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,
@@ -15,7 +15,7 @@
       "logits/rejected": -1.7377450466156006,
       "logps/chosen": -29.553977966308594,
       "logps/rejected": -42.813133239746094,
-      "loss": 1.5625,
+      "loss": 1.0,
       "rewards/accuracies": 0.0,
       "rewards/chosen": 0.0,
       "rewards/margins": 0.0,
@@ -25,2537 +25,597 @@
     {
       "epoch": 0.03,
       "learning_rate": 1.282051282051282e-06,
-      "logits/chosen": -1.8663330078125,
-      "logits/rejected": -1.870657205581665,
-      "logps/chosen": -36.98442077636719,
-      "logps/rejected": -33.6639289855957,
-      "loss": 1.4725,
-      "rewards/accuracies": 0.5416666865348816,
-      "rewards/chosen": 0.008871171623468399,
-      "rewards/margins": 0.02241421490907669,
-      "rewards/rejected": -0.013543044216930866,
+      "logits/chosen": -1.8664777278900146,
+      "logits/rejected": -1.8707849979400635,
+      "logps/chosen": -36.99364471435547,
+      "logps/rejected": -33.650604248046875,
+      "loss": 0.9766,
+      "rewards/accuracies": 0.5277777910232544,
+      "rewards/chosen": 0.00906434003263712,
+      "rewards/margins": 0.023435616865754128,
+      "rewards/rejected": -0.014371277764439583,
       "step": 10
     },
     {
       "epoch": 0.05,
       "learning_rate": 2.564102564102564e-06,
-      "logits/chosen": -1.9977548122406006,
-      "logits/rejected": -2.0003960132598877,
-      "logps/chosen": -29.64242172241211,
-      "logps/rejected": -29.04966163635254,
-      "loss": 1.6323,
-      "rewards/accuracies": 0.4124999940395355,
-      "rewards/chosen": -9.128078818321228e-05,
-      "rewards/margins": -0.005604482255876064,
-      "rewards/rejected": 0.005513200536370277,
+      "logits/chosen": -1.9982150793075562,
+      "logits/rejected": -2.0008621215820312,
+      "logps/chosen": -29.64394760131836,
+      "logps/rejected": -29.04986000061035,
+      "loss": 1.0107,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": -0.0012267641723155975,
+      "rewards/margins": -0.010734880343079567,
+      "rewards/rejected": 0.00950811617076397,
       "step": 20
     },
     {
       "epoch": 0.08,
       "learning_rate": 3.846153846153847e-06,
-      "logits/chosen": -1.9204576015472412,
-      "logits/rejected": -1.917781114578247,
-      "logps/chosen": -31.413188934326172,
-      "logps/rejected": -33.22998046875,
-      "loss": 1.5666,
-      "rewards/accuracies": 0.5625,
-      "rewards/chosen": 0.0011628343490883708,
-      "rewards/margins": 0.005770112853497267,
-      "rewards/rejected": -0.0046072788536548615,
+      "logits/chosen": -1.920768141746521,
+      "logits/rejected": -1.9180870056152344,
+      "logps/chosen": -31.416461944580078,
+      "logps/rejected": -33.2098274230957,
+      "loss": 1.0063,
+      "rewards/accuracies": 0.512499988079071,
+      "rewards/chosen": -0.000256747764069587,
+      "rewards/margins": -0.0063001858070492744,
+      "rewards/rejected": 0.00604343693703413,
       "step": 30
     },
     {
       "epoch": 0.1,
       "learning_rate": 4.999896948438434e-06,
-      "logits/chosen": -2.0172924995422363,
-      "logits/rejected": -2.0085625648498535,
-      "logps/chosen": -32.583003997802734,
-      "logps/rejected": -32.50304412841797,
-      "loss": 1.6493,
-      "rewards/accuracies": 0.48750001192092896,
-      "rewards/chosen": -0.0024785413406789303,
-      "rewards/margins": -0.006329345051199198,
-      "rewards/rejected": 0.00385080324485898,
+      "logits/chosen": -2.0177602767944336,
+      "logits/rejected": -2.009014129638672,
+      "logps/chosen": -32.56236267089844,
+      "logps/rejected": -32.517822265625,
+      "loss": 0.9863,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.010110180824995041,
+      "rewards/margins": 0.013717299327254295,
+      "rewards/rejected": -0.003607118036597967,
       "step": 40
     },
     {
       "epoch": 0.13,
       "learning_rate": 4.987541037542187e-06,
-      "logits/chosen": -1.8622093200683594,
-      "logits/rejected": -1.851438283920288,
-      "logps/chosen": -33.5463981628418,
-      "logps/rejected": -35.46432113647461,
-      "loss": 1.569,
-      "rewards/accuracies": 0.5,
-      "rewards/chosen": 0.004430866800248623,
-      "rewards/margins": 0.009221290238201618,
-      "rewards/rejected": -0.004790422972291708,
+      "logits/chosen": -1.862694501876831,
+      "logits/rejected": -1.8519262075424194,
+      "logps/chosen": -33.541160583496094,
+      "logps/rejected": -35.44048309326172,
+      "loss": 0.9969,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.011423684656620026,
+      "rewards/margins": 0.003121361369267106,
+      "rewards/rejected": 0.00830232072621584,
       "step": 50
     },
     {
       "epoch": 0.16,
       "learning_rate": 4.954691471941119e-06,
-      "logits/chosen": -1.939429521560669,
-      "logits/rejected": -1.941382646560669,
-      "logps/chosen": -32.560157775878906,
-      "logps/rejected": -33.21000671386719,
-      "loss": 1.5224,
-      "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": 0.016012197360396385,
-      "rewards/margins": 0.03146379441022873,
-      "rewards/rejected": -0.015451604500412941,
+      "logits/chosen": -1.9417282342910767,
+      "logits/rejected": -1.9436867237091064,
+      "logps/chosen": -32.52958679199219,
+      "logps/rejected": -33.216880798339844,
+      "loss": 0.9296,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.04941769689321518,
+      "rewards/margins": 0.08126799017190933,
+      "rewards/rejected": -0.031850285828113556,
       "step": 60
     },
     {
       "epoch": 0.18,
       "learning_rate": 4.901618883413549e-06,
-      "logits/chosen": -2.0719385147094727,
-      "logits/rejected": -2.0768914222717285,
-      "logps/chosen": -33.97148513793945,
-      "logps/rejected": -36.61540985107422,
-      "loss": 1.5627,
-      "rewards/accuracies": 0.5,
-      "rewards/chosen": 0.006800562143325806,
-      "rewards/margins": 0.025414401665329933,
-      "rewards/rejected": -0.018613843247294426,
+      "logits/chosen": -2.0729386806488037,
+      "logits/rejected": -2.0779125690460205,
+      "logps/chosen": -33.99254608154297,
+      "logps/rejected": -36.62586212158203,
+      "loss": 0.9629,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": -0.0028400986921042204,
+      "rewards/margins": 0.037055134773254395,
+      "rewards/rejected": -0.03989524394273758,
       "step": 70
     },
     {
       "epoch": 0.21,
       "learning_rate": 4.828760511501322e-06,
-      "logits/chosen": -1.9341154098510742,
-      "logits/rejected": -1.9372339248657227,
-      "logps/chosen": -34.31813049316406,
-      "logps/rejected": -34.64340591430664,
-      "loss": 1.3398,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.043871380388736725,
-      "rewards/margins": 0.06426960229873657,
-      "rewards/rejected": -0.020398225635290146,
+      "logits/chosen": -1.9335737228393555,
+      "logits/rejected": -1.9366981983184814,
+      "logps/chosen": -34.332157135009766,
+      "logps/rejected": -34.641021728515625,
+      "loss": 0.9026,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.06695752590894699,
+      "rewards/margins": 0.10098665952682495,
+      "rewards/rejected": -0.03402913734316826,
       "step": 80
     },
     {
       "epoch": 0.23,
       "learning_rate": 4.7367166013034295e-06,
-      "logits/chosen": -1.9430510997772217,
-      "logits/rejected": -1.9475826025009155,
-      "logps/chosen": -32.405487060546875,
-      "logps/rejected": -32.3362922668457,
-      "loss": 1.6048,
-      "rewards/accuracies": 0.612500011920929,
-      "rewards/chosen": 0.02477400377392769,
-      "rewards/margins": 0.01572720892727375,
-      "rewards/rejected": 0.009046795777976513,
+      "logits/chosen": -1.9401309490203857,
+      "logits/rejected": -1.9446433782577515,
+      "logps/chosen": -32.37213897705078,
+      "logps/rejected": -32.343849182128906,
+      "loss": 0.9438,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.06670050323009491,
+      "rewards/margins": 0.05615914613008499,
+      "rewards/rejected": 0.010541360825300217,
       "step": 90
     },
     {
       "epoch": 0.26,
       "learning_rate": 4.626245458345211e-06,
-      "logits/chosen": -2.040834426879883,
-      "logits/rejected": -2.0388479232788086,
-      "logps/chosen": -32.189510345458984,
-      "logps/rejected": -31.297454833984375,
-      "loss": 1.4583,
-      "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": 0.020529210567474365,
-      "rewards/margins": 0.0330255962908268,
-      "rewards/rejected": -0.012496383860707283,
+      "logits/chosen": -2.037087917327881,
+      "logits/rejected": -2.035101890563965,
+      "logps/chosen": -32.13945388793945,
+      "logps/rejected": -31.313283920288086,
+      "loss": 0.8961,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.07096613943576813,
+      "rewards/margins": 0.10391455888748169,
+      "rewards/rejected": -0.032948415726423264,
       "step": 100
     },
     {
       "epoch": 0.26,
-      "eval_logits/chosen": -2.2356765270233154,
-      "eval_logits/rejected": -2.2308189868927,
-      "eval_logps/chosen": -34.03032302856445,
-      "eval_logps/rejected": -37.523555755615234,
-      "eval_loss": 1.6161867380142212,
-      "eval_rewards/accuracies": 0.5245016813278198,
-      "eval_rewards/chosen": 0.0016898562898859382,
-      "eval_rewards/margins": 0.004464337136596441,
-      "eval_rewards/rejected": -0.0027744807302951813,
-      "eval_runtime": 145.7904,
-      "eval_samples_per_second": 2.353,
+      "eval_logits/chosen": -2.231482982635498,
+      "eval_logits/rejected": -2.226637363433838,
+      "eval_logps/chosen": -34.02524185180664,
+      "eval_logps/rejected": -37.54085159301758,
+      "eval_loss": 0.9766585230827332,
+      "eval_rewards/accuracies": 0.5365448594093323,
+      "eval_rewards/chosen": 0.006516099441796541,
+      "eval_rewards/margins": 0.02347717247903347,
+      "eval_rewards/rejected": -0.016961071640253067,
+      "eval_runtime": 145.8279,
+      "eval_samples_per_second": 2.352,
       "eval_steps_per_second": 0.295,
       "step": 100
     },
     {
       "epoch": 0.29,
       "learning_rate": 4.498257201263691e-06,
-      "logits/chosen": -1.996921181678772,
-      "logits/rejected": -1.9945523738861084,
-      "logps/chosen": -33.126094818115234,
-      "logps/rejected": -34.01633071899414,
-      "loss": 1.6379,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.04708842188119888,
-      "rewards/margins": 0.041870661079883575,
-      "rewards/rejected": 0.005217757076025009,
+      "logits/chosen": -1.9920060634613037,
+      "logits/rejected": -1.9896419048309326,
+      "logps/chosen": -33.146766662597656,
+      "logps/rejected": -34.02008819580078,
+      "loss": 0.9486,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": 0.0679345652461052,
+      "rewards/margins": 0.061437882483005524,
+      "rewards/rejected": 0.006496679037809372,
       "step": 110
     },
     {
       "epoch": 0.31,
       "learning_rate": 4.353806263777678e-06,
-      "logits/chosen": -2.0087766647338867,
-      "logits/rejected": -2.0004258155822754,
-      "logps/chosen": -32.360713958740234,
-      "logps/rejected": -32.135719299316406,
-      "loss": 1.5576,
-      "rewards/accuracies": 0.512499988079071,
-      "rewards/chosen": 0.03367554023861885,
-      "rewards/margins": 0.023361390456557274,
-      "rewards/rejected": 0.010314146056771278,
+      "logits/chosen": -2.003952741622925,
+      "logits/rejected": -1.9956319332122803,
+      "logps/chosen": -32.33639144897461,
+      "logps/rejected": -32.133079528808594,
+      "loss": 0.9488,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.07595954835414886,
+      "rewards/margins": 0.05606143921613693,
+      "rewards/rejected": 0.019898109138011932,
       "step": 120
     },
     {
       "epoch": 0.34,
       "learning_rate": 4.1940827077152755e-06,
-      "logits/chosen": -2.036501407623291,
-      "logits/rejected": -2.0285134315490723,
-      "logps/chosen": -30.350208282470703,
-      "logps/rejected": -32.078697204589844,
-      "loss": 1.4709,
+      "logits/chosen": -2.0316150188446045,
+      "logits/rejected": -2.0236544609069824,
+      "logps/chosen": -30.298206329345703,
+      "logps/rejected": -32.07080841064453,
+      "loss": 0.8911,
       "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": 0.05020326375961304,
-      "rewards/margins": 0.06047834828495979,
-      "rewards/rejected": -0.010275078006088734,
+      "rewards/chosen": 0.12425784766674042,
+      "rewards/margins": 0.1367165446281433,
+      "rewards/rejected": -0.0124586820602417,
       "step": 130
     },
     {
       "epoch": 0.36,
       "learning_rate": 4.0204024186666215e-06,
-      "logits/chosen": -1.966408371925354,
-      "logits/rejected": -1.9766807556152344,
