818e-06,
+      "logits/chosen": -0.49799761176109314,
+      "logits/rejected": -0.3481315076351166,
+      "logps/chosen": -192.59019470214844,
+      "logps/rejected": -155.11587524414062,
+      "loss": 0.6338,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.6473811268806458,
+      "rewards/margins": 0.15366099774837494,
+      "rewards/rejected": 0.4937201142311096,
+      "step": 3100
+    },
+    {
+      "epoch": 0.9794889964962009,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.513207211996831e-06,
+      "logits/chosen": -0.43314170837402344,
+      "logits/rejected": -0.3531089425086975,
+      "logps/chosen": -195.55174255371094,
+      "logps/rejected": -162.95797729492188,
+      "loss": 0.6499,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.6635746955871582,
+      "rewards/margins": 0.1176295131444931,
+      "rewards/rejected": 0.5459452271461487,
+      "step": 3110
+    },
+    {
+      "epoch": 0.9826384788000473,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.510077535841612e-06,
+      "logits/chosen": -0.4261040687561035,
+      "logits/rejected": -0.3906673192977905,
+      "logps/chosen": -192.201416015625,
+      "logps/rejected": -177.6776123046875,
+      "loss": 0.6474,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.7501270771026611,
+      "rewards/margins": 0.1394282579421997,
+      "rewards/rejected": 0.6106988191604614,
+      "step": 3120
+    },
+    {
+      "epoch": 0.9857879611038936,
+      "grad_norm": 4.0625,
+      "learning_rate": 4.506938923754342e-06,
+      "logits/chosen": -0.47423991560935974,
+      "logits/rejected": -0.36756032705307007,
+      "logps/chosen": -209.3015594482422,
+      "logps/rejected": -172.35110473632812,
+      "loss": 0.6551,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.6946024894714355,
+      "rewards/margins": 0.12966330349445343,
+      "rewards/rejected": 0.5649392008781433,
+      "step": 3130
+    },
+    {
+      "epoch": 0.9889374434077398,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.50379138968793e-06,
+      "logits/chosen": -0.4364834725856781,
+      "logits/rejected": -0.28871777653694153,
+      "logps/chosen": -206.86984252929688,
+      "logps/rejected": -161.70379638671875,
+      "loss": 0.6198,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.7661986351013184,
+      "rewards/margins": 0.20797491073608398,
+      "rewards/rejected": 0.5582237839698792,
+      "step": 3140
+    },
+    {
+      "epoch": 0.9920869257115862,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.500634947634943e-06,
+      "logits/chosen": -0.4126051962375641,
+      "logits/rejected": -0.3056962192058563,
+      "logps/chosen": -204.4892578125,
+      "logps/rejected": -179.78683471679688,
+      "loss": 0.6801,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.730519711971283,
+      "rewards/margins": 0.09220778197050095,
+      "rewards/rejected": 0.6383119821548462,
+      "step": 3150
+    },
+    {
+      "epoch": 0.9952364080154324,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.497469611627554e-06,
+      "logits/chosen": -0.4934718608856201,
+      "logits/rejected": -0.3333396911621094,
+      "logps/chosen": -200.66720581054688,
+      "logps/rejected": -161.30099487304688,
+      "loss": 0.6013,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.7085192203521729,
+      "rewards/margins": 0.23979604244232178,
+      "rewards/rejected": 0.46872320771217346,
+      "step": 3160
+    },
+    {
+      "epoch": 0.9983858903192788,
+      "grad_norm": 3.453125,
+      "learning_rate": 4.4942953957374724e-06,
+      "logits/chosen": -0.5716500878334045,
+      "logits/rejected": -0.4639313220977783,
+      "logps/chosen": -209.87631225585938,
+      "logps/rejected": -187.9324493408203,
+      "loss": 0.6631,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.7570088505744934,
+      "rewards/margins": 0.10784139484167099,
+      "rewards/rejected": 0.6491674184799194,
+      "step": 3170
+    },
+    {
+      "epoch": 1.001535372623125,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.491112314075883e-06,
+      "logits/chosen": -0.4782622456550598,
+      "logits/rejected": -0.406808078289032,
+      "logps/chosen": -200.47714233398438,
+      "logps/rejected": -190.09597778320312,
+      "loss": 0.6675,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.7260812520980835,
+      "rewards/margins": 0.08043357729911804,
+      "rewards/rejected": 0.6456476449966431,
+      "step": 3180
+    },
+    {
+      "epoch": 1.0046848549269713,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.487920380793386e-06,
+      "logits/chosen": -0.5357908010482788,
+      "logits/rejected": -0.40378838777542114,
+      "logps/chosen": -195.0863494873047,
+      "logps/rejected": -157.54249572753906,
+      "loss": 0.6263,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.7121897339820862,
+      "rewards/margins": 0.17122754454612732,
+      "rewards/rejected": 0.5409621596336365,
+      "step": 3190
+    },
+    {
+      "epoch": 1.0078343372308176,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.4847196100799305e-06,
+      "logits/chosen": -0.5046489834785461,
+      "logits/rejected": -0.38591229915618896,
+      "logps/chosen": -206.5364227294922,
+      "logps/rejected": -177.38302612304688,
+      "loss": 0.6648,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.698761522769928,
+      "rewards/margins": 0.10818042606115341,
+      "rewards/rejected": 0.5905810594558716,
+      "step": 3200
+    },
+    {
+      "epoch": 1.010983819534664,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.481510016164753e-06,
+      "logits/chosen": -0.4204939007759094,
+      "logits/rejected": -0.32506507635116577,
+      "logps/chosen": -186.22238159179688,
+      "logps/rejected": -157.1585235595703,
+      "loss": 0.6338,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.7055295705795288,
+      "rewards/margins": 0.16552086174488068,
+      "rewards/rejected": 0.5400087237358093,
+      "step": 3210
+    },
+    {
+      "epoch": 1.0141333018385104,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.478291613316316e-06,
+      "logits/chosen": -0.526237428188324,
+      "logits/rejected": -0.5045801997184753,
+      "logps/chosen": -179.8779754638672,
+      "logps/rejected": -179.01283264160156,
+      "loss": 0.6786,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.6495916247367859,
+      "rewards/margins": 0.07142230123281479,
+      "rewards/rejected": 0.5781692862510681,
+      "step": 3220
+    },
+    {
+      "epoch": 1.0172827841423566,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.47506441584224e-06,
+      "logits/chosen": -0.47785645723342896,
+      "logits/rejected": -0.37066784501075745,
+      "logps/chosen": -176.42306518554688,
+      "logps/rejected": -158.9258575439453,
+      "loss": 0.6722,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.6187280416488647,
+      "rewards/margins": 0.08662554621696472,
+      "rewards/rejected": 0.5321024656295776,
+      "step": 3230
+    },
+    {
+      "epoch": 1.0204322664462029,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.471828438089245e-06,
+      "logits/chosen": -0.5045980215072632,
+      "logits/rejected": -0.3146094083786011,
+      "logps/chosen": -192.99447631835938,
+      "logps/rejected": -160.2574462890625,
+      "loss": 0.6403,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.6895080208778381,
+      "rewards/margins": 0.1523585319519043,
+      "rewards/rejected": 0.5371494293212891,
+      "step": 3240
+    },
+    {
+      "epoch": 1.0235817487500491,
+      "grad_norm": 2.609375,
+      "learning_rate": 4.4685836944430815e-06,
+      "logits/chosen": -0.45219331979751587,
+      "logits/rejected": -0.3365115821361542,
+      "logps/chosen": -212.85543823242188,
+      "logps/rejected": -181.6865692138672,
+      "loss": 0.6293,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.7488403916358948,
+      "rewards/margins": 0.16951295733451843,
+      "rewards/rejected": 0.579327404499054,
+      "step": 3250
+    },
+    {
+      "epoch": 1.0267312310538954,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.465330199328473e-06,
+      "logits/chosen": -0.46741557121276855,
+      "logits/rejected": -0.4280362129211426,
+      "logps/chosen": -201.90011596679688,
+      "logps/rejected": -174.54420471191406,
+      "loss": 0.6809,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.692274808883667,
+      "rewards/margins": 0.0669262483716011,
+      "rewards/rejected": 0.625348687171936,
+      "step": 3260
+    },
+    {
+      "epoch": 1.029880713357742,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.462067967209045e-06,
+      "logits/chosen": -0.46599286794662476,
+      "logits/rejected": -0.40415382385253906,
+      "logps/chosen": -191.61109924316406,
+      "logps/rejected": -170.8994140625,
+      "loss": 0.637,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.694964587688446,
+      "rewards/margins": 0.15309689939022064,
+      "rewards/rejected": 0.541867733001709,
+      "step": 3270
+    },
+    {
+      "epoch": 1.0330301956615882,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.458797012587266e-06,
+      "logits/chosen": -0.45959001779556274,
+      "logits/rejected": -0.3480526804924011,
+      "logps/chosen": -194.1427764892578,
+      "logps/rejected": -164.58615112304688,
+      "loss": 0.6511,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.6329681277275085,
+      "rewards/margins": 0.11931748688220978,
+      "rewards/rejected": 0.51365065574646,
+      "step": 3280
+    },
+    {
+      "epoch": 1.0361796779654344,
+      "grad_norm": 2.5,
+      "learning_rate": 4.455517350004379e-06,
+      "logits/chosen": -0.4861987233161926,
+      "logits/rejected": -0.3476967513561249,
+      "logps/chosen": -218.5071563720703,
+      "logps/rejected": -177.85232543945312,
+      "loss": 0.6175,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.8144248723983765,
+      "rewards/margins": 0.19891203939914703,
+      "rewards/rejected": 0.6155128479003906,
+      "step": 3290
+    },
+    {
+      "epoch": 1.0393291602692807,
+      "grad_norm": 2.078125,
+      "learning_rate": 4.452228994040341e-06,
+      "logits/chosen": -0.4951728284358978,
+      "logits/rejected": -0.3645640015602112,
+      "logps/chosen": -196.14259338378906,
+      "logps/rejected": -167.96095275878906,
+      "loss": 0.6494,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.7171255946159363,
+      "rewards/margins": 0.12575490772724152,
+      "rewards/rejected": 0.591370701789856,
+      "step": 3300
+    },
+    {
+      "epoch": 1.042478642573127,
+      "grad_norm": 2.21875,
+      "learning_rate": 4.448931959313754e-06,
+      "logits/chosen": -0.4978647232055664,
+      "logits/rejected": -0.3358718454837799,
+      "logps/chosen": -200.82778930664062,
+      "logps/rejected": -170.21578979492188,
+      "loss": 0.6559,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.6936062574386597,
+      "rewards/margins": 0.12191645056009293,
+      "rewards/rejected": 0.5716897249221802,
+      "step": 3310
+    },
+    {
+      "epoch": 1.0456281248769734,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.4456262604818044e-06,
+      "logits/chosen": -0.5448322892189026,
+      "logits/rejected": -0.4078589379787445,
+      "logps/chosen": -202.68072509765625,
+      "logps/rejected": -172.15322875976562,
+      "loss": 0.65,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.6884664297103882,
+      "rewards/margins": 0.12657414376735687,
+      "rewards/rejected": 0.5618923306465149,
+      "step": 3320
+    },
+    {
+      "epoch": 1.0487776071808197,
+      "grad_norm": 4.0,
+      "learning_rate": 4.442311912240194e-06,
+      "logits/chosen": -0.46120721101760864,
+      "logits/rejected": -0.31904077529907227,
+      "logps/chosen": -204.01699829101562,
+      "logps/rejected": -172.23001098632812,
+      "loss": 0.6343,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6871546506881714,
+      "rewards/margins": 0.15920737385749817,
+      "rewards/rejected": 0.5279473066329956,
+      "step": 3330
+    },
+    {
+      "epoch": 1.051927089484666,
+      "grad_norm": 4.0625,
+      "learning_rate": 4.438988929323075e-06,
+      "logits/chosen": -0.4875815808773041,
+      "logits/rejected": -0.37348097562789917,
+      "logps/chosen": -197.5282745361328,
+      "logps/rejected": -167.55035400390625,
+      "loss": 0.6422,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.7295669317245483,
+      "rewards/margins": 0.15396803617477417,
+      "rewards/rejected": 0.5755988955497742,
+      "step": 3340
+    },
+    {
+      "epoch": 1.0550765717885122,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.435657326502986e-06,
+      "logits/chosen": -0.49304842948913574,
+      "logits/rejected": -0.3635968565940857,
+      "logps/chosen": -194.97134399414062,
+      "logps/rejected": -163.18759155273438,
+      "loss": 0.615,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.7523232698440552,
+      "rewards/margins": 0.2145862579345703,
+      "rewards/rejected": 0.5377371311187744,
+      "step": 3350
+    },
+    {
+      "epoch": 1.0582260540923585,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.432317118590789e-06,
+      "logits/chosen": -0.48891234397888184,
+      "logits/rejected": -0.3675716519355774,
+      "logps/chosen": -206.11074829101562,
+      "logps/rejected": -187.20938110351562,
+      "loss": 0.6612,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.6792635917663574,
+      "rewards/margins": 0.10087893903255463,
+      "rewards/rejected": 0.5783846974372864,
+      "step": 3360
+    },
+    {
+      "epoch": 1.061375536396205,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.428968320435597e-06,
+      "logits/chosen": -0.44766363501548767,
+      "logits/rejected": -0.442352294921875,
+      "logps/chosen": -195.1834716796875,
+      "logps/rejected": -194.13803100585938,
+      "loss": 0.6908,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.7168688178062439,
+      "rewards/margins": 0.04094035550951958,
+      "rewards/rejected": 0.6759284734725952,
+      "step": 3370
+    },
+    {
+      "epoch": 1.0645250187000512,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.425610946924714e-06,
+      "logits/chosen": -0.5080258250236511,
+      "logits/rejected": -0.3179413378238678,
+      "logps/chosen": -209.2963409423828,
+      "logps/rejected": -164.95855712890625,
+      "loss": 0.6388,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.693896472454071,
+      "rewards/margins": 0.1626008152961731,
+      "rewards/rejected": 0.5312955975532532,
+      "step": 3380
+    },
+    {
+      "epoch": 1.0676745010038975,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.422245012983563e-06,
+      "logits/chosen": -0.43999892473220825,
+      "logits/rejected": -0.34846392273902893,
+      "logps/chosen": -202.3621826171875,
+      "logps/rejected": -172.74575805664062,
+      "loss": 0.656,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.7476739287376404,
+      "rewards/margins": 0.1216663271188736,
+      "rewards/rejected": 0.6260076761245728,
+      "step": 3390
+    },
+    {
+      "epoch": 1.0708239833077438,
+      "grad_norm": 2.875,
+      "learning_rate": 4.418870533575626e-06,
+      "logits/chosen": -0.4500795006752014,
+      "logits/rejected": -0.35547685623168945,
+      "logps/chosen": -194.93258666992188,
+      "logps/rejected": -169.08042907714844,
+      "loss": 0.6452,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.6677473783493042,
+      "rewards/margins": 0.13091039657592773,
+      "rewards/rejected": 0.5368369221687317,
+      "step": 3400
+    },
+    {
+      "epoch": 1.07397346561159,
+      "grad_norm": 2.21875,
+      "learning_rate": 4.4154875237023725e-06,
+      "logits/chosen": -0.5026477575302124,
+      "logits/rejected": -0.42180952429771423,
+      "logps/chosen": -190.4978790283203,
+      "logps/rejected": -170.85133361816406,
+      "loss": 0.6539,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.6496149301528931,
+      "rewards/margins": 0.1192195862531662,
+      "rewards/rejected": 0.5303953289985657,
+      "step": 3410
+    },
+    {
+      "epoch": 1.0771229479154365,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.412095998403198e-06,
+      "logits/chosen": -0.467219740152359,
+      "logits/rejected": -0.36950141191482544,
+      "logps/chosen": -184.17282104492188,
+      "logps/rejected": -159.8118133544922,
+      "loss": 0.6296,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.6559081673622131,
+      "rewards/margins": 0.16591888666152954,
+      "rewards/rejected": 0.4899892210960388,
+      "step": 3420
+    },
+    {
+      "epoch": 1.0802724302192828,
