diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,4795 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 100,
+  "global_step": 2865,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0010471204188481676,
+      "grad_norm": 1.5253963524854417,
+      "learning_rate": 1.7421602787456448e-07,
+      "logits/chosen": -11.625,
+      "logits/rejected": -12.0,
+      "logps/chosen": -480.0,
+      "logps/rejected": -564.0,
+      "loss": 0.6914,
+      "rewards/accuracies": 0.0,
+      "rewards/chosen": 0.0,
+      "rewards/margins": 0.0,
+      "rewards/rejected": 0.0,
+      "step": 1
+    },
+    {
+      "epoch": 0.010471204188481676,
+      "grad_norm": 1.3922238355420165,
+      "learning_rate": 1.7421602787456445e-06,
+      "logits/chosen": -11.375,
+      "logits/rejected": -11.125,
+      "logps/chosen": -442.0,
+      "logps/rejected": -516.0,
+      "loss": 0.692,
+      "rewards/accuracies": 0.1388888955116272,
+      "rewards/chosen": 0.000850677490234375,
+      "rewards/margins": 0.001220703125,
+      "rewards/rejected": -0.0003719329833984375,
+      "step": 10
+    },
+    {
+      "epoch": 0.020942408376963352,
+      "grad_norm": 1.4106577412404244,
+      "learning_rate": 3.484320557491289e-06,
+      "logits/chosen": -11.125,
+      "logits/rejected": -11.125,
+      "logps/chosen": -406.0,
+      "logps/rejected": -418.0,
+      "loss": 0.6915,
+      "rewards/accuracies": 0.21875,
+      "rewards/chosen": -0.0002231597900390625,
+      "rewards/margins": 0.000438690185546875,
+      "rewards/rejected": -0.000659942626953125,
+      "step": 20
+    },
+    {
+      "epoch": 0.031413612565445025,
+      "grad_norm": 1.4712574165051036,
+      "learning_rate": 5.226480836236934e-06,
+      "logits/chosen": -11.3125,
+      "logits/rejected": -11.375,
+      "logps/chosen": -422.0,
+      "logps/rejected": -458.0,
+      "loss": 0.6906,
+      "rewards/accuracies": 0.3812499940395355,
+      "rewards/chosen": -0.00390625,
+      "rewards/margins": 0.003448486328125,
+      "rewards/rejected": -0.007354736328125,
+      "step": 30
+    },
+    {
+      "epoch": 0.041884816753926704,
+      "grad_norm": 1.6451473868588853,
+      "learning_rate": 6.968641114982578e-06,
+      "logits/chosen": -11.875,
+      "logits/rejected": -11.6875,
+      "logps/chosen": -436.0,
+      "logps/rejected": -456.0,
+      "loss": 0.6877,
+      "rewards/accuracies": 0.48750001192092896,
+      "rewards/chosen": -0.057861328125,
+      "rewards/margins": 0.01153564453125,
+      "rewards/rejected": -0.0693359375,
+      "step": 40
+    },
+    {
+      "epoch": 0.05235602094240838,
+      "grad_norm": 1.8752448553673084,
+      "learning_rate": 8.710801393728225e-06,
+      "logits/chosen": -12.5625,
+      "logits/rejected": -12.5625,
+      "logps/chosen": -448.0,
+      "logps/rejected": -512.0,
+      "loss": 0.6804,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": -0.1875,
+      "rewards/margins": 0.032470703125,
+      "rewards/rejected": -0.2197265625,
+      "step": 50
+    },
+    {
+      "epoch": 0.06282722513089005,
+      "grad_norm": 1.8923035427984414,
+      "learning_rate": 1.0452961672473868e-05,
+      "logits/chosen": -12.0,
+      "logits/rejected": -11.875,
+      "logps/chosen": -368.0,
+      "logps/rejected": -432.0,
+      "loss": 0.6669,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": -0.185546875,
+      "rewards/margins": 0.076171875,
+      "rewards/rejected": -0.26171875,
+      "step": 60
+    },
+    {
+      "epoch": 0.07329842931937172,
+      "grad_norm": 1.845377895263991,
+      "learning_rate": 1.2195121951219513e-05,
+      "logits/chosen": -11.75,
+      "logits/rejected": -11.75,
+      "logps/chosen": -474.0,
+      "logps/rejected": -544.0,
+      "loss": 0.663,
+      "rewards/accuracies": 0.581250011920929,
+      "rewards/chosen": -0.310546875,
+      "rewards/margins": 0.10693359375,
+      "rewards/rejected": -0.416015625,
+      "step": 70
+    },
+    {
+      "epoch": 0.08376963350785341,
+      "grad_norm": 1.8953695346226918,
+      "learning_rate": 1.3937282229965156e-05,
+      "logits/chosen": -11.3125,
+      "logits/rejected": -11.25,
+      "logps/chosen": -452.0,
+      "logps/rejected": -556.0,
+      "loss": 0.6377,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -0.53515625,
+      "rewards/margins": 0.216796875,
+      "rewards/rejected": -0.75390625,
+      "step": 80
+    },
+    {