-      "logps/chosen": -31.250171661376953,
-      "logps/rejected": -32.605186462402344,
-      "loss": 1.2633,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": 0.06779567152261734,
-      "rewards/margins": 0.09591736644506454,
-      "rewards/rejected": -0.028121691197156906,
+      "logits/chosen": -1.962376356124878,
+      "logits/rejected": -1.9726108312606812,
+      "logps/chosen": -31.235275268554688,
+      "logps/rejected": -32.56925964355469,
+      "loss": 0.8524,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.1290682703256607,
+      "rewards/margins": 0.15313370525836945,
+      "rewards/rejected": -0.024065453559160233,
       "step": 140
     },
     {
       "epoch": 0.39,
       "learning_rate": 3.834196265035119e-06,
-      "logits/chosen": -1.8785550594329834,
-      "logits/rejected": -1.8797214031219482,
-      "logps/chosen": -33.973148345947266,
-      "logps/rejected": -34.82178497314453,
-      "loss": 1.271,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": 0.08781836926937103,
-      "rewards/margins": 0.11363118886947632,
-      "rewards/rejected": -0.02581283450126648,
+      "logits/chosen": -1.8732143640518188,
+      "logits/rejected": -1.874371886253357,
+      "logps/chosen": -33.8985481262207,
+      "logps/rejected": -34.81908416748047,
+      "loss": 0.786,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.20590214431285858,
+      "rewards/margins": 0.24918103218078613,
+      "rewards/rejected": -0.04327889531850815,
       "step": 150
     },
     {
       "epoch": 0.42,
       "learning_rate": 3.636998309800573e-06,
-      "logits/chosen": -1.9306548833847046,
-      "logits/rejected": -1.927215337753296,
-      "logps/chosen": -36.05388259887695,
-      "logps/rejected": -32.733726501464844,
-      "loss": 1.3903,
-      "rewards/accuracies": 0.637499988079071,
-      "rewards/chosen": 0.04711749404668808,
-      "rewards/margins": 0.051506586372852325,
-      "rewards/rejected": -0.004389096051454544,
+      "logits/chosen": -1.924538016319275,
+      "logits/rejected": -1.921121597290039,
+      "logps/chosen": -36.01353454589844,
+      "logps/rejected": -32.723262786865234,
+      "loss": 0.8894,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.11069967597723007,
+      "rewards/margins": 0.11105670034885406,
+      "rewards/rejected": -0.0003570284752640873,
       "step": 160
     },
     {
       "epoch": 0.44,
       "learning_rate": 3.4304331721118078e-06,
-      "logits/chosen": -2.0320611000061035,
-      "logits/rejected": -2.0246951580047607,
-      "logps/chosen": -33.545772552490234,
-      "logps/rejected": -31.430078506469727,
-      "loss": 1.1164,
+      "logits/chosen": -2.0247209072113037,
+      "logits/rejected": -2.017392635345459,
+      "logps/chosen": -33.49879455566406,
+      "logps/rejected": -31.44363784790039,
+      "loss": 0.7434,
       "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.10083547979593277,
-      "rewards/margins": 0.13212835788726807,
-      "rewards/rejected": -0.0312928780913353,
+      "rewards/chosen": 0.2093469649553299,
+      "rewards/margins": 0.2735980153083801,
+      "rewards/rejected": -0.06425107270479202,
       "step": 170
     },
     {
       "epoch": 0.47,
       "learning_rate": 3.2162026428305436e-06,
-      "logits/chosen": -2.0383529663085938,
-      "logits/rejected": -2.0436065196990967,
-      "logps/chosen": -32.30766677856445,
-      "logps/rejected": -32.46817398071289,
-      "loss": 1.2603,
-      "rewards/accuracies": 0.625,
-      "rewards/chosen": 0.09484858065843582,
-      "rewards/margins": 0.08806191384792328,
-      "rewards/rejected": 0.006786657031625509,
+      "logits/chosen": -2.031324863433838,
+      "logits/rejected": -2.0365915298461914,
+      "logps/chosen": -32.253074645996094,
+      "logps/rejected": -32.45112609863281,
+      "loss": 0.8312,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.20420043170452118,
+      "rewards/margins": 0.18039169907569885,
+      "rewards/rejected": 0.023808732628822327,
       "step": 180
     },
     {
       "epoch": 0.49,
       "learning_rate": 2.996071664294641e-06,
-      "logits/chosen": -2.0393786430358887,
-      "logits/rejected": -2.036611795425415,
-      "logps/chosen": -31.301233291625977,
-      "logps/rejected": -31.35910415649414,
-      "loss": 1.2965,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.0750882551074028,
-      "rewards/margins": 0.09144291281700134,
-      "rewards/rejected": -0.016354653984308243,
+      "logits/chosen": -2.0317559242248535,
+      "logits/rejected": -2.0289719104766846,
+      "logps/chosen": -31.279537200927734,
+      "logps/rejected": -31.34115219116211,
+      "loss": 0.8405,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.14659160375595093,
+      "rewards/margins": 0.16264860332012177,
+      "rewards/rejected": -0.016057008877396584,
       "step": 190
     },
     {
       "epoch": 0.52,
       "learning_rate": 2.7718537898066833e-06,
-      "logits/chosen": -1.9094291925430298,
-      "logits/rejected": -1.9140870571136475,
-      "logps/chosen": -31.389511108398438,
-      "logps/rejected": -32.832645416259766,
-      "loss": 1.279,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.08670923858880997,
-      "rewards/margins": 0.10339467227458954,
-      "rewards/rejected": -0.01668544113636017,
+      "logits/chosen": -1.9018064737319946,
+      "logits/rejected": -1.9064457416534424,
+      "logps/chosen": -31.301830291748047,
+      "logps/rejected": -32.8339729309082,
+      "loss": 0.7699,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.21311891078948975,
+      "rewards/margins": 0.2432461678981781,
+      "rewards/rejected": -0.03012726828455925,
       "step": 200
     },
     {
       "epoch": 0.52,
-      "eval_logits/chosen": -2.2359395027160645,
-      "eval_logits/rejected": -2.231079339981079,
-      "eval_logps/chosen": -34.08246994018555,
-      "eval_logps/rejected": -37.59228515625,
-      "eval_loss": 1.6099865436553955,
-      "eval_rewards/accuracies": 0.525747537612915,
-      "eval_rewards/chosen": -0.019167287275195122,
-      "eval_rewards/margins": 0.011099184863269329,
-      "eval_rewards/rejected": -0.030266473069787025,
-      "eval_runtime": 145.6348,
-      "eval_samples_per_second": 2.355,
+      "eval_logits/chosen": -2.2282192707061768,
+      "eval_logits/rejected": -2.223379611968994,
+      "eval_logps/chosen": -34.05467224121094,
+      "eval_logps/rejected": -37.57374572753906,
+      "eval_loss": 0.974229633808136,
+      "eval_rewards/accuracies": 0.530315637588501,
+      "eval_rewards/chosen": -0.01408342458307743,
+      "eval_rewards/margins": 0.02590302750468254,
+      "eval_rewards/rejected": -0.03998645395040512,
+      "eval_runtime": 145.7826,
+      "eval_samples_per_second": 2.353,
       "eval_steps_per_second": 0.295,
       "step": 200
     },
     {
       "epoch": 0.55,
       "learning_rate": 2.5453962426402006e-06,
-      "logits/chosen": -2.021488666534424,
-      "logits/rejected": -2.0321521759033203,
-      "logps/chosen": -31.773427963256836,
-      "logps/rejected": -33.9339485168457,
-      "loss": 1.1843,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 0.07915131002664566,
-      "rewards/margins": 0.107650026679039,
-      "rewards/rejected": -0.02849872037768364,
+      "logits/chosen": -2.014596462249756,
+      "logits/rejected": -2.0252418518066406,
+      "logps/chosen": -31.797557830810547,
+      "logps/rejected": -33.982398986816406,
+      "loss": 0.8139,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.12162177264690399,
+      "rewards/margins": 0.20541362464427948,
+      "rewards/rejected": -0.08379185199737549,
       "step": 210
     },
     {
       "epoch": 0.57,
       "learning_rate": 2.3185646976551794e-06,
-      "logits/chosen": -1.9136203527450562,
-      "logits/rejected": -1.9284076690673828,
-      "logps/chosen": -29.9388370513916,
-      "logps/rejected": -31.61223793029785,
-      "loss": 1.2413,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.06908205896615982,
-      "rewards/margins": 0.09070716798305511,
-      "rewards/rejected": -0.021625112742185593,
+      "logits/chosen": -1.9068737030029297,
+      "logits/rejected": -1.9216482639312744,
+      "logps/chosen": -29.838830947875977,
+      "logps/rejected": -31.62994956970215,
+      "loss": 0.766,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.1908968687057495,
+      "rewards/margins": 0.24113738536834717,
+      "rewards/rejected": -0.05024053901433945,
       "step": 220
     },
     {
       "epoch": 0.6,
       "learning_rate": 2.0932279108998323e-06,
-      "logits/chosen": -1.9700790643692017,
-      "logits/rejected": -1.9740597009658813,
-      "logps/chosen": -33.1496467590332,
-      "logps/rejected": -31.65804100036621,
-      "loss": 1.1206,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.10712933540344238,
-      "rewards/margins": 0.15062502026557922,
-      "rewards/rejected": -0.043495677411556244,
+      "logits/chosen": -1.9636850357055664,
+      "logits/rejected": -1.967655897140503,
+      "logps/chosen": -33.12433624267578,
+      "logps/rejected": -31.637094497680664,
+      "loss": 0.7833,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.20519034564495087,
+      "rewards/margins": 0.2666449546813965,
+      "rewards/rejected": -0.06145460531115532,
       "step": 230
     },
     {
       "epoch": 0.62,
       "learning_rate": 1.8712423238279358e-06,
-      "logits/chosen": -1.9689995050430298,
-      "logits/rejected": -1.9470970630645752,
-      "logps/chosen": -33.9182243347168,
-      "logps/rejected": -35.09490966796875,
-      "loss": 1.073,
-      "rewards/accuracies": 0.75,
-      "rewards/chosen": 0.09421048313379288,
-      "rewards/margins": 0.15185722708702087,
-      "rewards/rejected": -0.0576467402279377,
+      "logits/chosen": -1.9611847400665283,
+      "logits/rejected": -1.9393657445907593,
+      "logps/chosen": -33.867958068847656,
+      "logps/rejected": -35.12390899658203,
+      "loss": 0.737,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.2000524252653122,
+      "rewards/margins": 0.32123422622680664,
+      "rewards/rejected": -0.12118180096149445,
       "step": 240
     },
     {
       "epoch": 0.65,
       "learning_rate": 1.6544367689701824e-06,
-      "logits/chosen": -2.010103702545166,
-      "logits/rejected": -2.006786346435547,
-      "logps/chosen": -32.75348663330078,
-      "logps/rejected": -36.29840087890625,
-      "loss": 1.2351,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.0686345100402832,
-      "rewards/margins": 0.09802549332380295,
-      "rewards/rejected": -0.029390990734100342,
+      "logits/chosen": -2.001960277557373,
+      "logits/rejected": -1.9986454248428345,
+      "logps/chosen": -32.73499298095703,
+      "logps/rejected": -36.28093719482422,
+      "loss": 0.8434,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.13304933905601501,
+      "rewards/margins": 0.17225751280784607,
+      "rewards/rejected": -0.03920816630125046,
       "step": 250
     },
     {
       "epoch": 0.68,
       "learning_rate": 1.4445974030621963e-06,
-      "logits/chosen": -1.877804160118103,
-      "logits/rejected": -1.8753639459609985,
-      "logps/chosen": -34.015769958496094,
-      "logps/rejected": -35.56401824951172,
-      "loss": 1.2236,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.06905192881822586,
-      "rewards/margins": 0.09281176328659058,
-      "rewards/rejected": -0.023759832605719566,
+      "logits/chosen": -1.8687576055526733,
+      "logits/rejected": -1.866320013999939,
+      "logps/chosen": -33.981781005859375,
+      "logps/rejected": -35.54584503173828,
+      "loss": 0.8296,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": 0.14463281631469727,
+      "rewards/margins": 0.1734902262687683,
+      "rewards/rejected": -0.028857415542006493,
       "step": 260
     },
     {
       "epoch": 0.7,
       "learning_rate": 1.243452991757889e-06,
-      "logits/chosen": -1.8633472919464111,
-      "logits/rejected": -1.860896348953247,
-      "logps/chosen": -34.244041442871094,
-      "logps/rejected": -31.82059669494629,
-      "loss": 1.2837,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 0.06142274662852287,
-      "rewards/margins": 0.08112844079732895,
-      "rewards/rejected": -0.019705694168806076,
+      "logits/chosen": -1.853939414024353,
+      "logits/rejected": -1.8515437841415405,
+      "logps/chosen": -34.234046936035156,
+      "logps/rejected": -31.837631225585938,
+      "loss": 0.8487,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": 0.11448518931865692,
+      "rewards/margins": 0.16089434921741486,
+      "rewards/rejected": -0.04640916362404823,
       "step": 270
     },
     {
       "epoch": 0.73,
       "learning_rate": 1.0526606671603523e-06,