+      "grad_norm": 2.0,
+      "learning_rate": 4.4086959727553484e-06,
+      "logits/chosen": -0.43744510412216187,
+      "logits/rejected": -0.4274655282497406,
+      "logps/chosen": -184.89468383789062,
+      "logps/rejected": -174.7428436279297,
+      "loss": 0.6674,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.6194896101951599,
+      "rewards/margins": 0.08610032498836517,
+      "rewards/rejected": 0.5333893299102783,
+      "step": 3430
+    },
+    {
+      "epoch": 1.083421912523129,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.4052874618738645e-06,
+      "logits/chosen": -0.42463669180870056,
+      "logits/rejected": -0.32110992074012756,
+      "logps/chosen": -198.18606567382812,
+      "logps/rejected": -166.31546020507812,
+      "loss": 0.6371,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.7485095262527466,
+      "rewards/margins": 0.16625897586345673,
+      "rewards/rejected": 0.5822504758834839,
+      "step": 3440
+    },
+    {
+      "epoch": 1.0865713948269753,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.401870480911505e-06,
+      "logits/chosen": -0.5295459032058716,
+      "logits/rejected": -0.3508453369140625,
+      "logps/chosen": -202.53451538085938,
+      "logps/rejected": -164.10360717773438,
+      "loss": 0.6324,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.7244966626167297,
+      "rewards/margins": 0.17259523272514343,
+      "rewards/rejected": 0.5519014000892639,
+      "step": 3450
+    },
+    {
+      "epoch": 1.0897208771308216,
+      "grad_norm": 2.171875,
+      "learning_rate": 4.398445045058682e-06,
+      "logits/chosen": -0.39355480670928955,
+      "logits/rejected": -0.3064250349998474,
+      "logps/chosen": -198.85391235351562,
+      "logps/rejected": -193.22476196289062,
+      "loss": 0.6711,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.7315341234207153,
+      "rewards/margins": 0.08587940782308578,
+      "rewards/rejected": 0.6456546783447266,
+      "step": 3460
+    },
+    {
+      "epoch": 1.0928703594346678,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.395011169543398e-06,
+      "logits/chosen": -0.5193358659744263,
+      "logits/rejected": -0.43463462591171265,
+      "logps/chosen": -221.0367431640625,
+      "logps/rejected": -189.76303100585938,
+      "loss": 0.6429,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.7745602130889893,
+      "rewards/margins": 0.15965554118156433,
+      "rewards/rejected": 0.6149047017097473,
+      "step": 3470
+    },
+    {
+      "epoch": 1.0960198417385143,
+      "grad_norm": 2.5,
+      "learning_rate": 4.3915688696311734e-06,
+      "logits/chosen": -0.47026944160461426,
+      "logits/rejected": -0.3491131067276001,
+      "logps/chosen": -193.53817749023438,
+      "logps/rejected": -180.67047119140625,
+      "loss": 0.6465,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.7157790660858154,
+      "rewards/margins": 0.1319868266582489,
+      "rewards/rejected": 0.5837923288345337,
+      "step": 3480
+    },
+    {
+      "epoch": 1.0991693240423606,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.3881181606249775e-06,
+      "logits/chosen": -0.4417082369327545,
+      "logits/rejected": -0.33755919337272644,
+      "logps/chosen": -205.1912841796875,
+      "logps/rejected": -168.96607971191406,
+      "loss": 0.6527,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.7606759071350098,
+      "rewards/margins": 0.125040203332901,
+      "rewards/rejected": 0.6356357336044312,
+      "step": 3490
+    },
+    {
+      "epoch": 1.1023188063462068,
+      "grad_norm": 3.5,
+      "learning_rate": 4.384659057865165e-06,
+      "logits/chosen": -0.4954513609409332,
+      "logits/rejected": -0.41869044303894043,
+      "logps/chosen": -207.7197265625,
+      "logps/rejected": -187.3145751953125,
+      "loss": 0.6409,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.7566835880279541,
+      "rewards/margins": 0.13307683169841766,
+      "rewards/rejected": 0.6236067414283752,
+      "step": 3500
+    },
+    {
+      "epoch": 1.105468288650053,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.381191576729404e-06,
+      "logits/chosen": -0.465837299823761,
+      "logits/rejected": -0.3935115933418274,
+      "logps/chosen": -189.66842651367188,
+      "logps/rejected": -181.11953735351562,
+      "loss": 0.67,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.6702840328216553,
+      "rewards/margins": 0.08853673934936523,
+      "rewards/rejected": 0.58174729347229,
+      "step": 3510
+    },
+    {
+      "epoch": 1.1086177709538994,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.377715732632613e-06,
+      "logits/chosen": -0.4160510003566742,
+      "logits/rejected": -0.34267672896385193,
+      "logps/chosen": -190.46127319335938,
+      "logps/rejected": -164.63645935058594,
+      "loss": 0.6486,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.7083535194396973,
+      "rewards/margins": 0.13883116841316223,
+      "rewards/rejected": 0.5695223808288574,
+      "step": 3520
+    },
+    {
+      "epoch": 1.1117672532577458,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.374231541026883e-06,
+      "logits/chosen": -0.5327466130256653,
+      "logits/rejected": -0.38102999329566956,
+      "logps/chosen": -208.6250457763672,
+      "logps/rejected": -187.94937133789062,
+      "loss": 0.6461,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.7454079985618591,
+      "rewards/margins": 0.1365070343017578,
+      "rewards/rejected": 0.6089010238647461,
+      "step": 3530
+    },
+    {
+      "epoch": 1.114916735561592,
+      "grad_norm": 3.703125,
+      "learning_rate": 4.370739017401417e-06,
+      "logits/chosen": -0.5027688145637512,
+      "logits/rejected": -0.4032517373561859,
+      "logps/chosen": -195.99122619628906,
+      "logps/rejected": -161.76515197753906,
+      "loss": 0.6612,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.7000411152839661,
+      "rewards/margins": 0.11141884326934814,
+      "rewards/rejected": 0.5886222720146179,
+      "step": 3540
+    },
+    {
+      "epoch": 1.1180662178654384,
+      "grad_norm": 4.1875,
+      "learning_rate": 4.367238177282462e-06,
+      "logits/chosen": -0.4465219974517822,
+      "logits/rejected": -0.3524473309516907,
+      "logps/chosen": -191.51556396484375,
+      "logps/rejected": -171.40370178222656,
+      "loss": 0.6437,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.6995719075202942,
+      "rewards/margins": 0.14049354195594788,
+      "rewards/rejected": 0.5590783953666687,
+      "step": 3550
+    },
+    {
+      "epoch": 1.1212157001692846,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.363729036233231e-06,
+      "logits/chosen": -0.440678209066391,
+      "logits/rejected": -0.35682061314582825,
+      "logps/chosen": -233.2520751953125,
+      "logps/rejected": -197.86093139648438,
+      "loss": 0.6151,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.7932190895080566,
+      "rewards/margins": 0.20204809308052063,
+      "rewards/rejected": 0.5911709666252136,
+      "step": 3560
+    },
+    {
+      "epoch": 1.1243651824731309,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.360211609853841e-06,
+      "logits/chosen": -0.46588149666786194,
+      "logits/rejected": -0.3290489614009857,
+      "logps/chosen": -189.09005737304688,
+      "logps/rejected": -161.47216796875,
+      "loss": 0.6275,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.7199033498764038,
+      "rewards/margins": 0.18104612827301025,
+      "rewards/rejected": 0.5388572216033936,
+      "step": 3570
+    },
+    {
+      "epoch": 1.1275146647769774,
+      "grad_norm": 3.53125,
+      "learning_rate": 4.356685913781243e-06,
+      "logits/chosen": -0.4571777284145355,
+      "logits/rejected": -0.2775833010673523,
+      "logps/chosen": -212.7926788330078,
+      "logps/rejected": -178.75543212890625,
+      "loss": 0.6815,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.7048720717430115,
+      "rewards/margins": 0.06307940185070038,
+      "rewards/rejected": 0.6417926549911499,
+      "step": 3580
+    },
+    {
+      "epoch": 1.1306641470808236,
+      "grad_norm": 1.96875,
+      "learning_rate": 4.353151963689153e-06,
+      "logits/chosen": -0.4465222954750061,
+      "logits/rejected": -0.33635538816452026,
+      "logps/chosen": -193.8217315673828,
+      "logps/rejected": -170.3684844970703,
+      "loss": 0.656,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.6877092719078064,
+      "rewards/margins": 0.12182395160198212,
+      "rewards/rejected": 0.5658854246139526,
+      "step": 3590
+    },
+    {
+      "epoch": 1.13381362938467,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.349609775287977e-06,
+      "logits/chosen": -0.45926880836486816,
+      "logits/rejected": -0.3848228454589844,
+      "logps/chosen": -188.42308044433594,
+      "logps/rejected": -178.51895141601562,
+      "loss": 0.6831,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.625271201133728,
+      "rewards/margins": 0.08226005733013153,
+      "rewards/rejected": 0.5430110692977905,
+      "step": 3600
+    },
+    {
+      "epoch": 1.1369631116885162,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.346059364324747e-06,
+      "logits/chosen": -0.5314878821372986,
+      "logits/rejected": -0.4037748873233795,
+      "logps/chosen": -198.21347045898438,
+      "logps/rejected": -172.4250946044922,
+      "loss": 0.6247,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.7435242533683777,
+      "rewards/margins": 0.1861993968486786,
+      "rewards/rejected": 0.5573248267173767,
+      "step": 3610
+    },
+    {
+      "epoch": 1.1401125939923624,
+      "grad_norm": 3.0,
+      "learning_rate": 4.342500746583049e-06,
+      "logits/chosen": -0.4643464982509613,
+      "logits/rejected": -0.39244017004966736,
+      "logps/chosen": -199.8033447265625,
+      "logps/rejected": -178.8501739501953,
+      "loss": 0.6568,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.7149983644485474,
+      "rewards/margins": 0.11290229856967926,
+      "rewards/rejected": 0.6020959615707397,
+      "step": 3620
+    },
+    {
+      "epoch": 1.143262076296209,
+      "grad_norm": 2.171875,
+      "learning_rate": 4.338933937882952e-06,
+      "logits/chosen": -0.48191675543785095,
+      "logits/rejected": -0.34128910303115845,
+      "logps/chosen": -192.5113983154297,
+      "logps/rejected": -171.77792358398438,
+      "loss": 0.6556,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.7115253210067749,
+      "rewards/margins": 0.11348304897546768,
+      "rewards/rejected": 0.5980421900749207,
+      "step": 3630
+    },
+    {
+      "epoch": 1.1464115586000552,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.335358954080939e-06,
+      "logits/chosen": -0.5026925802230835,
+      "logits/rejected": -0.3849731683731079,
+      "logps/chosen": -195.01315307617188,
+      "logps/rejected": -153.3354949951172,
+      "loss": 0.6349,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.664111316204071,
+      "rewards/margins": 0.1614895761013031,
+      "rewards/rejected": 0.5026217699050903,
+      "step": 3640
+    },
+    {
+      "epoch": 1.1495610409039014,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.331775811069837e-06,
+      "logits/chosen": -0.3999708294868469,
+      "logits/rejected": -0.3556813895702362,
+      "logps/chosen": -193.3236083984375,
+      "logps/rejected": -170.455810546875,
+      "loss": 0.6737,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.6599856019020081,
+      "rewards/margins": 0.08706656098365784,
+      "rewards/rejected": 0.5729190111160278,
+      "step": 3650
+    },
+    {
+      "epoch": 1.1527105232077477,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.328184524778743e-06,
+      "logits/chosen": -0.5279411673545837,
+      "logits/rejected": -0.4057609438896179,
+      "logps/chosen": -195.410400390625,
+      "logps/rejected": -158.7968292236328,
+      "loss": 0.6285,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.6674174070358276,
+      "rewards/margins": 0.166324645280838,
+      "rewards/rejected": 0.501092791557312,
+      "step": 3660
+    },
+    {
+      "epoch": 1.155860005511594,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.324585111172959e-06,
+      "logits/chosen": -0.46349841356277466,
+      "logits/rejected": -0.3997943699359894,
+      "logps/chosen": -198.0673370361328,
+      "logps/rejected": -179.5006103515625,
+      "loss": 0.6457,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.7249336242675781,
+      "rewards/margins": 0.15420496463775635,
+      "rewards/rejected": 0.5707286596298218,
+      "step": 3670
+    },
+    {
+      "epoch": 1.1590094878154402,
+      "grad_norm": 2.0,
+      "learning_rate": 4.320977586253911e-06,
+      "logits/chosen": -0.4722752571105957,
+      "logits/rejected": -0.3240264058113098,
+      "logps/chosen": -202.7415771484375,
+      "logps/rejected": -171.6900177001953,
+      "loss": 0.6431,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.669415295124054,
+      "rewards/margins": 0.1358814686536789,
+      "rewards/rejected": 0.5335337519645691,
+      "step": 3680
+    },
+    {
+      "epoch": 1.1621589701192867,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.317361966059092e-06,
+      "logits/chosen": -0.4404812753200531,
+      "logits/rejected": -0.32088351249694824,
+      "logps/chosen": -207.16323852539062,
+      "logps/rejected": -176.3754119873047,
+      "loss": 0.6309,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6894677877426147,
+      "rewards/margins": 0.16316859424114227,
+      "rewards/rejected": 0.5262991786003113,
+      "step": 3690
+    },
+    {
+      "epoch": 1.165308452423133,
+      "grad_norm": 6.875,
+      "learning_rate": 4.313738266661979e-06,
+      "logits/chosen": -0.4920511841773987,
+      "logits/rejected": -0.34299007058143616,
+      "logps/chosen": -201.88381958007812,
+      "logps/rejected": -172.03121948242188,
+      "loss": 0.6224,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.7374047636985779,
+      "rewards/margins": 0.18349048495292664,
+      "rewards/rejected": 0.5539143681526184,
+      "step": 3700
+    },
+    {
+      "epoch": 1.1684579347269792,
+      "grad_norm": 3.0,
+      "learning_rate": 4.310106504171966e-06,
+      "logits/chosen": -0.45534056425094604,
+      "logits/rejected": -0.35874494910240173,
+      "logps/chosen": -190.65248107910156,
+      "logps/rejected": -158.1346893310547,
+      "loss": 0.6446,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.6649999618530273,
+      "rewards/margins": 0.14345432817935944,
+      "rewards/rejected": 0.5215457081794739,
+      "step": 3710
+    },
+    {
+      "epoch": 1.1716074170308255,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.306466694734292e-06,
+      "logits/chosen": -0.4667026102542877,
+      "logits/rejected": -0.33577385544776917,
+      "logps/chosen": -194.20590209960938,
+      "logps/rejected": -177.63400268554688,
+      "loss": 0.6922,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.6832542419433594,
+      "rewards/margins": 0.034148164093494415,
+      "rewards/rejected": 0.6491062045097351,
+      "step": 3720
+    },
+    {
+      "epoch": 1.174756899334672,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.302818854529969e-06,
+      "logits/chosen": -0.5350615382194519,
+      "logits/rejected": -0.42588871717453003,
+      "logps/chosen": -198.84042358398438,
+      "logps/rejected": -164.7383270263672,
+      "loss": 0.6171,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.7313185334205627,
+      "rewards/margins": 0.19767650961875916,
+      "rewards/rejected": 0.5336421132087708,
+      "step": 3730
+    },
+    {
+      "epoch": 1.1779063816385182,
+      "grad_norm": 4.53125,
+      "learning_rate": 4.299162999775712e-06,
+      "logits/chosen": -0.40838590264320374,
+      "logits/rejected": -0.37127965688705444,
+      "logps/chosen": -196.0567169189453,
+      "logps/rejected": -196.9461669921875,
+      "loss": 0.69,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.7238589525222778,
+      "rewards/margins": 0.06422169506549835,
+      "rewards/rejected": 0.6596371531486511,
+      "step": 3740
+    },
+    {
+      "epoch": 1.1810558639423645,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.295499146723864e-06,
+      "logits/chosen": -0.5344425439834595,
+      "logits/rejected": -0.35419797897338867,
+      "logps/chosen": -202.3170623779297,
+      "logps/rejected": -161.15969848632812,
+      "loss": 0.6066,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.7712113261222839,
+      "rewards/margins": 0.229976087808609,
+      "rewards/rejected": 0.5412352681159973,
