diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -1,9 +1,9 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 4.0,
+  "epoch": 1.0,
   "eval_steps": 100,
-  "global_step": 1540,
+  "global_step": 385,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -25,2537 +25,597 @@
     {
       "epoch": 0.03,
       "learning_rate": 1.282051282051282e-06,
-      "logits/chosen": -1.866829752922058,
-      "logits/rejected": -1.8711390495300293,
-      "logps/chosen": -36.97040557861328,
-      "logps/rejected": -33.66280746459961,
-      "loss": 0.9793,
-      "rewards/accuracies": 0.5972222089767456,
-      "rewards/chosen": 0.010857796296477318,
-      "rewards/margins": 0.020676644518971443,
-      "rewards/rejected": -0.009818848222494125,
+      "logits/chosen": -1.8667490482330322,
+      "logits/rejected": -1.8710733652114868,
+      "logps/chosen": -36.97007369995117,
+      "logps/rejected": -33.66944885253906,
+      "loss": 0.9317,
+      "rewards/accuracies": 0.5694444179534912,
+      "rewards/chosen": 0.03287407010793686,
+      "rewards/margins": 0.06830974668264389,
+      "rewards/rejected": -0.03543568402528763,
       "step": 10
     },
     {
       "epoch": 0.05,
       "learning_rate": 2.564102564102564e-06,
-      "logits/chosen": -1.9976301193237305,
-      "logits/rejected": -2.0002708435058594,
-      "logps/chosen": -29.64252281188965,
-      "logps/rejected": -29.05857276916504,
-      "loss": 1.0016,
-      "rewards/accuracies": 0.4749999940395355,
-      "rewards/chosen": -9.895558468997478e-05,
-      "rewards/margins": -0.0015598980244249105,
-      "rewards/rejected": 0.0014609418576583266,
+      "logits/chosen": -1.9981460571289062,
+      "logits/rejected": -2.000789165496826,
+      "logps/chosen": -29.641231536865234,
+      "logps/rejected": -29.06744384765625,
+      "loss": 0.9955,
+      "rewards/accuracies": 0.44999998807907104,
+      "rewards/chosen": 0.0008672710391692817,
+      "rewards/margins": 0.004467610269784927,
+      "rewards/rejected": -0.0036003391724079847,
       "step": 20
     },
     {
       "epoch": 0.08,
       "learning_rate": 3.846153846153847e-06,
-      "logits/chosen": -1.9210872650146484,
-      "logits/rejected": -1.918402910232544,
-      "logps/chosen": -31.397838592529297,
-      "logps/rejected": -33.22719192504883,
-      "loss": 0.9919,
-      "rewards/accuracies": 0.612500011920929,
-      "rewards/chosen": 0.005476226564496756,
-      "rewards/margins": 0.00809534452855587,
-      "rewards/rejected": -0.0026191179640591145,
+      "logits/chosen": -1.920600175857544,
+      "logits/rejected": -1.917925238609314,
+      "logps/chosen": -31.395061492919922,
+      "logps/rejected": -33.240909576416016,
+      "loss": 0.9609,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.01893135905265808,
+      "rewards/margins": 0.03913776949048042,
+      "rewards/rejected": -0.020206410437822342,
       "step": 30
     },
     {
       "epoch": 0.1,
       "learning_rate": 4.999896948438434e-06,
-      "logits/chosen": -2.017606258392334,
-      "logits/rejected": -2.0088634490966797,
-      "logps/chosen": -32.566612243652344,
-      "logps/rejected": -32.52539825439453,
-      "loss": 0.9931,
-      "rewards/accuracies": 0.5375000238418579,
-      "rewards/chosen": 0.0030602235347032547,
-      "rewards/margins": 0.0068772160448133945,
-      "rewards/rejected": -0.0038169927429407835,
+      "logits/chosen": -2.017815113067627,
+      "logits/rejected": -2.0090720653533936,
+      "logps/chosen": -32.5806884765625,
+      "logps/rejected": -32.515098571777344,
+      "loss": 1.0013,
+      "rewards/accuracies": 0.44999998807907104,
+      "rewards/chosen": -0.003494268748909235,
+      "rewards/margins": -0.0013132141903042793,
+      "rewards/rejected": -0.002181055024266243,
       "step": 40
     },
     {
       "epoch": 0.13,
       "learning_rate": 4.987541037542187e-06,
-      "logits/chosen": -1.8625805377960205,
-      "logits/rejected": -1.8518139123916626,
-      "logps/chosen": -33.56818771362305,
-      "logps/rejected": -35.4713020324707,
-      "loss": 0.9975,
+      "logits/chosen": -1.8630876541137695,
+      "logits/rejected": -1.8523353338241577,
+      "logps/chosen": -33.549766540527344,
+      "logps/rejected": -35.46318435668945,
+      "loss": 0.9833,
       "rewards/accuracies": 0.550000011920929,
-      "rewards/chosen": -0.0032131080515682697,
-      "rewards/margins": 0.0024742281530052423,
-      "rewards/rejected": -0.005687335971742868,
+      "rewards/chosen": 0.006940312683582306,
+      "rewards/margins": 0.016694897785782814,
+      "rewards/rejected": -0.009754580445587635,
       "step": 50
     },
     {
       "epoch": 0.16,
       "learning_rate": 4.954691471941119e-06,
-      "logits/chosen": -1.9413166046142578,
-      "logits/rejected": -1.9432499408721924,
-      "logps/chosen": -32.55731964111328,
-      "logps/rejected": -33.21186065673828,
-      "loss": 0.975,
-      "rewards/accuracies": 0.5249999761581421,
-      "rewards/chosen": 0.012859411537647247,
-      "rewards/margins": 0.02500327304005623,
-      "rewards/rejected": -0.01214386522769928,
+      "logits/chosen": -1.9419746398925781,
+      "logits/rejected": -1.943914771080017,
+      "logps/chosen": -32.527896881103516,
+      "logps/rejected": -33.21547317504883,
+      "loss": 0.9153,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.06505907326936722,
+      "rewards/margins": 0.10474522411823273,
+      "rewards/rejected": -0.03968615084886551,
       "step": 60
     },
     {
       "epoch": 0.18,
       "learning_rate": 4.901618883413549e-06,
-      "logits/chosen": -2.0717270374298096,
-      "logits/rejected": -2.0766983032226562,
-      "logps/chosen": -34.008384704589844,
-      "logps/rejected": -36.6549186706543,
-      "loss": 0.9802,
-      "rewards/accuracies": 0.6000000238418579,
-      "rewards/chosen": -0.0059700957499444485,
-      "rewards/margins": 0.01984369568526745,
-      "rewards/rejected": -0.025813793763518333,
+      "logits/chosen": -2.07257080078125,
+      "logits/rejected": -2.0775399208068848,
+      "logps/chosen": -34.00202560424805,
+      "logps/rejected": -36.622886657714844,
+      "loss": 0.9636,
+      "rewards/accuracies": 0.512499988079071,
+      "rewards/chosen": -0.012188142165541649,
+      "rewards/margins": 0.03642461448907852,
+      "rewards/rejected": -0.04861275106668472,
       "step": 70
     },
     {
       "epoch": 0.21,
       "learning_rate": 4.828760511501322e-06,
-      "logits/chosen": -1.932422399520874,
-      "logits/rejected": -1.9355411529541016,
-      "logps/chosen": -34.327762603759766,
-      "logps/rejected": -34.66461944580078,
-      "loss": 0.9483,
+      "logits/chosen": -1.9333629608154297,
+      "logits/rejected": -1.9364970922470093,
+      "logps/chosen": -34.302101135253906,
+      "logps/rejected": -34.63160705566406,
+      "loss": 0.862,
       "rewards/accuracies": 0.5625,
-      "rewards/chosen": 0.03001437522470951,
-      "rewards/margins": 0.05167943984270096,
-      "rewards/rejected": -0.02166506089270115,
+      "rewards/chosen": 0.1131378561258316,
+      "rewards/margins": 0.1484164148569107,
+      "rewards/rejected": -0.0352785661816597,
       "step": 80
     },
     {
       "epoch": 0.23,
       "learning_rate": 4.7367166013034295e-06,
-      "logits/chosen": -1.9404243230819702,
-      "logits/rejected": -1.944933533668518,
-      "logps/chosen": -32.38780975341797,
-      "logps/rejected": -32.35973358154297,
-      "loss": 0.9759,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.02388269081711769,
-      "rewards/margins": 0.024130593985319138,
-      "rewards/rejected": -0.00024790168390609324,
+      "logits/chosen": -1.9408857822418213,
+      "logits/rejected": -1.945412039756775,
+      "logps/chosen": -32.36528015136719,
+      "logps/rejected": -32.34526824951172,
+      "loss": 0.9225,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.09192506223917007,
+      "rewards/margins": 0.07964853197336197,
+      "rewards/rejected": 0.012276534922420979,
       "step": 90
     },
     {
       "epoch": 0.26,
       "learning_rate": 4.626245458345211e-06,
-      "logits/chosen": -2.0375094413757324,
-      "logits/rejected": -2.0355331897735596,
-      "logps/chosen": -32.14101028442383,
-      "logps/rejected": -31.31320571899414,
-      "loss": 0.956,
-      "rewards/accuracies": 0.6000000238418579,
-      "rewards/chosen": 0.029947593808174133,
-      "rewards/margins": 0.04404589533805847,
-      "rewards/rejected": -0.014098303392529488,
+      "logits/chosen": -2.037550210952759,
+      "logits/rejected": -2.0355725288391113,
+      "logps/chosen": -32.142730712890625,
+      "logps/rejected": -31.29366683959961,
+      "loss": 0.8913,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.08829358220100403,
+      "rewards/margins": 0.1130049005150795,
+      "rewards/rejected": -0.02471131458878517,
       "step": 100
     },
     {
       "epoch": 0.26,
-      "eval_logits/chosen": -2.232485294342041,
-      "eval_logits/rejected": -2.227635622024536,
-      "eval_logps/chosen": -34.049468994140625,
-      "eval_logps/rejected": -37.54916000366211,
-      "eval_loss": 0.994844913482666,
-      "eval_rewards/accuracies": 0.5631229281425476,
-      "eval_rewards/chosen": -0.004474184010177851,
-      "eval_rewards/margins": 0.0052884831093251705,
-      "eval_rewards/rejected": -0.009762667119503021,
-      "eval_runtime": 146.0056,
-      "eval_samples_per_second": 2.349,
+      "eval_logits/chosen": -2.232161283493042,
+      "eval_logits/rejected": -2.2273108959198,
+      "eval_logps/chosen": -34.040714263916016,
+      "eval_logps/rejected": -37.54047775268555,
+      "eval_loss": 0.9844526052474976,
+      "eval_rewards/accuracies": 0.5195183157920837,
+      "eval_rewards/chosen": -0.005542654078453779,
+      "eval_rewards/margins": 0.015924591571092606,
+      "eval_rewards/rejected": -0.021467244252562523,
+      "eval_runtime": 145.9018,
+      "eval_samples_per_second": 2.351,
       "eval_steps_per_second": 0.295,
       "step": 100
     },
     {
       "epoch": 0.29,
       "learning_rate": 4.498257201263691e-06,
-      "logits/chosen": -1.9923181533813477,
-      "logits/rejected": -1.989919662475586,
-      "logps/chosen": -33.152462005615234,
-      "logps/rejected": -34.039642333984375,
-      "loss": 0.9695,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.027407139539718628,
-      "rewards/margins": 0.03048800863325596,
-      "rewards/rejected": -0.003080862807109952,
+      "logits/chosen": -1.9926517009735107,
+      "logits/rejected": -1.9902753829956055,
+      "logps/chosen": -33.12412643432617,
+      "logps/rejected": -34.011417388916016,
+      "loss": 0.9361,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.10772128403186798,
+      "rewards/margins": 0.09156213700771332,
+      "rewards/rejected": 0.016159160062670708,
       "step": 110
     },
     {
       "epoch": 0.31,
       "learning_rate": 4.353806263777678e-06,
-      "logits/chosen": -2.0034618377685547,
-      "logits/rejected": -1.9951326847076416,
-      "logps/chosen": -32.344200134277344,
-      "logps/rejected": -32.13230895996094,
-      "loss": 0.9785,
-      "rewards/accuracies": 0.512499988079071,
-      "rewards/chosen": 0.03021082654595375,
-      "rewards/margins": 0.02145235612988472,
-      "rewards/rejected": 0.008758468553423882,
+      "logits/chosen": -2.00441312789917,
+      "logits/rejected": -1.996093988418579,
+      "logps/chosen": -32.33955383300781,
+      "logps/rejected": -32.13432312011719,
+      "loss": 0.9401,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.09481850266456604,
+      "rewards/margins": 0.07035262137651443,
+      "rewards/rejected": 0.024465877562761307,
       "step": 120
     },
     {
       "epoch": 0.34,
       "learning_rate": 4.1940827077152755e-06,
-      "logits/chosen": -2.031149387359619,
-      "logits/rejected": -2.0231785774230957,
-      "logps/chosen": -30.327457427978516,
-      "logps/rejected": -32.072906494140625,
-      "loss": 0.9496,
-      "rewards/accuracies": 0.5375000238418579,
-      "rewards/chosen": 0.044478606432676315,
-      "rewards/margins": 0.050448428839445114,
-      "rewards/rejected": -0.005969815421849489,
+      "logits/chosen": -2.0318965911865234,
+      "logits/rejected": -2.023927688598633,
+      "logps/chosen": -30.336984634399414,
+      "logps/rejected": -32.0634765625,
+      "loss": 0.9061,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.12486012279987335,
+      "rewards/margins": 0.1342838853597641,
+      "rewards/rejected": -0.009423775598406792,
       "step": 130
     },
     {
       "epoch": 0.36,
       "learning_rate": 4.0204024186666215e-06,
-      "logits/chosen": -1.961639642715454,
-      "logits/rejected": -1.9718501567840576,
-      "logps/chosen": -31.234424591064453,
-      "logps/rejected": -32.57652282714844,