-      "logits/chosen": -1.9670900106430054,
-      "logits/rejected": -1.9565184116363525,
-      "logps/chosen": -35.00214767456055,
-      "logps/rejected": -31.88429832458496,
-      "loss": 0.9577,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 0.14179818332195282,
-      "rewards/margins": 0.15912050008773804,
-      "rewards/rejected": -0.017322316765785217,
+      "logits/chosen": -1.9570705890655518,
+      "logits/rejected": -1.9465986490249634,
+      "logps/chosen": -35.030006408691406,
+      "logps/rejected": -31.88030433654785,
+      "loss": 0.7553,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.22864773869514465,
+      "rewards/margins": 0.2561652660369873,
+      "rewards/rejected": -0.02751758135855198,
       "step": 280
     },
     {
       "epoch": 0.75,
       "learning_rate": 8.737922755071455e-07,
-      "logits/chosen": -2.0620455741882324,
-      "logits/rejected": -2.0470833778381348,
-      "logps/chosen": -30.75039291381836,
-      "logps/rejected": -32.6963996887207,
-      "loss": 1.3545,
-      "rewards/accuracies": 0.5874999761581421,
-      "rewards/chosen": 0.06949276477098465,
-      "rewards/margins": 0.08731748908758163,
-      "rewards/rejected": -0.017824724316596985,
+      "logits/chosen": -2.0522685050964355,
+      "logits/rejected": -2.0373730659484863,
+      "logps/chosen": -30.7352352142334,
+      "logps/rejected": -32.61699676513672,
+      "loss": 0.9072,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.13222160935401917,
+      "rewards/margins": 0.10783363878726959,
+      "rewards/rejected": 0.024387964978814125,
       "step": 290
     },
     {
       "epoch": 0.78,
       "learning_rate": 7.08321427484816e-07,
-      "logits/chosen": -1.933719277381897,
-      "logits/rejected": -1.931168556213379,
-      "logps/chosen": -32.519874572753906,
-      "logps/rejected": -30.880207061767578,
-      "loss": 1.0043,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.1628943681716919,
-      "rewards/margins": 0.1882532835006714,
-      "rewards/rejected": -0.02535891905426979,
+      "logits/chosen": -1.9228973388671875,
+      "logits/rejected": -1.9203764200210571,
+      "logps/chosen": -32.44710159301758,
+      "logps/rejected": -30.934436798095703,
+      "loss": 0.6723,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.3360074460506439,
+      "rewards/margins": 0.418344646692276,
+      "rewards/rejected": -0.08233721554279327,
       "step": 300
     },
     {
       "epoch": 0.78,
-      "eval_logits/chosen": -2.2311501502990723,
-      "eval_logits/rejected": -2.226292610168457,
-      "eval_logps/chosen": -34.0844841003418,
-      "eval_logps/rejected": -37.60036087036133,
-      "eval_loss": 1.5962141752243042,
-      "eval_rewards/accuracies": 0.5166113376617432,
-      "eval_rewards/chosen": -0.019972510635852814,
-      "eval_rewards/margins": 0.013524654321372509,
-      "eval_rewards/rejected": -0.0334971658885479,
-      "eval_runtime": 145.5939,
-      "eval_samples_per_second": 2.356,
+      "eval_logits/chosen": -2.2234153747558594,
+      "eval_logits/rejected": -2.218602180480957,
+      "eval_logps/chosen": -34.08680725097656,
+      "eval_logps/rejected": -37.60466003417969,
+      "eval_loss": 0.976102888584137,
+      "eval_rewards/accuracies": 0.529900312423706,
+      "eval_rewards/chosen": -0.036580219864845276,
+      "eval_rewards/margins": 0.02504708059132099,
+      "eval_rewards/rejected": -0.06162729859352112,
+      "eval_runtime": 145.7665,
+      "eval_samples_per_second": 2.353,
       "eval_steps_per_second": 0.295,
       "step": 300
     },
     {
       "epoch": 0.81,
-      "grad_norm": 104.0,
-      "learning_rate": 4.84533120650964e-06,
-      "logits/chosen": -2.0658023357391357,
-      "logits/rejected": -2.0529816150665283,
-      "logps/chosen": -32.19905090332031,
-      "logps/rejected": -32.87395477294922,
-      "loss": 0.7465,
-      "rewards/accuracies": 0.887499988079071,
-      "rewards/chosen": 0.13173677027225494,
-      "rewards/margins": 0.22489500045776367,
-      "rewards/rejected": -0.09315825998783112,
+      "learning_rate": 5.576113578589035e-07,
+      "logits/chosen": -1.9082481861114502,
+      "logits/rejected": -1.9050118923187256,
+      "logps/chosen": -31.349285125732422,
+      "logps/rejected": -33.84658432006836,
+      "loss": 0.7796,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.17784307897090912,
+      "rewards/margins": 0.2509470283985138,
+      "rewards/rejected": -0.07310393452644348,
       "step": 310
     },
     {
       "epoch": 0.83,
-      "grad_norm": 42.5,
-      "learning_rate": 4.825108134172131e-06,
-      "logits/chosen": -1.9775402545928955,
-      "logits/rejected": -1.9688619375228882,
-      "logps/chosen": -31.920490264892578,
-      "logps/rejected": -30.43929672241211,
-      "loss": 0.712,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 0.17379958927631378,
-      "rewards/margins": 0.2529620826244354,
-      "rewards/rejected": -0.07916250824928284,
+      "learning_rate": 4.229036944380913e-07,
+      "logits/chosen": -1.9580894708633423,
+      "logits/rejected": -1.9458973407745361,
+      "logps/chosen": -34.3031005859375,
+      "logps/rejected": -33.67659378051758,
+      "loss": 0.7302,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.2050826996564865,
+      "rewards/margins": 0.3015114367008209,
+      "rewards/rejected": -0.09642868489027023,
       "step": 320
     },
     {
       "epoch": 0.86,
-      "grad_norm": 88.0,
-      "learning_rate": 4.80369052967602e-06,
-      "logits/chosen": -1.916343331336975,
-      "logits/rejected": -1.928449273109436,
-      "logps/chosen": -29.984766006469727,
-      "logps/rejected": -33.69157791137695,
-      "loss": 0.603,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 0.19780051708221436,
-      "rewards/margins": 0.3135302662849426,
-      "rewards/rejected": -0.11572977155447006,
+      "learning_rate": 3.053082288996112e-07,
+      "logits/chosen": -1.9932842254638672,
+      "logits/rejected": -1.9918495416641235,
+      "logps/chosen": -33.17847442626953,
+      "logps/rejected": -32.54157638549805,
+      "loss": 0.7677,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.21512338519096375,
+      "rewards/margins": 0.2585209906101227,
+      "rewards/rejected": -0.04339758679270744,
       "step": 330
     },
     {
       "epoch": 0.88,
-      "grad_norm": 93.0,
-      "learning_rate": 4.781089396387968e-06,
-      "logits/chosen": -1.8849292993545532,
-      "logits/rejected": -1.8756763935089111,
-      "logps/chosen": -34.16048812866211,
-      "logps/rejected": -36.131134033203125,
-      "loss": 0.854,
-      "rewards/accuracies": 0.887499988079071,
-      "rewards/chosen": 0.21537461876869202,
-      "rewards/margins": 0.32513895630836487,
-      "rewards/rejected": -0.10976432263851166,
+      "learning_rate": 2.0579377374915805e-07,
+      "logits/chosen": -2.0800719261169434,
+      "logits/rejected": -2.064396381378174,
+      "logps/chosen": -33.80484390258789,
+      "logps/rejected": -33.1123046875,
+      "loss": 0.7636,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.2621825039386749,
+      "rewards/margins": 0.25817227363586426,
+      "rewards/rejected": 0.004010227043181658,
       "step": 340
     },
     {
       "epoch": 0.91,
-      "grad_norm": 64.0,
-      "learning_rate": 4.757316345716554e-06,
-      "logits/chosen": -1.94269597530365,
-      "logits/rejected": -1.9433438777923584,
-      "logps/chosen": -33.91167068481445,
-      "logps/rejected": -34.086830139160156,
-      "loss": 0.6834,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 0.1974446177482605,
-      "rewards/margins": 0.29258665442466736,
-      "rewards/rejected": -0.09514202922582626,
+      "learning_rate": 1.2518018074041684e-07,
+      "logits/chosen": -1.9522559642791748,
+      "logits/rejected": -1.951424241065979,
+      "logps/chosen": -32.8499755859375,
+      "logps/rejected": -32.56407165527344,
+      "loss": 0.6881,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.3024839758872986,
+      "rewards/margins": 0.3717316687107086,
+      "rewards/rejected": -0.06924761831760406,
       "step": 350
     },
     {
       "epoch": 0.94,
-      "grad_norm": 81.0,
-      "learning_rate": 4.73238359114687e-06,
-      "logits/chosen": -2.0718166828155518,
-      "logits/rejected": -2.0780444145202637,
-      "logps/chosen": -31.210739135742188,
-      "logps/rejected": -32.90790557861328,
-      "loss": 0.7717,
-      "rewards/accuracies": 0.824999988079071,
-      "rewards/chosen": 0.14500285685062408,
-      "rewards/margins": 0.24085240066051483,
-      "rewards/rejected": -0.09584951400756836,
+      "learning_rate": 6.41315865106129e-08,
+      "logits/chosen": -1.9075695276260376,
+      "logits/rejected": -1.917851209640503,
+      "logps/chosen": -31.882221221923828,
+      "logps/rejected": -35.31555938720703,
+      "loss": 0.7689,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.22572879493236542,
+      "rewards/margins": 0.24449090659618378,
+      "rewards/rejected": -0.018762132152915,
       "step": 360
     },
     {
       "epoch": 0.96,
-      "grad_norm": 200.0,
-      "learning_rate": 4.706303941965804e-06,
-      "logits/chosen": -2.0039541721343994,
-      "logits/rejected": -2.0034594535827637,
-      "logps/chosen": -32.95136260986328,
-      "logps/rejected": -36.23757553100586,
-      "loss": 0.7959,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 0.19742996990680695,
-      "rewards/margins": 0.2725493013858795,
-      "rewards/rejected": -0.07511930912733078,
+      "learning_rate": 2.3150941078050325e-08,
+      "logits/chosen": -2.0478641986846924,
+      "logits/rejected": -2.041414737701416,
+      "logps/chosen": -33.331912994384766,
+      "logps/rejected": -29.259756088256836,
+      "loss": 0.7658,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.2240387201309204,
+      "rewards/margins": 0.2473684549331665,
+      "rewards/rejected": -0.02332974039018154,
       "step": 370
     },
     {
       "epoch": 0.99,
-      "grad_norm": 67.0,
-      "learning_rate": 4.679090796681225e-06,
-      "logits/chosen": -2.038403272628784,
-      "logits/rejected": -2.0337538719177246,
-      "logps/chosen": -30.219844818115234,
-      "logps/rejected": -29.44011878967285,
-      "loss": 0.6897,
-      "rewards/accuracies": 0.887499988079071,
-      "rewards/chosen": 0.1963024139404297,
-      "rewards/margins": 0.286792516708374,
-      "rewards/rejected": -0.09049010276794434,
+      "learning_rate": 2.575864278703266e-09,
+      "logits/chosen": -1.9081246852874756,
+      "logits/rejected": -1.9103105068206787,
+      "logps/chosen": -33.882568359375,
+      "logps/rejected": -30.96805191040039,
+      "loss": 0.741,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.24953576922416687,
+      "rewards/margins": 0.3163323998451233,
+      "rewards/rejected": -0.06679664552211761,
       "step": 380
     },
     {
-      "epoch": 1.01,
-      "grad_norm": 68.0,
-      "learning_rate": 4.650758136138454e-06,
-      "logits/chosen": -1.8102906942367554,
-      "logits/rejected": -1.8167164325714111,
-      "logps/chosen": -31.897790908813477,
-      "logps/rejected": -36.53219985961914,
-      "loss": 0.8214,
-      "rewards/accuracies": 0.9125000238418579,
-      "rewards/chosen": 0.1951691210269928,
-      "rewards/margins": 0.36073189973831177,
-      "rewards/rejected": -0.16556277871131897,
-      "step": 390
-    },
-    {
-      "epoch": 1.04,
-      "grad_norm": 54.0,
-      "learning_rate": 4.621320516337559e-06,
-      "logits/chosen": -1.9675865173339844,
-      "logits/rejected": -1.961277723312378,
-      "logps/chosen": -33.26747131347656,
-      "logps/rejected": -32.52979278564453,
-      "loss": 0.7239,
-      "rewards/accuracies": 0.824999988079071,
-      "rewards/chosen": 0.226090669631958,
-      "rewards/margins": 0.3144562244415283,
-      "rewards/rejected": -0.08836553245782852,
-      "step": 400
-    },
-    {
-      "epoch": 1.04,
-      "eval_logits/chosen": -2.2324159145355225,
-      "eval_logits/rejected": -2.227566957473755,
-      "eval_logps/chosen": -34.08934783935547,
-      "eval_logps/rejected": -37.594459533691406,
-      "eval_loss": 1.646134853363037,
-      "eval_rewards/accuracies": 0.5340532064437866,
-      "eval_rewards/chosen": -0.02191990241408348,
-      "eval_rewards/margins": 0.009215297177433968,
-      "eval_rewards/rejected": -0.03113519586622715,
-      "eval_runtime": 145.6073,
-      "eval_samples_per_second": 2.356,
-      "eval_steps_per_second": 0.295,
-      "step": 400
-    },
-    {
-      "epoch": 1.06,
-      "grad_norm": 67.5,
-      "learning_rate": 4.590793060955158e-06,
-      "logits/chosen": -1.9676673412322998,
-      "logits/rejected": -1.9748865365982056,
-      "logps/chosen": -28.631023406982422,