+      "epoch": 0.09424083769633508,
+      "grad_norm": 2.2139050071469604,
+      "learning_rate": 1.56794425087108e-05,
+      "logits/chosen": -10.4375,
+      "logits/rejected": -10.4375,
+      "logps/chosen": -454.0,
+      "logps/rejected": -492.0,
+      "loss": 0.6267,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -0.83984375,
+      "rewards/margins": 0.1650390625,
+      "rewards/rejected": -1.0078125,
+      "step": 90
+    },
+    {
+      "epoch": 0.10471204188481675,
+      "grad_norm": 2.2261764080692545,
+      "learning_rate": 1.742160278745645e-05,
+      "logits/chosen": -11.5625,
+      "logits/rejected": -11.6875,
+      "logps/chosen": -596.0,
+      "logps/rejected": -628.0,
+      "loss": 0.6518,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": -1.21875,
+      "rewards/margins": 0.138671875,
+      "rewards/rejected": -1.359375,
+      "step": 100
+    },
+    {
+      "epoch": 0.10471204188481675,
+      "eval_logits/chosen": -11.8125,
+      "eval_logits/rejected": -11.5,
+      "eval_logps/chosen": -398.0,
+      "eval_logps/rejected": -380.0,
+      "eval_loss": 0.6870937347412109,
+      "eval_rewards/accuracies": 0.595703125,
+      "eval_rewards/chosen": -0.796875,
+      "eval_rewards/margins": 0.11474609375,
+      "eval_rewards/rejected": -0.9140625,
+      "eval_runtime": 48.6163,
+      "eval_samples_per_second": 41.138,
+      "eval_steps_per_second": 0.658,
+      "step": 100
+    },
+    {
+      "epoch": 0.11518324607329843,
+      "grad_norm": 1.901609206013187,
+      "learning_rate": 1.9163763066202093e-05,
+      "logits/chosen": -11.6875,
+      "logits/rejected": -11.625,
+      "logps/chosen": -524.0,
+      "logps/rejected": -588.0,
+      "loss": 0.6452,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -0.96875,
+      "rewards/margins": 0.2197265625,
+      "rewards/rejected": -1.1875,
+      "step": 110
+    },
+    {
+      "epoch": 0.1256544502617801,
+      "grad_norm": 2.430125718642876,
+      "learning_rate": 2.0905923344947736e-05,
+      "logits/chosen": -12.375,
+      "logits/rejected": -12.3125,
+      "logps/chosen": -490.0,
+      "logps/rejected": -572.0,
+      "loss": 0.6289,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -1.125,
+      "rewards/margins": 0.2578125,
+      "rewards/rejected": -1.3828125,
+      "step": 120
+    },
+    {
+      "epoch": 0.13612565445026178,
+      "grad_norm": 2.2073773283758933,
+      "learning_rate": 2.264808362369338e-05,
+      "logits/chosen": -13.8125,
+      "logits/rejected": -13.75,
+      "logps/chosen": -508.0,
+      "logps/rejected": -540.0,
+      "loss": 0.6459,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": -0.9765625,
+      "rewards/margins": 0.181640625,
+      "rewards/rejected": -1.15625,
+      "step": 130
+    },
+    {
+      "epoch": 0.14659685863874344,
+      "grad_norm": 2.376907441544231,
+      "learning_rate": 2.4390243902439026e-05,
+      "logits/chosen": -14.125,
+      "logits/rejected": -13.9375,
+      "logps/chosen": -540.0,
+      "logps/rejected": -672.0,
+      "loss": 0.6177,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.84765625,
+      "rewards/margins": 0.318359375,
+      "rewards/rejected": -1.1640625,
+      "step": 140
+    },
+    {
+      "epoch": 0.15706806282722513,
+      "grad_norm": 2.304863388850328,
+      "learning_rate": 2.6132404181184672e-05,
+      "logits/chosen": -13.125,
+      "logits/rejected": -13.125,
+      "logps/chosen": -572.0,
+      "logps/rejected": -588.0,
+      "loss": 0.6431,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -1.3671875,
+      "rewards/margins": 0.166015625,
+      "rewards/rejected": -1.53125,
+      "step": 150
+    },
+    {
+      "epoch": 0.16753926701570682,
+      "grad_norm": 2.0750080864988663,
+      "learning_rate": 2.7874564459930312e-05,
+      "logits/chosen": -12.3125,
+      "logits/rejected": -12.125,
+      "logps/chosen": -516.0,
+      "logps/rejected": -564.0,
+      "loss": 0.6229,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -1.15625,
+      "rewards/margins": 0.193359375,
+      "rewards/rejected": -1.3515625,
+      "step": 160
+    },
+    {
+      "epoch": 0.17801047120418848,
+      "grad_norm": 2.0324550300364894,
+      "learning_rate": 2.9616724738675962e-05,
+      "logits/chosen": -10.8125,
+      "logits/rejected": -10.8125,
+      "logps/chosen": -552.0,
+      "logps/rejected": -656.0,