-      "loss": 0.9319,
+      "logits/chosen": -1.9620994329452515,
+      "logits/rejected": -1.9723354578018188,
+      "logps/chosen": -31.222240447998047,
+      "logps/rejected": -32.57916259765625,
+      "loss": 0.795,
       "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.05557093024253845,
-      "rewards/margins": 0.06806287914514542,
-      "rewards/rejected": -0.012491944245994091,
+      "rewards/chosen": 0.17767605185508728,
+      "rewards/margins": 0.21753115952014923,
+      "rewards/rejected": -0.03985511139035225,
       "step": 140
     },
     {
       "epoch": 0.39,
       "learning_rate": 3.834196265035119e-06,
-      "logits/chosen": -1.8733562231063843,
-      "logits/rejected": -1.8745276927947998,
-      "logps/chosen": -33.9085578918457,
-      "logps/rejected": -34.8173828125,
-      "loss": 0.8973,
-      "rewards/accuracies": 0.5874999761581421,
-      "rewards/chosen": 0.08524172008037567,
-      "rewards/margins": 0.10328052937984467,
-      "rewards/rejected": -0.018038809299468994,
+      "logits/chosen": -1.8727748394012451,
+      "logits/rejected": -1.87395441532135,
+      "logps/chosen": -33.931861877441406,
+      "logps/rejected": -34.79869842529297,
+      "loss": 0.7946,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.23474939167499542,
+      "rewards/margins": 0.272051066160202,
+      "rewards/rejected": -0.03730170056223869,
       "step": 150
     },
     {
       "epoch": 0.42,
       "learning_rate": 3.636998309800573e-06,
-      "logits/chosen": -1.9239585399627686,
-      "logits/rejected": -1.9205586910247803,
-      "logps/chosen": -35.99612045288086,
-      "logps/rejected": -32.70136260986328,
-      "loss": 0.9538,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.052666354924440384,
-      "rewards/margins": 0.04624979570508003,
-      "rewards/rejected": 0.006416561547666788,
+      "logits/chosen": -1.9248745441436768,
+      "logits/rejected": -1.9214649200439453,
+      "logps/chosen": -36.014469146728516,
+      "logps/rejected": -32.73783493041992,
+      "loss": 0.8532,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.14148668944835663,
+      "rewards/margins": 0.15506146848201752,
+      "rewards/rejected": -0.013574766926467419,
       "step": 160
     },
     {
       "epoch": 0.44,
       "learning_rate": 3.4304331721118078e-06,
-      "logits/chosen": -2.02461576461792,
-      "logits/rejected": -2.017319440841675,
-      "logps/chosen": -33.47701644897461,
-      "logps/rejected": -31.457340240478516,
-      "loss": 0.8721,
-      "rewards/accuracies": 0.7749999761581421,
-      "rewards/chosen": 0.09625236690044403,
-      "rewards/margins": 0.12790000438690186,
-      "rewards/rejected": -0.031647637486457825,
+      "logits/chosen": -2.025555372238159,
+      "logits/rejected": -2.01819109916687,
+      "logps/chosen": -33.50218200683594,
+      "logps/rejected": -31.41971206665039,
+      "loss": 0.7292,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.26611366868019104,
+      "rewards/margins": 0.3271873891353607,
+      "rewards/rejected": -0.06107370927929878,
       "step": 170
     },
     {
       "epoch": 0.47,
       "learning_rate": 3.2162026428305436e-06,
-      "logits/chosen": -2.030789613723755,
-      "logits/rejected": -2.0360310077667236,
-      "logps/chosen": -32.23139572143555,
-      "logps/rejected": -32.44371032714844,
-      "loss": 0.9184,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.0940173864364624,
-      "rewards/margins": 0.0815875232219696,
-      "rewards/rejected": 0.012429863214492798,
+      "logits/chosen": -2.0320167541503906,
+      "logits/rejected": -2.037261486053467,
+      "logps/chosen": -32.24850845336914,
+      "logps/rejected": -32.45344924926758,
+      "loss": 0.7865,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.2666531801223755,
+      "rewards/margins": 0.23813048005104065,
+      "rewards/rejected": 0.028522688895463943,
       "step": 180
     },
     {
       "epoch": 0.49,
       "learning_rate": 2.996071664294641e-06,
-      "logits/chosen": -2.0321707725524902,
-      "logits/rejected": -2.0294008255004883,
-      "logps/chosen": -31.273319244384766,
-      "logps/rejected": -31.368595123291016,
-      "loss": 0.9202,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.06469077616930008,
-      "rewards/margins": 0.0798054188489914,
-      "rewards/rejected": -0.015114650130271912,
+      "logits/chosen": -2.032525062561035,
+      "logits/rejected": -2.0297436714172363,
+      "logps/chosen": -31.313217163085938,
+      "logps/rejected": -31.349472045898438,
+      "loss": 0.8387,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.1581628918647766,
+      "rewards/margins": 0.18629543483257294,
+      "rewards/rejected": -0.02813255414366722,
       "step": 190
     },
     {
       "epoch": 0.52,
       "learning_rate": 2.7718537898066833e-06,
-      "logits/chosen": -1.9020545482635498,
-      "logits/rejected": -1.9066784381866455,
-      "logps/chosen": -31.315628051757812,
-      "logps/rejected": -32.859519958496094,
-      "loss": 0.8922,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": 0.0871955007314682,
-      "rewards/margins": 0.1077716127038002,
-      "rewards/rejected": -0.02057611010968685,
+      "logits/chosen": -1.902632713317871,
+      "logits/rejected": -1.907284140586853,
+      "logps/chosen": -31.320043563842773,
+      "logps/rejected": -32.85698699951172,
+      "loss": 0.7293,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.2576131224632263,
+      "rewards/margins": 0.31706100702285767,
+      "rewards/rejected": -0.05944784730672836,
       "step": 200
     },
     {
       "epoch": 0.52,
-      "eval_logits/chosen": -2.2279350757598877,
-      "eval_logits/rejected": -2.2231225967407227,
-      "eval_logps/chosen": -34.06590270996094,
-      "eval_logps/rejected": -37.589378356933594,
-      "eval_loss": 0.9875569939613342,
-      "eval_rewards/accuracies": 0.530315637588501,
-      "eval_rewards/chosen": -0.009404394775629044,
-      "eval_rewards/margins": 0.012423668056726456,
-      "eval_rewards/rejected": -0.0218280628323555,
-      "eval_runtime": 145.8494,
-      "eval_samples_per_second": 2.352,
+      "eval_logits/chosen": -2.228637933731079,
+      "eval_logits/rejected": -2.2238004207611084,
+      "eval_logps/chosen": -34.053680419921875,
+      "eval_logps/rejected": -37.581058502197266,
+      "eval_loss": 0.9601577520370483,
+      "eval_rewards/accuracies": 0.5714285373687744,
+      "eval_rewards/chosen": -0.01721162348985672,
+      "eval_rewards/margins": 0.040783192962408066,
+      "eval_rewards/rejected": -0.057994820177555084,
+      "eval_runtime": 145.5388,
+      "eval_samples_per_second": 2.357,
       "eval_steps_per_second": 0.295,
       "step": 200
     },
     {
       "epoch": 0.55,
       "learning_rate": 2.5453962426402006e-06,
-      "logits/chosen": -2.0143890380859375,
-      "logits/rejected": -2.0250439643859863,
-      "logps/chosen": -31.765193939208984,
-      "logps/rejected": -33.9805908203125,
-      "loss": 0.9046,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.061833299696445465,
-      "rewards/margins": 0.09720136225223541,
-      "rewards/rejected": -0.03536807745695114,
+      "logits/chosen": -2.0149245262145996,
+      "logits/rejected": -2.025560140609741,
+      "logps/chosen": -31.77438735961914,
+      "logps/rejected": -33.95419692993164,
+      "loss": 0.7666,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.17722666263580322,
+      "rewards/margins": 0.25957340002059937,
+      "rewards/rejected": -0.08234670013189316,
       "step": 210
     },
     {
       "epoch": 0.57,
       "learning_rate": 2.3185646976551794e-06,
-      "logits/chosen": -1.907122254371643,
-      "logits/rejected": -1.9218772649765015,
-      "logps/chosen": -29.810348510742188,
-      "logps/rejected": -31.652551651000977,
-      "loss": 0.8813,
-      "rewards/accuracies": 0.75,
-      "rewards/chosen": 0.0903574600815773,
-      "rewards/margins": 0.11866964399814606,
-      "rewards/rejected": -0.028312185779213905,
+      "logits/chosen": -1.906951904296875,
+      "logits/rejected": -1.9217418432235718,
+      "logps/chosen": -29.83829116821289,
+      "logps/rejected": -31.636096954345703,
+      "loss": 0.7204,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.2459266185760498,
+      "rewards/margins": 0.3160557448863983,
+      "rewards/rejected": -0.07012919336557388,
       "step": 220
     },
     {
       "epoch": 0.6,
       "learning_rate": 2.0932279108998323e-06,
-      "logits/chosen": -1.9638553857803345,
-      "logits/rejected": -1.9678163528442383,
-      "logps/chosen": -33.116092681884766,
-      "logps/rejected": -31.662506103515625,
-      "loss": 0.8826,
-      "rewards/accuracies": 0.637499988079071,
-      "rewards/chosen": 0.09041241556406021,
-      "rewards/margins": 0.1243731826543808,
-      "rewards/rejected": -0.03396075591444969,
+      "logits/chosen": -1.9629713296890259,
+      "logits/rejected": -1.9669532775878906,
+      "logps/chosen": -33.124656677246094,
+      "logps/rejected": -31.630443572998047,
+      "loss": 0.7348,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.2635299265384674,
+      "rewards/margins": 0.3365571200847626,
+      "rewards/rejected": -0.07302714884281158,
       "step": 230
     },
     {
       "epoch": 0.62,
       "learning_rate": 1.8712423238279358e-06,
-      "logits/chosen": -1.9616014957427979,
-      "logits/rejected": -1.9398181438446045,
-      "logps/chosen": -33.844024658203125,
-      "logps/rejected": -35.163543701171875,
-      "loss": 0.8473,
+      "logits/chosen": -1.9614177942276,
+      "logits/rejected": -1.9395818710327148,
+      "logps/chosen": -33.87095260620117,
+      "logps/rejected": -35.10104751586914,
+      "loss": 0.6865,
       "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.09291732311248779,
-      "rewards/margins": 0.15674278140068054,
-      "rewards/rejected": -0.06382545083761215,
+      "rewards/chosen": 0.254514217376709,
+      "rewards/margins": 0.38974156975746155,
+      "rewards/rejected": -0.13522735238075256,
       "step": 240
     },
     {
       "epoch": 0.65,
       "learning_rate": 1.6544367689701824e-06,
-      "logits/chosen": -2.002652883529663,
-      "logits/rejected": -1.9993493556976318,
-      "logps/chosen": -32.73945617675781,
-      "logps/rejected": -36.27147674560547,
-      "loss": 0.9305,
-      "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": 0.055684663355350494,
-      "rewards/margins": 0.0696500912308693,
-      "rewards/rejected": -0.013965430669486523,
+      "logits/chosen": -2.0029492378234863,
+      "logits/rejected": -1.999629020690918,
+      "logps/chosen": -32.730865478515625,
+      "logps/rejected": -36.28009796142578,
+      "loss": 0.8055,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.17478153109550476,
+      "rewards/margins": 0.2244400531053543,
+      "rewards/rejected": -0.04965851828455925,
       "step": 250
     },
     {
       "epoch": 0.68,
       "learning_rate": 1.4445974030621963e-06,
-      "logits/chosen": -1.869681715965271,
-      "logits/rejected": -1.8672786951065063,
-      "logps/chosen": -34.0054931640625,
-      "logps/rejected": -35.55742645263672,
-      "loss": 0.9293,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.05487124249339104,
-      "rewards/margins": 0.07071295380592346,
-      "rewards/rejected": -0.015841711312532425,
+      "logits/chosen": -1.8703190088272095,
+      "logits/rejected": -1.8679043054580688,
+      "logps/chosen": -33.98231887817383,
+      "logps/rejected": -35.54644775390625,
+      "loss": 0.8042,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.18547315895557404,
+      "rewards/margins": 0.22311437129974365,
+      "rewards/rejected": -0.03764120861887932,
       "step": 260
     },
     {
       "epoch": 0.7,
       "learning_rate": 1.243452991757889e-06,
-      "logits/chosen": -1.8544126749038696,
-      "logits/rejected": -1.852034330368042,
-      "logps/chosen": -34.22079086303711,
-      "logps/rejected": -31.864788055419922,
-      "loss": 0.9205,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.0530419759452343,
-      "rewards/margins": 0.08107996731996536,
-      "rewards/rejected": -0.028037995100021362,
+      "logits/chosen": -1.8551464080810547,
+      "logits/rejected": -1.852746605873108,
+      "logps/chosen": -34.20850372314453,
+      "logps/rejected": -31.803356170654297,
+      "loss": 0.8096,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.170187309384346,
+      "rewards/margins": 0.19900819659233093,
+      "rewards/rejected": -0.02882089652121067,
       "step": 270
     },
     {
       "epoch": 0.73,
       "learning_rate": 1.0526606671603523e-06,
-      "logits/chosen": -1.957960844039917,
-      "logits/rejected": -1.9474948644638062,
-      "logps/chosen": -35.02719497680664,