-      "logps/rejected": -29.46906089782715,
-      "loss": 0.6247,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.1957395225763321,
-      "rewards/margins": 0.34407514333724976,
-      "rewards/rejected": -0.14833566546440125,
-      "step": 410
-    },
-    {
-      "epoch": 1.09,
-      "grad_norm": 75.5,
-      "learning_rate": 4.559191453574582e-06,
-      "logits/chosen": -1.9874868392944336,
-      "logits/rejected": -1.986395239830017,
-      "logps/chosen": -33.74576187133789,
-      "logps/rejected": -31.112499237060547,
-      "loss": 0.8861,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": 0.20197901129722595,
-      "rewards/margins": 0.2697620093822479,
-      "rewards/rejected": -0.06778299063444138,
-      "step": 420
-    },
-    {
-      "epoch": 1.12,
-      "grad_norm": 117.0,
-      "learning_rate": 4.52653192962838e-06,
-      "logits/chosen": -1.9822708368301392,
-      "logits/rejected": -1.964942216873169,
-      "logps/chosen": -30.64545249938965,
-      "logps/rejected": -33.29753112792969,
-      "loss": 0.9065,
-      "rewards/accuracies": 0.887499988079071,
-      "rewards/chosen": 0.13270007073879242,
-      "rewards/margins": 0.2811988890171051,
-      "rewards/rejected": -0.1484988033771515,
-      "step": 430
-    },
-    {
-      "epoch": 1.14,
-      "grad_norm": 49.0,
-      "learning_rate": 4.492831268057307e-06,
-      "logits/chosen": -2.014035224914551,
-      "logits/rejected": -2.016022205352783,
-      "logps/chosen": -36.08784866333008,
-      "logps/rejected": -35.217376708984375,
-      "loss": 0.5593,
-      "rewards/accuracies": 0.887499988079071,
-      "rewards/chosen": 0.20980596542358398,
-      "rewards/margins": 0.35277143120765686,
-      "rewards/rejected": -0.14296546578407288,
-      "step": 440
-    },
-    {
-      "epoch": 1.17,
-      "grad_norm": 55.75,
-      "learning_rate": 4.458106782690094e-06,
-      "logits/chosen": -2.093276262283325,
-      "logits/rejected": -2.093129873275757,
-      "logps/chosen": -32.13650894165039,
-      "logps/rejected": -33.764305114746094,
-      "loss": 0.8816,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 0.198737233877182,
-      "rewards/margins": 0.2804142236709595,
-      "rewards/rejected": -0.08167698234319687,
-      "step": 450
-    },
-    {
-      "epoch": 1.19,
-      "grad_norm": 100.5,
-      "learning_rate": 4.422376313348405e-06,
-      "logits/chosen": -2.038951873779297,
-      "logits/rejected": -2.0315186977386475,
-      "logps/chosen": -31.86245346069336,
-      "logps/rejected": -36.433692932128906,
-      "loss": 0.7285,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/chosen": 0.21740083396434784,
-      "rewards/margins": 0.36926203966140747,
-      "rewards/rejected": -0.15186122059822083,
-      "step": 460
-    },
-    {
-      "epoch": 1.22,
-      "grad_norm": 126.5,
-      "learning_rate": 4.3856582166815696e-06,
-      "logits/chosen": -1.940791130065918,
-      "logits/rejected": -1.9372183084487915,
-      "logps/chosen": -33.40504837036133,
-      "logps/rejected": -33.189998626708984,
-      "loss": 0.609,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 0.21089871227741241,
-      "rewards/margins": 0.3173474669456482,
-      "rewards/rejected": -0.10644875466823578,
-      "step": 470
-    },
-    {
-      "epoch": 1.25,
-      "grad_norm": 104.5,
-      "learning_rate": 4.347971356735789e-06,
-      "logits/chosen": -2.065157651901245,
-      "logits/rejected": -2.0581443309783936,
-      "logps/chosen": -30.389184951782227,
-      "logps/rejected": -32.29039764404297,
-      "loss": 0.7324,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 0.17574059963226318,
-      "rewards/margins": 0.29450365900993347,
-      "rewards/rejected": -0.11876306682825089,
-      "step": 480
-    },
-    {
-      "epoch": 1.27,
-      "grad_norm": 68.0,
-      "learning_rate": 4.309335095262675e-06,
-      "logits/chosen": -2.0097293853759766,
-      "logits/rejected": -2.0112791061401367,
-      "logps/chosen": -34.72505569458008,
-      "logps/rejected": -33.918052673339844,
-      "loss": 0.5935,
-      "rewards/accuracies": 0.887499988079071,
-      "rewards/chosen": 0.2512527406215668,
-      "rewards/margins": 0.3618691861629486,
-      "rewards/rejected": -0.11061643064022064,
-      "step": 490
-    },
-    {
-      "epoch": 1.3,
-      "grad_norm": 107.0,
-      "learning_rate": 4.269769281772082e-06,
-      "logits/chosen": -1.8976560831069946,
-      "logits/rejected": -1.8952856063842773,
-      "logps/chosen": -32.58599090576172,
-      "logps/rejected": -36.84335708618164,
-      "loss": 0.6061,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.23451784253120422,
-      "rewards/margins": 0.38613152503967285,
-      "rewards/rejected": -0.15161368250846863,
-      "step": 500
-    },
-    {
-      "epoch": 1.3,
-      "eval_logits/chosen": -2.2330281734466553,
-      "eval_logits/rejected": -2.228160858154297,
-      "eval_logps/chosen": -34.10296630859375,
-      "eval_logps/rejected": -37.62388229370117,
-      "eval_loss": 1.6487250328063965,
-      "eval_rewards/accuracies": 0.5394518375396729,
-      "eval_rewards/chosen": -0.02736487239599228,
-      "eval_rewards/margins": 0.015538448467850685,
-      "eval_rewards/rejected": -0.04290332272648811,
-      "eval_runtime": 145.2589,
-      "eval_samples_per_second": 2.361,
-      "eval_steps_per_second": 0.296,
-      "step": 500
-    },
-    {
-      "epoch": 1.32,
-      "grad_norm": 63.5,
-      "learning_rate": 4.22929424333435e-06,
-      "logits/chosen": -2.002866744995117,
-      "logits/rejected": -2.007615566253662,
-      "logps/chosen": -32.78615188598633,
-      "logps/rejected": -31.923254013061523,
-      "loss": 0.7109,
-      "rewards/accuracies": 0.887499988079071,
-      "rewards/chosen": 0.2684818208217621,
-      "rewards/margins": 0.37934139370918274,
-      "rewards/rejected": -0.11085958778858185,
-      "step": 510
-    },
-    {
-      "epoch": 1.35,
-      "grad_norm": 90.0,
-      "learning_rate": 4.1879307741372085e-06,
-      "logits/chosen": -2.034381151199341,
-      "logits/rejected": -2.0453076362609863,
-      "logps/chosen": -30.794418334960938,
-      "logps/rejected": -31.83915138244629,
-      "loss": 0.6671,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.2383977472782135,
-      "rewards/margins": 0.37098294496536255,
-      "rewards/rejected": -0.13258522748947144,
-      "step": 520
-    },
-    {
-      "epoch": 1.38,
-      "grad_norm": 62.25,
-      "learning_rate": 4.145700124802693e-06,
-      "logits/chosen": -1.9620320796966553,
-      "logits/rejected": -1.9585500955581665,
-      "logps/chosen": -31.819278717041016,
-      "logps/rejected": -32.71375274658203,
-      "loss": 0.6704,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.24049386382102966,
-      "rewards/margins": 0.38088440895080566,
-      "rewards/rejected": -0.14039050042629242,
-      "step": 530
-    },
-    {
-      "epoch": 1.4,
-      "grad_norm": 50.5,
-      "learning_rate": 4.102623991469562e-06,
-      "logits/chosen": -1.8313617706298828,
-      "logits/rejected": -1.84051513671875,
-      "logps/chosen": -32.123390197753906,
-      "logps/rejected": -32.177490234375,
-      "loss": 0.6783,
-      "rewards/accuracies": 0.887499988079071,
-      "rewards/chosen": 0.17839959263801575,
-      "rewards/margins": 0.30700013041496277,
-      "rewards/rejected": -0.12860055267810822,
-      "step": 540
-    },
-    {
-      "epoch": 1.43,
-      "grad_norm": 72.5,
-      "learning_rate": 4.058724504646834e-06,
-      "logits/chosen": -1.9312222003936768,
-      "logits/rejected": -1.9248243570327759,
-      "logps/chosen": -33.194034576416016,
-      "logps/rejected": -31.20892906188965,
-      "loss": 0.6703,
-      "rewards/accuracies": 0.8374999761581421,
-      "rewards/chosen": 0.2057083547115326,
-      "rewards/margins": 0.29274117946624756,
-      "rewards/rejected": -0.08703283220529556,
-      "step": 550
-    },
-    {
-      "epoch": 1.45,
-      "grad_norm": 141.0,
-      "learning_rate": 4.014024217844167e-06,
-      "logits/chosen": -2.0215349197387695,
-      "logits/rejected": -2.019446849822998,
-      "logps/chosen": -33.937469482421875,
-      "logps/rejected": -31.723003387451172,
-      "loss": 0.9186,
-      "rewards/accuracies": 0.8374999761581421,
-      "rewards/chosen": 0.25349074602127075,
-      "rewards/margins": 0.34373739361763,
-      "rewards/rejected": -0.09024664759635925,
-      "step": 560
-    },
-    {
-      "epoch": 1.48,
-      "grad_norm": 62.25,
-      "learning_rate": 3.968546095984911e-06,
-      "logits/chosen": -1.8566452264785767,
-      "logits/rejected": -1.8544600009918213,
-      "logps/chosen": -32.25889587402344,
-      "logps/rejected": -31.205657958984375,
-      "loss": 0.8459,
-      "rewards/accuracies": 0.8125,
-      "rewards/chosen": 0.21500520408153534,
-      "rewards/margins": 0.2963467836380005,
-      "rewards/rejected": -0.08134157210588455,
-      "step": 570
-    },
-    {
-      "epoch": 1.51,
-      "grad_norm": 46.75,
-      "learning_rate": 3.922313503607806e-06,
-      "logits/chosen": -1.9905545711517334,
-      "logits/rejected": -1.987287163734436,
-      "logps/chosen": -30.5700740814209,
-      "logps/rejected": -34.977298736572266,
-      "loss": 0.721,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/chosen": 0.21382951736450195,
-      "rewards/margins": 0.3542833626270294,
-      "rewards/rejected": -0.14045384526252747,
-      "step": 580
-    },
-    {
-      "epoch": 1.53,
-      "grad_norm": 38.0,
-      "learning_rate": 3.875350192863368e-06,
-      "logits/chosen": -1.9324638843536377,
-      "logits/rejected": -1.9361374378204346,
-      "logps/chosen": -29.091140747070312,
-      "logps/rejected": -30.924423217773438,
-      "loss": 0.6955,
-      "rewards/accuracies": 0.862500011920929,
-      "rewards/chosen": 0.20955555140972137,
-      "rewards/margins": 0.31577685475349426,
-      "rewards/rejected": -0.1062212809920311,
-      "step": 590
-    },
-    {
-      "epoch": 1.56,
-      "grad_norm": 75.5,
-      "learning_rate": 3.8276802913111436e-06,
-      "logits/chosen": -1.9715423583984375,
-      "logits/rejected": -1.9713079929351807,
-      "logps/chosen": -31.609033584594727,
-      "logps/rejected": -31.35397720336914,
-      "loss": 0.9255,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 0.24499893188476562,
-      "rewards/margins": 0.3093833923339844,
-      "rewards/rejected": -0.06438445299863815,
-      "step": 600
-    },
-    {
-      "epoch": 1.56,
-      "eval_logits/chosen": -2.232149600982666,
-      "eval_logits/rejected": -2.227276086807251,
-      "eval_logps/chosen": -34.00742721557617,
-      "eval_logps/rejected": -37.54637908935547,
-      "eval_loss": 1.591224193572998,
-      "eval_rewards/accuracies": 0.5544019937515259,
-      "eval_rewards/chosen": 0.010849811136722565,
-      "eval_rewards/margins": 0.022752860561013222,
-      "eval_rewards/rejected": -0.011903048492968082,
-      "eval_runtime": 145.3935,
-      "eval_samples_per_second": 2.359,
-      "eval_steps_per_second": 0.296,
-      "step": 600
-    },
-    {
-      "epoch": 1.58,
-      "grad_norm": 159.0,
-      "learning_rate": 3.7793282895240927e-06,
-      "logits/chosen": -2.0326523780822754,
-      "logits/rejected": -2.0333547592163086,
-      "logps/chosen": -34.38473129272461,
-      "logps/rejected": -33.29423141479492,
-      "loss": 0.5891,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 0.25816404819488525,
-      "rewards/margins": 0.3421815037727356,
-      "rewards/rejected": -0.08401744067668915,
-      "step": 610
-    },
-    {
-      "epoch": 1.61,
-      "grad_norm": 91.0,
-      "learning_rate": 3.730319028506478e-06,
-      "logits/chosen": -1.9972617626190186,
-      "logits/rejected": -1.994637131690979,
-      "logps/chosen": -32.45311737060547,
-      "logps/rejected": -32.34403991699219,
-      "loss": 0.5185,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/chosen": 0.2689127027988434,
-      "rewards/margins": 0.378562867641449,
-      "rewards/rejected": -0.10965017974376678,
-      "step": 620
-    },
-    {
-      "epoch": 1.64,
-      "grad_norm": 113.5,
-      "learning_rate": 3.6806776869317074e-06,
-      "logits/chosen": -2.012416124343872,
-      "logits/rejected": -2.003248929977417,
-      "logps/chosen": -32.12730026245117,
-      "logps/rejected": -31.12014389038086,
-      "loss": 0.7937,
-      "rewards/accuracies": 0.824999988079071,
-      "rewards/chosen": 0.2595270574092865,
-      "rewards/margins": 0.33343806862831116,
-      "rewards/rejected": -0.07391100376844406,
-      "step": 630
-    },
-    {
-      "epoch": 1.66,
-      "grad_norm": 75.0,
-      "learning_rate": 3.6304297682067146e-06,
-      "logits/chosen": -2.016336441040039,
-      "logits/rejected": -2.013164758682251,
