diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,12780 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 3750,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.004,
+      "grad_norm": 18.786390881708563,
+      "learning_rate": 6.666666666666668e-08,
+      "logits/chosen": -0.5580782294273376,
+      "logits/rejected": -0.7519971132278442,
+      "logps/chosen": -1.739689588546753,
+      "logps/rejected": -2.5574848651885986,
+      "loss": 1.6179,
+      "odds_ratio_loss": 0.6882386207580566,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.08698447048664093,
+      "rewards/margins": 0.040889762341976166,
+      "rewards/rejected": -0.1278742551803589,
+      "sft_loss": 1.739689588546753,
+      "step": 5
+    },
+    {
+      "epoch": 0.008,
+      "grad_norm": 23.4203092044018,
+      "learning_rate": 1.3333333333333336e-07,
+      "logits/chosen": -0.33697596192359924,
+      "logits/rejected": -0.5306824445724487,
+      "logps/chosen": -1.5324140787124634,
+      "logps/rejected": -1.5423911809921265,
+      "loss": 1.6397,
+      "odds_ratio_loss": 0.6955539584159851,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.07662070542573929,
+      "rewards/margins": 0.0004988554865121841,
+      "rewards/rejected": -0.07711955904960632,
+      "sft_loss": 1.5324140787124634,
+      "step": 10
+    },
+    {
+      "epoch": 0.012,
+      "grad_norm": 115.39311996590018,
+      "learning_rate": 2.0000000000000002e-07,
+      "logits/chosen": -0.46800583600997925,
+      "logits/rejected": -0.6147341728210449,
+      "logps/chosen": -1.223116397857666,
+      "logps/rejected": -1.5311689376831055,
+      "loss": 1.8705,
+      "odds_ratio_loss": 0.5750418901443481,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.06115582585334778,
+      "rewards/margins": 0.015402625314891338,
+      "rewards/rejected": -0.07655844837427139,
+      "sft_loss": 1.223116397857666,
+      "step": 15
+    },
+    {
+      "epoch": 0.016,
+      "grad_norm": 60.195715309742106,
+      "learning_rate": 2.666666666666667e-07,
+      "logits/chosen": -0.699694812297821,
+      "logits/rejected": -0.5708433985710144,
+      "logps/chosen": -2.690779209136963,
+      "logps/rejected": -2.1934876441955566,
+      "loss": 1.7322,
+      "odds_ratio_loss": 1.3645074367523193,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.1345389485359192,
+      "rewards/margins": -0.02486458420753479,
+      "rewards/rejected": -0.1096743792295456,
+      "sft_loss": 2.690779209136963,
+      "step": 20
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 136.78792197913728,
+      "learning_rate": 3.3333333333333335e-07,
+      "logits/chosen": -0.4181106686592102,
+      "logits/rejected": -0.6348497867584229,
+      "logps/chosen": -1.2124601602554321,
+      "logps/rejected": -1.4138513803482056,
+      "loss": 1.8234,
+      "odds_ratio_loss": 0.6432029008865356,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.060623008757829666,
+      "rewards/margins": 0.010069566778838634,
+      "rewards/rejected": -0.07069256901741028,
+      "sft_loss": 1.2124601602554321,
+      "step": 25
+    },
+    {
+      "epoch": 0.024,
+      "grad_norm": 82.77355067386294,
+      "learning_rate": 4.0000000000000003e-07,
+      "logits/chosen": -0.43863534927368164,
+      "logits/rejected": -0.43456798791885376,
+      "logps/chosen": -1.6070562601089478,
+      "logps/rejected": -1.8459875583648682,
+      "loss": 1.6574,
+      "odds_ratio_loss": 0.6512861251831055,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.08035281300544739,
+      "rewards/margins": 0.011946573853492737,
+      "rewards/rejected": -0.09229937940835953,
+      "sft_loss": 1.6070562601089478,
+      "step": 30
+    },
+    {
+      "epoch": 0.028,
+      "grad_norm": 23.36137001588406,
+      "learning_rate": 4.666666666666667e-07,
+      "logits/chosen": -0.6327546238899231,
+      "logits/rejected": -0.5695880055427551,
+      "logps/chosen": -1.2451648712158203,
+      "logps/rejected": -1.3754527568817139,
+      "loss": 1.5305,
+      "odds_ratio_loss": 0.7118954658508301,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.062258243560791016,
+      "rewards/margins": 0.006514391396194696,
+      "rewards/rejected": -0.06877263635396957,
+      "sft_loss": 1.2451648712158203,
+      "step": 35
+    },
+    {
+      "epoch": 0.032,
+      "grad_norm": 36.18385215531129,
+      "learning_rate": 5.333333333333335e-07,
+      "logits/chosen": -0.5442869067192078,
+      "logits/rejected": -0.5034470558166504,
+      "logps/chosen": -1.5560173988342285,
+      "logps/rejected": -1.7619720697402954,
+      "loss": 1.4824,
+      "odds_ratio_loss": 0.7844841480255127,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.07780086249113083,
+      "rewards/margins": 0.010297740809619427,
+      "rewards/rejected": -0.08809860795736313,
+      "sft_loss": 1.5560173988342285,
+      "step": 40
+    },
+    {
+      "epoch": 0.036,
+      "grad_norm": 13.148829953179805,
+      "learning_rate": 6.000000000000001e-07,
+      "logits/chosen": -0.6464945673942566,
+      "logits/rejected": -0.6559021472930908,
+      "logps/chosen": -1.3128869533538818,
+      "logps/rejected": -1.5577877759933472,
+      "loss": 1.4002,
+      "odds_ratio_loss": 0.7039626836776733,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.06564434617757797,
+      "rewards/margins": 0.012245049700140953,
+      "rewards/rejected": -0.07788939774036407,
+      "sft_loss": 1.3128869533538818,
+      "step": 45
+    },
+    {
+      "epoch": 0.04,
+      "grad_norm": 31.644096608034133,
+      "learning_rate": 6.666666666666667e-07,
+      "logits/chosen": -0.6939803957939148,
+      "logits/rejected": -0.9392184019088745,
+      "logps/chosen": -1.2749038934707642,
+      "logps/rejected": -1.8819398880004883,
+      "loss": 1.4027,
+      "odds_ratio_loss": 0.5089942812919617,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.06374519318342209,
+      "rewards/margins": 0.030351802706718445,
+      "rewards/rejected": -0.09409699589014053,
+      "sft_loss": 1.2749038934707642,
+      "step": 50
+    },
+    {
+      "epoch": 0.044,
+      "grad_norm": 62.42905881663458,
+      "learning_rate": 7.333333333333334e-07,
+      "logits/chosen": -0.8882333636283875,
+      "logits/rejected": -0.8815867304801941,
+      "logps/chosen": -1.6295671463012695,
+      "logps/rejected": -1.6508338451385498,
+      "loss": 1.53,
+      "odds_ratio_loss": 0.8277268409729004,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": -0.08147837221622467,
+      "rewards/margins": 0.0010633214842528105,
+      "rewards/rejected": -0.08254168927669525,
+      "sft_loss": 1.6295671463012695,
+      "step": 55
+    },
+    {
+      "epoch": 0.048,
+      "grad_norm": 16.729835988532628,
+      "learning_rate": 8.000000000000001e-07,
+      "logits/chosen": -1.0203667879104614,
+      "logits/rejected": -0.8139181137084961,
+      "logps/chosen": -1.0937988758087158,
+      "logps/rejected": -1.4742703437805176,
+      "loss": 1.3953,
+      "odds_ratio_loss": 0.49027538299560547,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05468994379043579,
+      "rewards/margins": 0.019023580476641655,
+      "rewards/rejected": -0.073713518679142,
+      "sft_loss": 1.0937988758087158,
+      "step": 60
+    },
+    {
+      "epoch": 0.052,
+      "grad_norm": 13.431687226067643,
+      "learning_rate": 8.666666666666668e-07,
+      "logits/chosen": -0.6755761504173279,
+      "logits/rejected": -0.9210033416748047,
+      "logps/chosen": -1.2131409645080566,
+      "logps/rejected": -1.709071159362793,
+      "loss": 1.3022,
+      "odds_ratio_loss": 0.5299323201179504,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.060657043009996414,
+      "rewards/margins": 0.024796508252620697,
+      "rewards/rejected": -0.08545355498790741,
+      "sft_loss": 1.2131409645080566,
+      "step": 65
+    },
+    {
+      "epoch": 0.056,
+      "grad_norm": 14.183853012054012,
+      "learning_rate": 9.333333333333334e-07,
+      "logits/chosen": -0.6897495985031128,
+      "logits/rejected": -0.6041379570960999,
+      "logps/chosen": -1.1399095058441162,
+      "logps/rejected": -1.4036375284194946,
+      "loss": 1.2925,
+      "odds_ratio_loss": 0.5934557914733887,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.05699547380208969,
+      "rewards/margins": 0.01318640448153019,
+      "rewards/rejected": -0.07018186897039413,
+      "sft_loss": 1.1399095058441162,
+      "step": 70
+    },
+    {
+      "epoch": 0.06,
+      "grad_norm": 8.539096149562175,
+      "learning_rate": 1.0000000000000002e-06,
+      "logits/chosen": -0.8109081387519836,
+      "logits/rejected": -0.8047749400138855,
+      "logps/chosen": -1.396490216255188,
+      "logps/rejected": -1.7933557033538818,
+      "loss": 1.327,
+      "odds_ratio_loss": 0.5322818756103516,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.06982450932264328,
+      "rewards/margins": 0.019843269139528275,
+      "rewards/rejected": -0.08966778218746185,
+      "sft_loss": 1.396490216255188,
+      "step": 75
+    },
+    {
+      "epoch": 0.064,
+      "grad_norm": 16.572964152820063,
+      "learning_rate": 1.066666666666667e-06,
+      "logits/chosen": -0.6374383568763733,
+      "logits/rejected": -0.795744001865387,
+      "logps/chosen": -1.2772842645645142,
+      "logps/rejected": -1.3515651226043701,
+      "loss": 1.3445,
+      "odds_ratio_loss": 0.7885932922363281,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.06386421620845795,
+      "rewards/margins": 0.0037140310741961002,
+      "rewards/rejected": -0.06757824867963791,
+      "sft_loss": 1.2772842645645142,
+      "step": 80
+    },
+    {
+      "epoch": 0.068,
+      "grad_norm": 11.477331395703159,
+      "learning_rate": 1.1333333333333334e-06,
+      "logits/chosen": -0.655463695526123,
+      "logits/rejected": -0.6941269636154175,
+      "logps/chosen": -1.2637232542037964,
+      "logps/rejected": -1.4608389139175415,
+      "loss": 1.3153,
+      "odds_ratio_loss": 0.628704845905304,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.0631861612200737,
+      "rewards/margins": 0.009855778887867928,
+      "rewards/rejected": -0.07304193824529648,
+      "sft_loss": 1.2637232542037964,
+      "step": 85
+    },
+    {
+      "epoch": 0.072,
+      "grad_norm": 8.882050736167244,
+      "learning_rate": 1.2000000000000002e-06,
+      "logits/chosen": -0.7108498215675354,
+      "logits/rejected": -0.7867471575737,
+      "logps/chosen": -1.346906065940857,
+      "logps/rejected": -1.1983606815338135,
+      "loss": 1.233,
+      "odds_ratio_loss": 1.0082073211669922,
+      "rewards/accuracies": 0.20000000298023224,
+      "rewards/chosen": -0.06734530627727509,
+      "rewards/margins": -0.007427269127219915,
+      "rewards/rejected": -0.05991803854703903,
+      "sft_loss": 1.346906065940857,
+      "step": 90
+    },
+    {
+      "epoch": 0.076,
+      "grad_norm": 9.355470695768565,
+      "learning_rate": 1.2666666666666669e-06,
+      "logits/chosen": -0.5399163961410522,
+      "logits/rejected": -0.7329431772232056,
+      "logps/chosen": -1.2461333274841309,
+      "logps/rejected": -1.0014374256134033,
+      "loss": 1.2829,
+      "odds_ratio_loss": 0.9972267150878906,
+      "rewards/accuracies": 0.20000000298023224,
+      "rewards/chosen": -0.06230667233467102,
+      "rewards/margins": -0.012234793975949287,
+      "rewards/rejected": -0.050071872770786285,
+      "sft_loss": 1.2461333274841309,
+      "step": 95
+    },
+    {
+      "epoch": 0.08,
+      "grad_norm": 21.047192553416995,
+      "learning_rate": 1.3333333333333334e-06,
+      "logits/chosen": -0.4293244779109955,
+      "logits/rejected": -1.0099979639053345,
+      "logps/chosen": -1.2779626846313477,
+      "logps/rejected": -1.6887423992156982,
+      "loss": 1.2277,
+      "odds_ratio_loss": 0.48076191544532776,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.06389813870191574,
+      "rewards/margins": 0.020538978278636932,
+      "rewards/rejected": -0.08443711698055267,
+      "sft_loss": 1.2779626846313477,
+      "step": 100
+    },
+    {
+      "epoch": 0.084,
+      "grad_norm": 7.732739294319649,
+      "learning_rate": 1.4000000000000001e-06,
+      "logits/chosen": -0.48744726181030273,
+      "logits/rejected": -0.6798993349075317,
+      "logps/chosen": -1.0932471752166748,
+      "logps/rejected": -1.5973366498947144,
+      "loss": 1.1933,
+      "odds_ratio_loss": 0.6056200861930847,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.05466235801577568,
+      "rewards/margins": 0.025204479694366455,
+      "rewards/rejected": -0.07986684143543243,
+      "sft_loss": 1.0932471752166748,
+      "step": 105
+    },
+    {
+      "epoch": 0.088,
+      "grad_norm": 17.46330045101831,
+      "learning_rate": 1.4666666666666669e-06,
+      "logits/chosen": -0.7039790153503418,
+      "logits/rejected": -0.4986042380332947,
+      "logps/chosen": -1.3014500141143799,
+      "logps/rejected": -1.3300515413284302,
+      "loss": 1.2646,
+      "odds_ratio_loss": 0.7541639804840088,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.06507251411676407,
+      "rewards/margins": 0.0014300707262009382,
+      "rewards/rejected": -0.06650258600711823,
+      "sft_loss": 1.3014500141143799,
+      "step": 110
+    },
+    {
+      "epoch": 0.092,
+      "grad_norm": 9.681223889349653,
+      "learning_rate": 1.5333333333333334e-06,
+      "logits/chosen": -0.33655840158462524,
+      "logits/rejected": -0.45631903409957886,
+      "logps/chosen": -1.1225440502166748,
+      "logps/rejected": -1.4931358098983765,
+      "loss": 1.2771,
+      "odds_ratio_loss": 0.5696986317634583,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.05612720176577568,
+      "rewards/margins": 0.01852959208190441,
+      "rewards/rejected": -0.07465679943561554,
+      "sft_loss": 1.1225440502166748,
+      "step": 115
+    },
+    {
+      "epoch": 0.096,
+      "grad_norm": 6.114934666493648,
+      "learning_rate": 1.6000000000000001e-06,
+      "logits/chosen": -0.23245234787464142,
+      "logits/rejected": -0.36162400245666504,
+      "logps/chosen": -0.8415991067886353,
+      "logps/rejected": -1.6995826959609985,
+      "loss": 1.3043,
+      "odds_ratio_loss": 0.4082559049129486,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.04207995533943176,
+      "rewards/margins": 0.04289917275309563,
+      "rewards/rejected": -0.08497913181781769,
+      "sft_loss": 0.8415991067886353,
+      "step": 120
+    },
+    {
+      "epoch": 0.1,
+      "grad_norm": 29.566253249742736,
+      "learning_rate": 1.6666666666666667e-06,
+      "logits/chosen": -0.6754384636878967,
+      "logits/rejected": -0.5784372091293335,
+      "logps/chosen": -1.1139187812805176,
+      "logps/rejected": -1.0610531568527222,
+      "loss": 1.2067,
+      "odds_ratio_loss": 0.8141271471977234,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.05569593235850334,
+      "rewards/margins": -0.002643275773152709,
+      "rewards/rejected": -0.05305265635251999,
+      "sft_loss": 1.1139187812805176,
+      "step": 125
+    },
+    {
+      "epoch": 0.104,
+      "grad_norm": 18.345090279015356,
+      "learning_rate": 1.7333333333333336e-06,
+      "logits/chosen": -0.5198614597320557,
+      "logits/rejected": -0.5959798693656921,
+      "logps/chosen": -1.2163742780685425,
+      "logps/rejected": -1.0906116962432861,
+      "loss": 1.3261,
+      "odds_ratio_loss": 0.8466520309448242,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": -0.060818713158369064,
+      "rewards/margins": -0.006288123317062855,
+      "rewards/rejected": -0.054530590772628784,
+      "sft_loss": 1.2163742780685425,
+      "step": 130
+    },
+    {
+      "epoch": 0.108,
+      "grad_norm": 14.549007799191262,
+      "learning_rate": 1.8000000000000001e-06,
+      "logits/chosen": -0.43266358971595764,
+      "logits/rejected": -0.566035270690918,
+      "logps/chosen": -1.3842393159866333,
+      "logps/rejected": -1.4916588068008423,
+      "loss": 1.3058,
+      "odds_ratio_loss": 0.6889979243278503,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.06921195983886719,
+      "rewards/margins": 0.005370972212404013,
+      "rewards/rejected": -0.07458294183015823,
+      "sft_loss": 1.3842393159866333,
+      "step": 135
+    },
+    {
+      "epoch": 0.112,
+      "grad_norm": 23.355911973188775,
+      "learning_rate": 1.8666666666666669e-06,
+      "logits/chosen": -0.592883825302124,
+      "logits/rejected": -0.4132818281650543,
+      "logps/chosen": -1.100903868675232,
+      "logps/rejected": -1.2986562252044678,
+      "loss": 1.141,
+      "odds_ratio_loss": 0.5977322459220886,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.055045194923877716,
+      "rewards/margins": 0.009887613356113434,
+      "rewards/rejected": -0.06493280827999115,
+      "sft_loss": 1.100903868675232,
+      "step": 140
+    },
+    {
+      "epoch": 0.116,
+      "grad_norm": 9.702121972571497,
+      "learning_rate": 1.9333333333333336e-06,
+      "logits/chosen": -0.6822474598884583,
+      "logits/rejected": -0.49178019165992737,
+      "logps/chosen": -1.1203429698944092,
+      "logps/rejected": -1.9260637760162354,
+      "loss": 1.2472,
+      "odds_ratio_loss": 0.4579285979270935,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.05601715296506882,
+      "rewards/margins": 0.04028604179620743,
+      "rewards/rejected": -0.09630318731069565,
+      "sft_loss": 1.1203429698944092,
+      "step": 145
+    },
+    {
+      "epoch": 0.12,
+      "grad_norm": 16.918077923973826,
+      "learning_rate": 2.0000000000000003e-06,
+      "logits/chosen": -0.5237475633621216,
+      "logits/rejected": -0.8934429287910461,
+      "logps/chosen": -0.8749428987503052,
+      "logps/rejected": -1.3101131916046143,
+      "loss": 1.1565,
+      "odds_ratio_loss": 0.5730590224266052,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.04374714195728302,
+      "rewards/margins": 0.021758515387773514,
+      "rewards/rejected": -0.06550566107034683,
+      "sft_loss": 0.8749428987503052,
+      "step": 150
+    },
+    {
+      "epoch": 0.124,
+      "grad_norm": 11.087436159759044,
+      "learning_rate": 2.0666666666666666e-06,
+      "logits/chosen": -0.5964301824569702,
+      "logits/rejected": -0.8087562322616577,
+      "logps/chosen": -1.5669138431549072,
+      "logps/rejected": -1.6930387020111084,
+      "loss": 1.3477,
+      "odds_ratio_loss": 0.6572802066802979,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.07834570109844208,
+      "rewards/margins": 0.006306241266429424,
+      "rewards/rejected": -0.08465193212032318,
+      "sft_loss": 1.5669138431549072,
+      "step": 155
+    },
+    {
+      "epoch": 0.128,
+      "grad_norm": 11.340859017147654,
+      "learning_rate": 2.133333333333334e-06,
+      "logits/chosen": -0.7819768190383911,
+      "logits/rejected": -0.4041160047054291,
+      "logps/chosen": -1.2440321445465088,
+      "logps/rejected": -1.2431968450546265,
+      "loss": 1.2147,
+      "odds_ratio_loss": 0.7579740285873413,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.0622016116976738,
+      "rewards/margins": -4.1763483750401065e-05,
+      "rewards/rejected": -0.06215984374284744,
+      "sft_loss": 1.2440321445465088,
+      "step": 160
+    },
+    {
+      "epoch": 0.132,
+      "grad_norm": 11.290804145869599,
+      "learning_rate": 2.2e-06,
+      "logits/chosen": -0.3270031809806824,
+      "logits/rejected": -0.42402464151382446,
+      "logps/chosen": -1.2503063678741455,
+      "logps/rejected": -1.2082438468933105,
+      "loss": 1.3976,
+      "odds_ratio_loss": 0.7837561368942261,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.06251531839370728,
+      "rewards/margins": -0.00210312707349658,
+      "rewards/rejected": -0.06041219085454941,
+      "sft_loss": 1.2503063678741455,
+      "step": 165
+    },
+    {
+      "epoch": 0.136,
+      "grad_norm": 8.653604673632358,
+      "learning_rate": 2.266666666666667e-06,
+      "logits/chosen": -0.4413372874259949,
+      "logits/rejected": -0.7267617583274841,
+      "logps/chosen": -1.1855711936950684,
+      "logps/rejected": -1.0826096534729004,
+      "loss": 1.2368,
+      "odds_ratio_loss": 0.8635139465332031,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.059278566390275955,
+      "rewards/margins": -0.005148076917976141,
+      "rewards/rejected": -0.05413048714399338,
+      "sft_loss": 1.1855711936950684,
+      "step": 170
+    },
+    {
+      "epoch": 0.14,
+      "grad_norm": 18.84252613733039,
+      "learning_rate": 2.3333333333333336e-06,
+      "logits/chosen": -0.31860145926475525,
+      "logits/rejected": -0.42350611090660095,
+      "logps/chosen": -1.1006211042404175,
+      "logps/rejected": -1.26455557346344,
+      "loss": 1.2382,
+      "odds_ratio_loss": 0.6225001811981201,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05503106117248535,
+      "rewards/margins": 0.00819671992212534,
+      "rewards/rejected": -0.06322778016328812,
+      "sft_loss": 1.1006211042404175,
+      "step": 175
+    },
+    {
+      "epoch": 0.144,
+      "grad_norm": 8.60596229262369,
+      "learning_rate": 2.4000000000000003e-06,
+      "logits/chosen": -0.6520022749900818,
+      "logits/rejected": -0.5037721991539001,
+      "logps/chosen": -1.0799922943115234,
+      "logps/rejected": -1.3683429956436157,
+      "loss": 1.2163,
+      "odds_ratio_loss": 0.6034747362136841,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05399961397051811,
+      "rewards/margins": 0.014417541213333607,
+      "rewards/rejected": -0.06841715425252914,
+      "sft_loss": 1.0799922943115234,
+      "step": 180
+    },
+    {
+      "epoch": 0.148,
+      "grad_norm": 8.352997349835983,
+      "learning_rate": 2.466666666666667e-06,
+      "logits/chosen": -0.33638280630111694,
+      "logits/rejected": -0.3540743589401245,
+      "logps/chosen": -1.5841834545135498,
+      "logps/rejected": -1.106182336807251,
+      "loss": 1.274,
+      "odds_ratio_loss": 1.2082093954086304,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": -0.07920917868614197,
+      "rewards/margins": -0.02390005625784397,
+      "rewards/rejected": -0.05530911684036255,
+      "sft_loss": 1.5841834545135498,
+      "step": 185
+    },
+    {
+      "epoch": 0.152,
+      "grad_norm": 10.77421191597458,
+      "learning_rate": 2.5333333333333338e-06,
+      "logits/chosen": -0.35303565859794617,
+      "logits/rejected": -0.5206754803657532,
+      "logps/chosen": -1.4167323112487793,
+      "logps/rejected": -1.4747750759124756,
+      "loss": 1.2796,
+      "odds_ratio_loss": 0.7247415781021118,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.0708366185426712,
+      "rewards/margins": 0.002902137814089656,
+      "rewards/rejected": -0.07373875379562378,
+      "sft_loss": 1.4167323112487793,
+      "step": 190
+    },
+    {
+      "epoch": 0.156,
+      "grad_norm": 7.1177401499786415,
+      "learning_rate": 2.6e-06,
+      "logits/chosen": -0.2510073482990265,
+      "logits/rejected": -0.5589274168014526,
+      "logps/chosen": -1.442596197128296,
+      "logps/rejected": -1.4756193161010742,
+      "loss": 1.3657,
+      "odds_ratio_loss": 0.7003253102302551,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.07212980836629868,
+      "rewards/margins": 0.0016511573921889067,
+      "rewards/rejected": -0.07378096878528595,
+      "sft_loss": 1.442596197128296,
+      "step": 195
+    },
+    {
+      "epoch": 0.16,
+      "grad_norm": 16.763946753945877,
+      "learning_rate": 2.666666666666667e-06,
+      "logits/chosen": -0.45789265632629395,
+      "logits/rejected": -0.4571969509124756,
+      "logps/chosen": -1.0912725925445557,
+      "logps/rejected": -1.1674692630767822,
+      "loss": 1.3449,
+      "odds_ratio_loss": 0.7581279873847961,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.05456363409757614,
+      "rewards/margins": 0.0038098313380032778,
+      "rewards/rejected": -0.05837346240878105,
+      "sft_loss": 1.0912725925445557,
+      "step": 200
+    },
+    {
+      "epoch": 0.164,
+      "grad_norm": 7.633189094243207,
+      "learning_rate": 2.7333333333333336e-06,
+      "logits/chosen": -0.5795624852180481,
+      "logits/rejected": -0.6412222981452942,
+      "logps/chosen": -1.0281997919082642,
+      "logps/rejected": -1.2789350748062134,
+      "loss": 1.2025,
+      "odds_ratio_loss": 0.6383514404296875,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.05140998959541321,
+      "rewards/margins": 0.012536766938865185,
+      "rewards/rejected": -0.06394675374031067,
+      "sft_loss": 1.0281997919082642,
+      "step": 205
+    },
+    {
+      "epoch": 0.168,
+      "grad_norm": 22.399131275217997,
+      "learning_rate": 2.8000000000000003e-06,
+      "logits/chosen": -0.3538805842399597,
+      "logits/rejected": -0.6279109120368958,
+      "logps/chosen": -1.0936148166656494,
+      "logps/rejected": -1.2315231561660767,
+      "loss": 1.273,
+      "odds_ratio_loss": 0.6472857594490051,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.05468074232339859,
+      "rewards/margins": 0.0068954164162278175,
+      "rewards/rejected": -0.06157616525888443,
+      "sft_loss": 1.0936148166656494,
+      "step": 210
+    },
+    {
+      "epoch": 0.172,
+      "grad_norm": 12.511784021141086,
+      "learning_rate": 2.866666666666667e-06,
+      "logits/chosen": -0.6461108326911926,
+      "logits/rejected": -0.8027406930923462,
+      "logps/chosen": -1.279821515083313,
+      "logps/rejected": -1.2249929904937744,
+      "loss": 1.1567,
+      "odds_ratio_loss": 0.7108494639396667,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": -0.06399108469486237,
+      "rewards/margins": -0.002741418778896332,
+      "rewards/rejected": -0.06124965474009514,
+      "sft_loss": 1.279821515083313,
+      "step": 215
+    },
+    {
+      "epoch": 0.176,
+      "grad_norm": 18.111790971616198,
+      "learning_rate": 2.9333333333333338e-06,
+      "logits/chosen": -0.5222848057746887,
+      "logits/rejected": -0.5374074578285217,
+      "logps/chosen": -1.1454452276229858,
+      "logps/rejected": -1.541313648223877,
+      "loss": 1.3047,
+      "odds_ratio_loss": 0.46772265434265137,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.05727226287126541,
+      "rewards/margins": 0.01979340985417366,
+      "rewards/rejected": -0.07706567645072937,
+      "sft_loss": 1.1454452276229858,
+      "step": 220
+    },
+    {
+      "epoch": 0.18,
+      "grad_norm": 12.734975127781638,
+      "learning_rate": 3e-06,
+      "logits/chosen": -0.5977376699447632,
+      "logits/rejected": -0.2883208096027374,
+      "logps/chosen": -1.3275381326675415,
+      "logps/rejected": -1.18563711643219,
+      "loss": 1.2512,
+      "odds_ratio_loss": 0.9130814671516418,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.06637690961360931,
+      "rewards/margins": -0.00709505146369338,
+      "rewards/rejected": -0.0592818558216095,
+      "sft_loss": 1.3275381326675415,
+      "step": 225
+    },
+    {
+      "epoch": 0.184,
+      "grad_norm": 21.003312916139876,
+      "learning_rate": 3.066666666666667e-06,
+      "logits/chosen": -0.5674949288368225,
+      "logits/rejected": -0.49964016675949097,
+      "logps/chosen": -1.375934362411499,
+      "logps/rejected": -1.5092494487762451,
+      "loss": 1.3693,
+      "odds_ratio_loss": 0.7147940993309021,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": -0.06879671663045883,
+      "rewards/margins": 0.006665749941021204,
+      "rewards/rejected": -0.0754624754190445,
+      "sft_loss": 1.375934362411499,
+      "step": 230
+    },
+    {
+      "epoch": 0.188,
+      "grad_norm": 11.904735065181235,
+      "learning_rate": 3.133333333333334e-06,
+      "logits/chosen": -0.6791194677352905,
+      "logits/rejected": -0.7436596155166626,
+      "logps/chosen": -0.8784946203231812,
+      "logps/rejected": -1.1814043521881104,
+      "loss": 1.248,
+      "odds_ratio_loss": 0.5295432806015015,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.043924733996391296,
+      "rewards/margins": 0.015145489946007729,
+      "rewards/rejected": -0.059070222079753876,
+      "sft_loss": 0.8784946203231812,
+      "step": 235
+    },
+    {
+      "epoch": 0.192,
+      "grad_norm": 15.144263155627513,
+      "learning_rate": 3.2000000000000003e-06,
+      "logits/chosen": -0.39098218083381653,
+      "logits/rejected": -0.48908573389053345,
+      "logps/chosen": -1.4278444051742554,
+      "logps/rejected": -1.5091187953948975,
+      "loss": 1.3214,
+      "odds_ratio_loss": 0.6652258634567261,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.07139221578836441,
+      "rewards/margins": 0.004063720349222422,
+      "rewards/rejected": -0.0754559338092804,
+      "sft_loss": 1.4278444051742554,
+      "step": 240
+    },
+    {
+      "epoch": 0.196,
+      "grad_norm": 7.701817646797151,
+      "learning_rate": 3.266666666666667e-06,
+      "logits/chosen": -0.39734160900115967,
+      "logits/rejected": -0.6708099246025085,
+      "logps/chosen": -1.1432991027832031,
+      "logps/rejected": -1.2521193027496338,
+      "loss": 1.2041,
+      "odds_ratio_loss": 0.7118695974349976,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.057164955884218216,
+      "rewards/margins": 0.005441013723611832,
+      "rewards/rejected": -0.06260596960783005,
+      "sft_loss": 1.1432991027832031,
+      "step": 245
+    },
+    {
+      "epoch": 0.2,
+      "grad_norm": 9.281984255417264,
+      "learning_rate": 3.3333333333333333e-06,
+      "logits/chosen": -0.3979361057281494,
+      "logits/rejected": -0.5376076102256775,
+      "logps/chosen": -0.9498193860054016,
+      "logps/rejected": -1.696179986000061,
+      "loss": 1.3049,
+      "odds_ratio_loss": 0.5728334188461304,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.04749097302556038,
+      "rewards/margins": 0.03731803596019745,
+      "rewards/rejected": -0.08480901271104813,
+      "sft_loss": 0.9498193860054016,
+      "step": 250
+    },
+    {
+      "epoch": 0.204,
+      "grad_norm": 9.425798997764154,
+      "learning_rate": 3.4000000000000005e-06,
+      "logits/chosen": -0.58461594581604,
+      "logits/rejected": -0.7774965763092041,
+      "logps/chosen": -1.144986867904663,
+      "logps/rejected": -1.2843310832977295,
+      "loss": 1.3384,
+      "odds_ratio_loss": 0.6974171996116638,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.057249344885349274,
+      "rewards/margins": 0.006967212073504925,
+      "rewards/rejected": -0.06421655416488647,
+      "sft_loss": 1.144986867904663,
+      "step": 255
+    },
+    {
+      "epoch": 0.208,
+      "grad_norm": 6.829873424161352,
+      "learning_rate": 3.4666666666666672e-06,
+      "logits/chosen": -0.4852909445762634,
+      "logits/rejected": -0.8059079051017761,
+      "logps/chosen": -1.2967959642410278,
+      "logps/rejected": -1.1163851022720337,
+      "loss": 1.2336,
+      "odds_ratio_loss": 0.985200047492981,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": -0.06483979523181915,
+      "rewards/margins": -0.009020542725920677,
+      "rewards/rejected": -0.05581926181912422,
+      "sft_loss": 1.2967959642410278,
+      "step": 260
+    },
+    {
+      "epoch": 0.212,
+      "grad_norm": 12.53260839362661,
+      "learning_rate": 3.5333333333333335e-06,
+      "logits/chosen": -0.8439911603927612,
+      "logits/rejected": -0.4962303042411804,
+      "logps/chosen": -1.4761368036270142,
+      "logps/rejected": -1.4275569915771484,
+      "loss": 1.2936,
+      "odds_ratio_loss": 0.8170219659805298,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": -0.07380684465169907,
+      "rewards/margins": -0.0024289849679917097,
+      "rewards/rejected": -0.07137785851955414,
+      "sft_loss": 1.4761368036270142,
+      "step": 265
+    },
+    {
+      "epoch": 0.216,
+      "grad_norm": 10.909364520509888,
+      "learning_rate": 3.6000000000000003e-06,
+      "logits/chosen": -0.7634373307228088,
+      "logits/rejected": -0.5086906552314758,
+      "logps/chosen": -1.543953537940979,
+      "logps/rejected": -1.3793667554855347,
+      "loss": 1.4597,
+      "odds_ratio_loss": 0.8944965600967407,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.07719766348600388,
+      "rewards/margins": -0.008229334838688374,
+      "rewards/rejected": -0.06896833330392838,
+      "sft_loss": 1.543953537940979,
+      "step": 270
+    },
+    {
+      "epoch": 0.22,
+      "grad_norm": 14.036081295560992,
+      "learning_rate": 3.6666666666666666e-06,
+      "logits/chosen": -0.920630931854248,
+      "logits/rejected": -0.5563604831695557,
+      "logps/chosen": -1.1286569833755493,
+      "logps/rejected": -1.3310692310333252,
+      "loss": 1.3055,
+      "odds_ratio_loss": 0.621965765953064,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.056432850658893585,
+      "rewards/margins": 0.010120616294443607,
+      "rewards/rejected": -0.06655346602201462,
+      "sft_loss": 1.1286569833755493,
+      "step": 275
+    },
+    {
+      "epoch": 0.224,
+      "grad_norm": 7.314465975057194,
+      "learning_rate": 3.7333333333333337e-06,
+      "logits/chosen": -0.39631861448287964,
+      "logits/rejected": -0.6269375681877136,
+      "logps/chosen": -1.2827950716018677,
+      "logps/rejected": -1.5547349452972412,
+      "loss": 1.2209,
+      "odds_ratio_loss": 0.7333937883377075,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.06413975358009338,
+      "rewards/margins": 0.013596994802355766,
+      "rewards/rejected": -0.0777367502450943,
+      "sft_loss": 1.2827950716018677,
+      "step": 280
+    },
+    {
+      "epoch": 0.228,
+      "grad_norm": 8.448913656983327,
+      "learning_rate": 3.8000000000000005e-06,
+      "logits/chosen": -0.465139240026474,
+      "logits/rejected": -0.5822895169258118,
+      "logps/chosen": -1.4464404582977295,
+      "logps/rejected": -1.5029900074005127,
+      "loss": 1.2867,
+      "odds_ratio_loss": 0.6958650350570679,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.07232202589511871,
+      "rewards/margins": 0.002827476244419813,
+      "rewards/rejected": -0.07514949887990952,
+      "sft_loss": 1.4464404582977295,
+      "step": 285
+    },
+    {
+      "epoch": 0.232,
+      "grad_norm": 5.845937122945277,
+      "learning_rate": 3.866666666666667e-06,
+      "logits/chosen": -0.43053460121154785,
+      "logits/rejected": -0.5996862649917603,
+      "logps/chosen": -1.004393458366394,
+      "logps/rejected": -1.2151196002960205,
+      "loss": 1.235,
+      "odds_ratio_loss": 0.6134498715400696,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.05021967366337776,
+      "rewards/margins": 0.010536303743720055,
+      "rewards/rejected": -0.060755979269742966,
+      "sft_loss": 1.004393458366394,
+      "step": 290
+    },
+    {
+      "epoch": 0.236,
+      "grad_norm": 7.768195625026772,
+      "learning_rate": 3.9333333333333335e-06,
+      "logits/chosen": -0.3257526755332947,
+      "logits/rejected": -0.6037822365760803,
+      "logps/chosen": -2.014780282974243,
+      "logps/rejected": -1.3312056064605713,
+      "loss": 1.3802,
+      "odds_ratio_loss": 1.3934136629104614,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.1007390022277832,
+      "rewards/margins": -0.034178726375103,
+      "rewards/rejected": -0.0665602907538414,
+      "sft_loss": 2.014780282974243,
+      "step": 295
+    },
+    {
+      "epoch": 0.24,
+      "grad_norm": 7.746680686909107,
+      "learning_rate": 4.000000000000001e-06,
+      "logits/chosen": -0.733782947063446,
+      "logits/rejected": -0.6740893125534058,
+      "logps/chosen": -1.2197449207305908,
+      "logps/rejected": -1.2922364473342896,
+      "loss": 1.2703,
+      "odds_ratio_loss": 0.665755569934845,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.06098724529147148,
+      "rewards/margins": 0.003624574514105916,
+      "rewards/rejected": -0.06461182236671448,
+      "sft_loss": 1.2197449207305908,
+      "step": 300
+    },
+    {
+      "epoch": 0.244,
+      "grad_norm": 13.0956632135183,
+      "learning_rate": 4.066666666666667e-06,
+      "logits/chosen": -0.7939602732658386,
+      "logits/rejected": -0.3044114410877228,
+      "logps/chosen": -1.1555862426757812,
+      "logps/rejected": -1.36032235622406,
+      "loss": 1.321,
+      "odds_ratio_loss": 0.639162003993988,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.05777931213378906,
+      "rewards/margins": 0.010236804373562336,
+      "rewards/rejected": -0.06801611930131912,
+      "sft_loss": 1.1555862426757812,
+      "step": 305
+    },
+    {
+      "epoch": 0.248,
+      "grad_norm": 11.678408623894509,
+      "learning_rate": 4.133333333333333e-06,
+      "logits/chosen": -0.7341225743293762,
+      "logits/rejected": -0.7172056436538696,
+      "logps/chosen": -1.210506796836853,
+      "logps/rejected": -1.2130941152572632,
+      "loss": 1.3283,
+      "odds_ratio_loss": 0.751409113407135,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.06052535027265549,
+      "rewards/margins": 0.00012936182611156255,
+      "rewards/rejected": -0.06065471097826958,
+      "sft_loss": 1.210506796836853,
+      "step": 310
+    },
+    {
+      "epoch": 0.252,
+      "grad_norm": 10.485708060104384,
+      "learning_rate": 4.2000000000000004e-06,
+      "logits/chosen": -0.7123367190361023,
+      "logits/rejected": -0.5732913613319397,
+      "logps/chosen": -1.216386079788208,
+      "logps/rejected": -1.3553041219711304,
+      "loss": 1.2446,
+      "odds_ratio_loss": 0.649107038974762,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.06081929802894592,
+      "rewards/margins": 0.006945909466594458,
+      "rewards/rejected": -0.06776521354913712,
+      "sft_loss": 1.216386079788208,
+      "step": 315
+    },
+    {
+      "epoch": 0.256,
+      "grad_norm": 5.890587593911549,
+      "learning_rate": 4.266666666666668e-06,
+      "logits/chosen": -0.5388275384902954,
+      "logits/rejected": -0.5508286356925964,
+      "logps/chosen": -1.1097781658172607,
+      "logps/rejected": -1.4719569683074951,
+      "loss": 1.1982,
+      "odds_ratio_loss": 0.5467421412467957,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05548890680074692,
+      "rewards/margins": 0.01810893975198269,
+      "rewards/rejected": -0.07359784096479416,
+      "sft_loss": 1.1097781658172607,
+      "step": 320
+    },
+    {
+      "epoch": 0.26,
+      "grad_norm": 16.06285773185922,
+      "learning_rate": 4.333333333333334e-06,
+      "logits/chosen": -0.7908880114555359,
+      "logits/rejected": -0.20292505621910095,
+      "logps/chosen": -2.2549867630004883,
+      "logps/rejected": -1.2224481105804443,
+      "loss": 1.5111,
+      "odds_ratio_loss": 1.8231436014175415,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.11274933815002441,
+      "rewards/margins": -0.05162693187594414,
+      "rewards/rejected": -0.061122406274080276,
+      "sft_loss": 2.2549867630004883,
+      "step": 325
+    },
+    {
+      "epoch": 0.264,
+      "grad_norm": 8.454262342240037,
+      "learning_rate": 4.4e-06,
+      "logits/chosen": -0.4531725347042084,
+      "logits/rejected": -0.431458055973053,
+      "logps/chosen": -1.2020156383514404,
+      "logps/rejected": -1.3703901767730713,
+      "loss": 1.23,
+      "odds_ratio_loss": 0.5940185785293579,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.06010078266263008,
+      "rewards/margins": 0.008418736979365349,
+      "rewards/rejected": -0.06851951032876968,
+      "sft_loss": 1.2020156383514404,
+      "step": 330
+    },
+    {
+      "epoch": 0.268,
+      "grad_norm": 9.90145361257035,
+      "learning_rate": 4.4666666666666665e-06,
+      "logits/chosen": -0.5263561010360718,
+      "logits/rejected": -0.6940143704414368,
+      "logps/chosen": -0.9874106645584106,
+      "logps/rejected": -1.1406878232955933,
+      "loss": 1.2596,
+      "odds_ratio_loss": 0.8020124435424805,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.04937053471803665,
+      "rewards/margins": 0.007663858123123646,
+      "rewards/rejected": -0.05703439190983772,
+      "sft_loss": 0.9874106645584106,
+      "step": 335
+    },
+    {
+      "epoch": 0.272,
+      "grad_norm": 7.2942480975024475,
+      "learning_rate": 4.533333333333334e-06,
+      "logits/chosen": -0.4240873456001282,
+      "logits/rejected": -0.6028670072555542,
+      "logps/chosen": -1.3075716495513916,
+      "logps/rejected": -2.017928123474121,
+      "loss": 1.2217,
+      "odds_ratio_loss": 0.5069721341133118,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.0653785839676857,
+      "rewards/margins": 0.035517822951078415,
+      "rewards/rejected": -0.10089640319347382,
+      "sft_loss": 1.3075716495513916,
+      "step": 340
+    },
+    {
+      "epoch": 0.276,
+      "grad_norm": 8.866201564442973,
+      "learning_rate": 4.600000000000001e-06,
+      "logits/chosen": -0.7544479370117188,
+      "logits/rejected": -0.21181067824363708,
+      "logps/chosen": -1.218145489692688,
+      "logps/rejected": -1.204552412033081,
+      "loss": 1.1941,
+      "odds_ratio_loss": 0.8569790720939636,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.0609072744846344,
+      "rewards/margins": -0.0006796553498134017,
+      "rewards/rejected": -0.06022762134671211,
+      "sft_loss": 1.218145489692688,
+      "step": 345
+    },
+    {
+      "epoch": 0.28,
+      "grad_norm": 8.192936983091547,
+      "learning_rate": 4.666666666666667e-06,
+      "logits/chosen": -0.32254576683044434,
+      "logits/rejected": -0.6721733212471008,
+      "logps/chosen": -1.1165757179260254,
+      "logps/rejected": -1.2479455471038818,
+      "loss": 1.2225,
+      "odds_ratio_loss": 0.6464930176734924,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.05582878738641739,
+      "rewards/margins": 0.006568485405296087,
+      "rewards/rejected": -0.062397271394729614,
+      "sft_loss": 1.1165757179260254,
+      "step": 350
+    },
+    {
+      "epoch": 0.284,
+      "grad_norm": 13.020714692622663,
+      "learning_rate": 4.7333333333333335e-06,
+      "logits/chosen": -0.4439846873283386,
+      "logits/rejected": -0.48574358224868774,
+      "logps/chosen": -1.4092596769332886,
+      "logps/rejected": -1.8674914836883545,
+      "loss": 1.1975,
+      "odds_ratio_loss": 0.6004700064659119,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.07046298682689667,
+      "rewards/margins": 0.022911589592695236,
+      "rewards/rejected": -0.0933745726943016,
+      "sft_loss": 1.4092596769332886,
+      "step": 355
+    },
+    {
+      "epoch": 0.288,
+      "grad_norm": 11.400235583060827,
+      "learning_rate": 4.800000000000001e-06,
+      "logits/chosen": -0.3304193615913391,
+      "logits/rejected": -0.539570689201355,
+      "logps/chosen": -1.2983877658843994,
+      "logps/rejected": -1.2585299015045166,
+      "loss": 1.2775,
+      "odds_ratio_loss": 0.7338186502456665,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": -0.06491939723491669,
+      "rewards/margins": -0.0019928955007344484,
+      "rewards/rejected": -0.06292649358510971,
+      "sft_loss": 1.2983877658843994,
+      "step": 360
+    },
+    {
+      "epoch": 0.292,
+      "grad_norm": 8.196409587525967,
+      "learning_rate": 4.866666666666667e-06,
+      "logits/chosen": -0.4542488157749176,
+      "logits/rejected": -0.49449652433395386,
+      "logps/chosen": -1.022592544555664,
+      "logps/rejected": -1.0563173294067383,
+      "loss": 1.3053,
+      "odds_ratio_loss": 0.7796744704246521,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.05112962797284126,
+      "rewards/margins": 0.0016862439224496484,
+      "rewards/rejected": -0.05281587317585945,
+      "sft_loss": 1.022592544555664,
+      "step": 365
+    },
+    {
+      "epoch": 0.296,
+      "grad_norm": 10.509812724188748,
+      "learning_rate": 4.933333333333334e-06,
+      "logits/chosen": -0.46988096833229065,
+      "logits/rejected": -0.31580036878585815,
+      "logps/chosen": -1.3959112167358398,
+      "logps/rejected": -1.6540298461914062,
+      "loss": 1.2666,
+      "odds_ratio_loss": 0.5971593856811523,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.06979555636644363,
+      "rewards/margins": 0.012905935756862164,
+      "rewards/rejected": -0.08270148932933807,
+      "sft_loss": 1.3959112167358398,
+      "step": 370
+    },
+    {
+      "epoch": 0.3,
+      "grad_norm": 6.579280238314472,
+      "learning_rate": 5e-06,
+      "logits/chosen": -0.3907073736190796,
+      "logits/rejected": -0.5983395576477051,
+      "logps/chosen": -1.2141754627227783,
+      "logps/rejected": -1.3564534187316895,
+      "loss": 1.2967,
+      "odds_ratio_loss": 0.6791958808898926,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.06070876866579056,
+      "rewards/margins": 0.007113890256732702,
+      "rewards/rejected": -0.06782267242670059,
+      "sft_loss": 1.2141754627227783,
+      "step": 375
+    },
+    {
+      "epoch": 0.304,
+      "grad_norm": 8.913119169147508,
+      "learning_rate": 4.999972922944898e-06,
+      "logits/chosen": -0.7129698395729065,
+      "logits/rejected": -0.48502880334854126,
+      "logps/chosen": -1.3781096935272217,
+      "logps/rejected": -1.5335767269134521,
+      "loss": 1.3143,
+      "odds_ratio_loss": 0.6898760795593262,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.06890548765659332,
+      "rewards/margins": 0.007773351855576038,
+      "rewards/rejected": -0.07667883485555649,
+      "sft_loss": 1.3781096935272217,
+      "step": 380
+    },
+    {
+      "epoch": 0.308,
+      "grad_norm": 8.218029726787362,
+      "learning_rate": 4.999891692366121e-06,
+      "logits/chosen": -0.26474082469940186,
+      "logits/rejected": -0.49545183777809143,
+      "logps/chosen": -0.9910273551940918,
+      "logps/rejected": -0.944848895072937,
+      "loss": 1.2702,
+      "odds_ratio_loss": 0.7946783304214478,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.04955137148499489,
+      "rewards/margins": -0.0023089235182851553,
+      "rewards/rejected": -0.04724244400858879,
+      "sft_loss": 0.9910273551940918,
+      "step": 385
+    },
+    {
+      "epoch": 0.312,
+      "grad_norm": 7.274203044505213,
+      "learning_rate": 4.999756310023261e-06,
+      "logits/chosen": -0.2605069875717163,
+      "logits/rejected": -0.868638813495636,
+      "logps/chosen": -1.2463948726654053,
+      "logps/rejected": -1.2507517337799072,
+      "loss": 1.3006,
+      "odds_ratio_loss": 0.741990327835083,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.06231974810361862,
+      "rewards/margins": 0.00021784081764053553,
+      "rewards/rejected": -0.06253758817911148,
+      "sft_loss": 1.2463948726654053,
+      "step": 390
+    },
+    {
+      "epoch": 0.316,
+      "grad_norm": 7.536549818073618,
+      "learning_rate": 4.99956677884892e-06,
+      "logits/chosen": -0.9248331785202026,
+      "logits/rejected": -0.4915364384651184,
+      "logps/chosen": -1.2074693441390991,
+      "logps/rejected": -1.51399827003479,
+      "loss": 1.2689,
+      "odds_ratio_loss": 0.590670645236969,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.060373466461896896,
+      "rewards/margins": 0.015326438471674919,
+      "rewards/rejected": -0.07569991052150726,
+      "sft_loss": 1.2074693441390991,
+      "step": 395
+    },
+    {
+      "epoch": 0.32,
+      "grad_norm": 7.302200763333058,
+      "learning_rate": 4.999323102948655e-06,
+      "logits/chosen": -1.0176507234573364,
+      "logits/rejected": -0.4075535833835602,
+      "logps/chosen": -1.2042930126190186,
+      "logps/rejected": -1.1218703985214233,
+      "loss": 1.3829,
+      "odds_ratio_loss": 0.8366869688034058,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.06021466106176376,
+      "rewards/margins": -0.004121133126318455,
+      "rewards/rejected": -0.056093525141477585,
+      "sft_loss": 1.2042930126190186,
+      "step": 400
+    },
+    {
+      "epoch": 0.324,
+      "grad_norm": 9.822662210243099,
+      "learning_rate": 4.999025287600886e-06,
+      "logits/chosen": -0.6442962884902954,
+      "logits/rejected": -1.0240824222564697,
+      "logps/chosen": -1.1904699802398682,
+      "logps/rejected": -1.3809891939163208,
+      "loss": 1.192,
+      "odds_ratio_loss": 0.5962144732475281,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.059523504227399826,
+      "rewards/margins": 0.009525952860713005,
+      "rewards/rejected": -0.06904946267604828,
+      "sft_loss": 1.1904699802398682,
+      "step": 405
+    },
+    {
+      "epoch": 0.328,
+      "grad_norm": 7.990885197694433,
+      "learning_rate": 4.998673339256785e-06,
+      "logits/chosen": -0.47288981080055237,
+      "logits/rejected": -1.0867726802825928,
+      "logps/chosen": -1.395422101020813,
+      "logps/rejected": -1.6224483251571655,
+      "loss": 1.4465,
+      "odds_ratio_loss": 0.728722870349884,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.06977111101150513,
+      "rewards/margins": 0.011351310648024082,
+      "rewards/rejected": -0.08112241327762604,
+      "sft_loss": 1.395422101020813,
+      "step": 410
+    },
+    {
+      "epoch": 0.332,
+      "grad_norm": 22.49371675260937,
+      "learning_rate": 4.99826726554013e-06,
+      "logits/chosen": -0.7810731530189514,
+      "logits/rejected": -0.6283494234085083,
+      "logps/chosen": -1.206227421760559,
+      "logps/rejected": -1.0297901630401611,
+      "loss": 1.3452,
+      "odds_ratio_loss": 0.9216625094413757,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": -0.06031137704849243,
+      "rewards/margins": -0.008821866475045681,
+      "rewards/rejected": -0.051489509642124176,
+      "sft_loss": 1.206227421760559,
+      "step": 415
+    },
+    {
+      "epoch": 0.336,
+      "grad_norm": 7.038901075406193,
+      "learning_rate": 4.997807075247147e-06,
+      "logits/chosen": -0.4292398989200592,
+      "logits/rejected": -0.7970374226570129,
+      "logps/chosen": -1.2103397846221924,
+      "logps/rejected": -1.492582082748413,
+      "loss": 1.301,
+      "odds_ratio_loss": 0.5660519003868103,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.06051699072122574,
+      "rewards/margins": 0.014112117700278759,
+      "rewards/rejected": -0.07462911307811737,
+      "sft_loss": 1.2103397846221924,
+      "step": 420
+    },
+    {
+      "epoch": 0.34,
+      "grad_norm": 13.51899355096485,
+      "learning_rate": 4.997292778346312e-06,
+      "logits/chosen": -0.3543488681316376,
+      "logits/rejected": -0.5476067662239075,
+      "logps/chosen": -1.12038254737854,
+      "logps/rejected": -1.4835479259490967,
+      "loss": 1.2928,
+      "odds_ratio_loss": 0.621757984161377,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.056019127368927,
+      "rewards/margins": 0.018158262595534325,
+      "rewards/rejected": -0.07417739927768707,
+      "sft_loss": 1.12038254737854,
+      "step": 425
+    },
+    {
+      "epoch": 0.344,
+      "grad_norm": 18.36630863667148,
+      "learning_rate": 4.996724385978142e-06,
+      "logits/chosen": -0.6492398977279663,
+      "logits/rejected": -0.38872796297073364,
+      "logps/chosen": -1.6991106271743774,
+      "logps/rejected": -1.4718869924545288,
+      "loss": 1.3601,
+      "odds_ratio_loss": 1.0099233388900757,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.08495552837848663,
+      "rewards/margins": -0.011361182667315006,
+      "rewards/rejected": -0.0735943466424942,
+      "sft_loss": 1.6991106271743774,
+      "step": 430
+    },
+    {
+      "epoch": 0.348,
+      "grad_norm": 13.64259911410447,
+      "learning_rate": 4.996101910454953e-06,
+      "logits/chosen": -0.5003396272659302,
+      "logits/rejected": -0.9429534673690796,
+      "logps/chosen": -1.7060340642929077,
+      "logps/rejected": -1.4147275686264038,
+      "loss": 1.3717,
+      "odds_ratio_loss": 0.9737583994865417,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.0853017121553421,
+      "rewards/margins": -0.014565333724021912,
+      "rewards/rejected": -0.07073638588190079,
+      "sft_loss": 1.7060340642929077,
+      "step": 435
+    },
+    {
+      "epoch": 0.352,
+      "grad_norm": 14.592917079690519,
+      "learning_rate": 4.995425365260585e-06,
+      "logits/chosen": -0.40251150727272034,
+      "logits/rejected": -0.9722174406051636,
+      "logps/chosen": -1.0912220478057861,
+      "logps/rejected": -1.363419771194458,
+      "loss": 1.2888,
+      "odds_ratio_loss": 0.5721346735954285,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05456110090017319,
+      "rewards/margins": 0.01360989362001419,
+      "rewards/rejected": -0.06817099452018738,
+      "sft_loss": 1.0912220478057861,
+      "step": 440
+    },
+    {
+      "epoch": 0.356,
+      "grad_norm": 6.937137874320563,
+      "learning_rate": 4.994694765050121e-06,
+      "logits/chosen": -0.8574808835983276,
+      "logits/rejected": -0.7796967625617981,
+      "logps/chosen": -1.3558590412139893,
+      "logps/rejected": -1.6645358800888062,
+      "loss": 1.2991,
+      "odds_ratio_loss": 0.5890018939971924,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.06779295206069946,
+      "rewards/margins": 0.015433847904205322,
+      "rewards/rejected": -0.08322679251432419,
+      "sft_loss": 1.3558590412139893,
+      "step": 445
+    },
+    {
+      "epoch": 0.36,
+      "grad_norm": 13.634295128279398,
+      "learning_rate": 4.993910125649561e-06,
+      "logits/chosen": -0.35561490058898926,
+      "logits/rejected": -0.9158598184585571,
+      "logps/chosen": -1.2657678127288818,
+      "logps/rejected": -1.4462788105010986,
+      "loss": 1.3292,
+      "odds_ratio_loss": 0.7243040800094604,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.06328839808702469,
+      "rewards/margins": 0.009025548584759235,
+      "rewards/rejected": -0.07231394946575165,
+      "sft_loss": 1.2657678127288818,
+      "step": 450
+    },
+    {
+      "epoch": 0.364,
+      "grad_norm": 15.588418308947864,
+      "learning_rate": 4.993071464055486e-06,
+      "logits/chosen": -0.7908333539962769,
+      "logits/rejected": -0.9238386154174805,
+      "logps/chosen": -1.464706301689148,
+      "logps/rejected": -1.4646415710449219,
+      "loss": 1.3062,
+      "odds_ratio_loss": 0.7673660516738892,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.07323531806468964,
+      "rewards/margins": -3.232434437450138e-06,
+      "rewards/rejected": -0.07323208451271057,
+      "sft_loss": 1.464706301689148,
+      "step": 455
+    },
+    {
+      "epoch": 0.368,
+      "grad_norm": 7.67023021612323,
+      "learning_rate": 4.992178798434684e-06,
+      "logits/chosen": -0.46548470854759216,
+      "logits/rejected": -1.2474113702774048,
+      "logps/chosen": -0.9748779535293579,
+      "logps/rejected": -1.329563856124878,
+      "loss": 1.2079,
+      "odds_ratio_loss": 0.5502606630325317,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.048743896186351776,
+      "rewards/margins": 0.01773429848253727,
+      "rewards/rejected": -0.0664781928062439,
+      "sft_loss": 0.9748779535293579,
+      "step": 460
+    },
+    {
+      "epoch": 0.372,
+      "grad_norm": 9.396021890858217,
+      "learning_rate": 4.9912321481237616e-06,
+      "logits/chosen": -0.3827522397041321,
+      "logits/rejected": -0.7908186912536621,
+      "logps/chosen": -1.2393275499343872,
+      "logps/rejected": -1.1069507598876953,
+      "loss": 1.1528,
+      "odds_ratio_loss": 0.8488165140151978,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.06196638196706772,
+      "rewards/margins": -0.00661883782595396,
+      "rewards/rejected": -0.055347543209791183,
+      "sft_loss": 1.2393275499343872,
+      "step": 465
+    },
+    {
+      "epoch": 0.376,
+      "grad_norm": 15.667684557323762,
+      "learning_rate": 4.990231533628719e-06,
+      "logits/chosen": -0.7983174920082092,
+      "logits/rejected": -0.942952036857605,
+      "logps/chosen": -1.61797297000885,
+      "logps/rejected": -1.5470950603485107,
+      "loss": 1.3583,
+      "odds_ratio_loss": 0.9691923260688782,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.08089864999055862,
+      "rewards/margins": -0.003543901490047574,