-      "logps/rejected": -31.89957046508789,
-      "loss": 0.8836,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": 0.09883318096399307,
-      "rewards/margins": 0.11640648543834686,
-      "rewards/rejected": -0.017573300749063492,
+      "logits/chosen": -1.9582526683807373,
+      "logits/rejected": -1.947749376296997,
+      "logps/chosen": -35.0114631652832,
+      "logps/rejected": -31.88564682006836,
+      "loss": 0.692,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.3106640875339508,
+      "rewards/margins": 0.3508565425872803,
+      "rewards/rejected": -0.040192440152168274,
       "step": 280
     },
     {
       "epoch": 0.75,
       "learning_rate": 8.737922755071455e-07,
-      "logits/chosen": -2.0539374351501465,
-      "logits/rejected": -2.0390708446502686,
-      "logps/chosen": -30.736724853515625,
-      "logps/rejected": -32.64698791503906,
-      "loss": 0.9452,
+      "logits/chosen": -2.053699493408203,
+      "logits/rejected": -2.038789749145508,
+      "logps/chosen": -30.727243423461914,
+      "logps/rejected": -32.641685485839844,
+      "loss": 0.8827,
       "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": 0.05622049421072006,
-      "rewards/margins": 0.05476574972271919,
-      "rewards/rejected": 0.001454742974601686,
+      "rewards/chosen": 0.17719359695911407,
+      "rewards/margins": 0.16805905103683472,
+      "rewards/rejected": 0.009134533815085888,
       "step": 290
     },
     {
       "epoch": 0.78,
       "learning_rate": 7.08321427484816e-07,
-      "logits/chosen": -1.9243535995483398,
-      "logits/rejected": -1.9218591451644897,
-      "logps/chosen": -32.35340118408203,
-      "logps/rejected": -30.938573837280273,
-      "loss": 0.8033,
-      "rewards/accuracies": 0.75,
-      "rewards/chosen": 0.17211246490478516,
-      "rewards/margins": 0.20864196121692657,
-      "rewards/rejected": -0.036529459059238434,
+      "logits/chosen": -1.9243850708007812,
+      "logits/rejected": -1.9218356609344482,
+      "logps/chosen": -32.43050003051758,
+      "logps/rejected": -30.8950138092041,
+      "loss": 0.6144,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": 0.446951299905777,
+      "rewards/margins": 0.5173346400260925,
+      "rewards/rejected": -0.07038338482379913,
       "step": 300
     },
     {
       "epoch": 0.78,
-      "eval_logits/chosen": -2.224973440170288,
-      "eval_logits/rejected": -2.2201507091522217,
-      "eval_logps/chosen": -34.1029052734375,
-      "eval_logps/rejected": -37.6210823059082,
-      "eval_loss": 0.989064633846283,
-      "eval_rewards/accuracies": 0.5182723999023438,
-      "eval_rewards/chosen": -0.020507192239165306,
-      "eval_rewards/margins": 0.010832280851900578,
-      "eval_rewards/rejected": -0.03133947402238846,
-      "eval_runtime": 145.7799,
-      "eval_samples_per_second": 2.353,
+      "eval_logits/chosen": -2.2248916625976562,
+      "eval_logits/rejected": -2.2200686931610107,
+      "eval_logps/chosen": -34.08660125732422,
+      "eval_logps/rejected": -37.603153228759766,
+      "eval_loss": 0.9712583422660828,
+      "eval_rewards/accuracies": 0.5282392501831055,
+      "eval_rewards/chosen": -0.046843186020851135,
+      "eval_rewards/margins": 0.031036507338285446,
+      "eval_rewards/rejected": -0.07787969708442688,
+      "eval_runtime": 145.8399,
+      "eval_samples_per_second": 2.352,
       "eval_steps_per_second": 0.295,
       "step": 300
     },
     {
       "epoch": 0.81,
-      "grad_norm": 14.5,
-      "learning_rate": 4.84533120650964e-06,
-      "logits/chosen": -2.058703660964966,
-      "logits/rejected": -2.045941114425659,
-      "logps/chosen": -32.090511322021484,
-      "logps/rejected": -32.93180465698242,
-      "loss": 0.7824,
-      "rewards/accuracies": 0.8374999761581421,
-      "rewards/chosen": 0.1313646137714386,
-      "rewards/margins": 0.2185889184474945,
-      "rewards/rejected": -0.08722428977489471,
+      "learning_rate": 5.576113578589035e-07,
+      "logits/chosen": -1.9094560146331787,
+      "logits/rejected": -1.9062097072601318,
+      "logps/chosen": -31.33370018005371,
+      "logps/rejected": -33.83475112915039,
+      "loss": 0.728,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.24268333613872528,
+      "rewards/margins": 0.32601919770240784,
+      "rewards/rejected": -0.08333584666252136,
       "step": 310
     },
     {
       "epoch": 0.83,
-      "grad_norm": 13.75,
-      "learning_rate": 4.825108134172131e-06,
-      "logits/chosen": -1.9678691625595093,
-      "logits/rejected": -1.9592756032943726,
-      "logps/chosen": -31.680923461914062,
-      "logps/rejected": -30.48342514038086,
-      "loss": 0.7409,
-      "rewards/accuracies": 0.8374999761581421,
-      "rewards/chosen": 0.20221905410289764,
-      "rewards/margins": 0.2748294472694397,
-      "rewards/rejected": -0.07261036336421967,
+      "learning_rate": 4.229036944380913e-07,
+      "logits/chosen": -1.9593706130981445,
+      "logits/rejected": -1.9471466541290283,
+      "logps/chosen": -34.3392333984375,
+      "logps/rejected": -33.68544387817383,
+      "loss": 0.6954,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.2311573475599289,
+      "rewards/margins": 0.36310091614723206,
+      "rewards/rejected": -0.13194358348846436,
       "step": 320
     },
     {
       "epoch": 0.86,
-      "grad_norm": 18.625,
-      "learning_rate": 4.80369052967602e-06,
-      "logits/chosen": -1.902902364730835,
-      "logits/rejected": -1.9149367809295654,
-      "logps/chosen": -29.743581771850586,
-      "logps/rejected": -33.734004974365234,
-      "loss": 0.6801,
-      "rewards/accuracies": 0.8125,
-      "rewards/chosen": 0.2207062542438507,
-      "rewards/margins": 0.32023119926452637,
-      "rewards/rejected": -0.09952493757009506,
+      "learning_rate": 3.053082288996112e-07,
+      "logits/chosen": -1.9940307140350342,
+      "logits/rejected": -1.9925845861434937,
+      "logps/chosen": -33.20854568481445,
+      "logps/rejected": -32.552764892578125,
+      "loss": 0.7359,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.24952416121959686,
+      "rewards/margins": 0.31538745760917664,
+      "rewards/rejected": -0.06586329638957977,
       "step": 330
     },
     {
       "epoch": 0.88,
-      "grad_norm": 17.625,
-      "learning_rate": 4.781089396387968e-06,
-      "logits/chosen": -1.8662229776382446,
-      "logits/rejected": -1.8570550680160522,
-      "logps/chosen": -33.95461654663086,
-      "logps/rejected": -36.284881591796875,
-      "loss": 0.6995,
-      "rewards/accuracies": 0.862500011920929,
-      "rewards/chosen": 0.22329291701316833,
-      "rewards/margins": 0.3517391085624695,
-      "rewards/rejected": -0.12844614684581757,
+      "learning_rate": 2.0579377374915805e-07,
+      "logits/chosen": -2.080763578414917,
+      "logits/rejected": -2.065063953399658,
+      "logps/chosen": -33.82006072998047,
+      "logps/rejected": -33.105167388916016,
+      "loss": 0.7407,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.32339948415756226,
+      "rewards/margins": 0.31181785464286804,
+      "rewards/rejected": 0.011581619270145893,
       "step": 340
     },
     {
       "epoch": 0.91,
-      "grad_norm": 13.8125,
-      "learning_rate": 4.757316345716554e-06,
-      "logits/chosen": -1.915514588356018,
-      "logits/rejected": -1.9161659479141235,
-      "logps/chosen": -33.57290267944336,
-      "logps/rejected": -34.201316833496094,
-      "loss": 0.6661,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": 0.2497142255306244,
-      "rewards/margins": 0.35541585087776184,
-      "rewards/rejected": -0.10570164024829865,
+      "learning_rate": 1.2518018074041684e-07,
+      "logits/chosen": -1.9535402059555054,
+      "logits/rejected": -1.9526821374893188,
+      "logps/chosen": -32.8734130859375,
+      "logps/rejected": -32.565185546875,
+      "loss": 0.645,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.36781299114227295,
+      "rewards/margins": 0.4578477442264557,
+      "rewards/rejected": -0.09003473073244095,
       "step": 350
     },
     {
       "epoch": 0.94,
-      "grad_norm": 13.25,
-      "learning_rate": 4.73238359114687e-06,
-      "logits/chosen": -2.041379928588867,
-      "logits/rejected": -2.0475096702575684,
-      "logps/chosen": -31.086139678955078,
-      "logps/rejected": -33.104515075683594,
-      "loss": 0.7501,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": 0.14613109827041626,
-      "rewards/margins": 0.27700015902519226,
-      "rewards/rejected": -0.130869060754776,
+      "learning_rate": 6.41315865106129e-08,
+      "logits/chosen": -1.909014344215393,
+      "logits/rejected": -1.919298768043518,
+      "logps/chosen": -31.879894256591797,
+      "logps/rejected": -35.34550857543945,
+      "loss": 0.6942,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.29231759905815125,
+      "rewards/margins": 0.34339430928230286,
+      "rewards/rejected": -0.051076728850603104,
       "step": 360
     },
     {
       "epoch": 0.96,
-      "grad_norm": 15.5625,
-      "learning_rate": 4.706303941965804e-06,
-      "logits/chosen": -1.9701541662216187,
-      "logits/rejected": -1.9697401523590088,
-      "logps/chosen": -32.81611633300781,
-      "logps/rejected": -36.49714279174805,
-      "loss": 0.6811,
-      "rewards/accuracies": 0.8374999761581421,
-      "rewards/chosen": 0.1886475384235382,
-      "rewards/margins": 0.3228580951690674,
-      "rewards/rejected": -0.1342105269432068,
+      "learning_rate": 2.3150941078050325e-08,
+      "logits/chosen": -2.0480055809020996,
+      "logits/rejected": -2.0415189266204834,
+      "logps/chosen": -33.37665939331055,
+      "logps/rejected": -29.24251937866211,
+      "loss": 0.7583,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.24777868390083313,
+      "rewards/margins": 0.262265145778656,
+      "rewards/rejected": -0.014486486092209816,
       "step": 370
     },
     {
       "epoch": 0.99,
-      "grad_norm": 14.25,
-      "learning_rate": 4.679090796681225e-06,
-      "logits/chosen": -1.9989287853240967,
-      "logits/rejected": -1.9943631887435913,
-      "logps/chosen": -30.109970092773438,
-      "logps/rejected": -29.691638946533203,
-      "loss": 0.697,
-      "rewards/accuracies": 0.8374999761581421,
-      "rewards/chosen": 0.1801886111497879,
-      "rewards/margins": 0.3235127031803131,
-      "rewards/rejected": -0.1433240920305252,
+      "learning_rate": 2.575864278703266e-09,
+      "logits/chosen": -1.9075886011123657,
+      "logits/rejected": -1.9097877740859985,
+      "logps/chosen": -33.8558464050293,
+      "logps/rejected": -30.982568740844727,
+      "loss": 0.6567,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.3448841869831085,
+      "rewards/margins": 0.44383174180984497,
+      "rewards/rejected": -0.09894753992557526,
       "step": 380
     },
     {
-      "epoch": 1.01,
-      "grad_norm": 17.0,
-      "learning_rate": 4.650758136138454e-06,
-      "logits/chosen": -1.7691841125488281,
-      "logits/rejected": -1.775526762008667,
-      "logps/chosen": -31.6003475189209,
-      "logps/rejected": -36.804012298583984,
-      "loss": 0.5984,
-      "rewards/accuracies": 0.862500011920929,
-      "rewards/chosen": 0.23561076819896698,
-      "rewards/margins": 0.4413274824619293,
-      "rewards/rejected": -0.20571665465831757,
-      "step": 390
-    },
-    {
-      "epoch": 1.04,
-      "grad_norm": 15.0,
-      "learning_rate": 4.621320516337559e-06,
-      "logits/chosen": -1.91928231716156,
-      "logits/rejected": -1.9130380153656006,
-      "logps/chosen": -32.96872329711914,
-      "logps/rejected": -32.79420471191406,
-      "loss": 0.6314,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": 0.2591918110847473,
-      "rewards/margins": 0.40479007363319397,
-      "rewards/rejected": -0.14559821784496307,
-      "step": 400
-    },
-    {
-      "epoch": 1.04,
-      "eval_logits/chosen": -2.1853952407836914,
-      "eval_logits/rejected": -2.180595636367798,
-      "eval_logps/chosen": -34.25471115112305,
-      "eval_logps/rejected": -37.85035705566406,
-      "eval_loss": 0.9658516049385071,
-      "eval_rewards/accuracies": 0.5332226157188416,
-      "eval_rewards/chosen": -0.06604895740747452,
-      "eval_rewards/margins": 0.034071628004312515,
-      "eval_rewards/rejected": -0.10012059658765793,
-      "eval_runtime": 145.8361,
-      "eval_samples_per_second": 2.352,
-      "eval_steps_per_second": 0.295,
-      "step": 400
-    },
-    {
-      "epoch": 1.06,
-      "grad_norm": 11.375,
-      "learning_rate": 4.590793060955158e-06,
-      "logits/chosen": -1.9149179458618164,
-      "logits/rejected": -1.922204613685608,
-      "logps/chosen": -28.569345474243164,
-      "logps/rejected": -29.881052017211914,
-      "loss": 0.6408,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.16530932486057281,
-      "rewards/margins": 0.40015679597854614,