+      "loss": 0.5885,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -1.21875,
+      "rewards/margins": 0.369140625,
+      "rewards/rejected": -1.5859375,
+      "step": 170
+    },
+    {
+      "epoch": 0.18848167539267016,
+      "grad_norm": 2.148846401144063,
+      "learning_rate": 3.13588850174216e-05,
+      "logits/chosen": -10.75,
+      "logits/rejected": -10.6875,
+      "logps/chosen": -612.0,
+      "logps/rejected": -684.0,
+      "loss": 0.6081,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -1.625,
+      "rewards/margins": 0.458984375,
+      "rewards/rejected": -2.078125,
+      "step": 180
+    },
+    {
+      "epoch": 0.19895287958115182,
+      "grad_norm": 2.0967921324218293,
+      "learning_rate": 3.310104529616725e-05,
+      "logits/chosen": -11.5,
+      "logits/rejected": -11.5,
+      "logps/chosen": -544.0,
+      "logps/rejected": -612.0,
+      "loss": 0.6339,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -1.390625,
+      "rewards/margins": 0.279296875,
+      "rewards/rejected": -1.6640625,
+      "step": 190
+    },
+    {
+      "epoch": 0.2094240837696335,
+      "grad_norm": 2.227471408580189,
+      "learning_rate": 3.48432055749129e-05,
+      "logits/chosen": -12.6875,
+      "logits/rejected": -12.6875,
+      "logps/chosen": -568.0,
+      "logps/rejected": -660.0,
+      "loss": 0.6218,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -1.09375,
+      "rewards/margins": 0.32421875,
+      "rewards/rejected": -1.4140625,
+      "step": 200
+    },
+    {
+      "epoch": 0.2094240837696335,
+      "eval_logits/chosen": -14.75,
+      "eval_logits/rejected": -14.375,
+      "eval_logps/chosen": -420.0,
+      "eval_logps/rejected": -406.0,
+      "eval_loss": 0.6912500262260437,
+      "eval_rewards/accuracies": 0.58203125,
+      "eval_rewards/chosen": -1.015625,
+      "eval_rewards/margins": 0.1474609375,
+      "eval_rewards/rejected": -1.1640625,
+      "eval_runtime": 46.6174,
+      "eval_samples_per_second": 42.902,
+      "eval_steps_per_second": 0.686,
+      "step": 200
+    },
+    {
+      "epoch": 0.2198952879581152,
+      "grad_norm": 2.301299378701334,
+      "learning_rate": 3.6585365853658535e-05,
+      "logits/chosen": -14.0,
+      "logits/rejected": -13.8125,
+      "logps/chosen": -486.0,
+      "logps/rejected": -608.0,
+      "loss": 0.6267,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": -1.1640625,
+      "rewards/margins": 0.376953125,
+      "rewards/rejected": -1.5390625,
+      "step": 210
+    },
+    {
+      "epoch": 0.23036649214659685,
+      "grad_norm": 2.1665016843996012,
+      "learning_rate": 3.8327526132404185e-05,
+      "logits/chosen": -11.0,
+      "logits/rejected": -10.6875,
+      "logps/chosen": -560.0,
+      "logps/rejected": -616.0,
+      "loss": 0.6275,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": -1.265625,
+      "rewards/margins": 0.298828125,
+      "rewards/rejected": -1.5625,
+      "step": 220
+    },
+    {
+      "epoch": 0.24083769633507854,
+      "grad_norm": 1.9329287136652065,
+      "learning_rate": 4.006968641114983e-05,
+      "logits/chosen": -9.8125,
+      "logits/rejected": -9.875,
+      "logps/chosen": -536.0,
+      "logps/rejected": -600.0,
+      "loss": 0.6208,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -1.4921875,
+      "rewards/margins": 0.333984375,
+      "rewards/rejected": -1.828125,
+      "step": 230
+    },
+    {
+      "epoch": 0.2513089005235602,
+      "grad_norm": 2.26072058310108,
+      "learning_rate": 4.181184668989547e-05,
+      "logits/chosen": -9.25,
+      "logits/rejected": -9.125,
+      "logps/chosen": -544.0,
+      "logps/rejected": -640.0,
+      "loss": 0.6081,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -1.609375,
+      "rewards/margins": 0.375,
+      "rewards/rejected": -1.984375,
+      "step": 240
+    },
+    {
+      "epoch": 0.2617801047120419,
+      "grad_norm": 2.6127162131741275,
+      "learning_rate": 4.3554006968641115e-05,
+      "logits/chosen": -10.3125,
+      "logits/rejected": -10.0,
+      "logps/chosen": -664.0,
+      "logps/rejected": -724.0,
+      "loss": 0.6107,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -1.9921875,
+      "rewards/margins": 0.400390625,
+      "rewards/rejected": -2.390625,
+      "step": 250
+    },
+    {
+      "epoch": 0.27225130890052357,
+      "grad_norm": 3.319947401779959,
+      "learning_rate": 4.529616724738676e-05,