+      "rewards/rejected": -0.07735475152730942,
+      "sft_loss": 1.61797297000885,
+      "step": 470
+    },
+    {
+      "epoch": 0.38,
+      "grad_norm": 10.262155313138615,
+      "learning_rate": 4.989176976624511e-06,
+      "logits/chosen": -0.5470207929611206,
+      "logits/rejected": -0.8008524179458618,
+      "logps/chosen": -1.2335219383239746,
+      "logps/rejected": -1.537672758102417,
+      "loss": 1.3457,
+      "odds_ratio_loss": 0.6203063130378723,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.06167609617114067,
+      "rewards/margins": 0.015207541175186634,
+      "rewards/rejected": -0.07688363641500473,
+      "sft_loss": 1.2335219383239746,
+      "step": 475
+    },
+    {
+      "epoch": 0.384,
+      "grad_norm": 8.161147591651558,
+      "learning_rate": 4.988068499954578e-06,
+      "logits/chosen": -0.3918524384498596,
+      "logits/rejected": -0.5939075946807861,
+      "logps/chosen": -1.3265020847320557,
+      "logps/rejected": -1.2360165119171143,
+      "loss": 1.2193,
+      "odds_ratio_loss": 0.7953775525093079,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": -0.0663251057267189,
+      "rewards/margins": -0.004524278454482555,
+      "rewards/rejected": -0.06180082634091377,
+      "sft_loss": 1.3265020847320557,
+      "step": 480
+    },
+    {
+      "epoch": 0.388,
+      "grad_norm": 7.500330838134909,
+      "learning_rate": 4.986906127630346e-06,
+      "logits/chosen": -0.4039885401725769,
+      "logits/rejected": -0.5603979825973511,
+      "logps/chosen": -1.1305885314941406,
+      "logps/rejected": -1.2077453136444092,
+      "loss": 1.31,
+      "odds_ratio_loss": 0.710327684879303,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.05652942508459091,
+      "rewards/margins": 0.0038578410167247057,
+      "rewards/rejected": -0.0603872649371624,
+      "sft_loss": 1.1305885314941406,
+      "step": 485
+    },
+    {
+      "epoch": 0.392,
+      "grad_norm": 7.430782365855567,
+      "learning_rate": 4.985689884830711e-06,
+      "logits/chosen": -0.5145419836044312,
+      "logits/rejected": -0.5283325910568237,
+      "logps/chosen": -0.8873690366744995,
+      "logps/rejected": -1.0988765954971313,
+      "loss": 1.2163,
+      "odds_ratio_loss": 0.5778387784957886,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.044368453323841095,
+      "rewards/margins": 0.010575374588370323,
+      "rewards/rejected": -0.05494382977485657,
+      "sft_loss": 0.8873690366744995,
+      "step": 490
+    },
+    {
+      "epoch": 0.396,
+      "grad_norm": 7.833399890164202,
+      "learning_rate": 4.984419797901491e-06,
+      "logits/chosen": -0.6590821743011475,
+      "logits/rejected": -0.7595802545547485,
+      "logps/chosen": -1.2602180242538452,
+      "logps/rejected": -1.3176265954971313,
+      "loss": 1.2691,
+      "odds_ratio_loss": 0.8493593335151672,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.06301090866327286,
+      "rewards/margins": 0.0028704279102385044,
+      "rewards/rejected": -0.06588132679462433,
+      "sft_loss": 1.2602180242538452,
+      "step": 495
+    },
+    {
+      "epoch": 0.4,
+      "grad_norm": 18.632214535791704,
+      "learning_rate": 4.983095894354858e-06,
+      "logits/chosen": -0.4741068482398987,
+      "logits/rejected": -0.6550690531730652,
+      "logps/chosen": -1.0877597332000732,
+      "logps/rejected": -1.4105587005615234,
+      "loss": 1.2655,
+      "odds_ratio_loss": 0.6428815126419067,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.05438799411058426,
+      "rewards/margins": 0.01613995060324669,
+      "rewards/rejected": -0.07052794098854065,
+      "sft_loss": 1.0877597332000732,
+      "step": 500
+    },
+    {
+      "epoch": 0.404,
+      "grad_norm": 13.093628531972609,
+      "learning_rate": 4.981718202868738e-06,
+      "logits/chosen": -0.37520939111709595,
+      "logits/rejected": -0.5333417654037476,
+      "logps/chosen": -1.2307997941970825,
+      "logps/rejected": -1.3670756816864014,
+      "loss": 1.3233,
+      "odds_ratio_loss": 0.7391397356987,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.06153998523950577,
+      "rewards/margins": 0.006813795771449804,
+      "rewards/rejected": -0.06835378706455231,
+      "sft_loss": 1.2307997941970825,
+      "step": 505
+    },
+    {
+      "epoch": 0.408,
+      "grad_norm": 32.02171447844317,
+      "learning_rate": 4.980286753286196e-06,
+      "logits/chosen": -1.0221941471099854,
+      "logits/rejected": -0.6870493292808533,
+      "logps/chosen": -1.1637510061264038,
+      "logps/rejected": -1.1249791383743286,
+      "loss": 1.2777,
+      "odds_ratio_loss": 0.7576397061347961,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.05818755179643631,
+      "rewards/margins": -0.0019385985797271132,
+      "rewards/rejected": -0.05624895542860031,
+      "sft_loss": 1.1637510061264038,
+      "step": 510
+    },
+    {
+      "epoch": 0.412,
+      "grad_norm": 8.922962337548432,
+      "learning_rate": 4.978801576614779e-06,
+      "logits/chosen": -0.5233002305030823,
+      "logits/rejected": -0.7131946086883545,
+      "logps/chosen": -1.1369407176971436,
+      "logps/rejected": -1.2576231956481934,
+      "loss": 1.3731,
+      "odds_ratio_loss": 0.630257248878479,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.05684703588485718,
+      "rewards/margins": 0.00603412976488471,
+      "rewards/rejected": -0.06288117170333862,
+      "sft_loss": 1.1369407176971436,
+      "step": 515
+    },
+    {
+      "epoch": 0.416,
+      "grad_norm": 6.071002612452945,
+      "learning_rate": 4.97726270502586e-06,
+      "logits/chosen": -0.751970112323761,
+      "logits/rejected": -0.5403101444244385,
+      "logps/chosen": -1.125468373298645,
+      "logps/rejected": -1.4324285984039307,
+      "loss": 1.2763,
+      "odds_ratio_loss": 0.6366511583328247,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.05627341940999031,
+      "rewards/margins": 0.015348007902503014,
+      "rewards/rejected": -0.07162143290042877,
+      "sft_loss": 1.125468373298645,
+      "step": 520
+    },
+    {
+      "epoch": 0.42,
+      "grad_norm": 19.360606964673867,
+      "learning_rate": 4.975670171853926e-06,
+      "logits/chosen": -0.6735143065452576,
+      "logits/rejected": -0.5386877059936523,
+      "logps/chosen": -1.3962455987930298,
+      "logps/rejected": -1.3698430061340332,
+      "loss": 1.3162,
+      "odds_ratio_loss": 0.7852639555931091,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.06981228291988373,
+      "rewards/margins": -0.0013201329857110977,
+      "rewards/rejected": -0.06849215179681778,
+      "sft_loss": 1.3962455987930298,
+      "step": 525
+    },
+    {
+      "epoch": 0.424,
+      "grad_norm": 12.60459466219797,
+      "learning_rate": 4.974024011595864e-06,
+      "logits/chosen": -0.726483941078186,
+      "logits/rejected": -0.45278626680374146,
+      "logps/chosen": -1.1809227466583252,
+      "logps/rejected": -1.4415466785430908,
+      "loss": 1.2782,
+      "odds_ratio_loss": 0.5937660932540894,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.05904613807797432,
+      "rewards/margins": 0.013031196780502796,
+      "rewards/rejected": -0.07207732647657394,
+      "sft_loss": 1.1809227466583252,
+      "step": 530
+    },
+    {
+      "epoch": 0.428,
+      "grad_norm": 16.735047153184535,
+      "learning_rate": 4.97232425991021e-06,
+      "logits/chosen": -0.7554991245269775,
+      "logits/rejected": -0.6952065825462341,
+      "logps/chosen": -1.2135225534439087,
+      "logps/rejected": -1.2788511514663696,
+      "loss": 1.2871,
+      "odds_ratio_loss": 0.6615133881568909,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.060676127672195435,
+      "rewards/margins": 0.0032664339523762465,
+      "rewards/rejected": -0.0639425590634346,
+      "sft_loss": 1.2135225534439087,
+      "step": 535
+    },
+    {
+      "epoch": 0.432,
+      "grad_norm": 28.636993191829472,
+      "learning_rate": 4.970570953616383e-06,
+      "logits/chosen": -0.3668878972530365,
+      "logits/rejected": -0.47576600313186646,
+      "logps/chosen": -1.254703402519226,
+      "logps/rejected": -1.4109866619110107,
+      "loss": 1.3499,
+      "odds_ratio_loss": 0.6886258125305176,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.0627351701259613,
+      "rewards/margins": 0.007814161479473114,
+      "rewards/rejected": -0.07054933160543442,
+      "sft_loss": 1.254703402519226,
+      "step": 540
+    },
+    {
+      "epoch": 0.436,
+      "grad_norm": 8.20123877269534,
+      "learning_rate": 4.9687641306938766e-06,
+      "logits/chosen": -0.596192479133606,
+      "logits/rejected": -0.34622710943222046,
+      "logps/chosen": -1.258967399597168,
+      "logps/rejected": -1.168402910232544,
+      "loss": 1.2366,
+      "odds_ratio_loss": 0.9441742897033691,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.06294836848974228,
+      "rewards/margins": -0.004528213292360306,
+      "rewards/rejected": -0.058420151472091675,
+      "sft_loss": 1.258967399597168,
+      "step": 545
+    },
+    {
+      "epoch": 0.44,
+      "grad_norm": 19.0599247063106,
+      "learning_rate": 4.966903830281449e-06,
+      "logits/chosen": -0.550312876701355,
+      "logits/rejected": -0.7260756492614746,
+      "logps/chosen": -0.8798558115959167,
+      "logps/rejected": -0.9841960072517395,
+      "loss": 1.2888,
+      "odds_ratio_loss": 0.6650792360305786,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.043992795050144196,
+      "rewards/margins": 0.00521700968965888,
+      "rewards/rejected": -0.049209803342819214,
+      "sft_loss": 0.8798558115959167,
+      "step": 550
+    },
+    {
+      "epoch": 0.444,
+      "grad_norm": 9.539583052659502,
+      "learning_rate": 4.964990092676263e-06,
+      "logits/chosen": -0.5485479235649109,
+      "logits/rejected": -0.2632920444011688,
+      "logps/chosen": -1.2334239482879639,
+      "logps/rejected": -1.2576792240142822,
+      "loss": 1.1946,
+      "odds_ratio_loss": 0.7366870641708374,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.06167120486497879,
+      "rewards/margins": 0.0012127698864787817,
+      "rewards/rejected": -0.06288396567106247,
+      "sft_loss": 1.2334239482879639,
+      "step": 555
+    },
+    {
+      "epoch": 0.448,
+      "grad_norm": 6.575286386431191,
+      "learning_rate": 4.9630229593330226e-06,
+      "logits/chosen": -0.5613077878952026,
+      "logits/rejected": -0.9460631608963013,
+      "logps/chosen": -1.4012134075164795,
+      "logps/rejected": -1.474373698234558,
+      "loss": 1.356,
+      "odds_ratio_loss": 0.7623913884162903,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.07006067782640457,
+      "rewards/margins": 0.0036580085288733244,
+      "rewards/rejected": -0.07371868193149567,
+      "sft_loss": 1.4012134075164795,
+      "step": 560
+    },
+    {
+      "epoch": 0.452,
+      "grad_norm": 8.260756058648278,
+      "learning_rate": 4.96100247286307e-06,
+      "logits/chosen": -0.3762677311897278,
+      "logits/rejected": -0.681728184223175,
+      "logps/chosen": -1.2107107639312744,
+      "logps/rejected": -1.336549162864685,
+      "loss": 1.2595,
+      "odds_ratio_loss": 0.6657492518424988,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.06053553894162178,
+      "rewards/margins": 0.006291926838457584,
+      "rewards/rejected": -0.06682746857404709,
+      "sft_loss": 1.2107107639312744,
+      "step": 565
+    },
+    {
+      "epoch": 0.456,
+      "grad_norm": 5.224482900966935,
+      "learning_rate": 4.958928677033465e-06,
+      "logits/chosen": -0.5366466045379639,
+      "logits/rejected": -0.2813822627067566,
+      "logps/chosen": -1.0963077545166016,
+      "logps/rejected": -1.9244539737701416,
+      "loss": 1.1666,
+      "odds_ratio_loss": 0.4927639961242676,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.0548153892159462,
+      "rewards/margins": 0.04140730947256088,
+      "rewards/rejected": -0.09622270613908768,
+      "sft_loss": 1.0963077545166016,
+      "step": 570
+    },
+    {
+      "epoch": 0.46,
+      "grad_norm": 9.231089411843929,
+      "learning_rate": 4.956801616766033e-06,
+      "logits/chosen": -0.41479817032814026,
+      "logits/rejected": -1.2379553318023682,
+      "logps/chosen": -1.3129132986068726,
+      "logps/rejected": -1.4864296913146973,
+      "loss": 1.2148,
+      "odds_ratio_loss": 0.6478667259216309,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.06564567983150482,
+      "rewards/margins": 0.008675819262862206,
+      "rewards/rejected": -0.07432149350643158,
+      "sft_loss": 1.3129132986068726,
+      "step": 575
+    },
+    {
+      "epoch": 0.464,
+      "grad_norm": 12.571799699194248,
+      "learning_rate": 4.954621338136399e-06,
+      "logits/chosen": -0.43359094858169556,
+      "logits/rejected": -0.7193197011947632,
+      "logps/chosen": -1.1483484506607056,
+      "logps/rejected": -1.4485629796981812,
+      "loss": 1.217,
+      "odds_ratio_loss": 0.5974001288414001,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05741741508245468,
+      "rewards/margins": 0.015010732226073742,
+      "rewards/rejected": -0.0724281519651413,
+      "sft_loss": 1.1483484506607056,
+      "step": 580
+    },
+    {
+      "epoch": 0.468,
+      "grad_norm": 13.14113848877051,
+      "learning_rate": 4.9523878883729794e-06,
+      "logits/chosen": -0.3140491545200348,
+      "logits/rejected": -1.1171773672103882,
+      "logps/chosen": -1.1811208724975586,
+      "logps/rejected": -1.2877556085586548,
+      "loss": 1.3736,
+      "odds_ratio_loss": 0.6588929295539856,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.05905604362487793,
+      "rewards/margins": 0.005331738851964474,
+      "rewards/rejected": -0.06438778340816498,
+      "sft_loss": 1.1811208724975586,
+      "step": 585
+    },
+    {
+      "epoch": 0.472,
+      "grad_norm": 6.741763650848641,
+      "learning_rate": 4.95010131585597e-06,
+      "logits/chosen": -0.41005969047546387,
+      "logits/rejected": -1.1937333345413208,
+      "logps/chosen": -1.1939094066619873,
+      "logps/rejected": -1.519858717918396,
+      "loss": 1.3015,
+      "odds_ratio_loss": 0.5203220844268799,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05969547480344772,
+      "rewards/margins": 0.016297463327646255,
+      "rewards/rejected": -0.07599293440580368,
+      "sft_loss": 1.1939094066619873,
+      "step": 590
+    },
+    {
+      "epoch": 0.476,
+      "grad_norm": 11.415912203136902,
+      "learning_rate": 4.94776167011629e-06,
+      "logits/chosen": -0.8332249522209167,
+      "logits/rejected": -1.193943738937378,
+      "logps/chosen": -1.5026148557662964,
+      "logps/rejected": -1.3563969135284424,
+      "loss": 1.318,
+      "odds_ratio_loss": 0.8462100028991699,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": -0.07513075321912766,
+      "rewards/margins": -0.007310903165489435,
+      "rewards/rejected": -0.06781984865665436,
+      "sft_loss": 1.5026148557662964,
+      "step": 595
+    },
+    {
+      "epoch": 0.48,
+      "grad_norm": 6.517733955264292,
+      "learning_rate": 4.9453690018345144e-06,
+      "logits/chosen": -0.7536684274673462,
+      "logits/rejected": -0.5858591794967651,
+      "logps/chosen": -1.0818665027618408,
+      "logps/rejected": -0.9140428304672241,
+      "loss": 1.2555,
+      "odds_ratio_loss": 0.847333550453186,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.05409332364797592,
+      "rewards/margins": -0.008391184732317924,
+      "rewards/rejected": -0.045702140778303146,
+      "sft_loss": 1.0818665027618408,
+      "step": 600
+    },
+    {
+      "epoch": 0.484,
+      "grad_norm": 11.605525874938747,
+      "learning_rate": 4.94292336283977e-06,
+      "logits/chosen": -0.4545044004917145,
+      "logits/rejected": -0.9466699361801147,
+      "logps/chosen": -1.3461689949035645,
+      "logps/rejected": -1.257264256477356,
+      "loss": 1.2515,
+      "odds_ratio_loss": 0.7972198724746704,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": -0.0673084557056427,
+      "rewards/margins": -0.0044452352449297905,
+      "rewards/rejected": -0.06286321580410004,
+      "sft_loss": 1.3461689949035645,
+      "step": 605
+    },
+    {
+      "epoch": 0.488,
+      "grad_norm": 16.623733933538222,
+      "learning_rate": 4.940424806108619e-06,
+      "logits/chosen": -0.44307154417037964,
+      "logits/rejected": -1.1300188302993774,
+      "logps/chosen": -1.5469882488250732,
+      "logps/rejected": -1.546287178993225,
+      "loss": 1.296,
+      "odds_ratio_loss": 0.7446098923683167,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.07734941691160202,
+      "rewards/margins": -3.506392386043444e-05,
+      "rewards/rejected": -0.0773143544793129,
+      "sft_loss": 1.5469882488250732,
+      "step": 610
+    },
+    {
+      "epoch": 0.492,
+      "grad_norm": 14.50351733733393,
+      "learning_rate": 4.937873385763909e-06,
+      "logits/chosen": -0.4979380667209625,
+      "logits/rejected": -0.7332254648208618,
+      "logps/chosen": -1.186477780342102,
+      "logps/rejected": -1.4961540699005127,
+      "loss": 1.2214,
+      "odds_ratio_loss": 0.5873562097549438,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.05932389572262764,
+      "rewards/margins": 0.015483811497688293,
+      "rewards/rejected": -0.07480770349502563,
+      "sft_loss": 1.186477780342102,
+      "step": 615
+    },
+    {
+      "epoch": 0.496,
+      "grad_norm": 10.189101808498776,
+      "learning_rate": 4.935269157073597e-06,
+      "logits/chosen": -0.30127787590026855,
+      "logits/rejected": -0.2239534556865692,
+      "logps/chosen": -1.0941071510314941,
+      "logps/rejected": -1.155369520187378,
+      "loss": 1.2658,
+      "odds_ratio_loss": 0.7786355018615723,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": -0.054705359041690826,
+      "rewards/margins": 0.0030631190165877342,
+      "rewards/rejected": -0.057768482714891434,
+      "sft_loss": 1.0941071510314941,
+      "step": 620
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 6.717391051471845,
+      "learning_rate": 4.93261217644956e-06,
+      "logits/chosen": -0.6986908316612244,
+      "logits/rejected": -0.6756628751754761,
+      "logps/chosen": -1.0286588668823242,
+      "logps/rejected": -1.368683099746704,
+      "loss": 1.2814,
+      "odds_ratio_loss": 0.5862299203872681,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05143294483423233,
+      "rewards/margins": 0.017001213505864143,
+      "rewards/rejected": -0.06843416392803192,
+      "sft_loss": 1.0286588668823242,
+      "step": 625
+    },
+    {
+      "epoch": 0.504,
+      "grad_norm": 4.84256770780635,
+      "learning_rate": 4.9299025014463665e-06,
+      "logits/chosen": -0.6707471609115601,
+      "logits/rejected": -0.4917599558830261,
+      "logps/chosen": -1.401749849319458,
+      "logps/rejected": -1.291032075881958,
+      "loss": 1.2497,
+      "odds_ratio_loss": 0.8192909359931946,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.0700874850153923,
+      "rewards/margins": -0.005535887088626623,
+      "rewards/rejected": -0.06455160677433014,
+      "sft_loss": 1.401749849319458,
+      "step": 630
+    },
+    {
+      "epoch": 0.508,
+      "grad_norm": 5.259599717240724,
+      "learning_rate": 4.92714019076003e-06,
+      "logits/chosen": -0.6145176291465759,
+      "logits/rejected": -0.6850901246070862,
+      "logps/chosen": -1.1875053644180298,
+      "logps/rejected": -1.5674359798431396,
+      "loss": 1.2434,
+      "odds_ratio_loss": 0.4647662043571472,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.059375274926424026,
+      "rewards/margins": 0.018996523693203926,
+      "rewards/rejected": -0.0783717930316925,
+      "sft_loss": 1.1875053644180298,
+      "step": 635
+    },
+    {
+      "epoch": 0.512,
+      "grad_norm": 8.97887669520296,
+      "learning_rate": 4.924325304226745e-06,
+      "logits/chosen": -0.509232223033905,
+      "logits/rejected": -0.804535984992981,
+      "logps/chosen": -0.9804172515869141,
+      "logps/rejected": -1.4311110973358154,
+      "loss": 1.2258,
+      "odds_ratio_loss": 0.47105565667152405,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.04902086406946182,
+      "rewards/margins": 0.022534683346748352,
+      "rewards/rejected": -0.07155554741621017,
+      "sft_loss": 0.9804172515869141,
+      "step": 640
+    },
+    {
+      "epoch": 0.516,
+      "grad_norm": 12.339274811909346,
+      "learning_rate": 4.921457902821578e-06,
+      "logits/chosen": -0.6703714728355408,
+      "logits/rejected": -0.7211524844169617,
+      "logps/chosen": -1.412858247756958,
+      "logps/rejected": -1.7014133930206299,
+      "loss": 1.2609,
+      "odds_ratio_loss": 0.6318767666816711,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.07064291089773178,
+      "rewards/margins": 0.014427749440073967,
+      "rewards/rejected": -0.0850706622004509,
+      "sft_loss": 1.412858247756958,
+      "step": 645
+    },
+    {
+      "epoch": 0.52,
+      "grad_norm": 6.902631133651239,
+      "learning_rate": 4.91853804865716e-06,
+      "logits/chosen": -0.423556387424469,
+      "logits/rejected": -0.6117393970489502,
+      "logps/chosen": -0.9919289350509644,
+      "logps/rejected": -1.5979692935943604,
+      "loss": 1.2347,
+      "odds_ratio_loss": 0.4265903830528259,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.049596451222896576,
+      "rewards/margins": 0.0303020179271698,
+      "rewards/rejected": -0.07989846915006638,
+      "sft_loss": 0.9919289350509644,
+      "step": 650
+    },
+    {
+      "epoch": 0.524,
+      "grad_norm": 20.833564963024443,
+      "learning_rate": 4.915565804982332e-06,
+      "logits/chosen": -0.6421648263931274,
+      "logits/rejected": -0.2954399287700653,
+      "logps/chosen": -1.2931578159332275,
+      "logps/rejected": -1.3003164529800415,
+      "loss": 1.2488,
+      "odds_ratio_loss": 0.7667359113693237,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.06465788185596466,
+      "rewards/margins": 0.0003579244075808674,
+      "rewards/rejected": -0.06501581519842148,
+      "sft_loss": 1.2931578159332275,
+      "step": 655
+    },
+    {
+      "epoch": 0.528,
+      "grad_norm": 5.965871281558726,
+      "learning_rate": 4.912541236180779e-06,
+      "logits/chosen": -0.6520070433616638,
+      "logits/rejected": -0.7092788815498352,
+      "logps/chosen": -1.151734471321106,
+      "logps/rejected": -1.186388611793518,
+      "loss": 1.1441,
+      "odds_ratio_loss": 0.7639530301094055,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.05758672207593918,
+      "rewards/margins": 0.0017327029490843415,
+      "rewards/rejected": -0.059319429099559784,
+      "sft_loss": 1.151734471321106,
+      "step": 660
+    },
+    {
+      "epoch": 0.532,
+      "grad_norm": 6.664910493167132,
+      "learning_rate": 4.909464407769633e-06,
+      "logits/chosen": -0.5549234747886658,
+      "logits/rejected": -0.8910662531852722,
+      "logps/chosen": -1.2270019054412842,
+      "logps/rejected": -1.200127363204956,
+      "loss": 1.23,
+      "odds_ratio_loss": 0.8013579249382019,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.06135009601712227,
+      "rewards/margins": -0.0013437264133244753,
+      "rewards/rejected": -0.060006361454725266,
+      "sft_loss": 1.2270019054412842,
+      "step": 665
+    },
+    {
+      "epoch": 0.536,
+      "grad_norm": 5.1055719510419335,
+      "learning_rate": 4.9063353863980565e-06,
+      "logits/chosen": -0.4341130256652832,
+      "logits/rejected": -0.791808545589447,
+      "logps/chosen": -1.1596591472625732,
+      "logps/rejected": -1.5821669101715088,
+      "loss": 1.2846,
+      "odds_ratio_loss": 0.5709139108657837,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.05798295885324478,
+      "rewards/margins": 0.021125389263033867,
+      "rewards/rejected": -0.0791083499789238,
+      "sft_loss": 1.1596591472625732,
+      "step": 670
+    },
+    {
+      "epoch": 0.54,
+      "grad_norm": 7.5751885145020506,
+      "learning_rate": 4.903154239845798e-06,
+      "logits/chosen": -0.4769849181175232,
+      "logits/rejected": -0.5648764967918396,
+      "logps/chosen": -1.1110217571258545,
+      "logps/rejected": -1.4287192821502686,
+      "loss": 1.2388,
+      "odds_ratio_loss": 0.5047262907028198,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.05555109307169914,
+      "rewards/margins": 0.015884879976511,
+      "rewards/rejected": -0.07143596559762955,
+      "sft_loss": 1.1110217571258545,
+      "step": 675
+    },
+    {
+      "epoch": 0.544,
+      "grad_norm": 7.921191214189487,
+      "learning_rate": 4.899921037021719e-06,
+      "logits/chosen": -0.36087316274642944,
+      "logits/rejected": -0.1616624891757965,
+      "logps/chosen": -1.1064417362213135,
+      "logps/rejected": -0.9575250744819641,
+      "loss": 1.2096,
+      "odds_ratio_loss": 0.8907485008239746,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": -0.05532208830118179,
+      "rewards/margins": -0.007445829920470715,
+      "rewards/rejected": -0.047876257449388504,
+      "sft_loss": 1.1064417362213135,
+      "step": 680
+    },
+    {
+      "epoch": 0.548,
+      "grad_norm": 12.574180795892904,
+      "learning_rate": 4.896635847962311e-06,
+      "logits/chosen": -0.6308740973472595,
+      "logits/rejected": -0.6598986387252808,
+      "logps/chosen": -1.0816127061843872,
+      "logps/rejected": -1.5012505054473877,
+      "loss": 1.2037,
+      "odds_ratio_loss": 0.6682350635528564,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.05408062785863876,
+      "rewards/margins": 0.02098189666867256,
+      "rewards/rejected": -0.07506252825260162,
+      "sft_loss": 1.0816127061843872,
+      "step": 685
+    },
+    {
+      "epoch": 0.552,
+      "grad_norm": 11.992784752793785,
+      "learning_rate": 4.893298743830168e-06,
+      "logits/chosen": -0.44944000244140625,
+      "logits/rejected": -0.8219810724258423,
+      "logps/chosen": -1.5772783756256104,
+      "logps/rejected": -1.2028121948242188,
+      "loss": 1.2864,
+      "odds_ratio_loss": 1.1701303720474243,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.07886390388011932,
+      "rewards/margins": -0.018723303452134132,
+      "rewards/rejected": -0.06014060229063034,
+      "sft_loss": 1.5772783756256104,
+      "step": 690
+    },
+    {
+      "epoch": 0.556,
+      "grad_norm": 8.358855854874342,
+      "learning_rate": 4.889909796912454e-06,
+      "logits/chosen": -0.28839197754859924,
+      "logits/rejected": -0.44521206617355347,
+      "logps/chosen": -1.1581156253814697,
+      "logps/rejected": -1.1767852306365967,
+      "loss": 1.2989,
+      "odds_ratio_loss": 0.6894288063049316,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.057905782014131546,
+      "rewards/margins": 0.0009334773058071733,
+      "rewards/rejected": -0.058839261531829834,
+      "sft_loss": 1.1581156253814697,
+      "step": 695
+    },
+    {
+      "epoch": 0.56,
+      "grad_norm": 6.831254137005319,
+      "learning_rate": 4.88646908061933e-06,
+      "logits/chosen": -0.6353722214698792,
+      "logits/rejected": -0.7715078592300415,
+      "logps/chosen": -1.5473369359970093,
+      "logps/rejected": -1.5499107837677002,
+      "loss": 1.2584,
+      "odds_ratio_loss": 0.7936614751815796,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.07736685127019882,
+      "rewards/margins": 0.00012868604972027242,
+      "rewards/rejected": -0.0774955302476883,
+      "sft_loss": 1.5473369359970093,
+      "step": 700
+    },
+    {
+      "epoch": 0.564,
+      "grad_norm": 10.218090713494131,
+      "learning_rate": 4.882976669482368e-06,
+      "logits/chosen": -0.6931608319282532,
+      "logits/rejected": -0.6439999341964722,
+      "logps/chosen": -1.328382968902588,
+      "logps/rejected": -1.6422935724258423,
+      "loss": 1.3709,
+      "odds_ratio_loss": 0.5778505802154541,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.06641916185617447,
+      "rewards/margins": 0.015695523470640182,
+      "rewards/rejected": -0.08211468160152435,
+      "sft_loss": 1.328382968902588,
+      "step": 705
+    },
+    {
+      "epoch": 0.568,
+      "grad_norm": 5.768384836009137,
+      "learning_rate": 4.879432639152935e-06,
+      "logits/chosen": -0.5731868147850037,
+      "logits/rejected": -0.9456470608711243,
+      "logps/chosen": -0.9817520976066589,
+      "logps/rejected": -1.3865406513214111,
+      "loss": 1.2994,
+      "odds_ratio_loss": 0.48160356283187866,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.049087610095739365,
+      "rewards/margins": 0.02023942954838276,
+      "rewards/rejected": -0.06932704150676727,
+      "sft_loss": 0.9817520976066589,
+      "step": 710
+    },
+    {
+      "epoch": 0.572,
+      "grad_norm": 7.509810654555548,
+      "learning_rate": 4.875837066400553e-06,
+      "logits/chosen": -0.691825807094574,
+      "logits/rejected": -0.5271461606025696,
+      "logps/chosen": -0.9871689677238464,
+      "logps/rejected": -1.324852705001831,
+      "loss": 1.2221,
+      "odds_ratio_loss": 0.5907098054885864,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.04935844987630844,
+      "rewards/margins": 0.016884196549654007,
+      "rewards/rejected": -0.06624264270067215,
+      "sft_loss": 0.9871689677238464,
+      "step": 715
+    },
+    {
+      "epoch": 0.576,
+      "grad_norm": 24.694839731628768,
+      "learning_rate": 4.8721900291112415e-06,
+      "logits/chosen": -0.8862325549125671,
+      "logits/rejected": -0.5018847584724426,
+      "logps/chosen": -1.0031644105911255,
+      "logps/rejected": -1.247604250907898,
+      "loss": 1.2261,
+      "odds_ratio_loss": 0.6080331802368164,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05015822499990463,
+      "rewards/margins": 0.012221988290548325,
+      "rewards/rejected": -0.06238021329045296,
+      "sft_loss": 1.0031644105911255,
+      "step": 720
+    },
+    {
+      "epoch": 0.58,
+      "grad_norm": 26.923254409620736,
+      "learning_rate": 4.868491606285823e-06,
+      "logits/chosen": -0.5057476162910461,
+      "logits/rejected": -0.7251291275024414,
+      "logps/chosen": -1.470099687576294,
+      "logps/rejected": -1.3734233379364014,
+      "loss": 1.3296,
+      "odds_ratio_loss": 0.9267051815986633,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.07350499927997589,
+      "rewards/margins": -0.004833821672946215,
+      "rewards/rejected": -0.06867116689682007,
+      "sft_loss": 1.470099687576294,
+      "step": 725
+    },
+    {
+      "epoch": 0.584,
+      "grad_norm": 7.574149129374884,
+      "learning_rate": 4.864741878038218e-06,
+      "logits/chosen": -0.46996521949768066,
+      "logits/rejected": -0.6176045536994934,
+      "logps/chosen": -1.454437017440796,
+      "logps/rejected": -1.3104689121246338,
+      "loss": 1.2605,
+      "odds_ratio_loss": 0.8346797823905945,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.07272185385227203,
+      "rewards/margins": -0.007198403589427471,
+      "rewards/rejected": -0.06552345305681229,
+      "sft_loss": 1.454437017440796,
+      "step": 730
+    },
+    {
+      "epoch": 0.588,
+      "grad_norm": 8.305352840446517,
+      "learning_rate": 4.860940925593703e-06,
+      "logits/chosen": -0.30859681963920593,
+      "logits/rejected": -1.0121045112609863,
+      "logps/chosen": -1.1916425228118896,
+      "logps/rejected": -1.6442506313323975,
+      "loss": 1.2909,
+      "odds_ratio_loss": 0.5468284487724304,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.05958213284611702,
+      "rewards/margins": 0.02263040840625763,
+      "rewards/rejected": -0.08221253752708435,
+      "sft_loss": 1.1916425228118896,
+      "step": 735
+    },
+    {
+      "epoch": 0.592,
+      "grad_norm": 30.679914716790897,
+      "learning_rate": 4.857088831287158e-06,
+      "logits/chosen": -0.3152746558189392,
+      "logits/rejected": -0.498546838760376,
+      "logps/chosen": -1.2811144590377808,
+      "logps/rejected": -1.4738355875015259,
+      "loss": 1.242,
+      "odds_ratio_loss": 0.6398525834083557,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.06405572593212128,
+      "rewards/margins": 0.009636052884161472,
+      "rewards/rejected": -0.07369177788496017,
+      "sft_loss": 1.2811144590377808,
+      "step": 740
+    },
+    {
+      "epoch": 0.596,
+      "grad_norm": 7.06148057428711,
+      "learning_rate": 4.85318567856128e-06,
+      "logits/chosen": -0.6131590604782104,
+      "logits/rejected": -0.22784848511219025,
+      "logps/chosen": -1.190834879875183,
+      "logps/rejected": -1.4315545558929443,
+      "loss": 1.211,
+      "odds_ratio_loss": 0.6610680818557739,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.059541743248701096,
+      "rewards/margins": 0.012035989202558994,
+      "rewards/rejected": -0.07157773524522781,
+      "sft_loss": 1.190834879875183,
+      "step": 745
+    },
+    {
+      "epoch": 0.6,
+      "grad_norm": 10.554654256963408,
+      "learning_rate": 4.849231551964771e-06,
+      "logits/chosen": -0.30865949392318726,
+      "logits/rejected": -0.6756810545921326,
+      "logps/chosen": -0.9886703491210938,
+      "logps/rejected": -1.524072289466858,
+      "loss": 1.149,
+      "odds_ratio_loss": 0.4154701828956604,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.049433521926403046,
+      "rewards/margins": 0.0267700906842947,
+      "rewards/rejected": -0.0762036144733429,
+      "sft_loss": 0.9886703491210938,
+      "step": 750
+    },
+    {
+      "epoch": 0.604,
+      "grad_norm": 8.5013064131026,
+      "learning_rate": 4.8452265371505176e-06,
+      "logits/chosen": -0.3945849537849426,
+      "logits/rejected": -0.6222106218338013,
+      "logps/chosen": -1.2622836828231812,
+      "logps/rejected": -1.3229162693023682,
+      "loss": 1.2813,
+      "odds_ratio_loss": 0.7592854499816895,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.06311418116092682,
+      "rewards/margins": 0.0030316333286464214,
+      "rewards/rejected": -0.06614581495523453,
+      "sft_loss": 1.2622836828231812,
+      "step": 755
+    },
+    {
+      "epoch": 0.608,
+      "grad_norm": 6.339074653596066,
+      "learning_rate": 4.841170720873723e-06,
+      "logits/chosen": -0.6454433798789978,
+      "logits/rejected": -0.5933985710144043,
+      "logps/chosen": -1.0739322900772095,
+      "logps/rejected": -1.2623248100280762,
+      "loss": 1.2882,
+      "odds_ratio_loss": 0.6267717480659485,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.053696613758802414,
+      "rewards/margins": 0.009419633075594902,
+      "rewards/rejected": -0.06311623752117157,
+      "sft_loss": 1.0739322900772095,
+      "step": 760
+    },
+    {
+      "epoch": 0.612,
+      "grad_norm": 9.528245360044645,
+      "learning_rate": 4.837064190990036e-06,
+      "logits/chosen": -0.5372971296310425,
+      "logits/rejected": -0.4304935932159424,
+      "logps/chosen": -1.3918521404266357,
+      "logps/rejected": -1.2833397388458252,
+      "loss": 1.2469,
+      "odds_ratio_loss": 0.8755296468734741,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.06959259510040283,
+      "rewards/margins": -0.005425615236163139,
+      "rewards/rejected": -0.06416698545217514,
+      "sft_loss": 1.3918521404266357,
+      "step": 765
+    },
+    {
+      "epoch": 0.616,
+      "grad_norm": 45.4719267607946,
+      "learning_rate": 4.832907036453647e-06,
+      "logits/chosen": -0.6772319078445435,
+      "logits/rejected": -0.29380422830581665,
+      "logps/chosen": -1.3806759119033813,
+      "logps/rejected": -1.4640812873840332,
+      "loss": 1.2546,
+      "odds_ratio_loss": 0.7038690447807312,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.06903379410505295,
+      "rewards/margins": 0.00417026923969388,
+      "rewards/rejected": -0.07320406287908554,
+      "sft_loss": 1.3806759119033813,
+      "step": 770
+    },
+    {
+      "epoch": 0.62,
+      "grad_norm": 6.67487476307786,
+      "learning_rate": 4.828699347315357e-06,
+      "logits/chosen": -0.43924084305763245,
+      "logits/rejected": -0.685263991355896,
+      "logps/chosen": -1.3941123485565186,
+      "logps/rejected": -1.5404117107391357,
+      "loss": 1.3004,
+      "odds_ratio_loss": 0.6843758821487427,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.06970562040805817,
+      "rewards/margins": 0.0073149665258824825,
+      "rewards/rejected": -0.07702059298753738,
+      "sft_loss": 1.3941123485565186,
+      "step": 775
+    },
+    {
+      "epoch": 0.624,
+      "grad_norm": 8.063952217109858,
+      "learning_rate": 4.824441214720629e-06,
+      "logits/chosen": -0.4597851634025574,
+      "logits/rejected": -0.7097247242927551,
+      "logps/chosen": -1.4838628768920898,
+      "logps/rejected": -1.552299976348877,
+      "loss": 1.3045,
+      "odds_ratio_loss": 0.7198641300201416,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.07419314980506897,
+      "rewards/margins": 0.0034218481741845608,
+      "rewards/rejected": -0.07761499285697937,
+      "sft_loss": 1.4838628768920898,
+      "step": 780
+    },
+    {
+      "epoch": 0.628,
+      "grad_norm": 8.186475314952034,
+      "learning_rate": 4.8201327309076176e-06,
+      "logits/chosen": -0.6055313944816589,
+      "logits/rejected": -0.5660229325294495,
+      "logps/chosen": -1.3449480533599854,
+      "logps/rejected": -1.3947374820709229,
+      "loss": 1.2475,
+      "odds_ratio_loss": 0.701248824596405,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.0672474056482315,
+      "rewards/margins": 0.0024894648231565952,
+      "rewards/rejected": -0.06973687559366226,
+      "sft_loss": 1.3449480533599854,
+      "step": 785
+    },
+    {
+      "epoch": 0.632,
+      "grad_norm": 11.28386887635978,
+      "learning_rate": 4.815773989205165e-06,
+      "logits/chosen": -0.4913761615753174,
+      "logits/rejected": -0.7349493503570557,
+      "logps/chosen": -1.3903801441192627,
+      "logps/rejected": -1.261516809463501,
+      "loss": 1.2877,
+      "odds_ratio_loss": 0.858357310295105,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.06951900571584702,
+      "rewards/margins": -0.006443170364946127,
+      "rewards/rejected": -0.06307584047317505,
+      "sft_loss": 1.3903801441192627,
+      "step": 790
+    },
+    {
+      "epoch": 0.636,
+      "grad_norm": 15.974062267930158,
+      "learning_rate": 4.811365084030784e-06,
+      "logits/chosen": -0.6103811264038086,
+      "logits/rejected": -0.6908004879951477,
+      "logps/chosen": -1.1836979389190674,
+      "logps/rejected": -1.4412634372711182,
+      "loss": 1.2186,
+      "odds_ratio_loss": 0.599416971206665,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.05918489769101143,
+      "rewards/margins": 0.012878276407718658,
+      "rewards/rejected": -0.07206316292285919,
+      "sft_loss": 1.1836979389190674,
+      "step": 795
+    },
+    {
+      "epoch": 0.64,
+      "grad_norm": 9.332345967135447,
+      "learning_rate": 4.806906110888606e-06,
+      "logits/chosen": -0.10099928081035614,
+      "logits/rejected": -0.21330437064170837,
+      "logps/chosen": -0.9208562970161438,
+      "logps/rejected": -1.2308391332626343,
+      "loss": 1.2217,
+      "odds_ratio_loss": 0.5371675491333008,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.04604281857609749,
+      "rewards/margins": 0.01549914013594389,
+      "rewards/rejected": -0.061541955918073654,
+      "sft_loss": 0.9208562970161438,
+      "step": 800
+    },
+    {
+      "epoch": 0.644,
+      "grad_norm": 6.712188908944328,
+      "learning_rate": 4.8023971663673235e-06,
+      "logits/chosen": -0.45073431730270386,
+      "logits/rejected": -0.6796762347221375,
+      "logps/chosen": -1.2248754501342773,
+      "logps/rejected": -1.1859427690505981,
+      "loss": 1.2317,
+      "odds_ratio_loss": 0.9216899871826172,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.06124377250671387,
+      "rewards/margins": -0.0019466314697638154,
+      "rewards/rejected": -0.059297144412994385,
+      "sft_loss": 1.2248754501342773,
+      "step": 805
+    },
+    {
+      "epoch": 0.648,
+      "grad_norm": 6.089598640062157,
+      "learning_rate": 4.7978383481380865e-06,
+      "logits/chosen": -0.7606781125068665,
+      "logits/rejected": -0.34865307807922363,
+      "logps/chosen": -1.3903577327728271,
+      "logps/rejected": -1.3126481771469116,
+      "loss": 1.2731,
+      "odds_ratio_loss": 0.8095407485961914,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": -0.06951788067817688,
+      "rewards/margins": -0.003885473357513547,
+      "rewards/rejected": -0.0656324103474617,
+      "sft_loss": 1.3903577327728271,
+      "step": 810
+    },
+    {
+      "epoch": 0.652,
+      "grad_norm": 8.775544099907432,
+      "learning_rate": 4.793229754952393e-06,
+      "logits/chosen": -0.9465087652206421,
+      "logits/rejected": -0.8528397679328918,
+      "logps/chosen": -1.3370808362960815,
+      "logps/rejected": -1.6353280544281006,
+      "loss": 1.2299,
+      "odds_ratio_loss": 0.5765669941902161,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.06685404479503632,
+      "rewards/margins": 0.01491236686706543,
+      "rewards/rejected": -0.08176640421152115,
+      "sft_loss": 1.3370808362960815,
+      "step": 815
+    },
+    {
+      "epoch": 0.656,
+      "grad_norm": 9.946349607495078,
+      "learning_rate": 4.788571486639948e-06,
+      "logits/chosen": -0.538291335105896,
+      "logits/rejected": -0.8991764783859253,
+      "logps/chosen": -1.3249986171722412,
+      "logps/rejected": -1.2387126684188843,
+      "loss": 1.1861,
+      "odds_ratio_loss": 0.8351901173591614,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.06624993681907654,
+      "rewards/margins": -0.004314306192100048,
+      "rewards/rejected": -0.061935633420944214,
+      "sft_loss": 1.3249986171722412,
+      "step": 820
+    },
+    {
+      "epoch": 0.66,
+      "grad_norm": 13.684922078518843,
+      "learning_rate": 4.783863644106502e-06,
+      "logits/chosen": -0.40957608819007874,
+      "logits/rejected": -1.0564932823181152,
+      "logps/chosen": -1.2083319425582886,
+      "logps/rejected": -1.3241065740585327,
+      "loss": 1.2727,
+      "odds_ratio_loss": 0.7248133420944214,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.06041660159826279,
+      "rewards/margins": 0.005788723472505808,
+      "rewards/rejected": -0.06620533019304276,
+      "sft_loss": 1.2083319425582886,
+      "step": 825
+    },
+    {
+      "epoch": 0.664,
+      "grad_norm": 14.240624119385929,
+      "learning_rate": 4.779106329331665e-06,
+      "logits/chosen": -0.6208971738815308,
+      "logits/rejected": -0.53315269947052,
+      "logps/chosen": -1.327461838722229,
+      "logps/rejected": -1.1612827777862549,
+      "loss": 1.3201,
+      "odds_ratio_loss": 0.9040085077285767,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.06637309491634369,
+      "rewards/margins": -0.008308951742947102,
+      "rewards/rejected": -0.058064140379428864,
+      "sft_loss": 1.327461838722229,
+      "step": 830
+    },
+    {
+      "epoch": 0.668,
+      "grad_norm": 6.175231699145468,
+      "learning_rate": 4.774299645366696e-06,
+      "logits/chosen": -0.5456127524375916,
+      "logits/rejected": -0.816623330116272,
+      "logps/chosen": -1.1675437688827515,
+      "logps/rejected": -1.4906387329101562,
+      "loss": 1.2384,
+      "odds_ratio_loss": 0.5253651142120361,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.058377187699079514,
+      "rewards/margins": 0.016154741868376732,
+      "rewards/rejected": -0.0745319351553917,
+      "sft_loss": 1.1675437688827515,
+      "step": 835
+    },
+    {
+      "epoch": 0.672,
+      "grad_norm": 9.640667146514005,
+      "learning_rate": 4.769443696332272e-06,
+      "logits/chosen": -0.2830110788345337,
+      "logits/rejected": -0.8570443391799927,
+      "logps/chosen": -1.2739179134368896,
+      "logps/rejected": -1.4723364114761353,
+      "loss": 1.223,
+      "odds_ratio_loss": 0.7033334374427795,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.06369589269161224,
+      "rewards/margins": 0.009920930489897728,
+      "rewards/rejected": -0.07361682504415512,
+      "sft_loss": 1.2739179134368896,
+      "step": 840
+    },
+    {
+      "epoch": 0.676,
+      "grad_norm": 12.645968569930028,
+      "learning_rate": 4.764538587416233e-06,
+      "logits/chosen": -0.8710149526596069,
+      "logits/rejected": -0.8611608743667603,
+      "logps/chosen": -1.452150583267212,
+      "logps/rejected": -2.0592668056488037,
+      "loss": 1.3282,
+      "odds_ratio_loss": 0.6635312438011169,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.07260753214359283,
+      "rewards/margins": 0.03035581484436989,
+      "rewards/rejected": -0.10296335071325302,
+      "sft_loss": 1.452150583267212,
+      "step": 845
+    },
+    {
+      "epoch": 0.68,
+      "grad_norm": 9.432527411843669,
+      "learning_rate": 4.759584424871302e-06,
+      "logits/chosen": -0.5329976081848145,
+      "logits/rejected": -0.8772062063217163,
+      "logps/chosen": -0.9151167869567871,
+      "logps/rejected": -1.6539852619171143,
+      "loss": 1.2911,
+      "odds_ratio_loss": 0.4387364387512207,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.045755840837955475,
+      "rewards/margins": 0.03694342449307442,
+      "rewards/rejected": -0.082699254155159,
+      "sft_loss": 0.9151167869567871,
+      "step": 850
+    },
+    {
+      "epoch": 0.684,
+      "grad_norm": 7.726661197517004,
+      "learning_rate": 4.754581316012785e-06,
+      "logits/chosen": -0.4633726477622986,
+      "logits/rejected": -0.5861400961875916,
+      "logps/chosen": -0.6932090520858765,
+      "logps/rejected": -1.0552204847335815,
+      "loss": 1.1924,
+      "odds_ratio_loss": 0.4030598998069763,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.034660451114177704,
+      "rewards/margins": 0.018100565299391747,
+      "rewards/rejected": -0.0527610182762146,
+      "sft_loss": 0.6932090520858765,
+      "step": 855
+    },
+    {
+      "epoch": 0.688,
+      "grad_norm": 9.894504014422436,
+      "learning_rate": 4.749529369216246e-06,
+      "logits/chosen": -0.42406773567199707,
+      "logits/rejected": -0.4842946529388428,
+      "logps/chosen": -1.0714144706726074,
+      "logps/rejected": -1.1619442701339722,
+      "loss": 1.352,
+      "odds_ratio_loss": 0.6463958024978638,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05357072502374649,
+      "rewards/margins": 0.0045264954678714275,
+      "rewards/rejected": -0.05809721350669861,
+      "sft_loss": 1.0714144706726074,
+      "step": 860
+    },
+    {
+      "epoch": 0.692,
+      "grad_norm": 9.153094977007724,
+      "learning_rate": 4.744428693915158e-06,
+      "logits/chosen": -0.2847048342227936,
+      "logits/rejected": -0.4418734610080719,
+      "logps/chosen": -1.2593896389007568,
+      "logps/rejected": -0.9671368598937988,
+      "loss": 1.3217,
+      "odds_ratio_loss": 0.9432659149169922,
+      "rewards/accuracies": 0.10000000149011612,
+      "rewards/chosen": -0.06296949088573456,
+      "rewards/margins": -0.014612642116844654,
+      "rewards/rejected": -0.04835684597492218,
+      "sft_loss": 1.2593896389007568,
+      "step": 865
+    },
+    {
+      "epoch": 0.696,
+      "grad_norm": 7.840444386866316,
+      "learning_rate": 4.7392794005985324e-06,
+      "logits/chosen": -0.4391769468784332,
+      "logits/rejected": -0.9956706762313843,
+      "logps/chosen": -1.0819652080535889,
+      "logps/rejected": -1.3912889957427979,
+      "loss": 1.1336,
+      "odds_ratio_loss": 0.5440673232078552,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.054098259657621384,
+      "rewards/margins": 0.01546618901193142,
+      "rewards/rejected": -0.06956445425748825,
+      "sft_loss": 1.0819652080535889,
+      "step": 870
+    },
+    {
+      "epoch": 0.7,
+      "grad_norm": 12.411494501286214,
+      "learning_rate": 4.734081600808531e-06,
+      "logits/chosen": -0.9742234349250793,
+      "logits/rejected": -0.4317501485347748,
+      "logps/chosen": -1.1081851720809937,
+      "logps/rejected": -1.167130947113037,
+      "loss": 1.2256,
+      "odds_ratio_loss": 0.6975724697113037,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.0554092600941658,
+      "rewards/margins": 0.002947291126474738,
+      "rewards/rejected": -0.058356545865535736,
+      "sft_loss": 1.1081851720809937,
+      "step": 875
+    },
+    {
+      "epoch": 0.704,
+      "grad_norm": 8.306111344864336,
+      "learning_rate": 4.7288354071380415e-06,
+      "logits/chosen": -0.5642324686050415,
+      "logits/rejected": -0.8995448350906372,
+      "logps/chosen": -1.0954984426498413,
+      "logps/rejected": -1.2880576848983765,
+      "loss": 1.2105,
+      "odds_ratio_loss": 0.6666485071182251,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.0547749288380146,
+      "rewards/margins": 0.009627962484955788,
+      "rewards/rejected": -0.06440288573503494,
+      "sft_loss": 1.0954984426498413,
+      "step": 880
+    },
+    {
+      "epoch": 0.708,
+      "grad_norm": 10.907946303534318,
+      "learning_rate": 4.723540933228245e-06,
+      "logits/chosen": -0.8124354481697083,
+      "logits/rejected": -0.5376805067062378,
+      "logps/chosen": -1.105232834815979,
+      "logps/rejected": -1.4322589635849,
+      "loss": 1.2672,
+      "odds_ratio_loss": 0.5457401871681213,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05526164174079895,
+      "rewards/margins": 0.016351303085684776,
+      "rewards/rejected": -0.07161294668912888,
+      "sft_loss": 1.105232834815979,
+      "step": 885
+    },
+    {
+      "epoch": 0.712,
+      "grad_norm": 11.344492337732651,
+      "learning_rate": 4.7181982937661485e-06,
+      "logits/chosen": -0.6914325952529907,
+      "logits/rejected": -0.9896718859672546,
+      "logps/chosen": -1.3581817150115967,
+      "logps/rejected": -1.8004964590072632,
+      "loss": 1.2577,
+      "odds_ratio_loss": 0.5711563229560852,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.06790908426046371,
+      "rewards/margins": 0.02211574651300907,
+      "rewards/rejected": -0.09002482891082764,
+      "sft_loss": 1.3581817150115967,
+      "step": 890
+    },
+    {
+      "epoch": 0.716,
+      "grad_norm": 14.944195714469451,
+      "learning_rate": 4.712807604482108e-06,
+      "logits/chosen": -0.580794632434845,
+      "logits/rejected": -0.6542994379997253,
+      "logps/chosen": -1.1799992322921753,
+      "logps/rejected": -1.1197493076324463,
+      "loss": 1.2151,
+      "odds_ratio_loss": 0.8679493069648743,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.058999963104724884,
+      "rewards/margins": -0.0030124965123832226,
+      "rewards/rejected": -0.05598746985197067,
+      "sft_loss": 1.1799992322921753,
+      "step": 895
+    },
+    {
+      "epoch": 0.72,
+      "grad_norm": 6.69431914838341,
+      "learning_rate": 4.707368982147318e-06,
+      "logits/chosen": -0.26036280393600464,
+      "logits/rejected": -0.8896979093551636,
+      "logps/chosen": -1.174317479133606,
+      "logps/rejected": -0.9866326451301575,
+      "loss": 1.2563,
+      "odds_ratio_loss": 1.064286470413208,
+      "rewards/accuracies": 0.20000000298023224,
+      "rewards/chosen": -0.058715879917144775,
+      "rewards/margins": -0.009384247474372387,
+      "rewards/rejected": -0.049331631511449814,
+      "sft_loss": 1.174317479133606,
+      "step": 900
+    },
+    {
+      "epoch": 0.724,
+      "grad_norm": 4.7835304167015975,
+      "learning_rate": 4.701882544571277e-06,
+      "logits/chosen": -0.9345399141311646,
+      "logits/rejected": -0.6917155385017395,
+      "logps/chosen": -1.392803430557251,
+      "logps/rejected": -1.371321439743042,
+      "loss": 1.1668,
+      "odds_ratio_loss": 0.7409270405769348,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.06964017450809479,
+      "rewards/margins": -0.001074108062312007,
+      "rewards/rejected": -0.06856606900691986,
+      "sft_loss": 1.392803430557251,
+      "step": 905
+    },
+    {
+      "epoch": 0.728,
+      "grad_norm": 11.02550077384499,
+      "learning_rate": 4.696348410599244e-06,
+      "logits/chosen": -0.5705204010009766,
+      "logits/rejected": -0.6748573184013367,
+      "logps/chosen": -1.0039920806884766,
+      "logps/rejected": -1.592974066734314,
+      "loss": 1.2679,
+      "odds_ratio_loss": 0.5323291420936584,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05019960552453995,
+      "rewards/margins": 0.029449105262756348,
+      "rewards/rejected": -0.0796487107872963,
+      "sft_loss": 1.0039920806884766,
+      "step": 910
+    },
+    {
+      "epoch": 0.732,
+      "grad_norm": 10.692588852434419,
+      "learning_rate": 4.690766700109659e-06,
+      "logits/chosen": -0.4835734963417053,
+      "logits/rejected": -0.7319347858428955,
+      "logps/chosen": -1.1647623777389526,
+      "logps/rejected": -2.5557546615600586,
+      "loss": 1.2441,
+      "odds_ratio_loss": 0.41198450326919556,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05823811888694763,
+      "rewards/margins": 0.06954962015151978,
+      "rewards/rejected": -0.1277877390384674,
+      "sft_loss": 1.1647623777389526,
+      "step": 915
+    },
+    {
+      "epoch": 0.736,
+      "grad_norm": 10.396971419498014,
+      "learning_rate": 4.685137534011549e-06,
+      "logits/chosen": -0.8939110040664673,
+      "logits/rejected": -0.3123689293861389,
+      "logps/chosen": -1.2236140966415405,
+      "logps/rejected": -1.3663610219955444,
+      "loss": 1.1922,
+      "odds_ratio_loss": 0.7821038365364075,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.06118069961667061,
+      "rewards/margins": 0.007137349806725979,
+      "rewards/rejected": -0.06831805408000946,
+      "sft_loss": 1.2236140966415405,
+      "step": 920
+    },
+    {
+      "epoch": 0.74,
+      "grad_norm": 8.738002331287573,
+      "learning_rate": 4.679461034241906e-06,
+      "logits/chosen": -0.6164524555206299,
+      "logits/rejected": -0.7329466938972473,
+      "logps/chosen": -1.4917128086090088,
+      "logps/rejected": -1.479864478111267,
+      "loss": 1.2453,
+      "odds_ratio_loss": 0.770171046257019,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.07458565384149551,
+      "rewards/margins": -0.0005924153956584632,
+      "rewards/rejected": -0.07399322092533112,
+      "sft_loss": 1.4917128086090088,
+      "step": 925
+    },
+    {
+      "epoch": 0.744,
+      "grad_norm": 23.09910613175578,
+      "learning_rate": 4.673737323763048e-06,
+      "logits/chosen": -0.6520587205886841,
+      "logits/rejected": -0.7833204865455627,
+      "logps/chosen": -1.1136635541915894,
+      "logps/rejected": -1.1649823188781738,
+      "loss": 1.1591,
+      "odds_ratio_loss": 0.724600613117218,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.05568317696452141,
+      "rewards/margins": 0.0025659373495727777,
+      "rewards/rejected": -0.05824911594390869,
+      "sft_loss": 1.1136635541915894,
+      "step": 930
+    },
+    {
+      "epoch": 0.748,
+      "grad_norm": 9.576889593824372,
+      "learning_rate": 4.667966526559953e-06,
+      "logits/chosen": -0.7509557008743286,
+      "logits/rejected": -0.49857956171035767,
+      "logps/chosen": -1.3980618715286255,
+      "logps/rejected": -1.2008607387542725,
+      "loss": 1.3096,
+      "odds_ratio_loss": 0.9694040417671204,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": -0.06990310549736023,
+      "rewards/margins": -0.009860062971711159,
+      "rewards/rejected": -0.06004303693771362,
+      "sft_loss": 1.3980618715286255,
+      "step": 935
+    },
+    {
+      "epoch": 0.752,
+      "grad_norm": 8.475227307177539,
+      "learning_rate": 4.662148767637578e-06,
+      "logits/chosen": -0.5015154480934143,
+      "logits/rejected": -0.7089800834655762,
+      "logps/chosen": -1.1982038021087646,
+      "logps/rejected": -1.4972496032714844,
+      "loss": 1.2186,
+      "odds_ratio_loss": 0.5610604286193848,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.059910185635089874,
+      "rewards/margins": 0.014952296391129494,
+      "rewards/rejected": -0.07486248016357422,