-      "rewards/rejected": -0.23484745621681213,
-      "step": 410
-    },
-    {
-      "epoch": 1.09,
-      "grad_norm": 14.5,
-      "learning_rate": 4.559191453574582e-06,
-      "logits/chosen": -1.931968092918396,
-      "logits/rejected": -1.9309993982315063,
-      "logps/chosen": -33.57292938232422,
-      "logps/rejected": -31.347585678100586,
-      "loss": 0.7222,
-      "rewards/accuracies": 0.7749999761581421,
-      "rewards/chosen": 0.2033342868089676,
-      "rewards/margins": 0.3246961236000061,
-      "rewards/rejected": -0.1213618740439415,
-      "step": 420
-    },
-    {
-      "epoch": 1.12,
-      "grad_norm": 12.9375,
-      "learning_rate": 4.52653192962838e-06,
-      "logits/chosen": -1.923240303993225,
-      "logits/rejected": -1.906058669090271,
-      "logps/chosen": -30.46048355102539,
-      "logps/rejected": -33.793663024902344,
-      "loss": 0.6241,
-      "rewards/accuracies": 0.8374999761581421,
-      "rewards/chosen": 0.15501730144023895,
-      "rewards/margins": 0.4152299761772156,
-      "rewards/rejected": -0.26021263003349304,
-      "step": 430
-    },
-    {
-      "epoch": 1.14,
-      "grad_norm": 14.4375,
-      "learning_rate": 4.492831268057307e-06,
-      "logits/chosen": -1.9517921209335327,
-      "logits/rejected": -1.9540119171142578,
-      "logps/chosen": -35.7196044921875,
-      "logps/rejected": -35.67481231689453,
-      "loss": 0.5249,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/chosen": 0.26782849431037903,
-      "rewards/margins": 0.5122817158699036,
-      "rewards/rejected": -0.24445319175720215,
-      "step": 440
-    },
-    {
-      "epoch": 1.17,
-      "grad_norm": 14.5,
-      "learning_rate": 4.458106782690094e-06,
-      "logits/chosen": -2.025313138961792,
-      "logits/rejected": -2.0250680446624756,
-      "logps/chosen": -31.915420532226562,
-      "logps/rejected": -34.20271301269531,
-      "loss": 0.6338,
-      "rewards/accuracies": 0.7875000238418579,
-      "rewards/chosen": 0.21537843346595764,
-      "rewards/margins": 0.40815800428390503,
-      "rewards/rejected": -0.192779541015625,
-      "step": 450
-    },
-    {
-      "epoch": 1.19,
-      "grad_norm": 14.1875,
-      "learning_rate": 4.422376313348405e-06,
-      "logits/chosen": -1.966787576675415,
-      "logits/rejected": -1.959297776222229,
-      "logps/chosen": -31.532024383544922,
-      "logps/rejected": -36.94917678833008,
-      "loss": 0.5198,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 0.26218029856681824,
-      "rewards/margins": 0.530721127986908,
-      "rewards/rejected": -0.2685408294200897,
-      "step": 460
-    },
-    {
-      "epoch": 1.22,
-      "grad_norm": 12.9375,
-      "learning_rate": 4.3856582166815696e-06,
-      "logits/chosen": -1.8692712783813477,
-      "logits/rejected": -1.8658641576766968,
-      "logps/chosen": -33.09648895263672,
-      "logps/rejected": -33.70581817626953,
-      "loss": 0.5796,
-      "rewards/accuracies": 0.8125,
-      "rewards/chosen": 0.25074246525764465,
-      "rewards/margins": 0.4853232502937317,
-      "rewards/rejected": -0.23458079993724823,
-      "step": 470
-    },
-    {
-      "epoch": 1.25,
-      "grad_norm": 12.875,
-      "learning_rate": 4.347971356735789e-06,
-      "logits/chosen": -1.9908945560455322,
-      "logits/rejected": -1.9839731454849243,
-      "logps/chosen": -30.299631118774414,
-      "logps/rejected": -32.82419967651367,
-      "loss": 0.6571,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.15867193043231964,
-      "rewards/margins": 0.407886266708374,
-      "rewards/rejected": -0.24921436607837677,
-      "step": 480
-    },
-    {
-      "epoch": 1.27,
-      "grad_norm": 13.6875,
-      "learning_rate": 4.309335095262675e-06,
-      "logits/chosen": -1.9327518939971924,
-      "logits/rejected": -1.9343318939208984,
-      "logps/chosen": -34.46928024291992,
-      "logps/rejected": -34.476844787597656,
-      "loss": 0.5353,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 0.2651732265949249,
-      "rewards/margins": 0.5157736539840698,
-      "rewards/rejected": -0.2506003975868225,
-      "step": 490
-    },
-    {
-      "epoch": 1.3,
-      "grad_norm": 14.375,
-      "learning_rate": 4.269769281772082e-06,
-      "logits/chosen": -1.822686791419983,
-      "logits/rejected": -1.8202558755874634,
-      "logps/chosen": -32.31444549560547,
-      "logps/rejected": -37.581722259521484,
-      "loss": 0.5138,
-      "rewards/accuracies": 0.862500011920929,
-      "rewards/chosen": 0.25735390186309814,
-      "rewards/margins": 0.5925731062889099,
-      "rewards/rejected": -0.33521920442581177,
-      "step": 500
-    },
-    {
-      "epoch": 1.3,
-      "eval_logits/chosen": -2.1644699573516846,
-      "eval_logits/rejected": -2.1597023010253906,
-      "eval_logps/chosen": -34.53253936767578,
-      "eval_logps/rejected": -38.092079162597656,
-      "eval_loss": 0.9765273332595825,
-      "eval_rewards/accuracies": 0.5361295938491821,
-      "eval_rewards/chosen": -0.14939545094966888,
-      "eval_rewards/margins": 0.023241618648171425,
-      "eval_rewards/rejected": -0.17263707518577576,
-      "eval_runtime": 146.0192,
-      "eval_samples_per_second": 2.349,
-      "eval_steps_per_second": 0.294,
-      "step": 500
-    },
-    {
-      "epoch": 1.32,
-      "grad_norm": 10.875,
-      "learning_rate": 4.22929424333435e-06,
-      "logits/chosen": -1.9213244915008545,
-      "logits/rejected": -1.9261335134506226,
-      "logps/chosen": -32.475154876708984,
-      "logps/rejected": -32.32494354248047,
-      "loss": 0.5603,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": 0.29466065764427185,
-      "rewards/margins": 0.49831128120422363,
-      "rewards/rejected": -0.20365062355995178,
-      "step": 510
-    },
-    {
-      "epoch": 1.35,
-      "grad_norm": 13.0625,
-      "learning_rate": 4.1879307741372085e-06,
-      "logits/chosen": -1.9467157125473022,
-      "logits/rejected": -1.9576940536499023,
-      "logps/chosen": -30.570581436157227,
-      "logps/rejected": -32.42475509643555,
-      "loss": 0.5459,
-      "rewards/accuracies": 0.8125,
-      "rewards/chosen": 0.24594974517822266,
-      "rewards/margins": 0.5210699439048767,
-      "rewards/rejected": -0.27512019872665405,
-      "step": 520
-    },
-    {
-      "epoch": 1.38,
-      "grad_norm": 8.25,
-      "learning_rate": 4.145700124802693e-06,
-      "logits/chosen": -1.8762938976287842,
-      "logits/rejected": -1.8731403350830078,
-      "logps/chosen": -31.770130157470703,
-      "logps/rejected": -33.37158966064453,
-      "loss": 0.5711,
-      "rewards/accuracies": 0.862500011920929,
-      "rewards/chosen": 0.19511549174785614,
-      "rewards/margins": 0.49776020646095276,
-      "rewards/rejected": -0.30264467000961304,
-      "step": 530
-    },
-    {
-      "epoch": 1.4,
-      "grad_norm": 13.125,
-      "learning_rate": 4.102623991469562e-06,
-      "logits/chosen": -1.7426817417144775,
-      "logits/rejected": -1.7519454956054688,
-      "logps/chosen": -31.739831924438477,
-      "logps/rejected": -32.71033477783203,
-      "loss": 0.5853,
-      "rewards/accuracies": 0.7875000238418579,
-      "rewards/chosen": 0.24886760115623474,
-      "rewards/margins": 0.5051722526550293,
-      "rewards/rejected": -0.25630468130111694,
-      "step": 540
-    },
-    {
-      "epoch": 1.43,
-      "grad_norm": 12.5625,
-      "learning_rate": 4.058724504646834e-06,
-      "logits/chosen": -1.8341865539550781,
-      "logits/rejected": -1.8280948400497437,
-      "logps/chosen": -32.86029815673828,
-      "logps/rejected": -31.735515594482422,
-      "loss": 0.5944,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.254402756690979,
-      "rewards/margins": 0.47765302658081055,
-      "rewards/rejected": -0.22325026988983154,
-      "step": 550
-    },
-    {
-      "epoch": 1.45,
-      "grad_norm": 10.25,
-      "learning_rate": 4.014024217844167e-06,
-      "logits/chosen": -1.9202044010162354,
-      "logits/rejected": -1.918296217918396,
-      "logps/chosen": -33.7692985534668,
-      "logps/rejected": -32.30878448486328,
-      "loss": 0.5946,
-      "rewards/accuracies": 0.8374999761581421,
-      "rewards/chosen": 0.24056951701641083,
-      "rewards/margins": 0.48398923873901367,
-      "rewards/rejected": -0.24341969192028046,
-      "step": 560
-    },
-    {
-      "epoch": 1.48,
-      "grad_norm": 14.875,
-      "learning_rate": 3.968546095984911e-06,
-      "logits/chosen": -1.7542064189910889,
-      "logits/rejected": -1.7522131204605103,
-      "logps/chosen": -32.00287628173828,
-      "logps/rejected": -31.731969833374023,
-      "loss": 0.5988,
-      "rewards/accuracies": 0.8125,
-      "rewards/chosen": 0.23806098103523254,
-      "rewards/margins": 0.45696067810058594,
-      "rewards/rejected": -0.2188996523618698,
-      "step": 570
-    },
-    {
-      "epoch": 1.51,
-      "grad_norm": 19.75,
-      "learning_rate": 3.922313503607806e-06,
-      "logits/chosen": -1.8884613513946533,
-      "logits/rejected": -1.8850774765014648,
-      "logps/chosen": -30.179805755615234,
-      "logps/rejected": -35.7034797668457,
-      "loss": 0.5011,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 0.27745240926742554,
-      "rewards/margins": 0.6006461381912231,
-      "rewards/rejected": -0.3231937289237976,
-      "step": 580
-    },
-    {
-      "epoch": 1.53,
-      "grad_norm": 11.1875,
-      "learning_rate": 3.875350192863368e-06,
-      "logits/chosen": -1.8237355947494507,
-      "logits/rejected": -1.8271907567977905,
-      "logps/chosen": -28.952396392822266,
-      "logps/rejected": -31.399667739868164,
-      "loss": 0.6153,
-      "rewards/accuracies": 0.8125,
-      "rewards/chosen": 0.19879142940044403,
-      "rewards/margins": 0.42103099822998047,
-      "rewards/rejected": -0.22223956882953644,
-      "step": 590
-    },
-    {
-      "epoch": 1.56,
-      "grad_norm": 12.1875,
-      "learning_rate": 3.8276802913111436e-06,
-      "logits/chosen": -1.8603225946426392,
-      "logits/rejected": -1.8602790832519531,
-      "logps/chosen": -31.146831512451172,
-      "logps/rejected": -32.076019287109375,
-      "loss": 0.5152,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/chosen": 0.32240957021713257,
-      "rewards/margins": 0.5873108506202698,
-      "rewards/rejected": -0.2649012804031372,
-      "step": 600
-    },
-    {
-      "epoch": 1.56,
-      "eval_logits/chosen": -2.1365857124328613,
-      "eval_logits/rejected": -2.131840229034424,
-      "eval_logps/chosen": -34.447776794433594,
-      "eval_logps/rejected": -38.098880767822266,
-      "eval_loss": 0.9495607614517212,
-      "eval_rewards/accuracies": 0.5884551405906677,
-      "eval_rewards/chosen": -0.12396734952926636,
-      "eval_rewards/margins": 0.05070924386382103,
-      "eval_rewards/rejected": -0.17467659711837769,
-      "eval_runtime": 145.8984,
-      "eval_samples_per_second": 2.351,
-      "eval_steps_per_second": 0.295,
-      "step": 600
-    },
-    {
-      "epoch": 1.58,
-      "grad_norm": 14.4375,
-      "learning_rate": 3.7793282895240927e-06,
-      "logits/chosen": -1.9220027923583984,
-      "logits/rejected": -1.9227746725082397,
-      "logps/chosen": -33.91392135620117,
-      "logps/rejected": -34.14142608642578,
-      "loss": 0.4821,
-      "rewards/accuracies": 0.8374999761581421,
-      "rewards/chosen": 0.334866464138031,
-      "rewards/margins": 0.6520383358001709,
-      "rewards/rejected": -0.3171718716621399,
-      "step": 610
-    },
-    {
-      "epoch": 1.61,
-      "grad_norm": 12.8125,
-      "learning_rate": 3.730319028506478e-06,
-      "logits/chosen": -1.8861291408538818,
-      "logits/rejected": -1.8837579488754272,
-      "logps/chosen": -32.04036331176758,
-      "logps/rejected": -33.019126892089844,
-      "loss": 0.5179,
-      "rewards/accuracies": 0.862500011920929,
-      "rewards/chosen": 0.32551074028015137,
-      "rewards/margins": 0.6102737784385681,
-      "rewards/rejected": -0.284762978553772,
-      "step": 620
-    },
-    {
-      "epoch": 1.64,
-      "grad_norm": 16.625,
-      "learning_rate": 3.6806776869317074e-06,
-      "logits/chosen": -1.8974058628082275,
-      "logits/rejected": -1.8884598016738892,
-      "logps/chosen": -31.70094871520996,
-      "logps/rejected": -31.9660701751709,
-      "loss": 0.5081,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 0.32255104184150696,
-      "rewards/margins": 0.6317622065544128,
-      "rewards/rejected": -0.30921119451522827,
-      "step": 630
-    },
-    {
-      "epoch": 1.66,
-      "grad_norm": 14.9375,
-      "learning_rate": 3.6304297682067146e-06,
-      "logits/chosen": -1.8995754718780518,
-      "logits/rejected": -1.8962815999984741,
-      "logps/chosen": -31.30817222595215,
-      "logps/rejected": -33.31178283691406,
-      "loss": 0.5456,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 0.2670743465423584,