+      "step": 3750
+    },
+    {
+      "epoch": 1.1842053462462108,
+      "grad_norm": 2.453125,
+      "learning_rate": 4.2918273116623245e-06,
+      "logits/chosen": -0.39341261982917786,
+      "logits/rejected": -0.3262158930301666,
+      "logps/chosen": -178.946533203125,
+      "logps/rejected": -171.49209594726562,
+      "loss": 0.6667,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.6648285388946533,
+      "rewards/margins": 0.08790337294340134,
+      "rewards/rejected": 0.5769251585006714,
+      "step": 3760
+    },
+    {
+      "epoch": 1.187354828550057,
+      "grad_norm": 2.359375,
+      "learning_rate": 4.288147510914477e-06,
+      "logits/chosen": -0.46092867851257324,
+      "logits/rejected": -0.44449153542518616,
+      "logps/chosen": -199.6840362548828,
+      "logps/rejected": -191.94036865234375,
+      "loss": 0.6985,
+      "rewards/accuracies": 0.512499988079071,
+      "rewards/chosen": 0.6821948885917664,
+      "rewards/margins": 0.02965412102639675,
+      "rewards/rejected": 0.6525408029556274,
+      "step": 3770
+    },
+    {
+      "epoch": 1.1905043108539033,
+      "grad_norm": 2.453125,
+      "learning_rate": 4.284459760839122e-06,
+      "logits/chosen": -0.5228853225708008,
+      "logits/rejected": -0.3966136574745178,
+      "logps/chosen": -197.607177734375,
+      "logps/rejected": -169.21734619140625,
+      "loss": 0.6785,
+      "rewards/accuracies": 0.48750001192092896,
+      "rewards/chosen": 0.6453790664672852,
+      "rewards/margins": 0.06496746838092804,
+      "rewards/rejected": 0.5804116129875183,
+      "step": 3780
+    },
+    {
+      "epoch": 1.1936537931577498,
+      "grad_norm": 2.015625,
+      "learning_rate": 4.28076407783039e-06,
+      "logits/chosen": -0.4347180724143982,
+      "logits/rejected": -0.3091747760772705,
+      "logps/chosen": -183.70968627929688,
+      "logps/rejected": -151.46176147460938,
+      "loss": 0.5928,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.7321251034736633,
+      "rewards/margins": 0.25292515754699707,
+      "rewards/rejected": 0.4791998863220215,
+      "step": 3790
+    },
+    {
+      "epoch": 1.196803275461596,
+      "grad_norm": 2.375,
+      "learning_rate": 4.277060478317687e-06,
+      "logits/chosen": -0.5005819797515869,
+      "logits/rejected": -0.36348724365234375,
+      "logps/chosen": -205.44021606445312,
+      "logps/rejected": -170.33944702148438,
+      "loss": 0.6353,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.6576865911483765,
+      "rewards/margins": 0.1621154546737671,
+      "rewards/rejected": 0.49557122588157654,
+      "step": 3800
+    },
+    {
+      "epoch": 1.1999527577654423,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.2733489787656075e-06,
+      "logits/chosen": -0.4653560221195221,
+      "logits/rejected": -0.3435734808444977,
+      "logps/chosen": -185.8004913330078,
+      "logps/rejected": -159.63597106933594,
+      "loss": 0.6312,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.6951313018798828,
+      "rewards/margins": 0.1663820743560791,
+      "rewards/rejected": 0.5287492871284485,
+      "step": 3810
+    },
+    {
+      "epoch": 1.2031022400692886,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.269629595673867e-06,
+      "logits/chosen": -0.478889524936676,
+      "logits/rejected": -0.4129001200199127,
+      "logps/chosen": -221.73828125,
+      "logps/rejected": -191.03372192382812,
+      "loss": 0.6142,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.8134725689888,
+      "rewards/margins": 0.20523671805858612,
+      "rewards/rejected": 0.6082358360290527,
+      "step": 3820
+    },
+    {
+      "epoch": 1.2062517223731348,
+      "grad_norm": 2.515625,
+      "learning_rate": 4.265902345577227e-06,
+      "logits/chosen": -0.4593692421913147,
+      "logits/rejected": -0.3384065628051758,
+      "logps/chosen": -186.61167907714844,
+      "logps/rejected": -161.26182556152344,
+      "loss": 0.6454,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.660686731338501,
+      "rewards/margins": 0.13643305003643036,
+      "rewards/rejected": 0.5242536664009094,
+      "step": 3830
+    },
+    {
+      "epoch": 1.2094012046769813,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.262167245045424e-06,
+      "logits/chosen": -0.503960132598877,
+      "logits/rejected": -0.3568868339061737,
+      "logps/chosen": -189.79818725585938,
+      "logps/rejected": -160.52206420898438,
+      "loss": 0.6303,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.7322565317153931,
+      "rewards/margins": 0.18888680636882782,
+      "rewards/rejected": 0.5433696508407593,
+      "step": 3840
+    },
+    {
+      "epoch": 1.2125506869808276,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.258424310683094e-06,
+      "logits/chosen": -0.5178459882736206,
+      "logits/rejected": -0.4422483444213867,
+      "logps/chosen": -202.4190216064453,
+      "logps/rejected": -191.25418090820312,
+      "loss": 0.6634,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.6625019311904907,
+      "rewards/margins": 0.09516273438930511,
+      "rewards/rejected": 0.5673390626907349,
+      "step": 3850
+    },
+    {
+      "epoch": 1.2157001692846738,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.254673559129698e-06,
+      "logits/chosen": -0.4428611695766449,
+      "logits/rejected": -0.3638356328010559,
+      "logps/chosen": -177.161376953125,
+      "logps/rejected": -158.397216796875,
+      "loss": 0.6791,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.660545825958252,
+      "rewards/margins": 0.06867508590221405,
+      "rewards/rejected": 0.5918707251548767,
+      "step": 3860
+    },
+    {
+      "epoch": 1.21884965158852,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.250915007059448e-06,
+      "logits/chosen": -0.4426344037055969,
+      "logits/rejected": -0.39190584421157837,
+      "logps/chosen": -182.5616455078125,
+      "logps/rejected": -166.43875122070312,
+      "loss": 0.6467,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.6761131286621094,
+      "rewards/margins": 0.147123783826828,
+      "rewards/rejected": 0.5289894342422485,
+      "step": 3870
+    },
+    {
+      "epoch": 1.2219991338923664,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.247148671181237e-06,
+      "logits/chosen": -0.5092406868934631,
+      "logits/rejected": -0.41020458936691284,
+      "logps/chosen": -190.69708251953125,
+      "logps/rejected": -165.26943969726562,
+      "loss": 0.6436,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.7014779448509216,
+      "rewards/margins": 0.13978223502635956,
+      "rewards/rejected": 0.5616958141326904,
+      "step": 3880
+    },
+    {
+      "epoch": 1.2251486161962126,
+      "grad_norm": 2.0625,
+      "learning_rate": 4.243374568238556e-06,
+      "logits/chosen": -0.49269500374794006,
+      "logits/rejected": -0.34552276134490967,
+      "logps/chosen": -197.5285186767578,
+      "logps/rejected": -163.1229248046875,
+      "loss": 0.6266,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.686238706111908,
+      "rewards/margins": 0.17949660122394562,
+      "rewards/rejected": 0.5067421197891235,
+      "step": 3890
+    },
+    {
+      "epoch": 1.228298098500059,
+      "grad_norm": 2.453125,
+      "learning_rate": 4.23959271500943e-06,
+      "logits/chosen": -0.4887842535972595,
+      "logits/rejected": -0.32892632484436035,
+      "logps/chosen": -213.9662322998047,
+      "logps/rejected": -175.28543090820312,
+      "loss": 0.6495,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.7332077026367188,
+      "rewards/margins": 0.14412468671798706,
+      "rewards/rejected": 0.5890830755233765,
+      "step": 3900
+    },
+    {
+      "epoch": 1.2314475808039054,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.235803128306337e-06,
+      "logits/chosen": -0.4606234133243561,
+      "logits/rejected": -0.3164953291416168,
+      "logps/chosen": -217.72933959960938,
+      "logps/rejected": -188.64517211914062,
+      "loss": 0.6548,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.7374159693717957,
+      "rewards/margins": 0.12077488750219345,
+      "rewards/rejected": 0.6166411638259888,
+      "step": 3910
+    },
+    {
+      "epoch": 1.2345970631077516,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.232005824976133e-06,
+      "logits/chosen": -0.4276387095451355,
+      "logits/rejected": -0.34980887174606323,
+      "logps/chosen": -212.1448516845703,
+      "logps/rejected": -174.04776000976562,
+      "loss": 0.6255,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.7644809484481812,
+      "rewards/margins": 0.19336572289466858,
+      "rewards/rejected": 0.5711151361465454,
+      "step": 3920
+    },
+    {
+      "epoch": 1.237746545411598,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.22820082189998e-06,
+      "logits/chosen": -0.4039751887321472,
+      "logits/rejected": -0.44209712743759155,
+      "logps/chosen": -198.69827270507812,
+      "logps/rejected": -201.2398681640625,
+      "loss": 0.6825,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.6999807357788086,
+      "rewards/margins": 0.07114030420780182,
+      "rewards/rejected": 0.6288403868675232,
+      "step": 3930
+    },
+    {
+      "epoch": 1.2408960277154444,
+      "grad_norm": 3.25,
+      "learning_rate": 4.224388135993271e-06,
+      "logits/chosen": -0.4334556460380554,
+      "logits/rejected": -0.30388978123664856,
+      "logps/chosen": -206.3880157470703,
+      "logps/rejected": -177.33868408203125,
+      "loss": 0.6462,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.745646595954895,
+      "rewards/margins": 0.1241118311882019,
+      "rewards/rejected": 0.6215347051620483,
+      "step": 3940
+    },
+    {
+      "epoch": 1.2440455100192906,
+      "grad_norm": 3.5,
+      "learning_rate": 4.220567784205551e-06,
+      "logits/chosen": -0.38715043663978577,
+      "logits/rejected": -0.2959139049053192,
+      "logps/chosen": -204.772216796875,
+      "logps/rejected": -177.6510772705078,
+      "loss": 0.649,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.7557042837142944,
+      "rewards/margins": 0.1426253467798233,
+      "rewards/rejected": 0.6130789518356323,
+      "step": 3950
+    },
+    {
+      "epoch": 1.247194992323137,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.216739783520447e-06,
+      "logits/chosen": -0.44497212767601013,
+      "logits/rejected": -0.3760547935962677,
+      "logps/chosen": -193.57406616210938,
+      "logps/rejected": -169.6040802001953,
+      "loss": 0.6622,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.6512136459350586,
+      "rewards/margins": 0.10468528419733047,
+      "rewards/rejected": 0.5465283989906311,
+      "step": 3960
+    },
+    {
+      "epoch": 1.2503444746269832,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.212904150955587e-06,
+      "logits/chosen": -0.512496292591095,
+      "logits/rejected": -0.40937352180480957,
+      "logps/chosen": -215.65908813476562,
+      "logps/rejected": -184.07162475585938,
+      "loss": 0.6385,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.8019892573356628,
+      "rewards/margins": 0.16033987700939178,
+      "rewards/rejected": 0.6416494250297546,
+      "step": 3970
+    },
+    {
+      "epoch": 1.2534939569308294,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.209060903562528e-06,
+      "logits/chosen": -0.4832407832145691,
+      "logits/rejected": -0.37063470482826233,
+      "logps/chosen": -211.16928100585938,
+      "logps/rejected": -176.49264526367188,
+      "loss": 0.6414,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.7144185900688171,
+      "rewards/margins": 0.13941380381584167,
+      "rewards/rejected": 0.5750047564506531,
+      "step": 3980
+    },
+    {
+      "epoch": 1.2566434392346757,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.20521005842668e-06,
+      "logits/chosen": -0.4679936468601227,
+      "logits/rejected": -0.320305734872818,
+      "logps/chosen": -198.4337615966797,
+      "logps/rejected": -156.4415283203125,
+      "loss": 0.615,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.717679500579834,
+      "rewards/margins": 0.20881009101867676,
+      "rewards/rejected": 0.5088694095611572,
+      "step": 3990
+    },
+    {
+      "epoch": 1.2597929215385222,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.201351632667227e-06,
+      "logits/chosen": -0.4357661306858063,
+      "logits/rejected": -0.33136463165283203,
+      "logps/chosen": -215.5205535888672,
+      "logps/rejected": -194.84225463867188,
+      "loss": 0.6584,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.7862467169761658,
+      "rewards/margins": 0.1381584107875824,
+      "rewards/rejected": 0.6480883359909058,
+      "step": 4000
+    },
+    {
+      "epoch": 1.2597929215385222,
+      "eval_logits/chosen": -0.6012239456176758,
+      "eval_logits/rejected": -0.4762186110019684,
+      "eval_logps/chosen": -243.07638549804688,
+      "eval_logps/rejected": -222.3102569580078,
+      "eval_loss": 0.6651261448860168,
+      "eval_rewards/accuracies": 0.5946148037910461,
+      "eval_rewards/chosen": 0.8024876117706299,
+      "eval_rewards/margins": 0.10073534399271011,
+      "eval_rewards/rejected": 0.701752245426178,
+      "eval_runtime": 3657.1235,
+      "eval_samples_per_second": 0.366,
+      "eval_steps_per_second": 0.366,
+      "step": 4000
+    },
+    {
+      "epoch": 1.2629424038423684,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.197485643437058e-06,
+      "logits/chosen": -0.45790332555770874,
+      "logits/rejected": -0.3080021142959595,
+      "logps/chosen": -194.04660034179688,
+      "logps/rejected": -152.27664184570312,
+      "loss": 0.6051,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.7683263421058655,
+      "rewards/margins": 0.22982630133628845,
+      "rewards/rejected": 0.5384999513626099,
+      "step": 4010
+    },
+    {
+      "epoch": 1.2660918861462147,
+      "grad_norm": 3.765625,
+      "learning_rate": 4.19361210792268e-06,
+      "logits/chosen": -0.45886415243148804,
+      "logits/rejected": -0.36696118116378784,
+      "logps/chosen": -197.91018676757812,
+      "logps/rejected": -165.89205932617188,
+      "loss": 0.6525,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.649300217628479,
+      "rewards/margins": 0.12859824299812317,
+      "rewards/rejected": 0.5207020044326782,
+      "step": 4020
+    },
+    {
+      "epoch": 1.269241368450061,
+      "grad_norm": 2.0625,
+      "learning_rate": 4.189731043344151e-06,
+      "logits/chosen": -0.5488325953483582,
+      "logits/rejected": -0.3945949673652649,
+      "logps/chosen": -198.27198791503906,
+      "logps/rejected": -160.57528686523438,
+      "loss": 0.6517,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.6978427767753601,
+      "rewards/margins": 0.1346282958984375,
+      "rewards/rejected": 0.5632144808769226,
+      "step": 4030
+    },
+    {
+      "epoch": 1.2723908507539075,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.185842466954998e-06,
+      "logits/chosen": -0.5013604760169983,
+      "logits/rejected": -0.40894627571105957,
+      "logps/chosen": -192.42051696777344,
+      "logps/rejected": -167.13221740722656,
+      "loss": 0.6322,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.6890965700149536,
+      "rewards/margins": 0.17353561520576477,
+      "rewards/rejected": 0.5155609846115112,
+      "step": 4040
+    },
+    {
+      "epoch": 1.2755403330577537,
+      "grad_norm": 3.265625,
+      "learning_rate": 4.181946396042146e-06,
+      "logits/chosen": -0.46542900800704956,
+      "logits/rejected": -0.3752228319644928,
+      "logps/chosen": -204.59176635742188,
+      "logps/rejected": -184.3190155029297,
+      "loss": 0.646,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.712192952632904,
+      "rewards/margins": 0.12876209616661072,
+      "rewards/rejected": 0.5834308862686157,
+      "step": 4050
+    },
+    {
+      "epoch": 1.2786898153616,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.178042847925833e-06,
+      "logits/chosen": -0.4636574387550354,
+      "logits/rejected": -0.4096450209617615,
+      "logps/chosen": -207.9044952392578,
+      "logps/rejected": -191.27566528320312,
+      "loss": 0.6599,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.7198086380958557,
+      "rewards/margins": 0.10640084743499756,
+      "rewards/rejected": 0.6134077906608582,
+      "step": 4060
+    },
+    {
+      "epoch": 1.2818392976654462,
+      "grad_norm": 2.609375,
+      "learning_rate": 4.174131839959539e-06,
+      "logits/chosen": -0.5049449801445007,
+      "logits/rejected": -0.37073415517807007,
+      "logps/chosen": -201.08135986328125,