+      "logits/chosen": -9.0625,
+      "logits/rejected": -8.75,
+      "logps/chosen": -612.0,
+      "logps/rejected": -732.0,
+      "loss": 0.6145,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -2.265625,
+      "rewards/margins": 0.482421875,
+      "rewards/rejected": -2.75,
+      "step": 260
+    },
+    {
+      "epoch": 0.28272251308900526,
+      "grad_norm": 2.0537214398988892,
+      "learning_rate": 4.703832752613241e-05,
+      "logits/chosen": -11.375,
+      "logits/rejected": -11.4375,
+      "logps/chosen": -540.0,
+      "logps/rejected": -600.0,
+      "loss": 0.6153,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -1.5703125,
+      "rewards/margins": 0.296875,
+      "rewards/rejected": -1.8671875,
+      "step": 270
+    },
+    {
+      "epoch": 0.2931937172774869,
+      "grad_norm": 2.0669920889522904,
+      "learning_rate": 4.878048780487805e-05,
+      "logits/chosen": -10.9375,
+      "logits/rejected": -10.875,
+      "logps/chosen": -564.0,
+      "logps/rejected": -680.0,
+      "loss": 0.6282,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -1.953125,
+      "rewards/margins": 0.53515625,
+      "rewards/rejected": -2.484375,
+      "step": 280
+    },
+    {
+      "epoch": 0.3036649214659686,
+      "grad_norm": 2.5222188552660647,
+      "learning_rate": 4.999983293481418e-05,
+      "logits/chosen": -12.0625,
+      "logits/rejected": -12.0,
+      "logps/chosen": -596.0,
+      "logps/rejected": -660.0,
+      "loss": 0.6623,
+      "rewards/accuracies": 0.59375,
+      "rewards/chosen": -2.046875,
+      "rewards/margins": 0.34375,
+      "rewards/rejected": -2.390625,
+      "step": 290
+    },
+    {
+      "epoch": 0.31413612565445026,
+      "grad_norm": 2.3291185215255017,
+      "learning_rate": 4.9996862949181595e-05,
+      "logits/chosen": -10.9375,
+      "logits/rejected": -10.875,
+      "logps/chosen": -656.0,
+      "logps/rejected": -676.0,
+      "loss": 0.6888,
+      "rewards/accuracies": 0.59375,
+      "rewards/chosen": -2.4375,
+      "rewards/margins": 0.0986328125,
+      "rewards/rejected": -2.53125,
+      "step": 300
+    },
+    {
+      "epoch": 0.31413612565445026,
+      "eval_logits/chosen": -9.125,
+      "eval_logits/rejected": -8.4375,
+      "eval_logps/chosen": -556.0,
+      "eval_logps/rejected": -544.0,
+      "eval_loss": 0.7233906388282776,
+      "eval_rewards/accuracies": 0.578125,
+      "eval_rewards/chosen": -2.375,
+      "eval_rewards/margins": 0.1845703125,
+      "eval_rewards/rejected": -2.5625,
+      "eval_runtime": 46.6254,
+      "eval_samples_per_second": 42.895,
+      "eval_steps_per_second": 0.686,
+      "step": 300
+    },
+    {
+      "epoch": 0.32460732984293195,
+      "grad_norm": 2.4230207533476813,
+      "learning_rate": 4.9990180911526896e-05,
+      "logits/chosen": -10.8125,
+      "logits/rejected": -10.625,
+      "logps/chosen": -540.0,
+      "logps/rejected": -612.0,
+      "loss": 0.6297,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -1.6796875,
+      "rewards/margins": 0.330078125,
+      "rewards/rejected": -2.015625,
+      "step": 310
+    },
+    {
+      "epoch": 0.33507853403141363,
+      "grad_norm": 2.616511841060664,
+      "learning_rate": 4.997978781413745e-05,
+      "logits/chosen": -9.5625,
+      "logits/rejected": -9.5,
+      "logps/chosen": -684.0,
+      "logps/rejected": -724.0,
+      "loss": 0.6457,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": -2.40625,
+      "rewards/margins": 0.33984375,
+      "rewards/rejected": -2.75,
+      "step": 320
+    },
+    {
+      "epoch": 0.34554973821989526,
+      "grad_norm": 2.7004225196482374,
+      "learning_rate": 4.996568520039565e-05,
+      "logits/chosen": -7.875,
+      "logits/rejected": -7.96875,
+      "logps/chosen": -740.0,
+      "logps/rejected": -868.0,
+      "loss": 0.5962,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -3.203125,
+      "rewards/margins": 0.5390625,
+      "rewards/rejected": -3.75,
+      "step": 330
+    },
+    {
+      "epoch": 0.35602094240837695,
+      "grad_norm": 2.1577957803065257,
+      "learning_rate": 4.994787516454973e-05,
+      "logits/chosen": -9.125,
+      "logits/rejected": -9.125,
+      "logps/chosen": -652.0,
+      "logps/rejected": -736.0,
+      "loss": 0.6146,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -2.484375,
+      "rewards/margins": 0.41015625,
+      "rewards/rejected": -2.890625,