+      "sft_loss": 1.1982038021087646,
+      "step": 940
+    },
+    {
+      "epoch": 0.756,
+      "grad_norm": 10.570484526302579,
+      "learning_rate": 4.656284173018144e-06,
+      "logits/chosen": -0.6629132628440857,
+      "logits/rejected": -0.8142105340957642,
+      "logps/chosen": -1.2457422018051147,
+      "logps/rejected": -1.447788953781128,
+      "loss": 1.2879,
+      "odds_ratio_loss": 0.5766414403915405,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.0622871108353138,
+      "rewards/margins": 0.010102340951561928,
+      "rewards/rejected": -0.07238946110010147,
+      "sft_loss": 1.2457422018051147,
+      "step": 945
+    },
+    {
+      "epoch": 0.76,
+      "grad_norm": 8.792884078555856,
+      "learning_rate": 4.650372869738415e-06,
+      "logits/chosen": -0.38089218735694885,
+      "logits/rejected": -0.724978506565094,
+      "logps/chosen": -1.0249511003494263,
+      "logps/rejected": -1.4063374996185303,
+      "loss": 1.3351,
+      "odds_ratio_loss": 0.5218092799186707,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05124755948781967,
+      "rewards/margins": 0.01906932331621647,
+      "rewards/rejected": -0.07031688839197159,
+      "sft_loss": 1.0249511003494263,
+      "step": 950
+    },
+    {
+      "epoch": 0.764,
+      "grad_norm": 9.019548926117617,
+      "learning_rate": 4.644414985846934e-06,
+      "logits/chosen": -0.5280020833015442,
+      "logits/rejected": -1.183532953262329,
+      "logps/chosen": -1.1438463926315308,
+      "logps/rejected": -1.5815012454986572,
+      "loss": 1.311,
+      "odds_ratio_loss": 0.4890708029270172,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.057192325592041016,
+      "rewards/margins": 0.02188274636864662,
+      "rewards/rejected": -0.07907506823539734,
+      "sft_loss": 1.1438463926315308,
+      "step": 955
+    },
+    {
+      "epoch": 0.768,
+      "grad_norm": 10.010010873228056,
+      "learning_rate": 4.638410650401267e-06,
+      "logits/chosen": -0.6906821131706238,
+      "logits/rejected": -0.42066025733947754,
+      "logps/chosen": -0.9353683590888977,
+      "logps/rejected": -1.1939796209335327,
+      "loss": 1.2082,
+      "odds_ratio_loss": 0.6122977137565613,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.046768419444561005,
+      "rewards/margins": 0.012930569238960743,
+      "rewards/rejected": -0.059698980301618576,
+      "sft_loss": 0.9353683590888977,
+      "step": 960
+    },
+    {
+      "epoch": 0.772,
+      "grad_norm": 13.636523513835607,
+      "learning_rate": 4.632359993465188e-06,
+      "logits/chosen": -0.7019423842430115,
+      "logits/rejected": -0.8399526476860046,
+      "logps/chosen": -1.0443174839019775,
+      "logps/rejected": -1.1119390726089478,
+      "loss": 1.2281,
+      "odds_ratio_loss": 0.7568255066871643,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.05221586674451828,
+      "rewards/margins": 0.0033810839522629976,
+      "rewards/rejected": -0.055596958845853806,
+      "sft_loss": 1.0443174839019775,
+      "step": 965
+    },
+    {
+      "epoch": 0.776,
+      "grad_norm": 10.680468517829564,
+      "learning_rate": 4.626263146105875e-06,
+      "logits/chosen": -0.7569029927253723,
+      "logits/rejected": -1.1442753076553345,
+      "logps/chosen": -1.0383312702178955,
+      "logps/rejected": -1.2682640552520752,
+      "loss": 1.2608,
+      "odds_ratio_loss": 0.715472936630249,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.051916562020778656,
+      "rewards/margins": 0.011496638879179955,
+      "rewards/rejected": -0.06341320276260376,
+      "sft_loss": 1.0383312702178955,
+      "step": 970
+    },
+    {
+      "epoch": 0.78,
+      "grad_norm": 16.270504514158986,
+      "learning_rate": 4.620120240391065e-06,
+      "logits/chosen": -0.5434718728065491,
+      "logits/rejected": -0.8585309982299805,
+      "logps/chosen": -1.0939325094223022,
+      "logps/rejected": -1.3599112033843994,
+      "loss": 1.2639,
+      "odds_ratio_loss": 0.7648831605911255,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": -0.05469663068652153,
+      "rewards/margins": 0.01329893060028553,
+      "rewards/rejected": -0.06799556314945221,
+      "sft_loss": 1.0939325094223022,
+      "step": 975
+    },
+    {
+      "epoch": 0.784,
+      "grad_norm": 5.804422084174317,
+      "learning_rate": 4.613931409386196e-06,
+      "logits/chosen": -0.7977913022041321,
+      "logits/rejected": -0.9274693727493286,
+      "logps/chosen": -1.18310546875,
+      "logps/rejected": -1.8053483963012695,
+      "loss": 1.2574,
+      "odds_ratio_loss": 0.4495466351509094,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.059155285358428955,
+      "rewards/margins": 0.031112130731344223,
+      "rewards/rejected": -0.09026740491390228,
+      "sft_loss": 1.18310546875,
+      "step": 980
+    },
+    {
+      "epoch": 0.788,
+      "grad_norm": 9.788416898738078,
+      "learning_rate": 4.607696787151522e-06,
+      "logits/chosen": -0.9438807368278503,
+      "logits/rejected": -0.6228753328323364,
+      "logps/chosen": -1.5850467681884766,
+      "logps/rejected": -1.510814905166626,
+      "loss": 1.1717,
+      "odds_ratio_loss": 0.8166835904121399,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.07925233989953995,
+      "rewards/margins": -0.0037115835584700108,
+      "rewards/rejected": -0.07554075121879578,
+      "sft_loss": 1.5850467681884766,
+      "step": 985
+    },
+    {
+      "epoch": 0.792,
+      "grad_norm": 12.55818469153721,
+      "learning_rate": 4.601416508739211e-06,
+      "logits/chosen": -0.731402575969696,
+      "logits/rejected": -1.0273559093475342,
+      "logps/chosen": -1.129149317741394,
+      "logps/rejected": -1.219930648803711,
+      "loss": 1.2695,
+      "odds_ratio_loss": 0.6738437414169312,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05645746737718582,
+      "rewards/margins": 0.0045390622690320015,
+      "rewards/rejected": -0.06099652498960495,
+      "sft_loss": 1.129149317741394,
+      "step": 990
+    },
+    {
+      "epoch": 0.796,
+      "grad_norm": 9.153552767046978,
+      "learning_rate": 4.595090710190419e-06,
+      "logits/chosen": -0.7405273914337158,
+      "logits/rejected": -1.1718242168426514,
+      "logps/chosen": -1.2145724296569824,
+      "logps/rejected": -1.4828448295593262,
+      "loss": 1.1849,
+      "odds_ratio_loss": 0.5643509030342102,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.06072862073779106,
+      "rewards/margins": 0.013413618318736553,
+      "rewards/rejected": -0.07414223998785019,
+      "sft_loss": 1.2145724296569824,
+      "step": 995
+    },
+    {
+      "epoch": 0.8,
+      "grad_norm": 6.94868849947181,
+      "learning_rate": 4.588719528532342e-06,
+      "logits/chosen": -0.7134780883789062,
+      "logits/rejected": -0.905554473400116,
+      "logps/chosen": -1.4175317287445068,
+      "logps/rejected": -1.3552976846694946,
+      "loss": 1.3012,
+      "odds_ratio_loss": 0.8563035130500793,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": -0.0708765834569931,
+      "rewards/margins": -0.003111699130386114,
+      "rewards/rejected": -0.06776488572359085,
+      "sft_loss": 1.4175317287445068,
+      "step": 1000
+    },
+    {
+      "epoch": 0.804,
+      "grad_norm": 12.921749906218944,
+      "learning_rate": 4.582303101775249e-06,
+      "logits/chosen": -0.7810646891593933,
+      "logits/rejected": -1.1520473957061768,
+      "logps/chosen": -1.2127867937088013,
+      "logps/rejected": -1.4734896421432495,
+      "loss": 1.2652,
+      "odds_ratio_loss": 0.5705010890960693,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.06063934043049812,
+      "rewards/margins": 0.013035142794251442,
+      "rewards/rejected": -0.07367448508739471,
+      "sft_loss": 1.2127867937088013,
+      "step": 1005
+    },
+    {
+      "epoch": 0.808,
+      "grad_norm": 9.197435023909987,
+      "learning_rate": 4.575841568909494e-06,
+      "logits/chosen": -0.6398459672927856,
+      "logits/rejected": -0.5644738078117371,
+      "logps/chosen": -1.268541932106018,
+      "logps/rejected": -1.5668294429779053,
+      "loss": 1.2232,
+      "odds_ratio_loss": 0.6429783701896667,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.06342709064483643,
+      "rewards/margins": 0.014914381317794323,
+      "rewards/rejected": -0.07834147661924362,
+      "sft_loss": 1.268541932106018,
+      "step": 1010
+    },
+    {
+      "epoch": 0.812,
+      "grad_norm": 5.987738310425259,
+      "learning_rate": 4.569335069902502e-06,
+      "logits/chosen": -0.8146150708198547,
+      "logits/rejected": -0.5185267925262451,
+      "logps/chosen": -1.2354018688201904,
+      "logps/rejected": -1.3597724437713623,
+      "loss": 1.2304,
+      "odds_ratio_loss": 0.7136649489402771,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.06177009269595146,
+      "rewards/margins": 0.006218533962965012,
+      "rewards/rejected": -0.06798862665891647,
+      "sft_loss": 1.2354018688201904,
+      "step": 1015
+    },
+    {
+      "epoch": 0.816,
+      "grad_norm": 11.820317098577457,
+      "learning_rate": 4.562783745695738e-06,
+      "logits/chosen": -0.6676985621452332,
+      "logits/rejected": -0.3732604682445526,
+      "logps/chosen": -1.0979827642440796,
+      "logps/rejected": -1.6110618114471436,
+      "loss": 1.2048,
+      "odds_ratio_loss": 0.6128841638565063,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.05489913746714592,
+      "rewards/margins": 0.025653952732682228,
+      "rewards/rejected": -0.0805530920624733,
+      "sft_loss": 1.0979827642440796,
+      "step": 1020
+    },
+    {
+      "epoch": 0.82,
+      "grad_norm": 12.430320372288909,
+      "learning_rate": 4.556187738201656e-06,
+      "logits/chosen": -0.3898774981498718,
+      "logits/rejected": -1.3098582029342651,
+      "logps/chosen": -0.9205626249313354,
+      "logps/rejected": -1.4675109386444092,
+      "loss": 1.2001,
+      "odds_ratio_loss": 0.6006077527999878,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.04602813720703125,
+      "rewards/margins": 0.02734740450978279,
+      "rewards/rejected": -0.07337555289268494,
+      "sft_loss": 0.9205626249313354,
+      "step": 1025
+    },
+    {
+      "epoch": 0.824,
+      "grad_norm": 8.240787854800642,
+      "learning_rate": 4.549547190300622e-06,
+      "logits/chosen": -0.749883234500885,
+      "logits/rejected": -0.5567277669906616,
+      "logps/chosen": -1.1923110485076904,
+      "logps/rejected": -1.448547124862671,
+      "loss": 1.2761,
+      "odds_ratio_loss": 0.5901089906692505,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05961555242538452,
+      "rewards/margins": 0.012811805121600628,
+      "rewards/rejected": -0.07242736220359802,
+      "sft_loss": 1.1923110485076904,
+      "step": 1030
+    },
+    {
+      "epoch": 0.828,
+      "grad_norm": 8.576070028214076,
+      "learning_rate": 4.542862245837821e-06,
+      "logits/chosen": -0.4225188195705414,
+      "logits/rejected": -0.694170355796814,
+      "logps/chosen": -1.115942358970642,
+      "logps/rejected": -1.1966840028762817,
+      "loss": 1.3326,
+      "odds_ratio_loss": 0.7018226385116577,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.055797118693590164,
+      "rewards/margins": 0.004037079401314259,
+      "rewards/rejected": -0.05983419343829155,
+      "sft_loss": 1.115942358970642,
+      "step": 1035
+    },
+    {
+      "epoch": 0.832,
+      "grad_norm": 8.003040390950462,
+      "learning_rate": 4.536133049620143e-06,
+      "logits/chosen": -0.614066481590271,
+      "logits/rejected": -0.9744704365730286,
+      "logps/chosen": -1.1512901782989502,
+      "logps/rejected": -1.686820387840271,
+      "loss": 1.3049,
+      "odds_ratio_loss": 0.5957285761833191,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05756450816988945,
+      "rewards/margins": 0.0267765112221241,
+      "rewards/rejected": -0.08434101194143295,
+      "sft_loss": 1.1512901782989502,
+      "step": 1040
+    },
+    {
+      "epoch": 0.836,
+      "grad_norm": 7.214120995777482,
+      "learning_rate": 4.529359747413038e-06,
+      "logits/chosen": -0.602199912071228,
+      "logits/rejected": -1.299626350402832,
+      "logps/chosen": -1.2445770502090454,
+      "logps/rejected": -1.3810675144195557,
+      "loss": 1.3109,
+      "odds_ratio_loss": 0.7491277456283569,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.062228865921497345,
+      "rewards/margins": 0.00682451855391264,
+      "rewards/rejected": -0.06905338168144226,
+      "sft_loss": 1.2445770502090454,
+      "step": 1045
+    },
+    {
+      "epoch": 0.84,
+      "grad_norm": 7.796605464265155,
+      "learning_rate": 4.522542485937369e-06,
+      "logits/chosen": -0.3931456208229065,
+      "logits/rejected": -0.8093290328979492,
+      "logps/chosen": -1.3564033508300781,
+      "logps/rejected": -1.3207740783691406,
+      "loss": 1.2757,
+      "odds_ratio_loss": 0.8102067708969116,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.06782017648220062,
+      "rewards/margins": -0.0017814624588936567,
+      "rewards/rejected": -0.06603871285915375,
+      "sft_loss": 1.3564033508300781,
+      "step": 1050
+    },
+    {
+      "epoch": 0.844,
+      "grad_norm": 11.038841006173273,
+      "learning_rate": 4.515681412866228e-06,
+      "logits/chosen": -0.8583968281745911,
+      "logits/rejected": -0.5115640163421631,
+      "logps/chosen": -1.3405869007110596,
+      "logps/rejected": -1.1792690753936768,
+      "loss": 1.2427,
+      "odds_ratio_loss": 1.1266233921051025,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.06702934950590134,
+      "rewards/margins": -0.008065891452133656,
+      "rewards/rejected": -0.058963458985090256,
+      "sft_loss": 1.3405869007110596,
+      "step": 1055
+    },
+    {
+      "epoch": 0.848,
+      "grad_norm": 37.58723049143948,
+      "learning_rate": 4.508776676821739e-06,
+      "logits/chosen": -0.6703466773033142,
+      "logits/rejected": -0.9433542490005493,
+      "logps/chosen": -1.2843058109283447,
+      "logps/rejected": -1.6363897323608398,
+      "loss": 1.359,
+      "odds_ratio_loss": 0.5473671555519104,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.0642152950167656,
+      "rewards/margins": 0.017604198306798935,
+      "rewards/rejected": -0.08181948959827423,
+      "sft_loss": 1.2843058109283447,
+      "step": 1060
+    },
+    {
+      "epoch": 0.852,
+      "grad_norm": 12.74608944254742,
+      "learning_rate": 4.501828427371834e-06,
+      "logits/chosen": -0.3732682168483734,
+      "logits/rejected": -0.7298794984817505,
+      "logps/chosen": -1.0062689781188965,
+      "logps/rejected": -1.6333110332489014,
+      "loss": 1.225,
+      "odds_ratio_loss": 0.45508384704589844,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.050313450396060944,
+      "rewards/margins": 0.031352099031209946,
+      "rewards/rejected": -0.08166555315256119,
+      "sft_loss": 1.0062689781188965,
+      "step": 1065
+    },
+    {
+      "epoch": 0.856,
+      "grad_norm": 4.647181122267829,
+      "learning_rate": 4.494836815027022e-06,
+      "logits/chosen": -0.744674026966095,
+      "logits/rejected": -0.5776953101158142,
+      "logps/chosen": -1.1960818767547607,
+      "logps/rejected": -1.277769684791565,
+      "loss": 1.1808,
+      "odds_ratio_loss": 0.6548304557800293,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.05980409309267998,
+      "rewards/margins": 0.0040843975730240345,
+      "rewards/rejected": -0.06388849020004272,
+      "sft_loss": 1.1960818767547607,
+      "step": 1070
+    },
+    {
+      "epoch": 0.86,
+      "grad_norm": 9.112232033291143,
+      "learning_rate": 4.48780199123712e-06,
+      "logits/chosen": -0.48859691619873047,
+      "logits/rejected": -0.8226814270019531,
+      "logps/chosen": -0.9840165972709656,
+      "logps/rejected": -1.3010642528533936,
+      "loss": 1.3173,
+      "odds_ratio_loss": 0.5319587588310242,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.04920082911849022,
+      "rewards/margins": 0.01585238054394722,
+      "rewards/rejected": -0.06505320966243744,
+      "sft_loss": 0.9840165972709656,
+      "step": 1075
+    },
+    {
+      "epoch": 0.864,
+      "grad_norm": 8.357848437646894,
+      "learning_rate": 4.4807241083879774e-06,
+      "logits/chosen": -0.7419109344482422,
+      "logits/rejected": -0.5743356347084045,
+      "logps/chosen": -1.0393415689468384,
+      "logps/rejected": -1.860235571861267,
+      "loss": 1.1895,
+      "odds_ratio_loss": 0.3943214416503906,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.0519670769572258,
+      "rewards/margins": 0.0410446934401989,
+      "rewards/rejected": -0.0930117815732956,
+      "sft_loss": 1.0393415689468384,
+      "step": 1080
+    },
+    {
+      "epoch": 0.868,
+      "grad_norm": 7.001917916478807,
+      "learning_rate": 4.473603319798173e-06,
+      "logits/chosen": -0.8395317196846008,
+      "logits/rejected": -0.760552704334259,
+      "logps/chosen": -1.2476990222930908,
+      "logps/rejected": -1.3389023542404175,
+      "loss": 1.2252,
+      "odds_ratio_loss": 0.6428620219230652,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.0623849555850029,
+      "rewards/margins": 0.004560163244605064,
+      "rewards/rejected": -0.06694512069225311,
+      "sft_loss": 1.2476990222930908,
+      "step": 1085
+    },
+    {
+      "epoch": 0.872,
+      "grad_norm": 7.900019571893691,
+      "learning_rate": 4.466439779715696e-06,
+      "logits/chosen": -0.68291836977005,
+      "logits/rejected": -0.886945903301239,
+      "logps/chosen": -1.321554183959961,
+      "logps/rejected": -1.8369743824005127,
+      "loss": 1.2689,
+      "odds_ratio_loss": 0.5644279718399048,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.06607771664857864,
+      "rewards/margins": 0.02577100321650505,
+      "rewards/rejected": -0.0918487161397934,
+      "sft_loss": 1.321554183959961,
+      "step": 1090
+    },
+    {
+      "epoch": 0.876,
+      "grad_norm": 8.642864900662888,
+      "learning_rate": 4.4592336433146e-06,
+      "logits/chosen": -0.633298397064209,
+      "logits/rejected": -1.161464810371399,
+      "logps/chosen": -1.0371453762054443,
+      "logps/rejected": -1.3207931518554688,
+      "loss": 1.1633,
+      "odds_ratio_loss": 0.5788105726242065,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.051857270300388336,
+      "rewards/margins": 0.01418238878250122,
+      "rewards/rejected": -0.06603965908288956,
+      "sft_loss": 1.0371453762054443,
+      "step": 1095
+    },
+    {
+      "epoch": 0.88,
+      "grad_norm": 9.05386821771176,
+      "learning_rate": 4.451985066691649e-06,
+      "logits/chosen": -0.7766702771186829,
+      "logits/rejected": -0.5547041893005371,
+      "logps/chosen": -1.1104316711425781,
+      "logps/rejected": -1.3220654726028442,
+      "loss": 1.3197,
+      "odds_ratio_loss": 0.6292269229888916,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.055521585047245026,
+      "rewards/margins": 0.010581688955426216,
+      "rewards/rejected": -0.06610327214002609,
+      "sft_loss": 1.1104316711425781,
+      "step": 1100
+    },
+    {
+      "epoch": 0.884,
+      "grad_norm": 6.120618129770535,
+      "learning_rate": 4.444694206862929e-06,
+      "logits/chosen": -0.7643265128135681,
+      "logits/rejected": -0.858683705329895,
+      "logps/chosen": -1.2096381187438965,
+      "logps/rejected": -1.3979839086532593,
+      "loss": 1.1843,
+      "odds_ratio_loss": 0.7661855220794678,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.06048189848661423,
+      "rewards/margins": 0.00941728986799717,
+      "rewards/rejected": -0.06989918649196625,
+      "sft_loss": 1.2096381187438965,
+      "step": 1105
+    },
+    {
+      "epoch": 0.888,
+      "grad_norm": 11.782139121174437,
+      "learning_rate": 4.437361221760449e-06,
+      "logits/chosen": -0.5270341634750366,
+      "logits/rejected": -0.7325950264930725,
+      "logps/chosen": -1.3420865535736084,
+      "logps/rejected": -1.8302757740020752,
+      "loss": 1.2491,
+      "odds_ratio_loss": 0.6886317133903503,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.06710433214902878,
+      "rewards/margins": 0.024409450590610504,
+      "rewards/rejected": -0.09151377528905869,
+      "sft_loss": 1.3420865535736084,
+      "step": 1110
+    },
+    {
+      "epoch": 0.892,
+      "grad_norm": 18.68798830441005,
+      "learning_rate": 4.4299862702287255e-06,
+      "logits/chosen": -0.5396694540977478,
+      "logits/rejected": -0.6563701033592224,
+      "logps/chosen": -1.2778229713439941,
+      "logps/rejected": -1.229236364364624,
+      "loss": 1.2489,
+      "odds_ratio_loss": 0.869365394115448,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.06389115005731583,
+      "rewards/margins": -0.0024293330498039722,
+      "rewards/rejected": -0.06146181747317314,
+      "sft_loss": 1.2778229713439941,
+      "step": 1115
+    },
+    {
+      "epoch": 0.896,
+      "grad_norm": 8.493094923954416,
+      "learning_rate": 4.422569512021332e-06,
+      "logits/chosen": -0.43230313062667847,
+      "logits/rejected": -1.0870712995529175,
+      "logps/chosen": -1.1300268173217773,
+      "logps/rejected": -1.6185880899429321,
+      "loss": 1.186,
+      "odds_ratio_loss": 0.4703669548034668,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.05650133639574051,
+      "rewards/margins": 0.02442805841565132,
+      "rewards/rejected": -0.08092939108610153,
+      "sft_loss": 1.1300268173217773,
+      "step": 1120
+    },
+    {
+      "epoch": 0.9,
+      "grad_norm": 5.267928354352151,
+      "learning_rate": 4.415111107797445e-06,
+      "logits/chosen": -0.5481151342391968,
+      "logits/rejected": -0.9923849105834961,
+      "logps/chosen": -1.230312466621399,
+      "logps/rejected": -1.5415050983428955,
+      "loss": 1.2303,
+      "odds_ratio_loss": 0.5594199299812317,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.061515629291534424,
+      "rewards/margins": 0.015559625811874866,
+      "rewards/rejected": -0.07707525044679642,
+      "sft_loss": 1.230312466621399,
+      "step": 1125
+    },
+    {
+      "epoch": 0.904,
+      "grad_norm": 8.086520043398188,
+      "learning_rate": 4.407611219118363e-06,
+      "logits/chosen": -0.3550862669944763,
+      "logits/rejected": -0.8939773440361023,
+      "logps/chosen": -1.110655665397644,
+      "logps/rejected": -1.5090737342834473,
+      "loss": 1.2223,
+      "odds_ratio_loss": 0.6366047263145447,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.0555327832698822,
+      "rewards/margins": 0.01992090605199337,
+      "rewards/rejected": -0.07545368373394012,
+      "sft_loss": 1.110655665397644,
+      "step": 1130
+    },
+    {
+      "epoch": 0.908,
+      "grad_norm": 8.346110146833315,
+      "learning_rate": 4.4000700084440046e-06,
+      "logits/chosen": -0.468745619058609,
+      "logits/rejected": -1.0615545511245728,
+      "logps/chosen": -1.219507098197937,
+      "logps/rejected": -1.5035933256149292,
+      "loss": 1.2581,
+      "odds_ratio_loss": 0.9526287317276001,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.06097535416483879,
+      "rewards/margins": 0.014204311184585094,
+      "rewards/rejected": -0.07517966628074646,
+      "sft_loss": 1.219507098197937,
+      "step": 1135
+    },
+    {
+      "epoch": 0.912,
+      "grad_norm": 21.962006553169825,
+      "learning_rate": 4.3924876391293915e-06,
+      "logits/chosen": -0.5484793782234192,
+      "logits/rejected": -0.46796178817749023,
+      "logps/chosen": -1.108927845954895,
+      "logps/rejected": -1.0250526666641235,
+      "loss": 1.2947,
+      "odds_ratio_loss": 0.8142274618148804,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.05544638633728027,
+      "rewards/margins": -0.004193754401057959,
+      "rewards/rejected": -0.05125263333320618,
+      "sft_loss": 1.108927845954895,
+      "step": 1140
+    },
+    {
+      "epoch": 0.916,
+      "grad_norm": 15.332345311015494,
+      "learning_rate": 4.384864275421109e-06,
+      "logits/chosen": -0.5269834399223328,
+      "logits/rejected": -0.967495322227478,
+      "logps/chosen": -1.288408875465393,
+      "logps/rejected": -1.6966063976287842,
+      "loss": 1.2474,
+      "odds_ratio_loss": 0.5316591262817383,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.06442044675350189,
+      "rewards/margins": 0.020409878343343735,
+      "rewards/rejected": -0.08483032882213593,
+      "sft_loss": 1.288408875465393,
+      "step": 1145
+    },
+    {
+      "epoch": 0.92,
+      "grad_norm": 10.130919066814135,
+      "learning_rate": 4.377200082453748e-06,
+      "logits/chosen": -0.5598348379135132,
+      "logits/rejected": -0.8047167062759399,
+      "logps/chosen": -1.2670339345932007,
+      "logps/rejected": -1.4398491382598877,
+      "loss": 1.2942,
+      "odds_ratio_loss": 0.6326471567153931,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.06335169821977615,
+      "rewards/margins": 0.008640759624540806,
+      "rewards/rejected": -0.07199246436357498,
+      "sft_loss": 1.2670339345932007,
+      "step": 1150
+    },
+    {
+      "epoch": 0.924,
+      "grad_norm": 6.4155914839593855,
+      "learning_rate": 4.36949522624633e-06,
+      "logits/chosen": -0.8152793645858765,
+      "logits/rejected": -0.6962675452232361,
+      "logps/chosen": -1.3344051837921143,
+      "logps/rejected": -1.4677374362945557,
+      "loss": 1.2375,
+      "odds_ratio_loss": 0.715882420539856,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.06672026216983795,
+      "rewards/margins": 0.006666617002338171,
+      "rewards/rejected": -0.07338687032461166,
+      "sft_loss": 1.3344051837921143,
+      "step": 1155
+    },
+    {
+      "epoch": 0.928,
+      "grad_norm": 7.363210525745847,
+      "learning_rate": 4.361749873698707e-06,
+      "logits/chosen": -0.5409479737281799,
+      "logits/rejected": -0.722708523273468,
+      "logps/chosen": -1.1098625659942627,
+      "logps/rejected": -1.3639353513717651,
+      "loss": 1.2489,
+      "odds_ratio_loss": 0.7049247026443481,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.05549313500523567,
+      "rewards/margins": 0.012703632935881615,
+      "rewards/rejected": -0.06819676607847214,
+      "sft_loss": 1.1098625659942627,
+      "step": 1160
+    },
+    {
+      "epoch": 0.932,
+      "grad_norm": 10.93689863299814,
+      "learning_rate": 4.353964192587949e-06,
+      "logits/chosen": -0.7145729064941406,
+      "logits/rejected": -0.849127471446991,
+      "logps/chosen": -1.3707650899887085,
+      "logps/rejected": -1.4907456636428833,
+      "loss": 1.2284,
+      "odds_ratio_loss": 0.6671477556228638,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.06853825598955154,
+      "rewards/margins": 0.005999024957418442,
+      "rewards/rejected": -0.07453727722167969,
+      "sft_loss": 1.3707650899887085,
+      "step": 1165
+    },
+    {
+      "epoch": 0.936,
+      "grad_norm": 6.2833560952473375,
+      "learning_rate": 4.346138351564711e-06,
+      "logits/chosen": -0.5584260821342468,
+      "logits/rejected": -1.0391467809677124,
+      "logps/chosen": -1.1828453540802002,
+      "logps/rejected": -1.3176859617233276,
+      "loss": 1.2247,
+      "odds_ratio_loss": 0.6230691075325012,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.05914226919412613,
+      "rewards/margins": 0.006742039229720831,
+      "rewards/rejected": -0.0658842995762825,
+      "sft_loss": 1.1828453540802002,
+      "step": 1170
+    },
+    {
+      "epoch": 0.94,
+      "grad_norm": 7.981319588041361,
+      "learning_rate": 4.338272520149572e-06,
+      "logits/chosen": -0.48743829131126404,
+      "logits/rejected": -0.8333004713058472,
+      "logps/chosen": -1.7361185550689697,
+      "logps/rejected": -1.4733507633209229,
+      "loss": 1.2669,
+      "odds_ratio_loss": 0.959330677986145,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.08680592477321625,
+      "rewards/margins": -0.013138381764292717,
+      "rewards/rejected": -0.07366754859685898,
+      "sft_loss": 1.7361185550689697,
+      "step": 1175
+    },
+    {
+      "epoch": 0.944,
+      "grad_norm": 7.109380757572563,
+      "learning_rate": 4.330366868729376e-06,
+      "logits/chosen": -0.22841350734233856,
+      "logits/rejected": -0.941692054271698,
+      "logps/chosen": -1.0154824256896973,
+      "logps/rejected": -1.5618747472763062,
+      "loss": 1.2833,
+      "odds_ratio_loss": 0.5690563321113586,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.050774119794368744,
+      "rewards/margins": 0.02731963060796261,
+      "rewards/rejected": -0.0780937448143959,
+      "sft_loss": 1.0154824256896973,
+      "step": 1180
+    },
+    {
+      "epoch": 0.948,
+      "grad_norm": 13.318206186445765,
+      "learning_rate": 4.322421568553529e-06,
+      "logits/chosen": -0.38624149560928345,
+      "logits/rejected": -1.0594637393951416,
+      "logps/chosen": -1.07871413230896,
+      "logps/rejected": -1.562281847000122,
+      "loss": 1.2659,
+      "odds_ratio_loss": 0.5322721600532532,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.053935706615448,
+      "rewards/margins": 0.02417839504778385,
+      "rewards/rejected": -0.0781140998005867,
+      "sft_loss": 1.07871413230896,
+      "step": 1185
+    },
+    {
+      "epoch": 0.952,
+      "grad_norm": 9.11517027692733,
+      "learning_rate": 4.3144367917302964e-06,
+      "logits/chosen": -0.8500941395759583,
+      "logits/rejected": -0.7091385722160339,
+      "logps/chosen": -1.1739693880081177,
+      "logps/rejected": -1.5754871368408203,
+      "loss": 1.2802,
+      "odds_ratio_loss": 0.6279300451278687,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.058698467910289764,
+      "rewards/margins": 0.02007589302957058,
+      "rewards/rejected": -0.0787743553519249,
+      "sft_loss": 1.1739693880081177,
+      "step": 1190
+    },
+    {
+      "epoch": 0.956,
+      "grad_norm": 7.697301408661275,
+      "learning_rate": 4.30641271122307e-06,
+      "logits/chosen": -0.4521522521972656,
+      "logits/rejected": -0.8235799670219421,
+      "logps/chosen": -1.2677797079086304,
+      "logps/rejected": -1.3189507722854614,
+      "loss": 1.2587,
+      "odds_ratio_loss": 0.691112756729126,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.06338898837566376,
+      "rewards/margins": 0.002558555454015732,
+      "rewards/rejected": -0.06594754755496979,
+      "sft_loss": 1.2677797079086304,
+      "step": 1195
+    },
+    {
+      "epoch": 0.96,
+      "grad_norm": 9.997266919643533,
+      "learning_rate": 4.2983495008466285e-06,
+      "logits/chosen": -0.8174387216567993,
+      "logits/rejected": -0.7150999307632446,
+      "logps/chosen": -0.8913282155990601,
+      "logps/rejected": -1.1885263919830322,
+      "loss": 1.2154,
+      "odds_ratio_loss": 0.6406058073043823,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.04456641152501106,
+      "rewards/margins": 0.014859904535114765,
+      "rewards/rejected": -0.05942631885409355,
+      "sft_loss": 0.8913282155990601,
+      "step": 1200
+    },
+    {
+      "epoch": 0.964,
+      "grad_norm": 5.221880235557841,
+      "learning_rate": 4.290247335263362e-06,
+      "logits/chosen": -1.0014302730560303,
+      "logits/rejected": -0.9239674806594849,
+      "logps/chosen": -1.0359283685684204,
+      "logps/rejected": -1.3768165111541748,
+      "loss": 1.2759,
+      "odds_ratio_loss": 0.6447513103485107,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05179642513394356,
+      "rewards/margins": 0.017044400796294212,
+      "rewards/rejected": -0.06884082406759262,
+      "sft_loss": 1.0359283685684204,
+      "step": 1205
+    },
+    {
+      "epoch": 0.968,
+      "grad_norm": 6.187170258551322,
+      "learning_rate": 4.2821063899795015e-06,
+      "logits/chosen": -0.39787670969963074,
+      "logits/rejected": -1.0457478761672974,
+      "logps/chosen": -1.1205229759216309,
+      "logps/rejected": -1.4367693662643433,
+      "loss": 1.2251,
+      "odds_ratio_loss": 0.5175265073776245,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.0560261532664299,
+      "rewards/margins": 0.015812328085303307,
+      "rewards/rejected": -0.07183848321437836,
+      "sft_loss": 1.1205229759216309,
+      "step": 1210
+    },
+    {
+      "epoch": 0.972,
+      "grad_norm": 5.076678498483747,
+      "learning_rate": 4.273926841341303e-06,
+      "logits/chosen": -0.9177725911140442,
+      "logits/rejected": -0.9428675770759583,
+      "logps/chosen": -1.2810909748077393,
+      "logps/rejected": -1.7306153774261475,
+      "loss": 1.2798,
+      "odds_ratio_loss": 0.6016807556152344,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.06405454128980637,
+      "rewards/margins": 0.02247622422873974,
+      "rewards/rejected": -0.08653075993061066,
+      "sft_loss": 1.2810909748077393,
+      "step": 1215
+    },
+    {
+      "epoch": 0.976,
+      "grad_norm": 6.982415041589051,
+      "learning_rate": 4.265708866531238e-06,
+      "logits/chosen": -0.8390816450119019,
+      "logits/rejected": -0.5559083223342896,
+      "logps/chosen": -0.9034653902053833,
+      "logps/rejected": -1.1532642841339111,
+      "loss": 1.2121,
+      "odds_ratio_loss": 0.5824159383773804,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.0451732762157917,
+      "rewards/margins": 0.01248995028436184,
+      "rewards/rejected": -0.057663220912218094,
+      "sft_loss": 0.9034653902053833,
+      "step": 1220
+    },
+    {
+      "epoch": 0.98,
+      "grad_norm": 8.854119024299864,
+      "learning_rate": 4.257452643564155e-06,
+      "logits/chosen": -0.40446987748146057,
+      "logits/rejected": -0.9087270498275757,
+      "logps/chosen": -0.8603678941726685,
+      "logps/rejected": -1.3758715391159058,
+      "loss": 1.1785,
+      "odds_ratio_loss": 0.416604608297348,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.0430183932185173,
+      "rewards/margins": 0.025775188580155373,
+      "rewards/rejected": -0.06879357993602753,
+      "sft_loss": 0.8603678941726685,
+      "step": 1225
+    },
+    {
+      "epoch": 0.984,
+      "grad_norm": 8.750528387170686,
+      "learning_rate": 4.249158351283414e-06,
+      "logits/chosen": -1.0308736562728882,
+      "logits/rejected": -0.8786516189575195,
+      "logps/chosen": -0.9902039766311646,
+      "logps/rejected": -1.3153424263000488,
+      "loss": 1.2155,
+      "odds_ratio_loss": 0.4901192784309387,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.04951019585132599,
+      "rewards/margins": 0.016256922855973244,
+      "rewards/rejected": -0.06576712429523468,
+      "sft_loss": 0.9902039766311646,
+      "step": 1230
+    },
+    {
+      "epoch": 0.988,
+      "grad_norm": 10.866038696658444,
+      "learning_rate": 4.240826169357024e-06,
+      "logits/chosen": -1.0902023315429688,
+      "logits/rejected": -1.481338620185852,
+      "logps/chosen": -1.419777750968933,
+      "logps/rejected": -1.2114101648330688,
+      "loss": 1.3133,
+      "odds_ratio_loss": 0.9872404932975769,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.07098889350891113,
+      "rewards/margins": -0.01041838526725769,
+      "rewards/rejected": -0.06057050824165344,
+      "sft_loss": 1.419777750968933,
+      "step": 1235
+    },
+    {
+      "epoch": 0.992,
+      "grad_norm": 7.937224210186921,
+      "learning_rate": 4.232456278273743e-06,
+      "logits/chosen": -0.6265432834625244,
+      "logits/rejected": -0.510567843914032,
+      "logps/chosen": -1.0876511335372925,
+      "logps/rejected": -1.4437576532363892,
+      "loss": 1.1532,
+      "odds_ratio_loss": 0.5487557649612427,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.054382555186748505,
+      "rewards/margins": 0.017805328592658043,
+      "rewards/rejected": -0.0721878856420517,
+      "sft_loss": 1.0876511335372925,
+      "step": 1240
+    },
+    {
+      "epoch": 0.996,
+      "grad_norm": 18.05729223024994,
+      "learning_rate": 4.224048859339175e-06,
+      "logits/chosen": -0.4967614710330963,
+      "logits/rejected": -0.7329981327056885,
+      "logps/chosen": -1.0347087383270264,
+      "logps/rejected": -1.6826976537704468,
+      "loss": 1.3198,
+      "odds_ratio_loss": 0.5197392702102661,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.051735442131757736,
+      "rewards/margins": 0.03239942714571953,
+      "rewards/rejected": -0.08413486927747726,
+      "sft_loss": 1.0347087383270264,
+      "step": 1245
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 8.219444479887624,
+      "learning_rate": 4.215604094671835e-06,
+      "logits/chosen": -0.5567103624343872,
+      "logits/rejected": -0.5455074310302734,
+      "logps/chosen": -1.3113296031951904,
+      "logps/rejected": -1.5509188175201416,
+      "loss": 1.1873,
+      "odds_ratio_loss": 0.6092044711112976,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.06556646525859833,
+      "rewards/margins": 0.01197946909815073,
+      "rewards/rejected": -0.07754594087600708,
+      "sft_loss": 1.3113296031951904,
+      "step": 1250
+    },
+    {
+      "epoch": 1.004,
+      "grad_norm": 6.382865998528307,
+      "learning_rate": 4.207122167199209e-06,
+      "logits/chosen": -0.6577066779136658,
+      "logits/rejected": -0.7173200845718384,
+      "logps/chosen": -0.8776922225952148,
+      "logps/rejected": -1.7570756673812866,
+      "loss": 0.8296,
+      "odds_ratio_loss": 0.33190062642097473,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.04388461261987686,
+      "rewards/margins": 0.04396916925907135,
+      "rewards/rejected": -0.08785378187894821,
+      "sft_loss": 0.8776922225952148,
+      "step": 1255
+    },
+    {
+      "epoch": 1.008,
+      "grad_norm": 5.105708309165987,
+      "learning_rate": 4.198603260653792e-06,
+      "logits/chosen": -0.5839263796806335,
+      "logits/rejected": -0.866306483745575,
+      "logps/chosen": -0.8336095809936523,
+      "logps/rejected": -1.26575767993927,
+      "loss": 0.8358,
+      "odds_ratio_loss": 0.46527212858200073,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.041680481284856796,
+      "rewards/margins": 0.021607402712106705,
+      "rewards/rejected": -0.0632878839969635,
+      "sft_loss": 0.8336095809936523,
+      "step": 1260
+    },
+    {
+      "epoch": 1.012,
+      "grad_norm": 4.8302181441829966,
+      "learning_rate": 4.1900475595691044e-06,
+      "logits/chosen": -0.599847674369812,
+      "logits/rejected": -1.0592182874679565,
+      "logps/chosen": -0.5498633980751038,
+      "logps/rejected": -1.3975508213043213,
+      "loss": 0.7752,
+      "odds_ratio_loss": 0.25286543369293213,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.027493169531226158,
+      "rewards/margins": 0.042384374886751175,
+      "rewards/rejected": -0.06987754255533218,
+      "sft_loss": 0.5498633980751038,
+      "step": 1265
+    },
+    {
+      "epoch": 1.016,
+      "grad_norm": 6.056131985323491,
+      "learning_rate": 4.181455249275701e-06,
+      "logits/chosen": -0.2955573499202728,
+      "logits/rejected": -0.5213496088981628,
+      "logps/chosen": -0.7416674494743347,
+      "logps/rejected": -0.9686793088912964,
+      "loss": 0.8138,
+      "odds_ratio_loss": 0.6542133092880249,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.037083376199007034,
+      "rewards/margins": 0.011350592598319054,
+      "rewards/rejected": -0.04843396693468094,
+      "sft_loss": 0.7416674494743347,
+      "step": 1270
+    },
+    {
+      "epoch": 1.02,
+      "grad_norm": 5.064697244932875,
+      "learning_rate": 4.172826515897146e-06,
+      "logits/chosen": -0.6576083302497864,
+      "logits/rejected": -0.812027096748352,
+      "logps/chosen": -0.6591814160346985,
+      "logps/rejected": -1.5505030155181885,
+      "loss": 0.7844,
+      "odds_ratio_loss": 0.30150288343429565,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03295907378196716,
+      "rewards/margins": 0.0445660725235939,
+      "rewards/rejected": -0.07752513885498047,
+      "sft_loss": 0.6591814160346985,
+      "step": 1275
+    },
+    {
+      "epoch": 1.024,
+      "grad_norm": 5.807191452802592,
+      "learning_rate": 4.1641615463459926e-06,
+      "logits/chosen": -0.4745884835720062,
+      "logits/rejected": -0.4976702332496643,
+      "logps/chosen": -0.6141924858093262,
+      "logps/rejected": -1.0756144523620605,
+      "loss": 0.7885,
+      "odds_ratio_loss": 0.458048015832901,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03070961870253086,
+      "rewards/margins": 0.023071100935339928,
+      "rewards/rejected": -0.05378072336316109,
+      "sft_loss": 0.6141924858093262,
+      "step": 1280
+    },
+    {
+      "epoch": 1.028,
+      "grad_norm": 5.371739030260733,
+      "learning_rate": 4.1554605283197255e-06,
+      "logits/chosen": -0.8033881187438965,
+      "logits/rejected": -0.8408336639404297,
+      "logps/chosen": -0.478267103433609,
+      "logps/rejected": -1.3151143789291382,
+      "loss": 0.8676,
+      "odds_ratio_loss": 0.30922359228134155,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.02391335740685463,
+      "rewards/margins": 0.04184236004948616,
+      "rewards/rejected": -0.06575571000576019,
+      "sft_loss": 0.478267103433609,
+      "step": 1285
+    },
+    {
+      "epoch": 1.032,
+      "grad_norm": 6.568180405098848,
+      "learning_rate": 4.146723650296701e-06,
+      "logits/chosen": -0.6499618887901306,
+      "logits/rejected": -0.931594967842102,
+      "logps/chosen": -1.0891497135162354,
+      "logps/rejected": -1.2812997102737427,
+      "loss": 0.9537,
+      "odds_ratio_loss": 0.7874979972839355,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.05445748567581177,
+      "rewards/margins": 0.009607489220798016,
+      "rewards/rejected": -0.06406497955322266,
+      "sft_loss": 1.0891497135162354,
+      "step": 1290
+    },
+    {
+      "epoch": 1.036,
+      "grad_norm": 8.474892928899138,
+      "learning_rate": 4.1379511015320625e-06,
+      "logits/chosen": -0.4418262839317322,
+      "logits/rejected": -0.8559747934341431,
+      "logps/chosen": -0.9219955205917358,
+      "logps/rejected": -1.7327539920806885,
+      "loss": 0.8688,
+      "odds_ratio_loss": 0.3014797568321228,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.04609977453947067,
+      "rewards/margins": 0.04053793102502823,
+      "rewards/rejected": -0.0866376981139183,
+      "sft_loss": 0.9219955205917358,
+      "step": 1295
+    },
+    {
+      "epoch": 1.04,
+      "grad_norm": 5.782082477060897,
+      "learning_rate": 4.129143072053639e-06,
+      "logits/chosen": -0.7809382677078247,
+      "logits/rejected": -0.8675028085708618,
+      "logps/chosen": -0.8315264582633972,
+      "logps/rejected": -1.3996779918670654,
+      "loss": 0.899,
+      "odds_ratio_loss": 0.3759022355079651,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.0415763296186924,
+      "rewards/margins": 0.02840757742524147,
+      "rewards/rejected": -0.06998389959335327,
+      "sft_loss": 0.8315264582633972,
+      "step": 1300
+    },
+    {
+      "epoch": 1.044,
+      "grad_norm": 6.243618609939396,
+      "learning_rate": 4.120299752657828e-06,
+      "logits/chosen": -0.7213522791862488,
+      "logits/rejected": -0.5738562345504761,
+      "logps/chosen": -0.927670955657959,
+      "logps/rejected": -1.4523080587387085,
+      "loss": 0.8292,
+      "odds_ratio_loss": 0.45550140738487244,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.04638354852795601,
+      "rewards/margins": 0.026231860741972923,
+      "rewards/rejected": -0.07261540740728378,
+      "sft_loss": 0.927670955657959,
+      "step": 1305
+    },
+    {
+      "epoch": 1.048,
+      "grad_norm": 9.073750027816232,
+      "learning_rate": 4.111421334905468e-06,
+      "logits/chosen": -0.6499530673027039,
+      "logits/rejected": -0.8254686594009399,
+      "logps/chosen": -0.9249345660209656,
+      "logps/rejected": -1.39388906955719,
+      "loss": 0.7753,
+      "odds_ratio_loss": 0.43550366163253784,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.0462467297911644,
+      "rewards/margins": 0.023447733372449875,
+      "rewards/rejected": -0.06969445943832397,
+      "sft_loss": 0.9249345660209656,
+      "step": 1310
+    },
+    {
+      "epoch": 1.052,
+      "grad_norm": 10.11337598302246,
+      "learning_rate": 4.102508011117684e-06,
+      "logits/chosen": -0.44864311814308167,
+      "logits/rejected": -0.7308858633041382,
+      "logps/chosen": -0.9878190755844116,
+      "logps/rejected": -1.7209956645965576,
+      "loss": 0.7796,
+      "odds_ratio_loss": 0.357673704624176,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.04939096421003342,
+      "rewards/margins": 0.03665883094072342,
+      "rewards/rejected": -0.08604978024959564,
+      "sft_loss": 0.9878190755844116,
+      "step": 1315
+    },
+    {
+      "epoch": 1.056,
+      "grad_norm": 4.673425946179824,
+      "learning_rate": 4.093559974371725e-06,
+      "logits/chosen": -0.6861502528190613,
+      "logits/rejected": -0.45734572410583496,
+      "logps/chosen": -0.5693352818489075,
+      "logps/rejected": -1.291886568069458,
+      "loss": 0.8203,
+      "odds_ratio_loss": 0.2426222562789917,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.028466764837503433,
+      "rewards/margins": 0.036127571016550064,
+      "rewards/rejected": -0.0645943284034729,
+      "sft_loss": 0.5693352818489075,
+      "step": 1320
+    },
+    {
+      "epoch": 1.06,
+      "grad_norm": 5.847314097304619,
+      "learning_rate": 4.084577418496775e-06,
+      "logits/chosen": -0.8275073766708374,
+      "logits/rejected": -0.6742109060287476,
+      "logps/chosen": -0.5419312119483948,
+      "logps/rejected": -1.4742541313171387,
+      "loss": 0.8365,
+      "odds_ratio_loss": 0.3019058406352997,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.02709656022489071,
+      "rewards/margins": 0.046616148203611374,
+      "rewards/rejected": -0.07371271401643753,
+      "sft_loss": 0.5419312119483948,
+      "step": 1325
+    },
+    {
+      "epoch": 1.064,
+      "grad_norm": 4.856747702455789,
+      "learning_rate": 4.075560538069767e-06,
+      "logits/chosen": -0.5532726049423218,
+      "logits/rejected": -0.6589607000350952,
+      "logps/chosen": -0.7456861734390259,
+      "logps/rejected": -1.2630527019500732,
+      "loss": 0.8495,
+      "odds_ratio_loss": 0.54130619764328,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03728431463241577,
+      "rewards/margins": 0.025868326425552368,
+      "rewards/rejected": -0.06315263360738754,
+      "sft_loss": 0.7456861734390259,
+      "step": 1330
+    },
+    {
+      "epoch": 1.068,
+      "grad_norm": 7.765532059632631,
+      "learning_rate": 4.066509528411151e-06,
+      "logits/chosen": -0.5705880522727966,
+      "logits/rejected": -0.7791796922683716,
+      "logps/chosen": -0.867768406867981,
+      "logps/rejected": -1.508888602256775,
+      "loss": 0.8959,
+      "odds_ratio_loss": 0.4603050649166107,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.04338841885328293,
+      "rewards/margins": 0.032056018710136414,
+      "rewards/rejected": -0.07544443756341934,
+      "sft_loss": 0.867768406867981,
+      "step": 1335
+    },
+    {
+      "epoch": 1.072,
+      "grad_norm": 12.916315689810858,
+      "learning_rate": 4.05742458558068e-06,
+      "logits/chosen": -0.8875905275344849,
+      "logits/rejected": -0.7213624119758606,
+      "logps/chosen": -0.7063192129135132,
+      "logps/rejected": -1.434460997581482,
+      "loss": 0.7654,
+      "odds_ratio_loss": 0.3284406065940857,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03531596064567566,
+      "rewards/margins": 0.03640709072351456,
+      "rewards/rejected": -0.07172305136919022,
+      "sft_loss": 0.7063192129135132,
+      "step": 1340
+    },
+    {
+      "epoch": 1.076,
+      "grad_norm": 7.30216329780713,
+      "learning_rate": 4.048305906373151e-06,
+      "logits/chosen": -0.6724140644073486,
+      "logits/rejected": -0.686550498008728,
+      "logps/chosen": -0.9202359914779663,
+      "logps/rejected": -1.229242205619812,
+      "loss": 0.8455,
+      "odds_ratio_loss": 0.5194590091705322,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.046011798083782196,
+      "rewards/margins": 0.015450313687324524,
+      "rewards/rejected": -0.06146211549639702,
+      "sft_loss": 0.9202359914779663,
+      "step": 1345
+    },
+    {
+      "epoch": 1.08,
+      "grad_norm": 7.376368865831852,
+      "learning_rate": 4.039153688314146e-06,
+      "logits/chosen": -0.6789649724960327,
+      "logits/rejected": -0.7698289155960083,
+      "logps/chosen": -1.0499086380004883,
+      "logps/rejected": -1.4064363241195679,
+      "loss": 0.8224,
+      "odds_ratio_loss": 0.5207637548446655,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05249543860554695,
+      "rewards/margins": 0.017826389521360397,
+      "rewards/rejected": -0.07032182067632675,
+      "sft_loss": 1.0499086380004883,
+      "step": 1350
+    },
+    {
+      "epoch": 1.084,
+      "grad_norm": 10.69540570354578,
+      "learning_rate": 4.029968129655757e-06,
+      "logits/chosen": -0.7770857214927673,
+      "logits/rejected": -0.9475702047348022,
+      "logps/chosen": -0.7179058790206909,
+      "logps/rejected": -1.5098683834075928,
+      "loss": 0.8206,
+      "odds_ratio_loss": 0.33577170968055725,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.035895295441150665,
+      "rewards/margins": 0.03959812596440315,
+      "rewards/rejected": -0.07549341022968292,
+      "sft_loss": 0.7179058790206909,
+      "step": 1355
+    },
+    {
+      "epoch": 1.088,
+      "grad_norm": 8.204731502525116,
+      "learning_rate": 4.020749429372286e-06,
+      "logits/chosen": -0.851569652557373,
+      "logits/rejected": -0.98048335313797,
+      "logps/chosen": -0.7297436594963074,
+      "logps/rejected": -1.4111974239349365,
+      "loss": 0.8462,
+      "odds_ratio_loss": 0.3803647458553314,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.036487188190221786,
+      "rewards/margins": 0.03407268971204758,
+      "rewards/rejected": -0.07055987417697906,
+      "sft_loss": 0.7297436594963074,
+      "step": 1360
+    },
+    {
+      "epoch": 1.092,
+      "grad_norm": 8.235008694291277,
+      "learning_rate": 4.011497787155938e-06,
+      "logits/chosen": -0.7461971044540405,
+      "logits/rejected": -1.4180670976638794,
+      "logps/chosen": -0.9519845843315125,
+      "logps/rejected": -1.9593369960784912,
+      "loss": 0.9038,
+      "odds_ratio_loss": 0.356905996799469,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.04759923368692398,
+      "rewards/margins": 0.050367631018161774,
+      "rewards/rejected": -0.09796686470508575,
+      "sft_loss": 0.9519845843315125,
+      "step": 1365
+    },
+    {
+      "epoch": 1.096,
+      "grad_norm": 15.389585033573624,
+      "learning_rate": 4.002213403412492e-06,
+      "logits/chosen": -0.6717543005943298,
+      "logits/rejected": -0.9089921116828918,
+      "logps/chosen": -1.0161564350128174,
+      "logps/rejected": -1.3220138549804688,
+      "loss": 0.8947,
+      "odds_ratio_loss": 0.5496761202812195,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05080781504511833,
+      "rewards/margins": 0.01529287826269865,
+      "rewards/rejected": -0.06610070168972015,
+      "sft_loss": 1.0161564350128174,
+      "step": 1370
+    },
+    {
+      "epoch": 1.1,
+      "grad_norm": 7.445613929034652,
+      "learning_rate": 3.992896479256966e-06,
+      "logits/chosen": -0.680860161781311,
+      "logits/rejected": -0.6606763601303101,
+      "logps/chosen": -0.6550648808479309,
+      "logps/rejected": -1.352782964706421,
+      "loss": 0.8223,
+      "odds_ratio_loss": 0.3205656111240387,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.032753244042396545,
+      "rewards/margins": 0.03488590195775032,
+      "rewards/rejected": -0.06763914227485657,
+      "sft_loss": 0.6550648808479309,
+      "step": 1375
+    },
+    {
+      "epoch": 1.104,
+      "grad_norm": 5.348676916256749,
+      "learning_rate": 3.983547216509254e-06,
+      "logits/chosen": -0.5789454579353333,
+      "logits/rejected": -1.2544965744018555,
+      "logps/chosen": -0.796095609664917,
+      "logps/rejected": -1.5642998218536377,
+      "loss": 0.7846,
+      "odds_ratio_loss": 0.32175213098526,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03980477899312973,
+      "rewards/margins": 0.038410209119319916,
+      "rewards/rejected": -0.07821498811244965,
+      "sft_loss": 0.796095609664917,
+      "step": 1380
+    },
+    {
+      "epoch": 1.108,
+      "grad_norm": 11.505326821901532,
+      "learning_rate": 3.974165817689758e-06,
+      "logits/chosen": -0.956534206867218,
+      "logits/rejected": -1.1092783212661743,
+      "logps/chosen": -1.0736840963363647,
+      "logps/rejected": -1.5605031251907349,
+      "loss": 0.8767,
+      "odds_ratio_loss": 0.49855270981788635,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.05368421599268913,
+      "rewards/margins": 0.02434094250202179,
+      "rewards/rejected": -0.07802516222000122,
+      "sft_loss": 1.0736840963363647,
+      "step": 1385
+    },
+    {
+      "epoch": 1.112,
+      "grad_norm": 7.11996683938393,
+      "learning_rate": 3.964752486015001e-06,
+      "logits/chosen": -0.702655553817749,
+      "logits/rejected": -1.0277684926986694,
+      "logps/chosen": -0.8141889572143555,
+      "logps/rejected": -1.9155117273330688,
+      "loss": 0.7498,
+      "odds_ratio_loss": 0.26831483840942383,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.04070945456624031,
+      "rewards/margins": 0.05506613105535507,
+      "rewards/rejected": -0.09577558934688568,
+      "sft_loss": 0.8141889572143555,
+      "step": 1390
+    },
+    {
+      "epoch": 1.116,
+      "grad_norm": 5.797590260170669,
+      "learning_rate": 3.955307425393224e-06,
+      "logits/chosen": -0.7000407576560974,
+      "logits/rejected": -0.9680187106132507,
+      "logps/chosen": -0.9160090684890747,
+      "logps/rejected": -1.364323616027832,
+      "loss": 0.8904,
+      "odds_ratio_loss": 0.45110782980918884,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.045800451189279556,
+      "rewards/margins": 0.022415729239583015,
+      "rewards/rejected": -0.06821618974208832,
+      "sft_loss": 0.9160090684890747,
+      "step": 1395
+    },
+    {
+      "epoch": 1.12,
+      "grad_norm": 8.433254983665345,
+      "learning_rate": 3.945830840419966e-06,
+      "logits/chosen": -0.7009550929069519,
+      "logits/rejected": -1.0051246881484985,
+      "logps/chosen": -0.4989989697933197,
+      "logps/rejected": -1.407712697982788,
+      "loss": 0.7925,
+      "odds_ratio_loss": 0.27720946073532104,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.024949947372078896,
+      "rewards/margins": 0.045435696840286255,
+      "rewards/rejected": -0.07038564234972,
+      "sft_loss": 0.4989989697933197,
+      "step": 1400
+    },
+    {
+      "epoch": 1.124,
+      "grad_norm": 12.99123162983583,
+      "learning_rate": 3.936322936373641e-06,
+      "logits/chosen": -0.6947557330131531,
+      "logits/rejected": -0.905806839466095,
+      "logps/chosen": -0.9586040377616882,
+      "logps/rejected": -1.2324825525283813,
+      "loss": 0.8352,
+      "odds_ratio_loss": 0.5849367380142212,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.04793020337820053,
+      "rewards/margins": 0.013693928718566895,
+      "rewards/rejected": -0.06162412837147713,
+      "sft_loss": 0.9586040377616882,
+      "step": 1405
+    },
+    {
+      "epoch": 1.1280000000000001,
+      "grad_norm": 8.170518488538953,
+      "learning_rate": 3.92678391921108e-06,
+      "logits/chosen": -1.1551212072372437,
+      "logits/rejected": -0.8484439849853516,
+      "logps/chosen": -0.5776058435440063,
+      "logps/rejected": -1.4134206771850586,
+      "loss": 0.8291,
+      "odds_ratio_loss": 0.31984660029411316,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.028880292549729347,
+      "rewards/margins": 0.04179074615240097,
+      "rewards/rejected": -0.07067102938890457,
+      "sft_loss": 0.5776058435440063,
+      "step": 1410