+      "logps/rejected": -178.99465942382812,
+      "loss": 0.6637,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.7069438099861145,
+      "rewards/margins": 0.11059415340423584,
+      "rewards/rejected": 0.5963497161865234,
+      "step": 4070
+    },
+    {
+      "epoch": 1.2849887799692925,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.170213389529908e-06,
+      "logits/chosen": -0.4413929879665375,
+      "logits/rejected": -0.37840536236763,
+      "logps/chosen": -186.3639678955078,
+      "logps/rejected": -180.01528930664062,
+      "loss": 0.657,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.664527416229248,
+      "rewards/margins": 0.11193283647298813,
+      "rewards/rejected": 0.5525946617126465,
+      "step": 4080
+    },
+    {
+      "epoch": 1.2881382622731388,
+      "grad_norm": 2.359375,
+      "learning_rate": 4.16628751405667e-06,
+      "logits/chosen": -0.4986829161643982,
+      "logits/rejected": -0.34226298332214355,
+      "logps/chosen": -184.44613647460938,
+      "logps/rejected": -149.0967254638672,
+      "loss": 0.6311,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6766608953475952,
+      "rewards/margins": 0.16855625808238983,
+      "rewards/rejected": 0.508104681968689,
+      "step": 4090
+    },
+    {
+      "epoch": 1.291287744576985,
+      "grad_norm": 2.375,
+      "learning_rate": 4.162354230992562e-06,
+      "logits/chosen": -0.5013774633407593,
+      "logits/rejected": -0.4048996865749359,
+      "logps/chosen": -198.087646484375,
+      "logps/rejected": -170.59884643554688,
+      "loss": 0.6539,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.7187263369560242,
+      "rewards/margins": 0.12955203652381897,
+      "rewards/rejected": 0.5891742706298828,
+      "step": 4100
+    },
+    {
+      "epoch": 1.2944372268808315,
+      "grad_norm": 2.625,
+      "learning_rate": 4.158413557823253e-06,
+      "logits/chosen": -0.5318306684494019,
+      "logits/rejected": -0.3859057128429413,
+      "logps/chosen": -198.9587860107422,
+      "logps/rejected": -168.4477081298828,
+      "loss": 0.638,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.7599405646324158,
+      "rewards/margins": 0.16434063017368317,
+      "rewards/rejected": 0.5955999493598938,
+      "step": 4110
+    },
+    {
+      "epoch": 1.2975867091846778,
+      "grad_norm": 2.875,
+      "learning_rate": 4.154465512067266e-06,
+      "logits/chosen": -0.4665352702140808,
+      "logits/rejected": -0.3361702561378479,
+      "logps/chosen": -219.12088012695312,
+      "logps/rejected": -174.34353637695312,
+      "loss": 0.6203,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.7928240895271301,
+      "rewards/margins": 0.19742821156978607,
+      "rewards/rejected": 0.5953959226608276,
+      "step": 4120
+    },
+    {
+      "epoch": 1.300736191488524,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.1505101112758975e-06,
+      "logits/chosen": -0.5339746475219727,
+      "logits/rejected": -0.45769819617271423,
+      "logps/chosen": -180.9729766845703,
+      "logps/rejected": -161.6527557373047,
+      "loss": 0.666,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.6436579823493958,
+      "rewards/margins": 0.08750168979167938,
+      "rewards/rejected": 0.5561562180519104,
+      "step": 4130
+    },
+    {
+      "epoch": 1.3038856737923703,
+      "grad_norm": 2.625,
+      "learning_rate": 4.146547373033142e-06,
+      "logits/chosen": -0.44753965735435486,
+      "logits/rejected": -0.33734267950057983,
+      "logps/chosen": -184.26077270507812,
+      "logps/rejected": -155.27151489257812,
+      "loss": 0.6196,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.7223949432373047,
+      "rewards/margins": 0.19382601976394653,
+      "rewards/rejected": 0.5285689234733582,
+      "step": 4140
+    },
+    {
+      "epoch": 1.3070351560962168,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.142577314955614e-06,
+      "logits/chosen": -0.4315119683742523,
+      "logits/rejected": -0.27412423491477966,
+      "logps/chosen": -203.50393676757812,
+      "logps/rejected": -160.0498504638672,
+      "loss": 0.6151,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.7396818995475769,
+      "rewards/margins": 0.2151108682155609,
+      "rewards/rejected": 0.5245710015296936,
+      "step": 4150
+    },
+    {
+      "epoch": 1.310184638400063,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.138599954692467e-06,
+      "logits/chosen": -0.4760667681694031,
+      "logits/rejected": -0.4465053677558899,
+      "logps/chosen": -204.943359375,
+      "logps/rejected": -194.9350128173828,
+      "loss": 0.6528,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.7579629421234131,
+      "rewards/margins": 0.136540025472641,
+      "rewards/rejected": 0.6214228868484497,
+      "step": 4160
+    },
+    {
+      "epoch": 1.3133341207039093,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.13461530992532e-06,
+      "logits/chosen": -0.5714303255081177,
+      "logits/rejected": -0.39394524693489075,
+      "logps/chosen": -186.2847137451172,
+      "logps/rejected": -156.30838012695312,
+      "loss": 0.6444,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.6628353595733643,
+      "rewards/margins": 0.14596715569496155,
+      "rewards/rejected": 0.5168682336807251,
+      "step": 4170
+    },
+    {
+      "epoch": 1.3164836030077556,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.130623398368171e-06,
+      "logits/chosen": -0.4178102910518646,
+      "logits/rejected": -0.3646093010902405,
+      "logps/chosen": -174.4610595703125,
+      "logps/rejected": -152.55731201171875,
+      "loss": 0.6464,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.6118752360343933,
+      "rewards/margins": 0.13116849958896637,
+      "rewards/rejected": 0.48070669174194336,
+      "step": 4180
+    },
+    {
+      "epoch": 1.3196330853116018,
+      "grad_norm": 2.21875,
+      "learning_rate": 4.126624237767328e-06,
+      "logits/chosen": -0.4415621757507324,
+      "logits/rejected": -0.3622170686721802,
+      "logps/chosen": -186.65716552734375,
+      "logps/rejected": -164.0797576904297,
+      "loss": 0.6271,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.6536270380020142,
+      "rewards/margins": 0.1805739849805832,
+      "rewards/rejected": 0.47305306792259216,
+      "step": 4190
+    },
+    {
+      "epoch": 1.322782567615448,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.122617845901322e-06,
+      "logits/chosen": -0.5048421025276184,
+      "logits/rejected": -0.41297560930252075,
+      "logps/chosen": -190.35025024414062,
+      "logps/rejected": -161.85830688476562,
+      "loss": 0.6162,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.731613278388977,
+      "rewards/margins": 0.20245853066444397,
+      "rewards/rejected": 0.5291547179222107,
+      "step": 4200
+    },
+    {
+      "epoch": 1.3259320499192946,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.118604240580832e-06,
+      "logits/chosen": -0.40836066007614136,
+      "logits/rejected": -0.35586100816726685,
+      "logps/chosen": -199.27098083496094,
+      "logps/rejected": -172.96218872070312,
+      "loss": 0.6595,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.7109326720237732,
+      "rewards/margins": 0.11674892902374268,
+      "rewards/rejected": 0.5941838026046753,
+      "step": 4210
+    },
+    {
+      "epoch": 1.3290815322231408,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.114583439648604e-06,
+      "logits/chosen": -0.47792333364486694,
+      "logits/rejected": -0.3548354506492615,
+      "logps/chosen": -212.23178100585938,
+      "logps/rejected": -169.51315307617188,
+      "loss": 0.6391,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.6976093053817749,
+      "rewards/margins": 0.15473678708076477,
+      "rewards/rejected": 0.5428725481033325,
+      "step": 4220
+    },
+    {
+      "epoch": 1.3322310145269871,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.110555460979374e-06,
+      "logits/chosen": -0.5251237154006958,
+      "logits/rejected": -0.4099500775337219,
+      "logps/chosen": -200.9913787841797,
+      "logps/rejected": -171.32870483398438,
+      "loss": 0.6298,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.7174925804138184,
+      "rewards/margins": 0.1686524748802185,
+      "rewards/rejected": 0.5488401055335999,
+      "step": 4230
+    },
+    {
+      "epoch": 1.3353804968308334,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.106520322479786e-06,
+      "logits/chosen": -0.4651850163936615,
+      "logits/rejected": -0.34185513854026794,
+      "logps/chosen": -196.94589233398438,
+      "logps/rejected": -159.74307250976562,
+      "loss": 0.6523,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.6821771860122681,
+      "rewards/margins": 0.11817701160907745,
+      "rewards/rejected": 0.564000129699707,
+      "step": 4240
+    },
+    {
+      "epoch": 1.3385299791346799,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.102478042088315e-06,
+      "logits/chosen": -0.49223145842552185,
+      "logits/rejected": -0.35867035388946533,
+      "logps/chosen": -202.394775390625,
+      "logps/rejected": -171.31448364257812,
+      "loss": 0.6207,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.725380539894104,
+      "rewards/margins": 0.2017165720462799,
+      "rewards/rejected": 0.5236639380455017,
+      "step": 4250
+    },
+    {
+      "epoch": 1.3416794614385261,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.098428637775183e-06,
+      "logits/chosen": -0.5037646293640137,
+      "logits/rejected": -0.46415749192237854,
+      "logps/chosen": -181.4263916015625,
+      "logps/rejected": -168.53955078125,
+      "loss": 0.674,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.6651412844657898,
+      "rewards/margins": 0.08603055775165558,
+      "rewards/rejected": 0.5791107416152954,
+      "step": 4260
+    },
+    {
+      "epoch": 1.3448289437423724,
+      "grad_norm": 2.3125,
+      "learning_rate": 4.094372127542285e-06,
+      "logits/chosen": -0.44330787658691406,
+      "logits/rejected": -0.3730103075504303,
+      "logps/chosen": -190.1846923828125,
+      "logps/rejected": -176.09255981445312,
+      "loss": 0.6639,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.6509902477264404,
+      "rewards/margins": 0.11924519389867783,
+      "rewards/rejected": 0.5317450761795044,
+      "step": 4270
+    },
+    {
+      "epoch": 1.3479784260462186,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.0903085294231035e-06,
+      "logits/chosen": -0.4798402190208435,
+      "logits/rejected": -0.27211225032806396,
+      "logps/chosen": -221.6403045654297,
+      "logps/rejected": -164.9480743408203,
+      "loss": 0.5944,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.7979739308357239,
+      "rewards/margins": 0.26428526639938354,
+      "rewards/rejected": 0.5336886644363403,
+      "step": 4280
+    },
+    {
+      "epoch": 1.351127908350065,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.086237861482632e-06,
+      "logits/chosen": -0.45586076378822327,
+      "logits/rejected": -0.37053224444389343,
+      "logps/chosen": -212.56216430664062,
+      "logps/rejected": -183.00344848632812,
+      "loss": 0.6408,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.7277430295944214,
+      "rewards/margins": 0.1431419998407364,
+      "rewards/rejected": 0.5846010446548462,
+      "step": 4290
+    },
+    {
+      "epoch": 1.3542773906539112,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.0821601418172926e-06,
+      "logits/chosen": -0.4372914731502533,
+      "logits/rejected": -0.362470418214798,
+      "logps/chosen": -194.07884216308594,
+      "logps/rejected": -176.6485137939453,
+      "loss": 0.6393,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.6592785716056824,
+      "rewards/margins": 0.14654000103473663,
+      "rewards/rejected": 0.5127385854721069,
+      "step": 4300
+    },
+    {
+      "epoch": 1.3574268729577574,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.078075388554857e-06,
+      "logits/chosen": -0.4022819995880127,
+      "logits/rejected": -0.3265857398509979,
+      "logps/chosen": -197.20260620117188,
+      "logps/rejected": -163.478515625,
+      "loss": 0.6253,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.741704523563385,
+      "rewards/margins": 0.17545071244239807,
+      "rewards/rejected": 0.5662537217140198,
+      "step": 4310
+    },
+    {
+      "epoch": 1.360576355261604,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.0739836198543634e-06,
+      "logits/chosen": -0.5151673555374146,
+      "logits/rejected": -0.3456025719642639,
+      "logps/chosen": -217.0493927001953,
+      "logps/rejected": -183.4969024658203,
+      "loss": 0.619,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.7682887315750122,
+      "rewards/margins": 0.19945594668388367,
+      "rewards/rejected": 0.5688328146934509,
+      "step": 4320
+    },
+    {
+      "epoch": 1.3637258375654502,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.069884853906041e-06,
+      "logits/chosen": -0.5158268809318542,
+      "logits/rejected": -0.333678662776947,
+      "logps/chosen": -213.3292236328125,
+      "logps/rejected": -159.3480987548828,
+      "loss": 0.6005,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.801051914691925,
+      "rewards/margins": 0.23268017172813416,
+      "rewards/rejected": 0.5683717131614685,
+      "step": 4330
+    },
+    {
+      "epoch": 1.3668753198692964,
+      "grad_norm": 4.34375,
+      "learning_rate": 4.065779108931222e-06,
+      "logits/chosen": -0.4823933243751526,
+      "logits/rejected": -0.3266783654689789,
+      "logps/chosen": -199.21498107910156,
+      "logps/rejected": -166.51553344726562,
+      "loss": 0.6243,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.7152196764945984,
+      "rewards/margins": 0.1862681806087494,
+      "rewards/rejected": 0.5289515256881714,
+      "step": 4340
+    },
+    {
+      "epoch": 1.3700248021731427,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.0616664031822686e-06,
+      "logits/chosen": -0.43618011474609375,
+      "logits/rejected": -0.28981930017471313,
+      "logps/chosen": -192.85598754882812,
+      "logps/rejected": -157.85415649414062,
+      "loss": 0.6548,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6323038935661316,
+      "rewards/margins": 0.11133924871683121,
+      "rewards/rejected": 0.5209646821022034,
+      "step": 4350
+    },
+    {
+      "epoch": 1.3731742844769892,
+      "grad_norm": 2.5,
+      "learning_rate": 4.057546754942482e-06,
+      "logits/chosen": -0.5119596719741821,
+      "logits/rejected": -0.3471509516239166,
+      "logps/chosen": -196.14657592773438,
+      "logps/rejected": -154.673583984375,
+      "loss": 0.6247,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.6863905191421509,
+      "rewards/margins": 0.18415386974811554,
+      "rewards/rejected": 0.5022366642951965,
+      "step": 4360
+    },
+    {
+      "epoch": 1.3763237667808355,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.053420182526031e-06,
+      "logits/chosen": -0.4351939260959625,
+      "logits/rejected": -0.35209694504737854,
+      "logps/chosen": -194.9961700439453,
+      "logps/rejected": -161.25656127929688,
+      "loss": 0.6406,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.725360095500946,
+      "rewards/margins": 0.15019366145133972,
+      "rewards/rejected": 0.5751665234565735,
+      "step": 4370
+    },
+    {
+      "epoch": 1.3794732490846817,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.049286704277865e-06,
+      "logits/chosen": -0.4273145794868469,
+      "logits/rejected": -0.32500097155570984,
+      "logps/chosen": -198.8387908935547,
+      "logps/rejected": -177.99790954589844,
+      "loss": 0.6608,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6882764101028442,
+      "rewards/margins": 0.1112000122666359,
+      "rewards/rejected": 0.5770763158798218,
+      "step": 4380
+    },
+    {
+      "epoch": 1.382622731388528,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.045146338573634e-06,
+      "logits/chosen": -0.4316592216491699,
+      "logits/rejected": -0.33414119482040405,
+      "logps/chosen": -183.42718505859375,
+      "logps/rejected": -173.8665313720703,
+      "loss": 0.6702,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.6961274743080139,
+      "rewards/margins": 0.10534685850143433,
+      "rewards/rejected": 0.5907806158065796,
+      "step": 4390
+    },
+    {
+      "epoch": 1.3857722136923742,
+      "grad_norm": 3.125,
+      "learning_rate": 4.040999103819606e-06,
+      "logits/chosen": -0.5050019025802612,