-      "logps/chosen": -31.525562286376953,
-      "logps/rejected": -32.66303634643555,
-      "loss": 0.4959,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.26914453506469727,
-      "rewards/margins": 0.36659201979637146,
-      "rewards/rejected": -0.097447469830513,
-      "step": 640
-    },
-    {
-      "epoch": 1.69,
-      "grad_norm": 48.0,
-      "learning_rate": 3.579601087369492e-06,
-      "logits/chosen": -2.0207347869873047,
-      "logits/rejected": -2.023106575012207,
-      "logps/chosen": -32.97084426879883,
-      "logps/rejected": -34.033267974853516,
-      "loss": 0.4312,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.2731453478336334,
-      "rewards/margins": 0.413092702627182,
-      "rewards/rejected": -0.1399472951889038,
-      "step": 650
-    },
-    {
-      "epoch": 1.71,
-      "grad_norm": 48.0,
-      "learning_rate": 3.5282177578265295e-06,
-      "logits/chosen": -1.934343695640564,
-      "logits/rejected": -1.9348455667495728,
-      "logps/chosen": -33.32109832763672,
-      "logps/rejected": -31.806873321533203,
-      "loss": 0.6159,
-      "rewards/accuracies": 0.9125000238418579,
-      "rewards/chosen": 0.25029245018959045,
-      "rewards/margins": 0.33013930916786194,
-      "rewards/rejected": -0.0798468366265297,
-      "step": 660
-    },
-    {
-      "epoch": 1.74,
-      "grad_norm": 51.0,
-      "learning_rate": 3.476306177936961e-06,
-      "logits/chosen": -1.9735260009765625,
-      "logits/rejected": -1.9637184143066406,
-      "logps/chosen": -33.36211013793945,
-      "logps/rejected": -32.63333511352539,
-      "loss": 0.4666,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.25566938519477844,
-      "rewards/margins": 0.38588738441467285,
-      "rewards/rejected": -0.13021798431873322,
-      "step": 670
-    },
-    {
-      "epoch": 1.77,
-      "grad_norm": 56.0,
-      "learning_rate": 3.423893017450324e-06,
-      "logits/chosen": -1.8723722696304321,
-      "logits/rejected": -1.8691799640655518,
-      "logps/chosen": -30.749767303466797,
-      "logps/rejected": -34.51530838012695,
-      "loss": 0.5406,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/chosen": 0.2844772934913635,
-      "rewards/margins": 0.3631672263145447,
-      "rewards/rejected": -0.07868987321853638,
-      "step": 680
-    },
-    {
-      "epoch": 1.79,
-      "grad_norm": 96.0,
-      "learning_rate": 3.3710052038048794e-06,
-      "logits/chosen": -1.931316614151001,
-      "logits/rejected": -1.9304965734481812,
-      "logps/chosen": -34.53691864013672,
-      "logps/rejected": -35.59252166748047,
-      "loss": 0.5253,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.2933480739593506,
-      "rewards/margins": 0.38882699608802795,
-      "rewards/rejected": -0.09547892212867737,
-      "step": 690
-    },
-    {
-      "epoch": 1.82,
-      "grad_norm": 94.5,
-      "learning_rate": 3.3176699082935546e-06,
-      "logits/chosen": -1.9118010997772217,
-      "logits/rejected": -1.9148480892181396,
-      "logps/chosen": -31.930761337280273,
-      "logps/rejected": -35.87628173828125,
-      "loss": 0.8252,
-      "rewards/accuracies": 0.8374999761581421,
-      "rewards/chosen": 0.2688351273536682,
-      "rewards/margins": 0.36267900466918945,
-      "rewards/rejected": -0.09384389221668243,
-      "step": 700
-    },
-    {
-      "epoch": 1.82,
-      "eval_logits/chosen": -2.2346324920654297,
-      "eval_logits/rejected": -2.2297565937042236,
-      "eval_logps/chosen": -33.97812271118164,
-      "eval_logps/rejected": -37.50526428222656,
-      "eval_loss": 1.6334478855133057,
-      "eval_rewards/accuracies": 0.5215947031974792,
-      "eval_rewards/chosen": 0.022571848705410957,
-      "eval_rewards/margins": 0.018028803169727325,
-      "eval_rewards/rejected": 0.004543047398328781,
-      "eval_runtime": 145.3597,
-      "eval_samples_per_second": 2.36,
-      "eval_steps_per_second": 0.296,
-      "step": 700
-    },
-    {
-      "epoch": 1.84,
-      "grad_norm": 62.75,
-      "learning_rate": 3.2639145321045933e-06,
-      "logits/chosen": -2.0267205238342285,
-      "logits/rejected": -2.0294930934906006,
-      "logps/chosen": -34.24456024169922,
-      "logps/rejected": -34.45062255859375,
-      "loss": 0.8568,
-      "rewards/accuracies": 0.8374999761581421,
-      "rewards/chosen": 0.2604122459888458,
-      "rewards/margins": 0.3325996696949005,
-      "rewards/rejected": -0.0721873790025711,
-      "step": 710
-    },
-    {
-      "epoch": 1.87,
-      "grad_norm": 105.0,
-      "learning_rate": 3.2097666922441107e-06,
-      "logits/chosen": -1.8829209804534912,
-      "logits/rejected": -1.877000093460083,
-      "logps/chosen": -34.198673248291016,
-      "logps/rejected": -32.5337028503418,
-      "loss": 0.9437,
-      "rewards/accuracies": 0.8374999761581421,
-      "rewards/chosen": 0.25382402539253235,
-      "rewards/margins": 0.3265684247016907,
-      "rewards/rejected": -0.07274442911148071,
-      "step": 720
-    },
-    {
-      "epoch": 1.9,
-      "grad_norm": 35.75,
-      "learning_rate": 3.1552542073477554e-06,
-      "logits/chosen": -2.0574429035186768,
-      "logits/rejected": -2.0544135570526123,
-      "logps/chosen": -30.041122436523438,
-      "logps/rejected": -31.79475975036621,
-      "loss": 0.5771,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.24474552273750305,
-      "rewards/margins": 0.38337013125419617,
-      "rewards/rejected": -0.13862457871437073,
-      "step": 730
-    },
-    {
-      "epoch": 1.92,
-      "grad_norm": 46.75,
-      "learning_rate": 3.100405083388799e-06,
-      "logits/chosen": -1.8992414474487305,
-      "logits/rejected": -1.8990885019302368,
-      "logps/chosen": -32.76268768310547,
-      "logps/rejected": -37.8664436340332,
-      "loss": 0.5278,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.28019440174102783,
-      "rewards/margins": 0.3756488263607025,
-      "rewards/rejected": -0.09545440971851349,
-      "step": 740
-    },
-    {
-      "epoch": 1.95,
-      "grad_norm": 87.0,
-      "learning_rate": 3.0452474992899645e-06,
-      "logits/chosen": -1.7812280654907227,
-      "logits/rejected": -1.7864673137664795,
-      "logps/chosen": -36.23480987548828,
-      "logps/rejected": -34.09480667114258,
-      "loss": 0.6358,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 0.2790614366531372,
-      "rewards/margins": 0.32804808020591736,
-      "rewards/rejected": -0.048986706882715225,
-      "step": 750
-    },
-    {
-      "epoch": 1.97,
-      "grad_norm": 37.5,
-      "learning_rate": 2.989809792446417e-06,
-      "logits/chosen": -1.9918644428253174,
-      "logits/rejected": -1.9932422637939453,
-      "logps/chosen": -31.911479949951172,
-      "logps/rejected": -32.96862030029297,
-      "loss": 0.5173,
-      "rewards/accuracies": 0.9125000238418579,
-      "rewards/chosen": 0.2776804268360138,
-      "rewards/margins": 0.3534439206123352,
-      "rewards/rejected": -0.0757635161280632,
-      "step": 760
-    },
-    {
-      "epoch": 2.0,
-      "grad_norm": 101.0,
-      "learning_rate": 2.9341204441673267e-06,
-      "logits/chosen": -1.9615776538848877,
-      "logits/rejected": -1.960758924484253,
-      "logps/chosen": -31.484752655029297,
-      "logps/rejected": -35.08818817138672,
-      "loss": 0.5913,
-      "rewards/accuracies": 0.8708333969116211,
-      "rewards/chosen": 0.27634382247924805,
-      "rewards/margins": 0.30035752058029175,
-      "rewards/rejected": -0.024013686925172806,
-      "step": 770
-    },
-    {
-      "epoch": 2.03,
-      "grad_norm": 33.25,
-      "learning_rate": 2.878208065043501e-06,
-      "logits/chosen": -1.9611775875091553,
-      "logits/rejected": -1.960415244102478,
-      "logps/chosen": -33.879024505615234,
-      "logps/rejected": -31.674047470092773,
-      "loss": 0.3971,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.30777519941329956,
-      "rewards/margins": 0.4095862805843353,
-      "rewards/rejected": -0.10181108862161636,
-      "step": 780
-    },
-    {
-      "epoch": 2.05,
-      "grad_norm": 40.75,
-      "learning_rate": 2.8221013802485974e-06,
-      "logits/chosen": -1.9905250072479248,
-      "logits/rejected": -1.989000916481018,
-      "logps/chosen": -28.83389663696289,
-      "logps/rejected": -32.96330261230469,
-      "loss": 0.3133,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.29171431064605713,
-      "rewards/margins": 0.4225086271762848,
-      "rewards/rejected": -0.13079434633255005,
-      "step": 790
-    },
-    {
-      "epoch": 2.08,
-      "grad_norm": 76.0,
-      "learning_rate": 2.76582921478147e-06,
-      "logits/chosen": -2.0426809787750244,
-      "logits/rejected": -2.039111375808716,
-      "logps/chosen": -31.69073486328125,
-      "logps/rejected": -34.53063201904297,
-      "loss": 0.2848,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.2945896089076996,
-      "rewards/margins": 0.42722758650779724,
-      "rewards/rejected": -0.13263803720474243,
-      "step": 800
-    },
-    {
-      "epoch": 2.08,
-      "eval_logits/chosen": -2.2361207008361816,
-      "eval_logits/rejected": -2.231266736984253,
-      "eval_logps/chosen": -33.99641799926758,
-      "eval_logps/rejected": -37.524444580078125,
-      "eval_loss": 1.603250503540039,
-      "eval_rewards/accuracies": 0.52491694688797,
-      "eval_rewards/chosen": 0.015252894721925259,
-      "eval_rewards/margins": 0.018381815403699875,
-      "eval_rewards/rejected": -0.003128920216113329,
-      "eval_runtime": 145.2046,
-      "eval_samples_per_second": 2.362,
-      "eval_steps_per_second": 0.296,
-      "step": 800
-    },
-    {
-      "epoch": 2.1,
-      "grad_norm": 60.5,
-      "learning_rate": 2.7094204786572254e-06,
-      "logits/chosen": -1.8801984786987305,
-      "logits/rejected": -1.8722867965698242,
-      "logps/chosen": -33.65568542480469,
-      "logps/rejected": -34.81199645996094,
-      "loss": 0.3333,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.32022127509117126,
-      "rewards/margins": 0.42097288370132446,
-      "rewards/rejected": -0.10075156390666962,
-      "step": 810
-    },
-    {
-      "epoch": 2.13,
-      "grad_norm": 61.25,
-      "learning_rate": 2.6529041520546072e-06,
-      "logits/chosen": -1.955896019935608,
-      "logits/rejected": -1.9664968252182007,
-      "logps/chosen": -34.896785736083984,
-      "logps/rejected": -32.63465118408203,
-      "loss": 0.4459,
-      "rewards/accuracies": 0.9125000238418579,
-      "rewards/chosen": 0.31956198811531067,
-      "rewards/margins": 0.4008457660675049,
-      "rewards/rejected": -0.0812838152050972,
-      "step": 820
-    },
-    {
-      "epoch": 2.16,
-      "grad_norm": 84.0,
-      "learning_rate": 2.5963092704273302e-06,
-      "logits/chosen": -2.0049898624420166,
-      "logits/rejected": -2.0092310905456543,
-      "logps/chosen": -34.0383415222168,
-      "logps/rejected": -29.4811954498291,
-      "loss": 0.2762,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.34577542543411255,
-      "rewards/margins": 0.43095555901527405,
-      "rewards/rejected": -0.0851801335811615,
-      "step": 830
-    },
-    {
-      "epoch": 2.18,
-      "grad_norm": 82.0,
-      "learning_rate": 2.53966490958702e-06,
-      "logits/chosen": -2.0014491081237793,
-      "logits/rejected": -2.009239912033081,
-      "logps/chosen": -34.01618194580078,
-      "logps/rejected": -30.190011978149414,
-      "loss": 0.496,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.30874118208885193,
-      "rewards/margins": 0.4456659257411957,
-      "rewards/rejected": -0.13692477345466614,
-      "step": 840
-    },
-    {
-      "epoch": 2.21,
-      "grad_norm": 40.0,
-      "learning_rate": 2.4830001707654135e-06,
-      "logits/chosen": -1.930978536605835,
-      "logits/rejected": -1.9214372634887695,
-      "logps/chosen": -31.05685043334961,
-      "logps/rejected": -32.36315155029297,
-      "loss": 0.2819,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.29453787207603455,
-      "rewards/margins": 0.43427667021751404,
-      "rewards/rejected": -0.1397387832403183,
-      "step": 850
-    },
-    {
-      "epoch": 2.23,
-      "grad_norm": 34.75,
-      "learning_rate": 2.4263441656635054e-06,
-      "logits/chosen": -2.072263717651367,
-      "logits/rejected": -2.062443971633911,
-      "logps/chosen": -25.44430923461914,
-      "logps/rejected": -30.206756591796875,
-      "loss": 0.4507,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.3119046986103058,
-      "rewards/margins": 0.44567227363586426,
-      "rewards/rejected": -0.13376764953136444,
-      "step": 860
-    },
-    {
-      "epoch": 2.26,
-      "grad_norm": 89.0,
-      "learning_rate": 2.3697260014953107e-06,
-      "logits/chosen": -1.9230930805206299,
-      "logits/rejected": -1.9242877960205078,
-      "logps/chosen": -33.56352233886719,