+    },
+    {
+      "epoch": 1.1320000000000001,
+      "grad_norm": 8.09496861727856,
+      "learning_rate": 3.9172139955630774e-06,
+      "logits/chosen": -0.713034987449646,
+      "logits/rejected": -0.9983466863632202,
+      "logps/chosen": -0.6458183526992798,
+      "logps/rejected": -1.828302025794983,
+      "loss": 0.866,
+      "odds_ratio_loss": 0.30388593673706055,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03229091316461563,
+      "rewards/margins": 0.0591241791844368,
+      "rewards/rejected": -0.09141509234905243,
+      "sft_loss": 0.6458183526992798,
+      "step": 1415
+    },
+    {
+      "epoch": 1.1360000000000001,
+      "grad_norm": 16.205423769664655,
+      "learning_rate": 3.907613372729916e-06,
+      "logits/chosen": -0.6726616024971008,
+      "logits/rejected": -1.3900980949401855,
+      "logps/chosen": -0.7454610466957092,
+      "logps/rejected": -2.1414051055908203,
+      "loss": 0.8471,
+      "odds_ratio_loss": 0.2859545350074768,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03727305680513382,
+      "rewards/margins": 0.06979719549417496,
+      "rewards/rejected": -0.10707025229930878,
+      "sft_loss": 0.7454610466957092,
+      "step": 1420
+    },
+    {
+      "epoch": 1.1400000000000001,
+      "grad_norm": 10.269281839827107,
+      "learning_rate": 3.897982258676867e-06,
+      "logits/chosen": -0.7214896082878113,
+      "logits/rejected": -0.8336132168769836,
+      "logps/chosen": -0.9054323434829712,
+      "logps/rejected": -2.1648507118225098,
+      "loss": 0.9165,
+      "odds_ratio_loss": 0.38728705048561096,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.0452716164290905,
+      "rewards/margins": 0.06297094374895096,
+      "rewards/rejected": -0.10824254900217056,
+      "sft_loss": 0.9054323434829712,
+      "step": 1425
+    },
+    {
+      "epoch": 1.144,
+      "grad_norm": 6.605575368297014,
+      "learning_rate": 3.888320862029699e-06,
+      "logits/chosen": -0.5836726427078247,
+      "logits/rejected": -0.7385787963867188,
+      "logps/chosen": -0.9469176530838013,
+      "logps/rejected": -1.4403495788574219,
+      "loss": 0.8601,
+      "odds_ratio_loss": 0.4318571984767914,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.04734588414430618,
+      "rewards/margins": 0.02467160113155842,
+      "rewards/rejected": -0.07201748341321945,
+      "sft_loss": 0.9469176530838013,
+      "step": 1430
+    },
+    {
+      "epoch": 1.148,
+      "grad_norm": 5.846266947718309,
+      "learning_rate": 3.878629392070143e-06,
+      "logits/chosen": -0.7644879817962646,
+      "logits/rejected": -0.9813248515129089,
+      "logps/chosen": -0.8727389574050903,
+      "logps/rejected": -1.5859434604644775,
+      "loss": 0.8655,
+      "odds_ratio_loss": 0.47635746002197266,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.043636951595544815,
+      "rewards/margins": 0.03566022962331772,
+      "rewards/rejected": -0.07929717004299164,
+      "sft_loss": 0.8727389574050903,
+      "step": 1435
+    },
+    {
+      "epoch": 1.152,
+      "grad_norm": 8.026595177727547,
+      "learning_rate": 3.868908058731376e-06,
+      "logits/chosen": -1.0093854665756226,
+      "logits/rejected": -0.8972814679145813,
+      "logps/chosen": -0.641372799873352,
+      "logps/rejected": -1.255726933479309,
+      "loss": 0.7964,
+      "odds_ratio_loss": 0.33428263664245605,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.0320686399936676,
+      "rewards/margins": 0.030717704445123672,
+      "rewards/rejected": -0.06278634071350098,
+      "sft_loss": 0.641372799873352,
+      "step": 1440
+    },
+    {
+      "epoch": 1.156,
+      "grad_norm": 7.131166323627883,
+      "learning_rate": 3.859157072593459e-06,
+      "logits/chosen": -0.8095647096633911,
+      "logits/rejected": -1.309104323387146,
+      "logps/chosen": -0.9154605865478516,
+      "logps/rejected": -1.6045910120010376,
+      "loss": 0.8756,
+      "odds_ratio_loss": 0.3739989399909973,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.04577303305268288,
+      "rewards/margins": 0.0344565287232399,
+      "rewards/rejected": -0.08022955805063248,
+      "sft_loss": 0.9154605865478516,
+      "step": 1445
+    },
+    {
+      "epoch": 1.16,
+      "grad_norm": 11.191454450304596,
+      "learning_rate": 3.849376644878783e-06,
+      "logits/chosen": -0.7892749905586243,
+      "logits/rejected": -0.7882648706436157,
+      "logps/chosen": -1.122315764427185,
+      "logps/rejected": -1.2966439723968506,
+      "loss": 0.8137,
+      "odds_ratio_loss": 0.6968892812728882,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.05611578747630119,
+      "rewards/margins": 0.008716410025954247,
+      "rewards/rejected": -0.06483219563961029,
+      "sft_loss": 1.122315764427185,
+      "step": 1450
+    },
+    {
+      "epoch": 1.164,
+      "grad_norm": 9.51974427968147,
+      "learning_rate": 3.839566987447492e-06,
+      "logits/chosen": -1.2010838985443115,
+      "logits/rejected": -0.9590059518814087,
+      "logps/chosen": -0.6959027051925659,
+      "logps/rejected": -1.873453140258789,
+      "loss": 0.831,
+      "odds_ratio_loss": 0.36710458993911743,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.0347951278090477,
+      "rewards/margins": 0.05887751653790474,
+      "rewards/rejected": -0.09367264807224274,
+      "sft_loss": 0.6959027051925659,
+      "step": 1455
+    },
+    {
+      "epoch": 1.168,
+      "grad_norm": 6.6227065501638815,
+      "learning_rate": 3.829728312792895e-06,
+      "logits/chosen": -0.8780848383903503,
+      "logits/rejected": -0.8581530451774597,
+      "logps/chosen": -0.9142974019050598,
+      "logps/rejected": -1.3824703693389893,
+      "loss": 0.78,
+      "odds_ratio_loss": 0.4631693363189697,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.04571487009525299,
+      "rewards/margins": 0.023408645763993263,
+      "rewards/rejected": -0.0691235214471817,
+      "sft_loss": 0.9142974019050598,
+      "step": 1460
+    },
+    {
+      "epoch": 1.172,
+      "grad_norm": 6.363999703267043,
+      "learning_rate": 3.819860834036859e-06,
+      "logits/chosen": -0.5977569222450256,
+      "logits/rejected": -0.7959061861038208,
+      "logps/chosen": -1.1498180627822876,
+      "logps/rejected": -1.42813241481781,
+      "loss": 0.8887,
+      "odds_ratio_loss": 0.5259816646575928,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.05749090388417244,
+      "rewards/margins": 0.013915717601776123,
+      "rewards/rejected": -0.07140661776065826,
+      "sft_loss": 1.1498180627822876,
+      "step": 1465
+    },
+    {
+      "epoch": 1.176,
+      "grad_norm": 6.246391614781854,
+      "learning_rate": 3.8099647649251984e-06,
+      "logits/chosen": -0.7605501413345337,
+      "logits/rejected": -1.1908546686172485,
+      "logps/chosen": -0.8100579380989075,
+      "logps/rejected": -1.775541067123413,
+      "loss": 0.8348,
+      "odds_ratio_loss": 0.42932015657424927,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.04050289839506149,
+      "rewards/margins": 0.04827415570616722,
+      "rewards/rejected": -0.08877705037593842,
+      "sft_loss": 0.8100579380989075,
+      "step": 1470
+    },
+    {
+      "epoch": 1.18,
+      "grad_norm": 5.893619858051333,
+      "learning_rate": 3.8000403198230385e-06,
+      "logits/chosen": -0.787771463394165,
+      "logits/rejected": -1.0664321184158325,
+      "logps/chosen": -0.8280608057975769,
+      "logps/rejected": -1.8852030038833618,
+      "loss": 0.822,
+      "odds_ratio_loss": 0.3562834858894348,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.041403044015169144,
+      "rewards/margins": 0.05285710096359253,
+      "rewards/rejected": -0.09426014125347137,
+      "sft_loss": 0.8280608057975769,
+      "step": 1475
+    },
+    {
+      "epoch": 1.184,
+      "grad_norm": 7.183406495752281,
+      "learning_rate": 3.790087713710179e-06,
+      "logits/chosen": -0.7213200330734253,
+      "logits/rejected": -0.6918378472328186,
+      "logps/chosen": -0.5293585658073425,
+      "logps/rejected": -1.8362514972686768,
+      "loss": 0.7973,
+      "odds_ratio_loss": 0.19014953076839447,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.026467930525541306,
+      "rewards/margins": 0.06534464657306671,
+      "rewards/rejected": -0.09181257337331772,
+      "sft_loss": 0.5293585658073425,
+      "step": 1480
+    },
+    {
+      "epoch": 1.188,
+      "grad_norm": 17.119549825566487,
+      "learning_rate": 3.780107162176429e-06,
+      "logits/chosen": -0.6820231676101685,
+      "logits/rejected": -1.1885521411895752,
+      "logps/chosen": -0.7856122255325317,
+      "logps/rejected": -1.2408835887908936,
+      "loss": 0.8931,
+      "odds_ratio_loss": 0.4235461354255676,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03928060457110405,
+      "rewards/margins": 0.022763576358556747,
+      "rewards/rejected": -0.0620441809296608,
+      "sft_loss": 0.7856122255325317,
+      "step": 1485
+    },
+    {
+      "epoch": 1.192,
+      "grad_norm": 6.561999759540379,
+      "learning_rate": 3.770098881416945e-06,
+      "logits/chosen": -0.8763113021850586,
+      "logits/rejected": -1.155576229095459,
+      "logps/chosen": -0.5533097386360168,
+      "logps/rejected": -1.3931071758270264,
+      "loss": 0.8147,
+      "odds_ratio_loss": 0.2342720478773117,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02766549028456211,
+      "rewards/margins": 0.041989874094724655,
+      "rewards/rejected": -0.06965536624193192,
+      "sft_loss": 0.5533097386360168,
+      "step": 1490
+    },
+    {
+      "epoch": 1.196,
+      "grad_norm": 5.852667265636353,
+      "learning_rate": 3.760063088227542e-06,
+      "logits/chosen": -0.7204443216323853,
+      "logits/rejected": -1.0452661514282227,
+      "logps/chosen": -0.6752656698226929,
+      "logps/rejected": -1.6966698169708252,
+      "loss": 0.842,
+      "odds_ratio_loss": 0.3053438067436218,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.033763282001018524,
+      "rewards/margins": 0.0510701946914196,
+      "rewards/rejected": -0.08483348786830902,
+      "sft_loss": 0.6752656698226929,
+      "step": 1495
+    },
+    {
+      "epoch": 1.2,
+      "grad_norm": 7.666952370293206,
+      "learning_rate": 3.7500000000000005e-06,
+      "logits/chosen": -1.2530326843261719,
+      "logits/rejected": -1.17472243309021,
+      "logps/chosen": -0.8737776875495911,
+      "logps/rejected": -1.5337140560150146,
+      "loss": 0.8295,
+      "odds_ratio_loss": 0.4370272159576416,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.043688882142305374,
+      "rewards/margins": 0.032996825873851776,
+      "rewards/rejected": -0.07668570429086685,
+      "sft_loss": 0.8737776875495911,
+      "step": 1500
+    },
+    {
+      "epoch": 1.204,
+      "grad_norm": 8.643591741745004,
+      "learning_rate": 3.739909834717356e-06,
+      "logits/chosen": -1.0214110612869263,
+      "logits/rejected": -1.1113232374191284,
+      "logps/chosen": -1.0257785320281982,
+      "logps/rejected": -1.4637627601623535,
+      "loss": 0.84,
+      "odds_ratio_loss": 0.4033835530281067,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.05128893256187439,
+      "rewards/margins": 0.021899200975894928,
+      "rewards/rejected": -0.07318813353776932,
+      "sft_loss": 1.0257785320281982,
+      "step": 1505
+    },
+    {
+      "epoch": 1.208,
+      "grad_norm": 6.581431751480548,
+      "learning_rate": 3.7297928109491765e-06,
+      "logits/chosen": -0.735799252986908,
+      "logits/rejected": -0.750560462474823,
+      "logps/chosen": -0.838580310344696,
+      "logps/rejected": -1.3117971420288086,
+      "loss": 0.8373,
+      "odds_ratio_loss": 0.4745975434780121,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.04192901775240898,
+      "rewards/margins": 0.023660842329263687,
+      "rewards/rejected": -0.06558986008167267,
+      "sft_loss": 0.838580310344696,
+      "step": 1510
+    },
+    {
+      "epoch": 1.212,
+      "grad_norm": 8.38248048628384,
+      "learning_rate": 3.7196491478468322e-06,
+      "logits/chosen": -0.7478394508361816,
+      "logits/rejected": -0.9995372891426086,
+      "logps/chosen": -0.8661327362060547,
+      "logps/rejected": -1.3012148141860962,
+      "loss": 0.7344,
+      "odds_ratio_loss": 0.4485481381416321,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.043306637555360794,
+      "rewards/margins": 0.021754100918769836,
+      "rewards/rejected": -0.06506074219942093,
+      "sft_loss": 0.8661327362060547,
+      "step": 1515
+    },
+    {
+      "epoch": 1.216,
+      "grad_norm": 7.766215821807629,
+      "learning_rate": 3.7094790651387414e-06,
+      "logits/chosen": -0.7365530729293823,
+      "logits/rejected": -0.8527762293815613,
+      "logps/chosen": -0.9115310907363892,
+      "logps/rejected": -1.5248000621795654,
+      "loss": 0.8285,
+      "odds_ratio_loss": 0.4467516839504242,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.0455765500664711,
+      "rewards/margins": 0.030663449317216873,
+      "rewards/rejected": -0.07623999565839767,
+      "sft_loss": 0.9115310907363892,
+      "step": 1520
+    },
+    {
+      "epoch": 1.22,
+      "grad_norm": 6.4955437270373135,
+      "learning_rate": 3.699282783125616e-06,
+      "logits/chosen": -0.7742820978164673,
+      "logits/rejected": -0.937674880027771,
+      "logps/chosen": -0.8425240516662598,
+      "logps/rejected": -1.2824028730392456,
+      "loss": 0.779,
+      "odds_ratio_loss": 0.4461577534675598,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.04212620481848717,
+      "rewards/margins": 0.021993935108184814,
+      "rewards/rejected": -0.06412014365196228,
+      "sft_loss": 0.8425240516662598,
+      "step": 1525
+    },
+    {
+      "epoch": 1.224,
+      "grad_norm": 6.362003415761398,
+      "learning_rate": 3.689060522675689e-06,
+      "logits/chosen": -0.7660902142524719,
+      "logits/rejected": -1.5068974494934082,
+      "logps/chosen": -0.6618943810462952,
+      "logps/rejected": -1.425063133239746,
+      "loss": 0.845,
+      "odds_ratio_loss": 0.4632844030857086,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.03309471905231476,
+      "rewards/margins": 0.038158439099788666,
+      "rewards/rejected": -0.07125315815210342,
+      "sft_loss": 0.6618943810462952,
+      "step": 1530
+    },
+    {
+      "epoch": 1.228,
+      "grad_norm": 6.895766693871366,
+      "learning_rate": 3.6788125052199264e-06,
+      "logits/chosen": -0.7660204172134399,
+      "logits/rejected": -1.0502755641937256,
+      "logps/chosen": -0.8439900279045105,
+      "logps/rejected": -1.3339447975158691,
+      "loss": 0.8453,
+      "odds_ratio_loss": 0.40678563714027405,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.042199499905109406,
+      "rewards/margins": 0.024497732520103455,
+      "rewards/rejected": -0.06669723987579346,
+      "sft_loss": 0.8439900279045105,
+      "step": 1535
+    },
+    {
+      "epoch": 1.232,
+      "grad_norm": 7.095832854877261,
+      "learning_rate": 3.668538952747236e-06,
+      "logits/chosen": -0.8988308906555176,
+      "logits/rejected": -1.370678186416626,
+      "logps/chosen": -0.7767564058303833,
+      "logps/rejected": -0.9345728158950806,
+      "loss": 0.8361,
+      "odds_ratio_loss": 0.6314736008644104,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.03883781656622887,
+      "rewards/margins": 0.007890825159847736,
+      "rewards/rejected": -0.04672864452004433,
+      "sft_loss": 0.7767564058303833,
+      "step": 1540
+    },
+    {
+      "epoch": 1.236,
+      "grad_norm": 15.297702452409231,
+      "learning_rate": 3.658240087799655e-06,
+      "logits/chosen": -1.128368854522705,
+      "logits/rejected": -0.9741285443305969,
+      "logps/chosen": -0.9071061015129089,
+      "logps/rejected": -1.4348928928375244,
+      "loss": 0.8299,
+      "odds_ratio_loss": 0.4918842315673828,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.04535530507564545,
+      "rewards/margins": 0.026389339938759804,
+      "rewards/rejected": -0.0717446431517601,
+      "sft_loss": 0.9071061015129089,
+      "step": 1545
+    },
+    {
+      "epoch": 1.24,
+      "grad_norm": 15.626108640705025,
+      "learning_rate": 3.6479161334675294e-06,
+      "logits/chosen": -0.7529286742210388,
+      "logits/rejected": -0.9413396716117859,
+      "logps/chosen": -0.6087412238121033,
+      "logps/rejected": -1.2803432941436768,
+      "loss": 0.7919,
+      "odds_ratio_loss": 0.30838295817375183,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.030437063425779343,
+      "rewards/margins": 0.033580102026462555,
+      "rewards/rejected": -0.0640171617269516,
+      "sft_loss": 0.6087412238121033,
+      "step": 1550
+    },
+    {
+      "epoch": 1.244,
+      "grad_norm": 7.081745221757999,
+      "learning_rate": 3.6375673133846847e-06,
+      "logits/chosen": -0.8344209790229797,
+      "logits/rejected": -0.9608109593391418,
+      "logps/chosen": -1.0645172595977783,
+      "logps/rejected": -1.4052292108535767,
+      "loss": 0.8832,
+      "odds_ratio_loss": 0.5416877269744873,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05322585627436638,
+      "rewards/margins": 0.017035599797964096,
+      "rewards/rejected": -0.07026146352291107,
+      "sft_loss": 1.0645172595977783,
+      "step": 1555
+    },
+    {
+      "epoch": 1.248,
+      "grad_norm": 7.328229540786072,
+      "learning_rate": 3.627193851723577e-06,
+      "logits/chosen": -0.8219780921936035,
+      "logits/rejected": -1.0412352085113525,
+      "logps/chosen": -0.8243353962898254,
+      "logps/rejected": -1.1440980434417725,
+      "loss": 0.7661,
+      "odds_ratio_loss": 0.48546719551086426,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.04121676832437515,
+      "rewards/margins": 0.01598813384771347,
+      "rewards/rejected": -0.05720490217208862,
+      "sft_loss": 0.8243353962898254,
+      "step": 1560
+    },
+    {
+      "epoch": 1.252,
+      "grad_norm": 8.481385406459447,
+      "learning_rate": 3.616795973190442e-06,
+      "logits/chosen": -0.6780751943588257,
+      "logits/rejected": -1.5924546718597412,
+      "logps/chosen": -0.7332156896591187,
+      "logps/rejected": -1.6106353998184204,
+      "loss": 0.8628,
+      "odds_ratio_loss": 0.2695372700691223,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03666078671813011,
+      "rewards/margins": 0.04387098178267479,
+      "rewards/rejected": -0.0805317685008049,
+      "sft_loss": 0.7332156896591187,
+      "step": 1565
+    },
+    {
+      "epoch": 1.256,
+      "grad_norm": 4.8716559657268625,
+      "learning_rate": 3.6063739030204226e-06,
+      "logits/chosen": -0.7873969078063965,
+      "logits/rejected": -1.142225980758667,
+      "logps/chosen": -0.7441651821136475,
+      "logps/rejected": -1.4604860544204712,
+      "loss": 0.7917,
+      "odds_ratio_loss": 0.44381627440452576,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.03720825910568237,
+      "rewards/margins": 0.03581603989005089,
+      "rewards/rejected": -0.07302430272102356,
+      "sft_loss": 0.7441651821136475,
+      "step": 1570
+    },
+    {
+      "epoch": 1.26,
+      "grad_norm": 5.809355093226559,
+      "learning_rate": 3.595927866972694e-06,
+      "logits/chosen": -0.8129979968070984,
+      "logits/rejected": -0.8149601817131042,
+      "logps/chosen": -0.7339099645614624,
+      "logps/rejected": -1.4441230297088623,
+      "loss": 0.8216,
+      "odds_ratio_loss": 0.3208310604095459,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03669549524784088,
+      "rewards/margins": 0.035510655492544174,
+      "rewards/rejected": -0.07220615446567535,
+      "sft_loss": 0.7339099645614624,
+      "step": 1575
+    },
+    {
+      "epoch": 1.264,
+      "grad_norm": 6.200919368514555,
+      "learning_rate": 3.5854580913255706e-06,
+      "logits/chosen": -0.5802955627441406,
+      "logits/rejected": -0.772422194480896,
+      "logps/chosen": -1.0532017946243286,
+      "logps/rejected": -1.506858468055725,
+      "loss": 0.8307,
+      "odds_ratio_loss": 0.4608602523803711,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05266008898615837,
+      "rewards/margins": 0.022682830691337585,
+      "rewards/rejected": -0.07534292340278625,
+      "sft_loss": 1.0532017946243286,
+      "step": 1580
+    },
+    {
+      "epoch": 1.268,
+      "grad_norm": 5.8218405882663085,
+      "learning_rate": 3.574964802871607e-06,
+      "logits/chosen": -0.5340670347213745,
+      "logits/rejected": -1.2797355651855469,
+      "logps/chosen": -0.8620258569717407,
+      "logps/rejected": -1.4560054540634155,
+      "loss": 0.8596,
+      "odds_ratio_loss": 0.3901941776275635,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.043101292103528976,
+      "rewards/margins": 0.029698973521590233,
+      "rewards/rejected": -0.07280026376247406,
+      "sft_loss": 0.8620258569717407,
+      "step": 1585
+    },
+    {
+      "epoch": 1.272,
+      "grad_norm": 6.664692745829687,
+      "learning_rate": 3.564448228912682e-06,
+      "logits/chosen": -0.5848423838615417,
+      "logits/rejected": -0.9934035539627075,
+      "logps/chosen": -0.7198655009269714,
+      "logps/rejected": -1.534895896911621,
+      "loss": 0.6926,
+      "odds_ratio_loss": 0.38178950548171997,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03599327430129051,
+      "rewards/margins": 0.04075152799487114,
+      "rewards/rejected": -0.07674480974674225,
+      "sft_loss": 0.7198655009269714,
+      "step": 1590
+    },
+    {
+      "epoch": 1.276,
+      "grad_norm": 6.854701270720263,
+      "learning_rate": 3.5539085972550786e-06,
+      "logits/chosen": -0.5101417303085327,
+      "logits/rejected": -1.0358482599258423,
+      "logps/chosen": -0.7137448191642761,
+      "logps/rejected": -1.2877639532089233,
+      "loss": 0.7468,
+      "odds_ratio_loss": 0.388786256313324,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.035687245428562164,
+      "rewards/margins": 0.02870095707476139,
+      "rewards/rejected": -0.0643882006406784,
+      "sft_loss": 0.7137448191642761,
+      "step": 1595
+    },
+    {
+      "epoch": 1.28,
+      "grad_norm": 6.779174368552459,
+      "learning_rate": 3.543346136204545e-06,
+      "logits/chosen": -0.9765304327011108,
+      "logits/rejected": -0.6466418504714966,
+      "logps/chosen": -0.9283822178840637,
+      "logps/rejected": -1.2504222393035889,
+      "loss": 0.8717,
+      "odds_ratio_loss": 0.5800614953041077,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.04641910642385483,
+      "rewards/margins": 0.016102004796266556,
+      "rewards/rejected": -0.06252111494541168,
+      "sft_loss": 0.9283822178840637,
+      "step": 1600
+    },
+    {
+      "epoch": 1.284,
+      "grad_norm": 9.786348784846165,
+      "learning_rate": 3.532761074561355e-06,
+      "logits/chosen": -0.6300225853919983,
+      "logits/rejected": -1.1493359804153442,
+      "logps/chosen": -1.0870106220245361,
+      "logps/rejected": -1.61102294921875,
+      "loss": 0.801,
+      "odds_ratio_loss": 0.4377533793449402,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.054350532591342926,
+      "rewards/margins": 0.026200611144304276,
+      "rewards/rejected": -0.0805511474609375,
+      "sft_loss": 1.0870106220245361,
+      "step": 1605
+    },
+    {
+      "epoch": 1.288,
+      "grad_norm": 5.913885782177144,
+      "learning_rate": 3.522153641615345e-06,
+      "logits/chosen": -0.5452396273612976,
+      "logits/rejected": -1.095402479171753,
+      "logps/chosen": -0.9149934649467468,
+      "logps/rejected": -1.1826355457305908,
+      "loss": 0.9447,
+      "odds_ratio_loss": 0.5239506959915161,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.04574967175722122,
+      "rewards/margins": 0.01338210143148899,
+      "rewards/rejected": -0.05913177877664566,
+      "sft_loss": 0.9149934649467468,
+      "step": 1610
+    },
+    {
+      "epoch": 1.292,
+      "grad_norm": 5.896924307562007,
+      "learning_rate": 3.5115240671409534e-06,
+      "logits/chosen": -0.6529924273490906,
+      "logits/rejected": -0.7851907014846802,
+      "logps/chosen": -0.8748595118522644,
+      "logps/rejected": -1.2012172937393188,
+      "loss": 0.8603,
+      "odds_ratio_loss": 0.5565796494483948,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.04374297708272934,
+      "rewards/margins": 0.016317885369062424,
+      "rewards/rejected": -0.060060858726501465,
+      "sft_loss": 0.8748595118522644,
+      "step": 1615
+    },
+    {
+      "epoch": 1.296,
+      "grad_norm": 9.095017819397759,
+      "learning_rate": 3.5008725813922383e-06,
+      "logits/chosen": -0.5774226188659668,
+      "logits/rejected": -1.0028631687164307,
+      "logps/chosen": -1.284346342086792,
+      "logps/rejected": -1.4374910593032837,
+      "loss": 0.8847,
+      "odds_ratio_loss": 0.6303257346153259,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.06421732157468796,
+      "rewards/margins": 0.007657224778085947,
+      "rewards/rejected": -0.07187455147504807,
+      "sft_loss": 1.284346342086792,
+      "step": 1620
+    },
+    {
+      "epoch": 1.3,
+      "grad_norm": 9.33528794237972,
+      "learning_rate": 3.4901994150978926e-06,
+      "logits/chosen": -0.878362774848938,
+      "logits/rejected": -0.7265375256538391,
+      "logps/chosen": -0.8903997540473938,
+      "logps/rejected": -1.3938519954681396,
+      "loss": 0.8503,
+      "odds_ratio_loss": 0.45942550897598267,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.04451999068260193,
+      "rewards/margins": 0.025172609835863113,
+      "rewards/rejected": -0.06969259679317474,
+      "sft_loss": 0.8903997540473938,
+      "step": 1625
+    },
+    {
+      "epoch": 1.304,
+      "grad_norm": 10.158405753265898,
+      "learning_rate": 3.4795047994562463e-06,
+      "logits/chosen": -0.8939259648323059,
+      "logits/rejected": -0.750242292881012,
+      "logps/chosen": -0.9898282885551453,
+      "logps/rejected": -1.685683250427246,
+      "loss": 0.8473,
+      "odds_ratio_loss": 0.5104387402534485,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.049491409212350845,
+      "rewards/margins": 0.03479275852441788,
+      "rewards/rejected": -0.08428415656089783,
+      "sft_loss": 0.9898282885551453,
+      "step": 1630
+    },
+    {
+      "epoch": 1.308,
+      "grad_norm": 5.954588970837507,
+      "learning_rate": 3.4687889661302577e-06,
+      "logits/chosen": -0.9046002626419067,
+      "logits/rejected": -0.9883207082748413,
+      "logps/chosen": -0.8091312646865845,
+      "logps/rejected": -1.3389207124710083,
+      "loss": 0.8038,
+      "odds_ratio_loss": 0.5021311640739441,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.040456559509038925,
+      "rewards/margins": 0.02648947760462761,
+      "rewards/rejected": -0.06694603711366653,
+      "sft_loss": 0.8091312646865845,
+      "step": 1635
+    },
+    {
+      "epoch": 1.312,
+      "grad_norm": 6.811003884561586,
+      "learning_rate": 3.458052147242494e-06,
+      "logits/chosen": -0.7402358651161194,
+      "logits/rejected": -0.9504525065422058,
+      "logps/chosen": -0.9438158273696899,
+      "logps/rejected": -2.0763096809387207,
+      "loss": 0.8385,
+      "odds_ratio_loss": 0.28224462270736694,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.047190792858600616,
+      "rewards/margins": 0.05662469193339348,
+      "rewards/rejected": -0.1038154810667038,
+      "sft_loss": 0.9438158273696899,
+      "step": 1640
+    },
+    {
+      "epoch": 1.316,
+      "grad_norm": 8.65594926906311,
+      "learning_rate": 3.4472945753701038e-06,
+      "logits/chosen": -0.7355834245681763,
+      "logits/rejected": -0.7831618189811707,
+      "logps/chosen": -0.8326643705368042,
+      "logps/rejected": -1.457876205444336,
+      "loss": 0.8597,
+      "odds_ratio_loss": 0.43533915281295776,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.04163322225213051,
+      "rewards/margins": 0.031260598450899124,
+      "rewards/rejected": -0.07289381325244904,
+      "sft_loss": 0.8326643705368042,
+      "step": 1645
+    },
+    {
+      "epoch": 1.32,
+      "grad_norm": 5.787660518955911,
+      "learning_rate": 3.436516483539781e-06,
+      "logits/chosen": -0.6674588918685913,
+      "logits/rejected": -1.381403923034668,
+      "logps/chosen": -0.6492362022399902,
+      "logps/rejected": -1.1699539422988892,
+      "loss": 0.7273,
+      "odds_ratio_loss": 0.35149624943733215,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03246181458234787,
+      "rewards/margins": 0.026035884395241737,
+      "rewards/rejected": -0.05849768966436386,
+      "sft_loss": 0.6492362022399902,
+      "step": 1650
+    },
+    {
+      "epoch": 1.324,
+      "grad_norm": 6.447720144022753,
+      "learning_rate": 3.4257181052227133e-06,
+      "logits/chosen": -0.76915043592453,
+      "logits/rejected": -1.1846320629119873,
+      "logps/chosen": -0.8949073553085327,
+      "logps/rejected": -1.5921183824539185,
+      "loss": 0.8674,
+      "odds_ratio_loss": 0.47409844398498535,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.04474537447094917,
+      "rewards/margins": 0.03486054390668869,
+      "rewards/rejected": -0.07960591465234756,
+      "sft_loss": 0.8949073553085327,
+      "step": 1655
+    },
+    {
+      "epoch": 1.328,
+      "grad_norm": 16.496650507435916,
+      "learning_rate": 3.4148996743295305e-06,
+      "logits/chosen": -0.8497360348701477,
+      "logits/rejected": -1.0253467559814453,
+      "logps/chosen": -0.7608711123466492,
+      "logps/rejected": -1.4417269229888916,
+      "loss": 0.7469,
+      "odds_ratio_loss": 0.33891811966896057,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.038043562322854996,
+      "rewards/margins": 0.034042783081531525,
+      "rewards/rejected": -0.07208634912967682,
+      "sft_loss": 0.7608711123466492,
+      "step": 1660
+    },
+    {
+      "epoch": 1.332,
+      "grad_norm": 12.21871272581458,
+      "learning_rate": 3.4040614252052305e-06,
+      "logits/chosen": -0.7546567916870117,
+      "logits/rejected": -1.1986757516860962,
+      "logps/chosen": -0.6745157241821289,
+      "logps/rejected": -1.5598911046981812,
+      "loss": 0.7454,
+      "odds_ratio_loss": 0.3794470429420471,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03372578322887421,
+      "rewards/margins": 0.04426876828074455,
+      "rewards/rejected": -0.07799455523490906,
+      "sft_loss": 0.6745157241821289,
+      "step": 1665
+    },
+    {
+      "epoch": 1.336,
+      "grad_norm": 5.398708828190264,
+      "learning_rate": 3.3932035926241103e-06,
+      "logits/chosen": -0.9119020700454712,
+      "logits/rejected": -0.9367231130599976,
+      "logps/chosen": -0.8278735280036926,
+      "logps/rejected": -1.6131706237792969,
+      "loss": 0.7558,
+      "odds_ratio_loss": 0.4295481741428375,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.04139367491006851,
+      "rewards/margins": 0.03926485404372215,
+      "rewards/rejected": -0.08065854012966156,
+      "sft_loss": 0.8278735280036926,
+      "step": 1670
+    },
+    {
+      "epoch": 1.34,
+      "grad_norm": 7.998094668850599,
+      "learning_rate": 3.3823264117846722e-06,
+      "logits/chosen": -0.7095149159431458,
+      "logits/rejected": -0.9372469186782837,
+      "logps/chosen": -0.6504486799240112,
+      "logps/rejected": -1.4709304571151733,
+      "loss": 0.7999,
+      "odds_ratio_loss": 0.3226034343242645,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03252243250608444,
+      "rewards/margins": 0.041024088859558105,
+      "rewards/rejected": -0.07354652881622314,
+      "sft_loss": 0.6504486799240112,
+      "step": 1675
+    },
+    {
+      "epoch": 1.3439999999999999,
+      "grad_norm": 13.402324415016235,
+      "learning_rate": 3.3714301183045382e-06,
+      "logits/chosen": -1.0852770805358887,
+      "logits/rejected": -1.0294393301010132,
+      "logps/chosen": -0.6565207242965698,
+      "logps/rejected": -2.4157521724700928,
+      "loss": 0.835,
+      "odds_ratio_loss": 0.23401963710784912,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03282603248953819,
+      "rewards/margins": 0.0879615768790245,
+      "rewards/rejected": -0.1207876056432724,
+      "sft_loss": 0.6565207242965698,
+      "step": 1680
+    },
+    {
+      "epoch": 1.3479999999999999,
+      "grad_norm": 6.178693388099064,
+      "learning_rate": 3.360514948215339e-06,
+      "logits/chosen": -0.9066254496574402,
+      "logits/rejected": -1.067453384399414,
+      "logps/chosen": -0.5945879817008972,
+      "logps/rejected": -1.1769797801971436,
+      "loss": 0.7498,
+      "odds_ratio_loss": 0.4167722761631012,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.02972939983010292,
+      "rewards/margins": 0.029119592159986496,
+      "rewards/rejected": -0.05884898826479912,
+      "sft_loss": 0.5945879817008972,
+      "step": 1685
+    },
+    {
+      "epoch": 1.3519999999999999,
+      "grad_norm": 8.662276656949278,
+      "learning_rate": 3.349581137957604e-06,
+      "logits/chosen": -0.7470632791519165,
+      "logits/rejected": -1.1178592443466187,
+      "logps/chosen": -0.7460489869117737,
+      "logps/rejected": -1.3396974802017212,
+      "loss": 0.7988,
+      "odds_ratio_loss": 0.39526912569999695,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.037302445620298386,
+      "rewards/margins": 0.029682422056794167,
+      "rewards/rejected": -0.0669848695397377,
+      "sft_loss": 0.7460489869117737,
+      "step": 1690
+    },
+    {
+      "epoch": 1.3559999999999999,
+      "grad_norm": 5.759001631003996,
+      "learning_rate": 3.338628924375638e-06,
+      "logits/chosen": -0.5028613805770874,
+      "logits/rejected": -0.9039660692214966,
+      "logps/chosen": -0.7187774181365967,
+      "logps/rejected": -1.8051927089691162,
+      "loss": 0.7071,
+      "odds_ratio_loss": 0.32172414660453796,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03593887761235237,
+      "rewards/margins": 0.05432076379656792,
+      "rewards/rejected": -0.09025964140892029,
+      "sft_loss": 0.7187774181365967,
+      "step": 1695
+    },
+    {
+      "epoch": 1.3599999999999999,
+      "grad_norm": 10.588167123809765,
+      "learning_rate": 3.3276585447123957e-06,
+      "logits/chosen": -1.002184510231018,
+      "logits/rejected": -1.065515160560608,
+      "logps/chosen": -0.9078339338302612,
+      "logps/rejected": -1.463555097579956,
+      "loss": 0.829,
+      "odds_ratio_loss": 0.39932265877723694,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.04539170488715172,
+      "rewards/margins": 0.02778605744242668,
+      "rewards/rejected": -0.0731777623295784,
+      "sft_loss": 0.9078339338302612,
+      "step": 1700
+    },
+    {
+      "epoch": 1.3639999999999999,
+      "grad_norm": 6.065572924023139,
+      "learning_rate": 3.3166702366043364e-06,
+      "logits/chosen": -0.9326748847961426,
+      "logits/rejected": -1.246891736984253,
+      "logps/chosen": -0.949301540851593,
+      "logps/rejected": -1.7571815252304077,
+      "loss": 0.8751,
+      "odds_ratio_loss": 0.5060352087020874,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.04746507853269577,
+      "rewards/margins": 0.040393996983766556,
+      "rewards/rejected": -0.08785907924175262,
+      "sft_loss": 0.949301540851593,
+      "step": 1705
+    },
+    {
+      "epoch": 1.3679999999999999,
+      "grad_norm": 7.242528374663146,
+      "learning_rate": 3.3056642380762783e-06,
+      "logits/chosen": -0.7751462459564209,
+      "logits/rejected": -0.972857654094696,
+      "logps/chosen": -1.051584243774414,
+      "logps/rejected": -1.275604486465454,
+      "loss": 0.8721,
+      "odds_ratio_loss": 0.57765793800354,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.05257921293377876,
+      "rewards/margins": 0.011201009154319763,
+      "rewards/rejected": -0.06378022581338882,
+      "sft_loss": 1.051584243774414,
+      "step": 1710
+    },
+    {
+      "epoch": 1.3719999999999999,
+      "grad_norm": 5.790819655822902,
+      "learning_rate": 3.294640787536245e-06,
+      "logits/chosen": -0.8153125643730164,
+      "logits/rejected": -1.0614861249923706,
+      "logps/chosen": -0.6195241212844849,
+      "logps/rejected": -1.4311720132827759,
+      "loss": 0.8206,
+      "odds_ratio_loss": 0.3184904158115387,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03097620978951454,
+      "rewards/margins": 0.040582384914159775,
+      "rewards/rejected": -0.07155859470367432,
+      "sft_loss": 0.6195241212844849,
+      "step": 1715
+    },
+    {
+      "epoch": 1.376,
+      "grad_norm": 11.980247948452748,
+      "learning_rate": 3.2836001237702993e-06,
+      "logits/chosen": -1.001387596130371,
+      "logits/rejected": -1.2363550662994385,
+      "logps/chosen": -0.6902490854263306,
+      "logps/rejected": -1.1932241916656494,
+      "loss": 0.8108,
+      "odds_ratio_loss": 0.368974506855011,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.034512460231781006,
+      "rewards/margins": 0.025148753076791763,
+      "rewards/rejected": -0.05966120958328247,
+      "sft_loss": 0.6902490854263306,
+      "step": 1720
+    },
+    {
+      "epoch": 1.38,
+      "grad_norm": 5.0828283091407735,
+      "learning_rate": 3.272542485937369e-06,
+      "logits/chosen": -1.0046998262405396,
+      "logits/rejected": -0.8059650659561157,
+      "logps/chosen": -0.7130595445632935,
+      "logps/rejected": -1.1542508602142334,
+      "loss": 0.8238,
+      "odds_ratio_loss": 0.4597712457180023,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.03565298020839691,
+      "rewards/margins": 0.02205955982208252,
+      "rewards/rejected": -0.05771253630518913,
+      "sft_loss": 0.7130595445632935,
+      "step": 1725
+    },
+    {
+      "epoch": 1.384,
+      "grad_norm": 5.35699654066337,
+      "learning_rate": 3.2614681135640696e-06,
+      "logits/chosen": -0.863267719745636,
+      "logits/rejected": -1.239708423614502,
+      "logps/chosen": -0.877366840839386,
+      "logps/rejected": -1.5349080562591553,
+      "loss": 0.805,
+      "odds_ratio_loss": 0.3669111132621765,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.04386834427714348,
+      "rewards/margins": 0.03287705034017563,
+      "rewards/rejected": -0.07674539089202881,
+      "sft_loss": 0.877366840839386,
+      "step": 1730
+    },
+    {
+      "epoch": 1.388,
+      "grad_norm": 7.300604070779147,
+      "learning_rate": 3.2503772465395143e-06,
+      "logits/chosen": -0.8804348111152649,
+      "logits/rejected": -0.6921446323394775,
+      "logps/chosen": -0.6613166332244873,
+      "logps/rejected": -1.4078587293624878,
+      "loss": 0.7645,
+      "odds_ratio_loss": 0.3147312104701996,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.033065833151340485,
+      "rewards/margins": 0.037327107042074203,
+      "rewards/rejected": -0.07039294391870499,
+      "sft_loss": 0.6613166332244873,
+      "step": 1735
+    },
+    {
+      "epoch": 1.392,
+      "grad_norm": 7.544399455719484,
+      "learning_rate": 3.2392701251101172e-06,
+      "logits/chosen": -0.749683678150177,
+      "logits/rejected": -1.0036853551864624,
+      "logps/chosen": -0.8362616300582886,
+      "logps/rejected": -1.4140605926513672,
+      "loss": 0.8505,
+      "odds_ratio_loss": 0.3677152693271637,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.04181307926774025,
+      "rewards/margins": 0.028889944776892662,
+      "rewards/rejected": -0.07070302218198776,
+      "sft_loss": 0.8362616300582886,
+      "step": 1740
+    },
+    {
+      "epoch": 1.396,
+      "grad_norm": 5.220293895574105,
+      "learning_rate": 3.228146989874389e-06,
+      "logits/chosen": -0.8896152377128601,
+      "logits/rejected": -0.9276788830757141,
+      "logps/chosen": -0.9979864358901978,
+      "logps/rejected": -1.3124749660491943,
+      "loss": 0.9121,
+      "odds_ratio_loss": 0.5718216896057129,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.049899328500032425,
+      "rewards/margins": 0.015724416822195053,
+      "rewards/rejected": -0.06562374532222748,
+      "sft_loss": 0.9979864358901978,
+      "step": 1745
+    },
+    {
+      "epoch": 1.4,
+      "grad_norm": 8.96935863952734,
+      "learning_rate": 3.217008081777726e-06,
+      "logits/chosen": -0.8349775075912476,
+      "logits/rejected": -0.8859192132949829,
+      "logps/chosen": -0.7498109340667725,
+      "logps/rejected": -1.4023478031158447,
+      "loss": 0.7615,
+      "odds_ratio_loss": 0.3978427052497864,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03749055042862892,
+      "rewards/margins": 0.03262684494256973,
+      "rewards/rejected": -0.07011739164590836,
+      "sft_loss": 0.7498109340667725,
+      "step": 1750
+    },
+    {
+      "epoch": 1.404,
+      "grad_norm": 4.812654034935553,
+      "learning_rate": 3.205853642107192e-06,
+      "logits/chosen": -0.8215142488479614,
+      "logits/rejected": -0.8328277468681335,
+      "logps/chosen": -0.7706052660942078,
+      "logps/rejected": -1.4792053699493408,
+      "loss": 0.7118,
+      "odds_ratio_loss": 0.3464731276035309,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.038530267775058746,
+      "rewards/margins": 0.03542999550700188,
+      "rewards/rejected": -0.07396025955677032,
+      "sft_loss": 0.7706052660942078,
+      "step": 1755
+    },
+    {
+      "epoch": 1.408,
+      "grad_norm": 5.327583776041735,
+      "learning_rate": 3.1946839124862873e-06,
+      "logits/chosen": -0.8214722871780396,
+      "logits/rejected": -0.9568880796432495,
+      "logps/chosen": -0.8749944567680359,
+      "logps/rejected": -1.4335380792617798,
+      "loss": 0.8792,
+      "odds_ratio_loss": 0.47741952538490295,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.04374972730875015,
+      "rewards/margins": 0.027927175164222717,
+      "rewards/rejected": -0.07167690247297287,
+      "sft_loss": 0.8749944567680359,
+      "step": 1760
+    },
+    {
+      "epoch": 1.412,
+      "grad_norm": 7.208526674502584,
+      "learning_rate": 3.183499134869721e-06,
+      "logits/chosen": -0.8680820465087891,
+      "logits/rejected": -0.9875283241271973,
+      "logps/chosen": -1.09853196144104,
+      "logps/rejected": -1.329644799232483,
+      "loss": 0.8253,
+      "odds_ratio_loss": 0.6291381120681763,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.05492659658193588,
+      "rewards/margins": 0.01155565120279789,
+      "rewards/rejected": -0.06648223847150803,
+      "sft_loss": 1.09853196144104,
+      "step": 1765
+    },
+    {
+      "epoch": 1.416,
+      "grad_norm": 12.79259892670254,
+      "learning_rate": 3.1722995515381644e-06,
+      "logits/chosen": -1.3741729259490967,
+      "logits/rejected": -0.9874537587165833,
+      "logps/chosen": -0.6389180421829224,
+      "logps/rejected": -1.135964035987854,
+      "loss": 0.8255,
+      "odds_ratio_loss": 0.34384432435035706,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.031945906579494476,
+      "rewards/margins": 0.024852296337485313,
+      "rewards/rejected": -0.05679820105433464,
+      "sft_loss": 0.6389180421829224,
+      "step": 1770
+    },
+    {
+      "epoch": 1.42,
+      "grad_norm": 8.588985311767232,
+      "learning_rate": 3.1610854050930063e-06,
+      "logits/chosen": -0.8712812662124634,
+      "logits/rejected": -1.1689540147781372,
+      "logps/chosen": -1.0062519311904907,
+      "logps/rejected": -1.973914384841919,
+      "loss": 0.9165,
+      "odds_ratio_loss": 0.35945457220077515,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.050312597304582596,
+      "rewards/margins": 0.04838312417268753,
+      "rewards/rejected": -0.09869572520256042,
+      "sft_loss": 1.0062519311904907,
+      "step": 1775
+    },
+    {
+      "epoch": 1.424,
+      "grad_norm": 5.412987589877953,
+      "learning_rate": 3.149856938451094e-06,
+      "logits/chosen": -0.5333024263381958,
+      "logits/rejected": -1.3137054443359375,
+      "logps/chosen": -0.8601778149604797,
+      "logps/rejected": -1.641958475112915,
+      "loss": 0.888,
+      "odds_ratio_loss": 0.7197853922843933,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.043008893728256226,
+      "rewards/margins": 0.039089031517505646,
+      "rewards/rejected": -0.08209791779518127,
+      "sft_loss": 0.8601778149604797,
+      "step": 1780
+    },
+    {
+      "epoch": 1.428,
+      "grad_norm": 6.165182680586835,
+      "learning_rate": 3.1386143948394764e-06,
+      "logits/chosen": -1.0411813259124756,
+      "logits/rejected": -0.48357734084129333,
+      "logps/chosen": -0.7215684652328491,
+      "logps/rejected": -1.2277052402496338,
+      "loss": 0.774,
+      "odds_ratio_loss": 0.4655603766441345,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.03607841581106186,
+      "rewards/margins": 0.02530684694647789,
+      "rewards/rejected": -0.06138526648283005,
+      "sft_loss": 0.7215684652328491,
+      "step": 1785
+    },
+    {
+      "epoch": 1.432,
+      "grad_norm": 5.856667562992677,
+      "learning_rate": 3.127358017790132e-06,
+      "logits/chosen": -1.1004786491394043,
+      "logits/rejected": -0.8851817846298218,
+      "logps/chosen": -0.5895828008651733,
+      "logps/rejected": -1.2631773948669434,
+      "loss": 0.7728,
+      "odds_ratio_loss": 0.329679012298584,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.029479142278432846,
+      "rewards/margins": 0.03367973119020462,
+      "rewards/rejected": -0.06315887719392776,
+      "sft_loss": 0.5895828008651733,
+      "step": 1790
+    },
+    {
+      "epoch": 1.436,
+      "grad_norm": 5.146001537629413,
+      "learning_rate": 3.116088051134695e-06,
+      "logits/chosen": -1.0093369483947754,
+      "logits/rejected": -1.3544859886169434,
+      "logps/chosen": -0.7406253218650818,
+      "logps/rejected": -1.5723696947097778,
+      "loss": 0.8389,
+      "odds_ratio_loss": 0.31857210397720337,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03703127056360245,
+      "rewards/margins": 0.041587214916944504,
+      "rewards/rejected": -0.07861848175525665,
+      "sft_loss": 0.7406253218650818,
+      "step": 1795
+    },
+    {
+      "epoch": 1.44,
+      "grad_norm": 4.843927894063411,
+      "learning_rate": 3.1048047389991693e-06,
+      "logits/chosen": -0.903844952583313,
+      "logits/rejected": -0.8139872550964355,
+      "logps/chosen": -1.0251444578170776,
+      "logps/rejected": -1.2178661823272705,
+      "loss": 0.9374,
+      "odds_ratio_loss": 0.6262407302856445,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05125723034143448,
+      "rewards/margins": 0.00963608454912901,
+      "rewards/rejected": -0.060893308371305466,
+      "sft_loss": 1.0251444578170776,
+      "step": 1800
+    },
+    {
+      "epoch": 1.444,
+      "grad_norm": 7.431201496879773,
+      "learning_rate": 3.0935083257986493e-06,
+      "logits/chosen": -0.6094223260879517,
+      "logits/rejected": -0.899307370185852,
+      "logps/chosen": -0.7733746767044067,
+      "logps/rejected": -1.2325997352600098,
+      "loss": 0.8234,
+      "odds_ratio_loss": 0.44773632287979126,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.03866872936487198,
+      "rewards/margins": 0.022961260750889778,
+      "rewards/rejected": -0.06162998825311661,
+      "sft_loss": 0.7733746767044067,
+      "step": 1805
+    },
+    {
+      "epoch": 1.448,
+      "grad_norm": 11.586614332572152,
+      "learning_rate": 3.082199056232015e-06,
+      "logits/chosen": -0.7563143372535706,
+      "logits/rejected": -1.069698452949524,
+      "logps/chosen": -1.023844599723816,
+      "logps/rejected": -1.5895153284072876,
+      "loss": 0.7528,
+      "odds_ratio_loss": 0.4808468222618103,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.051192231476306915,
+      "rewards/margins": 0.028283540159463882,
+      "rewards/rejected": -0.0794757753610611,
+      "sft_loss": 1.023844599723816,
+      "step": 1810
+    },
+    {
+      "epoch": 1.452,
+      "grad_norm": 8.9164649839478,
+      "learning_rate": 3.0708771752766397e-06,
+      "logits/chosen": -1.1078320741653442,
+      "logits/rejected": -1.0441960096359253,
+      "logps/chosen": -0.8418495059013367,
+      "logps/rejected": -1.273721694946289,
+      "loss": 0.8566,
+      "odds_ratio_loss": 0.4950701594352722,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.04209247976541519,
+      "rewards/margins": 0.0215936116874218,
+      "rewards/rejected": -0.06368608772754669,
+      "sft_loss": 0.8418495059013367,
+      "step": 1815
+    },
+    {
+      "epoch": 1.456,
+      "grad_norm": 5.634450444011399,
+      "learning_rate": 3.059542928183079e-06,
+      "logits/chosen": -1.020975112915039,
+      "logits/rejected": -1.0432868003845215,
+      "logps/chosen": -0.8400118947029114,
+      "logps/rejected": -1.9318656921386719,
+      "loss": 0.8575,
+      "odds_ratio_loss": 0.45713549852371216,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.04200059548020363,
+      "rewards/margins": 0.05459269881248474,
+      "rewards/rejected": -0.09659329801797867,
+      "sft_loss": 0.8400118947029114,
+      "step": 1820
+    },
+    {
+      "epoch": 1.46,
+      "grad_norm": 10.1775941550842,
+      "learning_rate": 3.0481965604697582e-06,
+      "logits/chosen": -0.6371676325798035,
+      "logits/rejected": -1.1217305660247803,
+      "logps/chosen": -0.7022415995597839,
+      "logps/rejected": -1.5772149562835693,
+      "loss": 0.7258,
+      "odds_ratio_loss": 0.3305204510688782,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.035112082958221436,
+      "rewards/margins": 0.04374866932630539,
+      "rewards/rejected": -0.07886074483394623,
+      "sft_loss": 0.7022415995597839,
+      "step": 1825
+    },
+    {
+      "epoch": 1.464,
+      "grad_norm": 6.898491609991041,
+      "learning_rate": 3.0368383179176584e-06,
+      "logits/chosen": -0.5923871994018555,
+      "logits/rejected": -1.568169355392456,
+      "logps/chosen": -0.7653383016586304,
+      "logps/rejected": -1.7148897647857666,
+      "loss": 0.7246,
+      "odds_ratio_loss": 0.36337292194366455,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03826691955327988,
+      "rewards/margins": 0.04747757315635681,
+      "rewards/rejected": -0.08574448525905609,
+      "sft_loss": 0.7653383016586304,
+      "step": 1830
+    },
+    {
+      "epoch": 1.468,
+      "grad_norm": 5.964977560470005,
+      "learning_rate": 3.025468446564985e-06,
+      "logits/chosen": -1.165166974067688,
+      "logits/rejected": -0.8749963641166687,
+      "logps/chosen": -0.8071148991584778,
+      "logps/rejected": -1.657152533531189,
+      "loss": 0.865,
+      "odds_ratio_loss": 0.321114182472229,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.04035574942827225,
+      "rewards/margins": 0.04250188171863556,
+      "rewards/rejected": -0.08285762369632721,
+      "sft_loss": 0.8071148991584778,
+      "step": 1835
+    },
+    {
+      "epoch": 1.472,
+      "grad_norm": 9.069724398190369,
+      "learning_rate": 3.0140871927018466e-06,
+      "logits/chosen": -1.0001251697540283,
+      "logits/rejected": -0.8858006596565247,
+      "logps/chosen": -0.8963924646377563,
+      "logps/rejected": -1.671795129776001,
+      "loss": 0.9639,
+      "odds_ratio_loss": 0.41972383856773376,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.04481962323188782,
+      "rewards/margins": 0.03877013176679611,
+      "rewards/rejected": -0.08358974754810333,
+      "sft_loss": 0.8963924646377563,
+      "step": 1840
+    },
+    {
+      "epoch": 1.476,
+      "grad_norm": 8.923998975352527,
+      "learning_rate": 3.002694802864912e-06,
+      "logits/chosen": -0.8372132182121277,
+      "logits/rejected": -0.897018551826477,
+      "logps/chosen": -0.6927198171615601,
+      "logps/rejected": -1.5746911764144897,
+      "loss": 0.7381,
+      "odds_ratio_loss": 0.30051150918006897,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.034635990858078,
+      "rewards/margins": 0.044098567217588425,
+      "rewards/rejected": -0.07873455435037613,
+      "sft_loss": 0.6927198171615601,
+      "step": 1845
+    },
+    {
+      "epoch": 1.48,
+      "grad_norm": 11.770053955264393,
+      "learning_rate": 2.9912915238320755e-06,
+      "logits/chosen": -0.6851986646652222,
+      "logits/rejected": -1.1784414052963257,
+      "logps/chosen": -0.833079993724823,
+      "logps/rejected": -1.4270697832107544,
+      "loss": 0.8667,
+      "odds_ratio_loss": 0.42909732460975647,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.04165399819612503,
+      "rewards/margins": 0.02969948947429657,
+      "rewards/rejected": -0.0713534951210022,
+      "sft_loss": 0.833079993724823,
+      "step": 1850
+    },
+    {
+      "epoch": 1.484,
+      "grad_norm": 6.233141067464782,
+      "learning_rate": 2.9798776026171087e-06,
+      "logits/chosen": -0.842139720916748,
+      "logits/rejected": -1.1131788492202759,
+      "logps/chosen": -0.9360324740409851,
+      "logps/rejected": -1.4683505296707153,
+      "loss": 0.8339,
+      "odds_ratio_loss": 0.5007175207138062,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.04680163040757179,
+      "rewards/margins": 0.026615899056196213,
+      "rewards/rejected": -0.07341752201318741,
+      "sft_loss": 0.9360324740409851,
+      "step": 1855
+    },
+    {
+      "epoch": 1.488,
+      "grad_norm": 6.003031745272665,
+      "learning_rate": 2.9684532864643123e-06,
+      "logits/chosen": -0.8800287246704102,
+      "logits/rejected": -1.1693894863128662,
+      "logps/chosen": -0.8431293368339539,
+      "logps/rejected": -1.1915092468261719,
+      "loss": 0.8388,
+      "odds_ratio_loss": 0.5465279817581177,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.04215647280216217,
+      "rewards/margins": 0.017418991774320602,
+      "rewards/rejected": -0.059575460851192474,
+      "sft_loss": 0.8431293368339539,
+      "step": 1860
+    },
+    {
+      "epoch": 1.492,
+      "grad_norm": 4.835110290839885,
+      "learning_rate": 2.957018822843154e-06,
+      "logits/chosen": -0.8184400796890259,
+      "logits/rejected": -1.3122928142547607,
+      "logps/chosen": -0.6195321679115295,
+      "logps/rejected": -1.447097897529602,
+      "loss": 0.8414,
+      "odds_ratio_loss": 0.3124271035194397,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.030976608395576477,
+      "rewards/margins": 0.04137829318642616,
+      "rewards/rejected": -0.07235489785671234,
+      "sft_loss": 0.6195321679115295,
+      "step": 1865
+    },
+    {
+      "epoch": 1.496,
+      "grad_norm": 6.054406343408761,
+      "learning_rate": 2.945574459442917e-06,
+      "logits/chosen": -0.9596401453018188,
+      "logits/rejected": -1.1060923337936401,
+      "logps/chosen": -0.6389337778091431,
+      "logps/rejected": -1.4341099262237549,
+      "loss": 0.8757,
+      "odds_ratio_loss": 0.32751792669296265,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03194668889045715,
+      "rewards/margins": 0.03975880891084671,
+      "rewards/rejected": -0.07170549780130386,
+      "sft_loss": 0.6389337778091431,
+      "step": 1870
+    },
+    {
+      "epoch": 1.5,
+      "grad_norm": 5.445524941655018,
+      "learning_rate": 2.9341204441673267e-06,
+      "logits/chosen": -0.647433876991272,
+      "logits/rejected": -1.0591230392456055,
+      "logps/chosen": -0.6938437223434448,
+      "logps/rejected": -1.3513716459274292,
+      "loss": 0.7662,
+      "odds_ratio_loss": 0.5081362128257751,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.0346921868622303,
+      "rewards/margins": 0.0328763946890831,
+      "rewards/rejected": -0.0675685852766037,
+      "sft_loss": 0.6938437223434448,
+      "step": 1875
+    },
+    {
+      "epoch": 1.504,
+      "grad_norm": 5.6546570414563,
+      "learning_rate": 2.922657025129185e-06,
+      "logits/chosen": -0.6220548748970032,
+      "logits/rejected": -0.8203363418579102,
+      "logps/chosen": -0.8300703167915344,
+      "logps/rejected": -1.3827760219573975,
+      "loss": 0.7737,
+      "odds_ratio_loss": 0.40472111105918884,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.04150351509451866,
+      "rewards/margins": 0.02763528749346733,
+      "rewards/rejected": -0.06913881003856659,