+      "logits/rejected": -0.3844815194606781,
+      "logps/chosen": -205.3455352783203,
+      "logps/rejected": -164.32203674316406,
+      "loss": 0.6283,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.7454142570495605,
+      "rewards/margins": 0.17193658649921417,
+      "rewards/rejected": 0.5734776854515076,
+      "step": 4400
+    },
+    {
+      "epoch": 1.3889216959962205,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.036845018452586e-06,
+      "logits/chosen": -0.4365948736667633,
+      "logits/rejected": -0.3728296458721161,
+      "logps/chosen": -188.38137817382812,
+      "logps/rejected": -169.47296142578125,
+      "loss": 0.6519,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.65342777967453,
+      "rewards/margins": 0.1305466592311859,
+      "rewards/rejected": 0.5228811502456665,
+      "step": 4410
+    },
+    {
+      "epoch": 1.392071178300067,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.0326841009398354e-06,
+      "logits/chosen": -0.46066349744796753,
+      "logits/rejected": -0.38210171461105347,
+      "logps/chosen": -189.90225219726562,
+      "logps/rejected": -175.32156372070312,
+      "loss": 0.6905,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.6591070890426636,
+      "rewards/margins": 0.048674892634153366,
+      "rewards/rejected": 0.6104320883750916,
+      "step": 4420
+    },
+    {
+      "epoch": 1.3952206606039133,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.028516369778987e-06,
+      "logits/chosen": -0.43766292929649353,
+      "logits/rejected": -0.29349422454833984,
+      "logps/chosen": -188.58241271972656,
+      "logps/rejected": -164.57627868652344,
+      "loss": 0.655,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.6774150729179382,
+      "rewards/margins": 0.10568390041589737,
+      "rewards/rejected": 0.5717312693595886,
+      "step": 4430
+    },
+    {
+      "epoch": 1.3983701429077595,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.0243418434979605e-06,
+      "logits/chosen": -0.43426451086997986,
+      "logits/rejected": -0.3612423539161682,
+      "logps/chosen": -201.9038543701172,
+      "logps/rejected": -177.84640502929688,
+      "loss": 0.6486,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.7348337173461914,
+      "rewards/margins": 0.1346462517976761,
+      "rewards/rejected": 0.6001874208450317,
+      "step": 4440
+    },
+    {
+      "epoch": 1.4015196252116058,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.020160540654892e-06,
+      "logits/chosen": -0.42761626839637756,
+      "logits/rejected": -0.31236857175827026,
+      "logps/chosen": -192.8739471435547,
+      "logps/rejected": -171.3946990966797,
+      "loss": 0.6802,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.6831879615783691,
+      "rewards/margins": 0.06777564436197281,
+      "rewards/rejected": 0.6154123544692993,
+      "step": 4450
+    },
+    {
+      "epoch": 1.4046691075154523,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.015972479838035e-06,
+      "logits/chosen": -0.4452191889286041,
+      "logits/rejected": -0.3907052278518677,
+      "logps/chosen": -210.70761108398438,
+      "logps/rejected": -191.74087524414062,
+      "loss": 0.6637,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.7690273523330688,
+      "rewards/margins": 0.09637213498353958,
+      "rewards/rejected": 0.6726552248001099,
+      "step": 4460
+    },
+    {
+      "epoch": 1.4078185898192985,
+      "grad_norm": 2.515625,
+      "learning_rate": 4.011777679665693e-06,
+      "logits/chosen": -0.5560199022293091,
+      "logits/rejected": -0.4294883608818054,
+      "logps/chosen": -188.9731903076172,
+      "logps/rejected": -155.7578125,
+      "loss": 0.6421,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.6779214143753052,
+      "rewards/margins": 0.14832261204719543,
+      "rewards/rejected": 0.5295988321304321,
+      "step": 4470
+    },
+    {
+      "epoch": 1.4109680721231448,
+      "grad_norm": 2.75,
+      "learning_rate": 4.007576158786123e-06,
+      "logits/chosen": -0.5267634987831116,
+      "logits/rejected": -0.4555323123931885,
+      "logps/chosen": -215.6820526123047,
+      "logps/rejected": -179.27700805664062,
+      "loss": 0.6235,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.7598739266395569,
+      "rewards/margins": 0.1784120500087738,
+      "rewards/rejected": 0.5814618468284607,
+      "step": 4480
+    },
+    {
+      "epoch": 1.414117554426991,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.003367935877466e-06,
+      "logits/chosen": -0.49042144417762756,
+      "logits/rejected": -0.40049901604652405,
+      "logps/chosen": -205.4383544921875,
+      "logps/rejected": -174.7517547607422,
+      "loss": 0.6269,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.725433349609375,
+      "rewards/margins": 0.18734149634838104,
+      "rewards/rejected": 0.538091778755188,
+      "step": 4490
+    },
+    {
+      "epoch": 1.4172670367308373,
+      "grad_norm": 2.484375,
+      "learning_rate": 3.999153029647651e-06,
+      "logits/chosen": -0.43651509284973145,
+      "logits/rejected": -0.32142752408981323,
+      "logps/chosen": -216.54519653320312,
+      "logps/rejected": -169.64065551757812,
+      "loss": 0.6045,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.8090263605117798,
+      "rewards/margins": 0.24071991443634033,
+      "rewards/rejected": 0.5683062076568604,
+      "step": 4500
+    },
+    {
+      "epoch": 1.4204165190346836,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.994931458834323e-06,
+      "logits/chosen": -0.4635355472564697,
+      "logits/rejected": -0.374891459941864,
+      "logps/chosen": -176.76364135742188,
+      "logps/rejected": -150.61795043945312,
+      "loss": 0.6502,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.6052500605583191,
+      "rewards/margins": 0.12052911520004272,
+      "rewards/rejected": 0.48472094535827637,
+      "step": 4510
+    },
+    {
+      "epoch": 1.4235660013385298,
+      "grad_norm": 3.484375,
+      "learning_rate": 3.990703242204754e-06,
+      "logits/chosen": -0.4779212474822998,
+      "logits/rejected": -0.3155384659767151,
+      "logps/chosen": -203.7576904296875,
+      "logps/rejected": -177.79849243164062,
+      "loss": 0.6564,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.805811882019043,
+      "rewards/margins": 0.1246214359998703,
+      "rewards/rejected": 0.6811904907226562,
+      "step": 4520
+    },
+    {
+      "epoch": 1.4267154836423763,
+      "grad_norm": 2.53125,
+      "learning_rate": 3.986468398555758e-06,
+      "logits/chosen": -0.4679221212863922,
+      "logits/rejected": -0.3618343472480774,
+      "logps/chosen": -198.53536987304688,
+      "logps/rejected": -171.15133666992188,
+      "loss": 0.6479,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.7284259796142578,
+      "rewards/margins": 0.14506401121616364,
+      "rewards/rejected": 0.5833619832992554,
+      "step": 4530
+    },
+    {
+      "epoch": 1.4298649659462226,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.98222694671361e-06,
+      "logits/chosen": -0.45645904541015625,
+      "logits/rejected": -0.36011576652526855,
+      "logps/chosen": -218.33425903320312,
+      "logps/rejected": -187.9898681640625,
+      "loss": 0.6464,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.8267934918403625,
+      "rewards/margins": 0.15051567554473877,
+      "rewards/rejected": 0.6762778162956238,
+      "step": 4540
+    },
+    {
+      "epoch": 1.4330144482500689,
+      "grad_norm": 3.453125,
+      "learning_rate": 3.977978905533966e-06,
+      "logits/chosen": -0.49182215332984924,
+      "logits/rejected": -0.40443873405456543,
+      "logps/chosen": -203.3308868408203,
+      "logps/rejected": -182.2737274169922,
+      "loss": 0.6645,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.6901066303253174,
+      "rewards/margins": 0.09435725957155228,
+      "rewards/rejected": 0.5957493185997009,
+      "step": 4550
+    },
+    {
+      "epoch": 1.4361639305539153,
+      "grad_norm": 2.125,
+      "learning_rate": 3.973724293901772e-06,
+      "logits/chosen": -0.4056168496608734,
+      "logits/rejected": -0.38391467928886414,
+      "logps/chosen": -195.6358184814453,
+      "logps/rejected": -183.57254028320312,
+      "loss": 0.6455,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.682805061340332,
+      "rewards/margins": 0.12637227773666382,
+      "rewards/rejected": 0.5564327836036682,
+      "step": 4560
+    },
+    {
+      "epoch": 1.4393134128577616,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.969463130731183e-06,
+      "logits/chosen": -0.4418698847293854,
+      "logits/rejected": -0.39611127972602844,
+      "logps/chosen": -185.42196655273438,
+      "logps/rejected": -161.41970825195312,
+      "loss": 0.6608,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.6986794471740723,
+      "rewards/margins": 0.10915372520685196,
+      "rewards/rejected": 0.5895256996154785,
+      "step": 4570
+    },
+    {
+      "epoch": 1.4424628951616079,
+      "grad_norm": 2.28125,
+      "learning_rate": 3.965195434965482e-06,
+      "logits/chosen": -0.49892979860305786,
+      "logits/rejected": -0.3777271807193756,
+      "logps/chosen": -214.09317016601562,
+      "logps/rejected": -188.13845825195312,
+      "loss": 0.6567,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.7749571800231934,
+      "rewards/margins": 0.1210152879357338,
+      "rewards/rejected": 0.6539419889450073,
+      "step": 4580
+    },
+    {
+      "epoch": 1.4456123774654541,
+      "grad_norm": 3.453125,
+      "learning_rate": 3.960921225576991e-06,
+      "logits/chosen": -0.5049037933349609,
+      "logits/rejected": -0.34426528215408325,
+      "logps/chosen": -215.4214630126953,
+      "logps/rejected": -190.46994018554688,
+      "loss": 0.6566,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.7792149782180786,
+      "rewards/margins": 0.11181477457284927,
+      "rewards/rejected": 0.6674002408981323,
+      "step": 4590
+    },
+    {
+      "epoch": 1.4487618597693004,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.956640521566989e-06,
+      "logits/chosen": -0.48397397994995117,
+      "logits/rejected": -0.4080958962440491,
+      "logps/chosen": -187.16818237304688,
+      "logps/rejected": -166.5101318359375,
+      "loss": 0.6367,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.7400224804878235,
+      "rewards/margins": 0.15180279314517975,
+      "rewards/rejected": 0.5882197022438049,
+      "step": 4600
+    },
+    {
+      "epoch": 1.4519113420731466,
+      "grad_norm": 2.75,
+      "learning_rate": 3.952353341965628e-06,
+      "logits/chosen": -0.44713473320007324,
+      "logits/rejected": -0.3727934956550598,
+      "logps/chosen": -202.4422607421875,
+      "logps/rejected": -174.40975952148438,
+      "loss": 0.6574,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6708841323852539,
+      "rewards/margins": 0.10317204892635345,
+      "rewards/rejected": 0.5677120089530945,
+      "step": 4610
+    },
+    {
+      "epoch": 1.455060824376993,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.948059705831847e-06,
+      "logits/chosen": -0.4646250605583191,
+      "logits/rejected": -0.34977811574935913,
+      "logps/chosen": -175.93081665039062,
+      "logps/rejected": -161.12667846679688,
+      "loss": 0.6457,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.6609789729118347,
+      "rewards/margins": 0.13796725869178772,
+      "rewards/rejected": 0.5230117440223694,
+      "step": 4620
+    },
+    {
+      "epoch": 1.4582103066808394,
+      "grad_norm": 2.90625,
+      "learning_rate": 3.943759632253289e-06,
+      "logits/chosen": -0.44996100664138794,
+      "logits/rejected": -0.3352479934692383,
+      "logps/chosen": -191.62132263183594,
+      "logps/rejected": -161.776611328125,
+      "loss": 0.6196,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.7225370407104492,
+      "rewards/margins": 0.20392043888568878,
+      "rewards/rejected": 0.5186166167259216,
+      "step": 4630
+    },
+    {
+      "epoch": 1.4613597889846857,
+      "grad_norm": 2.421875,
+      "learning_rate": 3.939453140346212e-06,
+      "logits/chosen": -0.4620143473148346,
+      "logits/rejected": -0.2916569113731384,
+      "logps/chosen": -198.8416748046875,
+      "logps/rejected": -158.47796630859375,
+      "loss": 0.6241,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6989172697067261,
+      "rewards/margins": 0.20297345519065857,
+      "rewards/rejected": 0.4959437847137451,
+      "step": 4640
+    },
+    {
+      "epoch": 1.464509271288532,
+      "grad_norm": 2.4375,
+      "learning_rate": 3.935140249255412e-06,
+      "logits/chosen": -0.4222196638584137,
+      "logits/rejected": -0.3922198414802551,
+      "logps/chosen": -191.1973114013672,
+      "logps/rejected": -182.5751495361328,
+      "loss": 0.7209,
+      "rewards/accuracies": 0.4749999940395355,
+      "rewards/chosen": 0.6637114882469177,
+      "rewards/margins": 0.014695653691887856,
+      "rewards/rejected": 0.6490157842636108,
+      "step": 4650
+    },
+    {
+      "epoch": 1.4676587535923782,
+      "grad_norm": 4.71875,
+      "learning_rate": 3.930820978154129e-06,
+      "logits/chosen": -0.4839719831943512,
+      "logits/rejected": -0.36961930990219116,
+      "logps/chosen": -199.48463439941406,
+      "logps/rejected": -167.02774047851562,
+      "loss": 0.6264,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.7042496204376221,
+      "rewards/margins": 0.18996970355510712,
+      "rewards/rejected": 0.5142799615859985,
+      "step": 4660
+    },
+    {
+      "epoch": 1.4708082358962247,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.926495346243967e-06,
+      "logits/chosen": -0.5004483461380005,
+      "logits/rejected": -0.40153947472572327,
+      "logps/chosen": -210.6992645263672,
+      "logps/rejected": -190.9122314453125,
+      "loss": 0.6394,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.7591875791549683,
+      "rewards/margins": 0.1497870534658432,
+      "rewards/rejected": 0.6094005107879639,
+      "step": 4670
+    },
+    {
+      "epoch": 1.473957718200071,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.922163372754807e-06,
+      "logits/chosen": -0.41180485486984253,
+      "logits/rejected": -0.31235820055007935,
+      "logps/chosen": -203.12744140625,
+      "logps/rejected": -167.7096710205078,
+      "loss": 0.6562,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.6816965937614441,
+      "rewards/margins": 0.12079987674951553,
+      "rewards/rejected": 0.5608968138694763,
+      "step": 4680
+    },
+    {
+      "epoch": 1.4771072005039172,
+      "grad_norm": 1.78125,
+      "learning_rate": 3.9178250769447245e-06,
+      "logits/chosen": -0.4602780342102051,
+      "logits/rejected": -0.352649986743927,
+      "logps/chosen": -178.3946075439453,
+      "logps/rejected": -165.46885681152344,
+      "loss": 0.6532,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.6400049924850464,
+      "rewards/margins": 0.12722846865653992,
+      "rewards/rejected": 0.5127764940261841,
+      "step": 4690
+    },
+    {
+      "epoch": 1.4802566828077635,
+      "grad_norm": 2.359375,
+      "learning_rate": 3.913480478099898e-06,
+      "logits/chosen": -0.4489854872226715,
+      "logits/rejected": -0.3600943684577942,
+      "logps/chosen": -215.1140594482422,
+      "logps/rejected": -183.04026794433594,
+      "loss": 0.6386,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.7823598980903625,
+      "rewards/margins": 0.1600048840045929,
+      "rewards/rejected": 0.6223549842834473,
+      "step": 4700
+    },
+    {
+      "epoch": 1.4834061651116097,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.909129595534527e-06,
+      "logits/chosen": -0.4606572091579437,
+      "logits/rejected": -0.3265830874443054,
+      "logps/chosen": -197.79678344726562,
+      "logps/rejected": -165.95912170410156,
+      "loss": 0.6541,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.6801806688308716,
+      "rewards/margins": 0.11523494869470596,
+      "rewards/rejected": 0.5649456977844238,
+      "step": 4710
+    },
+    {
+      "epoch": 1.486555647415456,
+      "grad_norm": 3.625,
+      "learning_rate": 3.904772448590747e-06,
+      "logits/chosen": -0.46349745988845825,
+      "logits/rejected": -0.4300110936164856,
+      "logps/chosen": -190.7307891845703,
+      "logps/rejected": -179.4709014892578,
+      "loss": 0.6772,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.6604156494140625,
+      "rewards/margins": 0.06815408170223236,
+      "rewards/rejected": 0.5922616124153137,
+      "step": 4720
+    },
+    {