+      "step": 340
+    },
+    {
+      "epoch": 0.36649214659685864,
+      "grad_norm": 2.3799329956425286,
+      "learning_rate": 4.992636035140273e-05,
+      "logits/chosen": -13.5,
+      "logits/rejected": -13.4375,
+      "logps/chosen": -524.0,
+      "logps/rejected": -596.0,
+      "loss": 0.6368,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": -1.5703125,
+      "rewards/margins": 0.26953125,
+      "rewards/rejected": -1.8359375,
+      "step": 350
+    },
+    {
+      "epoch": 0.3769633507853403,
+      "grad_norm": 3.000077045968203,
+      "learning_rate": 4.99011439559198e-05,
+      "logits/chosen": -13.1875,
+      "logits/rejected": -13.0,
+      "logps/chosen": -640.0,
+      "logps/rejected": -724.0,
+      "loss": 0.6183,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -2.28125,
+      "rewards/margins": 0.375,
+      "rewards/rejected": -2.671875,
+      "step": 360
+    },
+    {
+      "epoch": 0.387434554973822,
+      "grad_norm": 2.417425247354359,
+      "learning_rate": 4.987222972275369e-05,
+      "logits/chosen": -10.0,
+      "logits/rejected": -9.8125,
+      "logps/chosen": -652.0,
+      "logps/rejected": -740.0,
+      "loss": 0.6537,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -2.40625,
+      "rewards/margins": 0.4375,
+      "rewards/rejected": -2.84375,
+      "step": 370
+    },
+    {
+      "epoch": 0.39790575916230364,
+      "grad_norm": 2.0624934167282456,
+      "learning_rate": 4.983962194568872e-05,
+      "logits/chosen": -11.9375,
+      "logits/rejected": -11.5625,
+      "logps/chosen": -616.0,
+      "logps/rejected": -716.0,
+      "loss": 0.6303,
+      "rewards/accuracies": 0.59375,
+      "rewards/chosen": -1.953125,
+      "rewards/margins": 0.3203125,
+      "rewards/rejected": -2.265625,
+      "step": 380
+    },
+    {
+      "epoch": 0.4083769633507853,
+      "grad_norm": 1.9989611135894672,
+      "learning_rate": 4.9803325467003084e-05,
+      "logits/chosen": -11.75,
+      "logits/rejected": -11.375,
+      "logps/chosen": -640.0,
+      "logps/rejected": -720.0,
+      "loss": 0.6066,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -2.09375,
+      "rewards/margins": 0.421875,
+      "rewards/rejected": -2.515625,
+      "step": 390
+    },
+    {
+      "epoch": 0.418848167539267,
+      "grad_norm": 1.979079545060936,
+      "learning_rate": 4.9763345676749816e-05,
+      "logits/chosen": -9.625,
+      "logits/rejected": -9.5625,
+      "logps/chosen": -632.0,
+      "logps/rejected": -704.0,
+      "loss": 0.6369,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": -2.34375,
+      "rewards/margins": 0.328125,
+      "rewards/rejected": -2.671875,
+      "step": 400
+    },
+    {
+      "epoch": 0.418848167539267,
+      "eval_logits/chosen": -8.4375,
+      "eval_logits/rejected": -7.5625,
+      "eval_logps/chosen": -552.0,
+      "eval_logps/rejected": -548.0,
+      "eval_loss": 0.7125937342643738,
+      "eval_rewards/accuracies": 0.580078125,
+      "eval_rewards/chosen": -2.34375,
+      "eval_rewards/margins": 0.26171875,
+      "eval_rewards/rejected": -2.609375,
+      "eval_runtime": 46.6073,
+      "eval_samples_per_second": 42.912,
+      "eval_steps_per_second": 0.687,
+      "step": 400
+    },
+    {
+      "epoch": 0.4293193717277487,
+      "grad_norm": 1.8619719942593533,
+      "learning_rate": 4.9719688511956374e-05,
+      "logits/chosen": -9.125,
+      "logits/rejected": -9.0625,
+      "logps/chosen": -652.0,
+      "logps/rejected": -772.0,
+      "loss": 0.635,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -2.25,
+      "rewards/margins": 0.408203125,
+      "rewards/rejected": -2.65625,
+      "step": 410
+    },
+    {
+      "epoch": 0.4397905759162304,
+      "grad_norm": 2.572204311727596,
+      "learning_rate": 4.967236045574294e-05,
+      "logits/chosen": -8.9375,
+      "logits/rejected": -8.75,
+      "logps/chosen": -716.0,
+      "logps/rejected": -784.0,
+      "loss": 0.6008,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -3.0,
+      "rewards/margins": 0.357421875,
+      "rewards/rejected": -3.359375,
+      "step": 420
+    },
+    {
+      "epoch": 0.450261780104712,
+      "grad_norm": 2.447309969073372,
+      "learning_rate": 4.96213685363597e-05,
+      "logits/chosen": -8.75,
+      "logits/rejected": -8.8125,
+      "logps/chosen": -696.0,
+      "logps/rejected": -816.0,