+      "sft_loss": 0.8300703167915344,
+      "step": 1880
+    },
+    {
+      "epoch": 1.508,
+      "grad_norm": 6.230171438098299,
+      "learning_rate": 2.9111844506449973e-06,
+      "logits/chosen": -0.7947267293930054,
+      "logits/rejected": -0.9040530920028687,
+      "logps/chosen": -0.9311960935592651,
+      "logps/rejected": -1.4013652801513672,
+      "loss": 0.7731,
+      "odds_ratio_loss": 0.44902342557907104,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.04655980318784714,
+      "rewards/margins": 0.02350846491754055,
+      "rewards/rejected": -0.07006827741861343,
+      "sft_loss": 0.9311960935592651,
+      "step": 1885
+    },
+    {
+      "epoch": 1.512,
+      "grad_norm": 6.7332266058939405,
+      "learning_rate": 2.8997029692295875e-06,
+      "logits/chosen": -0.814238429069519,
+      "logits/rejected": -0.9660609364509583,
+      "logps/chosen": -0.7938546538352966,
+      "logps/rejected": -1.457929253578186,
+      "loss": 0.9477,
+      "odds_ratio_loss": 0.4195871949195862,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03969273716211319,
+      "rewards/margins": 0.033203721046447754,
+      "rewards/rejected": -0.07289645820856094,
+      "sft_loss": 0.7938546538352966,
+      "step": 1890
+    },
+    {
+      "epoch": 1.516,
+      "grad_norm": 5.752938476673707,
+      "learning_rate": 2.888212829590719e-06,
+      "logits/chosen": -0.6543909311294556,
+      "logits/rejected": -0.5943178534507751,
+      "logps/chosen": -0.5880377888679504,
+      "logps/rejected": -1.3045456409454346,
+      "loss": 0.8214,
+      "odds_ratio_loss": 0.3315754234790802,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02940189279615879,
+      "rewards/margins": 0.03582539036870003,
+      "rewards/rejected": -0.06522727757692337,
+      "sft_loss": 0.5880377888679504,
+      "step": 1895
+    },
+    {
+      "epoch": 1.52,
+      "grad_norm": 4.713391058679103,
+      "learning_rate": 2.876714280623708e-06,
+      "logits/chosen": -0.8652218580245972,
+      "logits/rejected": -0.9359322786331177,
+      "logps/chosen": -0.6846807599067688,
+      "logps/rejected": -1.5125911235809326,
+      "loss": 0.7733,
+      "odds_ratio_loss": 0.2784159481525421,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03423403576016426,
+      "rewards/margins": 0.04139552637934685,
+      "rewards/rejected": -0.07562955468893051,
+      "sft_loss": 0.6846807599067688,
+      "step": 1900
+    },
+    {
+      "epoch": 1.524,
+      "grad_norm": 9.06130869139701,
+      "learning_rate": 2.8652075714060296e-06,
+      "logits/chosen": -0.920418381690979,
+      "logits/rejected": -0.9699466824531555,
+      "logps/chosen": -0.49238044023513794,
+      "logps/rejected": -1.3032811880111694,
+      "loss": 0.7794,
+      "odds_ratio_loss": 0.21321973204612732,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.024619024246931076,
+      "rewards/margins": 0.040545038878917694,
+      "rewards/rejected": -0.06516405940055847,
+      "sft_loss": 0.49238044023513794,
+      "step": 1905
+    },
+    {
+      "epoch": 1.528,
+      "grad_norm": 6.44994215725063,
+      "learning_rate": 2.8536929511919227e-06,
+      "logits/chosen": -0.5823496580123901,
+      "logits/rejected": -1.490532636642456,
+      "logps/chosen": -0.7626686096191406,
+      "logps/rejected": -1.4420548677444458,
+      "loss": 0.8321,
+      "odds_ratio_loss": 0.3727583587169647,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.03813342750072479,
+      "rewards/margins": 0.03396930545568466,
+      "rewards/rejected": -0.07210274040699005,
+      "sft_loss": 0.7626686096191406,
+      "step": 1910
+    },
+    {
+      "epoch": 1.532,
+      "grad_norm": 10.576382626683463,
+      "learning_rate": 2.842170669406993e-06,
+      "logits/chosen": -0.9067608714103699,
+      "logits/rejected": -1.0328947305679321,
+      "logps/chosen": -0.7888859510421753,
+      "logps/rejected": -1.5981676578521729,
+      "loss": 0.9028,
+      "odds_ratio_loss": 0.3959670960903168,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.039444297552108765,
+      "rewards/margins": 0.04046408459544182,
+      "rewards/rejected": -0.07990838587284088,
+      "sft_loss": 0.7888859510421753,
+      "step": 1915
+    },
+    {
+      "epoch": 1.536,
+      "grad_norm": 10.142634875571975,
+      "learning_rate": 2.8306409756428067e-06,
+      "logits/chosen": -0.8129776120185852,
+      "logits/rejected": -1.2146799564361572,
+      "logps/chosen": -0.631020188331604,
+      "logps/rejected": -1.7902923822402954,
+      "loss": 0.7791,
+      "odds_ratio_loss": 0.4324275851249695,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.03155101090669632,
+      "rewards/margins": 0.05796360969543457,
+      "rewards/rejected": -0.08951462805271149,
+      "sft_loss": 0.631020188331604,
+      "step": 1920
+    },
+    {
+      "epoch": 1.54,
+      "grad_norm": 6.955192952345903,
+      "learning_rate": 2.8191041196514874e-06,
+      "logits/chosen": -0.9502149820327759,
+      "logits/rejected": -1.0196278095245361,
+      "logps/chosen": -0.8367894291877747,
+      "logps/rejected": -1.5817101001739502,
+      "loss": 0.8259,
+      "odds_ratio_loss": 0.31859907507896423,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.04183947294950485,
+      "rewards/margins": 0.03724603354930878,
+      "rewards/rejected": -0.07908550649881363,
+      "sft_loss": 0.8367894291877747,
+      "step": 1925
+    },
+    {
+      "epoch": 1.544,
+      "grad_norm": 8.19557532330742,
+      "learning_rate": 2.807560351340302e-06,
+      "logits/chosen": -1.2878282070159912,
+      "logits/rejected": -0.9704602956771851,
+      "logps/chosen": -0.6887832880020142,
+      "logps/rejected": -1.2875627279281616,
+      "loss": 0.7609,
+      "odds_ratio_loss": 0.40261825919151306,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.034439168870449066,
+      "rewards/margins": 0.029938969761133194,
+      "rewards/rejected": -0.06437813490629196,
+      "sft_loss": 0.6887832880020142,
+      "step": 1930
+    },
+    {
+      "epoch": 1.548,
+      "grad_norm": 11.200304687438543,
+      "learning_rate": 2.7960099207662535e-06,
+      "logits/chosen": -1.1493829488754272,
+      "logits/rejected": -1.0676811933517456,
+      "logps/chosen": -0.7964752316474915,
+      "logps/rejected": -1.2786619663238525,
+      "loss": 0.8397,
+      "odds_ratio_loss": 0.4415219724178314,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.039823759347200394,
+      "rewards/margins": 0.024109339341521263,
+      "rewards/rejected": -0.0639331042766571,
+      "sft_loss": 0.7964752316474915,
+      "step": 1935
+    },
+    {
+      "epoch": 1.552,
+      "grad_norm": 6.369713349556111,
+      "learning_rate": 2.7844530781306544e-06,
+      "logits/chosen": -0.8758190870285034,
+      "logits/rejected": -0.826004147529602,
+      "logps/chosen": -0.7934783697128296,
+      "logps/rejected": -1.5418189764022827,
+      "loss": 0.7721,
+      "odds_ratio_loss": 0.3249477446079254,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03967391699552536,
+      "rewards/margins": 0.037417031824588776,
+      "rewards/rejected": -0.07709096372127533,
+      "sft_loss": 0.7934783697128296,
+      "step": 1940
+    },
+    {
+      "epoch": 1.556,
+      "grad_norm": 7.211892711878723,
+      "learning_rate": 2.77289007377372e-06,
+      "logits/chosen": -0.8543888926506042,
+      "logits/rejected": -0.8026968240737915,
+      "logps/chosen": -0.5843006372451782,
+      "logps/rejected": -1.1217896938323975,
+      "loss": 0.8268,
+      "odds_ratio_loss": 0.388207346200943,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.02921503409743309,
+      "rewards/margins": 0.02687445841729641,
+      "rewards/rejected": -0.05608949065208435,
+      "sft_loss": 0.5843006372451782,
+      "step": 1945
+    },
+    {
+      "epoch": 1.56,
+      "grad_norm": 6.68682091618046,
+      "learning_rate": 2.761321158169134e-06,
+      "logits/chosen": -1.3163114786148071,
+      "logits/rejected": -0.9326937794685364,
+      "logps/chosen": -0.6405404210090637,
+      "logps/rejected": -1.8683983087539673,
+      "loss": 0.8001,
+      "odds_ratio_loss": 0.3494071960449219,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.032027024775743484,
+      "rewards/margins": 0.0613928958773613,
+      "rewards/rejected": -0.09341991692781448,
+      "sft_loss": 0.6405404210090637,
+      "step": 1950
+    },
+    {
+      "epoch": 1.564,
+      "grad_norm": 8.853441517495,
+      "learning_rate": 2.749746581918629e-06,
+      "logits/chosen": -0.7279099225997925,
+      "logits/rejected": -0.7794903516769409,
+      "logps/chosen": -0.6024842262268066,
+      "logps/rejected": -1.2369228601455688,
+      "loss": 0.8124,
+      "odds_ratio_loss": 0.40761059522628784,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.03012421354651451,
+      "rewards/margins": 0.031721919775009155,
+      "rewards/rejected": -0.061846137046813965,
+      "sft_loss": 0.6024842262268066,
+      "step": 1955
+    },
+    {
+      "epoch": 1.568,
+      "grad_norm": 4.771111743627134,
+      "learning_rate": 2.738166595746554e-06,
+      "logits/chosen": -0.6174139976501465,
+      "logits/rejected": -0.9298956990242004,
+      "logps/chosen": -0.8409628868103027,
+      "logps/rejected": -1.2658686637878418,
+      "loss": 0.8408,
+      "odds_ratio_loss": 0.46160784363746643,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.0420481413602829,
+      "rewards/margins": 0.021245291456580162,
+      "rewards/rejected": -0.0632934421300888,
+      "sft_loss": 0.8409628868103027,
+      "step": 1960
+    },
+    {
+      "epoch": 1.572,
+      "grad_norm": 7.568042308319601,
+      "learning_rate": 2.726581450494451e-06,
+      "logits/chosen": -0.5073860883712769,
+      "logits/rejected": -0.8094170689582825,
+      "logps/chosen": -0.6778584718704224,
+      "logps/rejected": -1.0111925601959229,
+      "loss": 0.7613,
+      "odds_ratio_loss": 0.44990649819374084,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.033892922103405,
+      "rewards/margins": 0.016666702926158905,
+      "rewards/rejected": -0.050559621304273605,
+      "sft_loss": 0.6778584718704224,
+      "step": 1965
+    },
+    {
+      "epoch": 1.576,
+      "grad_norm": 9.06697099963702,
+      "learning_rate": 2.7149913971156105e-06,
+      "logits/chosen": -0.8449646234512329,
+      "logits/rejected": -1.0290577411651611,
+      "logps/chosen": -0.7261394262313843,
+      "logps/rejected": -1.0704646110534668,
+      "loss": 0.9347,
+      "odds_ratio_loss": 0.4873872697353363,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.036306969821453094,
+      "rewards/margins": 0.017216259613633156,
+      "rewards/rejected": -0.0535232312977314,
+      "sft_loss": 0.7261394262313843,
+      "step": 1970
+    },
+    {
+      "epoch": 1.58,
+      "grad_norm": 7.716541168169069,
+      "learning_rate": 2.703396686669646e-06,
+      "logits/chosen": -0.43536773324012756,
+      "logits/rejected": -1.230164647102356,
+      "logps/chosen": -0.9158234596252441,
+      "logps/rejected": -1.4675378799438477,
+      "loss": 0.8256,
+      "odds_ratio_loss": 0.4452730119228363,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.045791175216436386,
+      "rewards/margins": 0.027585718780755997,
+      "rewards/rejected": -0.07337689399719238,
+      "sft_loss": 0.9158234596252441,
+      "step": 1975
+    },
+    {
+      "epoch": 1.584,
+      "grad_norm": 11.344687962958782,
+      "learning_rate": 2.6917975703170466e-06,
+      "logits/chosen": -0.7123010158538818,
+      "logits/rejected": -1.2155797481536865,
+      "logps/chosen": -0.6988154649734497,
+      "logps/rejected": -1.039637565612793,
+      "loss": 0.7954,
+      "odds_ratio_loss": 0.4463338255882263,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.034940771758556366,
+      "rewards/margins": 0.017041107639670372,
+      "rewards/rejected": -0.05198187753558159,
+      "sft_loss": 0.6988154649734497,
+      "step": 1980
+    },
+    {
+      "epoch": 1.588,
+      "grad_norm": 7.131197917094538,
+      "learning_rate": 2.6801942993137435e-06,
+      "logits/chosen": -0.5410433411598206,
+      "logits/rejected": -1.1760857105255127,
+      "logps/chosen": -0.8232455253601074,
+      "logps/rejected": -1.0818697214126587,
+      "loss": 0.8893,
+      "odds_ratio_loss": 0.5539475679397583,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.04116227477788925,
+      "rewards/margins": 0.012931210920214653,
+      "rewards/rejected": -0.05409349128603935,
+      "sft_loss": 0.8232455253601074,
+      "step": 1985
+    },
+    {
+      "epoch": 1.592,
+      "grad_norm": 5.5964963412533555,
+      "learning_rate": 2.668587125005663e-06,
+      "logits/chosen": -0.8460969924926758,
+      "logits/rejected": -1.182850956916809,
+      "logps/chosen": -0.6967536807060242,
+      "logps/rejected": -1.3319151401519775,
+      "loss": 0.7849,
+      "odds_ratio_loss": 0.41814327239990234,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.03483767807483673,
+      "rewards/margins": 0.03175807744264603,
+      "rewards/rejected": -0.06659575551748276,
+      "sft_loss": 0.6967536807060242,
+      "step": 1990
+    },
+    {
+      "epoch": 1.596,
+      "grad_norm": 5.763026931052832,
+      "learning_rate": 2.6569762988232838e-06,
+      "logits/chosen": -0.8818701505661011,
+      "logits/rejected": -1.0999513864517212,
+      "logps/chosen": -0.7708557844161987,
+      "logps/rejected": -1.50910222530365,
+      "loss": 0.7027,
+      "odds_ratio_loss": 0.3464539647102356,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.03854278475046158,
+      "rewards/margins": 0.03691232204437256,
+      "rewards/rejected": -0.07545509934425354,
+      "sft_loss": 0.7708557844161987,
+      "step": 1995
+    },
+    {
+      "epoch": 1.6,
+      "grad_norm": 4.594815757134377,
+      "learning_rate": 2.6453620722761897e-06,
+      "logits/chosen": -0.7889989018440247,
+      "logits/rejected": -1.0313599109649658,
+      "logps/chosen": -1.0150768756866455,
+      "logps/rejected": -1.3293017148971558,
+      "loss": 0.8486,
+      "odds_ratio_loss": 0.5205736756324768,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.050753843039274216,
+      "rewards/margins": 0.01571125164628029,
+      "rewards/rejected": -0.0664650946855545,
+      "sft_loss": 1.0150768756866455,
+      "step": 2000
+    },
+    {
+      "epoch": 1.604,
+      "grad_norm": 5.622181491776131,
+      "learning_rate": 2.6337446969476234e-06,
+      "logits/chosen": -0.6143472790718079,
+      "logits/rejected": -0.8800684213638306,
+      "logps/chosen": -0.70451819896698,
+      "logps/rejected": -1.127333641052246,
+      "loss": 0.7694,
+      "odds_ratio_loss": 0.47300809621810913,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.03522591292858124,
+      "rewards/margins": 0.021140772849321365,
+      "rewards/rejected": -0.056366682052612305,
+      "sft_loss": 0.70451819896698,
+      "step": 2005
+    },
+    {
+      "epoch": 1.608,
+      "grad_norm": 5.4411456874419635,
+      "learning_rate": 2.6221244244890336e-06,
+      "logits/chosen": -0.8750918507575989,
+      "logits/rejected": -1.0322843790054321,
+      "logps/chosen": -1.0601109266281128,
+      "logps/rejected": -1.424378514289856,
+      "loss": 0.8386,
+      "odds_ratio_loss": 0.5982974767684937,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": -0.05300554633140564,
+      "rewards/margins": 0.018213381990790367,
+      "rewards/rejected": -0.07121893018484116,
+      "sft_loss": 1.0601109266281128,
+      "step": 2010
+    },
+    {
+      "epoch": 1.612,
+      "grad_norm": 6.189805248319622,
+      "learning_rate": 2.6105015066146266e-06,
+      "logits/chosen": -1.0919370651245117,
+      "logits/rejected": -0.9859377145767212,
+      "logps/chosen": -0.7465149164199829,
+      "logps/rejected": -1.1737520694732666,
+      "loss": 0.7624,
+      "odds_ratio_loss": 0.41314896941185,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.037325747311115265,
+      "rewards/margins": 0.021361857652664185,
+      "rewards/rejected": -0.05868760868906975,
+      "sft_loss": 0.7465149164199829,
+      "step": 2015
+    },
+    {
+      "epoch": 1.616,
+      "grad_norm": 9.164029729819344,
+      "learning_rate": 2.5988761950959133e-06,
+      "logits/chosen": -0.8967428207397461,
+      "logits/rejected": -0.6835768818855286,
+      "logps/chosen": -0.7614465951919556,
+      "logps/rejected": -1.4321725368499756,
+      "loss": 0.8553,
+      "odds_ratio_loss": 0.4023088812828064,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03807232901453972,
+      "rewards/margins": 0.03353629633784294,
+      "rewards/rejected": -0.07160861790180206,
+      "sft_loss": 0.7614465951919556,
+      "step": 2020
+    },
+    {
+      "epoch": 1.62,
+      "grad_norm": 8.085049537513227,
+      "learning_rate": 2.587248741756253e-06,
+      "logits/chosen": -0.7910897731781006,
+      "logits/rejected": -1.1293184757232666,
+      "logps/chosen": -0.7620183229446411,
+      "logps/rejected": -1.539175033569336,
+      "loss": 0.7589,
+      "odds_ratio_loss": 0.3549695909023285,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.038100920617580414,
+      "rewards/margins": 0.0388578325510025,
+      "rewards/rejected": -0.07695874571800232,
+      "sft_loss": 0.7620183229446411,
+      "step": 2025
+    },
+    {
+      "epoch": 1.624,
+      "grad_norm": 10.097540227811598,
+      "learning_rate": 2.575619398465402e-06,
+      "logits/chosen": -0.8627431988716125,
+      "logits/rejected": -0.9311240315437317,
+      "logps/chosen": -0.877997875213623,
+      "logps/rejected": -1.0316495895385742,
+      "loss": 0.8384,
+      "odds_ratio_loss": 0.5710136890411377,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.04389990121126175,
+      "rewards/margins": 0.007682584226131439,
+      "rewards/rejected": -0.05158247798681259,
+      "sft_loss": 0.877997875213623,
+      "step": 2030
+    },
+    {
+      "epoch": 1.6280000000000001,
+      "grad_norm": 7.090714107623907,
+      "learning_rate": 2.563988417134056e-06,
+      "logits/chosen": -0.9987133145332336,
+      "logits/rejected": -0.8781261444091797,
+      "logps/chosen": -0.5755537748336792,
+      "logps/rejected": -1.2042471170425415,
+      "loss": 0.8191,
+      "odds_ratio_loss": 0.31655603647232056,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02877769246697426,
+      "rewards/margins": 0.03143466264009476,
+      "rewards/rejected": -0.06021235138177872,
+      "sft_loss": 0.5755537748336792,
+      "step": 2035
+    },
+    {
+      "epoch": 1.6320000000000001,
+      "grad_norm": 7.424256408913974,
+      "learning_rate": 2.5523560497083927e-06,
+      "logits/chosen": -0.9880864024162292,
+      "logits/rejected": -1.3933836221694946,
+      "logps/chosen": -0.6906462907791138,
+      "logps/rejected": -1.200685739517212,
+      "loss": 0.8062,
+      "odds_ratio_loss": 0.36662763357162476,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03453231602907181,
+      "rewards/margins": 0.025501970201730728,
+      "rewards/rejected": -0.060034286230802536,
+      "sft_loss": 0.6906462907791138,
+      "step": 2040
+    },
+    {
+      "epoch": 1.6360000000000001,
+      "grad_norm": 5.662873396280538,
+      "learning_rate": 2.5407225481646146e-06,
+      "logits/chosen": -0.638294517993927,
+      "logits/rejected": -0.9844516515731812,
+      "logps/chosen": -0.8093409538269043,
+      "logps/rejected": -1.5629360675811768,
+      "loss": 0.8729,
+      "odds_ratio_loss": 0.4065137505531311,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.040467046201229095,
+      "rewards/margins": 0.037679754197597504,
+      "rewards/rejected": -0.0781468003988266,
+      "sft_loss": 0.8093409538269043,
+      "step": 2045
+    },
+    {
+      "epoch": 1.6400000000000001,
+      "grad_norm": 4.8043102192548925,
+      "learning_rate": 2.5290881645034932e-06,
+      "logits/chosen": -0.5350710153579712,
+      "logits/rejected": -1.2690541744232178,
+      "logps/chosen": -0.8825578689575195,
+      "logps/rejected": -1.1409144401550293,
+      "loss": 0.8776,
+      "odds_ratio_loss": 0.7036653757095337,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.04412789270281792,
+      "rewards/margins": 0.01291782595217228,
+      "rewards/rejected": -0.057045720517635345,
+      "sft_loss": 0.8825578689575195,
+      "step": 2050
+    },
+    {
+      "epoch": 1.6440000000000001,
+      "grad_norm": 5.31433548093374,
+      "learning_rate": 2.517453150744904e-06,
+      "logits/chosen": -0.8365262746810913,
+      "logits/rejected": -0.594868540763855,
+      "logps/chosen": -0.7498319745063782,
+      "logps/rejected": -1.5040110349655151,
+      "loss": 0.7332,
+      "odds_ratio_loss": 0.33074039220809937,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03749159723520279,
+      "rewards/margins": 0.03770895302295685,
+      "rewards/rejected": -0.07520055025815964,
+      "sft_loss": 0.7498319745063782,
+      "step": 2055
+    },
+    {
+      "epoch": 1.6480000000000001,
+      "grad_norm": 6.655827286565878,
+      "learning_rate": 2.5058177589223766e-06,
+      "logits/chosen": -0.8108224868774414,
+      "logits/rejected": -0.8111976385116577,
+      "logps/chosen": -0.8283792734146118,
+      "logps/rejected": -1.2744197845458984,
+      "loss": 0.874,
+      "odds_ratio_loss": 0.4357910752296448,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.04141896218061447,
+      "rewards/margins": 0.022302016615867615,
+      "rewards/rejected": -0.06372098624706268,
+      "sft_loss": 0.8283792734146118,
+      "step": 2060
+    },
+    {
+      "epoch": 1.6520000000000001,
+      "grad_norm": 8.036123405257694,
+      "learning_rate": 2.4941822410776247e-06,
+      "logits/chosen": -1.0825451612472534,
+      "logits/rejected": -1.170818567276001,
+      "logps/chosen": -0.9788244366645813,
+      "logps/rejected": -1.9613186120986938,
+      "loss": 0.7772,
+      "odds_ratio_loss": 0.3012261986732483,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.04894121736288071,
+      "rewards/margins": 0.04912471026182175,
+      "rewards/rejected": -0.09806593507528305,
+      "sft_loss": 0.9788244366645813,
+      "step": 2065
+    },
+    {
+      "epoch": 1.6560000000000001,
+      "grad_norm": 6.061444627487459,
+      "learning_rate": 2.482546849255096e-06,
+      "logits/chosen": -0.7679504752159119,
+      "logits/rejected": -0.9109829664230347,
+      "logps/chosen": -0.6975770592689514,
+      "logps/rejected": -1.4365979433059692,
+      "loss": 0.8016,
+      "odds_ratio_loss": 0.29892677068710327,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03487885370850563,
+      "rewards/margins": 0.03695103898644447,
+      "rewards/rejected": -0.0718298926949501,
+      "sft_loss": 0.6975770592689514,
+      "step": 2070
+    },
+    {
+      "epoch": 1.6600000000000001,
+      "grad_norm": 4.509609302945273,
+      "learning_rate": 2.470911835496508e-06,
+      "logits/chosen": -0.9424687623977661,
+      "logits/rejected": -1.1301240921020508,
+      "logps/chosen": -0.6970169544219971,
+      "logps/rejected": -1.6458070278167725,
+      "loss": 0.8016,
+      "odds_ratio_loss": 0.3055366277694702,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.034850846976041794,
+      "rewards/margins": 0.04743950814008713,
+      "rewards/rejected": -0.08229035139083862,
+      "sft_loss": 0.6970169544219971,
+      "step": 2075
+    },
+    {
+      "epoch": 1.6640000000000001,
+      "grad_norm": 7.765550436552941,
+      "learning_rate": 2.4592774518353858e-06,
+      "logits/chosen": -0.8300328254699707,
+      "logits/rejected": -1.3914577960968018,
+      "logps/chosen": -0.7543548345565796,
+      "logps/rejected": -1.325756311416626,
+      "loss": 0.8444,
+      "odds_ratio_loss": 0.37441879510879517,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03771774470806122,
+      "rewards/margins": 0.028570080175995827,
+      "rewards/rejected": -0.0662878230214119,
+      "sft_loss": 0.7543548345565796,
+      "step": 2080
+    },
+    {
+      "epoch": 1.6680000000000001,
+      "grad_norm": 5.908452371029711,
+      "learning_rate": 2.447643950291608e-06,
+      "logits/chosen": -0.6916254758834839,
+      "logits/rejected": -1.0722490549087524,
+      "logps/chosen": -0.8863399624824524,
+      "logps/rejected": -1.6229625940322876,
+      "loss": 0.7622,
+      "odds_ratio_loss": 0.4502854347229004,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.04431699588894844,
+      "rewards/margins": 0.036831121891736984,
+      "rewards/rejected": -0.08114812523126602,
+      "sft_loss": 0.8863399624824524,
+      "step": 2085
+    },
+    {
+      "epoch": 1.6720000000000002,
+      "grad_norm": 12.568326286094635,
+      "learning_rate": 2.436011582865945e-06,
+      "logits/chosen": -0.764611542224884,
+      "logits/rejected": -1.219498872756958,
+      "logps/chosen": -0.8522571325302124,
+      "logps/rejected": -1.5530402660369873,
+      "loss": 0.8979,
+      "odds_ratio_loss": 0.4483674466609955,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.04261285811662674,
+      "rewards/margins": 0.035039156675338745,
+      "rewards/rejected": -0.07765202224254608,
+      "sft_loss": 0.8522571325302124,
+      "step": 2090
+    },
+    {
+      "epoch": 1.6760000000000002,
+      "grad_norm": 5.284478090573637,
+      "learning_rate": 2.4243806015345988e-06,
+      "logits/chosen": -1.1326720714569092,
+      "logits/rejected": -0.9926662445068359,
+      "logps/chosen": -0.636461615562439,
+      "logps/rejected": -1.5491859912872314,
+      "loss": 0.7837,
+      "odds_ratio_loss": 0.3587147295475006,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.03182308003306389,
+      "rewards/margins": 0.04563622921705246,
+      "rewards/rejected": -0.07745930552482605,
+      "sft_loss": 0.636461615562439,
+      "step": 2095
+    },
+    {
+      "epoch": 1.6800000000000002,
+      "grad_norm": 5.748225305959376,
+      "learning_rate": 2.4127512582437486e-06,
+      "logits/chosen": -0.8698973655700684,
+      "logits/rejected": -1.1889925003051758,
+      "logps/chosen": -0.825188934803009,
+      "logps/rejected": -1.2668731212615967,
+      "loss": 0.916,
+      "odds_ratio_loss": 0.5938987731933594,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.041259441524744034,
+      "rewards/margins": 0.022084210067987442,
+      "rewards/rejected": -0.06334365904331207,
+      "sft_loss": 0.825188934803009,
+      "step": 2100
+    },
+    {
+      "epoch": 1.6840000000000002,
+      "grad_norm": 13.816645361412702,
+      "learning_rate": 2.4011238049040875e-06,
+      "logits/chosen": -0.9653146862983704,
+      "logits/rejected": -1.273425817489624,
+      "logps/chosen": -0.48048973083496094,
+      "logps/rejected": -1.9898446798324585,
+      "loss": 0.7611,
+      "odds_ratio_loss": 0.20829610526561737,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.024024488404393196,
+      "rewards/margins": 0.07546775043010712,
+      "rewards/rejected": -0.09949223697185516,
+      "sft_loss": 0.48048973083496094,
+      "step": 2105
+    },
+    {
+      "epoch": 1.688,
+      "grad_norm": 7.128418390551076,
+      "learning_rate": 2.3894984933853734e-06,
+      "logits/chosen": -1.2187275886535645,
+      "logits/rejected": -0.9569811820983887,
+      "logps/chosen": -0.7580611109733582,
+      "logps/rejected": -1.5453588962554932,
+      "loss": 0.7746,
+      "odds_ratio_loss": 0.32491961121559143,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.037903059273958206,
+      "rewards/margins": 0.03936489298939705,
+      "rewards/rejected": -0.07726794481277466,
+      "sft_loss": 0.7580611109733582,
+      "step": 2110
+    },
+    {
+      "epoch": 1.692,
+      "grad_norm": 9.905564606255297,
+      "learning_rate": 2.377875575510967e-06,
+      "logits/chosen": -0.6468038558959961,
+      "logits/rejected": -1.1671807765960693,
+      "logps/chosen": -0.8859814405441284,
+      "logps/rejected": -1.915808916091919,
+      "loss": 0.8705,
+      "odds_ratio_loss": 0.3132167458534241,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.04429907351732254,
+      "rewards/margins": 0.05149136856198311,
+      "rewards/rejected": -0.09579044580459595,
+      "sft_loss": 0.8859814405441284,
+      "step": 2115
+    },
+    {
+      "epoch": 1.696,
+      "grad_norm": 5.125087558911786,
+      "learning_rate": 2.366255303052377e-06,
+      "logits/chosen": -0.7792321443557739,
+      "logits/rejected": -0.8570452928543091,
+      "logps/chosen": -1.0220623016357422,
+      "logps/rejected": -1.2880709171295166,
+      "loss": 0.7697,
+      "odds_ratio_loss": 0.534646213054657,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05110311508178711,
+      "rewards/margins": 0.013300428166985512,
+      "rewards/rejected": -0.06440354138612747,
+      "sft_loss": 1.0220623016357422,
+      "step": 2120
+    },
+    {
+      "epoch": 1.7,
+      "grad_norm": 7.717361248878697,
+      "learning_rate": 2.3546379277238107e-06,
+      "logits/chosen": -0.8779585957527161,
+      "logits/rejected": -1.2526636123657227,
+      "logps/chosen": -1.0405645370483398,
+      "logps/rejected": -1.4684456586837769,
+      "loss": 0.8437,
+      "odds_ratio_loss": 0.5074051022529602,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.05202822759747505,
+      "rewards/margins": 0.02139405533671379,
+      "rewards/rejected": -0.07342227548360825,
+      "sft_loss": 1.0405645370483398,
+      "step": 2125
+    },
+    {
+      "epoch": 1.704,
+      "grad_norm": 8.595275466816915,
+      "learning_rate": 2.3430237011767166e-06,
+      "logits/chosen": -0.7775979042053223,
+      "logits/rejected": -1.551299810409546,
+      "logps/chosen": -0.7821834683418274,
+      "logps/rejected": -2.03926944732666,
+      "loss": 0.8241,
+      "odds_ratio_loss": 0.37603259086608887,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03910917788743973,
+      "rewards/margins": 0.06285430490970612,
+      "rewards/rejected": -0.10196347534656525,
+      "sft_loss": 0.7821834683418274,
+      "step": 2130
+    },
+    {
+      "epoch": 1.708,
+      "grad_norm": 5.594173003610877,
+      "learning_rate": 2.3314128749943376e-06,
+      "logits/chosen": -1.2585442066192627,
+      "logits/rejected": -0.8739719390869141,
+      "logps/chosen": -1.027630090713501,
+      "logps/rejected": -1.4903634786605835,
+      "loss": 0.7552,
+      "odds_ratio_loss": 0.51094651222229,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.05138150602579117,
+      "rewards/margins": 0.023136669769883156,
+      "rewards/rejected": -0.07451816648244858,
+      "sft_loss": 1.027630090713501,
+      "step": 2135
+    },
+    {
+      "epoch": 1.712,
+      "grad_norm": 6.179490163786233,
+      "learning_rate": 2.319805700686257e-06,
+      "logits/chosen": -0.7619789838790894,
+      "logits/rejected": -1.402626633644104,
+      "logps/chosen": -1.0751920938491821,
+      "logps/rejected": -1.9567829370498657,
+      "loss": 0.8477,
+      "odds_ratio_loss": 0.3091539442539215,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.053759604692459106,
+      "rewards/margins": 0.04407954216003418,
+      "rewards/rejected": -0.09783915430307388,
+      "sft_loss": 1.0751920938491821,
+      "step": 2140
+    },
+    {
+      "epoch": 1.716,
+      "grad_norm": 5.014263826927916,
+      "learning_rate": 2.3082024296829538e-06,
+      "logits/chosen": -1.2588303089141846,
+      "logits/rejected": -0.8437775373458862,
+      "logps/chosen": -0.7372163534164429,
+      "logps/rejected": -1.173533320426941,
+      "loss": 0.7309,
+      "odds_ratio_loss": 0.4651245176792145,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.036860816180706024,
+      "rewards/margins": 0.021815849468111992,
+      "rewards/rejected": -0.058676671236753464,
+      "sft_loss": 0.7372163534164429,
+      "step": 2145
+    },
+    {
+      "epoch": 1.72,
+      "grad_norm": 12.671731572893654,
+      "learning_rate": 2.296603313330355e-06,
+      "logits/chosen": -0.940277099609375,
+      "logits/rejected": -0.9961813688278198,
+      "logps/chosen": -0.6746289134025574,
+      "logps/rejected": -1.4196538925170898,
+      "loss": 0.79,
+      "odds_ratio_loss": 0.29419276118278503,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.033731453120708466,
+      "rewards/margins": 0.037251245230436325,
+      "rewards/rejected": -0.07098269462585449,
+      "sft_loss": 0.6746289134025574,
+      "step": 2150
+    },
+    {
+      "epoch": 1.724,
+      "grad_norm": 8.807564564413038,
+      "learning_rate": 2.2850086028843894e-06,
+      "logits/chosen": -0.5990036129951477,
+      "logits/rejected": -1.1123206615447998,
+      "logps/chosen": -0.7771926522254944,
+      "logps/rejected": -1.3297137022018433,
+      "loss": 0.7942,
+      "odds_ratio_loss": 0.3976249098777771,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03885963559150696,
+      "rewards/margins": 0.027626052498817444,
+      "rewards/rejected": -0.0664856880903244,
+      "sft_loss": 0.7771926522254944,
+      "step": 2155
+    },
+    {
+      "epoch": 1.728,
+      "grad_norm": 6.458841654862515,
+      "learning_rate": 2.2734185495055503e-06,
+      "logits/chosen": -1.034919261932373,
+      "logits/rejected": -0.8573856353759766,
+      "logps/chosen": -0.7438480257987976,
+      "logps/rejected": -1.2708542346954346,
+      "loss": 0.6921,
+      "odds_ratio_loss": 0.4702334403991699,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.03719240427017212,
+      "rewards/margins": 0.026350298896431923,
+      "rewards/rejected": -0.06354270875453949,
+      "sft_loss": 0.7438480257987976,
+      "step": 2160
+    },
+    {
+      "epoch": 1.732,
+      "grad_norm": 10.704475448149418,
+      "learning_rate": 2.2618334042534464e-06,
+      "logits/chosen": -0.9798957705497742,
+      "logits/rejected": -0.9897342920303345,
+      "logps/chosen": -0.6244773864746094,
+      "logps/rejected": -1.1638376712799072,
+      "loss": 0.7625,
+      "odds_ratio_loss": 0.35523417592048645,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03122386895120144,
+      "rewards/margins": 0.026968013495206833,
+      "rewards/rejected": -0.05819188430905342,
+      "sft_loss": 0.6244773864746094,
+      "step": 2165
+    },
+    {
+      "epoch": 1.736,
+      "grad_norm": 8.192553193640592,
+      "learning_rate": 2.250253418081373e-06,
+      "logits/chosen": -0.960060715675354,
+      "logits/rejected": -0.8580840229988098,
+      "logps/chosen": -0.7168663740158081,
+      "logps/rejected": -1.152551531791687,
+      "loss": 0.7857,
+      "odds_ratio_loss": 0.495716392993927,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.03584332391619682,
+      "rewards/margins": 0.021784260869026184,
+      "rewards/rejected": -0.05762758105993271,
+      "sft_loss": 0.7168663740158081,
+      "step": 2170
+    },
+    {
+      "epoch": 1.74,
+      "grad_norm": 8.321261439994862,
+      "learning_rate": 2.238678841830867e-06,
+      "logits/chosen": -0.7686837911605835,
+      "logits/rejected": -1.141378402709961,
+      "logps/chosen": -1.0380868911743164,
+      "logps/rejected": -1.489746332168579,
+      "loss": 0.806,
+      "odds_ratio_loss": 0.4990416467189789,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.0519043430685997,
+      "rewards/margins": 0.022582972422242165,
+      "rewards/rejected": -0.07448731362819672,
+      "sft_loss": 1.0380868911743164,
+      "step": 2175
+    },
+    {
+      "epoch": 1.744,
+      "grad_norm": 6.016319364654182,
+      "learning_rate": 2.22710992622628e-06,
+      "logits/chosen": -1.1537871360778809,
+      "logits/rejected": -0.9235156774520874,
+      "logps/chosen": -1.0170602798461914,
+      "logps/rejected": -2.0361647605895996,
+      "loss": 0.8776,
+      "odds_ratio_loss": 0.5887846350669861,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.05085300654172897,
+      "rewards/margins": 0.05095522850751877,
+      "rewards/rejected": -0.10180824995040894,
+      "sft_loss": 1.0170602798461914,
+      "step": 2180
+    },
+    {
+      "epoch": 1.748,
+      "grad_norm": 10.90847815557463,
+      "learning_rate": 2.2155469218693464e-06,
+      "logits/chosen": -0.8062864542007446,
+      "logits/rejected": -0.944413959980011,
+      "logps/chosen": -0.8070542216300964,
+      "logps/rejected": -1.5685523748397827,
+      "loss": 0.873,
+      "odds_ratio_loss": 0.5560692548751831,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.0403527170419693,
+      "rewards/margins": 0.038074906915426254,
+      "rewards/rejected": -0.07842762768268585,
+      "sft_loss": 0.8070542216300964,
+      "step": 2185
+    },
+    {
+      "epoch": 1.752,
+      "grad_norm": 12.956912471620969,
+      "learning_rate": 2.2039900792337477e-06,
+      "logits/chosen": -0.9245501756668091,
+      "logits/rejected": -0.7286485433578491,
+      "logps/chosen": -1.1378872394561768,
+      "logps/rejected": -1.3941930532455444,
+      "loss": 0.8626,
+      "odds_ratio_loss": 0.5717015266418457,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.056894369423389435,
+      "rewards/margins": 0.012815283611416817,
+      "rewards/rejected": -0.0697096511721611,
+      "sft_loss": 1.1378872394561768,
+      "step": 2190
+    },
+    {
+      "epoch": 1.756,
+      "grad_norm": 11.281398520267489,
+      "learning_rate": 2.192439648659699e-06,
+      "logits/chosen": -1.042218804359436,
+      "logits/rejected": -0.6264259219169617,
+      "logps/chosen": -1.035028338432312,
+      "logps/rejected": -1.542610764503479,
+      "loss": 0.8083,
+      "odds_ratio_loss": 0.5132459998130798,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05175141617655754,
+      "rewards/margins": 0.025379126891493797,
+      "rewards/rejected": -0.07713054120540619,
+      "sft_loss": 1.035028338432312,
+      "step": 2195
+    },
+    {
+      "epoch": 1.76,
+      "grad_norm": 7.2793892053250735,
+      "learning_rate": 2.1808958803485134e-06,
+      "logits/chosen": -1.051295280456543,
+      "logits/rejected": -0.6670552492141724,
+      "logps/chosen": -0.6154602766036987,
+      "logps/rejected": -1.3951618671417236,
+      "loss": 0.8288,
+      "odds_ratio_loss": 0.2627789378166199,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.030773013830184937,
+      "rewards/margins": 0.038985081017017365,
+      "rewards/rejected": -0.0697580948472023,
+      "sft_loss": 0.6154602766036987,
+      "step": 2200
+    },
+    {
+      "epoch": 1.764,
+      "grad_norm": 6.172083283993613,
+      "learning_rate": 2.1693590243571937e-06,
+      "logits/chosen": -0.8260824084281921,
+      "logits/rejected": -1.0545563697814941,
+      "logps/chosen": -0.8104951977729797,
+      "logps/rejected": -1.3007080554962158,
+      "loss": 0.7999,
+      "odds_ratio_loss": 0.3759697377681732,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.04052475839853287,
+      "rewards/margins": 0.024510642513632774,
+      "rewards/rejected": -0.0650353953242302,
+      "sft_loss": 0.8104951977729797,
+      "step": 2205
+    },
+    {
+      "epoch": 1.768,
+      "grad_norm": 5.796258673720241,
+      "learning_rate": 2.157829330593008e-06,
+      "logits/chosen": -0.5958111882209778,
+      "logits/rejected": -1.353058099746704,
+      "logps/chosen": -0.7925723791122437,
+      "logps/rejected": -1.575204610824585,
+      "loss": 0.8336,
+      "odds_ratio_loss": 0.31617340445518494,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03962862119078636,
+      "rewards/margins": 0.03913161903619766,
+      "rewards/rejected": -0.07876023650169373,
+      "sft_loss": 0.7925723791122437,
+      "step": 2210
+    },
+    {
+      "epoch": 1.772,
+      "grad_norm": 5.853650964962732,
+      "learning_rate": 2.1463070488080777e-06,
+      "logits/chosen": -0.7320128679275513,
+      "logits/rejected": -1.4626697301864624,
+      "logps/chosen": -0.920146107673645,
+      "logps/rejected": -1.587083101272583,
+      "loss": 0.912,
+      "odds_ratio_loss": 0.36855971813201904,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.046007297933101654,
+      "rewards/margins": 0.03334684669971466,
+      "rewards/rejected": -0.07935415208339691,
+      "sft_loss": 0.920146107673645,
+      "step": 2215
+    },
+    {
+      "epoch": 1.776,
+      "grad_norm": 7.102475221339332,
+      "learning_rate": 2.134792428593971e-06,
+      "logits/chosen": -1.0143307447433472,
+      "logits/rejected": -1.0723531246185303,
+      "logps/chosen": -0.6754562854766846,
+      "logps/rejected": -1.2191624641418457,
+      "loss": 0.8002,
+      "odds_ratio_loss": 0.4698086380958557,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.03377281501889229,
+      "rewards/margins": 0.027185309678316116,
+      "rewards/rejected": -0.060958124697208405,
+      "sft_loss": 0.6754562854766846,
+      "step": 2220
+    },
+    {
+      "epoch": 1.78,
+      "grad_norm": 5.132375215504824,
+      "learning_rate": 2.1232857193762923e-06,
+      "logits/chosen": -0.8438177108764648,
+      "logits/rejected": -0.8187214136123657,
+      "logps/chosen": -0.9790526628494263,
+      "logps/rejected": -1.4144220352172852,
+      "loss": 0.8596,
+      "odds_ratio_loss": 0.5008438229560852,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.048952627927064896,
+      "rewards/margins": 0.021768469363451004,
+      "rewards/rejected": -0.0707211047410965,
+      "sft_loss": 0.9790526628494263,
+      "step": 2225
+    },
+    {
+      "epoch": 1.784,
+      "grad_norm": 5.864011412934527,
+      "learning_rate": 2.1117871704092818e-06,
+      "logits/chosen": -0.6806105375289917,
+      "logits/rejected": -1.0006288290023804,
+      "logps/chosen": -0.6304813027381897,
+      "logps/rejected": -2.105440378189087,
+      "loss": 0.8079,
+      "odds_ratio_loss": 0.289185106754303,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.031524065881967545,
+      "rewards/margins": 0.07374795526266098,
+      "rewards/rejected": -0.10527201741933823,
+      "sft_loss": 0.6304813027381897,
+      "step": 2230
+    },
+    {
+      "epoch": 1.788,
+      "grad_norm": 5.2792769917747755,
+      "learning_rate": 2.1002970307704134e-06,
+      "logits/chosen": -1.001686930656433,
+      "logits/rejected": -0.7307732105255127,
+      "logps/chosen": -0.4030815660953522,
+      "logps/rejected": -2.1336350440979004,
+      "loss": 0.8382,
+      "odds_ratio_loss": 0.0816396027803421,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.020154079422354698,
+      "rewards/margins": 0.08652767539024353,
+      "rewards/rejected": -0.10668174922466278,
+      "sft_loss": 0.4030815660953522,
+      "step": 2235
+    },
+    {
+      "epoch": 1.792,
+      "grad_norm": 6.219063011453153,
+      "learning_rate": 2.0888155493550027e-06,
+      "logits/chosen": -0.6964353322982788,
+      "logits/rejected": -0.9690049290657043,
+      "logps/chosen": -0.9550386667251587,
+      "logps/rejected": -1.2657750844955444,
+      "loss": 0.9063,
+      "odds_ratio_loss": 0.5276353359222412,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.047751929610967636,
+      "rewards/margins": 0.015536829829216003,
+      "rewards/rejected": -0.06328876316547394,
+      "sft_loss": 0.9550386667251587,
+      "step": 2240
+    },
+    {
+      "epoch": 1.796,
+      "grad_norm": 6.217554170161715,
+      "learning_rate": 2.0773429748708153e-06,
+      "logits/chosen": -1.1106126308441162,
+      "logits/rejected": -0.8946078419685364,
+      "logps/chosen": -0.7019718289375305,
+      "logps/rejected": -2.1581504344940186,
+      "loss": 0.7725,
+      "odds_ratio_loss": 0.34209513664245605,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.035098589956760406,
+      "rewards/margins": 0.07280893623828888,
+      "rewards/rejected": -0.10790753364562988,
+      "sft_loss": 0.7019718289375305,
+      "step": 2245
+    },
+    {
+      "epoch": 1.8,
+      "grad_norm": 6.803417111045092,
+      "learning_rate": 2.0658795558326745e-06,
+      "logits/chosen": -1.081549048423767,
+      "logits/rejected": -0.8075908422470093,
+      "logps/chosen": -1.0295336246490479,
+      "logps/rejected": -1.406858205795288,
+      "loss": 0.8343,
+      "odds_ratio_loss": 0.48274731636047363,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.05147667974233627,
+      "rewards/margins": 0.018866227939724922,
+      "rewards/rejected": -0.07034290581941605,
+      "sft_loss": 1.0295336246490479,
+      "step": 2250
+    },
+    {
+      "epoch": 1.804,
+      "grad_norm": 4.989031637803273,
+      "learning_rate": 2.0544255405570843e-06,
+      "logits/chosen": -0.9463821649551392,
+      "logits/rejected": -1.047937035560608,
+      "logps/chosen": -0.6406688690185547,
+      "logps/rejected": -1.53047776222229,
+      "loss": 0.8074,
+      "odds_ratio_loss": 0.3679446578025818,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.032033443450927734,
+      "rewards/margins": 0.04449043795466423,
+      "rewards/rejected": -0.07652387768030167,
+      "sft_loss": 0.6406688690185547,
+      "step": 2255
+    },
+    {
+      "epoch": 1.808,
+      "grad_norm": 6.658453228798975,
+      "learning_rate": 2.0429811771568468e-06,
+      "logits/chosen": -0.7510989308357239,
+      "logits/rejected": -1.236093282699585,
+      "logps/chosen": -0.8555262684822083,
+      "logps/rejected": -1.224966049194336,
+      "loss": 0.8413,
+      "odds_ratio_loss": 0.4976680874824524,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.04277631267905235,
+      "rewards/margins": 0.01847199536859989,
+      "rewards/rejected": -0.061248309910297394,
+      "sft_loss": 0.8555262684822083,
+      "step": 2260
+    },
+    {
+      "epoch": 1.812,
+      "grad_norm": 4.401496364800286,
+      "learning_rate": 2.031546713535688e-06,
+      "logits/chosen": -0.9858955144882202,
+      "logits/rejected": -0.9150092005729675,
+      "logps/chosen": -0.7890244722366333,
+      "logps/rejected": -1.2879811525344849,
+      "loss": 0.7905,
+      "odds_ratio_loss": 0.3429797291755676,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.039451222866773605,
+      "rewards/margins": 0.02494782581925392,
+      "rewards/rejected": -0.06439904868602753,
+      "sft_loss": 0.7890244722366333,
+      "step": 2265
+    },
+    {
+      "epoch": 1.8159999999999998,
+      "grad_norm": 9.174834943476265,
+      "learning_rate": 2.0201223973828917e-06,
+      "logits/chosen": -0.8151395916938782,
+      "logits/rejected": -1.2962208986282349,
+      "logps/chosen": -0.7152162790298462,
+      "logps/rejected": -1.581583857536316,
+      "loss": 0.7903,
+      "odds_ratio_loss": 0.385868102312088,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.03576081246137619,
+      "rewards/margins": 0.04331838712096214,
+      "rewards/rejected": -0.07907919585704803,
+      "sft_loss": 0.7152162790298462,
+      "step": 2270
+    },
+    {
+      "epoch": 1.8199999999999998,
+      "grad_norm": 6.3991437024550475,
+      "learning_rate": 2.0087084761679245e-06,
+      "logits/chosen": -0.8196107149124146,
+      "logits/rejected": -1.1247520446777344,
+      "logps/chosen": -0.8312146067619324,
+      "logps/rejected": -1.3900336027145386,
+      "loss": 0.7936,
+      "odds_ratio_loss": 0.44879013299942017,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.04156073182821274,
+      "rewards/margins": 0.027940943837165833,
+      "rewards/rejected": -0.06950168311595917,
+      "sft_loss": 0.8312146067619324,
+      "step": 2275
+    },
+    {
+      "epoch": 1.8239999999999998,
+      "grad_norm": 7.113083017693733,
+      "learning_rate": 1.997305197135089e-06,
+      "logits/chosen": -0.6806960701942444,
+      "logits/rejected": -0.8038619160652161,
+      "logps/chosen": -0.6782374382019043,
+      "logps/rejected": -1.2599354982376099,
+      "loss": 0.8263,
+      "odds_ratio_loss": 0.379936158657074,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.033911872655153275,
+      "rewards/margins": 0.02908490039408207,
+      "rewards/rejected": -0.0629967749118805,
+      "sft_loss": 0.6782374382019043,
+      "step": 2280
+    },
+    {
+      "epoch": 1.8279999999999998,
+      "grad_norm": 7.6235951885998015,
+      "learning_rate": 1.985912807298154e-06,
+      "logits/chosen": -1.0690265893936157,
+      "logits/rejected": -0.9779754877090454,
+      "logps/chosen": -0.7425954937934875,
+      "logps/rejected": -1.426913857460022,
+      "loss": 0.7348,
+      "odds_ratio_loss": 0.357009619474411,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03712977468967438,
+      "rewards/margins": 0.03421591967344284,
+      "rewards/rejected": -0.07134570181369781,
+      "sft_loss": 0.7425954937934875,
+      "step": 2285
+    },
+    {
+      "epoch": 1.8319999999999999,
+      "grad_norm": 5.615671437369734,
+      "learning_rate": 1.9745315534350157e-06,
+      "logits/chosen": -0.9129589200019836,
+      "logits/rejected": -0.6167286038398743,
+      "logps/chosen": -1.065497636795044,
+      "logps/rejected": -1.2868871688842773,
+      "loss": 0.8386,
+      "odds_ratio_loss": 0.6246089339256287,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.05327488109469414,
+      "rewards/margins": 0.011069480329751968,
+      "rewards/rejected": -0.0643443614244461,
+      "sft_loss": 1.065497636795044,
+      "step": 2290
+    },
+    {
+      "epoch": 1.8359999999999999,
+      "grad_norm": 6.084128369072109,
+      "learning_rate": 1.963161682082342e-06,
+      "logits/chosen": -0.9707215428352356,
+      "logits/rejected": -0.7441933751106262,
+      "logps/chosen": -0.9074664115905762,
+      "logps/rejected": -1.3849105834960938,
+      "loss": 0.8658,
+      "odds_ratio_loss": 0.4366206228733063,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.04537331685423851,
+      "rewards/margins": 0.023872217163443565,
+      "rewards/rejected": -0.06924553215503693,
+      "sft_loss": 0.9074664115905762,
+      "step": 2295
+    },
+    {
+      "epoch": 1.8399999999999999,
+      "grad_norm": 7.352748854504632,
+      "learning_rate": 1.9518034395302413e-06,
+      "logits/chosen": -0.7245827317237854,
+      "logits/rejected": -1.2959226369857788,
+      "logps/chosen": -0.7377294301986694,
+      "logps/rejected": -1.4373664855957031,
+      "loss": 0.9049,
+      "odds_ratio_loss": 0.33754152059555054,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03688646852970123,
+      "rewards/margins": 0.034981850534677505,
+      "rewards/rejected": -0.07186831533908844,
+      "sft_loss": 0.7377294301986694,
+      "step": 2300
+    },
+    {
+      "epoch": 1.8439999999999999,
+      "grad_norm": 4.828613164349784,
+      "learning_rate": 1.940457071816922e-06,
+      "logits/chosen": -0.7781765460968018,
+      "logits/rejected": -1.140925407409668,
+      "logps/chosen": -0.775480329990387,
+      "logps/rejected": -1.096097707748413,
+      "loss": 0.8761,
+      "odds_ratio_loss": 0.48161450028419495,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03877401351928711,
+      "rewards/margins": 0.016030868515372276,
+      "rewards/rejected": -0.054804880172014236,
+      "sft_loss": 0.775480329990387,
+      "step": 2305
+    },
+    {
+      "epoch": 1.8479999999999999,
+      "grad_norm": 6.702848454346183,
+      "learning_rate": 1.9291228247233607e-06,
+      "logits/chosen": -0.786065936088562,
+      "logits/rejected": -1.1700165271759033,
+      "logps/chosen": -0.8791561126708984,
+      "logps/rejected": -1.355006217956543,
+      "loss": 0.7591,
+      "odds_ratio_loss": 0.46694913506507874,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.04395780712366104,
+      "rewards/margins": 0.023792508989572525,
+      "rewards/rejected": -0.06775031983852386,
+      "sft_loss": 0.8791561126708984,
+      "step": 2310
+    },
+    {
+      "epoch": 1.8519999999999999,
+      "grad_norm": 7.711938205346599,
+      "learning_rate": 1.9178009437679855e-06,
+      "logits/chosen": -1.0333006381988525,
+      "logits/rejected": -0.5370423197746277,
+      "logps/chosen": -0.3634825050830841,
+      "logps/rejected": -1.1746513843536377,
+      "loss": 0.7101,
+      "odds_ratio_loss": 0.2676122784614563,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.018174124881625175,
+      "rewards/margins": 0.04055844619870186,
+      "rewards/rejected": -0.05873257666826248,
+      "sft_loss": 0.3634825050830841,
+      "step": 2315
+    },
+    {
+      "epoch": 1.8559999999999999,
+      "grad_norm": 8.761311945720486,
+      "learning_rate": 1.9064916742013515e-06,
+      "logits/chosen": -0.33012256026268005,
+      "logits/rejected": -1.075441598892212,
+      "logps/chosen": -0.9848030209541321,
+      "logps/rejected": -1.3406559228897095,
+      "loss": 0.8125,
+      "odds_ratio_loss": 0.5339032411575317,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.04924015700817108,
+      "rewards/margins": 0.01779264211654663,
+      "rewards/rejected": -0.06703279912471771,
+      "sft_loss": 0.9848030209541321,
+      "step": 2320
+    },
+    {
+      "epoch": 1.8599999999999999,
+      "grad_norm": 8.996579586998216,
+      "learning_rate": 1.895195261000831e-06,
+      "logits/chosen": -1.030765414237976,
+      "logits/rejected": -0.9685811996459961,
+      "logps/chosen": -0.8394074440002441,
+      "logps/rejected": -1.4255101680755615,
+      "loss": 0.7645,
+      "odds_ratio_loss": 0.4348466992378235,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.04197037220001221,
+      "rewards/margins": 0.029305141419172287,
+      "rewards/rejected": -0.07127551734447479,
+      "sft_loss": 0.8394074440002441,
+      "step": 2325
+    },
+    {
+      "epoch": 1.8639999999999999,
+      "grad_norm": 7.188662135547713,
+      "learning_rate": 1.883911948865306e-06,
+      "logits/chosen": -0.7353323698043823,
+      "logits/rejected": -0.9462113380432129,
+      "logps/chosen": -0.6895793676376343,
+      "logps/rejected": -1.388048529624939,
+      "loss": 0.7353,
+      "odds_ratio_loss": 0.40559130907058716,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03447896987199783,
+      "rewards/margins": 0.034923456609249115,
+      "rewards/rejected": -0.06940243393182755,
+      "sft_loss": 0.6895793676376343,
+      "step": 2330
+    },
+    {
+      "epoch": 1.8679999999999999,
+      "grad_norm": 7.322809013987031,
+      "learning_rate": 1.872641982209868e-06,
+      "logits/chosen": -0.9933965802192688,
+      "logits/rejected": -0.6932544112205505,
+      "logps/chosen": -0.6619799137115479,
+      "logps/rejected": -1.2481772899627686,
+      "loss": 0.7704,
+      "odds_ratio_loss": 0.38309139013290405,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.033098991960287094,
+      "rewards/margins": 0.029309872537851334,
+      "rewards/rejected": -0.06240885704755783,
+      "sft_loss": 0.6619799137115479,
+      "step": 2335
+    },
+    {
+      "epoch": 1.8719999999999999,
+      "grad_norm": 4.882755527025201,
+      "learning_rate": 1.8613856051605242e-06,
+      "logits/chosen": -0.7146080732345581,
+      "logits/rejected": -1.3212826251983643,
+      "logps/chosen": -0.970722496509552,
+      "logps/rejected": -1.4650259017944336,
+      "loss": 0.7754,
+      "odds_ratio_loss": 0.4080827236175537,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.04853612929582596,
+      "rewards/margins": 0.02471516653895378,
+      "rewards/rejected": -0.07325129956007004,
+      "sft_loss": 0.970722496509552,
+      "step": 2340
+    },
+    {
+      "epoch": 1.876,
+      "grad_norm": 6.023565399066244,
+      "learning_rate": 1.850143061548907e-06,
+      "logits/chosen": -0.8632639050483704,
+      "logits/rejected": -1.1179392337799072,
+      "logps/chosen": -0.7041983008384705,
+      "logps/rejected": -2.4500975608825684,
+      "loss": 0.815,
+      "odds_ratio_loss": 0.27313196659088135,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03520992025732994,
+      "rewards/margins": 0.08729497343301773,
+      "rewards/rejected": -0.12250488996505737,
+      "sft_loss": 0.7041983008384705,
+      "step": 2345
+    },
+    {
+      "epoch": 1.88,
+      "grad_norm": 7.758596635756381,
+      "learning_rate": 1.8389145949069953e-06,
+      "logits/chosen": -1.2442716360092163,