+      "epoch": 1.4897051297193025,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.900409056638542e-06,
+      "logits/chosen": -0.4786251485347748,
+      "logits/rejected": -0.3198348581790924,
+      "logps/chosen": -210.8201904296875,
+      "logps/rejected": -178.6912078857422,
+      "loss": 0.6503,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.7678909301757812,
+      "rewards/margins": 0.12642471492290497,
+      "rewards/rejected": 0.6414662003517151,
+      "step": 4730
+    },
+    {
+      "epoch": 1.4928546120231487,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.896039439075659e-06,
+      "logits/chosen": -0.5011542439460754,
+      "logits/rejected": -0.36163219809532166,
+      "logps/chosen": -197.447021484375,
+      "logps/rejected": -147.70782470703125,
+      "loss": 0.629,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6751397848129272,
+      "rewards/margins": 0.1756085753440857,
+      "rewards/rejected": 0.49953117966651917,
+      "step": 4740
+    },
+    {
+      "epoch": 1.496004094326995,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.891663615327518e-06,
+      "logits/chosen": -0.47268587350845337,
+      "logits/rejected": -0.3960368037223816,
+      "logps/chosen": -188.75637817382812,
+      "logps/rejected": -167.4852752685547,
+      "loss": 0.6569,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.6986013054847717,
+      "rewards/margins": 0.12062957137823105,
+      "rewards/rejected": 0.5779717564582825,
+      "step": 4750
+    },
+    {
+      "epoch": 1.4991535766308413,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.887281604847134e-06,
+      "logits/chosen": -0.510342001914978,
+      "logits/rejected": -0.36850231885910034,
+      "logps/chosen": -192.31849670410156,
+      "logps/rejected": -166.56289672851562,
+      "loss": 0.64,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.6647813320159912,
+      "rewards/margins": 0.15008948743343353,
+      "rewards/rejected": 0.5146918296813965,
+      "step": 4760
+    },
+    {
+      "epoch": 1.5023030589346877,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.8828934271150225e-06,
+      "logits/chosen": -0.460064172744751,
+      "logits/rejected": -0.4005635678768158,
+      "logps/chosen": -222.4634246826172,
+      "logps/rejected": -201.12998962402344,
+      "loss": 0.6611,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.7676762938499451,
+      "rewards/margins": 0.10998652130365372,
+      "rewards/rejected": 0.6576897501945496,
+      "step": 4770
+    },
+    {
+      "epoch": 1.505452541238534,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.878499101639116e-06,
+      "logits/chosen": -0.5471469759941101,
+      "logits/rejected": -0.4315733313560486,
+      "logps/chosen": -210.2429962158203,
+      "logps/rejected": -175.91360473632812,
+      "loss": 0.6623,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.7174965143203735,
+      "rewards/margins": 0.10746297985315323,
+      "rewards/rejected": 0.6100335717201233,
+      "step": 4780
+    },
+    {
+      "epoch": 1.5086020235423803,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.8740986479546796e-06,
+      "logits/chosen": -0.45277899503707886,
+      "logits/rejected": -0.4165739119052887,
+      "logps/chosen": -194.2645263671875,
+      "logps/rejected": -179.887939453125,
+      "loss": 0.6491,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.7052034139633179,
+      "rewards/margins": 0.1254056990146637,
+      "rewards/rejected": 0.5797977447509766,
+      "step": 4790
+    },
+    {
+      "epoch": 1.5117515058462265,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.869692085624218e-06,
+      "logits/chosen": -0.45139655470848083,
+      "logits/rejected": -0.3898164629936218,
+      "logps/chosen": -209.1693572998047,
+      "logps/rejected": -194.74021911621094,
+      "loss": 0.6479,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.7721225619316101,
+      "rewards/margins": 0.12638486921787262,
+      "rewards/rejected": 0.6457376480102539,
+      "step": 4800
+    },
+    {
+      "epoch": 1.5149009881500728,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.865279434237394e-06,
+      "logits/chosen": -0.4812788963317871,
+      "logits/rejected": -0.39241549372673035,
+      "logps/chosen": -190.20193481445312,
+      "logps/rejected": -166.25454711914062,
+      "loss": 0.6459,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.6281577944755554,
+      "rewards/margins": 0.12559270858764648,
+      "rewards/rejected": 0.5025650858879089,
+      "step": 4810
+    },
+    {
+      "epoch": 1.518050470453919,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.860860713410941e-06,
+      "logits/chosen": -0.38653573393821716,
+      "logits/rejected": -0.361195832490921,
+      "logps/chosen": -204.06520080566406,
+      "logps/rejected": -209.9178009033203,
+      "loss": 0.6876,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.7754232287406921,
+      "rewards/margins": 0.045936040580272675,
+      "rewards/rejected": 0.7294871211051941,
+      "step": 4820
+    },
+    {
+      "epoch": 1.5211999527577653,
+      "grad_norm": 2.453125,
+      "learning_rate": 3.8564359427885735e-06,
+      "logits/chosen": -0.469553142786026,
+      "logits/rejected": -0.41318267583847046,
+      "logps/chosen": -207.17117309570312,
+      "logps/rejected": -177.3662872314453,
+      "loss": 0.6248,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.7327437400817871,
+      "rewards/margins": 0.1847001016139984,
+      "rewards/rejected": 0.5480436086654663,
+      "step": 4830
+    },
+    {
+      "epoch": 1.5243494350616118,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.852005142040901e-06,
+      "logits/chosen": -0.4887300133705139,
+      "logits/rejected": -0.38054460287094116,
+      "logps/chosen": -189.93447875976562,
+      "logps/rejected": -159.81796264648438,
+      "loss": 0.635,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6666679978370667,
+      "rewards/margins": 0.1536942720413208,
+      "rewards/rejected": 0.5129736661911011,
+      "step": 4840
+    },
+    {
+      "epoch": 1.527498917365458,
+      "grad_norm": 2.453125,
+      "learning_rate": 3.8475683308653385e-06,
+      "logits/chosen": -0.4928087294101715,
+      "logits/rejected": -0.3735254406929016,
+      "logps/chosen": -204.97787475585938,
+      "logps/rejected": -168.03756713867188,
+      "loss": 0.6341,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.772538423538208,
+      "rewards/margins": 0.1797189563512802,
+      "rewards/rejected": 0.5928195118904114,
+      "step": 4850
+    },
+    {
+      "epoch": 1.5306483996693043,
+      "grad_norm": 2.375,
+      "learning_rate": 3.8431255289860225e-06,
+      "logits/chosen": -0.47978100180625916,
+      "logits/rejected": -0.3051653504371643,
+      "logps/chosen": -217.0726776123047,
+      "logps/rejected": -169.8887939453125,
+      "loss": 0.5989,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": 0.799135148525238,
+      "rewards/margins": 0.24375374615192413,
+      "rewards/rejected": 0.5553814172744751,
+      "step": 4860
+    },
+    {
+      "epoch": 1.5337978819731508,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.838676756153723e-06,
+      "logits/chosen": -0.4916785657405853,
+      "logits/rejected": -0.4126752018928528,
+      "logps/chosen": -204.36715698242188,
+      "logps/rejected": -171.4047088623047,
+      "loss": 0.6373,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.7680164575576782,
+      "rewards/margins": 0.1488504558801651,
+      "rewards/rejected": 0.6191660761833191,
+      "step": 4870
+    },
+    {
+      "epoch": 1.536947364276997,
+      "grad_norm": 2.328125,
+      "learning_rate": 3.834222032145751e-06,
+      "logits/chosen": -0.4622599184513092,
+      "logits/rejected": -0.3681351840496063,
+      "logps/chosen": -192.53379821777344,
+      "logps/rejected": -161.97409057617188,
+      "loss": 0.6518,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.6855158805847168,
+      "rewards/margins": 0.11872188746929169,
+      "rewards/rejected": 0.5667939782142639,
+      "step": 4880
+    },
+    {
+      "epoch": 1.5400968465808433,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.829761376765875e-06,
+      "logits/chosen": -0.5002883076667786,
+      "logits/rejected": -0.37738004326820374,
+      "logps/chosen": -208.83547973632812,
+      "logps/rejected": -180.82960510253906,
+      "loss": 0.6302,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.7079477906227112,
+      "rewards/margins": 0.172250896692276,
+      "rewards/rejected": 0.5356968641281128,
+      "step": 4890
+    },
+    {
+      "epoch": 1.5432463288846896,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.825294809844234e-06,
+      "logits/chosen": -0.4677095413208008,
+      "logits/rejected": -0.43684762716293335,
+      "logps/chosen": -204.42538452148438,
+      "logps/rejected": -189.5218505859375,
+      "loss": 0.6914,
+      "rewards/accuracies": 0.4749999940395355,
+      "rewards/chosen": 0.697405219078064,
+      "rewards/margins": 0.04717016965150833,
+      "rewards/rejected": 0.6502350568771362,
+      "step": 4900
+    },
+    {
+      "epoch": 1.5463958111885359,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.820822351237245e-06,
+      "logits/chosen": -0.44419798254966736,
+      "logits/rejected": -0.3508889377117157,
+      "logps/chosen": -192.20458984375,
+      "logps/rejected": -182.95726013183594,
+      "loss": 0.665,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.6904107928276062,
+      "rewards/margins": 0.09415493160486221,
+      "rewards/rejected": 0.596255898475647,
+      "step": 4910
+    },
+    {
+      "epoch": 1.5495452934923821,
+      "grad_norm": 2.59375,
+      "learning_rate": 3.816344020827516e-06,
+      "logits/chosen": -0.5454775094985962,
+      "logits/rejected": -0.39591822028160095,
+      "logps/chosen": -196.39208984375,
+      "logps/rejected": -159.05348205566406,
+      "loss": 0.6543,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.6435393691062927,
+      "rewards/margins": 0.11705289781093597,
+      "rewards/rejected": 0.5264865159988403,
+      "step": 4920
+    },
+    {
+      "epoch": 1.5526947757962284,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.8118598385237604e-06,
+      "logits/chosen": -0.439689576625824,
+      "logits/rejected": -0.4039355218410492,
+      "logps/chosen": -186.1417236328125,
+      "logps/rejected": -170.55740356445312,
+      "loss": 0.6797,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.6710829138755798,
+      "rewards/margins": 0.06706185638904572,
+      "rewards/rejected": 0.6040210127830505,
+      "step": 4930
+    },
+    {
+      "epoch": 1.5558442581000747,
+      "grad_norm": 2.28125,
+      "learning_rate": 3.807369824260706e-06,
+      "logits/chosen": -0.45275792479515076,
+      "logits/rejected": -0.3387320935726166,
+      "logps/chosen": -205.11349487304688,
+      "logps/rejected": -165.2199249267578,
+      "loss": 0.626,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.7621325850486755,
+      "rewards/margins": 0.1830775886774063,
+      "rewards/rejected": 0.5790549516677856,
+      "step": 4940
+    },
+    {
+      "epoch": 1.5589937404039211,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.8028739979990072e-06,
+      "logits/chosen": -0.4740076959133148,
+      "logits/rejected": -0.36817610263824463,
+      "logps/chosen": -217.1376190185547,
+      "logps/rejected": -184.6923065185547,
+      "loss": 0.617,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.7812771797180176,
+      "rewards/margins": 0.19556982815265656,
+      "rewards/rejected": 0.5857073068618774,
+      "step": 4950
+    },
+    {
+      "epoch": 1.5621432227077674,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.798372379725155e-06,
+      "logits/chosen": -0.47232404351234436,
+      "logits/rejected": -0.30332669615745544,
+      "logps/chosen": -179.11349487304688,
+      "logps/rejected": -152.26527404785156,
+      "loss": 0.6515,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.6055221557617188,
+      "rewards/margins": 0.1263987123966217,
+      "rewards/rejected": 0.47912344336509705,
+      "step": 4960
+    },
+    {
+      "epoch": 1.5652927050116137,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.79386498945139e-06,
+      "logits/chosen": -0.512363612651825,
+      "logits/rejected": -0.3525320291519165,
+      "logps/chosen": -199.22694396972656,
+      "logps/rejected": -158.5713348388672,
+      "loss": 0.6495,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.7140517234802246,
+      "rewards/margins": 0.13541573286056519,
+      "rewards/rejected": 0.5786360502243042,
+      "step": 4970
+    },
+    {
+      "epoch": 1.5684421873154601,
+      "grad_norm": 3.125,
+      "learning_rate": 3.789351847215613e-06,
+      "logits/chosen": -0.4590677320957184,
+      "logits/rejected": -0.2966582179069519,
+      "logps/chosen": -211.01083374023438,
+      "logps/rejected": -171.15493774414062,
+      "loss": 0.6124,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.7449513673782349,
+      "rewards/margins": 0.21209721267223358,
+      "rewards/rejected": 0.5328541994094849,
+      "step": 4980
+    },
+    {
+      "epoch": 1.5715916696193064,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.784832973081295e-06,
+      "logits/chosen": -0.4136572778224945,
+      "logits/rejected": -0.3735829293727875,
+      "logps/chosen": -190.76473999023438,
+      "logps/rejected": -176.0132293701172,
+      "loss": 0.6683,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.7110625505447388,
+      "rewards/margins": 0.08584681898355484,
+      "rewards/rejected": 0.6252157688140869,
+      "step": 4990
+    },
+    {
+      "epoch": 1.5747411519231527,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.7803083871373876e-06,
+      "logits/chosen": -0.42973631620407104,
+      "logits/rejected": -0.2951328754425049,
+      "logps/chosen": -206.06332397460938,
+      "logps/rejected": -174.61618041992188,
+      "loss": 0.6725,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.7120364308357239,
+      "rewards/margins": 0.0878576785326004,
+      "rewards/rejected": 0.6241787672042847,
+      "step": 5000
+    },
+    {
+      "epoch": 1.577890634226999,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.775778109498237e-06,
+      "logits/chosen": -0.391795814037323,
+      "logits/rejected": -0.2272360622882843,
+      "logps/chosen": -214.9050750732422,
+      "logps/rejected": -175.29550170898438,
+      "loss": 0.6073,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.7749017477035522,
+      "rewards/margins": 0.23047152161598206,
+      "rewards/rejected": 0.544430136680603,
+      "step": 5010
+    },
+    {
+      "epoch": 1.5810401165308452,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.7712421603034894e-06,
+      "logits/chosen": -0.43912237882614136,
+      "logits/rejected": -0.34138351678848267,
+      "logps/chosen": -204.3086395263672,
+      "logps/rejected": -182.76370239257812,
+      "loss": 0.7119,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": 0.6654126048088074,
+      "rewards/margins": 0.0014205619227141142,
+      "rewards/rejected": 0.6639919877052307,
+      "step": 5020
+    },
+    {
+      "epoch": 1.5841895988346915,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.766700559718006e-06,
+      "logits/chosen": -0.5014868974685669,
+      "logits/rejected": -0.4088328778743744,
+      "logps/chosen": -185.99476623535156,
+      "logps/rejected": -165.3468017578125,
+      "loss": 0.6596,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.6833370923995972,
+      "rewards/margins": 0.10189273208379745,
+      "rewards/rejected": 0.5814443826675415,
+      "step": 5030
+    },
+    {
+      "epoch": 1.5873390811385377,
+      "grad_norm": 2.390625,
+      "learning_rate": 3.762153327931772e-06,
+      "logits/chosen": -0.44292283058166504,
+      "logits/rejected": -0.3582010269165039,
+      "logps/chosen": -201.4560546875,
+      "logps/rejected": -181.69656372070312,
+      "loss": 0.6431,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.720178484916687,
+      "rewards/margins": 0.14207497239112854,
+      "rewards/rejected": 0.5781034231185913,
+      "step": 5040
+    },
+    {
+      "epoch": 1.5904885634423842,
+      "grad_norm": 3.375,
+      "learning_rate": 3.7576004851598052e-06,
+      "logits/chosen": -0.4733700156211853,
+      "logits/rejected": -0.3716031610965729,
+      "logps/chosen": -199.3379364013672,
+      "logps/rejected": -178.63943481445312,
+      "loss": 0.6621,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.75123131275177,
+      "rewards/margins": 0.10580404102802277,
+      "rewards/rejected": 0.6454272866249084,