-      "rewards/margins": 0.5347846150398254,
-      "rewards/rejected": -0.26771026849746704,
-      "step": 640
-    },
-    {
-      "epoch": 1.69,
-      "grad_norm": 15.25,
-      "learning_rate": 3.579601087369492e-06,
-      "logits/chosen": -1.9088016748428345,
-      "logits/rejected": -1.9110372066497803,
-      "logps/chosen": -32.70873260498047,
-      "logps/rejected": -34.64277648925781,
-      "loss": 0.4902,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 0.283491849899292,
-      "rewards/margins": 0.571306586265564,
-      "rewards/rejected": -0.28781479597091675,
-      "step": 650
-    },
-    {
-      "epoch": 1.71,
-      "grad_norm": 18.625,
-      "learning_rate": 3.5282177578265295e-06,
-      "logits/chosen": -1.8186346292495728,
-      "logits/rejected": -1.8190486431121826,
-      "logps/chosen": -32.91853332519531,
-      "logps/rejected": -32.44306564331055,
-      "loss": 0.501,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 0.30848973989486694,
-      "rewards/margins": 0.5592333078384399,
-      "rewards/rejected": -0.250743567943573,
-      "step": 660
-    },
-    {
-      "epoch": 1.74,
-      "grad_norm": 16.5,
-      "learning_rate": 3.476306177936961e-06,
-      "logits/chosen": -1.85702383518219,
-      "logits/rejected": -1.8474591970443726,
-      "logps/chosen": -32.971839904785156,
-      "logps/rejected": -33.291786193847656,
-      "loss": 0.4717,
-      "rewards/accuracies": 0.862500011920929,
-      "rewards/chosen": 0.3088335692882538,
-      "rewards/margins": 0.6040312051773071,
-      "rewards/rejected": -0.29519766569137573,
-      "step": 670
-    },
-    {
-      "epoch": 1.77,
-      "grad_norm": 14.8125,
-      "learning_rate": 3.423893017450324e-06,
-      "logits/chosen": -1.7545194625854492,
-      "logits/rejected": -1.7513306140899658,
-      "logps/chosen": -30.304126739501953,
-      "logps/rejected": -35.3206901550293,
-      "loss": 0.4469,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/chosen": 0.34705036878585815,
-      "rewards/margins": 0.6476843357086182,
-      "rewards/rejected": -0.3006339371204376,
-      "step": 680
-    },
-    {
-      "epoch": 1.79,
-      "grad_norm": 10.75,
-      "learning_rate": 3.3710052038048794e-06,
-      "logits/chosen": -1.8158613443374634,
-      "logits/rejected": -1.815222144126892,
-      "logps/chosen": -33.98094940185547,
-      "logps/rejected": -36.47856140136719,
-      "loss": 0.3944,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/chosen": 0.3868010938167572,
-      "rewards/margins": 0.7242231965065002,
-      "rewards/rejected": -0.33742210268974304,
-      "step": 690
-    },
-    {
-      "epoch": 1.82,
-      "grad_norm": 12.0,
-      "learning_rate": 3.3176699082935546e-06,
-      "logits/chosen": -1.7871694564819336,
-      "logits/rejected": -1.7897698879241943,
-      "logps/chosen": -31.470829010009766,
-      "logps/rejected": -36.80913162231445,
-      "loss": 0.4499,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 0.33960574865341187,
-      "rewards/margins": 0.6898441314697266,
-      "rewards/rejected": -0.3502383828163147,
-      "step": 700
-    },
-    {
-      "epoch": 1.82,
-      "eval_logits/chosen": -2.118246078491211,
-      "eval_logits/rejected": -2.1135261058807373,
-      "eval_logps/chosen": -34.49834442138672,
-      "eval_logps/rejected": -38.18844985961914,
-      "eval_loss": 0.9377100467681885,
-      "eval_rewards/accuracies": 0.5772424936294556,
-      "eval_rewards/chosen": -0.13913682103157043,
-      "eval_rewards/margins": 0.062411729246377945,
-      "eval_rewards/rejected": -0.20154856145381927,
-      "eval_runtime": 145.9395,
-      "eval_samples_per_second": 2.35,
-      "eval_steps_per_second": 0.295,
-      "step": 700
-    },
-    {
-      "epoch": 1.84,
-      "grad_norm": 15.125,
-      "learning_rate": 3.2639145321045933e-06,
-      "logits/chosen": -1.892765998840332,
-      "logits/rejected": -1.8953781127929688,
-      "logps/chosen": -33.85210037231445,
-      "logps/rejected": -35.282066345214844,
-      "loss": 0.5218,
-      "rewards/accuracies": 0.824999988079071,
-      "rewards/chosen": 0.3130471706390381,
-      "rewards/margins": 0.6166204214096069,
-      "rewards/rejected": -0.30357325077056885,
-      "step": 710
-    },
-    {
-      "epoch": 1.87,
-      "grad_norm": 16.375,
-      "learning_rate": 3.2097666922441107e-06,
-      "logits/chosen": -1.7477344274520874,
-      "logits/rejected": -1.7418196201324463,
-      "logps/chosen": -33.75777816772461,
-      "logps/rejected": -33.40064239501953,
-      "loss": 0.4735,
-      "rewards/accuracies": 0.862500011920929,
-      "rewards/chosen": 0.32263556122779846,
-      "rewards/margins": 0.6372753381729126,
-      "rewards/rejected": -0.31463971734046936,
-      "step": 720
-    },
-    {
-      "epoch": 1.9,
-      "grad_norm": 10.125,
-      "learning_rate": 3.1552542073477554e-06,
-      "logits/chosen": -1.9213536977767944,
-      "logits/rejected": -1.918287992477417,
-      "logps/chosen": -29.665637969970703,
-      "logps/rejected": -32.720909118652344,
-      "loss": 0.4566,
-      "rewards/accuracies": 0.9125000238418579,
-      "rewards/chosen": 0.29620498418807983,
-      "rewards/margins": 0.6780200004577637,
-      "rewards/rejected": -0.38181501626968384,
-      "step": 730
-    },
-    {
-      "epoch": 1.92,
-      "grad_norm": 13.1875,
-      "learning_rate": 3.100405083388799e-06,
-      "logits/chosen": -1.7594985961914062,
-      "logits/rejected": -1.7596817016601562,
-      "logps/chosen": -32.198509216308594,
-      "logps/rejected": -38.66451644897461,
-      "loss": 0.4607,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 0.37939977645874023,
-      "rewards/margins": 0.6904138326644897,
-      "rewards/rejected": -0.3110140264034271,
-      "step": 740
-    },
-    {
-      "epoch": 1.95,
-      "grad_norm": 12.875,
-      "learning_rate": 3.0452474992899645e-06,
-      "logits/chosen": -1.6468909978866577,
-      "logits/rejected": -1.6521577835083008,
-      "logps/chosen": -35.7952766418457,
-      "logps/rejected": -35.151161193847656,
-      "loss": 0.4738,
-      "rewards/accuracies": 0.8374999761581421,
-      "rewards/chosen": 0.3411572575569153,
-      "rewards/margins": 0.6948047876358032,
-      "rewards/rejected": -0.3536476194858551,
-      "step": 750
-    },
-    {
-      "epoch": 1.97,
-      "grad_norm": 16.625,
-      "learning_rate": 2.989809792446417e-06,
-      "logits/chosen": -1.843528151512146,
-      "logits/rejected": -1.8452043533325195,
-      "logps/chosen": -31.52127456665039,
-      "logps/rejected": -33.655906677246094,
-      "loss": 0.4756,
-      "rewards/accuracies": 0.887499988079071,
-      "rewards/chosen": 0.32532256841659546,
-      "rewards/margins": 0.5883294939994812,
-      "rewards/rejected": -0.26300689578056335,
-      "step": 760
-    },
-    {
-      "epoch": 2.0,
-      "grad_norm": 15.625,
-      "learning_rate": 2.9341204441673267e-06,
-      "logits/chosen": -1.8192428350448608,
-      "logits/rejected": -1.8181209564208984,
-      "logps/chosen": -31.00497055053711,
-      "logps/rejected": -35.86052703857422,
-      "loss": 0.5041,
-      "rewards/accuracies": 0.8833333849906921,
-      "rewards/chosen": 0.3511923551559448,
-      "rewards/margins": 0.6009055972099304,
-      "rewards/rejected": -0.24971330165863037,
-      "step": 770
-    },
-    {
-      "epoch": 2.03,
-      "grad_norm": 11.1875,
-      "learning_rate": 2.878208065043501e-06,
-      "logits/chosen": -1.8188778162002563,
-      "logits/rejected": -1.8182384967803955,
-      "logps/chosen": -33.38768005371094,
-      "logps/rejected": -32.7451057434082,
-      "loss": 0.3513,
-      "rewards/accuracies": 0.9125000238418579,
-      "rewards/chosen": 0.37823373079299927,
-      "rewards/margins": 0.7759095430374146,
-      "rewards/rejected": -0.39767593145370483,
-      "step": 780
-    },
-    {
-      "epoch": 2.05,
-      "grad_norm": 11.75,
-      "learning_rate": 2.8221013802485974e-06,
-      "logits/chosen": -1.8471590280532837,
-      "logits/rejected": -1.845741629600525,
-      "logps/chosen": -28.308391571044922,
-      "logps/rejected": -34.1221809387207,
-      "loss": 0.2972,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.37643638253211975,
-      "rewards/margins": 0.8221960067749023,
-      "rewards/rejected": -0.4457596242427826,
-      "step": 790
-    },
-    {
-      "epoch": 2.08,
-      "grad_norm": 16.125,
-      "learning_rate": 2.76582921478147e-06,
-      "logits/chosen": -1.8991243839263916,
-      "logits/rejected": -1.8957576751708984,
-      "logps/chosen": -30.958560943603516,
-      "logps/rejected": -35.7558479309082,
-      "loss": 0.2662,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.4405938982963562,
-      "rewards/margins": 0.9076371192932129,
-      "rewards/rejected": -0.46704310178756714,
-      "step": 800
-    },
-    {
-      "epoch": 2.08,
-      "eval_logits/chosen": -2.105231285095215,
-      "eval_logits/rejected": -2.1005356311798096,
-      "eval_logps/chosen": -34.60075759887695,
-      "eval_logps/rejected": -38.31148910522461,
-      "eval_loss": 0.9329387545585632,
-      "eval_rewards/accuracies": 0.574335515499115,
-      "eval_rewards/chosen": -0.16986209154129028,
-      "eval_rewards/margins": 0.06859838962554932,
-      "eval_rewards/rejected": -0.2384604513645172,
-      "eval_runtime": 145.8821,
-      "eval_samples_per_second": 2.351,
-      "eval_steps_per_second": 0.295,
-      "step": 800
-    },
-    {
-      "epoch": 2.1,
-      "grad_norm": 8.5625,
-      "learning_rate": 2.7094204786572254e-06,
-      "logits/chosen": -1.7284854650497437,
-      "logits/rejected": -1.7209049463272095,
-      "logps/chosen": -32.67493438720703,
-      "logps/rejected": -36.15032196044922,
-      "loss": 0.249,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.5343915224075317,
-      "rewards/margins": 1.0114519596099854,
-      "rewards/rejected": -0.4770604074001312,
-      "step": 810
-    },
-    {
-      "epoch": 2.13,
-      "grad_norm": 11.9375,
-      "learning_rate": 2.6529041520546072e-06,
-      "logits/chosen": -1.8052905797958374,
-      "logits/rejected": -1.815611481666565,
-      "logps/chosen": -34.110530853271484,
-      "logps/rejected": -33.74187088012695,
-      "loss": 0.3341,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.47554850578308105,
-      "rewards/margins": 0.8686782121658325,
-      "rewards/rejected": -0.3931296765804291,
-      "step": 820
-    },
-    {
-      "epoch": 2.16,
-      "grad_norm": 12.6875,
-      "learning_rate": 2.5963092704273302e-06,
-      "logits/chosen": -1.8559843301773071,
-      "logits/rejected": -1.8607652187347412,
-      "logps/chosen": -33.57707214355469,
-      "logps/rejected": -30.534915924072266,
-      "loss": 0.3746,
-      "rewards/accuracies": 0.887499988079071,
-      "rewards/chosen": 0.3977125585079193,
-      "rewards/margins": 0.7777142524719238,
-      "rewards/rejected": -0.3800017237663269,
-      "step": 830
-    },
-    {
-      "epoch": 2.18,
-      "grad_norm": 11.375,
-      "learning_rate": 2.53966490958702e-06,
-      "logits/chosen": -1.8550090789794922,
-      "logits/rejected": -1.8633220195770264,
-      "logps/chosen": -33.31046676635742,
-      "logps/rejected": -31.336292266845703,
-      "loss": 0.3163,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.4432702958583832,
-      "rewards/margins": 0.8898479342460632,
-      "rewards/rejected": -0.44657760858535767,
-      "step": 840
-    },
-    {
-      "epoch": 2.21,
-      "grad_norm": 12.5,
-      "learning_rate": 2.4830001707654135e-06,
-      "logits/chosen": -1.7852023839950562,
-      "logits/rejected": -1.7757289409637451,
-      "logps/chosen": -30.468114852905273,
-      "logps/rejected": -33.35639953613281,
-      "loss": 0.333,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.3975241780281067,
-      "rewards/margins": 0.8003025054931641,
-      "rewards/rejected": -0.402778297662735,
-      "step": 850
-    },
-    {
-      "epoch": 2.23,
-      "grad_norm": 9.3125,
-      "learning_rate": 2.4263441656635054e-06,
-      "logits/chosen": -1.922766923904419,
-      "logits/rejected": -1.9129383563995361,
-      "logps/chosen": -24.774024963378906,
-      "logps/rejected": -31.40387535095215,
-      "loss": 0.3162,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.43501368165016174,
-      "rewards/margins": 0.8944748640060425,
-      "rewards/rejected": -0.4594612121582031,
-      "step": 860
-    },
-    {
-      "epoch": 2.26,
-      "grad_norm": 10.9375,
-      "learning_rate": 2.3697260014953107e-06,
-      "logits/chosen": -1.7784268856048584,
-      "logits/rejected": -1.7797539234161377,
-      "logps/chosen": -32.7981071472168,
-      "logps/rejected": -31.451351165771484,
-      "loss": 0.2814,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.460468053817749,