+      "logits/rejected": -0.7619195580482483,
+      "logps/chosen": -0.8322132229804993,
+      "logps/rejected": -1.4952902793884277,
+      "loss": 0.9443,
+      "odds_ratio_loss": 0.39093518257141113,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.041610658168792725,
+      "rewards/margins": 0.03315385431051254,
+      "rewards/rejected": -0.07476451992988586,
+      "sft_loss": 0.8322132229804993,
+      "step": 2350
+    },
+    {
+      "epoch": 1.884,
+      "grad_norm": 6.603546187128215,
+      "learning_rate": 1.827700448461836e-06,
+      "logits/chosen": -0.8992365598678589,
+      "logits/rejected": -0.41766390204429626,
+      "logps/chosen": -0.7582792043685913,
+      "logps/rejected": -1.665892243385315,
+      "loss": 0.8475,
+      "odds_ratio_loss": 0.422523558139801,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.0379139669239521,
+      "rewards/margins": 0.04538065940141678,
+      "rewards/rejected": -0.08329462260007858,
+      "sft_loss": 0.7582792043685913,
+      "step": 2355
+    },
+    {
+      "epoch": 1.888,
+      "grad_norm": 4.094000393336399,
+      "learning_rate": 1.816500865130279e-06,
+      "logits/chosen": -1.0214028358459473,
+      "logits/rejected": -0.9226012229919434,
+      "logps/chosen": -0.78780597448349,
+      "logps/rejected": -1.5264439582824707,
+      "loss": 0.7332,
+      "odds_ratio_loss": 0.39362823963165283,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.03939030319452286,
+      "rewards/margins": 0.036931902170181274,
+      "rewards/rejected": -0.07632219791412354,
+      "sft_loss": 0.78780597448349,
+      "step": 2360
+    },
+    {
+      "epoch": 1.892,
+      "grad_norm": 5.145690592897293,
+      "learning_rate": 1.8053160875137137e-06,
+      "logits/chosen": -0.9314631223678589,
+      "logits/rejected": -0.9437487721443176,
+      "logps/chosen": -1.0313526391983032,
+      "logps/rejected": -1.4607595205307007,
+      "loss": 0.8259,
+      "odds_ratio_loss": 0.5963125228881836,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.05156763270497322,
+      "rewards/margins": 0.021470338106155396,
+      "rewards/rejected": -0.07303796708583832,
+      "sft_loss": 1.0313526391983032,
+      "step": 2365
+    },
+    {
+      "epoch": 1.896,
+      "grad_norm": 7.417881702928459,
+      "learning_rate": 1.7941463578928088e-06,
+      "logits/chosen": -0.797860324382782,
+      "logits/rejected": -1.1919324398040771,
+      "logps/chosen": -0.879246711730957,
+      "logps/rejected": -1.3754128217697144,
+      "loss": 0.8715,
+      "odds_ratio_loss": 0.4276247024536133,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.04396234080195427,
+      "rewards/margins": 0.024808308109641075,
+      "rewards/rejected": -0.0687706395983696,
+      "sft_loss": 0.879246711730957,
+      "step": 2370
+    },
+    {
+      "epoch": 1.9,
+      "grad_norm": 7.132781683007218,
+      "learning_rate": 1.7829919182222752e-06,
+      "logits/chosen": -0.972023606300354,
+      "logits/rejected": -1.1046335697174072,
+      "logps/chosen": -0.8717896342277527,
+      "logps/rejected": -1.0692293643951416,
+      "loss": 0.8009,
+      "odds_ratio_loss": 0.679745078086853,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.043589480221271515,
+      "rewards/margins": 0.009871991351246834,
+      "rewards/rejected": -0.0534614734351635,
+      "sft_loss": 0.8717896342277527,
+      "step": 2375
+    },
+    {
+      "epoch": 1.904,
+      "grad_norm": 4.092969776254415,
+      "learning_rate": 1.7718530101256115e-06,
+      "logits/chosen": -1.1578160524368286,
+      "logits/rejected": -1.0545146465301514,
+      "logps/chosen": -0.47031641006469727,
+      "logps/rejected": -1.8584105968475342,
+      "loss": 0.7973,
+      "odds_ratio_loss": 0.1941300332546234,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.023515818640589714,
+      "rewards/margins": 0.0694047138094902,
+      "rewards/rejected": -0.09292052686214447,
+      "sft_loss": 0.47031641006469727,
+      "step": 2380
+    },
+    {
+      "epoch": 1.908,
+      "grad_norm": 6.116647440253481,
+      "learning_rate": 1.7607298748898844e-06,
+      "logits/chosen": -1.042887568473816,
+      "logits/rejected": -0.9683266878128052,
+      "logps/chosen": -0.7862476706504822,
+      "logps/rejected": -1.1969187259674072,
+      "loss": 0.8179,
+      "odds_ratio_loss": 0.5028802752494812,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.03931238502264023,
+      "rewards/margins": 0.020533554255962372,
+      "rewards/rejected": -0.0598459430038929,
+      "sft_loss": 0.7862476706504822,
+      "step": 2385
+    },
+    {
+      "epoch": 1.912,
+      "grad_norm": 5.228060511447272,
+      "learning_rate": 1.7496227534604859e-06,
+      "logits/chosen": -0.8977655172348022,
+      "logits/rejected": -1.0492327213287354,
+      "logps/chosen": -0.6170272827148438,
+      "logps/rejected": -1.3571946620941162,
+      "loss": 0.7784,
+      "odds_ratio_loss": 0.2826383411884308,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.030851364135742188,
+      "rewards/margins": 0.0370083749294281,
+      "rewards/rejected": -0.06785973161458969,
+      "sft_loss": 0.6170272827148438,
+      "step": 2390
+    },
+    {
+      "epoch": 1.916,
+      "grad_norm": 10.090456032598201,
+      "learning_rate": 1.7385318864359304e-06,
+      "logits/chosen": -0.9312974214553833,
+      "logits/rejected": -1.235394835472107,
+      "logps/chosen": -0.9995916485786438,
+      "logps/rejected": -1.703447937965393,
+      "loss": 0.8131,
+      "odds_ratio_loss": 0.43445831537246704,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.04997957870364189,
+      "rewards/margins": 0.0351928174495697,
+      "rewards/rejected": -0.08517240732908249,
+      "sft_loss": 0.9995916485786438,
+      "step": 2395
+    },
+    {
+      "epoch": 1.92,
+      "grad_norm": 5.661494882202969,
+      "learning_rate": 1.7274575140626318e-06,
+      "logits/chosen": -1.0951160192489624,
+      "logits/rejected": -0.7472076416015625,
+      "logps/chosen": -0.819926917552948,
+      "logps/rejected": -1.4353911876678467,
+      "loss": 0.8619,
+      "odds_ratio_loss": 0.42657119035720825,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.04099633917212486,
+      "rewards/margins": 0.03077322244644165,
+      "rewards/rejected": -0.07176957279443741,
+      "sft_loss": 0.819926917552948,
+      "step": 2400
+    },
+    {
+      "epoch": 1.924,
+      "grad_norm": 6.119264571326813,
+      "learning_rate": 1.7163998762297013e-06,
+      "logits/chosen": -1.046919584274292,
+      "logits/rejected": -1.0116114616394043,
+      "logps/chosen": -0.6755749583244324,
+      "logps/rejected": -1.557366967201233,
+      "loss": 0.8711,
+      "odds_ratio_loss": 0.2980736196041107,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03377874568104744,
+      "rewards/margins": 0.04408959671854973,
+      "rewards/rejected": -0.07786835730075836,
+      "sft_loss": 0.6755749583244324,
+      "step": 2405
+    },
+    {
+      "epoch": 1.928,
+      "grad_norm": 7.45326513403781,
+      "learning_rate": 1.7053592124637557e-06,
+      "logits/chosen": -0.9731330871582031,
+      "logits/rejected": -1.1434471607208252,
+      "logps/chosen": -0.6788956522941589,
+      "logps/rejected": -1.4451775550842285,
+      "loss": 0.8022,
+      "odds_ratio_loss": 0.31722068786621094,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03394477814435959,
+      "rewards/margins": 0.0383140966296196,
+      "rewards/rejected": -0.07225887477397919,
+      "sft_loss": 0.6788956522941589,
+      "step": 2410
+    },
+    {
+      "epoch": 1.932,
+      "grad_norm": 5.554908477058027,
+      "learning_rate": 1.6943357619237227e-06,
+      "logits/chosen": -0.9072321653366089,
+      "logits/rejected": -1.343854546546936,
+      "logps/chosen": -0.7743789553642273,
+      "logps/rejected": -1.7294988632202148,
+      "loss": 0.7934,
+      "odds_ratio_loss": 0.27553829550743103,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03871894255280495,
+      "rewards/margins": 0.047755997627973557,
+      "rewards/rejected": -0.0864749401807785,
+      "sft_loss": 0.7743789553642273,
+      "step": 2415
+    },
+    {
+      "epoch": 1.936,
+      "grad_norm": 9.389343814394115,
+      "learning_rate": 1.6833297633956647e-06,
+      "logits/chosen": -1.0711301565170288,
+      "logits/rejected": -0.810192883014679,
+      "logps/chosen": -0.763390839099884,
+      "logps/rejected": -1.7045867443084717,
+      "loss": 0.8294,
+      "odds_ratio_loss": 0.2711796462535858,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03816954419016838,
+      "rewards/margins": 0.0470597967505455,
+      "rewards/rejected": -0.08522933721542358,
+      "sft_loss": 0.763390839099884,
+      "step": 2420
+    },
+    {
+      "epoch": 1.94,
+      "grad_norm": 6.732029543770066,
+      "learning_rate": 1.6723414552876052e-06,
+      "logits/chosen": -0.570145845413208,
+      "logits/rejected": -1.3672704696655273,
+      "logps/chosen": -1.0161564350128174,
+      "logps/rejected": -1.465152621269226,
+      "loss": 0.852,
+      "odds_ratio_loss": 0.5028510689735413,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.05080782249569893,
+      "rewards/margins": 0.022449806332588196,
+      "rewards/rejected": -0.07325764000415802,
+      "sft_loss": 1.0161564350128174,
+      "step": 2425
+    },
+    {
+      "epoch": 1.944,
+      "grad_norm": 6.884753284401447,
+      "learning_rate": 1.661371075624363e-06,
+      "logits/chosen": -0.8588671684265137,
+      "logits/rejected": -1.0025584697723389,
+      "logps/chosen": -0.7952778935432434,
+      "logps/rejected": -1.3232905864715576,
+      "loss": 0.8421,
+      "odds_ratio_loss": 0.44380640983581543,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03976389393210411,
+      "rewards/margins": 0.02640063688158989,
+      "rewards/rejected": -0.066164530813694,
+      "sft_loss": 0.7952778935432434,
+      "step": 2430
+    },
+    {
+      "epoch": 1.948,
+      "grad_norm": 10.032821854639238,
+      "learning_rate": 1.6504188620423977e-06,
+      "logits/chosen": -0.8894034624099731,
+      "logits/rejected": -1.096508502960205,
+      "logps/chosen": -0.9423715472221375,
+      "logps/rejected": -1.5046286582946777,
+      "loss": 0.83,
+      "odds_ratio_loss": 0.4647153317928314,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.04711857810616493,
+      "rewards/margins": 0.028112854808568954,
+      "rewards/rejected": -0.07523143291473389,
+      "sft_loss": 0.9423715472221375,
+      "step": 2435
+    },
+    {
+      "epoch": 1.952,
+      "grad_norm": 8.207321665297997,
+      "learning_rate": 1.6394850517846621e-06,
+      "logits/chosen": -0.9905040860176086,
+      "logits/rejected": -0.9356651306152344,
+      "logps/chosen": -0.8063279986381531,
+      "logps/rejected": -1.4606654644012451,
+      "loss": 0.7861,
+      "odds_ratio_loss": 0.4356115460395813,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.040316395461559296,
+      "rewards/margins": 0.03271687030792236,
+      "rewards/rejected": -0.07303327322006226,
+      "sft_loss": 0.8063279986381531,
+      "step": 2440
+    },
+    {
+      "epoch": 1.956,
+      "grad_norm": 6.452799460000441,
+      "learning_rate": 1.6285698816954626e-06,
+      "logits/chosen": -0.7353092432022095,
+      "logits/rejected": -1.1551518440246582,
+      "logps/chosen": -1.0718172788619995,
+      "logps/rejected": -1.5583527088165283,
+      "loss": 0.8719,
+      "odds_ratio_loss": 0.46111541986465454,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.05359087139368057,
+      "rewards/margins": 0.024326767772436142,
+      "rewards/rejected": -0.07791763544082642,
+      "sft_loss": 1.0718172788619995,
+      "step": 2445
+    },
+    {
+      "epoch": 1.96,
+      "grad_norm": 5.332170706332277,
+      "learning_rate": 1.6176735882153284e-06,
+      "logits/chosen": -0.9285340309143066,
+      "logits/rejected": -1.087899923324585,
+      "logps/chosen": -0.8756929636001587,
+      "logps/rejected": -1.3418347835540771,
+      "loss": 0.781,
+      "odds_ratio_loss": 0.5048962831497192,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.043784648180007935,
+      "rewards/margins": 0.023307088762521744,
+      "rewards/rejected": -0.06709174066781998,
+      "sft_loss": 0.8756929636001587,
+      "step": 2450
+    },
+    {
+      "epoch": 1.964,
+      "grad_norm": 6.2107913026430746,
+      "learning_rate": 1.6067964073758901e-06,
+      "logits/chosen": -0.6262539625167847,
+      "logits/rejected": -0.9137803316116333,
+      "logps/chosen": -0.7144273519515991,
+      "logps/rejected": -1.7826179265975952,
+      "loss": 0.7463,
+      "odds_ratio_loss": 0.21934516727924347,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.035721369087696075,
+      "rewards/margins": 0.05340953543782234,
+      "rewards/rejected": -0.08913090080022812,
+      "sft_loss": 0.7144273519515991,
+      "step": 2455
+    },
+    {
+      "epoch": 1.968,
+      "grad_norm": 6.679465106655453,
+      "learning_rate": 1.5959385747947697e-06,
+      "logits/chosen": -0.6637876629829407,
+      "logits/rejected": -1.1082921028137207,
+      "logps/chosen": -0.7338643074035645,
+      "logps/rejected": -1.9011523723602295,
+      "loss": 0.8968,
+      "odds_ratio_loss": 0.26652342081069946,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03669321909546852,
+      "rewards/margins": 0.05836440995335579,
+      "rewards/rejected": -0.09505762159824371,
+      "sft_loss": 0.7338643074035645,
+      "step": 2460
+    },
+    {
+      "epoch": 1.972,
+      "grad_norm": 13.019725966030125,
+      "learning_rate": 1.5851003256704697e-06,
+      "logits/chosen": -0.9775069355964661,
+      "logits/rejected": -1.155165195465088,
+      "logps/chosen": -0.771982729434967,
+      "logps/rejected": -1.2554118633270264,
+      "loss": 0.8792,
+      "odds_ratio_loss": 0.5085214376449585,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.03859913349151611,
+      "rewards/margins": 0.024171466007828712,
+      "rewards/rejected": -0.06277060508728027,
+      "sft_loss": 0.771982729434967,
+      "step": 2465
+    },
+    {
+      "epoch": 1.976,
+      "grad_norm": 6.415175624114794,
+      "learning_rate": 1.5742818947772875e-06,
+      "logits/chosen": -0.5195826888084412,
+      "logits/rejected": -0.9917033910751343,
+      "logps/chosen": -0.6547525525093079,
+      "logps/rejected": -1.6238784790039062,
+      "loss": 0.7833,
+      "odds_ratio_loss": 0.3622708320617676,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03273762762546539,
+      "rewards/margins": 0.04845630005002022,
+      "rewards/rejected": -0.08119393140077591,
+      "sft_loss": 0.6547525525093079,
+      "step": 2470
+    },
+    {
+      "epoch": 1.98,
+      "grad_norm": 16.881628339846156,
+      "learning_rate": 1.56348351646022e-06,
+      "logits/chosen": -1.1250643730163574,
+      "logits/rejected": -1.0977294445037842,
+      "logps/chosen": -0.7314361333847046,
+      "logps/rejected": -1.5915559530258179,
+      "loss": 0.8277,
+      "odds_ratio_loss": 0.3246075510978699,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.03657180815935135,
+      "rewards/margins": 0.043005991727113724,
+      "rewards/rejected": -0.07957780361175537,
+      "sft_loss": 0.7314361333847046,
+      "step": 2475
+    },
+    {
+      "epoch": 1.984,
+      "grad_norm": 5.6365603948727,
+      "learning_rate": 1.552705424629898e-06,
+      "logits/chosen": -0.6629668474197388,
+      "logits/rejected": -1.1308306455612183,
+      "logps/chosen": -0.657507061958313,
+      "logps/rejected": -1.4453434944152832,
+      "loss": 0.8399,
+      "odds_ratio_loss": 0.3202642798423767,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03287535160779953,
+      "rewards/margins": 0.03939182311296463,
+      "rewards/rejected": -0.07226717472076416,
+      "sft_loss": 0.657507061958313,
+      "step": 2480
+    },
+    {
+      "epoch": 1.988,
+      "grad_norm": 11.934551215608066,
+      "learning_rate": 1.5419478527575068e-06,
+      "logits/chosen": -1.0436184406280518,
+      "logits/rejected": -0.9820472002029419,
+      "logps/chosen": -1.0538418292999268,
+      "logps/rejected": -1.5438861846923828,
+      "loss": 0.763,
+      "odds_ratio_loss": 0.44970980286598206,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.05269209295511246,
+      "rewards/margins": 0.024502214044332504,
+      "rewards/rejected": -0.07719431817531586,
+      "sft_loss": 1.0538418292999268,
+      "step": 2485
+    },
+    {
+      "epoch": 1.992,
+      "grad_norm": 4.730272496249411,
+      "learning_rate": 1.5312110338697427e-06,
+      "logits/chosen": -0.7060618996620178,
+      "logits/rejected": -1.1768749952316284,
+      "logps/chosen": -0.6467851400375366,
+      "logps/rejected": -1.8637425899505615,
+      "loss": 0.6756,
+      "odds_ratio_loss": 0.24719473719596863,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03233925253152847,
+      "rewards/margins": 0.06084787845611572,
+      "rewards/rejected": -0.09318713843822479,
+      "sft_loss": 0.6467851400375366,
+      "step": 2490
+    },
+    {
+      "epoch": 1.996,
+      "grad_norm": 5.17945710231851,
+      "learning_rate": 1.520495200543754e-06,
+      "logits/chosen": -0.6287657022476196,
+      "logits/rejected": -1.2824218273162842,
+      "logps/chosen": -0.8325425982475281,
+      "logps/rejected": -1.3392772674560547,
+      "loss": 0.7908,
+      "odds_ratio_loss": 0.5226107835769653,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.041627127677202225,
+      "rewards/margins": 0.02533673867583275,
+      "rewards/rejected": -0.06696386635303497,
+      "sft_loss": 0.8325425982475281,
+      "step": 2495
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 7.185045771765787,
+      "learning_rate": 1.509800584902108e-06,
+      "logits/chosen": -0.7127388119697571,
+      "logits/rejected": -0.9128797650337219,
+      "logps/chosen": -0.9189082980155945,
+      "logps/rejected": -1.3660955429077148,
+      "loss": 0.7301,
+      "odds_ratio_loss": 0.41850653290748596,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.045945413410663605,
+      "rewards/margins": 0.022359367460012436,
+      "rewards/rejected": -0.06830477714538574,
+      "sft_loss": 0.9189082980155945,
+      "step": 2500
+    },
+    {
+      "epoch": 2.004,
+      "grad_norm": 5.779855004519885,
+      "learning_rate": 1.4991274186077632e-06,
+      "logits/chosen": -0.6605402827262878,
+      "logits/rejected": -0.8482527732849121,
+      "logps/chosen": -0.4051028788089752,
+      "logps/rejected": -1.4240385293960571,
+      "loss": 0.4714,
+      "odds_ratio_loss": 0.22203238308429718,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02025514282286167,
+      "rewards/margins": 0.05094677209854126,
+      "rewards/rejected": -0.07120192050933838,
+      "sft_loss": 0.4051028788089752,
+      "step": 2505
+    },
+    {
+      "epoch": 2.008,
+      "grad_norm": 5.911255709626764,
+      "learning_rate": 1.4884759328590476e-06,
+      "logits/chosen": -0.8963392376899719,
+      "logits/rejected": -1.2612377405166626,
+      "logps/chosen": -0.5547269582748413,
+      "logps/rejected": -1.2793185710906982,
+      "loss": 0.5443,
+      "odds_ratio_loss": 0.2325483113527298,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.027736347168684006,
+      "rewards/margins": 0.03622957691550255,
+      "rewards/rejected": -0.06396592408418655,
+      "sft_loss": 0.5547269582748413,
+      "step": 2510
+    },
+    {
+      "epoch": 2.012,
+      "grad_norm": 4.263744768867711,
+      "learning_rate": 1.4778463583846553e-06,
+      "logits/chosen": -0.6352235078811646,
+      "logits/rejected": -1.1501035690307617,
+      "logps/chosen": -0.6144863963127136,
+      "logps/rejected": -1.6748031377792358,
+      "loss": 0.5734,
+      "odds_ratio_loss": 0.24970802664756775,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03072432242333889,
+      "rewards/margins": 0.05301583930850029,
+      "rewards/rejected": -0.08374015986919403,
+      "sft_loss": 0.6144863963127136,
+      "step": 2515
+    },
+    {
+      "epoch": 2.016,
+      "grad_norm": 4.168544749574423,
+      "learning_rate": 1.467238925438646e-06,
+      "logits/chosen": -0.6912413835525513,
+      "logits/rejected": -0.9819846153259277,
+      "logps/chosen": -0.5293835401535034,
+      "logps/rejected": -1.6525404453277588,
+      "loss": 0.5187,
+      "odds_ratio_loss": 0.2445772886276245,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.02646917663514614,
+      "rewards/margins": 0.05615784972906113,
+      "rewards/rejected": -0.08262702077627182,
+      "sft_loss": 0.5293835401535034,
+      "step": 2520
+    },
+    {
+      "epoch": 2.02,
+      "grad_norm": 6.164215653537583,
+      "learning_rate": 1.4566538637954556e-06,
+      "logits/chosen": -1.3008034229278564,
+      "logits/rejected": -0.7624127864837646,
+      "logps/chosen": -0.40782418847084045,
+      "logps/rejected": -1.5719823837280273,
+      "loss": 0.5314,
+      "odds_ratio_loss": 0.13072045147418976,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.020391209051012993,
+      "rewards/margins": 0.058207906782627106,
+      "rewards/rejected": -0.07859911769628525,
+      "sft_loss": 0.40782418847084045,
+      "step": 2525
+    },
+    {
+      "epoch": 2.024,
+      "grad_norm": 4.4630247981790525,
+      "learning_rate": 1.446091402744923e-06,
+      "logits/chosen": -0.9515730738639832,
+      "logits/rejected": -1.204268217086792,
+      "logps/chosen": -0.748781144618988,
+      "logps/rejected": -1.6079699993133545,
+      "loss": 0.5591,
+      "odds_ratio_loss": 0.27362948656082153,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03743905574083328,
+      "rewards/margins": 0.04295944422483444,
+      "rewards/rejected": -0.08039849996566772,
+      "sft_loss": 0.748781144618988,
+      "step": 2530
+    },
+    {
+      "epoch": 2.028,
+      "grad_norm": 5.834985541054327,
+      "learning_rate": 1.4355517710873184e-06,
+      "logits/chosen": -0.761924147605896,
+      "logits/rejected": -1.0106163024902344,
+      "logps/chosen": -0.6143468618392944,
+      "logps/rejected": -1.4826791286468506,
+      "loss": 0.5242,
+      "odds_ratio_loss": 0.3327800929546356,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03071734681725502,
+      "rewards/margins": 0.043416619300842285,
+      "rewards/rejected": -0.074133962392807,
+      "sft_loss": 0.6143468618392944,
+      "step": 2535
+    },
+    {
+      "epoch": 2.032,
+      "grad_norm": 4.93330991376578,
+      "learning_rate": 1.4250351971283937e-06,
+      "logits/chosen": -0.7397106885910034,
+      "logits/rejected": -0.974991500377655,
+      "logps/chosen": -0.4890953600406647,
+      "logps/rejected": -1.5005598068237305,
+      "loss": 0.5051,
+      "odds_ratio_loss": 0.21572282910346985,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.024454768747091293,
+      "rewards/margins": 0.05057322978973389,
+      "rewards/rejected": -0.07502799481153488,
+      "sft_loss": 0.4890953600406647,
+      "step": 2540
+    },
+    {
+      "epoch": 2.036,
+      "grad_norm": 5.652910564521034,
+      "learning_rate": 1.41454190867443e-06,
+      "logits/chosen": -0.6541872024536133,
+      "logits/rejected": -1.2734535932540894,
+      "logps/chosen": -0.5036207437515259,
+      "logps/rejected": -1.882581353187561,
+      "loss": 0.4315,
+      "odds_ratio_loss": 0.25754693150520325,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.025181034579873085,
+      "rewards/margins": 0.06894804537296295,
+      "rewards/rejected": -0.09412907809019089,
+      "sft_loss": 0.5036207437515259,
+      "step": 2545
+    },
+    {
+      "epoch": 2.04,
+      "grad_norm": 5.023733060512668,
+      "learning_rate": 1.4040721330273063e-06,
+      "logits/chosen": -1.1533195972442627,
+      "logits/rejected": -0.9162420034408569,
+      "logps/chosen": -0.3805043697357178,
+      "logps/rejected": -1.3916971683502197,
+      "loss": 0.5004,
+      "odds_ratio_loss": 0.23528532683849335,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.01902521774172783,
+      "rewards/margins": 0.0505596399307251,
+      "rewards/rejected": -0.06958486139774323,
+      "sft_loss": 0.3805043697357178,
+      "step": 2550
+    },
+    {
+      "epoch": 2.044,
+      "grad_norm": 3.932026075807386,
+      "learning_rate": 1.3936260969795778e-06,
+      "logits/chosen": -0.8410050272941589,
+      "logits/rejected": -0.9105769991874695,
+      "logps/chosen": -0.6416537165641785,
+      "logps/rejected": -1.6031726598739624,
+      "loss": 0.5072,
+      "odds_ratio_loss": 0.26725801825523376,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.032082684338092804,
+      "rewards/margins": 0.04807594418525696,
+      "rewards/rejected": -0.08015862852334976,
+      "sft_loss": 0.6416537165641785,
+      "step": 2555
+    },
+    {
+      "epoch": 2.048,
+      "grad_norm": 6.870678133966881,
+      "learning_rate": 1.3832040268095589e-06,
+      "logits/chosen": -0.6729310750961304,
+      "logits/rejected": -0.8668330907821655,
+      "logps/chosen": -0.40369588136672974,
+      "logps/rejected": -1.3457889556884766,
+      "loss": 0.5096,
+      "odds_ratio_loss": 0.2014884054660797,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.020184790715575218,
+      "rewards/margins": 0.04710465669631958,
+      "rewards/rejected": -0.06728944927453995,
+      "sft_loss": 0.40369588136672974,
+      "step": 2560
+    },
+    {
+      "epoch": 2.052,
+      "grad_norm": 7.587293716398404,
+      "learning_rate": 1.3728061482764238e-06,
+      "logits/chosen": -0.843582808971405,
+      "logits/rejected": -0.9166304469108582,
+      "logps/chosen": -0.604731559753418,
+      "logps/rejected": -1.5095970630645752,
+      "loss": 0.5172,
+      "odds_ratio_loss": 0.2633668780326843,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03023657761514187,
+      "rewards/margins": 0.0452432855963707,
+      "rewards/rejected": -0.07547986507415771,
+      "sft_loss": 0.604731559753418,
+      "step": 2565
+    },
+    {
+      "epoch": 2.056,
+      "grad_norm": 4.0093961868186625,
+      "learning_rate": 1.362432686615316e-06,
+      "logits/chosen": -0.9857913851737976,
+      "logits/rejected": -0.7414765357971191,
+      "logps/chosen": -0.46555963158607483,
+      "logps/rejected": -1.3948227167129517,
+      "loss": 0.4983,
+      "odds_ratio_loss": 0.2733033299446106,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.023277979344129562,
+      "rewards/margins": 0.046463146805763245,
+      "rewards/rejected": -0.0697411373257637,
+      "sft_loss": 0.46555963158607483,
+      "step": 2570
+    },
+    {
+      "epoch": 2.06,
+      "grad_norm": 6.939466081035446,
+      "learning_rate": 1.3520838665324704e-06,
+      "logits/chosen": -1.1071815490722656,
+      "logits/rejected": -0.817790687084198,
+      "logps/chosen": -0.4723474085330963,
+      "logps/rejected": -1.3803852796554565,
+      "loss": 0.5202,
+      "odds_ratio_loss": 0.256916880607605,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.023617371916770935,
+      "rewards/margins": 0.04540189355611801,
+      "rewards/rejected": -0.06901925802230835,
+      "sft_loss": 0.4723474085330963,
+      "step": 2575
+    },
+    {
+      "epoch": 2.064,
+      "grad_norm": 4.140170449016894,
+      "learning_rate": 1.3417599122003464e-06,
+      "logits/chosen": -0.450253427028656,
+      "logits/rejected": -0.9482170343399048,
+      "logps/chosen": -0.6843463778495789,
+      "logps/rejected": -1.6336677074432373,
+      "loss": 0.5651,
+      "odds_ratio_loss": 0.27679651975631714,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03421732038259506,
+      "rewards/margins": 0.047466062009334564,
+      "rewards/rejected": -0.08168338239192963,
+      "sft_loss": 0.6843463778495789,
+      "step": 2580
+    },
+    {
+      "epoch": 2.068,
+      "grad_norm": 5.1100760056211545,
+      "learning_rate": 1.3314610472527645e-06,
+      "logits/chosen": -0.8113574981689453,
+      "logits/rejected": -1.0102758407592773,
+      "logps/chosen": -0.43412670493125916,
+      "logps/rejected": -1.9312794208526611,
+      "loss": 0.4464,
+      "odds_ratio_loss": 0.12802623212337494,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02170633338391781,
+      "rewards/margins": 0.07485763728618622,
+      "rewards/rejected": -0.09656397253274918,
+      "sft_loss": 0.43412670493125916,
+      "step": 2585
+    },
+    {
+      "epoch": 2.072,
+      "grad_norm": 7.56195378391692,
+      "learning_rate": 1.3211874947800747e-06,
+      "logits/chosen": -0.7834498286247253,
+      "logits/rejected": -0.8870722055435181,
+      "logps/chosen": -0.2993749678134918,
+      "logps/rejected": -1.3909119367599487,
+      "loss": 0.4961,
+      "odds_ratio_loss": 0.1264984905719757,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.014968748204410076,
+      "rewards/margins": 0.054576851427555084,
+      "rewards/rejected": -0.06954559683799744,
+      "sft_loss": 0.2993749678134918,
+      "step": 2590
+    },
+    {
+      "epoch": 2.076,
+      "grad_norm": 5.796090053121948,
+      "learning_rate": 1.3109394773243117e-06,
+      "logits/chosen": -0.9734644889831543,
+      "logits/rejected": -1.1090242862701416,
+      "logps/chosen": -0.729141891002655,
+      "logps/rejected": -1.5011317729949951,
+      "loss": 0.5283,
+      "odds_ratio_loss": 0.26416879892349243,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03645709902048111,
+      "rewards/margins": 0.038599494844675064,
+      "rewards/rejected": -0.07505659013986588,
+      "sft_loss": 0.729141891002655,
+      "step": 2595
+    },
+    {
+      "epoch": 2.08,
+      "grad_norm": 4.905492874489777,
+      "learning_rate": 1.3007172168743854e-06,
+      "logits/chosen": -0.6443847417831421,
+      "logits/rejected": -0.6509965658187866,
+      "logps/chosen": -0.4077952802181244,
+      "logps/rejected": -1.5444676876068115,
+      "loss": 0.4459,
+      "odds_ratio_loss": 0.1849530041217804,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02038976177573204,
+      "rewards/margins": 0.056833624839782715,
+      "rewards/rejected": -0.07722338289022446,
+      "sft_loss": 0.4077952802181244,
+      "step": 2600
+    },
+    {
+      "epoch": 2.084,
+      "grad_norm": 8.38385935812006,
+      "learning_rate": 1.2905209348612596e-06,
+      "logits/chosen": -1.0582338571548462,
+      "logits/rejected": -1.0338222980499268,
+      "logps/chosen": -0.3466675281524658,
+      "logps/rejected": -1.3382198810577393,
+      "loss": 0.4997,
+      "odds_ratio_loss": 0.17245514690876007,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.01733337715268135,
+      "rewards/margins": 0.049577612429857254,
+      "rewards/rejected": -0.0669109970331192,
+      "sft_loss": 0.3466675281524658,
+      "step": 2605
+    },
+    {
+      "epoch": 2.088,
+      "grad_norm": 4.39436373774201,
+      "learning_rate": 1.280350852153168e-06,
+      "logits/chosen": -0.9197877049446106,
+      "logits/rejected": -1.4192711114883423,
+      "logps/chosen": -0.5219482183456421,
+      "logps/rejected": -2.023850679397583,
+      "loss": 0.4535,
+      "odds_ratio_loss": 0.1568504273891449,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.026097416877746582,
+      "rewards/margins": 0.07509511709213257,
+      "rewards/rejected": -0.10119253396987915,
+      "sft_loss": 0.5219482183456421,
+      "step": 2610
+    },
+    {
+      "epoch": 2.092,
+      "grad_norm": 31.805321381556926,
+      "learning_rate": 1.2702071890508235e-06,
+      "logits/chosen": -0.8824436068534851,
+      "logits/rejected": -1.0908191204071045,
+      "logps/chosen": -0.49740344285964966,
+      "logps/rejected": -1.838228464126587,
+      "loss": 0.5843,
+      "odds_ratio_loss": 0.20825867354869843,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.024870174005627632,
+      "rewards/margins": 0.06704124808311462,
+      "rewards/rejected": -0.0919114202260971,
+      "sft_loss": 0.49740344285964966,
+      "step": 2615
+    },
+    {
+      "epoch": 2.096,
+      "grad_norm": 5.572501524698811,
+      "learning_rate": 1.260090165282645e-06,
+      "logits/chosen": -1.0808252096176147,
+      "logits/rejected": -0.8268556594848633,
+      "logps/chosen": -0.5673039555549622,
+      "logps/rejected": -1.536179780960083,
+      "loss": 0.4523,
+      "odds_ratio_loss": 0.24021944403648376,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.028365198522806168,
+      "rewards/margins": 0.048443786799907684,
+      "rewards/rejected": -0.07680898904800415,
+      "sft_loss": 0.5673039555549622,
+      "step": 2620
+    },
+    {
+      "epoch": 2.1,
+      "grad_norm": 5.981975287930625,
+      "learning_rate": 1.2500000000000007e-06,
+      "logits/chosen": -1.054109811782837,
+      "logits/rejected": -1.0586509704589844,
+      "logps/chosen": -0.7665703296661377,
+      "logps/rejected": -1.0670462846755981,
+      "loss": 0.5098,
+      "odds_ratio_loss": 0.5087045431137085,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.038328517228364944,
+      "rewards/margins": 0.015023794956505299,
+      "rewards/rejected": -0.05335230752825737,
+      "sft_loss": 0.7665703296661377,
+      "step": 2625
+    },
+    {
+      "epoch": 2.104,
+      "grad_norm": 7.544463465155528,
+      "learning_rate": 1.2399369117724582e-06,
+      "logits/chosen": -1.0681202411651611,
+      "logits/rejected": -1.0832048654556274,
+      "logps/chosen": -0.5446698069572449,
+      "logps/rejected": -1.5458736419677734,
+      "loss": 0.4264,
+      "odds_ratio_loss": 0.2728939950466156,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.027233490720391273,
+      "rewards/margins": 0.05006019398570061,
+      "rewards/rejected": -0.07729368656873703,
+      "sft_loss": 0.5446698069572449,
+      "step": 2630
+    },
+    {
+      "epoch": 2.108,
+      "grad_norm": 4.9815831235811325,
+      "learning_rate": 1.2299011185830557e-06,
+      "logits/chosen": -1.1409013271331787,
+      "logits/rejected": -0.8681640625,
+      "logps/chosen": -0.3343183994293213,
+      "logps/rejected": -1.3352359533309937,
+      "loss": 0.5583,
+      "odds_ratio_loss": 0.17289471626281738,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.016715919598937035,
+      "rewards/margins": 0.05004587024450302,
+      "rewards/rejected": -0.0667618066072464,
+      "sft_loss": 0.3343183994293213,
+      "step": 2635
+    },
+    {
+      "epoch": 2.112,
+      "grad_norm": 6.73634607533909,
+      "learning_rate": 1.2198928378235717e-06,
+      "logits/chosen": -1.1178267002105713,
+      "logits/rejected": -0.7387592196464539,
+      "logps/chosen": -0.44447723031044006,
+      "logps/rejected": -1.3598803281784058,
+      "loss": 0.5268,
+      "odds_ratio_loss": 0.21014420688152313,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.022223861888051033,
+      "rewards/margins": 0.04577016085386276,
+      "rewards/rejected": -0.06799402087926865,
+      "sft_loss": 0.44447723031044006,
+      "step": 2640
+    },
+    {
+      "epoch": 2.116,
+      "grad_norm": 4.379209925386878,
+      "learning_rate": 1.2099122862898214e-06,
+      "logits/chosen": -0.9532085657119751,
+      "logits/rejected": -1.0121229887008667,
+      "logps/chosen": -0.4848550856113434,
+      "logps/rejected": -1.2674915790557861,
+      "loss": 0.4112,
+      "odds_ratio_loss": 0.23585955798625946,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02424275316298008,
+      "rewards/margins": 0.039131827652454376,
+      "rewards/rejected": -0.0633745864033699,
+      "sft_loss": 0.4848550856113434,
+      "step": 2645
+    },
+    {
+      "epoch": 2.12,
+      "grad_norm": 5.5875611806210825,
+      "learning_rate": 1.1999596801769617e-06,
+      "logits/chosen": -0.9600692987442017,
+      "logits/rejected": -0.9517749547958374,
+      "logps/chosen": -0.24040523171424866,
+      "logps/rejected": -1.3553459644317627,
+      "loss": 0.5047,
+      "odds_ratio_loss": 0.10732688009738922,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.012020261958241463,
+      "rewards/margins": 0.05574704334139824,
+      "rewards/rejected": -0.06776730716228485,
+      "sft_loss": 0.24040523171424866,
+      "step": 2650
+    },
+    {
+      "epoch": 2.124,
+      "grad_norm": 5.665887523645101,
+      "learning_rate": 1.1900352350748026e-06,
+      "logits/chosen": -1.3023062944412231,
+      "logits/rejected": -0.6389235258102417,
+      "logps/chosen": -0.3508966863155365,
+      "logps/rejected": -1.516701102256775,
+      "loss": 0.4761,
+      "odds_ratio_loss": 0.20118489861488342,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.017544833943247795,
+      "rewards/margins": 0.05829022452235222,
+      "rewards/rejected": -0.07583504915237427,
+      "sft_loss": 0.3508966863155365,
+      "step": 2655
+    },
+    {
+      "epoch": 2.128,
+      "grad_norm": 4.849761975833439,
+      "learning_rate": 1.1801391659631423e-06,
+      "logits/chosen": -0.7300786375999451,
+      "logits/rejected": -1.2495789527893066,
+      "logps/chosen": -0.7808480262756348,
+      "logps/rejected": -1.481135606765747,
+      "loss": 0.5719,
+      "odds_ratio_loss": 0.3138294816017151,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.0390424020588398,
+      "rewards/margins": 0.035014379769563675,
+      "rewards/rejected": -0.07405678182840347,
+      "sft_loss": 0.7808480262756348,
+      "step": 2660
+    },
+    {
+      "epoch": 2.132,
+      "grad_norm": 7.470137375834024,
+      "learning_rate": 1.170271687207106e-06,
+      "logits/chosen": -0.8559409379959106,
+      "logits/rejected": -0.9146011471748352,
+      "logps/chosen": -0.42539578676223755,
+      "logps/rejected": -1.569939374923706,
+      "loss": 0.505,
+      "odds_ratio_loss": 0.17492997646331787,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.0212697871029377,
+      "rewards/margins": 0.05722718685865402,
+      "rewards/rejected": -0.07849697768688202,
+      "sft_loss": 0.42539578676223755,
+      "step": 2665
+    },
+    {
+      "epoch": 2.136,
+      "grad_norm": 4.739681467147201,
+      "learning_rate": 1.160433012552508e-06,
+      "logits/chosen": -0.571463942527771,
+      "logits/rejected": -1.121757984161377,
+      "logps/chosen": -0.5185319185256958,
+      "logps/rejected": -1.5073848962783813,
+      "loss": 0.5349,
+      "odds_ratio_loss": 0.20745344460010529,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.025926601141691208,
+      "rewards/margins": 0.04944263771176338,
+      "rewards/rejected": -0.07536924630403519,
+      "sft_loss": 0.5185319185256958,
+      "step": 2670
+    },
+    {
+      "epoch": 2.14,
+      "grad_norm": 4.496694368173494,
+      "learning_rate": 1.1506233551212186e-06,
+      "logits/chosen": -0.7737411856651306,
+      "logits/rejected": -1.0046793222427368,
+      "logps/chosen": -0.5965171456336975,
+      "logps/rejected": -1.5169528722763062,
+      "loss": 0.5752,
+      "odds_ratio_loss": 0.23371455073356628,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.029825860634446144,
+      "rewards/margins": 0.04602178931236267,
+      "rewards/rejected": -0.07584764063358307,
+      "sft_loss": 0.5965171456336975,
+      "step": 2675
+    },
+    {
+      "epoch": 2.144,
+      "grad_norm": 5.818039980551584,
+      "learning_rate": 1.1408429274065418e-06,
+      "logits/chosen": -1.1345255374908447,
+      "logits/rejected": -1.221318244934082,
+      "logps/chosen": -0.3457716703414917,
+      "logps/rejected": -1.645448088645935,
+      "loss": 0.3843,
+      "odds_ratio_loss": 0.146096870303154,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.017288584262132645,
+      "rewards/margins": 0.06498382240533829,
+      "rewards/rejected": -0.08227241039276123,
+      "sft_loss": 0.3457716703414917,
+      "step": 2680
+    },
+    {
+      "epoch": 2.148,
+      "grad_norm": 7.617489619432182,
+      "learning_rate": 1.1310919412686248e-06,
+      "logits/chosen": -0.6363608837127686,
+      "logits/rejected": -0.9788458943367004,
+      "logps/chosen": -0.31420546770095825,
+      "logps/rejected": -2.0080409049987793,
+      "loss": 0.4396,
+      "odds_ratio_loss": 0.08285339921712875,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.015710271894931793,
+      "rewards/margins": 0.08469177782535553,
+      "rewards/rejected": -0.10040205717086792,
+      "sft_loss": 0.31420546770095825,
+      "step": 2685
+    },
+    {
+      "epoch": 2.152,
+      "grad_norm": 5.306338004940538,
+      "learning_rate": 1.1213706079298566e-06,
+      "logits/chosen": -0.6837140917778015,
+      "logits/rejected": -1.3568503856658936,
+      "logps/chosen": -0.4734787046909332,
+      "logps/rejected": -1.4761061668395996,
+      "loss": 0.4861,
+      "odds_ratio_loss": 0.20139172673225403,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.02367393672466278,
+      "rewards/margins": 0.050131380558013916,
+      "rewards/rejected": -0.0738053172826767,
+      "sft_loss": 0.4734787046909332,
+      "step": 2690
+    },
+    {
+      "epoch": 2.156,
+      "grad_norm": 5.238611883596988,
+      "learning_rate": 1.1116791379703032e-06,
+      "logits/chosen": -0.6437733769416809,
+      "logits/rejected": -0.8825413584709167,
+      "logps/chosen": -0.3527492880821228,
+      "logps/rejected": -2.517010450363159,
+      "loss": 0.4288,
+      "odds_ratio_loss": 0.10402262210845947,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.0176374651491642,
+      "rewards/margins": 0.10821305215358734,
+      "rewards/rejected": -0.12585052847862244,
+      "sft_loss": 0.3527492880821228,
+      "step": 2695
+    },
+    {
+      "epoch": 2.16,
+      "grad_norm": 5.037423172133621,
+      "learning_rate": 1.1020177413231334e-06,
+      "logits/chosen": -0.7074111700057983,
+      "logits/rejected": -1.1180449724197388,
+      "logps/chosen": -0.706173300743103,
+      "logps/rejected": -1.7855567932128906,
+      "loss": 0.5407,
+      "odds_ratio_loss": 0.2965608835220337,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03530866652727127,
+      "rewards/margins": 0.05396916717290878,
+      "rewards/rejected": -0.08927784115076065,
+      "sft_loss": 0.706173300743103,
+      "step": 2700
+    },
+    {
+      "epoch": 2.164,
+      "grad_norm": 11.625726843094434,
+      "learning_rate": 1.0923866272700845e-06,
+      "logits/chosen": -0.6679056286811829,
+      "logits/rejected": -0.8790310025215149,
+      "logps/chosen": -0.37520498037338257,
+      "logps/rejected": -1.725489854812622,
+      "loss": 0.4112,
+      "odds_ratio_loss": 0.1370328813791275,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.018760250881314278,
+      "rewards/margins": 0.06751424074172974,
+      "rewards/rejected": -0.08627448976039886,
+      "sft_loss": 0.37520498037338257,
+      "step": 2705
+    },
+    {
+      "epoch": 2.168,
+      "grad_norm": 5.751387431303456,
+      "learning_rate": 1.0827860044369226e-06,
+      "logits/chosen": -1.1543610095977783,
+      "logits/rejected": -1.1389491558074951,
+      "logps/chosen": -0.8079617619514465,
+      "logps/rejected": -1.881744384765625,
+      "loss": 0.653,
+      "odds_ratio_loss": 0.21621274948120117,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.04039808362722397,
+      "rewards/margins": 0.0536891333758831,
+      "rewards/rejected": -0.09408722817897797,
+      "sft_loss": 0.8079617619514465,
+      "step": 2710
+    },
+    {
+      "epoch": 2.172,
+      "grad_norm": 13.56821195989023,
+      "learning_rate": 1.073216080788921e-06,
+      "logits/chosen": -0.7350510358810425,
+      "logits/rejected": -0.9062551259994507,
+      "logps/chosen": -0.41547298431396484,
+      "logps/rejected": -1.4787991046905518,
+      "loss": 0.4839,
+      "odds_ratio_loss": 0.17242933809757233,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02077365107834339,
+      "rewards/margins": 0.053166307508945465,
+      "rewards/rejected": -0.07393995672464371,
+      "sft_loss": 0.41547298431396484,
+      "step": 2715
+    },
+    {
+      "epoch": 2.176,
+      "grad_norm": 4.723019200073755,
+      "learning_rate": 1.06367706362636e-06,
+      "logits/chosen": -1.2677191495895386,
+      "logits/rejected": -1.3101189136505127,
+      "logps/chosen": -0.5442667603492737,
+      "logps/rejected": -1.3093074560165405,
+      "loss": 0.4889,
+      "odds_ratio_loss": 0.2581162452697754,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.027213340625166893,
+      "rewards/margins": 0.03825204446911812,
+      "rewards/rejected": -0.06546537578105927,
+      "sft_loss": 0.5442667603492737,
+      "step": 2720
+    },
+    {
+      "epoch": 2.18,
+      "grad_norm": 6.320991135595779,
+      "learning_rate": 1.0541691595800338e-06,
+      "logits/chosen": -0.7713707089424133,
+      "logits/rejected": -0.8052916526794434,
+      "logps/chosen": -0.8031466603279114,
+      "logps/rejected": -1.449881911277771,
+      "loss": 0.4879,
+      "odds_ratio_loss": 0.34610167145729065,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.04015733301639557,
+      "rewards/margins": 0.0323367640376091,
+      "rewards/rejected": -0.07249408960342407,
+      "sft_loss": 0.8031466603279114,
+      "step": 2725
+    },
+    {
+      "epoch": 2.184,
+      "grad_norm": 7.667541694530943,
+      "learning_rate": 1.0446925746067768e-06,
+      "logits/chosen": -0.735578179359436,
+      "logits/rejected": -1.290317416191101,
+      "logps/chosen": -0.4031465947628021,
+      "logps/rejected": -2.864250898361206,
+      "loss": 0.4889,
+      "odds_ratio_loss": 0.11299661546945572,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.020157333463430405,
+      "rewards/margins": 0.12305520474910736,
+      "rewards/rejected": -0.14321252703666687,
+      "sft_loss": 0.4031465947628021,
+      "step": 2730
+    },
+    {
+      "epoch": 2.188,
+      "grad_norm": 8.598577539363314,
+      "learning_rate": 1.0352475139849993e-06,
+      "logits/chosen": -1.1117980480194092,
+      "logits/rejected": -1.2077891826629639,
+      "logps/chosen": -0.6618624329566956,
+      "logps/rejected": -1.303166389465332,
+      "loss": 0.5357,
+      "odds_ratio_loss": 0.4117598533630371,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.03309311717748642,
+      "rewards/margins": 0.032065197825431824,
+      "rewards/rejected": -0.06515831500291824,
+      "sft_loss": 0.6618624329566956,
+      "step": 2735
+    },
+    {
+      "epoch": 2.192,
+      "grad_norm": 5.479651626100435,
+      "learning_rate": 1.0258341823102418e-06,
+      "logits/chosen": -0.873387336730957,
+      "logits/rejected": -1.0599385499954224,
+      "logps/chosen": -0.8911870121955872,
+      "logps/rejected": -1.9314569234848022,
+      "loss": 0.5608,
+      "odds_ratio_loss": 0.3484743535518646,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.04455935209989548,
+      "rewards/margins": 0.052013497799634933,
+      "rewards/rejected": -0.09657285362482071,
+      "sft_loss": 0.8911870121955872,
+      "step": 2740
+    },
+    {
+      "epoch": 2.196,
+      "grad_norm": 6.711233602443305,
+      "learning_rate": 1.0164527834907468e-06,
+      "logits/chosen": -1.1073827743530273,
+      "logits/rejected": -0.8712652325630188,
+      "logps/chosen": -0.3501274287700653,
+      "logps/rejected": -1.0953724384307861,
+      "loss": 0.463,
+      "odds_ratio_loss": 0.2384282350540161,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.017506374046206474,
+      "rewards/margins": 0.03726224601268768,
+      "rewards/rejected": -0.05476861447095871,
+      "sft_loss": 0.3501274287700653,
+      "step": 2745
+    },
+    {
+      "epoch": 2.2,
+      "grad_norm": 6.323686554397781,
+      "learning_rate": 1.0071035207430352e-06,
+      "logits/chosen": -0.5682858228683472,
+      "logits/rejected": -1.2733685970306396,
+      "logps/chosen": -0.5779908895492554,
+      "logps/rejected": -1.2694551944732666,
+      "loss": 0.4752,
+      "odds_ratio_loss": 0.37383320927619934,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.0288995448499918,
+      "rewards/margins": 0.03457321599125862,
+      "rewards/rejected": -0.06347276270389557,
+      "sft_loss": 0.5779908895492554,
+      "step": 2750
+    },
+    {
+      "epoch": 2.204,
+      "grad_norm": 5.639287689051844,
+      "learning_rate": 9.977865965875091e-07,
+      "logits/chosen": -0.7087677717208862,
+      "logits/rejected": -0.8449762463569641,
+      "logps/chosen": -0.5380462408065796,
+      "logps/rejected": -1.518744707107544,
+      "loss": 0.508,
+      "odds_ratio_loss": 0.23005056381225586,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02690231241285801,
+      "rewards/margins": 0.049034927040338516,
+      "rewards/rejected": -0.07593724131584167,
+      "sft_loss": 0.5380462408065796,
+      "step": 2755
+    },
+    {
+      "epoch": 2.208,
+      "grad_norm": 6.305391688848771,
+      "learning_rate": 9.88502212844063e-07,
+      "logits/chosen": -0.7394359111785889,
+      "logits/rejected": -1.2294960021972656,
+      "logps/chosen": -0.6039851307868958,
+      "logps/rejected": -1.388856291770935,
+      "loss": 0.5097,
+      "odds_ratio_loss": 0.2779824733734131,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.030199263244867325,
+      "rewards/margins": 0.039243556559085846,
+      "rewards/rejected": -0.06944280862808228,
+      "sft_loss": 0.6039851307868958,
+      "step": 2760
+    },
+    {
+      "epoch": 2.212,
+      "grad_norm": 5.741952540098281,
+      "learning_rate": 9.792505706277136e-07,
+      "logits/chosen": -0.8889573216438293,
+      "logits/rejected": -0.9258907437324524,
+      "logps/chosen": -0.5408663749694824,
+      "logps/rejected": -1.488094449043274,
+      "loss": 0.5131,
+      "odds_ratio_loss": 0.21949462592601776,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02704331837594509,
+      "rewards/margins": 0.047361403703689575,
+      "rewards/rejected": -0.07440472394227982,
+      "sft_loss": 0.5408663749694824,
+      "step": 2765
+    },
+    {
+      "epoch": 2.216,
+      "grad_norm": 11.523508801713763,
+      "learning_rate": 9.700318703442437e-07,
+      "logits/chosen": -1.087032437324524,
+      "logits/rejected": -0.9947022199630737,
+      "logps/chosen": -0.47177377343177795,
+      "logps/rejected": -1.4707008600234985,
+      "loss": 0.4983,
+      "odds_ratio_loss": 0.2322104275226593,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.02358868531882763,
+      "rewards/margins": 0.04994635283946991,
+      "rewards/rejected": -0.07353504002094269,
+      "sft_loss": 0.47177377343177795,
+      "step": 2770
+    },
+    {
+      "epoch": 2.22,
+      "grad_norm": 4.462198076439179,
+      "learning_rate": 9.608463116858544e-07,
+      "logits/chosen": -0.736533522605896,
+      "logits/rejected": -0.829948902130127,
+      "logps/chosen": -0.6742941737174988,
+      "logps/rejected": -1.4203133583068848,
+      "loss": 0.5412,
+      "odds_ratio_loss": 0.3509899079799652,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03371470421552658,
+      "rewards/margins": 0.037300966680049896,
+      "rewards/rejected": -0.07101567089557648,
+      "sft_loss": 0.6742941737174988,
+      "step": 2775
+    },
+    {
+      "epoch": 2.224,
+      "grad_norm": 4.228429825727303,
+      "learning_rate": 9.516940936268504e-07,
+      "logits/chosen": -1.1075184345245361,
+      "logits/rejected": -1.0531575679779053,
+      "logps/chosen": -0.21282517910003662,
+      "logps/rejected": -1.0983449220657349,
+      "loss": 0.4243,
+      "odds_ratio_loss": 0.24119925498962402,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.010641258209943771,
+      "rewards/margins": 0.04427598416805267,
+      "rewards/rejected": -0.05491724610328674,
+      "sft_loss": 0.21282517910003662,
+      "step": 2780
+    },
+    {
+      "epoch": 2.228,
+      "grad_norm": 5.45480141160989,
+      "learning_rate": 9.4257541441932e-07,
+      "logits/chosen": -0.9974383115768433,
+      "logits/rejected": -1.3215069770812988,
+      "logps/chosen": -0.42205625772476196,
+      "logps/rejected": -1.7309017181396484,
+      "loss": 0.3804,
+      "odds_ratio_loss": 0.16683904826641083,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02110280841588974,
+      "rewards/margins": 0.0654422789812088,
+      "rewards/rejected": -0.08654508739709854,
+      "sft_loss": 0.42205625772476196,
+      "step": 2785
+    },
+    {
+      "epoch": 2.232,
+      "grad_norm": 9.178748455728739,
+      "learning_rate": 9.334904715888496e-07,
+      "logits/chosen": -1.0844497680664062,
+      "logits/rejected": -1.070488691329956,
+      "logps/chosen": -0.32815462350845337,
+      "logps/rejected": -1.7352529764175415,
+      "loss": 0.4216,
+      "odds_ratio_loss": 0.09935733675956726,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.016407731920480728,
+      "rewards/margins": 0.07035491615533829,
+      "rewards/rejected": -0.08676265180110931,
+      "sft_loss": 0.32815462350845337,
+      "step": 2790
+    },
+    {
+      "epoch": 2.2359999999999998,
+      "grad_norm": 4.558311798079199,
+      "learning_rate": 9.244394619302338e-07,
+      "logits/chosen": -0.9466593861579895,
+      "logits/rejected": -0.8899604678153992,
+      "logps/chosen": -0.7910462617874146,
+      "logps/rejected": -1.691467523574829,
+      "loss": 0.5247,
+      "odds_ratio_loss": 0.33903276920318604,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.03955231234431267,
+      "rewards/margins": 0.04502106085419655,
+      "rewards/rejected": -0.08457337319850922,
+      "sft_loss": 0.7910462617874146,
+      "step": 2795
+    },
+    {
+      "epoch": 2.24,
+      "grad_norm": 5.935563711192816,
+      "learning_rate": 9.154225815032242e-07,
+      "logits/chosen": -0.7447048425674438,
+      "logits/rejected": -1.309199333190918,
+      "logps/chosen": -0.4879288673400879,
+      "logps/rejected": -1.4591110944747925,
+      "loss": 0.5058,
+      "odds_ratio_loss": 0.22252912819385529,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.024396440014243126,
+      "rewards/margins": 0.04855911061167717,
+      "rewards/rejected": -0.07295555621385574,
+      "sft_loss": 0.4879288673400879,
+      "step": 2800
+    },
+    {
+      "epoch": 2.2439999999999998,
+      "grad_norm": 5.083831381993147,
+      "learning_rate": 9.064400256282757e-07,
+      "logits/chosen": -0.4584059715270996,
+      "logits/rejected": -0.9872250556945801,
+      "logps/chosen": -0.5474129915237427,
+      "logps/rejected": -1.5436441898345947,
+      "loss": 0.4659,
+      "odds_ratio_loss": 0.20902732014656067,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.027370650321245193,
+      "rewards/margins": 0.049811553210020065,
+      "rewards/rejected": -0.07718220353126526,
+      "sft_loss": 0.5474129915237427,
+      "step": 2805
+    },
+    {
+      "epoch": 2.248,
+      "grad_norm": 6.37725598665606,
+      "learning_rate": 8.974919888823164e-07,
+      "logits/chosen": -1.1903154850006104,
+      "logits/rejected": -0.9519041180610657,
+      "logps/chosen": -0.407247930765152,
+      "logps/rejected": -1.9125579595565796,
+      "loss": 0.4918,
+      "odds_ratio_loss": 0.17733065783977509,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.02036239765584469,
+      "rewards/margins": 0.07526550441980362,
+      "rewards/rejected": -0.09562790393829346,
+      "sft_loss": 0.407247930765152,
+      "step": 2810
+    },
+    {
+      "epoch": 2.252,
+      "grad_norm": 23.791980041949664,
+      "learning_rate": 8.885786650945333e-07,
+      "logits/chosen": -0.9534207582473755,
+      "logits/rejected": -0.8568700551986694,
+      "logps/chosen": -0.4399814009666443,
+      "logps/rejected": -1.6407368183135986,
+      "loss": 0.4943,
+      "odds_ratio_loss": 0.2155378758907318,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.021999072283506393,
+      "rewards/margins": 0.0600377693772316,
+      "rewards/rejected": -0.08203683793544769,
+      "sft_loss": 0.4399814009666443,
+      "step": 2815
+    },
+    {
+      "epoch": 2.2560000000000002,
+      "grad_norm": 9.509659444717578,
+      "learning_rate": 8.797002473421729e-07,