+      "step": 5050
+    },
+    {
+      "epoch": 1.5936380457462305,
+      "grad_norm": 2.6875,
+      "learning_rate": 3.7530420516420676e-06,
+      "logits/chosen": -0.4269101023674011,
+      "logits/rejected": -0.37862318754196167,
+      "logps/chosen": -199.61886596679688,
+      "logps/rejected": -173.53884887695312,
+      "loss": 0.6456,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.7329145073890686,
+      "rewards/margins": 0.1574871987104416,
+      "rewards/rejected": 0.575427234172821,
+      "step": 5060
+    },
+    {
+      "epoch": 1.5967875280500767,
+      "grad_norm": 2.546875,
+      "learning_rate": 3.7484780476433764e-06,
+      "logits/chosen": -0.4474611282348633,
+      "logits/rejected": -0.35605159401893616,
+      "logps/chosen": -188.90847778320312,
+      "logps/rejected": -159.92965698242188,
+      "loss": 0.6346,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.6905893087387085,
+      "rewards/margins": 0.16502177715301514,
+      "rewards/rejected": 0.5255674123764038,
+      "step": 5070
+    },
+    {
+      "epoch": 1.5999370103539232,
+      "grad_norm": 2.75,
+      "learning_rate": 3.743908493453311e-06,
+      "logits/chosen": -0.43630313873291016,
+      "logits/rejected": -0.37972143292427063,
+      "logps/chosen": -223.48648071289062,
+      "logps/rejected": -194.86001586914062,
+      "loss": 0.6523,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.7133353352546692,
+      "rewards/margins": 0.11560231447219849,
+      "rewards/rejected": 0.5977329611778259,
+      "step": 5080
+    },
+    {
+      "epoch": 1.6030864926577695,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.739333409386126e-06,
+      "logits/chosen": -0.44998350739479065,
+      "logits/rejected": -0.3347667157649994,
+      "logps/chosen": -207.3821563720703,
+      "logps/rejected": -182.54971313476562,
+      "loss": 0.6532,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.7008602619171143,
+      "rewards/margins": 0.14106692373752594,
+      "rewards/rejected": 0.5597933530807495,
+      "step": 5090
+    },
+    {
+      "epoch": 1.6062359749616157,
+      "grad_norm": 2.25,
+      "learning_rate": 3.734752815780659e-06,
+      "logits/chosen": -0.514176070690155,
+      "logits/rejected": -0.3695998787879944,
+      "logps/chosen": -196.74203491210938,
+      "logps/rejected": -162.3852081298828,
+      "loss": 0.6246,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.7342701554298401,
+      "rewards/margins": 0.20258764922618866,
+      "rewards/rejected": 0.5316824913024902,
+      "step": 5100
+    },
+    {
+      "epoch": 1.609385457265462,
+      "grad_norm": 2.5,
+      "learning_rate": 3.7301667330002408e-06,
+      "logits/chosen": -0.439274400472641,
+      "logits/rejected": -0.3606041669845581,
+      "logps/chosen": -202.26397705078125,
+      "logps/rejected": -176.84947204589844,
+      "loss": 0.641,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.7874661684036255,
+      "rewards/margins": 0.1603676974773407,
+      "rewards/rejected": 0.6270985007286072,
+      "step": 5110
+    },
+    {
+      "epoch": 1.6125349395693083,
+      "grad_norm": 2.625,
+      "learning_rate": 3.7255751814326035e-06,
+      "logits/chosen": -0.4579714238643646,
+      "logits/rejected": -0.3923242390155792,
+      "logps/chosen": -189.69895935058594,
+      "logps/rejected": -181.8014373779297,
+      "loss": 0.6762,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.6748474836349487,
+      "rewards/margins": 0.07303015887737274,
+      "rewards/rejected": 0.6018173098564148,
+      "step": 5120
+    },
+    {
+      "epoch": 1.6156844218731545,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.720978181489792e-06,
+      "logits/chosen": -0.4986580014228821,
+      "logits/rejected": -0.4134606719017029,
+      "logps/chosen": -189.92276000976562,
+      "logps/rejected": -189.25656127929688,
+      "loss": 0.6926,
+      "rewards/accuracies": 0.4749999940395355,
+      "rewards/chosen": 0.7121160626411438,
+      "rewards/margins": 0.05977436155080795,
+      "rewards/rejected": 0.6523416042327881,
+      "step": 5130
+    },
+    {
+      "epoch": 1.6188339041770008,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.716375753608073e-06,
+      "logits/chosen": -0.45538026094436646,
+      "logits/rejected": -0.29467564821243286,
+      "logps/chosen": -217.4430694580078,
+      "logps/rejected": -182.09951782226562,
+      "loss": 0.6399,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.804099440574646,
+      "rewards/margins": 0.17206081748008728,
+      "rewards/rejected": 0.6320386528968811,
+      "step": 5140
+    },
+    {
+      "epoch": 1.621983386480847,
+      "grad_norm": 2.34375,
+      "learning_rate": 3.7117679182478415e-06,
+      "logits/chosen": -0.45015448331832886,
+      "logits/rejected": -0.3417063355445862,
+      "logps/chosen": -196.57412719726562,
+      "logps/rejected": -183.40467834472656,
+      "loss": 0.669,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.6573584675788879,
+      "rewards/margins": 0.0827714204788208,
+      "rewards/rejected": 0.5745870471000671,
+      "step": 5150
+    },
+    {
+      "epoch": 1.6251328687846935,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.707154695893535e-06,
+      "logits/chosen": -0.49352359771728516,
+      "logits/rejected": -0.3199608027935028,
+      "logps/chosen": -191.95172119140625,
+      "logps/rejected": -168.58969116210938,
+      "loss": 0.6372,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.743454098701477,
+      "rewards/margins": 0.16557399928569794,
+      "rewards/rejected": 0.5778801441192627,
+      "step": 5160
+    },
+    {
+      "epoch": 1.6282823510885398,
+      "grad_norm": 2.6875,
+      "learning_rate": 3.702536107053536e-06,
+      "logits/chosen": -0.402625173330307,
+      "logits/rejected": -0.3243730366230011,
+      "logps/chosen": -182.0989532470703,
+      "logps/rejected": -162.49978637695312,
+      "loss": 0.6791,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.6483687162399292,
+      "rewards/margins": 0.07241274416446686,
+      "rewards/rejected": 0.5759559869766235,
+      "step": 5170
+    },
+    {
+      "epoch": 1.6314318333923863,
+      "grad_norm": 2.5625,
+      "learning_rate": 3.697912172260085e-06,
+      "logits/chosen": -0.49576109647750854,
+      "logits/rejected": -0.39149078726768494,
+      "logps/chosen": -205.7623291015625,
+      "logps/rejected": -176.04811096191406,
+      "loss": 0.6461,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.7142454385757446,
+      "rewards/margins": 0.13653890788555145,
+      "rewards/rejected": 0.5777064561843872,
+      "step": 5180
+    },
+    {
+      "epoch": 1.6345813156962326,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.693282912069189e-06,
+      "logits/chosen": -0.4688878655433655,
+      "logits/rejected": -0.3151419162750244,
+      "logps/chosen": -222.38955688476562,
+      "logps/rejected": -187.0220489501953,
+      "loss": 0.6532,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.7705121636390686,
+      "rewards/margins": 0.11784696578979492,
+      "rewards/rejected": 0.6526652574539185,
+      "step": 5190
+    },
+    {
+      "epoch": 1.6377307980000788,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.6886483470605293e-06,
+      "logits/chosen": -0.4141194224357605,
+      "logits/rejected": -0.33433422446250916,
+      "logps/chosen": -192.88711547851562,
+      "logps/rejected": -161.22756958007812,
+      "loss": 0.6707,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.645660400390625,
+      "rewards/margins": 0.0918336808681488,
+      "rewards/rejected": 0.5538267493247986,
+      "step": 5200
+    },
+    {
+      "epoch": 1.640880280303925,
+      "grad_norm": 2.65625,
+      "learning_rate": 3.6840084978373704e-06,
+      "logits/chosen": -0.4529247283935547,
+      "logits/rejected": -0.23250219225883484,
+      "logps/chosen": -212.8038330078125,
+      "logps/rejected": -176.3964385986328,
+      "loss": 0.6297,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.7746697664260864,
+      "rewards/margins": 0.16910018026828766,
+      "rewards/rejected": 0.6055695414543152,
+      "step": 5210
+    },
+    {
+      "epoch": 1.6440297626077713,
+      "grad_norm": 2.328125,
+      "learning_rate": 3.6793633850264655e-06,
+      "logits/chosen": -0.4927656650543213,
+      "logits/rejected": -0.3135763704776764,
+      "logps/chosen": -213.3108673095703,
+      "logps/rejected": -167.95083618164062,
+      "loss": 0.6254,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.7371099591255188,
+      "rewards/margins": 0.19389715790748596,
+      "rewards/rejected": 0.5432127714157104,
+      "step": 5220
+    },
+    {
+      "epoch": 1.6471792449116176,
+      "grad_norm": 2.59375,
+      "learning_rate": 3.6747130292779715e-06,
+      "logits/chosen": -0.4238489270210266,
+      "logits/rejected": -0.3129653036594391,
+      "logps/chosen": -194.534912109375,
+      "logps/rejected": -174.8766632080078,
+      "loss": 0.6662,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.7266419529914856,
+      "rewards/margins": 0.09835498034954071,
+      "rewards/rejected": 0.6282869577407837,
+      "step": 5230
+    },
+    {
+      "epoch": 1.6503287272154639,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.6700574512653497e-06,
+      "logits/chosen": -0.4622649550437927,
+      "logits/rejected": -0.37696540355682373,
+      "logps/chosen": -199.86097717285156,
+      "logps/rejected": -181.8938446044922,
+      "loss": 0.6482,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.7182890176773071,
+      "rewards/margins": 0.12708035111427307,
+      "rewards/rejected": 0.5912087559700012,
+      "step": 5240
+    },
+    {
+      "epoch": 1.6534782095193101,
+      "grad_norm": 2.5,
+      "learning_rate": 3.66539667168528e-06,
+      "logits/chosen": -0.4936433732509613,
+      "logits/rejected": -0.3448127210140228,
+      "logps/chosen": -187.07192993164062,
+      "logps/rejected": -149.51683044433594,
+      "loss": 0.6323,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.6617879867553711,
+      "rewards/margins": 0.18173685669898987,
+      "rewards/rejected": 0.480051189661026,
+      "step": 5250
+    },
+    {
+      "epoch": 1.6566276918231566,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.6607307112575646e-06,
+      "logits/chosen": -0.5180322527885437,
+      "logits/rejected": -0.3746757507324219,
+      "logps/chosen": -201.63467407226562,
+      "logps/rejected": -166.1622314453125,
+      "loss": 0.5896,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.7969163656234741,
+      "rewards/margins": 0.26014357805252075,
+      "rewards/rejected": 0.5367728471755981,
+      "step": 5260
+    },
+    {
+      "epoch": 1.6597771741270029,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.6560595907250375e-06,
+      "logits/chosen": -0.3662557005882263,
+      "logits/rejected": -0.3333319425582886,
+      "logps/chosen": -191.13694763183594,
+      "logps/rejected": -189.7360382080078,
+      "loss": 0.6766,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.7387022972106934,
+      "rewards/margins": 0.08876340091228485,
+      "rewards/rejected": 0.6499389410018921,
+      "step": 5270
+    },
+    {
+      "epoch": 1.6629266564308491,
+      "grad_norm": 3.265625,
+      "learning_rate": 3.651383330853472e-06,
+      "logits/chosen": -0.4906235635280609,
+      "logits/rejected": -0.36935657262802124,
+      "logps/chosen": -217.0959014892578,
+      "logps/rejected": -180.4376220703125,
+      "loss": 0.6515,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.7411099076271057,
+      "rewards/margins": 0.1661885380744934,
+      "rewards/rejected": 0.5749213695526123,
+      "step": 5280
+    },
+    {
+      "epoch": 1.6660761387346956,
+      "grad_norm": 2.21875,
+      "learning_rate": 3.6467019524314905e-06,
+      "logits/chosen": -0.49627685546875,
+      "logits/rejected": -0.3914474844932556,
+      "logps/chosen": -189.37979125976562,
+      "logps/rejected": -160.41397094726562,
+      "loss": 0.6208,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.7144301533699036,
+      "rewards/margins": 0.19551965594291687,
+      "rewards/rejected": 0.5189104676246643,
+      "step": 5290
+    },
+    {
+      "epoch": 1.6692256210385419,
+      "grad_norm": 2.515625,
+      "learning_rate": 3.6420154762704685e-06,
+      "logits/chosen": -0.5125707983970642,
+      "logits/rejected": -0.3919418752193451,
+      "logps/chosen": -194.13682556152344,
+      "logps/rejected": -161.27906799316406,
+      "loss": 0.6436,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.6802747249603271,
+      "rewards/margins": 0.1379631608724594,
+      "rewards/rejected": 0.5423115491867065,
+      "step": 5300
+    },
+    {
+      "epoch": 1.6723751033423881,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.6373239232044445e-06,
+      "logits/chosen": -0.5417731404304504,
+      "logits/rejected": -0.3268618881702423,
+      "logps/chosen": -193.62232971191406,
+      "logps/rejected": -159.81134033203125,
+      "loss": 0.6213,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.6962756514549255,
+      "rewards/margins": 0.18256238102912903,
+      "rewards/rejected": 0.5137132406234741,
+      "step": 5310
+    },
+    {
+      "epoch": 1.6755245856462344,
+      "grad_norm": 2.359375,
+      "learning_rate": 3.632627314090026e-06,
+      "logits/chosen": -0.4504339098930359,
+      "logits/rejected": -0.38121432065963745,
+      "logps/chosen": -208.55001831054688,
+      "logps/rejected": -187.78811645507812,
+      "loss": 0.6453,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.7359059453010559,
+      "rewards/margins": 0.14206233620643616,
+      "rewards/rejected": 0.5938436388969421,
+      "step": 5320
+    },
+    {
+      "epoch": 1.6786740679500807,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.6279256698062986e-06,
+      "logits/chosen": -0.5013277530670166,
+      "logits/rejected": -0.33975881338119507,
+      "logps/chosen": -206.8636474609375,
+      "logps/rejected": -179.7592315673828,
+      "loss": 0.6196,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.7417197227478027,
+      "rewards/margins": 0.20961704850196838,
+      "rewards/rejected": 0.5321027040481567,
+      "step": 5330
+    },
+    {
+      "epoch": 1.681823550253927,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.6232190112547324e-06,
+      "logits/chosen": -0.5261915922164917,
+      "logits/rejected": -0.4154892861843109,
+      "logps/chosen": -196.72494506835938,
+      "logps/rejected": -167.34524536132812,
+      "loss": 0.6448,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.6392444968223572,
+      "rewards/margins": 0.13428232073783875,
+      "rewards/rejected": 0.5049622058868408,
+      "step": 5340
+    },
+    {
+      "epoch": 1.6849730325577732,
+      "grad_norm": 2.28125,
+      "learning_rate": 3.6185073593590868e-06,
+      "logits/chosen": -0.5182952880859375,
+      "logits/rejected": -0.32755130529403687,
+      "logps/chosen": -193.91754150390625,
+      "logps/rejected": -147.9218292236328,
+      "loss": 0.6283,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.6746249794960022,
+      "rewards/margins": 0.181734099984169,
+      "rewards/rejected": 0.4928908944129944,
+      "step": 5350
+    },
+    {
+      "epoch": 1.6881225148616195,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.613790735065321e-06,
+      "logits/chosen": -0.397797167301178,
+      "logits/rejected": -0.34379732608795166,
+      "logps/chosen": -196.30372619628906,
+      "logps/rejected": -173.0957489013672,
+      "loss": 0.653,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.7253559827804565,
+      "rewards/margins": 0.13321705162525177,
+      "rewards/rejected": 0.592138946056366,
+      "step": 5360
+    },
+    {
+      "epoch": 1.691271997165466,
+      "grad_norm": 2.46875,
+      "learning_rate": 3.6090691593414978e-06,
+      "logits/chosen": -0.49348416924476624,
+      "logits/rejected": -0.33678576350212097,
+      "logps/chosen": -201.0540313720703,
+      "logps/rejected": -170.12416076660156,
+      "loss": 0.611,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.7913371324539185,
+      "rewards/margins": 0.21138879656791687,
+      "rewards/rejected": 0.5799483060836792,
+      "step": 5370
+    },
+    {
+      "epoch": 1.6944214794693122,
+      "grad_norm": 2.546875,
+      "learning_rate": 3.604342653177695e-06,
+      "logits/chosen": -0.48641714453697205,
+      "logits/rejected": -0.40380674600601196,
+      "logps/chosen": -179.4705810546875,
+      "logps/rejected": -162.80154418945312,