+      "loss": 0.6498,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -2.890625,
+      "rewards/margins": 0.578125,
+      "rewards/rejected": -3.453125,
+      "step": 430
+    },
+    {
+      "epoch": 0.4607329842931937,
+      "grad_norm": 3.671854968122256,
+      "learning_rate": 4.956672032614315e-05,
+      "logits/chosen": -8.9375,
+      "logits/rejected": -9.0,
+      "logps/chosen": -664.0,
+      "logps/rejected": -744.0,
+      "loss": 0.6291,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -2.421875,
+      "rewards/margins": 0.43359375,
+      "rewards/rejected": -2.859375,
+      "step": 440
+    },
+    {
+      "epoch": 0.4712041884816754,
+      "grad_norm": 2.58429218543323,
+      "learning_rate": 4.950842394039157e-05,
+      "logits/chosen": -7.46875,
+      "logits/rejected": -7.125,
+      "logps/chosen": -620.0,
+      "logps/rejected": -712.0,
+      "loss": 0.6107,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -2.375,
+      "rewards/margins": 0.2890625,
+      "rewards/rejected": -2.65625,
+      "step": 450
+    },
+    {
+      "epoch": 0.4816753926701571,
+      "grad_norm": 2.3742951297661397,
+      "learning_rate": 4.944648803615994e-05,
+      "logits/chosen": -6.59375,
+      "logits/rejected": -6.375,
+      "logps/chosen": -704.0,
+      "logps/rejected": -808.0,
+      "loss": 0.5859,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -2.890625,
+      "rewards/margins": 0.6328125,
+      "rewards/rejected": -3.53125,
+      "step": 460
+    },
+    {
+      "epoch": 0.49214659685863876,
+      "grad_norm": 15.569327385684304,
+      "learning_rate": 4.938092181097433e-05,
+      "logits/chosen": -8.1875,
+      "logits/rejected": -8.125,
+      "logps/chosen": -700.0,
+      "logps/rejected": -824.0,
+      "loss": 0.5789,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -2.984375,
+      "rewards/margins": 0.482421875,
+      "rewards/rejected": -3.46875,
+      "step": 470
+    },
+    {
+      "epoch": 0.5026178010471204,
+      "grad_norm": 3.581155911705221,
+      "learning_rate": 4.931173500146607e-05,
+      "logits/chosen": -7.53125,
+      "logits/rejected": -7.375,
+      "logps/chosen": -688.0,
+      "logps/rejected": -820.0,
+      "loss": 0.6148,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -3.09375,
+      "rewards/margins": 0.63671875,
+      "rewards/rejected": -3.734375,
+      "step": 480
+    },
+    {
+      "epoch": 0.5130890052356021,
+      "grad_norm": 1.9629288150285007,
+      "learning_rate": 4.923893788192582e-05,
+      "logits/chosen": -7.3125,
+      "logits/rejected": -7.0,
+      "logps/chosen": -752.0,
+      "logps/rejected": -872.0,
+      "loss": 0.6318,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -3.296875,
+      "rewards/margins": 0.5234375,
+      "rewards/rejected": -3.8125,
+      "step": 490
+    },
+    {
+      "epoch": 0.5235602094240838,
+      "grad_norm": 2.062533215962185,
+      "learning_rate": 4.9162541262777906e-05,
+      "logits/chosen": -9.0,
+      "logits/rejected": -9.0,
+      "logps/chosen": -696.0,
+      "logps/rejected": -832.0,
+      "loss": 0.6137,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -2.75,
+      "rewards/margins": 0.50390625,
+      "rewards/rejected": -3.25,
+      "step": 500
+    },
+    {
+      "epoch": 0.5235602094240838,
+      "eval_logits/chosen": -8.125,
+      "eval_logits/rejected": -7.375,
+      "eval_logps/chosen": -588.0,
+      "eval_logps/rejected": -580.0,
+      "eval_loss": 0.7105468511581421,
+      "eval_rewards/accuracies": 0.556640625,
+      "eval_rewards/chosen": -2.703125,
+      "eval_rewards/margins": 0.23046875,
+      "eval_rewards/rejected": -2.921875,
+      "eval_runtime": 46.624,
+      "eval_samples_per_second": 42.896,
+      "eval_steps_per_second": 0.686,
+      "step": 500
+    },
+    {
+      "epoch": 0.5340314136125655,
+      "grad_norm": 2.2976542972737777,
+      "learning_rate": 4.90825564889749e-05,
+      "logits/chosen": -9.5625,
+      "logits/rejected": -9.375,
+      "logps/chosen": -712.0,
+      "logps/rejected": -716.0,
+      "loss": 0.6534,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -2.625,
+      "rewards/margins": 0.2353515625,
+      "rewards/rejected": -2.859375,
+      "step": 510
+    },
+    {
+      "epoch": 0.5445026178010471,
+      "grad_norm": 2.356702981265094,
+      "learning_rate": 4.899899543831288e-05,
+      "logits/chosen": -10.3125,
+      "logits/rejected": -10.125,