+      "logits/chosen": -0.47383326292037964,
+      "logits/rejected": -0.8560094833374023,
+      "logps/chosen": -0.532326340675354,
+      "logps/rejected": -1.574268102645874,
+      "loss": 0.4286,
+      "odds_ratio_loss": 0.2277209758758545,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02661631628870964,
+      "rewards/margins": 0.05209709331393242,
+      "rewards/rejected": -0.07871340215206146,
+      "sft_loss": 0.532326340675354,
+      "step": 2820
+    },
+    {
+      "epoch": 2.26,
+      "grad_norm": 5.1984524732684765,
+      "learning_rate": 8.708569279463622e-07,
+      "logits/chosen": -0.8231816291809082,
+      "logits/rejected": -1.3245340585708618,
+      "logps/chosen": -0.7534009218215942,
+      "logps/rejected": -1.5175855159759521,
+      "loss": 0.4659,
+      "odds_ratio_loss": 0.22750845551490784,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03767004236578941,
+      "rewards/margins": 0.03820923715829849,
+      "rewards/rejected": -0.0758792832493782,
+      "sft_loss": 0.7534009218215942,
+      "step": 2825
+    },
+    {
+      "epoch": 2.2640000000000002,
+      "grad_norm": 5.302329927455291,
+      "learning_rate": 8.620488984679378e-07,
+      "logits/chosen": -1.174373984336853,
+      "logits/rejected": -0.9231562614440918,
+      "logps/chosen": -0.49920812249183655,
+      "logps/rejected": -1.5488044023513794,
+      "loss": 0.4205,
+      "odds_ratio_loss": 0.1963232159614563,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.024960406124591827,
+      "rewards/margins": 0.0524798147380352,
+      "rewards/rejected": -0.07744021713733673,
+      "sft_loss": 0.49920812249183655,
+      "step": 2830
+    },
+    {
+      "epoch": 2.268,
+      "grad_norm": 8.784170043250743,
+      "learning_rate": 8.532763497032987e-07,
+      "logits/chosen": -0.7181037068367004,
+      "logits/rejected": -1.2467901706695557,
+      "logps/chosen": -0.5864667892456055,
+      "logps/rejected": -1.448925256729126,
+      "loss": 0.5551,
+      "odds_ratio_loss": 0.26310136914253235,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.029323343187570572,
+      "rewards/margins": 0.043122924864292145,
+      "rewards/rejected": -0.07244626432657242,
+      "sft_loss": 0.5864667892456055,
+      "step": 2835
+    },
+    {
+      "epoch": 2.2720000000000002,
+      "grad_norm": 7.046247884710958,
+      "learning_rate": 8.445394716802754e-07,
+      "logits/chosen": -0.8855420351028442,
+      "logits/rejected": -1.3163471221923828,
+      "logps/chosen": -0.4142325818538666,
+      "logps/rejected": -1.6013071537017822,
+      "loss": 0.468,
+      "odds_ratio_loss": 0.14958259463310242,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.020711630582809448,
+      "rewards/margins": 0.05935372784733772,
+      "rewards/rejected": -0.08006535470485687,
+      "sft_loss": 0.4142325818538666,
+      "step": 2840
+    },
+    {
+      "epoch": 2.276,
+      "grad_norm": 5.056482274476605,
+      "learning_rate": 8.35838453654009e-07,
+      "logits/chosen": -0.7100471258163452,
+      "logits/rejected": -0.6780123114585876,
+      "logps/chosen": -0.6010003685951233,
+      "logps/rejected": -1.8489612340927124,
+      "loss": 0.4035,
+      "odds_ratio_loss": 0.2834250032901764,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.030050018802285194,
+      "rewards/margins": 0.0623980388045311,
+      "rewards/rejected": -0.09244807064533234,
+      "sft_loss": 0.6010003685951233,
+      "step": 2845
+    },
+    {
+      "epoch": 2.2800000000000002,
+      "grad_norm": 5.899418760006874,
+      "learning_rate": 8.271734841028553e-07,
+      "logits/chosen": -0.8226686716079712,
+      "logits/rejected": -1.2473328113555908,
+      "logps/chosen": -0.7500003576278687,
+      "logps/rejected": -1.7512264251708984,
+      "loss": 0.4737,
+      "odds_ratio_loss": 0.24083253741264343,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03750002384185791,
+      "rewards/margins": 0.05006130784749985,
+      "rewards/rejected": -0.08756133168935776,
+      "sft_loss": 0.7500003576278687,
+      "step": 2850
+    },
+    {
+      "epoch": 2.284,
+      "grad_norm": 6.588276147321622,
+      "learning_rate": 8.185447507243e-07,
+      "logits/chosen": -0.8183411359786987,
+      "logits/rejected": -0.9986063838005066,
+      "logps/chosen": -0.5513178110122681,
+      "logps/rejected": -1.8742460012435913,
+      "loss": 0.5713,
+      "odds_ratio_loss": 0.18965277075767517,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.027565892785787582,
+      "rewards/margins": 0.06614641845226288,
+      "rewards/rejected": -0.09371231496334076,
+      "sft_loss": 0.5513178110122681,
+      "step": 2855
+    },
+    {
+      "epoch": 2.288,
+      "grad_norm": 4.435394123744806,
+      "learning_rate": 8.099524404308948e-07,
+      "logits/chosen": -0.8991169929504395,
+      "logits/rejected": -0.829195499420166,
+      "logps/chosen": -0.4634367525577545,
+      "logps/rejected": -1.410571813583374,
+      "loss": 0.4853,
+      "odds_ratio_loss": 0.23965725302696228,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.023171838372945786,
+      "rewards/margins": 0.047356750816106796,
+      "rewards/rejected": -0.07052858918905258,
+      "sft_loss": 0.4634367525577545,
+      "step": 2860
+    },
+    {
+      "epoch": 2.292,
+      "grad_norm": 5.612134177673955,
+      "learning_rate": 8.013967393462094e-07,
+      "logits/chosen": -0.6299402713775635,
+      "logits/rejected": -0.7464720010757446,
+      "logps/chosen": -0.24951812624931335,
+      "logps/rejected": -1.7939624786376953,
+      "loss": 0.4291,
+      "odds_ratio_loss": 0.07172398269176483,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.012475905939936638,
+      "rewards/margins": 0.07722222059965134,
+      "rewards/rejected": -0.08969812840223312,
+      "sft_loss": 0.24951812624931335,
+      "step": 2865
+    },
+    {
+      "epoch": 2.296,
+      "grad_norm": 5.034253606387231,
+      "learning_rate": 7.928778328007918e-07,
+      "logits/chosen": -0.8041081428527832,
+      "logits/rejected": -0.8288570642471313,
+      "logps/chosen": -0.4491206109523773,
+      "logps/rejected": -1.529018521308899,
+      "loss": 0.4568,
+      "odds_ratio_loss": 0.1833478957414627,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.022456031292676926,
+      "rewards/margins": 0.05399489402770996,
+      "rewards/rejected": -0.07645092159509659,
+      "sft_loss": 0.4491206109523773,
+      "step": 2870
+    },
+    {
+      "epoch": 2.3,
+      "grad_norm": 4.4545637866951004,
+      "learning_rate": 7.843959053281663e-07,
+      "logits/chosen": -0.6113277673721313,
+      "logits/rejected": -1.2742953300476074,
+      "logps/chosen": -0.3947676420211792,
+      "logps/rejected": -1.7495063543319702,
+      "loss": 0.4967,
+      "odds_ratio_loss": 0.12505455315113068,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.01973838172852993,
+      "rewards/margins": 0.06773693859577179,
+      "rewards/rejected": -0.08747532218694687,
+      "sft_loss": 0.3947676420211792,
+      "step": 2875
+    },
+    {
+      "epoch": 2.304,
+      "grad_norm": 4.327743928984781,
+      "learning_rate": 7.759511406608255e-07,
+      "logits/chosen": -0.8660491108894348,
+      "logits/rejected": -0.9698010683059692,
+      "logps/chosen": -0.4570358395576477,
+      "logps/rejected": -1.4516005516052246,
+      "loss": 0.4305,
+      "odds_ratio_loss": 0.17015303671360016,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.022851794958114624,
+      "rewards/margins": 0.049728237092494965,
+      "rewards/rejected": -0.07258002460002899,
+      "sft_loss": 0.4570358395576477,
+      "step": 2880
+    },
+    {
+      "epoch": 2.308,
+      "grad_norm": 4.831001325484035,
+      "learning_rate": 7.675437217262571e-07,
+      "logits/chosen": -0.9820590019226074,
+      "logits/rejected": -1.3118226528167725,
+      "logps/chosen": -0.6205871105194092,
+      "logps/rejected": -1.3952069282531738,
+      "loss": 0.5001,
+      "odds_ratio_loss": 0.38392138481140137,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.0310293547809124,
+      "rewards/margins": 0.038730986416339874,
+      "rewards/rejected": -0.06976033747196198,
+      "sft_loss": 0.6205871105194092,
+      "step": 2885
+    },
+    {
+      "epoch": 2.312,
+      "grad_norm": 12.399106444555985,
+      "learning_rate": 7.591738306429769e-07,
+      "logits/chosen": -0.8063400387763977,
+      "logits/rejected": -0.97333824634552,
+      "logps/chosen": -0.38419461250305176,
+      "logps/rejected": -1.590563178062439,
+      "loss": 0.5036,
+      "odds_ratio_loss": 0.1262492537498474,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.019209731370210648,
+      "rewards/margins": 0.06031842902302742,
+      "rewards/rejected": -0.07952816039323807,
+      "sft_loss": 0.38419461250305176,
+      "step": 2890
+    },
+    {
+      "epoch": 2.316,
+      "grad_norm": 5.421465099897958,
+      "learning_rate": 7.508416487165862e-07,
+      "logits/chosen": -0.9876044392585754,
+      "logits/rejected": -0.8423392176628113,
+      "logps/chosen": -0.2360585480928421,
+      "logps/rejected": -1.1289315223693848,
+      "loss": 0.3904,
+      "odds_ratio_loss": 0.1510620415210724,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.01180292759090662,
+      "rewards/margins": 0.04464365169405937,
+      "rewards/rejected": -0.056446582078933716,
+      "sft_loss": 0.2360585480928421,
+      "step": 2895
+    },
+    {
+      "epoch": 2.32,
+      "grad_norm": 3.870106030569181,
+      "learning_rate": 7.425473564358457e-07,
+      "logits/chosen": -0.6248672604560852,
+      "logits/rejected": -1.0867903232574463,
+      "logps/chosen": -0.4465237557888031,
+      "logps/rejected": -1.3280917406082153,
+      "loss": 0.4823,
+      "odds_ratio_loss": 0.23933851718902588,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.022326190024614334,
+      "rewards/margins": 0.04407840222120285,
+      "rewards/rejected": -0.06640458852052689,
+      "sft_loss": 0.4465237557888031,
+      "step": 2900
+    },
+    {
+      "epoch": 2.324,
+      "grad_norm": 6.574323626048454,
+      "learning_rate": 7.342911334687619e-07,
+      "logits/chosen": -0.8632572293281555,
+      "logits/rejected": -1.020462989807129,
+      "logps/chosen": -0.429226815700531,
+      "logps/rejected": -1.3181707859039307,
+      "loss": 0.488,
+      "odds_ratio_loss": 0.21809737384319305,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02146134153008461,
+      "rewards/margins": 0.044447191059589386,
+      "rewards/rejected": -0.0659085363149643,
+      "sft_loss": 0.429226815700531,
+      "step": 2905
+    },
+    {
+      "epoch": 2.328,
+      "grad_norm": 6.6512572207835525,
+      "learning_rate": 7.260731586586983e-07,
+      "logits/chosen": -0.9791855812072754,
+      "logits/rejected": -0.8407464027404785,
+      "logps/chosen": -0.16527701914310455,
+      "logps/rejected": -1.412431001663208,
+      "loss": 0.4797,
+      "odds_ratio_loss": 0.13517996668815613,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.008263850584626198,
+      "rewards/margins": 0.06235770136117935,
+      "rewards/rejected": -0.070621557533741,
+      "sft_loss": 0.16527701914310455,
+      "step": 2910
+    },
+    {
+      "epoch": 2.332,
+      "grad_norm": 8.3662208877826,
+      "learning_rate": 7.178936100204994e-07,
+      "logits/chosen": -0.7169278860092163,
+      "logits/rejected": -1.1434614658355713,
+      "logps/chosen": -0.619706392288208,
+      "logps/rejected": -1.5916235446929932,
+      "loss": 0.5525,
+      "odds_ratio_loss": 0.2509163022041321,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03098532184958458,
+      "rewards/margins": 0.0485958568751812,
+      "rewards/rejected": -0.07958117872476578,
+      "sft_loss": 0.619706392288208,
+      "step": 2915
+    },
+    {
+      "epoch": 2.336,
+      "grad_norm": 21.850965502390842,
+      "learning_rate": 7.097526647366379e-07,
+      "logits/chosen": -0.9213175773620605,
+      "logits/rejected": -0.9148879051208496,
+      "logps/chosen": -0.4506239891052246,
+      "logps/rejected": -1.790827751159668,
+      "loss": 0.5104,
+      "odds_ratio_loss": 0.1461685597896576,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02253119833767414,
+      "rewards/margins": 0.06701019406318665,
+      "rewards/rejected": -0.08954139798879623,
+      "sft_loss": 0.4506239891052246,
+      "step": 2920
+    },
+    {
+      "epoch": 2.34,
+      "grad_norm": 4.021479030561525,
+      "learning_rate": 7.016504991533727e-07,
+      "logits/chosen": -0.9241275787353516,
+      "logits/rejected": -1.168180227279663,
+      "logps/chosen": -0.34740298986434937,
+      "logps/rejected": -1.5019714832305908,
+      "loss": 0.3721,
+      "odds_ratio_loss": 0.15549840033054352,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.017370149493217468,
+      "rewards/margins": 0.057728420943021774,
+      "rewards/rejected": -0.07509858161211014,
+      "sft_loss": 0.34740298986434937,
+      "step": 2925
+    },
+    {
+      "epoch": 2.344,
+      "grad_norm": 5.884610068850595,
+      "learning_rate": 6.935872887769299e-07,
+      "logits/chosen": -0.7997936010360718,
+      "logits/rejected": -0.9736310839653015,
+      "logps/chosen": -0.5550605058670044,
+      "logps/rejected": -1.6123679876327515,
+      "loss": 0.4905,
+      "odds_ratio_loss": 0.2296057492494583,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.027753029018640518,
+      "rewards/margins": 0.05286537855863571,
+      "rewards/rejected": -0.08061840385198593,
+      "sft_loss": 0.5550605058670044,
+      "step": 2930
+    },
+    {
+      "epoch": 2.348,
+      "grad_norm": 4.469717951127329,
+      "learning_rate": 6.855632082697045e-07,
+      "logits/chosen": -0.8227846026420593,
+      "logits/rejected": -1.05718195438385,
+      "logps/chosen": -0.5576878786087036,
+      "logps/rejected": -1.503408670425415,
+      "loss": 0.5521,
+      "odds_ratio_loss": 0.23267917335033417,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02788439765572548,
+      "rewards/margins": 0.04728604108095169,
+      "rewards/rejected": -0.07517042756080627,
+      "sft_loss": 0.5576878786087036,
+      "step": 2935
+    },
+    {
+      "epoch": 2.352,
+      "grad_norm": 4.357069152595027,
+      "learning_rate": 6.775784314464717e-07,
+      "logits/chosen": -0.7742995023727417,
+      "logits/rejected": -0.7836570143699646,
+      "logps/chosen": -0.6989033222198486,
+      "logps/rejected": -1.754209280014038,
+      "loss": 0.4823,
+      "odds_ratio_loss": 0.251584529876709,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03494516760110855,
+      "rewards/margins": 0.05276529863476753,
+      "rewards/rejected": -0.08771046996116638,
+      "sft_loss": 0.6989033222198486,
+      "step": 2940
+    },
+    {
+      "epoch": 2.356,
+      "grad_norm": 5.097226495202278,
+      "learning_rate": 6.696331312706245e-07,
+      "logits/chosen": -0.8427531123161316,
+      "logits/rejected": -0.6410783529281616,
+      "logps/chosen": -0.4761292338371277,
+      "logps/rejected": -1.3721054792404175,
+      "loss": 0.4982,
+      "odds_ratio_loss": 0.25603288412094116,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.023806463927030563,
+      "rewards/margins": 0.04479881748557091,
+      "rewards/rejected": -0.06860526651144028,
+      "sft_loss": 0.4761292338371277,
+      "step": 2945
+    },
+    {
+      "epoch": 2.36,
+      "grad_norm": 9.533319054768027,
+      "learning_rate": 6.617274798504286e-07,
+      "logits/chosen": -1.0047314167022705,
+      "logits/rejected": -1.1931557655334473,
+      "logps/chosen": -0.30853766202926636,
+      "logps/rejected": -1.573937177658081,
+      "loss": 0.4144,
+      "odds_ratio_loss": 0.1045946478843689,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.015426883473992348,
+      "rewards/margins": 0.0632699728012085,
+      "rewards/rejected": -0.07869686186313629,
+      "sft_loss": 0.30853766202926636,
+      "step": 2950
+    },
+    {
+      "epoch": 2.364,
+      "grad_norm": 5.794627406886997,
+      "learning_rate": 6.538616484352902e-07,
+      "logits/chosen": -0.6078363656997681,
+      "logits/rejected": -0.9947171211242676,
+      "logps/chosen": -0.5081497430801392,
+      "logps/rejected": -1.6594918966293335,
+      "loss": 0.4509,
+      "odds_ratio_loss": 0.18001045286655426,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.025407487526535988,
+      "rewards/margins": 0.057567108422517776,
+      "rewards/rejected": -0.08297459036111832,
+      "sft_loss": 0.5081497430801392,
+      "step": 2955
+    },
+    {
+      "epoch": 2.368,
+      "grad_norm": 8.384862451811653,
+      "learning_rate": 6.460358074120518e-07,
+      "logits/chosen": -1.0173108577728271,
+      "logits/rejected": -0.7680644392967224,
+      "logps/chosen": -0.2868219017982483,
+      "logps/rejected": -1.2532503604888916,
+      "loss": 0.4436,
+      "odds_ratio_loss": 0.1625426560640335,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.014341096393764019,
+      "rewards/margins": 0.04832141846418381,
+      "rewards/rejected": -0.0626625195145607,
+      "sft_loss": 0.2868219017982483,
+      "step": 2960
+    },
+    {
+      "epoch": 2.372,
+      "grad_norm": 5.987701383797447,
+      "learning_rate": 6.382501263012936e-07,
+      "logits/chosen": -0.39140084385871887,
+      "logits/rejected": -1.0467764139175415,
+      "logps/chosen": -0.48364967107772827,
+      "logps/rejected": -1.6967146396636963,
+      "loss": 0.4764,
+      "odds_ratio_loss": 0.14926201105117798,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.024182487279176712,
+      "rewards/margins": 0.06065324693918228,
+      "rewards/rejected": -0.0848357304930687,
+      "sft_loss": 0.48364967107772827,
+      "step": 2965
+    },
+    {
+      "epoch": 2.376,
+      "grad_norm": 5.365533722525087,
+      "learning_rate": 6.305047737536707e-07,
+      "logits/chosen": -1.085931658744812,
+      "logits/rejected": -0.8201769590377808,
+      "logps/chosen": -0.31420257687568665,
+      "logps/rejected": -1.8399406671524048,
+      "loss": 0.4935,
+      "odds_ratio_loss": 0.16777533292770386,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.015710126608610153,
+      "rewards/margins": 0.07628689706325531,
+      "rewards/rejected": -0.09199702739715576,
+      "sft_loss": 0.31420257687568665,
+      "step": 2970
+    },
+    {
+      "epoch": 2.38,
+      "grad_norm": 4.187858613296553,
+      "learning_rate": 6.227999175462521e-07,
+      "logits/chosen": -1.0166877508163452,
+      "logits/rejected": -1.3475017547607422,
+      "logps/chosen": -0.45567312836647034,
+      "logps/rejected": -1.3064647912979126,
+      "loss": 0.4065,
+      "odds_ratio_loss": 0.20544198155403137,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.022783655673265457,
+      "rewards/margins": 0.0425395742058754,
+      "rewards/rejected": -0.06532323360443115,
+      "sft_loss": 0.45567312836647034,
+      "step": 2975
+    },
+    {
+      "epoch": 2.384,
+      "grad_norm": 5.441360920553305,
+      "learning_rate": 6.151357245788917e-07,
+      "logits/chosen": -0.4649627208709717,
+      "logits/rejected": -0.7259455323219299,
+      "logps/chosen": -0.32298606634140015,
+      "logps/rejected": -1.7776187658309937,
+      "loss": 0.421,
+      "odds_ratio_loss": 0.07744672149419785,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.016149302944540977,
+      "rewards/margins": 0.0727316364645958,
+      "rewards/rejected": -0.08888094127178192,
+      "sft_loss": 0.32298606634140015,
+      "step": 2980
+    },
+    {
+      "epoch": 2.388,
+      "grad_norm": 6.473770551803507,
+      "learning_rate": 6.075123608706093e-07,
+      "logits/chosen": -0.9025441408157349,
+      "logits/rejected": -0.8097003698348999,
+      "logps/chosen": -0.30892783403396606,
+      "logps/rejected": -1.4229497909545898,
+      "loss": 0.4896,
+      "odds_ratio_loss": 0.18760545551776886,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.015446392819285393,
+      "rewards/margins": 0.05570109561085701,
+      "rewards/rejected": -0.07114748656749725,
+      "sft_loss": 0.30892783403396606,
+      "step": 2985
+    },
+    {
+      "epoch": 2.392,
+      "grad_norm": 7.045700421865737,
+      "learning_rate": 5.999299915559956e-07,
+      "logits/chosen": -0.8020213842391968,
+      "logits/rejected": -0.9755544662475586,
+      "logps/chosen": -0.3655335307121277,
+      "logps/rejected": -1.7111154794692993,
+      "loss": 0.4826,
+      "odds_ratio_loss": 0.1401543915271759,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.018276674672961235,
+      "rewards/margins": 0.06727910041809082,
+      "rewards/rejected": -0.08555576950311661,
+      "sft_loss": 0.3655335307121277,
+      "step": 2990
+    },
+    {
+      "epoch": 2.396,
+      "grad_norm": 5.1559133299959345,
+      "learning_rate": 5.923887808816373e-07,
+      "logits/chosen": -0.7541839480400085,
+      "logits/rejected": -0.8241696357727051,
+      "logps/chosen": -0.32986879348754883,
+      "logps/rejected": -1.7209409475326538,
+      "loss": 0.4559,
+      "odds_ratio_loss": 0.11603020131587982,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.01649343967437744,
+      "rewards/margins": 0.06955362856388092,
+      "rewards/rejected": -0.08604706078767776,
+      "sft_loss": 0.32986879348754883,
+      "step": 2995
+    },
+    {
+      "epoch": 2.4,
+      "grad_norm": 5.5494435865172695,
+      "learning_rate": 5.848888922025553e-07,
+      "logits/chosen": -0.604158341884613,
+      "logits/rejected": -1.010777473449707,
+      "logps/chosen": -0.18639525771141052,
+      "logps/rejected": -1.7561891078948975,
+      "loss": 0.4748,
+      "odds_ratio_loss": 0.08248431980609894,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.009319763630628586,
+      "rewards/margins": 0.07848969846963882,
+      "rewards/rejected": -0.08780945837497711,
+      "sft_loss": 0.18639525771141052,
+      "step": 3000
+    },
+    {
+      "epoch": 2.404,
+      "grad_norm": 7.404224976766671,
+      "learning_rate": 5.774304879786688e-07,
+      "logits/chosen": -0.8347930908203125,
+      "logits/rejected": -1.0326844453811646,
+      "logps/chosen": -0.4041759967803955,
+      "logps/rejected": -1.3478691577911377,
+      "loss": 0.4676,
+      "odds_ratio_loss": 0.1940927505493164,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.020208800211548805,
+      "rewards/margins": 0.04718465730547905,
+      "rewards/rejected": -0.067393459379673,
+      "sft_loss": 0.4041759967803955,
+      "step": 3005
+    },
+    {
+      "epoch": 2.408,
+      "grad_norm": 6.3948612577090005,
+      "learning_rate": 5.700137297712749e-07,
+      "logits/chosen": -1.0781855583190918,
+      "logits/rejected": -0.7031400799751282,
+      "logps/chosen": -0.3962685465812683,
+      "logps/rejected": -1.2765676975250244,
+      "loss": 0.5032,
+      "odds_ratio_loss": 0.2406892478466034,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.019813427701592445,
+      "rewards/margins": 0.04401496425271034,
+      "rewards/rejected": -0.06382839381694794,
+      "sft_loss": 0.3962685465812683,
+      "step": 3010
+    },
+    {
+      "epoch": 2.412,
+      "grad_norm": 5.487144565387309,
+      "learning_rate": 5.626387782395512e-07,
+      "logits/chosen": -0.8737370371818542,
+      "logits/rejected": -1.1139321327209473,
+      "logps/chosen": -0.6051273345947266,
+      "logps/rejected": -1.6024173498153687,
+      "loss": 0.4723,
+      "odds_ratio_loss": 0.23752835392951965,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.0302563663572073,
+      "rewards/margins": 0.0498645082116127,
+      "rewards/rejected": -0.08012087643146515,
+      "sft_loss": 0.6051273345947266,
+      "step": 3015
+    },
+    {
+      "epoch": 2.416,
+      "grad_norm": 6.210678255713737,
+      "learning_rate": 5.553057931370729e-07,
+      "logits/chosen": -0.9348379373550415,
+      "logits/rejected": -0.7690949440002441,
+      "logps/chosen": -0.6480275392532349,
+      "logps/rejected": -1.3571795225143433,
+      "loss": 0.5108,
+      "odds_ratio_loss": 0.2786480188369751,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.032401375472545624,
+      "rewards/margins": 0.03545759990811348,
+      "rewards/rejected": -0.0678589791059494,
+      "sft_loss": 0.6480275392532349,
+      "step": 3020
+    },
+    {
+      "epoch": 2.42,
+      "grad_norm": 4.9274946658912615,
+      "learning_rate": 5.48014933308352e-07,
+      "logits/chosen": -0.7441031336784363,
+      "logits/rejected": -1.0764060020446777,
+      "logps/chosen": -0.38686519861221313,
+      "logps/rejected": -1.4046424627304077,
+      "loss": 0.5403,
+      "odds_ratio_loss": 0.14413750171661377,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.019343260675668716,
+      "rewards/margins": 0.05088886618614197,
+      "rewards/rejected": -0.07023213058710098,
+      "sft_loss": 0.38686519861221313,
+      "step": 3025
+    },
+    {
+      "epoch": 2.424,
+      "grad_norm": 4.697225271080028,
+      "learning_rate": 5.407663566854008e-07,
+      "logits/chosen": -0.9233636856079102,
+      "logits/rejected": -0.7391080260276794,
+      "logps/chosen": -0.3096584677696228,
+      "logps/rejected": -1.552175760269165,
+      "loss": 0.5123,
+      "odds_ratio_loss": 0.2237163782119751,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.015482926741242409,
+      "rewards/margins": 0.062125854194164276,
+      "rewards/rejected": -0.07760877907276154,
+      "sft_loss": 0.3096584677696228,
+      "step": 3030
+    },
+    {
+      "epoch": 2.428,
+      "grad_norm": 5.079507350370886,
+      "learning_rate": 5.335602202843054e-07,
+      "logits/chosen": -0.7503083944320679,
+      "logits/rejected": -0.8697785139083862,
+      "logps/chosen": -0.5030153393745422,
+      "logps/rejected": -1.383043885231018,
+      "loss": 0.4996,
+      "odds_ratio_loss": 0.20985543727874756,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.025150766596198082,
+      "rewards/margins": 0.04400142282247543,
+      "rewards/rejected": -0.06915219128131866,
+      "sft_loss": 0.5030153393745422,
+      "step": 3035
+    },
+    {
+      "epoch": 2.432,
+      "grad_norm": 5.622928495063832,
+      "learning_rate": 5.263966802018275e-07,
+      "logits/chosen": -0.8630277514457703,
+      "logits/rejected": -0.9782209396362305,
+      "logps/chosen": -0.364883691072464,
+      "logps/rejected": -3.0828795433044434,
+      "loss": 0.3849,
+      "odds_ratio_loss": 0.11438252776861191,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.01824418641626835,
+      "rewards/margins": 0.13589979708194733,
+      "rewards/rejected": -0.15414398908615112,
+      "sft_loss": 0.364883691072464,
+      "step": 3040
+    },
+    {
+      "epoch": 2.436,
+      "grad_norm": 6.518609406323272,
+      "learning_rate": 5.192758916120236e-07,
+      "logits/chosen": -0.720903754234314,
+      "logits/rejected": -0.9428263902664185,
+      "logps/chosen": -0.44252529740333557,
+      "logps/rejected": -1.551532506942749,
+      "loss": 0.5285,
+      "odds_ratio_loss": 0.19013534486293793,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.022126266732811928,
+      "rewards/margins": 0.05545036867260933,
+      "rewards/rejected": -0.0775766372680664,
+      "sft_loss": 0.44252529740333557,
+      "step": 3045
+    },
+    {
+      "epoch": 2.44,
+      "grad_norm": 6.898029710003963,
+      "learning_rate": 5.121980087628802e-07,
+      "logits/chosen": -1.102452039718628,
+      "logits/rejected": -1.1401493549346924,
+      "logps/chosen": -0.6420519351959229,
+      "logps/rejected": -1.6801929473876953,
+      "loss": 0.491,
+      "odds_ratio_loss": 0.22621390223503113,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03210259601473808,
+      "rewards/margins": 0.05190705135464668,
+      "rewards/rejected": -0.08400964736938477,
+      "sft_loss": 0.6420519351959229,
+      "step": 3050
+    },
+    {
+      "epoch": 2.444,
+      "grad_norm": 7.428665380081318,
+      "learning_rate": 5.051631849729785e-07,
+      "logits/chosen": -0.7611395120620728,
+      "logits/rejected": -0.9640461802482605,
+      "logps/chosen": -0.7123726606369019,
+      "logps/rejected": -1.2486783266067505,
+      "loss": 0.4901,
+      "odds_ratio_loss": 0.40984076261520386,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.03561863675713539,
+      "rewards/margins": 0.026815274730324745,
+      "rewards/rejected": -0.06243390962481499,
+      "sft_loss": 0.7123726606369019,
+      "step": 3055
+    },
+    {
+      "epoch": 2.448,
+      "grad_norm": 4.15877124152584,
+      "learning_rate": 4.981715726281666e-07,
+      "logits/chosen": -0.8188239336013794,
+      "logits/rejected": -1.3134756088256836,
+      "logps/chosen": -0.42243289947509766,
+      "logps/rejected": -1.7504783868789673,
+      "loss": 0.4158,
+      "odds_ratio_loss": 0.14152325689792633,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.021121647208929062,
+      "rewards/margins": 0.06640227138996124,
+      "rewards/rejected": -0.08752389997243881,
+      "sft_loss": 0.42243289947509766,
+      "step": 3060
+    },
+    {
+      "epoch": 2.452,
+      "grad_norm": 4.5394562483362435,
+      "learning_rate": 4.912233231782623e-07,
+      "logits/chosen": -0.844011127948761,
+      "logits/rejected": -0.77781081199646,
+      "logps/chosen": -0.5193901658058167,
+      "logps/rejected": -1.5341647863388062,
+      "loss": 0.5074,
+      "odds_ratio_loss": 0.34647947549819946,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.02596951089799404,
+      "rewards/margins": 0.050738729536533356,
+      "rewards/rejected": -0.07670824229717255,
+      "sft_loss": 0.5193901658058167,
+      "step": 3065
+    },
+    {
+      "epoch": 2.456,
+      "grad_norm": 5.9484966146690885,
+      "learning_rate": 4.843185871337722e-07,
+      "logits/chosen": -1.1236478090286255,
+      "logits/rejected": -1.187819242477417,
+      "logps/chosen": -0.3840157687664032,
+      "logps/rejected": -1.7428499460220337,
+      "loss": 0.5018,
+      "odds_ratio_loss": 0.13463526964187622,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.01920078694820404,
+      "rewards/margins": 0.06794170290231705,
+      "rewards/rejected": -0.08714248239994049,
+      "sft_loss": 0.3840157687664032,
+      "step": 3070
+    },
+    {
+      "epoch": 2.46,
+      "grad_norm": 5.253460866623653,
+      "learning_rate": 4.774575140626317e-07,
+      "logits/chosen": -0.7584226131439209,
+      "logits/rejected": -0.8733755350112915,
+      "logps/chosen": -0.6464089155197144,
+      "logps/rejected": -1.525433897972107,
+      "loss": 0.4441,
+      "odds_ratio_loss": 0.25305792689323425,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03232044726610184,
+      "rewards/margins": 0.04395125061273575,
+      "rewards/rejected": -0.07627169787883759,
+      "sft_loss": 0.6464089155197144,
+      "step": 3075
+    },
+    {
+      "epoch": 2.464,
+      "grad_norm": 12.350993791450163,
+      "learning_rate": 4.706402525869633e-07,
+      "logits/chosen": -0.7759664058685303,
+      "logits/rejected": -0.9143487811088562,
+      "logps/chosen": -0.3908182978630066,
+      "logps/rejected": -1.8124420642852783,
+      "loss": 0.4794,
+      "odds_ratio_loss": 0.12326017767190933,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.01954091526567936,
+      "rewards/margins": 0.07108117640018463,
+      "rewards/rejected": -0.09062208980321884,
+      "sft_loss": 0.3908182978630066,
+      "step": 3080
+    },
+    {
+      "epoch": 2.468,
+      "grad_norm": 14.144313900187102,
+      "learning_rate": 4.638669503798579e-07,
+      "logits/chosen": -0.8764970898628235,
+      "logits/rejected": -0.9682470560073853,
+      "logps/chosen": -0.3992885947227478,
+      "logps/rejected": -1.7859230041503906,
+      "loss": 0.5238,
+      "odds_ratio_loss": 0.14603550732135773,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.0199644286185503,
+      "rewards/margins": 0.06933172792196274,
+      "rewards/rejected": -0.08929616957902908,
+      "sft_loss": 0.3992885947227478,
+      "step": 3085
+    },
+    {
+      "epoch": 2.472,
+      "grad_norm": 4.717707133780473,
+      "learning_rate": 4.5713775416217884e-07,
+      "logits/chosen": -0.7459074854850769,
+      "logits/rejected": -0.9472630620002747,
+      "logps/chosen": -0.3936541676521301,
+      "logps/rejected": -1.3656264543533325,
+      "loss": 0.4566,
+      "odds_ratio_loss": 0.1686370074748993,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.019682709127664566,
+      "rewards/margins": 0.04859861359000206,
+      "rewards/rejected": -0.06828131526708603,
+      "sft_loss": 0.3936541676521301,
+      "step": 3090
+    },
+    {
+      "epoch": 2.476,
+      "grad_norm": 6.069217496310175,
+      "learning_rate": 4.5045280969937847e-07,
+      "logits/chosen": -0.7254356145858765,
+      "logits/rejected": -1.232452154159546,
+      "logps/chosen": -0.5967345833778381,
+      "logps/rejected": -1.7667179107666016,
+      "loss": 0.4779,
+      "odds_ratio_loss": 0.31498411297798157,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.029836729168891907,
+      "rewards/margins": 0.05849916860461235,
+      "rewards/rejected": -0.08833589404821396,
+      "sft_loss": 0.5967345833778381,
+      "step": 3095
+    },
+    {
+      "epoch": 2.48,
+      "grad_norm": 6.104980238131941,
+      "learning_rate": 4.438122617983442e-07,
+      "logits/chosen": -0.8972131609916687,
+      "logits/rejected": -0.9765474200248718,
+      "logps/chosen": -0.42376136779785156,
+      "logps/rejected": -1.4339518547058105,
+      "loss": 0.4595,
+      "odds_ratio_loss": 0.17065298557281494,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02118806727230549,
+      "rewards/margins": 0.05050952360033989,
+      "rewards/rejected": -0.07169759273529053,
+      "sft_loss": 0.42376136779785156,
+      "step": 3100
+    },
+    {
+      "epoch": 2.484,
+      "grad_norm": 6.781146830921427,
+      "learning_rate": 4.372162543042624e-07,
+      "logits/chosen": -0.8027432560920715,
+      "logits/rejected": -0.9535185694694519,
+      "logps/chosen": -0.5751426219940186,
+      "logps/rejected": -1.7298011779785156,
+      "loss": 0.443,
+      "odds_ratio_loss": 0.19305340945720673,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02875712886452675,
+      "rewards/margins": 0.05773293972015381,
+      "rewards/rejected": -0.08649007230997086,
+      "sft_loss": 0.5751426219940186,
+      "step": 3105
+    },
+    {
+      "epoch": 2.488,
+      "grad_norm": 5.981163589398859,
+      "learning_rate": 4.3066493009749853e-07,
+      "logits/chosen": -0.6171215772628784,
+      "logits/rejected": -0.874431312084198,
+      "logps/chosen": -0.6073184609413147,
+      "logps/rejected": -1.8347069025039673,
+      "loss": 0.3958,
+      "odds_ratio_loss": 0.24824929237365723,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.030365925282239914,
+      "rewards/margins": 0.061369407922029495,
+      "rewards/rejected": -0.09173533320426941,
+      "sft_loss": 0.6073184609413147,
+      "step": 3110
+    },
+    {
+      "epoch": 2.492,
+      "grad_norm": 5.642504715902896,
+      "learning_rate": 4.2415843109050667e-07,
+      "logits/chosen": -0.9500174522399902,
+      "logits/rejected": -1.0051153898239136,
+      "logps/chosen": -0.3405509889125824,
+      "logps/rejected": -1.6603267192840576,
+      "loss": 0.4716,
+      "odds_ratio_loss": 0.1337331086397171,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.01702754944562912,
+      "rewards/margins": 0.06598879396915436,
+      "rewards/rejected": -0.08301634341478348,
+      "sft_loss": 0.3405509889125824,
+      "step": 3115
+    },
+    {
+      "epoch": 2.496,
+      "grad_norm": 6.4696426005968775,
+      "learning_rate": 4.1769689822475147e-07,
+      "logits/chosen": -0.57563316822052,
+      "logits/rejected": -0.9418606758117676,
+      "logps/chosen": -0.5054479837417603,
+      "logps/rejected": -1.5817244052886963,
+      "loss": 0.4687,
+      "odds_ratio_loss": 0.21483473479747772,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.025272399187088013,
+      "rewards/margins": 0.05381382629275322,
+      "rewards/rejected": -0.07908622920513153,
+      "sft_loss": 0.5054479837417603,
+      "step": 3120
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 4.172914702601585,
+      "learning_rate": 4.1128047146765936e-07,
+      "logits/chosen": -0.7602171897888184,
+      "logits/rejected": -1.0409475564956665,
+      "logps/chosen": -0.40496939420700073,
+      "logps/rejected": -1.4085716009140015,
+      "loss": 0.4094,
+      "odds_ratio_loss": 0.21070821583271027,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.020248468965291977,
+      "rewards/margins": 0.050180114805698395,
+      "rewards/rejected": -0.07042858749628067,
+      "sft_loss": 0.40496939420700073,
+      "step": 3125
+    },
+    {
+      "epoch": 2.504,
+      "grad_norm": 5.627553812250054,
+      "learning_rate": 4.049092898095816e-07,
+      "logits/chosen": -0.8009653091430664,
+      "logits/rejected": -1.1446640491485596,
+      "logps/chosen": -0.302462637424469,
+      "logps/rejected": -1.758111596107483,
+      "loss": 0.5068,
+      "odds_ratio_loss": 0.08966059982776642,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.015123131684958935,
+      "rewards/margins": 0.07278244942426682,
+      "rewards/rejected": -0.08790557831525803,
+      "sft_loss": 0.302462637424469,
+      "step": 3130
+    },
+    {
+      "epoch": 2.508,
+      "grad_norm": 5.282423292390493,
+      "learning_rate": 3.9858349126078945e-07,
+      "logits/chosen": -1.0294119119644165,
+      "logits/rejected": -0.672359049320221,
+      "logps/chosen": -0.2604931592941284,
+      "logps/rejected": -1.511343002319336,
+      "loss": 0.4927,
+      "odds_ratio_loss": 0.10487516969442368,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.013024657964706421,
+      "rewards/margins": 0.06254249066114426,
+      "rewards/rejected": -0.07556714862585068,
+      "sft_loss": 0.2604931592941284,
+      "step": 3135
+    },
+    {
+      "epoch": 2.512,
+      "grad_norm": 5.2623944719501585,
+      "learning_rate": 3.9230321284847856e-07,
+      "logits/chosen": -0.9906686544418335,
+      "logits/rejected": -0.7455857992172241,
+      "logps/chosen": -0.599448561668396,
+      "logps/rejected": -1.501896619796753,
+      "loss": 0.4545,
+      "odds_ratio_loss": 0.2790692448616028,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.02997243031859398,
+      "rewards/margins": 0.045122403651475906,
+      "rewards/rejected": -0.07509483397006989,
+      "sft_loss": 0.599448561668396,
+      "step": 3140
+    },
+    {
+      "epoch": 2.516,
+      "grad_norm": 5.779281261694176,
+      "learning_rate": 3.86068590613804e-07,
+      "logits/chosen": -0.8702551126480103,
+      "logits/rejected": -1.0743459463119507,
+      "logps/chosen": -0.2949695885181427,
+      "logps/rejected": -1.3731797933578491,
+      "loss": 0.4727,
+      "odds_ratio_loss": 0.13363251090049744,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.01474847923964262,
+      "rewards/margins": 0.0539105124771595,
+      "rewards/rejected": -0.0686589926481247,
+      "sft_loss": 0.2949695885181427,
+      "step": 3145
+    },
+    {
+      "epoch": 2.52,
+      "grad_norm": 6.263065115062914,
+      "learning_rate": 3.798797596089351e-07,
+      "logits/chosen": -1.070436716079712,
+      "logits/rejected": -1.0166070461273193,
+      "logps/chosen": -0.2887745499610901,
+      "logps/rejected": -1.5660569667816162,
+      "loss": 0.4361,
+      "odds_ratio_loss": 0.11832698434591293,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.014438727870583534,
+      "rewards/margins": 0.06386412680149078,
+      "rewards/rejected": -0.07830285280942917,
+      "sft_loss": 0.2887745499610901,
+      "step": 3150
+    },
+    {
+      "epoch": 2.524,
+      "grad_norm": 6.821197735395411,
+      "learning_rate": 3.737368538941255e-07,
+      "logits/chosen": -0.9708150029182434,
+      "logits/rejected": -0.8317630887031555,
+      "logps/chosen": -0.41128820180892944,
+      "logps/rejected": -1.2074782848358154,
+      "loss": 0.4127,
+      "odds_ratio_loss": 0.2423263043165207,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.020564410835504532,
+      "rewards/margins": 0.03980950638651848,
+      "rewards/rejected": -0.06037392094731331,
+      "sft_loss": 0.41128820180892944,
+      "step": 3155
+    },
+    {
+      "epoch": 2.528,
+      "grad_norm": 5.542329936785543,
+      "learning_rate": 3.6764000653481263e-07,
+      "logits/chosen": -1.2484712600708008,
+      "logits/rejected": -0.9088461995124817,
+      "logps/chosen": -0.3338097929954529,
+      "logps/rejected": -1.6267837285995483,
+      "loss": 0.4552,
+      "odds_ratio_loss": 0.12026141583919525,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.016690488904714584,
+      "rewards/margins": 0.06464870274066925,
+      "rewards/rejected": -0.08133919537067413,
+      "sft_loss": 0.3338097929954529,
+      "step": 3160
+    },
+    {
+      "epoch": 2.532,
+      "grad_norm": 8.97897074618829,
+      "learning_rate": 3.615893495987335e-07,
+      "logits/chosen": -0.5620380640029907,
+      "logits/rejected": -0.8675304651260376,
+      "logps/chosen": -0.5898982882499695,
+      "logps/rejected": -1.3201124668121338,
+      "loss": 0.4837,
+      "odds_ratio_loss": 0.3108993470668793,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.029494917020201683,
+      "rewards/margins": 0.036510709673166275,
+      "rewards/rejected": -0.0660056322813034,
+      "sft_loss": 0.5898982882499695,
+      "step": 3165
+    },
+    {
+      "epoch": 2.536,
+      "grad_norm": 4.681514042861988,
+      "learning_rate": 3.555850141530659e-07,
+      "logits/chosen": -1.0148943662643433,
+      "logits/rejected": -0.982884407043457,
+      "logps/chosen": -0.22993119060993195,
+      "logps/rejected": -1.4043452739715576,
+      "loss": 0.5221,
+      "odds_ratio_loss": 0.08220822364091873,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.011496557854115963,
+      "rewards/margins": 0.058720700442790985,
+      "rewards/rejected": -0.07021726667881012,
+      "sft_loss": 0.22993119060993195,
+      "step": 3170
+    },
+    {
+      "epoch": 2.54,
+      "grad_norm": 7.722013429404472,
+      "learning_rate": 3.4962713026158697e-07,
+      "logits/chosen": -0.5106195211410522,
+      "logits/rejected": -1.0413014888763428,
+      "logps/chosen": -0.6414362788200378,
+      "logps/rejected": -1.3033192157745361,
+      "loss": 0.4655,
+      "odds_ratio_loss": 0.3088650107383728,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03207181394100189,
+      "rewards/margins": 0.0330941416323185,
+      "rewards/rejected": -0.06516595929861069,
+      "sft_loss": 0.6414362788200378,
+      "step": 3175
+    },
+    {
+      "epoch": 2.544,
+      "grad_norm": 4.914705494262134,
+      "learning_rate": 3.4371582698185636e-07,
+      "logits/chosen": -1.18616783618927,
+      "logits/rejected": -0.963121771812439,
+      "logps/chosen": -0.31630367040634155,
+      "logps/rejected": -1.29836106300354,
+      "loss": 0.5218,
+      "odds_ratio_loss": 0.20544719696044922,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.015815183520317078,
+      "rewards/margins": 0.049102868884801865,
+      "rewards/rejected": -0.06491805613040924,
+      "sft_loss": 0.31630367040634155,
+      "step": 3180
+    },
+    {
+      "epoch": 2.548,
+      "grad_norm": 10.854835141924323,
+      "learning_rate": 3.378512323624228e-07,
+      "logits/chosen": -0.5074478387832642,
+      "logits/rejected": -1.2903783321380615,
+      "logps/chosen": -0.7434813976287842,
+      "logps/rejected": -1.6036231517791748,
+      "loss": 0.5155,
+      "odds_ratio_loss": 0.265799343585968,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03717407211661339,
+      "rewards/margins": 0.04300709441304207,
+      "rewards/rejected": -0.08018116652965546,
+      "sft_loss": 0.7434813976287842,
+      "step": 3185
+    },
+    {
+      "epoch": 2.552,
+      "grad_norm": 5.764157322717205,
+      "learning_rate": 3.3203347344004737e-07,
+      "logits/chosen": -0.7570070028305054,
+      "logits/rejected": -1.1250250339508057,
+      "logps/chosen": -0.6759325265884399,
+      "logps/rejected": -1.439347267150879,
+      "loss": 0.5504,
+      "odds_ratio_loss": 0.3499813973903656,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.033796630799770355,
+      "rewards/margins": 0.03817072883248329,
+      "rewards/rejected": -0.07196736335754395,
+      "sft_loss": 0.6759325265884399,
+      "step": 3190
+    },
+    {
+      "epoch": 2.556,
+      "grad_norm": 4.888945884008844,
+      "learning_rate": 3.262626762369525e-07,
+      "logits/chosen": -1.1923449039459229,
+      "logits/rejected": -0.8670722842216492,
+      "logps/chosen": -0.31347471475601196,
+      "logps/rejected": -1.4067668914794922,
+      "loss": 0.4052,
+      "odds_ratio_loss": 0.12347825616598129,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.01567373424768448,
+      "rewards/margins": 0.05466460436582565,
+      "rewards/rejected": -0.07033834606409073,
+      "sft_loss": 0.31347471475601196,
+      "step": 3195
+    },
+    {
+      "epoch": 2.56,
+      "grad_norm": 4.380022841165363,
+      "learning_rate": 3.2053896575809426e-07,
+      "logits/chosen": -0.5602080821990967,
+      "logits/rejected": -1.0972874164581299,
+      "logps/chosen": -0.7473759651184082,
+      "logps/rejected": -1.6633918285369873,
+      "loss": 0.4408,
+      "odds_ratio_loss": 0.2667500674724579,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03736879676580429,
+      "rewards/margins": 0.04580079764127731,
+      "rewards/rejected": -0.0831695944070816,
+      "sft_loss": 0.7473759651184082,
+      "step": 3200
+    },
+    {
+      "epoch": 2.564,
+      "grad_norm": 5.674973517980157,
+      "learning_rate": 3.148624659884508e-07,
+      "logits/chosen": -0.7617183327674866,
+      "logits/rejected": -0.972303569316864,
+      "logps/chosen": -0.4322594702243805,
+      "logps/rejected": -1.4709136486053467,
+      "loss": 0.4488,
+      "odds_ratio_loss": 0.186048686504364,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.021612973883748055,
+      "rewards/margins": 0.05193271115422249,
+      "rewards/rejected": -0.0735456794500351,
+      "sft_loss": 0.4322594702243805,
+      "step": 3205
+    },
+    {
+      "epoch": 2.568,
+      "grad_norm": 4.977829670896469,
+      "learning_rate": 3.092332998903416e-07,
+      "logits/chosen": -0.8257712125778198,
+      "logits/rejected": -0.8016840219497681,
+      "logps/chosen": -0.4363314211368561,
+      "logps/rejected": -1.1737916469573975,
+      "loss": 0.4606,
+      "odds_ratio_loss": 0.22321203351020813,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.021816570311784744,
+      "rewards/margins": 0.03687301650643349,
+      "rewards/rejected": -0.05868958681821823,
+      "sft_loss": 0.4363314211368561,
+      "step": 3210
+    },
+    {
+      "epoch": 2.572,
+      "grad_norm": 5.0240652358480755,
+      "learning_rate": 3.0365158940075664e-07,
+      "logits/chosen": -1.0381290912628174,
+      "logits/rejected": -1.1659702062606812,
+      "logps/chosen": -0.5309845209121704,
+      "logps/rejected": -1.4535753726959229,
+      "loss": 0.6035,
+      "odds_ratio_loss": 0.28817451000213623,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.02654922381043434,
+      "rewards/margins": 0.04612954333424568,
+      "rewards/rejected": -0.07267877459526062,
+      "sft_loss": 0.5309845209121704,
+      "step": 3215
+    },
+    {
+      "epoch": 2.576,
+      "grad_norm": 5.408869231167505,
+      "learning_rate": 2.981174554287239e-07,
+      "logits/chosen": -1.015199899673462,
+      "logits/rejected": -1.3747820854187012,
+      "logps/chosen": -0.4069809317588806,
+      "logps/rejected": -1.5189340114593506,
+      "loss": 0.4696,
+      "odds_ratio_loss": 0.16449251770973206,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02034904807806015,
+      "rewards/margins": 0.05559765547513962,
+      "rewards/rejected": -0.07594670355319977,
+      "sft_loss": 0.4069809317588806,
+      "step": 3220
+    },
+    {
+      "epoch": 2.58,
+      "grad_norm": 7.760925759203226,
+      "learning_rate": 2.9263101785268253e-07,
+      "logits/chosen": -0.5573514699935913,
+      "logits/rejected": -1.252190113067627,
+      "logps/chosen": -0.41566723585128784,
+      "logps/rejected": -1.4632532596588135,
+      "loss": 0.4529,
+      "odds_ratio_loss": 0.16219457983970642,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02078336291015148,
+      "rewards/margins": 0.0523792989552021,
+      "rewards/rejected": -0.07316266000270844,
+      "sft_loss": 0.41566723585128784,
+      "step": 3225
+    },
+    {
+      "epoch": 2.584,
+      "grad_norm": 6.9168438538457995,
+      "learning_rate": 2.871923955178918e-07,
+      "logits/chosen": -0.7818273901939392,
+      "logits/rejected": -0.9009189605712891,
+      "logps/chosen": -0.5484607219696045,
+      "logps/rejected": -1.4074246883392334,
+      "loss": 0.4963,
+      "odds_ratio_loss": 0.2880396842956543,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.027423039078712463,
+      "rewards/margins": 0.042948197573423386,
+      "rewards/rejected": -0.07037124037742615,
+      "sft_loss": 0.5484607219696045,
+      "step": 3230
+    },
+    {
+      "epoch": 2.588,
+      "grad_norm": 5.722693546238487,
+      "learning_rate": 2.8180170623385213e-07,
+      "logits/chosen": -0.9606497883796692,
+      "logits/rejected": -0.9924991726875305,
+      "logps/chosen": -0.4993307590484619,
+      "logps/rejected": -2.206209659576416,
+      "loss": 0.4349,
+      "odds_ratio_loss": 0.2471529245376587,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.024966537952423096,
+      "rewards/margins": 0.08534395694732666,
+      "rewards/rejected": -0.11031049489974976,
+      "sft_loss": 0.4993307590484619,
+      "step": 3235
+    },
+    {
+      "epoch": 2.592,
+      "grad_norm": 5.892839113416792,
+      "learning_rate": 2.764590667717562e-07,
+      "logits/chosen": -0.47418349981307983,
+      "logits/rejected": -1.210965871810913,
+      "logps/chosen": -0.5353994965553284,
+      "logps/rejected": -1.8738822937011719,
+      "loss": 0.4341,
+      "odds_ratio_loss": 0.2881673276424408,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.02676997520029545,
+      "rewards/margins": 0.06692413985729218,
+      "rewards/rejected": -0.09369411319494247,
+      "sft_loss": 0.5353994965553284,
+      "step": 3240
+    },
+    {
+      "epoch": 2.596,
+      "grad_norm": 4.975415723996069,
+      "learning_rate": 2.7116459286195887e-07,
+      "logits/chosen": -1.0550581216812134,
+      "logits/rejected": -1.0994212627410889,
+      "logps/chosen": -0.29386892914772034,
+      "logps/rejected": -1.3806931972503662,
+      "loss": 0.4509,
+      "odds_ratio_loss": 0.1265828013420105,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.014693446457386017,
+      "rewards/margins": 0.054341208189725876,
+      "rewards/rejected": -0.0690346509218216,
+      "sft_loss": 0.29386892914772034,
+      "step": 3245
+    },
+    {
+      "epoch": 2.6,
+      "grad_norm": 4.814970729318306,
+      "learning_rate": 2.6591839919146963e-07,
+      "logits/chosen": -0.7558291554450989,
+      "logits/rejected": -0.8831847906112671,
+      "logps/chosen": -0.4392101764678955,
+      "logps/rejected": -1.304246187210083,
+      "loss": 0.494,
+      "odds_ratio_loss": 0.24885766208171844,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.021960508078336716,
+      "rewards/margins": 0.04325180500745773,
+      "rewards/rejected": -0.06521230936050415,
+      "sft_loss": 0.4392101764678955,
+      "step": 3250
+    },
+    {
+      "epoch": 2.604,
+      "grad_norm": 8.307336904912425,
+      "learning_rate": 2.6072059940146775e-07,
+      "logits/chosen": -1.1570188999176025,
+      "logits/rejected": -0.8211970329284668,
+      "logps/chosen": -0.4389687180519104,
+      "logps/rejected": -1.5756551027297974,
+      "loss": 0.4346,
+      "odds_ratio_loss": 0.1567719727754593,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.0219484381377697,
+      "rewards/margins": 0.05683432146906853,
+      "rewards/rejected": -0.07878275215625763,
+      "sft_loss": 0.4389687180519104,
+      "step": 3255
+    },
+    {
+      "epoch": 2.608,
+      "grad_norm": 5.842083846009801,
+      "learning_rate": 2.555713060848433e-07,
+      "logits/chosen": -0.9055282473564148,
+      "logits/rejected": -1.0206440687179565,
+      "logps/chosen": -0.6034643650054932,
+      "logps/rejected": -1.3517249822616577,
+      "loss": 0.4756,
+      "odds_ratio_loss": 0.37321245670318604,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.030173221603035927,
+      "rewards/margins": 0.037413034588098526,
+      "rewards/rejected": -0.067586250603199,
+      "sft_loss": 0.6034643650054932,
+      "step": 3260
+    },
+    {
+      "epoch": 2.612,
+      "grad_norm": 26.57275268189309,
+      "learning_rate": 2.504706307837551e-07,
+      "logits/chosen": -0.9575685262680054,
+      "logits/rejected": -1.1905924081802368,
+      "logps/chosen": -0.43124690651893616,
+      "logps/rejected": -1.5143071413040161,
+      "loss": 0.4799,
+      "odds_ratio_loss": 0.24718472361564636,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.021562347188591957,
+      "rewards/margins": 0.054153017699718475,
+      "rewards/rejected": -0.07571535557508469,
+      "sft_loss": 0.43124690651893616,
+      "step": 3265
+    },
+    {
+      "epoch": 2.616,
+      "grad_norm": 12.548046244131415,
+      "learning_rate": 2.454186839872158e-07,
+      "logits/chosen": -0.8983935117721558,
+      "logits/rejected": -0.9400730133056641,
+      "logps/chosen": -0.5530645251274109,
+      "logps/rejected": -1.652138113975525,
+      "loss": 0.4669,
+      "odds_ratio_loss": 0.21761877834796906,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.027653228491544724,
+      "rewards/margins": 0.054953683167696,
+      "rewards/rejected": -0.08260690420866013,
+      "sft_loss": 0.5530645251274109,
+      "step": 3270
+    },
+    {
+      "epoch": 2.62,
+      "grad_norm": 5.389650137843442,
+      "learning_rate": 2.404155751286988e-07,
+      "logits/chosen": -1.0610014200210571,
+      "logits/rejected": -1.133551836013794,
+      "logps/chosen": -0.39243531227111816,
+      "logps/rejected": -1.79599130153656,
+      "loss": 0.4603,
+      "odds_ratio_loss": 0.2275979220867157,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.01962176337838173,
+      "rewards/margins": 0.07017780840396881,
+      "rewards/rejected": -0.08979956060647964,
+      "sft_loss": 0.39243531227111816,
+      "step": 3275
+    },
+    {
+      "epoch": 2.624,
+      "grad_norm": 5.664088641927079,
+      "learning_rate": 2.3546141258376786e-07,
+      "logits/chosen": -0.7254642248153687,
+      "logits/rejected": -0.6703656911849976,
+      "logps/chosen": -0.3163744807243347,
+      "logps/rejected": -1.2279107570648193,
+      "loss": 0.5009,
+      "odds_ratio_loss": 0.15358732640743256,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.015818724408745766,
+      "rewards/margins": 0.04557682201266289,
+      "rewards/rejected": -0.0613955482840538,
+      "sft_loss": 0.3163744807243347,
+      "step": 3280
+    },
+    {
+      "epoch": 2.628,
+      "grad_norm": 4.6344681245327495,
+      "learning_rate": 2.3055630366772857e-07,
+      "logits/chosen": -0.7210429906845093,
+      "logits/rejected": -0.7038585543632507,
+      "logps/chosen": -0.45490559935569763,
+      "logps/rejected": -1.3154146671295166,
+      "loss": 0.5243,
+      "odds_ratio_loss": 0.2467753142118454,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.022745277732610703,
+      "rewards/margins": 0.04302545264363289,
+      "rewards/rejected": -0.06577073037624359,
+      "sft_loss": 0.45490559935569763,
+      "step": 3285
+    },
+    {
+      "epoch": 2.632,
+      "grad_norm": 4.8837015178641705,