-      "logps/rejected": -30.285165786743164,
-      "loss": 0.3499,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.3077928125858307,
-      "rewards/margins": 0.42885905504226685,
-      "rewards/rejected": -0.12106621265411377,
-      "step": 870
-    },
-    {
-      "epoch": 2.29,
-      "grad_norm": 29.5,
-      "learning_rate": 2.3131747660339396e-06,
-      "logits/chosen": -1.9387619495391846,
-      "logits/rejected": -1.9395242929458618,
-      "logps/chosen": -31.70212745666504,
-      "logps/rejected": -33.355247497558594,
-      "loss": 0.5159,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/chosen": 0.27484866976737976,
-      "rewards/margins": 0.3965802788734436,
-      "rewards/rejected": -0.12173161655664444,
-      "step": 880
-    },
-    {
-      "epoch": 2.31,
-      "grad_norm": 63.5,
-      "learning_rate": 2.256719512667651e-06,
-      "logits/chosen": -1.8424198627471924,
-      "logits/rejected": -1.8406569957733154,
-      "logps/chosen": -34.8422966003418,
-      "logps/rejected": -36.311790466308594,
-      "loss": 0.4473,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.3290487825870514,
-      "rewards/margins": 0.5028699636459351,
-      "rewards/rejected": -0.17382115125656128,
-      "step": 890
-    },
-    {
-      "epoch": 2.34,
-      "grad_norm": 120.5,
-      "learning_rate": 2.2003892454735786e-06,
-      "logits/chosen": -1.9881842136383057,
-      "logits/rejected": -1.9813495874404907,
-      "logps/chosen": -31.175704956054688,
-      "logps/rejected": -32.99733352661133,
-      "loss": 0.3671,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.3307729959487915,
-      "rewards/margins": 0.42647916078567505,
-      "rewards/rejected": -0.09570614993572235,
-      "step": 900
-    },
-    {
-      "epoch": 2.34,
-      "eval_logits/chosen": -2.235804796218872,
-      "eval_logits/rejected": -2.230922222137451,
-      "eval_logps/chosen": -33.96369552612305,
-      "eval_logps/rejected": -37.47232437133789,
-      "eval_loss": 1.6569337844848633,
-      "eval_rewards/accuracies": 0.5161960124969482,
-      "eval_rewards/chosen": 0.028342105448246002,
-      "eval_rewards/margins": 0.010623876005411148,
-      "eval_rewards/rejected": 0.017718229442834854,
-      "eval_runtime": 145.1552,
-      "eval_samples_per_second": 2.363,
-      "eval_steps_per_second": 0.296,
-      "step": 900
-    },
-    {
-      "epoch": 2.36,
-      "grad_norm": 49.25,
-      "learning_rate": 2.1442129043167877e-06,
-      "logits/chosen": -2.0350537300109863,
-      "logits/rejected": -2.0302882194519043,
-      "logps/chosen": -32.76664733886719,
-      "logps/rejected": -34.81679916381836,
-      "loss": 0.309,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.3164145350456238,
-      "rewards/margins": 0.4316856861114502,
-      "rewards/rejected": -0.1152711883187294,
-      "step": 910
-    },
-    {
-      "epoch": 2.39,
-      "grad_norm": 59.5,
-      "learning_rate": 2.088219349982323e-06,
-      "logits/chosen": -1.9527454376220703,
-      "logits/rejected": -1.9576810598373413,
-      "logps/chosen": -34.50981903076172,
-      "logps/rejected": -32.73914337158203,
-      "loss": 0.4281,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.3355061411857605,
-      "rewards/margins": 0.43119993805885315,
-      "rewards/rejected": -0.09569380432367325,
-      "step": 920
-    },
-    {
-      "epoch": 2.42,
-      "grad_norm": 58.25,
-      "learning_rate": 2.0324373493478803e-06,
-      "logits/chosen": -2.0501203536987305,
-      "logits/rejected": -2.0407142639160156,
-      "logps/chosen": -31.465560913085938,
-      "logps/rejected": -34.783172607421875,
-      "loss": 0.3184,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.3277474045753479,
-      "rewards/margins": 0.4638980031013489,
-      "rewards/rejected": -0.13615058362483978,
-      "step": 930
-    },
-    {
-      "epoch": 2.44,
-      "grad_norm": 45.5,
-      "learning_rate": 1.976895560604729e-06,
-      "logits/chosen": -1.9754626750946045,
-      "logits/rejected": -1.9723625183105469,
-      "logps/chosen": -30.506328582763672,
-      "logps/rejected": -32.73186111450195,
-      "loss": 0.3318,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.3128509521484375,
-      "rewards/margins": 0.41944509744644165,
-      "rewards/rejected": -0.10659414529800415,
-      "step": 940
-    },
-    {
-      "epoch": 2.47,
-      "grad_norm": 21.875,
-      "learning_rate": 1.921622518534466e-06,
-      "logits/chosen": -1.9184865951538086,
-      "logits/rejected": -1.9259870052337646,
-      "logps/chosen": -32.09298324584961,
-      "logps/rejected": -35.809776306152344,
-      "loss": 0.5459,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/chosen": 0.3112277388572693,
-      "rewards/margins": 0.4151647686958313,
-      "rewards/rejected": -0.1039370447397232,
-      "step": 950
-    },
-    {
-      "epoch": 2.49,
-      "grad_norm": 67.5,
-      "learning_rate": 1.8666466198491794e-06,
-      "logits/chosen": -1.9273126125335693,
-      "logits/rejected": -1.9204585552215576,
-      "logps/chosen": -33.03001403808594,
-      "logps/rejected": -36.92626190185547,
-      "loss": 0.4169,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.31939974427223206,
-      "rewards/margins": 0.45038121938705444,
-      "rewards/rejected": -0.13098154962062836,
-      "step": 960
-    },
-    {
-      "epoch": 2.52,
-      "grad_norm": 18.375,
-      "learning_rate": 1.8119961086025376e-06,
-      "logits/chosen": -1.9290339946746826,
-      "logits/rejected": -1.9297443628311157,
-      "logps/chosen": -30.42807960510254,
-      "logps/rejected": -33.275390625,
-      "loss": 0.3536,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.32332831621170044,
-      "rewards/margins": 0.44098013639450073,
-      "rewards/rejected": -0.11765184253454208,
-      "step": 970
-    },
-    {
-      "epoch": 2.55,
-      "grad_norm": 45.5,
-      "learning_rate": 1.7576990616793139e-06,
-      "logits/chosen": -1.9345155954360962,
-      "logits/rejected": -1.9458023309707642,
-      "logps/chosen": -32.152503967285156,
-      "logps/rejected": -34.349021911621094,
-      "loss": 0.346,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.36967429518699646,
-      "rewards/margins": 0.4567973017692566,
-      "rewards/rejected": -0.08712298423051834,
-      "step": 980
-    },
-    {
-      "epoch": 2.57,
-      "grad_norm": 79.0,
-      "learning_rate": 1.7037833743707892e-06,
-      "logits/chosen": -2.0078938007354736,
-      "logits/rejected": -2.0053062438964844,
-      "logps/chosen": -35.25431823730469,
-      "logps/rejected": -32.59493637084961,
-      "loss": 0.4808,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/chosen": 0.34669190645217896,
-      "rewards/margins": 0.42497485876083374,
-      "rewards/rejected": -0.07828302681446075,
-      "step": 990
-    },
-    {
-      "epoch": 2.6,
-      "grad_norm": 44.0,
-      "learning_rate": 1.6502767460434588e-06,
-      "logits/chosen": -1.9761663675308228,
-      "logits/rejected": -1.9800007343292236,
-      "logps/chosen": -33.687164306640625,
-      "logps/rejected": -34.747779846191406,
-      "loss": 0.3936,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.3104504346847534,
-      "rewards/margins": 0.41514307260513306,
-      "rewards/rejected": -0.10469266027212143,
-      "step": 1000
-    },
-    {
-      "epoch": 2.6,
-      "eval_logits/chosen": -2.2373571395874023,
-      "eval_logits/rejected": -2.2324929237365723,
-      "eval_logps/chosen": -33.947471618652344,
-      "eval_logps/rejected": -37.469818115234375,
-      "eval_loss": 1.6203227043151855,
-      "eval_rewards/accuracies": 0.5427741408348083,
-      "eval_rewards/chosen": 0.03483137488365173,
-      "eval_rewards/margins": 0.0161101296544075,
-      "eval_rewards/rejected": 0.018721245229244232,
-      "eval_runtime": 145.2271,
-      "eval_samples_per_second": 2.362,
-      "eval_steps_per_second": 0.296,
-      "step": 1000
-    },
-    {
-      "epoch": 2.62,
-      "grad_norm": 103.5,
-      "learning_rate": 1.5972066659083796e-06,
-      "logits/chosen": -1.9862302541732788,
-      "logits/rejected": -1.9912649393081665,
-      "logps/chosen": -32.02253341674805,
-      "logps/rejected": -32.78570556640625,
-      "loss": 0.3762,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.3111560046672821,
-      "rewards/margins": 0.4169352650642395,
-      "rewards/rejected": -0.10577923059463501,
-      "step": 1010
-    },
-    {
-      "epoch": 2.65,
-      "grad_norm": 54.25,
-      "learning_rate": 1.5446003988985041e-06,
-      "logits/chosen": -1.948102355003357,
-      "logits/rejected": -1.9508020877838135,
-      "logps/chosen": -29.719890594482422,
-      "logps/rejected": -31.992984771728516,
-      "loss": 0.3461,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.29029980301856995,
-      "rewards/margins": 0.407827228307724,
-      "rewards/rejected": -0.11752742528915405,
-      "step": 1020
-    },
-    {
-      "epoch": 2.68,
-      "grad_norm": 60.0,
-      "learning_rate": 1.4924849716612211e-06,
-      "logits/chosen": -2.005060911178589,
-      "logits/rejected": -1.9991967678070068,
-      "logps/chosen": -33.16535568237305,
-      "logps/rejected": -33.78788757324219,
-      "loss": 0.2867,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.32950255274772644,
-      "rewards/margins": 0.43745851516723633,
-      "rewards/rejected": -0.10795596987009048,
-      "step": 1030
-    },
-    {
-      "epoch": 2.7,
-      "grad_norm": 69.0,
-      "learning_rate": 1.440887158673332e-06,
-      "logits/chosen": -1.9332466125488281,
-      "logits/rejected": -1.936981201171875,
-      "logps/chosen": -34.967830657958984,
-      "logps/rejected": -34.762996673583984,
-      "loss": 0.3317,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.314359575510025,
-      "rewards/margins": 0.4517282545566559,
-      "rewards/rejected": -0.13736867904663086,
-      "step": 1040
-    },
-    {
-      "epoch": 2.73,
-      "grad_norm": 26.875,
-      "learning_rate": 1.3898334684855647e-06,
-      "logits/chosen": -1.9331417083740234,
-      "logits/rejected": -1.9462738037109375,
-      "logps/chosen": -31.087848663330078,
-      "logps/rejected": -33.28511428833008,
-      "loss": 0.3932,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.31001484394073486,
-      "rewards/margins": 0.4160439074039459,
-      "rewards/rejected": -0.10602907836437225,
-      "step": 1050
-    },
-    {
-      "epoch": 2.75,
-      "grad_norm": 97.5,
-      "learning_rate": 1.3393501301037245e-06,
-      "logits/chosen": -1.9240528345108032,
-      "logits/rejected": -1.917449951171875,
-      "logps/chosen": -31.343292236328125,
-      "logps/rejected": -33.56601333618164,
-      "loss": 0.3354,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.3264150619506836,
-      "rewards/margins": 0.4256357252597809,
-      "rewards/rejected": -0.0992206335067749,
-      "step": 1060
-    },
-    {
-      "epoch": 2.78,
-      "grad_norm": 62.75,
-      "learning_rate": 1.2894630795134454e-06,
-      "logits/chosen": -2.051302433013916,
-      "logits/rejected": -2.0521297454833984,
-      "logps/chosen": -32.764705657958984,
-      "logps/rejected": -33.20236587524414,
-      "loss": 0.2782,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.312948077917099,
-      "rewards/margins": 0.45601654052734375,
-      "rewards/rejected": -0.14306847751140594,
-      "step": 1070
-    },
-    {
-      "epoch": 2.81,
-      "grad_norm": 36.5,
-      "learning_rate": 1.2401979463554984e-06,
-      "logits/chosen": -1.982011079788208,
-      "logits/rejected": -1.9808781147003174,
-      "logps/chosen": -33.330299377441406,
-      "logps/rejected": -33.3953971862793,
-      "loss": 0.5045,
-      "rewards/accuracies": 0.887499988079071,
-      "rewards/chosen": 0.3038138747215271,
-      "rewards/margins": 0.38524001836776733,
-      "rewards/rejected": -0.08142612874507904,
-      "step": 1080
-    },
-    {
-      "epoch": 2.83,
-      "grad_norm": 31.875,
-      "learning_rate": 1.1915800407584705e-06,
-      "logits/chosen": -2.01176118850708,
-      "logits/rejected": -2.004225969314575,
-      "logps/chosen": -33.374778747558594,
-      "logps/rejected": -31.283130645751953,
-      "loss": 0.2661,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.34043893218040466,
-      "rewards/margins": 0.4370775818824768,
-      "rewards/rejected": -0.09663861989974976,
-      "step": 1090
-    },
-    {
-      "epoch": 2.86,
-      "grad_norm": 102.5,
-      "learning_rate": 1.1436343403356019e-06,
-      "logits/chosen": -1.9781728982925415,
-      "logits/rejected": -1.97735595703125,
-      "logps/chosen": -34.61248779296875,
-      "logps/rejected": -36.63856887817383,
-      "loss": 0.3156,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.3223697543144226,