+      "logps/chosen": -700.0,
+      "logps/rejected": -800.0,
+      "loss": 0.5769,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -2.515625,
+      "rewards/margins": 0.56640625,
+      "rewards/rejected": -3.078125,
+      "step": 520
+    },
+    {
+      "epoch": 0.5549738219895288,
+      "grad_norm": 2.7371704024933443,
+      "learning_rate": 4.891187051966765e-05,
+      "logits/chosen": -8.875,
+      "logits/rejected": -8.75,
+      "logps/chosen": -752.0,
+      "logps/rejected": -784.0,
+      "loss": 0.66,
+      "rewards/accuracies": 0.59375,
+      "rewards/chosen": -3.171875,
+      "rewards/margins": 0.2275390625,
+      "rewards/rejected": -3.390625,
+      "step": 530
+    },
+    {
+      "epoch": 0.5654450261780105,
+      "grad_norm": 2.2091281284383366,
+      "learning_rate": 4.882119467115191e-05,
+      "logits/chosen": -9.5625,
+      "logits/rejected": -9.4375,
+      "logps/chosen": -724.0,
+      "logps/rejected": -812.0,
+      "loss": 0.588,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": -3.140625,
+      "rewards/margins": 0.435546875,
+      "rewards/rejected": -3.578125,
+      "step": 540
+    },
+    {
+      "epoch": 0.5759162303664922,
+      "grad_norm": 2.9389275091389027,
+      "learning_rate": 4.872698135819402e-05,
+      "logits/chosen": -10.0625,
+      "logits/rejected": -9.75,
+      "logps/chosen": -708.0,
+      "logps/rejected": -776.0,
+      "loss": 0.5968,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -2.96875,
+      "rewards/margins": 0.443359375,
+      "rewards/rejected": -3.40625,
+      "step": 550
+    },
+    {
+      "epoch": 0.5863874345549738,
+      "grad_norm": 2.3808435860749366,
+      "learning_rate": 4.862924457153831e-05,
+      "logits/chosen": -8.375,
+      "logits/rejected": -8.1875,
+      "logps/chosen": -688.0,
+      "logps/rejected": -804.0,
+      "loss": 0.5923,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -3.390625,
+      "rewards/margins": 0.6328125,
+      "rewards/rejected": -4.03125,
+      "step": 560
+    },
+    {
+      "epoch": 0.5968586387434555,
+      "grad_norm": 2.4092607979284164,
+      "learning_rate": 4.8527998825167484e-05,
+      "logits/chosen": -9.25,
+      "logits/rejected": -9.25,
+      "logps/chosen": -712.0,
+      "logps/rejected": -828.0,
+      "loss": 0.6089,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -3.296875,
+      "rewards/margins": 0.59375,
+      "rewards/rejected": -3.890625,
+      "step": 570
+    },
+    {
+      "epoch": 0.6073298429319371,
+      "grad_norm": 1.986939887842181,
+      "learning_rate": 4.842325915414728e-05,
+      "logits/chosen": -10.5625,
+      "logits/rejected": -10.1875,
+      "logps/chosen": -648.0,
+      "logps/rejected": -756.0,
+      "loss": 0.623,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -2.71875,
+      "rewards/margins": 0.53125,
+      "rewards/rejected": -3.25,
+      "step": 580
+    },
+    {
+      "epoch": 0.6178010471204188,
+      "grad_norm": 2.4908357235700107,
+      "learning_rate": 4.831504111239373e-05,
+      "logits/chosen": -11.375,
+      "logits/rejected": -11.0,
+      "logps/chosen": -704.0,
+      "logps/rejected": -812.0,
+      "loss": 0.6223,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": -2.796875,
+      "rewards/margins": 0.328125,
+      "rewards/rejected": -3.140625,
+      "step": 590
+    },
+    {
+      "epoch": 0.6282722513089005,
+      "grad_norm": 2.745176728336525,
+      "learning_rate": 4.8203360770363425e-05,
+      "logits/chosen": -10.0,
+      "logits/rejected": -9.8125,
+      "logps/chosen": -728.0,
+      "logps/rejected": -768.0,
+      "loss": 0.6137,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -3.09375,
+      "rewards/margins": 0.310546875,
+      "rewards/rejected": -3.390625,
+      "step": 600
+    },
+    {
+      "epoch": 0.6282722513089005,
+      "eval_logits/chosen": -10.375,
+      "eval_logits/rejected": -9.3125,
+      "eval_logps/chosen": -584.0,
+      "eval_logps/rejected": -588.0,
+      "eval_loss": 0.7352656126022339,
+      "eval_rewards/accuracies": 0.576171875,
+      "eval_rewards/chosen": -2.65625,
+      "eval_rewards/margins": 0.33203125,
+      "eval_rewards/rejected": -2.984375,
+      "eval_runtime": 46.6188,
+      "eval_samples_per_second": 42.901,
+      "eval_steps_per_second": 0.686,
+      "step": 600
+    },
+    {
+      "epoch": 0.6387434554973822,
+      "grad_norm": 2.1115231948556974,