+      "loss": 0.6501,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.6522781252861023,
+      "rewards/margins": 0.11866960674524307,
+      "rewards/rejected": 0.5336084961891174,
+      "step": 5380
+    },
+    {
+      "epoch": 1.6975709617731587,
+      "grad_norm": 2.4375,
+      "learning_rate": 3.599611237585906e-06,
+      "logits/chosen": -0.45630472898483276,
+      "logits/rejected": -0.3485426902770996,
+      "logps/chosen": -171.63662719726562,
+      "logps/rejected": -156.93072509765625,
+      "loss": 0.6609,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.5976822972297668,
+      "rewards/margins": 0.1124814972281456,
+      "rewards/rejected": 0.48520079255104065,
+      "step": 5390
+    },
+    {
+      "epoch": 1.700720444077005,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.5948749335999493e-06,
+      "logits/chosen": -0.5297515988349915,
+      "logits/rejected": -0.44055309891700745,
+      "logps/chosen": -192.01625061035156,
+      "logps/rejected": -184.41848754882812,
+      "loss": 0.6709,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.6303820610046387,
+      "rewards/margins": 0.07556124031543732,
+      "rewards/rejected": 0.5548208951950073,
+      "step": 5400
+    },
+    {
+      "epoch": 1.7038699263808512,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.590133762275378e-06,
+      "logits/chosen": -0.5689431428909302,
+      "logits/rejected": -0.3852362334728241,
+      "logps/chosen": -212.37899780273438,
+      "logps/rejected": -171.7864227294922,
+      "loss": 0.6305,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.7726091742515564,
+      "rewards/margins": 0.1817570924758911,
+      "rewards/rejected": 0.5908521413803101,
+      "step": 5410
+    },
+    {
+      "epoch": 1.7070194086846975,
+      "grad_norm": 2.4375,
+      "learning_rate": 3.5853877446893802e-06,
+      "logits/chosen": -0.496450811624527,
+      "logits/rejected": -0.39372554421424866,
+      "logps/chosen": -183.16973876953125,
+      "logps/rejected": -166.51968383789062,
+      "loss": 0.6613,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.6342931985855103,
+      "rewards/margins": 0.08918163925409317,
+      "rewards/rejected": 0.5451115369796753,
+      "step": 5420
+    },
+    {
+      "epoch": 1.7101688909885437,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.5806369019406906e-06,
+      "logits/chosen": -0.5108271837234497,
+      "logits/rejected": -0.33603325486183167,
+      "logps/chosen": -210.8227996826172,
+      "logps/rejected": -167.770263671875,
+      "loss": 0.6393,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.7098419666290283,
+      "rewards/margins": 0.17859862744808197,
+      "rewards/rejected": 0.5312432646751404,
+      "step": 5430
+    },
+    {
+      "epoch": 1.71331837329239,
+      "grad_norm": 2.90625,
+      "learning_rate": 3.5758812551494926e-06,
+      "logits/chosen": -0.4778234362602234,
+      "logits/rejected": -0.3700777590274811,
+      "logps/chosen": -218.7685546875,
+      "logps/rejected": -184.45120239257812,
+      "loss": 0.6412,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.798811674118042,
+      "rewards/margins": 0.14647333323955536,
+      "rewards/rejected": 0.6523382663726807,
+      "step": 5440
+    },
+    {
+      "epoch": 1.7164678555962363,
+      "grad_norm": 2.4375,
+      "learning_rate": 3.571120825457327e-06,
+      "logits/chosen": -0.4903620779514313,
+      "logits/rejected": -0.37475109100341797,
+      "logps/chosen": -196.15283203125,
+      "logps/rejected": -187.24850463867188,
+      "loss": 0.6469,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.6965973377227783,
+      "rewards/margins": 0.13229303061962128,
+      "rewards/rejected": 0.5643042922019958,
+      "step": 5450
+    },
+    {
+      "epoch": 1.7196173379000825,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.5663556340269984e-06,
+      "logits/chosen": -0.5170494318008423,
+      "logits/rejected": -0.3688809871673584,
+      "logps/chosen": -198.2430419921875,
+      "logps/rejected": -165.7610321044922,
+      "loss": 0.6394,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.7186397314071655,
+      "rewards/margins": 0.14983563125133514,
+      "rewards/rejected": 0.568804144859314,
+      "step": 5460
+    },
+    {
+      "epoch": 1.722766820203929,
+      "grad_norm": 3.265625,
+      "learning_rate": 3.5615857020424786e-06,
+      "logits/chosen": -0.48056459426879883,
+      "logits/rejected": -0.42547035217285156,
+      "logps/chosen": -225.29281616210938,
+      "logps/rejected": -195.69827270507812,
+      "loss": 0.6488,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.7470188140869141,
+      "rewards/margins": 0.1338168829679489,
+      "rewards/rejected": 0.613201916217804,
+      "step": 5470
+    },
+    {
+      "epoch": 1.7259163025077753,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.5568110507088146e-06,
+      "logits/chosen": -0.42114323377609253,
+      "logits/rejected": -0.29093047976493835,
+      "logps/chosen": -188.66677856445312,
+      "logps/rejected": -163.27169799804688,
+      "loss": 0.6448,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.6538271307945251,
+      "rewards/margins": 0.14478826522827148,
+      "rewards/rejected": 0.5090388059616089,
+      "step": 5480
+    },
+    {
+      "epoch": 1.7290657848116215,
+      "grad_norm": 2.75,
+      "learning_rate": 3.5520317012520327e-06,
+      "logits/chosen": -0.4472483694553375,
+      "logits/rejected": -0.3266296684741974,
+      "logps/chosen": -198.31227111816406,
+      "logps/rejected": -158.91233825683594,
+      "loss": 0.6354,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.6903390884399414,
+      "rewards/margins": 0.15532377362251282,
+      "rewards/rejected": 0.5350152850151062,
+      "step": 5490
+    },
+    {
+      "epoch": 1.732215267115468,
+      "grad_norm": 1.9765625,
+      "learning_rate": 3.5472476749190465e-06,
+      "logits/chosen": -0.44366365671157837,
+      "logits/rejected": -0.3832937180995941,
+      "logps/chosen": -178.91236877441406,
+      "logps/rejected": -161.47250366210938,
+      "loss": 0.6579,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.6258020401000977,
+      "rewards/margins": 0.1169646754860878,
+      "rewards/rejected": 0.5088373422622681,
+      "step": 5500
+    },
+    {
+      "epoch": 1.7353647494193143,
+      "grad_norm": 3.53125,
+      "learning_rate": 3.5424589929775593e-06,
+      "logits/chosen": -0.4584302306175232,
+      "logits/rejected": -0.3267679810523987,
+      "logps/chosen": -193.6768798828125,
+      "logps/rejected": -170.361572265625,
+      "loss": 0.6402,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.7419657707214355,
+      "rewards/margins": 0.14626576006412506,
+      "rewards/rejected": 0.5956999063491821,
+      "step": 5510
+    },
+    {
+      "epoch": 1.7385142317231606,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.5376656767159724e-06,
+      "logits/chosen": -0.42630109190940857,
+      "logits/rejected": -0.33150094747543335,
+      "logps/chosen": -188.36764526367188,
+      "logps/rejected": -165.72805786132812,
+      "loss": 0.6567,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.6277583837509155,
+      "rewards/margins": 0.11112833023071289,
+      "rewards/rejected": 0.5166300535202026,
+      "step": 5520
+    },
+    {
+      "epoch": 1.7416637140270068,
+      "grad_norm": 2.40625,
+      "learning_rate": 3.5328677474432893e-06,
+      "logits/chosen": -0.4710637629032135,
+      "logits/rejected": -0.36444777250289917,
+      "logps/chosen": -193.68409729003906,
+      "logps/rejected": -158.17857360839844,
+      "loss": 0.6465,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.6602168083190918,
+      "rewards/margins": 0.13958847522735596,
+      "rewards/rejected": 0.5206283330917358,
+      "step": 5530
+    },
+    {
+      "epoch": 1.744813196330853,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.5280652264890197e-06,
+      "logits/chosen": -0.5480708479881287,
+      "logits/rejected": -0.4107402265071869,
+      "logps/chosen": -198.30174255371094,
+      "logps/rejected": -155.7358856201172,
+      "loss": 0.6093,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.7575246095657349,
+      "rewards/margins": 0.22439555823802948,
+      "rewards/rejected": 0.5331289768218994,
+      "step": 5540
+    },
+    {
+      "epoch": 1.7479626786346993,
+      "grad_norm": 2.375,
+      "learning_rate": 3.523258135203087e-06,
+      "logits/chosen": -0.49628472328186035,
+      "logits/rejected": -0.39903074502944946,
+      "logps/chosen": -206.2187957763672,
+      "logps/rejected": -172.8035888671875,
+      "loss": 0.6493,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.7126896977424622,
+      "rewards/margins": 0.13072232902050018,
+      "rewards/rejected": 0.5819673538208008,
+      "step": 5550
+    },
+    {
+      "epoch": 1.7511121609385456,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.518446494955732e-06,
+      "logits/chosen": -0.5156094431877136,
+      "logits/rejected": -0.38701295852661133,
+      "logps/chosen": -179.8013153076172,
+      "logps/rejected": -152.9771728515625,
+      "loss": 0.6506,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.6560216546058655,
+      "rewards/margins": 0.12969034910202026,
+      "rewards/rejected": 0.52633136510849,
+      "step": 5560
+    },
+    {
+      "epoch": 1.7542616432423919,
+      "grad_norm": 2.65625,
+      "learning_rate": 3.5136303271374185e-06,
+      "logits/chosen": -0.46732720732688904,
+      "logits/rejected": -0.33687490224838257,
+      "logps/chosen": -198.79708862304688,
+      "logps/rejected": -176.80389404296875,
+      "loss": 0.6617,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.7469144463539124,
+      "rewards/margins": 0.11790307611227036,
+      "rewards/rejected": 0.629011332988739,
+      "step": 5570
+    },
+    {
+      "epoch": 1.7574111255462384,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.5088096531587377e-06,
+      "logits/chosen": -0.4696858823299408,
+      "logits/rejected": -0.378410279750824,
+      "logps/chosen": -198.86094665527344,
+      "logps/rejected": -171.12442016601562,
+      "loss": 0.6509,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.6952942609786987,
+      "rewards/margins": 0.12442521750926971,
+      "rewards/rejected": 0.5708690285682678,
+      "step": 5580
+    },
+    {
+      "epoch": 1.7605606078500846,
+      "grad_norm": 2.265625,
+      "learning_rate": 3.5039844944503137e-06,
+      "logits/chosen": -0.44871068000793457,
+      "logits/rejected": -0.23819032311439514,
+      "logps/chosen": -208.89993286132812,
+      "logps/rejected": -163.7926025390625,
+      "loss": 0.6338,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.7412741184234619,
+      "rewards/margins": 0.17887642979621887,
+      "rewards/rejected": 0.5623977780342102,
+      "step": 5590
+    },
+    {
+      "epoch": 1.763710090153931,
+      "grad_norm": 2.484375,
+      "learning_rate": 3.4991548724627054e-06,
+      "logits/chosen": -0.5238919258117676,
+      "logits/rejected": -0.3585182726383209,
+      "logps/chosen": -222.5647430419922,
+      "logps/rejected": -178.90841674804688,
+      "loss": 0.5866,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.7905155420303345,
+      "rewards/margins": 0.27365249395370483,
+      "rewards/rejected": 0.5168629884719849,
+      "step": 5600
+    },
+    {
+      "epoch": 1.7668595724577774,
+      "grad_norm": 2.46875,
+      "learning_rate": 3.4943208086663183e-06,
+      "logits/chosen": -0.4847659170627594,
+      "logits/rejected": -0.33793026208877563,
+      "logps/chosen": -197.34933471679688,
+      "logps/rejected": -174.9829559326172,
+      "loss": 0.646,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.7014733552932739,
+      "rewards/margins": 0.12839707732200623,
+      "rewards/rejected": 0.5730762481689453,
+      "step": 5610
+    },
+    {
+      "epoch": 1.7700090547616236,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.4894823245512986e-06,
+      "logits/chosen": -0.506749153137207,
+      "logits/rejected": -0.45556968450546265,
+      "logps/chosen": -197.71902465820312,
+      "logps/rejected": -186.50241088867188,
+      "loss": 0.6803,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.7117626070976257,
+      "rewards/margins": 0.06737571209669113,
+      "rewards/rejected": 0.644386887550354,
+      "step": 5620
+    },
+    {
+      "epoch": 1.7731585370654699,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.484639441627448e-06,
+      "logits/chosen": -0.5070594549179077,
+      "logits/rejected": -0.3329693078994751,
+      "logps/chosen": -220.60986328125,
+      "logps/rejected": -183.98416137695312,
+      "loss": 0.6042,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.7875211834907532,
+      "rewards/margins": 0.2286391705274582,
+      "rewards/rejected": 0.5588821172714233,
+      "step": 5630
+    },
+    {
+      "epoch": 1.7763080193693161,
+      "grad_norm": 2.546875,
+      "learning_rate": 3.4797921814241196e-06,
+      "logits/chosen": -0.48938584327697754,
+      "logits/rejected": -0.37643399834632874,
+      "logps/chosen": -194.7692413330078,
+      "logps/rejected": -171.0836944580078,
+      "loss": 0.6345,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.7208179235458374,
+      "rewards/margins": 0.17952939867973328,
+      "rewards/rejected": 0.5412884950637817,
+      "step": 5640
+    },
+    {
+      "epoch": 1.7794575016731624,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.4749405654901297e-06,
+      "logits/chosen": -0.5021311044692993,
+      "logits/rejected": -0.3592470586299896,
+      "logps/chosen": -203.04798889160156,
+      "logps/rejected": -170.28916931152344,
+      "loss": 0.6468,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.7304830551147461,
+      "rewards/margins": 0.14200101792812347,
+      "rewards/rejected": 0.5884820222854614,
+      "step": 5650
+    },
+    {
+      "epoch": 1.7826069839770087,
+      "grad_norm": 1.8125,
+      "learning_rate": 3.470084615393655e-06,
+      "logits/chosen": -0.5099314451217651,
+      "logits/rejected": -0.36777496337890625,
+      "logps/chosen": -188.96286010742188,
+      "logps/rejected": -158.13487243652344,
+      "loss": 0.5854,
+      "rewards/accuracies": 0.8374999761581421,
+      "rewards/chosen": 0.7638787031173706,
+      "rewards/margins": 0.25533777475357056,
+      "rewards/rejected": 0.5085408687591553,
+      "step": 5660
+    },
+    {
+      "epoch": 1.785756466280855,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.4652243527221423e-06,
+      "logits/chosen": -0.4756031632423401,
+      "logits/rejected": -0.44920986890792847,
+      "logps/chosen": -185.1388397216797,
+      "logps/rejected": -172.55137634277344,
+      "loss": 0.6583,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.7094627618789673,
+      "rewards/margins": 0.13025884330272675,
+      "rewards/rejected": 0.5792039036750793,
+      "step": 5670
+    },
+    {
+      "epoch": 1.7889059485847014,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.460359799082209e-06,
+      "logits/chosen": -0.47689515352249146,
+      "logits/rejected": -0.34241801500320435,
+      "logps/chosen": -204.8109588623047,
+      "logps/rejected": -166.13514709472656,
+      "loss": 0.615,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.7595565915107727,
+      "rewards/margins": 0.21238622069358826,
+      "rewards/rejected": 0.5471702814102173,
+      "step": 5680
+    },
+    {
+      "epoch": 1.7920554308885477,
+      "grad_norm": 3.765625,
+      "learning_rate": 3.4554909760995485e-06,
+      "logits/chosen": -0.5418170094490051,
+      "logits/rejected": -0.41362690925598145,
+      "logps/chosen": -187.98043823242188,
+      "logps/rejected": -167.5854034423828,
+      "loss": 0.6338,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.7204712629318237,
+      "rewards/margins": 0.1737706959247589,
+      "rewards/rejected": 0.5467005968093872,
+      "step": 5690
+    },
+    {
+      "epoch": 1.795204913192394,
+      "grad_norm": 3.3125,
+      "learning_rate": 3.450617905418834e-06,
+      "logits/chosen": -0.442087322473526,
+      "logits/rejected": -0.3480719029903412,
+      "logps/chosen": -205.0787353515625,
+      "logps/rejected": -176.585693359375,
+      "loss": 0.6078,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.7968889474868774,
+      "rewards/margins": 0.2236328423023224,
+      "rewards/rejected": 0.5732561349868774,
+      "step": 5700
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 15000,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 100,
+  "total_flos": 0.0,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}