-      "rewards/margins": 0.4367140233516693,
-      "rewards/rejected": -0.11434426158666611,
-      "step": 1100
-    },
-    {
-      "epoch": 2.86,
-      "eval_logits/chosen": -2.237501621246338,
-      "eval_logits/rejected": -2.232618570327759,
-      "eval_logps/chosen": -33.959171295166016,
-      "eval_logps/rejected": -37.489627838134766,
-      "eval_loss": 1.6012389659881592,
-      "eval_rewards/accuracies": 0.5606312155723572,
-      "eval_rewards/chosen": 0.030150815844535828,
-      "eval_rewards/margins": 0.01935432478785515,
-      "eval_rewards/rejected": 0.01079649105668068,
-      "eval_runtime": 145.2896,
-      "eval_samples_per_second": 2.361,
-      "eval_steps_per_second": 0.296,
-      "step": 1100
-    },
-    {
-      "epoch": 2.88,
-      "grad_norm": 66.0,
-      "learning_rate": 1.0963854773524548e-06,
-      "logits/chosen": -2.0257740020751953,
-      "logits/rejected": -2.0311310291290283,
-      "logps/chosen": -34.69486999511719,
-      "logps/rejected": -35.754554748535156,
-      "loss": 0.4157,
-      "rewards/accuracies": 0.9125000238418579,
-      "rewards/chosen": 0.30889827013015747,
-      "rewards/margins": 0.39328455924987793,
-      "rewards/rejected": -0.08438632637262344,
-      "step": 1110
-    },
-    {
-      "epoch": 2.91,
-      "grad_norm": 90.0,
-      "learning_rate": 1.049857726072005e-06,
-      "logits/chosen": -1.9978212118148804,
-      "logits/rejected": -1.995883584022522,
-      "logps/chosen": -31.900583267211914,
-      "logps/rejected": -33.278594970703125,
-      "loss": 0.4591,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.2862483263015747,
-      "rewards/margins": 0.4045858383178711,
-      "rewards/rejected": -0.11833752691745758,
-      "step": 1120
-    },
-    {
-      "epoch": 2.94,
-      "grad_norm": 42.25,
-      "learning_rate": 1.0040749902836508e-06,
-      "logits/chosen": -1.9011417627334595,
-      "logits/rejected": -1.9039758443832397,
-      "logps/chosen": -28.964651107788086,
-      "logps/rejected": -30.513931274414062,
-      "loss": 0.2561,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.31359297037124634,
-      "rewards/margins": 0.42114973068237305,
-      "rewards/rejected": -0.1075567752122879,
-      "step": 1130
-    },
-    {
-      "epoch": 2.96,
-      "grad_norm": 19.875,
-      "learning_rate": 9.59060791022566e-07,
-      "logits/chosen": -1.9979721307754517,
-      "logits/rejected": -1.994855284690857,
-      "logps/chosen": -32.40828323364258,
-      "logps/rejected": -32.802223205566406,
-      "loss": 0.3667,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.3181864619255066,
-      "rewards/margins": 0.4283228814601898,
-      "rewards/rejected": -0.11013643443584442,
-      "step": 1140
-    },
-    {
-      "epoch": 2.99,
-      "grad_norm": 53.25,
-      "learning_rate": 9.148382544856885e-07,
-      "logits/chosen": -1.9201629161834717,
-      "logits/rejected": -1.9138599634170532,
-      "logps/chosen": -27.41982078552246,
-      "logps/rejected": -32.52082061767578,
-      "loss": 0.282,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.28726693987846375,
-      "rewards/margins": 0.43627747893333435,
-      "rewards/rejected": -0.14901044964790344,
-      "step": 1150
-    },
-    {
-      "epoch": 3.01,
-      "grad_norm": 52.75,
-      "learning_rate": 8.714301001505568e-07,
-      "logits/chosen": -2.038132905960083,
-      "logits/rejected": -2.0356593132019043,
-      "logps/chosen": -32.35755920410156,
-      "logps/rejected": -35.35866165161133,
-      "loss": 0.2873,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.29540252685546875,
-      "rewards/margins": 0.40225353837013245,
-      "rewards/rejected": -0.10685104131698608,
-      "step": 1160
-    },
-    {
-      "epoch": 3.04,
-      "grad_norm": 18.625,
-      "learning_rate": 8.288586291031025e-07,
-      "logits/chosen": -2.0191562175750732,
-      "logits/rejected": -2.0172648429870605,
-      "logps/chosen": -30.838241577148438,
-      "logps/rejected": -32.716129302978516,
-      "loss": 0.1723,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.3104664385318756,
-      "rewards/margins": 0.4513087272644043,
-      "rewards/rejected": -0.1408422887325287,
-      "step": 1170
-    },
-    {
-      "epoch": 3.06,
-      "grad_norm": 22.0,
-      "learning_rate": 7.871457125803897e-07,
-      "logits/chosen": -2.011491537094116,
-      "logits/rejected": -1.999138593673706,
-      "logps/chosen": -35.201385498046875,
-      "logps/rejected": -34.34071731567383,
-      "loss": 0.2661,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.33756545186042786,
-      "rewards/margins": 0.44526106119155884,
-      "rewards/rejected": -0.10769567638635635,
-      "step": 1180
-    },
-    {
-      "epoch": 3.09,
-      "grad_norm": 36.75,
-      "learning_rate": 7.463127807341966e-07,
-      "logits/chosen": -1.9028282165527344,
-      "logits/rejected": -1.8974478244781494,
-      "logps/chosen": -34.48936080932617,
-      "logps/rejected": -34.49340057373047,
-      "loss": 0.2986,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.34413009881973267,
-      "rewards/margins": 0.44636183977127075,
-      "rewards/rejected": -0.1022317185997963,
-      "step": 1190
-    },
-    {
-      "epoch": 3.12,
-      "grad_norm": 62.25,
-      "learning_rate": 7.063808116212021e-07,
-      "logits/chosen": -1.9445934295654297,
-      "logits/rejected": -1.9476436376571655,
-      "logps/chosen": -31.85054588317871,
-      "logps/rejected": -31.752578735351562,
-      "loss": 0.2893,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.30582019686698914,
-      "rewards/margins": 0.41237980127334595,
-      "rewards/rejected": -0.10655955970287323,
-      "step": 1200
-    },
-    {
-      "epoch": 3.12,
-      "eval_logits/chosen": -2.238332748413086,
-      "eval_logits/rejected": -2.233450174331665,
-      "eval_logps/chosen": -33.94802474975586,
-      "eval_logps/rejected": -37.49089431762695,
-      "eval_loss": 1.5704575777053833,
-      "eval_rewards/accuracies": 0.5365448594093323,
-      "eval_rewards/chosen": 0.03461090475320816,
-      "eval_rewards/margins": 0.024320118129253387,
-      "eval_rewards/rejected": 0.010290787555277348,
-      "eval_runtime": 145.3137,
-      "eval_samples_per_second": 2.36,
-      "eval_steps_per_second": 0.296,
-      "step": 1200
-    },
-    {
-      "epoch": 3.14,
-      "grad_norm": 39.0,
-      "learning_rate": 6.673703204254348e-07,
-      "logits/chosen": -1.976152777671814,
-      "logits/rejected": -1.971604347229004,
-      "logps/chosen": -30.516315460205078,
-      "logps/rejected": -31.238773345947266,
-      "loss": 0.2438,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.3029424548149109,
-      "rewards/margins": 0.4352136552333832,
-      "rewards/rejected": -0.1322712004184723,
-      "step": 1210
-    },
-    {
-      "epoch": 3.17,
-      "grad_norm": 28.875,
-      "learning_rate": 6.293013489185315e-07,
-      "logits/chosen": -1.988176703453064,
-      "logits/rejected": -1.982642412185669,
-      "logps/chosen": -34.4455680847168,
-      "logps/rejected": -34.812355041503906,
-      "loss": 0.2853,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.33134883642196655,
-      "rewards/margins": 0.4283396303653717,
-      "rewards/rejected": -0.09699074923992157,
-      "step": 1220
-    },
-    {
-      "epoch": 3.19,
-      "grad_norm": 40.0,
-      "learning_rate": 5.921934551632086e-07,
-      "logits/chosen": -1.9901288747787476,
-      "logits/rejected": -1.9769742488861084,
-      "logps/chosen": -32.32646560668945,
-      "logps/rejected": -34.40354537963867,
-      "loss": 0.3093,
-      "rewards/accuracies": 0.9125000238418579,
-      "rewards/chosen": 0.30702072381973267,
-      "rewards/margins": 0.4204995036125183,
-      "rewards/rejected": -0.11347881704568863,
-      "step": 1230
-    },
-    {
-      "epoch": 3.22,
-      "grad_norm": 44.0,
-      "learning_rate": 5.560657034652405e-07,
-      "logits/chosen": -2.034132957458496,
-      "logits/rejected": -2.0313124656677246,
-      "logps/chosen": -34.251712799072266,
-      "logps/rejected": -32.091835021972656,
-      "loss": 0.2462,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.3124735653400421,
-      "rewards/margins": 0.41325968503952026,
-      "rewards/rejected": -0.10078616440296173,
-      "step": 1240
-    },
-    {
-      "epoch": 3.25,
-      "grad_norm": 113.0,
-      "learning_rate": 5.2093665457911e-07,
-      "logits/chosen": -1.9658873081207275,
-      "logits/rejected": -1.9630600214004517,
-      "logps/chosen": -33.601524353027344,
-      "logps/rejected": -35.11139678955078,
-      "loss": 0.2643,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.3178528845310211,
-      "rewards/margins": 0.44355902075767517,
-      "rewards/rejected": -0.12570618093013763,
-      "step": 1250
-    },
-    {
-      "epoch": 3.27,
-      "grad_norm": 19.75,
-      "learning_rate": 4.868243561723535e-07,
-      "logits/chosen": -2.0280373096466064,
-      "logits/rejected": -2.022881031036377,
-      "logps/chosen": -29.97477149963379,
-      "logps/rejected": -32.641361236572266,
-      "loss": 0.2187,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.31440991163253784,
-      "rewards/margins": 0.41793423891067505,
-      "rewards/rejected": -0.103524349629879,
-      "step": 1260
-    },
-    {
-      "epoch": 3.3,
-      "grad_norm": 25.25,
-      "learning_rate": 4.537463335535161e-07,
-      "logits/chosen": -2.0543882846832275,
-      "logits/rejected": -2.0590672492980957,
-      "logps/chosen": -31.920276641845703,
-      "logps/rejected": -31.91879653930664,
-      "loss": 0.204,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.31628847122192383,
-      "rewards/margins": 0.41126084327697754,
-      "rewards/rejected": -0.09497231990098953,
-      "step": 1270
-    },
-    {
-      "epoch": 3.32,
-      "grad_norm": 28.25,
-      "learning_rate": 4.217195806684629e-07,
-      "logits/chosen": -1.9492976665496826,
-      "logits/rejected": -1.9564584493637085,
-      "logps/chosen": -34.14177703857422,
-      "logps/rejected": -32.934364318847656,
-      "loss": 0.2121,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.3061862885951996,
-      "rewards/margins": 0.4406258463859558,
-      "rewards/rejected": -0.13443957269191742,
-      "step": 1280
-    },
-    {
-      "epoch": 3.35,
-      "grad_norm": 30.75,
-      "learning_rate": 3.907605513696808e-07,
-      "logits/chosen": -1.8527300357818604,
-      "logits/rejected": -1.8549416065216064,
-      "logps/chosen": -32.315616607666016,
-      "logps/rejected": -36.80881118774414,
-      "loss": 0.2597,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.3265129327774048,
-      "rewards/margins": 0.4283279776573181,
-      "rewards/rejected": -0.10181504487991333,
-      "step": 1290
-    },
-    {
-      "epoch": 3.38,
-      "grad_norm": 157.0,
-      "learning_rate": 3.6088515096305675e-07,
-      "logits/chosen": -1.908172369003296,
-      "logits/rejected": -1.9115606546401978,
-      "logps/chosen": -32.43517303466797,
-      "logps/rejected": -32.76791763305664,
-      "loss": 0.277,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.3125966489315033,
-      "rewards/margins": 0.4340974688529968,
-      "rewards/rejected": -0.12150083482265472,
-      "step": 1300
-    },
-    {
-      "epoch": 3.38,
-      "eval_logits/chosen": -2.238203287124634,
-      "eval_logits/rejected": -2.233325719833374,
-      "eval_logps/chosen": -33.95594787597656,
-      "eval_logps/rejected": -37.486488342285156,
-      "eval_loss": 1.6102017164230347,
-      "eval_rewards/accuracies": 0.5402824282646179,
-      "eval_rewards/chosen": 0.03144202008843422,
-      "eval_rewards/margins": 0.019388414919376373,
-      "eval_rewards/rejected": 0.012053603306412697,
-      "eval_runtime": 145.3717,
-      "eval_samples_per_second": 2.359,
-      "eval_steps_per_second": 0.296,
-      "step": 1300
-    },
-    {
-      "epoch": 3.4,
-      "grad_norm": 53.0,
-      "learning_rate": 3.321087280364757e-07,
-      "logits/chosen": -2.0169308185577393,
-      "logits/rejected": -1.9967368841171265,
-      "logps/chosen": -30.350967407226562,
-      "logps/rejected": -35.431488037109375,
-      "loss": 0.2812,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.30497312545776367,
-      "rewards/margins": 0.41100645065307617,
-      "rewards/rejected": -0.1060333400964737,
-      "step": 1310
-    },
-    {
-      "epoch": 3.43,
-      "grad_norm": 93.0,
-      "learning_rate": 3.044460665744284e-07,
-      "logits/chosen": -1.9958242177963257,
-      "logits/rejected": -2.001598834991455,
-      "logps/chosen": -32.20707702636719,
-      "logps/rejected": -32.74049758911133,
-      "loss": 0.2256,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.33068421483039856,