-      "rewards/margins": 0.9011228680610657,
-      "rewards/rejected": -0.4406547546386719,
-      "step": 870
-    },
-    {
-      "epoch": 2.29,
-      "grad_norm": 16.625,
-      "learning_rate": 2.3131747660339396e-06,
-      "logits/chosen": -1.7835657596588135,
-      "logits/rejected": -1.7845604419708252,
-      "logps/chosen": -31.24370765686035,
-      "logps/rejected": -34.56098937988281,
-      "loss": 0.382,
-      "rewards/accuracies": 0.8374999761581421,
-      "rewards/chosen": 0.3436630368232727,
-      "rewards/margins": 0.7966837882995605,
-      "rewards/rejected": -0.45302075147628784,
-      "step": 880
-    },
-    {
-      "epoch": 2.31,
-      "grad_norm": 9.375,
-      "learning_rate": 2.256719512667651e-06,
-      "logits/chosen": -1.6883900165557861,
-      "logits/rejected": -1.6869096755981445,
-      "logps/chosen": -34.208251953125,
-      "logps/rejected": -37.77946090698242,
-      "loss": 0.2912,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.43700018525123596,
-      "rewards/margins": 1.0076673030853271,
-      "rewards/rejected": -0.570667028427124,
-      "step": 890
-    },
-    {
-      "epoch": 2.34,
-      "grad_norm": 12.0625,
-      "learning_rate": 2.2003892454735786e-06,
-      "logits/chosen": -1.8402154445648193,
-      "logits/rejected": -1.8334258794784546,
-      "logps/chosen": -30.50925636291504,
-      "logps/rejected": -34.355873107910156,
-      "loss": 0.2981,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.44801297783851624,
-      "rewards/margins": 0.9273546934127808,
-      "rewards/rejected": -0.47934168577194214,
-      "step": 900
-    },
-    {
-      "epoch": 2.34,
-      "eval_logits/chosen": -2.0982792377471924,
-      "eval_logits/rejected": -2.093583822250366,
-      "eval_logps/chosen": -34.59360122680664,
-      "eval_logps/rejected": -38.31861877441406,
-      "eval_loss": 0.9285210967063904,
-      "eval_rewards/accuracies": 0.5539867281913757,
-      "eval_rewards/chosen": -0.1677148938179016,
-      "eval_rewards/margins": 0.07288341969251633,
-      "eval_rewards/rejected": -0.24059830605983734,
-      "eval_runtime": 145.7917,
-      "eval_samples_per_second": 2.353,
-      "eval_steps_per_second": 0.295,
-      "step": 900
-    },
-    {
-      "epoch": 2.36,
-      "grad_norm": 11.875,
-      "learning_rate": 2.1442129043167877e-06,
-      "logits/chosen": -1.875894546508789,
-      "logits/rejected": -1.8712823390960693,
-      "logps/chosen": -32.074310302734375,
-      "logps/rejected": -36.22045135498047,
-      "loss": 0.2786,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.44501152634620667,
-      "rewards/margins": 0.9525611996650696,
-      "rewards/rejected": -0.5075496435165405,
-      "step": 910
-    },
-    {
-      "epoch": 2.39,
-      "grad_norm": 9.875,
-      "learning_rate": 2.088219349982323e-06,
-      "logits/chosen": -1.7960087060928345,
-      "logits/rejected": -1.8010889291763306,
-      "logps/chosen": -33.7156867980957,
-      "logps/rejected": -34.15389633178711,
-      "loss": 0.275,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.48986950516700745,
-      "rewards/margins": 0.9860662221908569,
-      "rewards/rejected": -0.49619680643081665,
-      "step": 920
-    },
-    {
-      "epoch": 2.42,
-      "grad_norm": 7.5,
-      "learning_rate": 2.0324373493478803e-06,
-      "logits/chosen": -1.8964351415634155,
-      "logits/rejected": -1.887263536453247,
-      "logps/chosen": -30.813091278076172,
-      "logps/rejected": -36.062828063964844,
-      "loss": 0.2684,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.4415518641471863,
-      "rewards/margins": 0.9275605082511902,
-      "rewards/rejected": -0.4860086441040039,
-      "step": 930
-    },
-    {
-      "epoch": 2.44,
-      "grad_norm": 13.75,
-      "learning_rate": 1.976895560604729e-06,
-      "logits/chosen": -1.824550986289978,
-      "logits/rejected": -1.8213589191436768,
-      "logps/chosen": -30.109264373779297,
-      "logps/rejected": -33.882484436035156,
-      "loss": 0.385,
-      "rewards/accuracies": 0.9125000238418579,
-      "rewards/chosen": 0.35375675559043884,
-      "rewards/margins": 0.7788890600204468,
-      "rewards/rejected": -0.4251323342323303,
-      "step": 940
-    },
-    {
-      "epoch": 2.47,
-      "grad_norm": 9.0625,
-      "learning_rate": 1.921622518534466e-06,
-      "logits/chosen": -1.7633476257324219,
-      "logits/rejected": -1.7706598043441772,
-      "logps/chosen": -31.719341278076172,
-      "logps/rejected": -37.02296447753906,
-      "loss": 0.3799,
-      "rewards/accuracies": 0.887499988079071,
-      "rewards/chosen": 0.34551283717155457,
-      "rewards/margins": 0.7874213457107544,
-      "rewards/rejected": -0.4419085383415222,
-      "step": 950
-    },
-    {
-      "epoch": 2.49,
-      "grad_norm": 11.8125,
-      "learning_rate": 1.8666466198491794e-06,
-      "logits/chosen": -1.775905966758728,
-      "logits/rejected": -1.7690013647079468,
-      "logps/chosen": -32.38079071044922,
-      "logps/rejected": -38.34041213989258,
-      "loss": 0.272,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.4343179166316986,
-      "rewards/margins": 0.9567987322807312,
-      "rewards/rejected": -0.522480845451355,
-      "step": 960
-    },
-    {
-      "epoch": 2.52,
-      "grad_norm": 12.5,
-      "learning_rate": 1.8119961086025376e-06,
-      "logits/chosen": -1.7786152362823486,
-      "logits/rejected": -1.7790699005126953,
-      "logps/chosen": -29.682214736938477,
-      "logps/rejected": -34.35015106201172,
-      "loss": 0.3109,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.4662559926509857,
-      "rewards/margins": 0.8769238591194153,
-      "rewards/rejected": -0.4106678068637848,
-      "step": 970
-    },
-    {
-      "epoch": 2.55,
-      "grad_norm": 11.375,
-      "learning_rate": 1.7576990616793139e-06,
-      "logits/chosen": -1.7818208932876587,
-      "logits/rejected": -1.7929811477661133,
-      "logps/chosen": -31.38067054748535,
-      "logps/rejected": -35.58013153076172,
-      "loss": 0.2729,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.5088068246841431,
-      "rewards/margins": 0.9434831738471985,
-      "rewards/rejected": -0.43467625975608826,
-      "step": 980
-    },
-    {
-      "epoch": 2.57,
-      "grad_norm": 8.8125,
-      "learning_rate": 1.7037833743707892e-06,
-      "logits/chosen": -1.8516921997070312,
-      "logits/rejected": -1.849347710609436,
-      "logps/chosen": -34.72385025024414,
-      "logps/rejected": -33.69477081298828,
-      "loss": 0.3997,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 0.41916027665138245,
-      "rewards/margins": 0.8078235387802124,
-      "rewards/rejected": -0.3886633515357971,
-      "step": 990
-    },
-    {
-      "epoch": 2.6,
-      "grad_norm": 11.3125,
-      "learning_rate": 1.6502767460434588e-06,
-      "logits/chosen": -1.8244616985321045,
-      "logits/rejected": -1.828687310218811,
-      "logps/chosen": -33.12670135498047,
-      "logps/rejected": -35.82239532470703,
-      "loss": 0.3625,
-      "rewards/accuracies": 0.887499988079071,
-      "rewards/chosen": 0.4009768068790436,
-      "rewards/margins": 0.8018827438354492,
-      "rewards/rejected": -0.4009060263633728,
-      "step": 1000
-    },
-    {
-      "epoch": 2.6,
-      "eval_logits/chosen": -2.0979790687561035,
-      "eval_logits/rejected": -2.0932867527008057,
-      "eval_logps/chosen": -34.648372650146484,
-      "eval_logps/rejected": -38.36067199707031,
-      "eval_loss": 0.9330434799194336,
-      "eval_rewards/accuracies": 0.5975913405418396,
-      "eval_rewards/chosen": -0.184145987033844,
-      "eval_rewards/margins": 0.06907007843255997,
-      "eval_rewards/rejected": -0.25321605801582336,
-      "eval_runtime": 146.1037,
-      "eval_samples_per_second": 2.348,
-      "eval_steps_per_second": 0.294,
-      "step": 1000
-    },
-    {
-      "epoch": 2.62,
-      "grad_norm": 14.8125,
-      "learning_rate": 1.5972066659083796e-06,
-      "logits/chosen": -1.8353703022003174,
-      "logits/rejected": -1.8402354717254639,
-      "logps/chosen": -31.5875186920166,
-      "logps/rejected": -33.825382232666016,
-      "loss": 0.4105,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 0.36387085914611816,
-      "rewards/margins": 0.7551089525222778,
-      "rewards/rejected": -0.39123815298080444,
-      "step": 1010
-    },
-    {
-      "epoch": 2.65,
-      "grad_norm": 13.75,
-      "learning_rate": 1.5446003988985041e-06,
-      "logits/chosen": -1.793535828590393,
-      "logits/rejected": -1.7959201335906982,
-      "logps/chosen": -29.006155014038086,
-      "logps/rejected": -33.393253326416016,
-      "loss": 0.2527,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.4318457543849945,
-      "rewards/margins": 0.9400718808174133,
-      "rewards/rejected": -0.5082260966300964,
-      "step": 1020
-    },
-    {
-      "epoch": 2.68,
-      "grad_norm": 12.375,
-      "learning_rate": 1.4924849716612211e-06,
-      "logits/chosen": -1.846716284751892,
-      "logits/rejected": -1.8414628505706787,
-      "logps/chosen": -32.6993522644043,
-      "logps/rejected": -35.04628372192383,
-      "loss": 0.3421,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.3869279623031616,
-      "rewards/margins": 0.845413863658905,
-      "rewards/rejected": -0.4584859013557434,
-      "step": 1030
-    },
-    {
-      "epoch": 2.7,
-      "grad_norm": 12.875,
-      "learning_rate": 1.440887158673332e-06,
-      "logits/chosen": -1.7852039337158203,
-      "logits/rejected": -1.789275884628296,
-      "logps/chosen": -34.4805793762207,
-      "logps/rejected": -35.91188049316406,
-      "loss": 0.3291,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.3819444179534912,
-      "rewards/margins": 0.8296347856521606,
-      "rewards/rejected": -0.44769033789634705,
-      "step": 1040
-    },
-    {
-      "epoch": 2.73,
-      "grad_norm": 11.5,
-      "learning_rate": 1.3898334684855647e-06,
-      "logits/chosen": -1.782046914100647,
-      "logits/rejected": -1.794843077659607,
-      "logps/chosen": -30.523595809936523,
-      "logps/rejected": -34.53136444091797,
-      "loss": 0.3096,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.4017874598503113,
-      "rewards/margins": 0.8551858067512512,
-      "rewards/rejected": -0.4533984065055847,
-      "step": 1050
-    },
-    {
-      "epoch": 2.75,
-      "grad_norm": 12.5625,
-      "learning_rate": 1.3393501301037245e-06,
-      "logits/chosen": -1.7690681219100952,
-      "logits/rejected": -1.7623764276504517,
-      "logps/chosen": -30.654958724975586,
-      "logps/rejected": -34.75595474243164,
-      "loss": 0.2791,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.4513107240200043,
-      "rewards/margins": 0.8827075958251953,
-      "rewards/rejected": -0.43139687180519104,
-      "step": 1060
-    },
-    {
-      "epoch": 2.78,
-      "grad_norm": 13.1875,
-      "learning_rate": 1.2894630795134454e-06,
-      "logits/chosen": -1.8938472270965576,
-      "logits/rejected": -1.8946069478988647,
-      "logps/chosen": -32.11028289794922,
-      "logps/rejected": -34.25971603393555,
-      "loss": 0.2991,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.431039035320282,
-      "rewards/margins": 0.8555458188056946,
-      "rewards/rejected": -0.4245067536830902,
-      "step": 1070
-    },
-    {
-      "epoch": 2.81,
-      "grad_norm": 12.5625,
-      "learning_rate": 1.2401979463554984e-06,
-      "logits/chosen": -1.8250017166137695,
-      "logits/rejected": -1.8238385915756226,
-      "logps/chosen": -32.814945220947266,
-      "logps/rejected": -34.437644958496094,
-      "loss": 0.3938,
-      "rewards/accuracies": 0.862500011920929,
-      "rewards/chosen": 0.38246551156044006,
-      "rewards/margins": 0.756209671497345,
-      "rewards/rejected": -0.37374407052993774,
-      "step": 1080
-    },
-    {
-      "epoch": 2.83,
-      "grad_norm": 10.8125,
-      "learning_rate": 1.1915800407584705e-06,
-      "logits/chosen": -1.848353385925293,
-      "logits/rejected": -1.8408327102661133,
-      "logps/chosen": -32.76283645629883,
-      "logps/rejected": -32.30727767944336,
-      "loss": 0.3359,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.438912570476532,
-      "rewards/margins": 0.8186356425285339,
-      "rewards/rejected": -0.3797229826450348,
-      "step": 1090
-    },
-    {
-      "epoch": 2.86,
-      "grad_norm": 9.9375,
-      "learning_rate": 1.1436343403356019e-06,
-      "logits/chosen": -1.8205063343048096,
-      "logits/rejected": -1.8194630146026611,
-      "logps/chosen": -33.94606399536133,
-      "logps/rejected": -37.99333953857422,
-      "loss": 0.2579,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.4417056143283844,
-      "rewards/margins": 0.9338957071304321,
-      "rewards/rejected": -0.49219006299972534,
-      "step": 1100
-    },
-    {
-      "epoch": 2.86,