+      "learning_rate": 2.257003546333042e-07,
+      "logits/chosen": -0.9644176363945007,
+      "logits/rejected": -1.0574061870574951,
+      "logps/chosen": -0.4861271381378174,
+      "logps/rejected": -1.230362892150879,
+      "loss": 0.506,
+      "odds_ratio_loss": 0.223617285490036,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02430635876953602,
+      "rewards/margins": 0.03721178323030472,
+      "rewards/rejected": -0.061518143862485886,
+      "sft_loss": 0.4861271381378174,
+      "step": 3290
+    },
+    {
+      "epoch": 2.636,
+      "grad_norm": 4.840338076295951,
+      "learning_rate": 2.208936706683351e-07,
+      "logits/chosen": -0.747097909450531,
+      "logits/rejected": -0.7051985859870911,
+      "logps/chosen": -0.5597037076950073,
+      "logps/rejected": -1.4610236883163452,
+      "loss": 0.4651,
+      "odds_ratio_loss": 0.3216997981071472,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.027985185384750366,
+      "rewards/margins": 0.045065999031066895,
+      "rewards/rejected": -0.07305117696523666,
+      "sft_loss": 0.5597037076950073,
+      "step": 3295
+    },
+    {
+      "epoch": 2.64,
+      "grad_norm": 7.554990598127512,
+      "learning_rate": 2.1613635589349756e-07,
+      "logits/chosen": -0.8587006330490112,
+      "logits/rejected": -0.9386578798294067,
+      "logps/chosen": -0.4001200199127197,
+      "logps/rejected": -1.5990378856658936,
+      "loss": 0.4853,
+      "odds_ratio_loss": 0.2407374083995819,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.020006002858281136,
+      "rewards/margins": 0.05994589254260063,
+      "rewards/rejected": -0.07995189726352692,
+      "sft_loss": 0.4001200199127197,
+      "step": 3300
+    },
+    {
+      "epoch": 2.644,
+      "grad_norm": 9.49274069461471,
+      "learning_rate": 2.1142851336005244e-07,
+      "logits/chosen": -1.1812444925308228,
+      "logits/rejected": -1.0167138576507568,
+      "logps/chosen": -0.39482712745666504,
+      "logps/rejected": -1.4087440967559814,
+      "loss": 0.4741,
+      "odds_ratio_loss": 0.14712780714035034,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.019741356372833252,
+      "rewards/margins": 0.05069585517048836,
+      "rewards/rejected": -0.07043720781803131,
+      "sft_loss": 0.39482712745666504,
+      "step": 3305
+    },
+    {
+      "epoch": 2.648,
+      "grad_norm": 6.995561448070307,
+      "learning_rate": 2.0677024504760752e-07,
+      "logits/chosen": -1.0624709129333496,
+      "logits/rejected": -1.4392796754837036,
+      "logps/chosen": -0.5113095641136169,
+      "logps/rejected": -1.4682908058166504,
+      "loss": 0.4876,
+      "odds_ratio_loss": 0.37565088272094727,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.025565480813384056,
+      "rewards/margins": 0.04784905165433884,
+      "rewards/rejected": -0.07341454178094864,
+      "sft_loss": 0.5113095641136169,
+      "step": 3310
+    },
+    {
+      "epoch": 2.652,
+      "grad_norm": 7.478983570395769,
+      "learning_rate": 2.0216165186191406e-07,
+      "logits/chosen": -0.7986949682235718,
+      "logits/rejected": -1.3443512916564941,
+      "logps/chosen": -0.6050113439559937,
+      "logps/rejected": -1.6553246974945068,
+      "loss": 0.4864,
+      "odds_ratio_loss": 0.1889444887638092,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.030250567942857742,
+      "rewards/margins": 0.052515674382448196,
+      "rewards/rejected": -0.08276623487472534,
+      "sft_loss": 0.6050113439559937,
+      "step": 3315
+    },
+    {
+      "epoch": 2.656,
+      "grad_norm": 5.628793749435426,
+      "learning_rate": 1.9760283363267684e-07,
+      "logits/chosen": -0.96771240234375,
+      "logits/rejected": -0.6631850004196167,
+      "logps/chosen": -0.5663673877716064,
+      "logps/rejected": -1.225140929222107,
+      "loss": 0.5049,
+      "odds_ratio_loss": 0.3630017340183258,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.028318369761109352,
+      "rewards/margins": 0.03293868154287338,
+      "rewards/rejected": -0.06125704571604729,
+      "sft_loss": 0.5663673877716064,
+      "step": 3320
+    },
+    {
+      "epoch": 2.66,
+      "grad_norm": 4.202178749975983,
+      "learning_rate": 1.9309388911139427e-07,
+      "logits/chosen": -0.8587250709533691,
+      "logits/rejected": -1.0437920093536377,
+      "logps/chosen": -0.5410333871841431,
+      "logps/rejected": -1.3033370971679688,
+      "loss": 0.4574,
+      "odds_ratio_loss": 0.25340771675109863,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.027051672339439392,
+      "rewards/margins": 0.038115184754133224,
+      "rewards/rejected": -0.06516685336828232,
+      "sft_loss": 0.5410333871841431,
+      "step": 3325
+    },
+    {
+      "epoch": 2.664,
+      "grad_norm": 4.996160979584086,
+      "learning_rate": 1.8863491596921745e-07,
+      "logits/chosen": -0.9663417935371399,
+      "logits/rejected": -1.0380603075027466,
+      "logps/chosen": -0.42533501982688904,
+      "logps/rejected": -1.5897793769836426,
+      "loss": 0.4536,
+      "odds_ratio_loss": 0.1942438781261444,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.0212667528539896,
+      "rewards/margins": 0.058222223073244095,
+      "rewards/rejected": -0.07948897033929825,
+      "sft_loss": 0.42533501982688904,
+      "step": 3330
+    },
+    {
+      "epoch": 2.668,
+      "grad_norm": 5.226973499413544,
+      "learning_rate": 1.8422601079483516e-07,
+      "logits/chosen": -0.8443425297737122,
+      "logits/rejected": -1.0523451566696167,
+      "logps/chosen": -0.6304419636726379,
+      "logps/rejected": -1.6065928936004639,
+      "loss": 0.4965,
+      "odds_ratio_loss": 0.25557130575180054,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03152209892868996,
+      "rewards/margins": 0.048807550221681595,
+      "rewards/rejected": -0.08032964915037155,
+      "sft_loss": 0.6304419636726379,
+      "step": 3335
+    },
+    {
+      "epoch": 2.672,
+      "grad_norm": 5.151887544991019,
+      "learning_rate": 1.798672690923828e-07,
+      "logits/chosen": -0.4008614122867584,
+      "logits/rejected": -1.028642177581787,
+      "logps/chosen": -0.4695549011230469,
+      "logps/rejected": -1.3987529277801514,
+      "loss": 0.4852,
+      "odds_ratio_loss": 0.2155243456363678,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.023477744311094284,
+      "rewards/margins": 0.046459902077913284,
+      "rewards/rejected": -0.06993765383958817,
+      "sft_loss": 0.4695549011230469,
+      "step": 3340
+    },
+    {
+      "epoch": 2.676,
+      "grad_norm": 4.960271872554221,
+      "learning_rate": 1.7555878527937164e-07,
+      "logits/chosen": -1.132819414138794,
+      "logits/rejected": -1.2651735544204712,
+      "logps/chosen": -0.2899089753627777,
+      "logps/rejected": -1.4967478513717651,
+      "loss": 0.4768,
+      "odds_ratio_loss": 0.12462642043828964,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.014495449140667915,
+      "rewards/margins": 0.06034195423126221,
+      "rewards/rejected": -0.07483740150928497,
+      "sft_loss": 0.2899089753627777,
+      "step": 3345
+    },
+    {
+      "epoch": 2.68,
+      "grad_norm": 5.252700869191946,
+      "learning_rate": 1.713006526846439e-07,
+      "logits/chosen": -1.0679802894592285,
+      "logits/rejected": -1.2805407047271729,
+      "logps/chosen": -0.6222777366638184,
+      "logps/rejected": -1.4826539754867554,
+      "loss": 0.4673,
+      "odds_ratio_loss": 0.23923306167125702,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.031113887205719948,
+      "rewards/margins": 0.04301881417632103,
+      "rewards/rejected": -0.07413269579410553,
+      "sft_loss": 0.6222777366638184,
+      "step": 3350
+    },
+    {
+      "epoch": 2.684,
+      "grad_norm": 6.088871700564241,
+      "learning_rate": 1.6709296354635335e-07,
+      "logits/chosen": -0.7418292760848999,
+      "logits/rejected": -1.1808269023895264,
+      "logps/chosen": -0.7619951963424683,
+      "logps/rejected": -1.662157654762268,
+      "loss": 0.514,
+      "odds_ratio_loss": 0.3287936747074127,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.038099758327007294,
+      "rewards/margins": 0.04500812292098999,
+      "rewards/rejected": -0.08310787379741669,
+      "sft_loss": 0.7619951963424683,
+      "step": 3355
+    },
+    {
+      "epoch": 2.6879999999999997,
+      "grad_norm": 4.838274349890116,
+      "learning_rate": 1.629358090099639e-07,
+      "logits/chosen": -1.0373485088348389,
+      "logits/rejected": -1.0439493656158447,
+      "logps/chosen": -0.42322856187820435,
+      "logps/rejected": -1.5781844854354858,
+      "loss": 0.4928,
+      "odds_ratio_loss": 0.1596982479095459,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.021161429584026337,
+      "rewards/margins": 0.05774780362844467,
+      "rewards/rejected": -0.07890923321247101,
+      "sft_loss": 0.42322856187820435,
+      "step": 3360
+    },
+    {
+      "epoch": 2.692,
+      "grad_norm": 4.434071150396496,
+      "learning_rate": 1.5882927912627772e-07,
+      "logits/chosen": -0.8480218648910522,
+      "logits/rejected": -0.9117224812507629,
+      "logps/chosen": -0.4945312440395355,
+      "logps/rejected": -1.384734034538269,
+      "loss": 0.4637,
+      "odds_ratio_loss": 0.23109391331672668,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.024726565927267075,
+      "rewards/margins": 0.0445101372897625,
+      "rewards/rejected": -0.06923670321702957,
+      "sft_loss": 0.4945312440395355,
+      "step": 3365
+    },
+    {
+      "epoch": 2.6959999999999997,
+      "grad_norm": 5.347217671653992,
+      "learning_rate": 1.5477346284948292e-07,
+      "logits/chosen": -1.1148836612701416,
+      "logits/rejected": -1.0766358375549316,
+      "logps/chosen": -0.2054683268070221,
+      "logps/rejected": -1.9612598419189453,
+      "loss": 0.352,
+      "odds_ratio_loss": 0.04080657660961151,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.010273417457938194,
+      "rewards/margins": 0.08778958767652512,
+      "rewards/rejected": -0.09806300699710846,
+      "sft_loss": 0.2054683268070221,
+      "step": 3370
+    },
+    {
+      "epoch": 2.7,
+      "grad_norm": 7.112953927848122,
+      "learning_rate": 1.507684480352292e-07,
+      "logits/chosen": -0.86543208360672,
+      "logits/rejected": -1.3143550157546997,
+      "logps/chosen": -0.7909175157546997,
+      "logps/rejected": -1.7691437005996704,
+      "loss": 0.4695,
+      "odds_ratio_loss": 0.23235614597797394,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03954587131738663,
+      "rewards/margins": 0.04891129583120346,
+      "rewards/rejected": -0.08845716714859009,
+      "sft_loss": 0.7909175157546997,
+      "step": 3375
+    },
+    {
+      "epoch": 2.7039999999999997,
+      "grad_norm": 7.9988782226086155,
+      "learning_rate": 1.4681432143872133e-07,
+      "logits/chosen": -0.6214197874069214,
+      "logits/rejected": -1.0236066579818726,
+      "logps/chosen": -0.5888375639915466,
+      "logps/rejected": -1.7203214168548584,
+      "loss": 0.4666,
+      "odds_ratio_loss": 0.1753186732530594,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02944187819957733,
+      "rewards/margins": 0.056574203073978424,
+      "rewards/rejected": -0.08601607382297516,
+      "sft_loss": 0.5888375639915466,
+      "step": 3380
+    },
+    {
+      "epoch": 2.708,
+      "grad_norm": 3.7336969896252983,
+      "learning_rate": 1.4291116871284205e-07,
+      "logits/chosen": -1.00608229637146,
+      "logits/rejected": -1.003975510597229,
+      "logps/chosen": -0.43859434127807617,
+      "logps/rejected": -1.8750841617584229,
+      "loss": 0.4176,
+      "odds_ratio_loss": 0.21985983848571777,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.02192971669137478,
+      "rewards/margins": 0.07182449102401733,
+      "rewards/rejected": -0.09375420212745667,
+      "sft_loss": 0.43859434127807617,
+      "step": 3385
+    },
+    {
+      "epoch": 2.7119999999999997,
+      "grad_norm": 9.218065040509618,
+      "learning_rate": 1.3905907440629752e-07,
+      "logits/chosen": -0.9918481111526489,
+      "logits/rejected": -1.3855509757995605,
+      "logps/chosen": -0.3834057152271271,
+      "logps/rejected": -1.7054436206817627,
+      "loss": 0.4777,
+      "odds_ratio_loss": 0.16741231083869934,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.019170286133885384,
+      "rewards/margins": 0.06610190868377686,
+      "rewards/rejected": -0.08527218550443649,
+      "sft_loss": 0.3834057152271271,
+      "step": 3390
+    },
+    {
+      "epoch": 2.716,
+      "grad_norm": 4.274440200102076,
+      "learning_rate": 1.352581219617824e-07,
+      "logits/chosen": -0.6903150677680969,
+      "logits/rejected": -1.1200445890426636,
+      "logps/chosen": -0.3996601402759552,
+      "logps/rejected": -1.8368583917617798,
+      "loss": 0.4716,
+      "odds_ratio_loss": 0.13882999122142792,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.01998300477862358,
+      "rewards/margins": 0.07185991108417511,
+      "rewards/rejected": -0.09184291958808899,
+      "sft_loss": 0.3996601402759552,
+      "step": 3395
+    },
+    {
+      "epoch": 2.7199999999999998,
+      "grad_norm": 4.630834348255805,
+      "learning_rate": 1.31508393714177e-07,
+      "logits/chosen": -1.1776244640350342,
+      "logits/rejected": -0.7361685037612915,
+      "logps/chosen": -0.2189209908246994,
+      "logps/rejected": -1.2366139888763428,
+      "loss": 0.3637,
+      "odds_ratio_loss": 0.09817551076412201,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.010946051217615604,
+      "rewards/margins": 0.050884656608104706,
+      "rewards/rejected": -0.061830706894397736,
+      "sft_loss": 0.2189209908246994,
+      "step": 3400
+    },
+    {
+      "epoch": 2.724,
+      "grad_norm": 5.857849838216993,
+      "learning_rate": 1.278099708887587e-07,
+      "logits/chosen": -0.8234980702400208,
+      "logits/rejected": -0.843535304069519,
+      "logps/chosen": -0.36947911977767944,
+      "logps/rejected": -1.4612740278244019,
+      "loss": 0.475,
+      "odds_ratio_loss": 0.1974354237318039,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.018473956733942032,
+      "rewards/margins": 0.054589755833148956,
+      "rewards/rejected": -0.07306370884180069,
+      "sft_loss": 0.36947911977767944,
+      "step": 3405
+    },
+    {
+      "epoch": 2.7279999999999998,
+      "grad_norm": 6.068229298253824,
+      "learning_rate": 1.241629335994471e-07,
+      "logits/chosen": -0.7435353994369507,
+      "logits/rejected": -1.0612701177597046,
+      "logps/chosen": -0.5406766533851624,
+      "logps/rejected": -1.4909696578979492,
+      "loss": 0.4116,
+      "odds_ratio_loss": 0.30309200286865234,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.027033831924200058,
+      "rewards/margins": 0.04751463979482651,
+      "rewards/rejected": -0.07454848289489746,
+      "sft_loss": 0.5406766533851624,
+      "step": 3410
+    },
+    {
+      "epoch": 2.732,
+      "grad_norm": 8.977086173159117,
+      "learning_rate": 1.2056736084706588e-07,
+      "logits/chosen": -0.9622576832771301,
+      "logits/rejected": -1.2189757823944092,
+      "logps/chosen": -0.43772053718566895,
+      "logps/rejected": -1.773707628250122,
+      "loss": 0.4551,
+      "odds_ratio_loss": 0.18512067198753357,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.021886030212044716,
+      "rewards/margins": 0.0667993575334549,
+      "rewards/rejected": -0.08868537843227386,
+      "sft_loss": 0.43772053718566895,
+      "step": 3415
+    },
+    {
+      "epoch": 2.7359999999999998,
+      "grad_norm": 5.513247634423349,
+      "learning_rate": 1.1702333051763271e-07,
+      "logits/chosen": -0.6844684481620789,
+      "logits/rejected": -1.1717528104782104,
+      "logps/chosen": -0.5392543077468872,
+      "logps/rejected": -1.2498254776000977,
+      "loss": 0.4714,
+      "odds_ratio_loss": 0.30782368779182434,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.02696271799504757,
+      "rewards/margins": 0.03552855923771858,
+      "rewards/rejected": -0.062491275370121,
+      "sft_loss": 0.5392543077468872,
+      "step": 3420
+    },
+    {
+      "epoch": 2.74,
+      "grad_norm": 3.9149151857473714,
+      "learning_rate": 1.1353091938067024e-07,
+      "logits/chosen": -0.6818164587020874,
+      "logits/rejected": -0.7006502151489258,
+      "logps/chosen": -0.4631117880344391,
+      "logps/rejected": -1.6541814804077148,
+      "loss": 0.434,
+      "odds_ratio_loss": 0.186067134141922,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.023155588656663895,
+      "rewards/margins": 0.059553492814302444,
+      "rewards/rejected": -0.08270907402038574,
+      "sft_loss": 0.4631117880344391,
+      "step": 3425
+    },
+    {
+      "epoch": 2.7439999999999998,
+      "grad_norm": 5.1831375438532135,
+      "learning_rate": 1.1009020308754587e-07,
+      "logits/chosen": -0.7273832559585571,
+      "logits/rejected": -0.6493693590164185,
+      "logps/chosen": -0.17281028628349304,
+      "logps/rejected": -1.4289897680282593,
+      "loss": 0.4975,
+      "odds_ratio_loss": 0.1179138645529747,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.008640513755381107,
+      "rewards/margins": 0.06280897557735443,
+      "rewards/rejected": -0.07144948095083237,
+      "sft_loss": 0.17281028628349304,
+      "step": 3430
+    },
+    {
+      "epoch": 2.748,
+      "grad_norm": 4.973491478841313,
+      "learning_rate": 1.067012561698319e-07,
+      "logits/chosen": -1.207350730895996,
+      "logits/rejected": -0.8244549632072449,
+      "logps/chosen": -0.5802727937698364,
+      "logps/rejected": -1.568420648574829,
+      "loss": 0.4946,
+      "odds_ratio_loss": 0.22998332977294922,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.029013637453317642,
+      "rewards/margins": 0.04940740019083023,
+      "rewards/rejected": -0.07842103391885757,
+      "sft_loss": 0.5802727937698364,
+      "step": 3435
+    },
+    {
+      "epoch": 2.752,
+      "grad_norm": 7.538216183918059,
+      "learning_rate": 1.0336415203768962e-07,
+      "logits/chosen": -0.7276453971862793,
+      "logits/rejected": -1.1800649166107178,
+      "logps/chosen": -0.268561452627182,
+      "logps/rejected": -1.8346469402313232,
+      "loss": 0.4621,
+      "odds_ratio_loss": 0.10011246055364609,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.013428074307739735,
+      "rewards/margins": 0.07830427587032318,
+      "rewards/rejected": -0.09173235297203064,
+      "sft_loss": 0.268561452627182,
+      "step": 3440
+    },
+    {
+      "epoch": 2.7560000000000002,
+      "grad_norm": 4.385995959035907,
+      "learning_rate": 1.0007896297828113e-07,
+      "logits/chosen": -0.9866234660148621,
+      "logits/rejected": -0.8177453875541687,
+      "logps/chosen": -0.2839185297489166,
+      "logps/rejected": -1.2978515625,
+      "loss": 0.4541,
+      "odds_ratio_loss": 0.13113179802894592,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.014195924624800682,
+      "rewards/margins": 0.05069665238261223,
+      "rewards/rejected": -0.06489257514476776,
+      "sft_loss": 0.2839185297489166,
+      "step": 3445
+    },
+    {
+      "epoch": 2.76,
+      "grad_norm": 5.040991958957209,
+      "learning_rate": 9.684576015420277e-08,
+      "logits/chosen": -1.0320391654968262,
+      "logits/rejected": -0.9190491437911987,
+      "logps/chosen": -0.49723321199417114,
+      "logps/rejected": -2.3937244415283203,
+      "loss": 0.4407,
+      "odds_ratio_loss": 0.21446876227855682,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.024861661717295647,
+      "rewards/margins": 0.09482455253601074,
+      "rewards/rejected": -0.11968620866537094,
+      "sft_loss": 0.49723321199417114,
+      "step": 3450
+    },
+    {
+      "epoch": 2.7640000000000002,
+      "grad_norm": 6.834196455601754,
+      "learning_rate": 9.36646136019434e-08,
+      "logits/chosen": -1.0055029392242432,
+      "logits/rejected": -0.663110077381134,
+      "logps/chosen": -0.2397165298461914,
+      "logps/rejected": -1.2084038257598877,
+      "loss": 0.4345,
+      "odds_ratio_loss": 0.11389895528554916,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.011985826306045055,
+      "rewards/margins": 0.04843435436487198,
+      "rewards/rejected": -0.06042018532752991,
+      "sft_loss": 0.2397165298461914,
+      "step": 3455
+    },
+    {
+      "epoch": 2.768,
+      "grad_norm": 6.622662942386174,
+      "learning_rate": 9.053559223036746e-08,
+      "logits/chosen": -0.5032299757003784,
+      "logits/rejected": -0.8415955305099487,
+      "logps/chosen": -0.7557247877120972,
+      "logps/rejected": -1.9935903549194336,
+      "loss": 0.4693,
+      "odds_ratio_loss": 0.29048803448677063,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03778623789548874,
+      "rewards/margins": 0.0618932731449604,
+      "rewards/rejected": -0.09967950731515884,
+      "sft_loss": 0.7557247877120972,
+      "step": 3460
+    },
+    {
+      "epoch": 2.7720000000000002,
+      "grad_norm": 5.779744351491692,
+      "learning_rate": 8.745876381922147e-08,
+      "logits/chosen": -0.9570730924606323,
+      "logits/rejected": -1.0343133211135864,
+      "logps/chosen": -0.3262855112552643,
+      "logps/rejected": -1.9049714803695679,
+      "loss": 0.4068,
+      "odds_ratio_loss": 0.1138528436422348,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.016314277425408363,
+      "rewards/margins": 0.07893429696559906,
+      "rewards/rejected": -0.09524857252836227,
+      "sft_loss": 0.3262855112552643,
+      "step": 3465
+    },
+    {
+      "epoch": 2.776,
+      "grad_norm": 6.075626717992663,
+      "learning_rate": 8.44341950176683e-08,
+      "logits/chosen": -0.8139813542366028,
+      "logits/rejected": -1.261385202407837,
+      "logps/chosen": -0.5494598150253296,
+      "logps/rejected": -1.436342477798462,
+      "loss": 0.474,
+      "odds_ratio_loss": 0.2713177800178528,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02747299149632454,
+      "rewards/margins": 0.044344138354063034,
+      "rewards/rejected": -0.07181712985038757,
+      "sft_loss": 0.5494598150253296,
+      "step": 3470
+    },
+    {
+      "epoch": 2.7800000000000002,
+      "grad_norm": 6.027330370816976,
+      "learning_rate": 8.146195134284052e-08,
+      "logits/chosen": -1.251039981842041,
+      "logits/rejected": -1.2202528715133667,
+      "logps/chosen": -0.4107237458229065,
+      "logps/rejected": -1.591752052307129,
+      "loss": 0.4984,
+      "odds_ratio_loss": 0.1679602563381195,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.020536188036203384,
+      "rewards/margins": 0.05905142426490784,
+      "rewards/rejected": -0.07958760112524033,
+      "sft_loss": 0.4107237458229065,
+      "step": 3475
+    },
+    {
+      "epoch": 2.784,
+      "grad_norm": 5.9048587954628315,
+      "learning_rate": 7.854209717842231e-08,
+      "logits/chosen": -0.8869150280952454,
+      "logits/rejected": -1.0556492805480957,
+      "logps/chosen": -0.4128708243370056,
+      "logps/rejected": -1.244500756263733,
+      "loss": 0.4525,
+      "odds_ratio_loss": 0.25618261098861694,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.02064354531466961,
+      "rewards/margins": 0.041581496596336365,
+      "rewards/rejected": -0.06222504377365112,
+      "sft_loss": 0.4128708243370056,
+      "step": 3480
+    },
+    {
+      "epoch": 2.7880000000000003,
+      "grad_norm": 5.609907193598404,
+      "learning_rate": 7.567469577325598e-08,
+      "logits/chosen": -1.0691337585449219,
+      "logits/rejected": -1.0011084079742432,
+      "logps/chosen": -0.2602657973766327,
+      "logps/rejected": -1.5600097179412842,
+      "loss": 0.4698,
+      "odds_ratio_loss": 0.10053672641515732,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.013013291172683239,
+      "rewards/margins": 0.0649871900677681,
+      "rewards/rejected": -0.07800048589706421,
+      "sft_loss": 0.2602657973766327,
+      "step": 3485
+    },
+    {
+      "epoch": 2.792,
+      "grad_norm": 6.288586916197363,
+      "learning_rate": 7.285980923996989e-08,
+      "logits/chosen": -0.814477801322937,
+      "logits/rejected": -1.0002424716949463,
+      "logps/chosen": -0.39793896675109863,
+      "logps/rejected": -1.5965421199798584,
+      "loss": 0.4963,
+      "odds_ratio_loss": 0.15374401211738586,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.01989694871008396,
+      "rewards/margins": 0.05993015691637993,
+      "rewards/rejected": -0.07982710003852844,
+      "sft_loss": 0.39793896675109863,
+      "step": 3490
+    },
+    {
+      "epoch": 2.7960000000000003,
+      "grad_norm": 7.196669432259674,
+      "learning_rate": 7.009749855363457e-08,
+      "logits/chosen": -0.8080123066902161,
+      "logits/rejected": -0.828715980052948,
+      "logps/chosen": -0.4427156448364258,
+      "logps/rejected": -1.4885919094085693,
+      "loss": 0.467,
+      "odds_ratio_loss": 0.21041274070739746,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.02213578298687935,
+      "rewards/margins": 0.0522938147187233,
+      "rewards/rejected": -0.07442959398031235,
+      "sft_loss": 0.4427156448364258,
+      "step": 3495
+    },
+    {
+      "epoch": 2.8,
+      "grad_norm": 6.610897880824554,
+      "learning_rate": 6.738782355044048e-08,
+      "logits/chosen": -0.7845413088798523,
+      "logits/rejected": -1.2522486448287964,
+      "logps/chosen": -0.19464164972305298,
+      "logps/rejected": -1.954445242881775,
+      "loss": 0.5121,
+      "odds_ratio_loss": 0.07869584858417511,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.009732084348797798,
+      "rewards/margins": 0.0879901796579361,
+      "rewards/rejected": -0.09772225469350815,
+      "sft_loss": 0.19464164972305298,
+      "step": 3500
+    },
+    {
+      "epoch": 2.8040000000000003,
+      "grad_norm": 5.994838172411329,
+      "learning_rate": 6.47308429264032e-08,
+      "logits/chosen": -0.6243001818656921,
+      "logits/rejected": -0.9205516576766968,
+      "logps/chosen": -0.45419034361839294,
+      "logps/rejected": -1.346792221069336,
+      "loss": 0.4879,
+      "odds_ratio_loss": 0.21579177677631378,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.022709516808390617,
+      "rewards/margins": 0.04463009163737297,
+      "rewards/rejected": -0.06733961403369904,
+      "sft_loss": 0.45419034361839294,
+      "step": 3505
+    },
+    {
+      "epoch": 2.808,
+      "grad_norm": 5.18964559816438,
+      "learning_rate": 6.212661423609184e-08,
+      "logits/chosen": -0.7687469720840454,
+      "logits/rejected": -0.9978445172309875,
+      "logps/chosen": -0.38948315382003784,
+      "logps/rejected": -1.690354585647583,
+      "loss": 0.4909,
+      "odds_ratio_loss": 0.1466546207666397,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.01947415992617607,
+      "rewards/margins": 0.06504356116056442,
+      "rewards/rejected": -0.08451772481203079,
+      "sft_loss": 0.38948315382003784,
+      "step": 3510
+    },
+    {
+      "epoch": 2.8120000000000003,
+      "grad_norm": 4.278585362133975,
+      "learning_rate": 5.957519389138106e-08,
+      "logits/chosen": -0.8727335929870605,
+      "logits/rejected": -0.9041558504104614,
+      "logps/chosen": -0.5910320281982422,
+      "logps/rejected": -1.4024748802185059,
+      "loss": 0.468,
+      "odds_ratio_loss": 0.2644711434841156,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.02955160103738308,
+      "rewards/margins": 0.0405721440911293,
+      "rewards/rejected": -0.07012374699115753,
+      "sft_loss": 0.5910320281982422,
+      "step": 3515
+    },
+    {
+      "epoch": 2.816,
+      "grad_norm": 8.621278806090325,
+      "learning_rate": 5.707663716023021e-08,
+      "logits/chosen": -0.7765305042266846,
+      "logits/rejected": -1.2130156755447388,
+      "logps/chosen": -0.4113030433654785,
+      "logps/rejected": -1.911913275718689,
+      "loss": 0.4462,
+      "odds_ratio_loss": 0.16190369427204132,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.020565154030919075,
+      "rewards/margins": 0.07503052800893784,
+      "rewards/rejected": -0.09559567272663116,
+      "sft_loss": 0.4113030433654785,
+      "step": 3520
+    },
+    {
+      "epoch": 2.82,
+      "grad_norm": 5.895614935922102,
+      "learning_rate": 5.463099816548578e-08,
+      "logits/chosen": -0.8122976422309875,
+      "logits/rejected": -1.2246438264846802,
+      "logps/chosen": -0.6956444978713989,
+      "logps/rejected": -1.742977499961853,
+      "loss": 0.4596,
+      "odds_ratio_loss": 0.2761508822441101,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03478222340345383,
+      "rewards/margins": 0.05236666277050972,
+      "rewards/rejected": -0.08714888989925385,
+      "sft_loss": 0.6956444978713989,
+      "step": 3525
+    },
+    {
+      "epoch": 2.824,
+      "grad_norm": 5.257809415391807,
+      "learning_rate": 5.22383298837098e-08,
+      "logits/chosen": -1.0394823551177979,
+      "logits/rejected": -1.370062232017517,
+      "logps/chosen": -0.6062598824501038,
+      "logps/rejected": -1.899864912033081,
+      "loss": 0.5637,
+      "odds_ratio_loss": 0.30006757378578186,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.030312994495034218,
+      "rewards/margins": 0.06468025594949722,
+      "rewards/rejected": -0.0949932411313057,
+      "sft_loss": 0.6062598824501038,
+      "step": 3530
+    },
+    {
+      "epoch": 2.828,
+      "grad_norm": 5.075614233899885,
+      "learning_rate": 4.989868414403048e-08,
+      "logits/chosen": -0.7696380615234375,
+      "logits/rejected": -0.9930199384689331,
+      "logps/chosen": -0.5346277952194214,
+      "logps/rejected": -2.1930575370788574,
+      "loss": 0.5216,
+      "odds_ratio_loss": 0.2663384974002838,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.02673139050602913,
+      "rewards/margins": 0.08292149752378464,
+      "rewards/rejected": -0.10965289175510406,
+      "sft_loss": 0.5346277952194214,
+      "step": 3535
+    },
+    {
+      "epoch": 2.832,
+      "grad_norm": 4.417337663334965,
+      "learning_rate": 4.761211162702117e-08,
+      "logits/chosen": -1.0091744661331177,
+      "logits/rejected": -0.9187496900558472,
+      "logps/chosen": -0.3657459616661072,
+      "logps/rejected": -1.539928674697876,
+      "loss": 0.5018,
+      "odds_ratio_loss": 0.1328873336315155,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.018287301063537598,
+      "rewards/margins": 0.05870913341641426,
+      "rewards/rejected": -0.07699643075466156,
+      "sft_loss": 0.3657459616661072,
+      "step": 3540
+    },
+    {
+      "epoch": 2.836,
+      "grad_norm": 4.546971390558208,
+      "learning_rate": 4.537866186360207e-08,
+      "logits/chosen": -0.7591463327407837,
+      "logits/rejected": -0.7273364067077637,
+      "logps/chosen": -0.42804569005966187,
+      "logps/rejected": -1.2514833211898804,
+      "loss": 0.4635,
+      "odds_ratio_loss": 0.2270592749118805,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.02140228822827339,
+      "rewards/margins": 0.04117188602685928,
+      "rewards/rejected": -0.06257417052984238,
+      "sft_loss": 0.42804569005966187,
+      "step": 3545
+    },
+    {
+      "epoch": 2.84,
+      "grad_norm": 8.001000670131893,
+      "learning_rate": 4.319838323396691e-08,
+      "logits/chosen": -1.2595059871673584,
+      "logits/rejected": -0.9697495698928833,
+      "logps/chosen": -0.4693234860897064,
+      "logps/rejected": -1.4664093255996704,
+      "loss": 0.5123,
+      "odds_ratio_loss": 0.27086126804351807,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.02346617542207241,
+      "rewards/margins": 0.04985428601503372,
+      "rewards/rejected": -0.07332046329975128,
+      "sft_loss": 0.4693234860897064,
+      "step": 3550
+    },
+    {
+      "epoch": 2.844,
+      "grad_norm": 15.587957013391982,
+      "learning_rate": 4.1071322966535487e-08,
+      "logits/chosen": -0.996113657951355,
+      "logits/rejected": -0.6731228828430176,
+      "logps/chosen": -0.2613915801048279,
+      "logps/rejected": -1.2313883304595947,
+      "loss": 0.432,
+      "odds_ratio_loss": 0.12515100836753845,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.013069577515125275,
+      "rewards/margins": 0.04849984496831894,
+      "rewards/rejected": -0.061569422483444214,
+      "sft_loss": 0.2613915801048279,
+      "step": 3555
+    },
+    {
+      "epoch": 2.848,
+      "grad_norm": 7.371630683017223,
+      "learning_rate": 3.8997527136930004e-08,
+      "logits/chosen": -0.9078682065010071,
+      "logits/rejected": -0.7578974962234497,
+      "logps/chosen": -0.3508548438549042,
+      "logps/rejected": -1.5415148735046387,
+      "loss": 0.4796,
+      "odds_ratio_loss": 0.122395358979702,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.01754274033010006,
+      "rewards/margins": 0.059532999992370605,
+      "rewards/rejected": -0.07707574218511581,
+      "sft_loss": 0.3508548438549042,
+      "step": 3560
+    },
+    {
+      "epoch": 2.852,
+      "grad_norm": 8.0726795268241,
+      "learning_rate": 3.6977040666977546e-08,
+      "logits/chosen": -1.1534029245376587,
+      "logits/rejected": -1.2099335193634033,
+      "logps/chosen": -0.5775087475776672,
+      "logps/rejected": -1.39055597782135,
+      "loss": 0.4869,
+      "odds_ratio_loss": 0.2733613848686218,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.028875436633825302,
+      "rewards/margins": 0.04065236449241638,
+      "rewards/rejected": -0.06952779740095139,
+      "sft_loss": 0.5775087475776672,
+      "step": 3565
+    },
+    {
+      "epoch": 2.856,
+      "grad_norm": 11.24865948956023,
+      "learning_rate": 3.5009907323737826e-08,
+      "logits/chosen": -0.552277684211731,
+      "logits/rejected": -1.1681157350540161,
+      "logps/chosen": -0.6316530108451843,
+      "logps/rejected": -1.6660516262054443,
+      "loss": 0.5061,
+      "odds_ratio_loss": 0.2643664479255676,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03158264979720116,
+      "rewards/margins": 0.05171992629766464,
+      "rewards/rejected": -0.0833025798201561,
+      "sft_loss": 0.6316530108451843,
+      "step": 3570
+    },
+    {
+      "epoch": 2.86,
+      "grad_norm": 8.187980250866474,
+      "learning_rate": 3.309616971855195e-08,
+      "logits/chosen": -0.9434603452682495,
+      "logits/rejected": -1.0135250091552734,
+      "logps/chosen": -0.23478789627552032,
+      "logps/rejected": -2.0454790592193604,
+      "loss": 0.4638,
+      "odds_ratio_loss": 0.12218568474054337,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.011739394627511501,
+      "rewards/margins": 0.09053455293178558,
+      "rewards/rejected": -0.10227394104003906,
+      "sft_loss": 0.23478789627552032,
+      "step": 3575
+    },
+    {
+      "epoch": 2.864,
+      "grad_norm": 10.077811295456423,
+      "learning_rate": 3.1235869306123766e-08,
+      "logits/chosen": -1.0300368070602417,
+      "logits/rejected": -0.8315317034721375,
+      "logps/chosen": -0.5715829133987427,
+      "logps/rejected": -1.4090276956558228,
+      "loss": 0.4626,
+      "odds_ratio_loss": 0.3393372893333435,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.028579145669937134,
+      "rewards/margins": 0.04187224060297012,
+      "rewards/rejected": -0.07045139372348785,
+      "sft_loss": 0.5715829133987427,
+      "step": 3580
+    },
+    {
+      "epoch": 2.868,
+      "grad_norm": 6.609706143071862,
+      "learning_rate": 2.9429046383618042e-08,
+      "logits/chosen": -1.0845788717269897,
+      "logits/rejected": -1.40879487991333,
+      "logps/chosen": -0.467085063457489,
+      "logps/rejected": -1.8588714599609375,
+      "loss": 0.5266,
+      "odds_ratio_loss": 0.16859348118305206,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02335425093770027,
+      "rewards/margins": 0.06958932429552078,
+      "rewards/rejected": -0.09294357895851135,
+      "sft_loss": 0.467085063457489,
+      "step": 3585
+    },
+    {
+      "epoch": 2.872,
+      "grad_norm": 4.974574218409692,
+      "learning_rate": 2.767574008979007e-08,
+      "logits/chosen": -0.8864003419876099,
+      "logits/rejected": -1.0464935302734375,
+      "logps/chosen": -0.5476582050323486,
+      "logps/rejected": -1.9629188776016235,
+      "loss": 0.4826,
+      "odds_ratio_loss": 0.19431143999099731,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02738291397690773,
+      "rewards/margins": 0.07076303660869598,
+      "rewards/rejected": -0.09814594686031342,
+      "sft_loss": 0.5476582050323486,
+      "step": 3590
+    },
+    {
+      "epoch": 2.876,
+      "grad_norm": 5.391207506724364,
+      "learning_rate": 2.59759884041369e-08,
+      "logits/chosen": -0.8431358337402344,
+      "logits/rejected": -0.9085556864738464,
+      "logps/chosen": -0.26856502890586853,
+      "logps/rejected": -1.4795863628387451,
+      "loss": 0.442,
+      "odds_ratio_loss": 0.1016441136598587,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.01342825312167406,
+      "rewards/margins": 0.060551077127456665,
+      "rewards/rejected": -0.07397932559251785,
+      "sft_loss": 0.26856502890586853,
+      "step": 3595
+    },
+    {
+      "epoch": 2.88,
+      "grad_norm": 4.308262440272032,
+      "learning_rate": 2.4329828146074096e-08,
+      "logits/chosen": -0.7681323289871216,
+      "logits/rejected": -0.8165963292121887,
+      "logps/chosen": -0.3159729242324829,
+      "logps/rejected": -1.6058391332626343,
+      "loss": 0.4946,
+      "odds_ratio_loss": 0.14674882590770721,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.015798646956682205,
+      "rewards/margins": 0.06449331343173981,
+      "rewards/rejected": -0.08029197156429291,
+      "sft_loss": 0.3159729242324829,
+      "step": 3600
+    },
+    {
+      "epoch": 2.884,
+      "grad_norm": 4.854650051917074,
+      "learning_rate": 2.2737294974140013e-08,
+      "logits/chosen": -0.9245132207870483,
+      "logits/rejected": -0.9439831972122192,
+      "logps/chosen": -0.5168892741203308,
+      "logps/rejected": -1.4722706079483032,
+      "loss": 0.6122,
+      "odds_ratio_loss": 0.26270395517349243,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.02584446594119072,
+      "rewards/margins": 0.04776906222105026,
+      "rewards/rejected": -0.07361352443695068,
+      "sft_loss": 0.5168892741203308,
+      "step": 3605
+    },
+    {
+      "epoch": 2.888,
+      "grad_norm": 4.810264838940049,
+      "learning_rate": 2.1198423385220822e-08,
+      "logits/chosen": -0.5693929195404053,
+      "logits/rejected": -0.8137924075126648,
+      "logps/chosen": -0.5562976598739624,
+      "logps/rejected": -1.870003342628479,
+      "loss": 0.4924,
+      "odds_ratio_loss": 0.1856372058391571,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02781488373875618,
+      "rewards/margins": 0.06568528711795807,
+      "rewards/rejected": -0.09350016713142395,
+      "sft_loss": 0.5562976598739624,
+      "step": 3610
+    },
+    {
+      "epoch": 2.892,
+      "grad_norm": 4.590085289404893,
+      "learning_rate": 1.9713246713805588e-08,
+      "logits/chosen": -0.7041809558868408,
+      "logits/rejected": -0.819964587688446,
+      "logps/chosen": -0.5815272331237793,
+      "logps/rejected": -1.5759763717651367,
+      "loss": 0.4648,
+      "odds_ratio_loss": 0.21994754672050476,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.029076358303427696,
+      "rewards/margins": 0.04972246289253235,
+      "rewards/rejected": -0.0787988156080246,
+      "sft_loss": 0.5815272331237793,
+      "step": 3615
+    },
+    {
+      "epoch": 2.896,
+      "grad_norm": 5.1032023997222655,
+      "learning_rate": 1.82817971312621e-08,
+      "logits/chosen": -0.9272521138191223,
+      "logits/rejected": -0.969031810760498,
+      "logps/chosen": -0.3067367672920227,
+      "logps/rejected": -1.5883128643035889,
+      "loss": 0.466,
+      "odds_ratio_loss": 0.13033534586429596,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.01533683855086565,
+      "rewards/margins": 0.06407880038022995,
+      "rewards/rejected": -0.07941563427448273,
+      "sft_loss": 0.3067367672920227,
+      "step": 3620
+    },
+    {
+      "epoch": 2.9,
+      "grad_norm": 5.133104020591015,
+      "learning_rate": 1.6904105645142443e-08,
+      "logits/chosen": -0.6860765218734741,
+      "logits/rejected": -1.004093885421753,
+      "logps/chosen": -0.313940167427063,
+      "logps/rejected": -1.984702706336975,
+      "loss": 0.4767,
+      "odds_ratio_loss": 0.05924931913614273,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.01569700986146927,
+      "rewards/margins": 0.08353812992572784,
+      "rewards/rejected": -0.09923513978719711,
+      "sft_loss": 0.313940167427063,
+      "step": 3625
+    },
+    {
+      "epoch": 2.904,
+      "grad_norm": 6.289524397383519,
+      "learning_rate": 1.5580202098509078e-08,
+      "logits/chosen": -0.6269059181213379,
+      "logits/rejected": -0.887751579284668,
+      "logps/chosen": -0.6534531116485596,
+      "logps/rejected": -1.3601219654083252,
+      "loss": 0.5362,
+      "odds_ratio_loss": 0.3122999668121338,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03267265111207962,
+      "rewards/margins": 0.03533344715833664,
+      "rewards/rejected": -0.06800609827041626,
+      "sft_loss": 0.6534531116485596,
+      "step": 3630
+    },
+    {
+      "epoch": 2.908,
+      "grad_norm": 5.321061246172241,
+      "learning_rate": 1.4310115169289263e-08,
+      "logits/chosen": -0.7741699814796448,
+      "logits/rejected": -0.8883988261222839,
+      "logps/chosen": -0.3918963074684143,
+      "logps/rejected": -1.3303663730621338,
+      "loss": 0.4475,
+      "odds_ratio_loss": 0.18605419993400574,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.019594816491007805,
+      "rewards/margins": 0.04692351073026657,
+      "rewards/rejected": -0.06651832163333893,
+      "sft_loss": 0.3918963074684143,
+      "step": 3635
+    },
+    {
+      "epoch": 2.912,
+      "grad_norm": 5.614615596322008,
+      "learning_rate": 1.3093872369654148e-08,
+      "logits/chosen": -0.813896656036377,
+      "logits/rejected": -0.9681650400161743,
+      "logps/chosen": -0.318438321352005,
+      "logps/rejected": -1.9196224212646484,
+      "loss": 0.5048,
+      "odds_ratio_loss": 0.14587683975696564,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.01592191681265831,
+      "rewards/margins": 0.08005920052528381,
+      "rewards/rejected": -0.09598111361265182,
+      "sft_loss": 0.318438321352005,
+      "step": 3640
+    },
+    {
+      "epoch": 2.916,
+      "grad_norm": 6.909146903510477,
+      "learning_rate": 1.193150004542204e-08,
+      "logits/chosen": -0.6660367250442505,
+      "logits/rejected": -1.0153348445892334,
+      "logps/chosen": -0.46080127358436584,
+      "logps/rejected": -1.8787386417388916,
+      "loss": 0.5142,
+      "odds_ratio_loss": 0.22567503154277802,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.023040063679218292,
+      "rewards/margins": 0.07089685648679733,
+      "rewards/rejected": -0.09393692761659622,
+      "sft_loss": 0.46080127358436584,
+      "step": 3645
+    },
+    {
+      "epoch": 2.92,
+      "grad_norm": 7.790549829630172,
+      "learning_rate": 1.0823023375489128e-08,
+      "logits/chosen": -0.817104697227478,
+      "logits/rejected": -0.8277327418327332,
+      "logps/chosen": -0.6163122057914734,
+      "logps/rejected": -1.7266231775283813,
+      "loss": 0.4548,
+      "odds_ratio_loss": 0.251958429813385,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.03081561252474785,
+      "rewards/margins": 0.05551555007696152,
+      "rewards/rejected": -0.08633115142583847,
+      "sft_loss": 0.6163122057914734,
+      "step": 3650
+    },
+    {
+      "epoch": 2.924,
+      "grad_norm": 4.529447800983838,
+      "learning_rate": 9.76846637128187e-09,
+      "logits/chosen": -1.1566288471221924,
+      "logits/rejected": -1.1448895931243896,
+      "logps/chosen": -0.6995830535888672,
+      "logps/rejected": -1.5597145557403564,
+      "loss": 0.4972,
+      "odds_ratio_loss": 0.30369362235069275,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03497915342450142,
+      "rewards/margins": 0.043006572872400284,
+      "rewards/rejected": -0.0779857262969017,
+      "sft_loss": 0.6995830535888672,
+      "step": 3655
+    },
+    {
+      "epoch": 2.928,
+      "grad_norm": 5.857785314987744,
+      "learning_rate": 8.767851876239075e-09,
+      "logits/chosen": -0.5766550898551941,
+      "logits/rejected": -0.9048866033554077,
+      "logps/chosen": -0.43968862295150757,
+      "logps/rejected": -1.336573839187622,
+      "loss": 0.4655,
+      "odds_ratio_loss": 0.20616094768047333,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.021984433755278587,
+      "rewards/margins": 0.04484425485134125,
+      "rewards/rejected": -0.06682869046926498,
+      "sft_loss": 0.43968862295150757,
+      "step": 3660
+    },
+    {
+      "epoch": 2.932,
+      "grad_norm": 5.159988160646394,
+      "learning_rate": 7.821201565316184e-09,
+      "logits/chosen": -1.0310778617858887,
+      "logits/rejected": -0.7640330791473389,
+      "logps/chosen": -0.3315303921699524,
+      "logps/rejected": -1.6290994882583618,
+      "loss": 0.4037,
+      "odds_ratio_loss": 0.10144094377756119,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.01657651923596859,
+      "rewards/margins": 0.064878448843956,
+      "rewards/rejected": -0.08145496994256973,
+      "sft_loss": 0.3315303921699524,
+      "step": 3665
+    },
+    {
+      "epoch": 2.936,
+      "grad_norm": 18.981926903930404,
+      "learning_rate": 6.9285359445145366e-09,
+      "logits/chosen": -0.8311254382133484,
+      "logits/rejected": -0.9923003911972046,
+      "logps/chosen": -0.5326222777366638,
+      "logps/rejected": -1.6575853824615479,
+      "loss": 0.4691,
+      "odds_ratio_loss": 0.2388612926006317,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.02663111314177513,
+      "rewards/margins": 0.056248150765895844,
+      "rewards/rejected": -0.08287926018238068,
+      "sft_loss": 0.5326222777366638,
+      "step": 3670
+    },
+    {
+      "epoch": 2.94,
+      "grad_norm": 4.65539553385166,
+      "learning_rate": 6.089874350439507e-09,
+      "logits/chosen": -0.8828865885734558,
+      "logits/rejected": -1.2277076244354248,
+      "logps/chosen": -0.44972023367881775,
+      "logps/rejected": -1.8230899572372437,
+      "loss": 0.4531,
+      "odds_ratio_loss": 0.0991957038640976,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.022486012428998947,
+      "rewards/margins": 0.06866848468780518,
+      "rewards/rejected": -0.09115449339151382,
+      "sft_loss": 0.44972023367881775,
+      "step": 3675
+    },
+    {
+      "epoch": 2.944,
+      "grad_norm": 7.036020184151846,
+      "learning_rate": 5.305234949880001e-09,
+      "logits/chosen": -0.8388309478759766,
+      "logits/rejected": -0.784132719039917,
+      "logps/chosen": -0.5297240018844604,
+      "logps/rejected": -1.3611842393875122,
+      "loss": 0.4606,
+      "odds_ratio_loss": 0.23987647891044617,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02648620307445526,
+      "rewards/margins": 0.04157300293445587,
+      "rewards/rejected": -0.06805920600891113,
+      "sft_loss": 0.5297240018844604,
+      "step": 3680
+    },
+    {
+      "epoch": 2.948,
+      "grad_norm": 8.642237728292242,
+      "learning_rate": 4.57463473941544e-09,
+      "logits/chosen": -1.135202407836914,
+      "logits/rejected": -0.9565531611442566,
+      "logps/chosen": -0.48526114225387573,
+      "logps/rejected": -1.3105952739715576,
+      "loss": 0.5389,
+      "odds_ratio_loss": 0.2372247278690338,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.024263057857751846,
+      "rewards/margins": 0.04126670956611633,
+      "rewards/rejected": -0.06552976369857788,
+      "sft_loss": 0.48526114225387573,
+      "step": 3685
+    },
+    {
+      "epoch": 2.952,
+      "grad_norm": 4.384021145091059,
+      "learning_rate": 3.8980895450474455e-09,
+      "logits/chosen": -0.7918862104415894,
+      "logits/rejected": -1.0930888652801514,
+      "logps/chosen": -0.47447633743286133,
+      "logps/rejected": -1.5464060306549072,
+      "loss": 0.4936,
+      "odds_ratio_loss": 0.19870418310165405,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.023723818361759186,
+      "rewards/margins": 0.053596485406160355,
+      "rewards/rejected": -0.07732030004262924,
+      "sft_loss": 0.47447633743286133,
+      "step": 3690
+    },
+    {
+      "epoch": 2.956,
+      "grad_norm": 5.606541344203649,
+      "learning_rate": 3.275614021857609e-09,
+      "logits/chosen": -1.0226024389266968,
+      "logits/rejected": -1.014661192893982,
+      "logps/chosen": -0.8057661056518555,
+      "logps/rejected": -1.672181487083435,
+      "loss": 0.5466,
+      "odds_ratio_loss": 0.33504363894462585,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.040288303047418594,
+      "rewards/margins": 0.04332076758146286,
+      "rewards/rejected": -0.08360908180475235,
+      "sft_loss": 0.8057661056518555,
+      "step": 3695
+    },
+    {
+      "epoch": 2.96,
+      "grad_norm": 6.6670138700200035,
+      "learning_rate": 2.7072216536885855e-09,
+      "logits/chosen": -0.6843828558921814,
+      "logits/rejected": -0.8663504719734192,
+      "logps/chosen": -0.4747343063354492,
+      "logps/rejected": -1.3466534614562988,
+      "loss": 0.3528,
+      "odds_ratio_loss": 0.2356620728969574,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02373671717941761,
+      "rewards/margins": 0.04359595477581024,
+      "rewards/rejected": -0.0673326700925827,
+      "sft_loss": 0.4747343063354492,
+      "step": 3700
+    },
+    {
+      "epoch": 2.964,
+      "grad_norm": 7.141276758716209,
+      "learning_rate": 2.192924752854042e-09,
+      "logits/chosen": -0.7393054962158203,
+      "logits/rejected": -1.508094072341919,
+      "logps/chosen": -0.49731236696243286,
+      "logps/rejected": -1.8641046285629272,
+      "loss": 0.4864,
+      "odds_ratio_loss": 0.12721852958202362,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02486562170088291,
+      "rewards/margins": 0.06833961606025696,
+      "rewards/rejected": -0.09320523589849472,
+      "sft_loss": 0.49731236696243286,
+      "step": 3705
+    },
+    {
+      "epoch": 2.968,
+      "grad_norm": 5.266798743387709,
+      "learning_rate": 1.7327344598702667e-09,
+      "logits/chosen": -1.0665310621261597,
+      "logits/rejected": -0.858269989490509,
+      "logps/chosen": -0.536304235458374,
+      "logps/rejected": -1.4411725997924805,
+      "loss": 0.4227,
+      "odds_ratio_loss": 0.328524649143219,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.026815209537744522,
+      "rewards/margins": 0.04524341970682144,
+      "rewards/rejected": -0.07205863296985626,
+      "sft_loss": 0.536304235458374,
+      "step": 3710
+    },
+    {
+      "epoch": 2.972,
+      "grad_norm": 6.750661652998483,
+      "learning_rate": 1.3266607432155243e-09,
+      "logits/chosen": -0.6614164113998413,
+      "logits/rejected": -0.7348114848136902,
+      "logps/chosen": -0.39280936121940613,
+      "logps/rejected": -1.426816463470459,
+      "loss": 0.3934,
+      "odds_ratio_loss": 0.1887359321117401,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.019640469923615456,
+      "rewards/margins": 0.05170035362243652,
+      "rewards/rejected": -0.07134082168340683,
+      "sft_loss": 0.39280936121940613,
+      "step": 3715
+    },
+    {
+      "epoch": 2.976,
+      "grad_norm": 6.207932687783427,
+      "learning_rate": 9.747123991141193e-10,
+      "logits/chosen": -0.7722820043563843,
+      "logits/rejected": -0.7826868295669556,
+      "logps/chosen": -0.45992159843444824,
+      "logps/rejected": -1.4466217756271362,
+      "loss": 0.5124,
+      "odds_ratio_loss": 0.19496043026447296,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.02299608290195465,
+      "rewards/margins": 0.04933501034975052,
+      "rewards/rejected": -0.07233108580112457,
+      "sft_loss": 0.45992159843444824,
+      "step": 3720
+    },
+    {
+      "epoch": 2.98,
+      "grad_norm": 5.007278723928615,
+      "learning_rate": 6.768970513457151e-10,
+      "logits/chosen": -0.8529748916625977,
+      "logits/rejected": -1.0278675556182861,
+      "logps/chosen": -0.3574032783508301,
+      "logps/rejected": -1.736302137374878,
+      "loss": 0.4984,
+      "odds_ratio_loss": 0.11382756382226944,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -0.017870163545012474,
+      "rewards/margins": 0.06894494593143463,
+      "rewards/rejected": -0.08681510388851166,
+      "sft_loss": 0.3574032783508301,
+      "step": 3725
+    },
+    {
+      "epoch": 2.984,
+      "grad_norm": 5.762199382824231,
+      "learning_rate": 4.332211510807427e-10,
+      "logits/chosen": -0.7780022025108337,
+      "logits/rejected": -1.1421865224838257,
+      "logps/chosen": -0.7099230885505676,
+      "logps/rejected": -1.55299711227417,
+      "loss": 0.546,
+      "odds_ratio_loss": 0.3306809067726135,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03549615666270256,
+      "rewards/margins": 0.042153697460889816,
+      "rewards/rejected": -0.07764985412359238,
+      "sft_loss": 0.7099230885505676,
+      "step": 3730
+    },
+    {
+      "epoch": 2.988,
+      "grad_norm": 4.682934344503231,
+      "learning_rate": 2.43689976739403e-10,
+      "logits/chosen": -0.8084946870803833,
+      "logits/rejected": -1.319000482559204,
+      "logps/chosen": -0.4620262682437897,
+      "logps/rejected": -1.2795708179473877,
+      "loss": 0.4644,
+      "odds_ratio_loss": 0.23649486899375916,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.023101316764950752,
+      "rewards/margins": 0.04087722301483154,
+      "rewards/rejected": -0.06397853791713715,
+      "sft_loss": 0.4620262682437897,
+      "step": 3735
+    },
+    {
+      "epoch": 2.992,
+      "grad_norm": 6.157538833749046,
+      "learning_rate": 1.0830763387897902e-10,
+      "logits/chosen": -0.8642560839653015,
+      "logits/rejected": -1.1312994956970215,
+      "logps/chosen": -0.3675539791584015,
+      "logps/rejected": -1.7602100372314453,
+      "loss": 0.5132,
+      "odds_ratio_loss": 0.23475000262260437,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.018377700820565224,
+      "rewards/margins": 0.06963280588388443,
+      "rewards/rejected": -0.0880105048418045,
+      "sft_loss": 0.3675539791584015,
+      "step": 3740
+    },
+    {
+      "epoch": 2.996,
+      "grad_norm": 5.903707133419837,
+      "learning_rate": 2.7077055103075233e-11,
+      "logits/chosen": -0.7607343792915344,
+      "logits/rejected": -1.1049892902374268,
+      "logps/chosen": -0.606769323348999,
+      "logps/rejected": -1.8843456506729126,
+      "loss": 0.4721,
+      "odds_ratio_loss": 0.20161142945289612,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.03033846616744995,
+      "rewards/margins": 0.06387881934642792,
+      "rewards/rejected": -0.09421730041503906,
+      "sft_loss": 0.606769323348999,
+      "step": 3745
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 4.814195311056904,
+      "learning_rate": 0.0,
+      "logits/chosen": -1.1040897369384766,
+      "logits/rejected": -1.2379401922225952,
+      "logps/chosen": -0.4253109097480774,
+      "logps/rejected": -1.3982548713684082,
+      "loss": 0.4811,
+      "odds_ratio_loss": 0.22021734714508057,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.02126554399728775,
+      "rewards/margins": 0.0486472025513649,
+      "rewards/rejected": -0.06991274654865265,
+      "sft_loss": 0.4253109097480774,
+      "step": 3750
+    },
+    {
+      "epoch": 3.0,
+      "step": 3750,
+      "total_flos": 126401337753600.0,
+      "train_loss": 0.8602390615145366,
+      "train_runtime": 16444.8476,
+      "train_samples_per_second": 3.649,
+      "train_steps_per_second": 0.228
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 3750,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 100.0,
+  "total_flos": 126401337753600.0,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}