-      "rewards/margins": 0.43058300018310547,
-      "rewards/rejected": -0.09989883005619049,
-      "step": 1320
-    },
-    {
-      "epoch": 3.45,
-      "grad_norm": 54.75,
-      "learning_rate": 2.779113783626916e-07,
-      "logits/chosen": -1.964477300643921,
-      "logits/rejected": -1.9595527648925781,
-      "logps/chosen": -32.27384567260742,
-      "logps/rejected": -34.89734649658203,
-      "loss": 0.2232,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.2994646430015564,
-      "rewards/margins": 0.42948657274246216,
-      "rewards/rejected": -0.13002195954322815,
-      "step": 1330
-    },
-    {
-      "epoch": 3.48,
-      "grad_norm": 28.625,
-      "learning_rate": 2.5251829568697204e-07,
-      "logits/chosen": -1.8306457996368408,
-      "logits/rejected": -1.840346097946167,
-      "logps/chosen": -32.9451789855957,
-      "logps/rejected": -31.277496337890625,
-      "loss": 0.2011,
-      "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.34176546335220337,
-      "rewards/margins": 0.43911829590797424,
-      "rewards/rejected": -0.09735285490751266,
-      "step": 1340
-    },
-    {
-      "epoch": 3.51,
-      "grad_norm": 21.375,
-      "learning_rate": 2.2827986432927774e-07,
-      "logits/chosen": -1.8909282684326172,
-      "logits/rejected": -1.8956327438354492,
-      "logps/chosen": -33.05854797363281,
-      "logps/rejected": -33.51481246948242,
-      "loss": 0.152,
-      "rewards/accuracies": 0.987500011920929,
-      "rewards/chosen": 0.35695692896842957,
-      "rewards/margins": 0.480957567691803,
-      "rewards/rejected": -0.12400064617395401,
-      "step": 1350
-    },
-    {
-      "epoch": 3.53,
-      "grad_norm": 135.0,
-      "learning_rate": 2.0520853686560177e-07,
-      "logits/chosen": -1.917721152305603,
-      "logits/rejected": -1.9233052730560303,
-      "logps/chosen": -33.081817626953125,
-      "logps/rejected": -34.85997772216797,
-      "loss": 0.3369,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.29993897676467896,
-      "rewards/margins": 0.41870397329330444,
-      "rewards/rejected": -0.11876501142978668,
-      "step": 1360
-    },
-    {
-      "epoch": 3.56,
-      "grad_norm": 40.0,
-      "learning_rate": 1.833161662683672e-07,
-      "logits/chosen": -2.0079197883605957,
-      "logits/rejected": -2.0034046173095703,
-      "logps/chosen": -31.874841690063477,
-      "logps/rejected": -31.1782169342041,
-      "loss": 0.2298,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.30433863401412964,
-      "rewards/margins": 0.4108657240867615,
-      "rewards/rejected": -0.10652711242437363,
-      "step": 1370
-    },
-    {
-      "epoch": 3.58,
-      "grad_norm": 39.5,
-      "learning_rate": 1.626139998169246e-07,
-      "logits/chosen": -1.930163025856018,
-      "logits/rejected": -1.9319047927856445,
-      "logps/chosen": -28.90609359741211,
-      "logps/rejected": -31.02655601501465,
-      "loss": 0.2334,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.3012833595275879,
-      "rewards/margins": 0.4142329692840576,
-      "rewards/rejected": -0.11294958740472794,
-      "step": 1380
-    },
-    {
-      "epoch": 3.61,
-      "grad_norm": 47.5,
-      "learning_rate": 1.4311267331922535e-07,
-      "logits/chosen": -1.9173237085342407,
-      "logits/rejected": -1.9218599796295166,
-      "logps/chosen": -31.18109703063965,
-      "logps/rejected": -33.24029541015625,
-      "loss": 0.1751,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.3353577256202698,
-      "rewards/margins": 0.43625006079673767,
-      "rewards/rejected": -0.10089237987995148,
-      "step": 1390
-    },
-    {
-      "epoch": 3.64,
-      "grad_norm": 39.5,
-      "learning_rate": 1.2482220564763669e-07,
-      "logits/chosen": -1.9880527257919312,
-      "logits/rejected": -1.9912986755371094,
-      "logps/chosen": -34.084754943847656,
-      "logps/rejected": -33.75811004638672,
-      "loss": 0.139,
-      "rewards/accuracies": 0.987500011920929,
-      "rewards/chosen": 0.33832791447639465,
-      "rewards/margins": 0.47214603424072266,
-      "rewards/rejected": -0.1338181048631668,
-      "step": 1400
-    },
-    {
-      "epoch": 3.64,
-      "eval_logits/chosen": -2.2381296157836914,
-      "eval_logits/rejected": -2.2332541942596436,
-      "eval_logps/chosen": -33.96630096435547,
-      "eval_logps/rejected": -37.49371337890625,
-      "eval_loss": 1.6181015968322754,
-      "eval_rewards/accuracies": 0.5307309031486511,
-      "eval_rewards/chosen": 0.027301045134663582,
-      "eval_rewards/margins": 0.01813744381070137,
-      "eval_rewards/rejected": 0.009163597598671913,
-      "eval_runtime": 145.1897,
-      "eval_samples_per_second": 2.362,
-      "eval_steps_per_second": 0.296,
-      "step": 1400
-    },
-    {
-      "epoch": 3.66,
-      "grad_norm": 16.375,
-      "learning_rate": 1.0775199359171346e-07,
-      "logits/chosen": -2.0566494464874268,
-      "logits/rejected": -2.049795150756836,
-      "logps/chosen": -32.942054748535156,
-      "logps/rejected": -34.50425338745117,
-      "loss": 0.2,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.3441971242427826,
-      "rewards/margins": 0.4480733871459961,
-      "rewards/rejected": -0.1038762778043747,
-      "step": 1410
-    },
-    {
-      "epoch": 3.69,
-      "grad_norm": 38.5,
-      "learning_rate": 9.191080703056604e-08,
-      "logits/chosen": -1.942500352859497,
-      "logits/rejected": -1.9532867670059204,
-      "logps/chosen": -33.3380126953125,
-      "logps/rejected": -33.94847869873047,
-      "loss": 0.1618,
-      "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.3245156705379486,
-      "rewards/margins": 0.4674775004386902,
-      "rewards/rejected": -0.1429617702960968,
-      "step": 1420
-    },
-    {
-      "epoch": 3.71,
-      "grad_norm": 35.0,
-      "learning_rate": 7.730678442730539e-08,
-      "logits/chosen": -2.0275931358337402,
-      "logits/rejected": -2.039466381072998,
-      "logps/chosen": -34.284996032714844,
-      "logps/rejected": -34.00452423095703,
-      "loss": 0.262,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.3103240728378296,
-      "rewards/margins": 0.45679956674575806,
-      "rewards/rejected": -0.14647547900676727,
-      "step": 1430
-    },
-    {
-      "epoch": 3.74,
-      "grad_norm": 47.25,
-      "learning_rate": 6.394742864787806e-08,
-      "logits/chosen": -2.0178730487823486,
-      "logits/rejected": -2.0198729038238525,
-      "logps/chosen": -32.42511749267578,
-      "logps/rejected": -34.111873626708984,
-      "loss": 0.2048,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.31330591440200806,
-      "rewards/margins": 0.4140680432319641,
-      "rewards/rejected": -0.10076216608285904,
-      "step": 1440
-    },
-    {
-      "epoch": 3.77,
-      "grad_norm": 31.75,
-      "learning_rate": 5.183960310644748e-08,
-      "logits/chosen": -1.9993807077407837,
-      "logits/rejected": -1.992196798324585,
-      "logps/chosen": -34.766136169433594,
-      "logps/rejected": -34.06538772583008,
-      "loss": 0.2089,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.32087528705596924,
-      "rewards/margins": 0.43886953592300415,
-      "rewards/rejected": -0.1179942637681961,
-      "step": 1450
-    },
-    {
-      "epoch": 3.79,
-      "grad_norm": 47.5,
-      "learning_rate": 4.098952823928693e-08,
-      "logits/chosen": -1.9658348560333252,
-      "logits/rejected": -1.9719345569610596,
-      "logps/chosen": -30.47488021850586,
-      "logps/rejected": -34.05385971069336,
-      "loss": 0.2022,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.29531940817832947,
-      "rewards/margins": 0.44886380434036255,
-      "rewards/rejected": -0.15354439616203308,
-      "step": 1460
-    },
-    {
-      "epoch": 3.82,
-      "grad_norm": 24.625,
-      "learning_rate": 3.1402778309014284e-08,
-      "logits/chosen": -1.9123966693878174,
-      "logits/rejected": -1.910093903541565,
-      "logps/chosen": -30.029949188232422,
-      "logps/rejected": -30.82769775390625,
-      "loss": 0.211,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.3262413442134857,
-      "rewards/margins": 0.41962748765945435,
-      "rewards/rejected": -0.09338615089654922,
-      "step": 1470
-    },
-    {
-      "epoch": 3.84,
-      "grad_norm": 25.5,
-      "learning_rate": 2.3084278540791427e-08,
-      "logits/chosen": -2.0873568058013916,
-      "logits/rejected": -2.0816843509674072,
-      "logps/chosen": -34.48334884643555,
-      "logps/rejected": -32.353607177734375,
-      "loss": 0.1471,
-      "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.32076385617256165,
-      "rewards/margins": 0.43952280282974243,
-      "rewards/rejected": -0.1187589168548584,
-      "step": 1480
-    },
-    {
-      "epoch": 3.87,
-      "grad_norm": 20.125,
-      "learning_rate": 1.6038302591975807e-08,
-      "logits/chosen": -1.9781821966171265,
-      "logits/rejected": -1.9803062677383423,
-      "logps/chosen": -27.70733642578125,
-      "logps/rejected": -28.307626724243164,
-      "loss": 0.2222,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.3076176047325134,
-      "rewards/margins": 0.4217551648616791,
-      "rewards/rejected": -0.11413755267858505,
-      "step": 1490
-    },
-    {
-      "epoch": 3.9,
-      "grad_norm": 43.0,
-      "learning_rate": 1.0268470356514237e-08,
-      "logits/chosen": -1.987160086631775,
-      "logits/rejected": -1.98151433467865,
-      "logps/chosen": -32.05828857421875,
-      "logps/rejected": -32.30227279663086,
-      "loss": 0.24,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.2970001995563507,
-      "rewards/margins": 0.43032747507095337,
-      "rewards/rejected": -0.13332727551460266,
-      "step": 1500
-    },
-    {
-      "epoch": 3.9,
-      "eval_logits/chosen": -2.2380595207214355,
-      "eval_logits/rejected": -2.2331745624542236,
-      "eval_logps/chosen": -33.95270919799805,
-      "eval_logps/rejected": -37.47934341430664,
-      "eval_loss": 1.6201497316360474,
-      "eval_rewards/accuracies": 0.52491694688797,
-      "eval_rewards/chosen": 0.032736361026763916,
-      "eval_rewards/margins": 0.017824459820985794,
-      "eval_rewards/rejected": 0.014911895617842674,
-      "eval_runtime": 145.1995,
-      "eval_samples_per_second": 2.362,
-      "eval_steps_per_second": 0.296,
-      "step": 1500
-    },
-    {
-      "epoch": 3.92,
-      "grad_norm": 26.25,
-      "learning_rate": 5.777746105209147e-09,
-      "logits/chosen": -1.9137132167816162,
-      "logits/rejected": -1.918095350265503,
-      "logps/chosen": -33.41365051269531,
-      "logps/rejected": -34.701316833496094,
-      "loss": 0.2691,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.3029375374317169,
-      "rewards/margins": 0.41931143403053284,
-      "rewards/rejected": -0.11637385189533234,
-      "step": 1510
-    },
-    {
-      "epoch": 3.95,
-      "grad_norm": 39.75,
-      "learning_rate": 2.5684369628148352e-09,
-      "logits/chosen": -1.973489761352539,
-      "logits/rejected": -1.9717988967895508,
-      "logps/chosen": -29.93434715270996,
-      "logps/rejected": -32.88115692138672,
-      "loss": 0.2589,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.31993287801742554,
-      "rewards/margins": 0.4365101456642151,
-      "rewards/rejected": -0.11657728254795074,
-      "step": 1520
-    },
-    {
-      "epoch": 3.97,
-      "grad_norm": 44.75,
-      "learning_rate": 6.421917227455999e-10,
-      "logits/chosen": -2.070810556411743,
-      "logits/rejected": -2.063199043273926,
-      "logps/chosen": -27.3148250579834,
-      "logps/rejected": -28.99945068359375,
-      "loss": 0.1829,
-      "rewards/accuracies": 0.987500011920929,
-      "rewards/chosen": 0.2992464005947113,
-      "rewards/margins": 0.4303843379020691,
-      "rewards/rejected": -0.1311378926038742,
-      "step": 1530
-    },
-    {
-      "epoch": 4.0,
-      "grad_norm": 28.0,
-      "learning_rate": 0.0,
-      "logits/chosen": -1.9703433513641357,
-      "logits/rejected": -1.9601656198501587,
-      "logps/chosen": -32.22865295410156,
-      "logps/rejected": -35.73883056640625,
-      "loss": 0.1937,
-      "rewards/accuracies": 0.987500011920929,
-      "rewards/chosen": 0.33501744270324707,
-      "rewards/margins": 0.4660801291465759,
-      "rewards/rejected": -0.13106267154216766,
-      "step": 1540
-    },
-    {
-      "epoch": 4.0,
-      "step": 1540,
+      "epoch": 1.0,
+      "step": 385,
       "total_flos": 0.0,
-      "train_loss": 0.36088944433571457,
-      "train_runtime": 10769.3493,
-      "train_samples_per_second": 1.144,
-      "train_steps_per_second": 0.143
+      "train_loss": 0.8438688600218142,
+      "train_runtime": 3250.9917,
+      "train_samples_per_second": 0.947,
+      "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,