-      "eval_logits/chosen": -2.09816837310791,
-      "eval_logits/rejected": -2.093458414077759,
-      "eval_logps/chosen": -34.628299713134766,
-      "eval_logps/rejected": -38.356414794921875,
-      "eval_loss": 0.9268556833267212,
-      "eval_rewards/accuracies": 0.5830564498901367,
-      "eval_rewards/chosen": -0.1781238317489624,
-      "eval_rewards/margins": 0.0738147422671318,
-      "eval_rewards/rejected": -0.2519385516643524,
-      "eval_runtime": 145.9377,
-      "eval_samples_per_second": 2.35,
-      "eval_steps_per_second": 0.295,
-      "step": 1100
-    },
-    {
-      "epoch": 2.88,
-      "grad_norm": 14.125,
-      "learning_rate": 1.0963854773524548e-06,
-      "logits/chosen": -1.8701127767562866,
-      "logits/rejected": -1.8755674362182617,
-      "logps/chosen": -34.26189041137695,
-      "logps/rejected": -37.125205993652344,
-      "loss": 0.3524,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/chosen": 0.3615680932998657,
-      "rewards/margins": 0.8360538482666016,
-      "rewards/rejected": -0.47448572516441345,
-      "step": 1110
-    },
-    {
-      "epoch": 2.91,
-      "grad_norm": 11.1875,
-      "learning_rate": 1.049857726072005e-06,
-      "logits/chosen": -1.8384907245635986,
-      "logits/rejected": -1.8364282846450806,
-      "logps/chosen": -31.351863861083984,
-      "logps/rejected": -34.544898986816406,
-      "loss": 0.3331,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.3793022334575653,
-      "rewards/margins": 0.8479455709457397,
-      "rewards/rejected": -0.4686434268951416,
-      "step": 1120
-    },
-    {
-      "epoch": 2.94,
-      "grad_norm": 8.3125,
-      "learning_rate": 1.0040749902836508e-06,
-      "logits/chosen": -1.7408651113510132,
-      "logits/rejected": -1.7435070276260376,
-      "logps/chosen": -28.371801376342773,
-      "logps/rejected": -32.005165100097656,
-      "loss": 0.2663,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.4130497872829437,
-      "rewards/margins": 0.9410881996154785,
-      "rewards/rejected": -0.5280382633209229,
-      "step": 1130
-    },
-    {
-      "epoch": 2.96,
-      "grad_norm": 15.5,
-      "learning_rate": 9.59060791022566e-07,
-      "logits/chosen": -1.8376386165618896,
-      "logits/rejected": -1.8344367742538452,
-      "logps/chosen": -31.69878578186035,
-      "logps/rejected": -33.702476501464844,
-      "loss": 0.3231,
-      "rewards/accuracies": 0.9125000238418579,
-      "rewards/chosen": 0.45148882269859314,
-      "rewards/margins": 0.8041674494743347,
-      "rewards/rejected": -0.3526785969734192,
-      "step": 1140
-    },
-    {
-      "epoch": 2.99,
-      "grad_norm": 12.25,
-      "learning_rate": 9.148382544856885e-07,
-      "logits/chosen": -1.757229208946228,
-      "logits/rejected": -1.7509183883666992,
-      "logps/chosen": -27.021270751953125,
-      "logps/rejected": -33.757110595703125,
-      "loss": 0.339,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.3350144624710083,
-      "rewards/margins": 0.8176575899124146,
-      "rewards/rejected": -0.48264312744140625,
-      "step": 1150
-    },
-    {
-      "epoch": 3.01,
-      "grad_norm": 10.375,
-      "learning_rate": 8.714301001505568e-07,
-      "logits/chosen": -1.8691253662109375,
-      "logits/rejected": -1.8666025400161743,
-      "logps/chosen": -31.70334815979004,
-      "logps/rejected": -36.7634391784668,
-      "loss": 0.3008,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.41781607270240784,
-      "rewards/margins": 0.9193881750106812,
-      "rewards/rejected": -0.5015721917152405,
-      "step": 1160
-    },
-    {
-      "epoch": 3.04,
-      "grad_norm": 11.75,
-      "learning_rate": 8.288586291031025e-07,
-      "logits/chosen": -1.8561022281646729,
-      "logits/rejected": -1.853996992111206,
-      "logps/chosen": -30.246707916259766,
-      "logps/rejected": -34.137977600097656,
-      "loss": 0.2548,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.41030970215797424,
-      "rewards/margins": 0.9424960017204285,
-      "rewards/rejected": -0.5321863293647766,
-      "step": 1170
-    },
-    {
-      "epoch": 3.06,
-      "grad_norm": 9.875,
-      "learning_rate": 7.871457125803897e-07,
-      "logits/chosen": -1.8501169681549072,
-      "logits/rejected": -1.8379520177841187,
-      "logps/chosen": -34.59787368774414,
-      "logps/rejected": -35.80955505371094,
-      "loss": 0.2586,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.4342281222343445,
-      "rewards/margins": 0.9556528329849243,
-      "rewards/rejected": -0.5214247703552246,
-      "step": 1180
-    },
-    {
-      "epoch": 3.09,
-      "grad_norm": 11.375,
-      "learning_rate": 7.463127807341966e-07,
-      "logits/chosen": -1.7463674545288086,
-      "logits/rejected": -1.741085410118103,
-      "logps/chosen": -33.60605239868164,
-      "logps/rejected": -35.72685241699219,
-      "loss": 0.2403,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.5230919122695923,
-      "rewards/margins": 0.9698019027709961,
-      "rewards/rejected": -0.4467100501060486,
-      "step": 1190
-    },
-    {
-      "epoch": 3.12,
-      "grad_norm": 8.0625,
-      "learning_rate": 7.063808116212021e-07,
-      "logits/chosen": -1.7915064096450806,
-      "logits/rejected": -1.7947463989257812,
-      "logps/chosen": -31.306697845458984,
-      "logps/rejected": -32.91527557373047,
-      "loss": 0.343,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/chosen": 0.3925189971923828,
-      "rewards/margins": 0.8212476968765259,
-      "rewards/rejected": -0.4287285804748535,
-      "step": 1200
-    },
-    {
-      "epoch": 3.12,
-      "eval_logits/chosen": -2.096956968307495,
-      "eval_logits/rejected": -2.0922605991363525,
-      "eval_logps/chosen": -34.60926055908203,
-      "eval_logps/rejected": -38.34450149536133,
-      "eval_loss": 0.9257969856262207,
-      "eval_rewards/accuracies": 0.5772424936294556,
-      "eval_rewards/chosen": -0.1724136769771576,
-      "eval_rewards/margins": 0.07595065981149673,
-      "eval_rewards/rejected": -0.24836434423923492,
-      "eval_runtime": 145.7776,
-      "eval_samples_per_second": 2.353,
-      "eval_steps_per_second": 0.295,
-      "step": 1200
-    },
-    {
-      "epoch": 3.14,
-      "grad_norm": 10.4375,
-      "learning_rate": 6.673703204254348e-07,
-      "logits/chosen": -1.8141043186187744,
-      "logits/rejected": -1.8096128702163696,
-      "logps/chosen": -29.880346298217773,
-      "logps/rejected": -32.50574493408203,
-      "loss": 0.2651,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.4179980158805847,
-      "rewards/margins": 0.8972921371459961,
-      "rewards/rejected": -0.4792940020561218,
-      "step": 1210
-    },
-    {
-      "epoch": 3.17,
-      "grad_norm": 11.0,
-      "learning_rate": 6.293013489185315e-07,
-      "logits/chosen": -1.8287605047225952,
-      "logits/rejected": -1.8233217000961304,
-      "logps/chosen": -33.81926345825195,
-      "logps/rejected": -36.30878829956055,
-      "loss": 0.274,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.43640297651290894,
-      "rewards/margins": 0.9580758213996887,
-      "rewards/rejected": -0.5216729044914246,
-      "step": 1220
-    },
-    {
-      "epoch": 3.19,
-      "grad_norm": 10.375,
-      "learning_rate": 5.921934551632086e-07,
-      "logits/chosen": -1.8299005031585693,
-      "logits/rejected": -1.8165563344955444,
-      "logps/chosen": -31.802743911743164,
-      "logps/rejected": -35.76459503173828,
-      "loss": 0.319,
-      "rewards/accuracies": 0.862500011920929,
-      "rewards/chosen": 0.3873817026615143,
-      "rewards/margins": 0.8808059692382812,
-      "rewards/rejected": -0.4934242367744446,
-      "step": 1230
-    },
-    {
-      "epoch": 3.22,
-      "grad_norm": 6.28125,
-      "learning_rate": 5.560657034652405e-07,
-      "logits/chosen": -1.876331090927124,
-      "logits/rejected": -1.8739484548568726,
-      "logps/chosen": -33.402381896972656,
-      "logps/rejected": -33.20888137817383,
-      "loss": 0.2775,
-      "rewards/accuracies": 0.9125000238418579,
-      "rewards/chosen": 0.489155113697052,
-      "rewards/margins": 0.899858832359314,
-      "rewards/rejected": -0.41070374846458435,
-      "step": 1240
-    },
-    {
-      "epoch": 3.25,
-      "grad_norm": 6.4375,
-      "learning_rate": 5.2093665457911e-07,
-      "logits/chosen": -1.7996089458465576,
-      "logits/rejected": -1.7967039346694946,
-      "logps/chosen": -32.854007720947266,
-      "logps/rejected": -36.70931625366211,
-      "loss": 0.2307,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.4626428186893463,
-      "rewards/margins": 1.036298155784607,
-      "rewards/rejected": -0.5736554861068726,
-      "step": 1250
-    },
-    {
-      "epoch": 3.27,
-      "grad_norm": 13.5,
-      "learning_rate": 4.868243561723535e-07,
-      "logits/chosen": -1.8704442977905273,
-      "logits/rejected": -1.86517333984375,
-      "logps/chosen": -29.182064056396484,
-      "logps/rejected": -33.91926193237305,
-      "loss": 0.2301,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.4736199975013733,
-      "rewards/margins": 0.9346338510513306,
-      "rewards/rejected": -0.46101388335227966,
-      "step": 1260
-    },
-    {
-      "epoch": 3.3,
-      "grad_norm": 14.25,
-      "learning_rate": 4.537463335535161e-07,
-      "logits/chosen": -1.8953670263290405,
-      "logits/rejected": -1.9002799987792969,
-      "logps/chosen": -31.160940170288086,
-      "logps/rejected": -33.14803695678711,
-      "loss": 0.2666,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.4650169014930725,
-      "rewards/margins": 0.9050186276435852,
-      "rewards/rejected": -0.44000187516212463,
-      "step": 1270
-    },
-    {
-      "epoch": 3.32,
-      "grad_norm": 11.8125,
-      "learning_rate": 4.217195806684629e-07,
-      "logits/chosen": -1.7883453369140625,
-      "logits/rejected": -1.7957513332366943,
-      "logps/chosen": -33.53427505493164,
-      "logps/rejected": -34.219120025634766,
-      "loss": 0.2995,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.41189074516296387,
-      "rewards/margins": 0.8981454968452454,
-      "rewards/rejected": -0.4862546920776367,
-      "step": 1280
-    },
-    {
-      "epoch": 3.35,
-      "grad_norm": 14.3125,
-      "learning_rate": 3.907605513696808e-07,
-      "logits/chosen": -1.7024030685424805,
-      "logits/rejected": -1.704781174659729,
-      "logps/chosen": -31.582706451416016,
-      "logps/rejected": -38.130348205566406,
-      "loss": 0.2925,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.46475711464881897,
-      "rewards/margins": 0.9375804662704468,
-      "rewards/rejected": -0.4728233218193054,
-      "step": 1290
-    },
-    {
-      "epoch": 3.38,
-      "grad_norm": 14.75,
-      "learning_rate": 3.6088515096305675e-07,
-      "logits/chosen": -1.7490756511688232,
-      "logits/rejected": -1.7525970935821533,
-      "logps/chosen": -31.550472259521484,
-      "logps/rejected": -34.04153060913086,
-      "loss": 0.23,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.4998587667942047,
-      "rewards/margins": 0.9730690717697144,
-      "rewards/rejected": -0.47321027517318726,
-      "step": 1300
-    },
-    {
-      "epoch": 3.38,
-      "eval_logits/chosen": -2.0966811180114746,
-      "eval_logits/rejected": -2.0919907093048096,
-      "eval_logps/chosen": -34.60612106323242,
-      "eval_logps/rejected": -38.34492111206055,
-      "eval_loss": 0.9245251417160034,
-      "eval_rewards/accuracies": 0.5830564498901367,
-      "eval_rewards/chosen": -0.17146992683410645,
-      "eval_rewards/margins": 0.07701896876096725,
-      "eval_rewards/rejected": -0.2484889179468155,
-      "eval_runtime": 145.7251,
-      "eval_samples_per_second": 2.354,
-      "eval_steps_per_second": 0.295,
-      "step": 1300
-    },
-    {
-      "epoch": 3.4,
-      "grad_norm": 6.4375,
-      "learning_rate": 3.321087280364757e-07,
-      "logits/chosen": -1.8596166372299194,
-      "logits/rejected": -1.8392422199249268,
-      "logps/chosen": -29.572362899780273,
-      "logps/rejected": -36.99042892456055,
-      "loss": 0.2607,
-      "rewards/accuracies": 0.9125000238418579,
-      "rewards/chosen": 0.4623106122016907,
-      "rewards/margins": 1.0095179080963135,
-      "rewards/rejected": -0.5472072958946228,
-      "step": 1310
-    },
-    {
-      "epoch": 3.43,
-      "grad_norm": 11.6875,
-      "learning_rate": 3.044460665744284e-07,
-      "logits/chosen": -1.8343448638916016,
-      "logits/rejected": -1.8401319980621338,
-      "logps/chosen": -31.251697540283203,
-      "logps/rejected": -34.008384704589844,
-      "loss": 0.22,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.5346270799636841,
-      "rewards/margins": 0.9899166226387024,
-      "rewards/rejected": -0.4552895128726959,
-      "step": 1320
-    },
-    {
-      "epoch": 3.45,
-      "grad_norm": 6.875,
-      "learning_rate": 2.779113783626916e-07,
-      "logits/chosen": -1.8131704330444336,
-      "logits/rejected": -1.808203101158142,
-      "logps/chosen": -31.631412506103516,
-      "logps/rejected": -36.38072967529297,
-      "loss": 0.2259,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.4173283576965332,
-      "rewards/margins": 0.9598596692085266,
-      "rewards/rejected": -0.5425313115119934,
-      "step": 1330
-    },
-    {
-      "epoch": 3.48,
-      "grad_norm": 8.4375,
-      "learning_rate": 2.5251829568697204e-07,
-      "logits/chosen": -1.673139214515686,
-      "logits/rejected": -1.682692289352417,
-      "logps/chosen": -32.161163330078125,
-      "logps/rejected": -32.22640609741211,
-      "loss": 0.3039,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.4915284216403961,
-      "rewards/margins": 0.849216639995575,
-      "rewards/rejected": -0.3576882481575012,
-      "step": 1340
-    },
-    {
-      "epoch": 3.51,
-      "grad_norm": 7.28125,
-      "learning_rate": 2.2827986432927774e-07,
-      "logits/chosen": -1.7312647104263306,
-      "logits/rejected": -1.7359787225723267,
-      "logps/chosen": -32.18891143798828,
-      "logps/rejected": -34.65907287597656,
-      "loss": 0.1974,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.5286074876785278,
-      "rewards/margins": 0.9648852348327637,
-      "rewards/rejected": -0.43627771735191345,
-      "step": 1350
-    },
-    {
-      "epoch": 3.53,
-      "grad_norm": 14.6875,
-      "learning_rate": 2.0520853686560177e-07,
-      "logits/chosen": -1.758639931678772,
-      "logits/rejected": -1.7640345096588135,
-      "logps/chosen": -32.392555236816406,
-      "logps/rejected": -36.23365020751953,
-      "loss": 0.2898,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.4317311644554138,
-      "rewards/margins": 0.9329068064689636,
-      "rewards/rejected": -0.5011756420135498,
-      "step": 1360
-    },
-    {
-      "epoch": 3.56,
-      "grad_norm": 8.6875,
-      "learning_rate": 1.833161662683672e-07,
-      "logits/chosen": -1.852590799331665,
-      "logits/rejected": -1.8480865955352783,
-      "logps/chosen": -31.320281982421875,
-      "logps/rejected": -32.50859832763672,
-      "loss": 0.2705,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.3946218490600586,
-      "rewards/margins": 0.8736333847045898,
-      "rewards/rejected": -0.479011595249176,
-      "step": 1370
-    },
-    {
-      "epoch": 3.58,
-      "grad_norm": 13.0,
-      "learning_rate": 1.626139998169246e-07,
-      "logits/chosen": -1.7653238773345947,
-      "logits/rejected": -1.767480492591858,
-      "logps/chosen": -28.307449340820312,
-      "logps/rejected": -32.36919021606445,
-      "loss": 0.2888,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.4055551588535309,
-      "rewards/margins": 0.8930577039718628,
-      "rewards/rejected": -0.4875025153160095,
-      "step": 1380
-    },
-    {
-      "epoch": 3.61,
-      "grad_norm": 11.0,
-      "learning_rate": 1.4311267331922535e-07,
-      "logits/chosen": -1.7655260562896729,
-      "logits/rejected": -1.7701616287231445,
-      "logps/chosen": -30.582462310791016,
-      "logps/rejected": -34.197975158691406,
-      "loss": 0.3033,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.4311094284057617,
-      "rewards/margins": 0.7940821051597595,
-      "rewards/rejected": -0.3629726767539978,
-      "step": 1390
-    },
-    {
-      "epoch": 3.64,
-      "grad_norm": 6.5625,
-      "learning_rate": 1.2482220564763669e-07,
-      "logits/chosen": -1.8291633129119873,
-      "logits/rejected": -1.8324148654937744,
-      "logps/chosen": -33.25137710571289,
-      "logps/rejected": -35.00487518310547,
-      "loss": 0.1873,
-      "rewards/accuracies": 0.987500011920929,
-      "rewards/chosen": 0.503759503364563,
-      "rewards/margins": 0.9781522750854492,
-      "rewards/rejected": -0.4743928015232086,
-      "step": 1400
-    },
-    {
-      "epoch": 3.64,
-      "eval_logits/chosen": -2.0966780185699463,
-      "eval_logits/rejected": -2.0919783115386963,
-      "eval_logps/chosen": -34.61306381225586,
-      "eval_logps/rejected": -38.35472106933594,
-      "eval_loss": 0.9230431914329529,
-      "eval_rewards/accuracies": 0.5834717750549316,
-      "eval_rewards/chosen": -0.17355351150035858,
-      "eval_rewards/margins": 0.07787670940160751,
-      "eval_rewards/rejected": -0.2514302134513855,
-      "eval_runtime": 145.942,
-      "eval_samples_per_second": 2.35,
-      "eval_steps_per_second": 0.295,
-      "step": 1400
-    },
-    {
-      "epoch": 3.66,
-      "grad_norm": 12.8125,
-      "learning_rate": 1.0775199359171346e-07,
-      "logits/chosen": -1.896807074546814,
-      "logits/rejected": -1.8899128437042236,
-      "logps/chosen": -32.11521911621094,
-      "logps/rejected": -35.75047302246094,
-      "loss": 0.2435,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.5061985850334167,
-      "rewards/margins": 0.9579696655273438,
-      "rewards/rejected": -0.4517711102962494,
-      "step": 1410
-    },
-    {
-      "epoch": 3.69,
-      "grad_norm": 12.75,
-      "learning_rate": 9.191080703056604e-08,
-      "logits/chosen": -1.7876735925674438,
-      "logits/rejected": -1.7987076044082642,
-      "logps/chosen": -32.546993255615234,
-      "logps/rejected": -35.14876937866211,
-      "loss": 0.2304,
-      "rewards/accuracies": 0.987500011920929,
-      "rewards/chosen": 0.48069244623184204,
-      "rewards/margins": 0.9479993581771851,
-      "rewards/rejected": -0.4673069417476654,
-      "step": 1420
-    },
-    {
-      "epoch": 3.71,
-      "grad_norm": 10.375,
-      "learning_rate": 7.730678442730539e-08,
-      "logits/chosen": -1.865950345993042,
-      "logits/rejected": -1.877820611000061,
-      "logps/chosen": -33.48182678222656,
-      "logps/rejected": -35.27021026611328,
-      "loss": 0.25,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.47369375824928284,
-      "rewards/margins": 0.9632562398910522,
-      "rewards/rejected": -0.4895625710487366,
-      "step": 1430
-    },
-    {
-      "epoch": 3.74,
-      "grad_norm": 9.0625,
-      "learning_rate": 6.394742864787806e-08,
-      "logits/chosen": -1.8556573390960693,
-      "logits/rejected": -1.8578948974609375,
-      "logps/chosen": -31.755971908569336,
-      "logps/rejected": -35.50471115112305,
-      "loss": 0.2889,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.43572336435317993,
-      "rewards/margins": 0.929146945476532,
-      "rewards/rejected": -0.49342355132102966,
-      "step": 1440
-    },
-    {
-      "epoch": 3.77,
-      "grad_norm": 11.625,
-      "learning_rate": 5.183960310644748e-08,
-      "logits/chosen": -1.8402214050292969,
-      "logits/rejected": -1.8331701755523682,
-      "logps/chosen": -34.008670806884766,
-      "logps/rejected": -35.25522994995117,
-      "loss": 0.2689,
-      "rewards/accuracies": 0.9125000238418579,
-      "rewards/chosen": 0.46789687871932983,
-      "rewards/margins": 0.9133448600769043,
-      "rewards/rejected": -0.4454478621482849,
-      "step": 1450
-    },
-    {
-      "epoch": 3.79,
-      "grad_norm": 15.25,
-      "learning_rate": 4.098952823928693e-08,
-      "logits/chosen": -1.806641936302185,
-      "logits/rejected": -1.8126341104507446,
-      "logps/chosen": -29.7572021484375,
-      "logps/rejected": -35.442596435546875,
-      "loss": 0.2346,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.43679237365722656,
-      "rewards/margins": 0.9685717821121216,
-      "rewards/rejected": -0.5317793488502502,
-      "step": 1460
-    },
-    {
-      "epoch": 3.82,
-      "grad_norm": 10.375,
-      "learning_rate": 3.1402778309014284e-08,
-      "logits/chosen": -1.7558095455169678,
-      "logits/rejected": -1.7539069652557373,
-      "logps/chosen": -29.265018463134766,
-      "logps/rejected": -32.04896545410156,
-      "loss": 0.2884,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.4741613268852234,
-      "rewards/margins": 0.9105805158615112,
-      "rewards/rejected": -0.43641918897628784,
-      "step": 1470
-    },
-    {
-      "epoch": 3.84,
-      "grad_norm": 12.0625,
-      "learning_rate": 2.3084278540791427e-08,
-      "logits/chosen": -1.9300537109375,
-      "logits/rejected": -1.9244730472564697,
-      "logps/chosen": -34.008506774902344,
-      "logps/rejected": -33.67087936401367,
-      "loss": 0.262,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.38302525877952576,
-      "rewards/margins": 0.8672767877578735,
-      "rewards/rejected": -0.4842514097690582,
-      "step": 1480
-    },
-    {
-      "epoch": 3.87,
-      "grad_norm": 10.6875,
-      "learning_rate": 1.6038302591975807e-08,
-      "logits/chosen": -1.8236973285675049,
-      "logits/rejected": -1.8260858058929443,
-      "logps/chosen": -27.153884887695312,
-      "logps/rejected": -29.336627960205078,
-      "loss": 0.3435,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 0.3967483341693878,
-      "rewards/margins": 0.7910512685775757,
-      "rewards/rejected": -0.39430302381515503,
-      "step": 1490
-    },
-    {
-      "epoch": 3.9,
-      "grad_norm": 10.875,
-      "learning_rate": 1.0268470356514237e-08,
-      "logits/chosen": -1.8308475017547607,
-      "logits/rejected": -1.8252521753311157,
-      "logps/chosen": -31.535476684570312,
-      "logps/rejected": -33.61229705810547,
-      "loss": 0.3099,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.37959426641464233,
-      "rewards/margins": 0.872596263885498,
-      "rewards/rejected": -0.4930020868778229,
-      "step": 1500
-    },
-    {
-      "epoch": 3.9,
-      "eval_logits/chosen": -2.097231864929199,
-      "eval_logits/rejected": -2.0925421714782715,
-      "eval_logps/chosen": -34.605594635009766,
-      "eval_logps/rejected": -38.34328842163086,
-      "eval_loss": 0.9251258969306946,
-      "eval_rewards/accuracies": 0.5772424936294556,
-      "eval_rewards/chosen": -0.17131240665912628,
-      "eval_rewards/margins": 0.07668941468000412,
-      "eval_rewards/rejected": -0.2480018138885498,
-      "eval_runtime": 145.9509,
-      "eval_samples_per_second": 2.35,
-      "eval_steps_per_second": 0.295,
-      "step": 1500
-    },
-    {
-      "epoch": 3.92,
-      "grad_norm": 13.9375,
-      "learning_rate": 5.777746105209147e-09,
-      "logits/chosen": -1.7565958499908447,
-      "logits/rejected": -1.7606948614120483,
-      "logps/chosen": -32.811988830566406,
-      "logps/rejected": -35.829490661621094,
-      "loss": 0.3182,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/chosen": 0.40770164132118225,
-      "rewards/margins": 0.8334344029426575,
-      "rewards/rejected": -0.4257327914237976,
-      "step": 1510
-    },
-    {
-      "epoch": 3.95,
-      "grad_norm": 4.40625,
-      "learning_rate": 2.5684369628148352e-09,
-      "logits/chosen": -1.8115999698638916,
-      "logits/rejected": -1.8100284337997437,
-      "logps/chosen": -29.23919105529785,
-      "logps/rejected": -34.142478942871094,
-      "loss": 0.3087,
-      "rewards/accuracies": 0.9125000238418579,
-      "rewards/chosen": 0.44849544763565063,
-      "rewards/margins": 0.9143233299255371,
-      "rewards/rejected": -0.46582794189453125,
-      "step": 1520
-    },
-    {
-      "epoch": 3.97,
-      "grad_norm": 12.5625,
-      "learning_rate": 6.421917227455999e-10,
-      "logits/chosen": -1.910882592201233,
-      "logits/rejected": -1.903089165687561,
-      "logps/chosen": -26.731464385986328,
-      "logps/rejected": -30.273723602294922,
-      "loss": 0.2875,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.39944320917129517,
-      "rewards/margins": 0.8800784349441528,
-      "rewards/rejected": -0.4806351661682129,
-      "step": 1530
-    },
-    {
-      "epoch": 4.0,
-      "grad_norm": 12.4375,
-      "learning_rate": 0.0,
-      "logits/chosen": -1.8143596649169922,
-      "logits/rejected": -1.8040344715118408,
-      "logps/chosen": -31.594379425048828,
-      "logps/rejected": -37.145050048828125,
-      "loss": 0.244,
-      "rewards/accuracies": 0.987500011920929,
-      "rewards/chosen": 0.44154614210128784,
-      "rewards/margins": 0.9617103338241577,
-      "rewards/rejected": -0.5201640129089355,
-      "step": 1540
-    },
-    {
-      "epoch": 4.0,
-      "step": 1540,
+      "epoch": 1.0,
+      "step": 385,
       "total_flos": 0.0,
-      "train_loss": 0.32053508061867253,
-      "train_runtime": 10790.1902,
-      "train_samples_per_second": 1.141,
-      "train_steps_per_second": 0.143
+      "train_loss": 0.8125879040012112,
+      "train_runtime": 3249.3137,
+      "train_samples_per_second": 0.948,
+      "train_steps_per_second": 0.118
     }
   ],
   "logging_steps": 10,
-  "max_steps": 1540,
+  "max_steps": 385,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 4,
+  "num_train_epochs": 1,
   "save_steps": 100,
   "total_flos": 0.0,
   "train_batch_size": 4,