+      "learning_rate": 4.808823471266702e-05,
+      "logits/chosen": -11.0,
+      "logits/rejected": -10.9375,
+      "logps/chosen": -648.0,
+      "logps/rejected": -712.0,
+      "loss": 0.6112,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -2.390625,
+      "rewards/margins": 0.46875,
+      "rewards/rejected": -2.859375,
+      "step": 610
+    },
+    {
+      "epoch": 0.6492146596858639,
+      "grad_norm": 2.4827072595363453,
+      "learning_rate": 4.796968003560639e-05,
+      "logits/chosen": -10.0625,
+      "logits/rejected": -9.8125,
+      "logps/chosen": -720.0,
+      "logps/rejected": -848.0,
+      "loss": 0.6256,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -3.03125,
+      "rewards/margins": 0.53515625,
+      "rewards/rejected": -3.5625,
+      "step": 620
+    },
+    {
+      "epoch": 0.6596858638743456,
+      "grad_norm": 2.006535224501023,
+      "learning_rate": 4.7847714344635864e-05,
+      "logits/chosen": -10.125,
+      "logits/rejected": -10.0,
+      "logps/chosen": -704.0,
+      "logps/rejected": -736.0,
+      "loss": 0.6463,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": -3.171875,
+      "rewards/margins": 0.38671875,
+      "rewards/rejected": -3.546875,
+      "step": 630
+    },
+    {
+      "epoch": 0.6701570680628273,
+      "grad_norm": 2.0982869503696198,
+      "learning_rate": 4.772235575174776e-05,
+      "logits/chosen": -10.4375,
+      "logits/rejected": -10.25,
+      "logps/chosen": -740.0,
+      "logps/rejected": -828.0,
+      "loss": 0.5672,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -3.140625,
+      "rewards/margins": 0.5390625,
+      "rewards/rejected": -3.671875,
+      "step": 640
+    },
+    {
+      "epoch": 0.680628272251309,
+      "grad_norm": 2.3525665482533804,
+      "learning_rate": 4.7593622872782754e-05,
+      "logits/chosen": -9.1875,
+      "logits/rejected": -9.0625,
+      "logps/chosen": -716.0,
+      "logps/rejected": -820.0,
+      "loss": 0.587,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -3.296875,
+      "rewards/margins": 0.6015625,
+      "rewards/rejected": -3.90625,
+      "step": 650
+    },
+    {
+      "epoch": 0.6910994764397905,
+      "grad_norm": 2.2945758463452726,
+      "learning_rate": 4.746153482466542e-05,
+      "logits/chosen": -10.0625,
+      "logits/rejected": -9.9375,
+      "logps/chosen": -716.0,
+      "logps/rejected": -808.0,
+      "loss": 0.6166,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": -3.15625,
+      "rewards/margins": 0.51171875,
+      "rewards/rejected": -3.671875,
+      "step": 660
+    },
+    {
+      "epoch": 0.7015706806282722,
+      "grad_norm": 2.0610003891890534,
+      "learning_rate": 4.732611122256537e-05,
+      "logits/chosen": -11.75,
+      "logits/rejected": -11.5,
+      "logps/chosen": -740.0,
+      "logps/rejected": -816.0,
+      "loss": 0.5886,
+      "rewards/accuracies": 0.581250011920929,
+      "rewards/chosen": -3.0,
+      "rewards/margins": 0.431640625,
+      "rewards/rejected": -3.421875,
+      "step": 670
+    },
+    {
+      "epoch": 0.7120418848167539,
+      "grad_norm": 2.2590551111166888,
+      "learning_rate": 4.718737217698435e-05,
+      "logits/chosen": -8.8125,
+      "logits/rejected": -8.8125,
+      "logps/chosen": -728.0,
+      "logps/rejected": -836.0,
+      "loss": 0.6212,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -3.390625,
+      "rewards/margins": 0.58984375,
+      "rewards/rejected": -3.984375,
+      "step": 680
+    },
+    {
+      "epoch": 0.7225130890052356,
+      "grad_norm": 2.1801154298219974,
+      "learning_rate": 4.704533829076982e-05,
+      "logits/chosen": -9.75,
+      "logits/rejected": -9.5,
+      "logps/chosen": -652.0,
+      "logps/rejected": -756.0,
+      "loss": 0.5912,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -2.515625,
+      "rewards/margins": 0.419921875,
+      "rewards/rejected": -2.921875,
+      "step": 690
+    },
+    {
+      "epoch": 0.7329842931937173,
+      "grad_norm": 2.5143357331963316,
+      "learning_rate": 4.6900030656055455e-05,
+      "logits/chosen": -7.96875,
+      "logits/rejected": -7.5625,
+      "logps/chosen": -716.0,
+      "logps/rejected": -820.0,
+      "loss": 0.5821,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -3.0,
+      "rewards/margins": 0.53125,
+      "rewards/rejected": -3.53125,
+      "step": 700
+    },
+    {
+      "epoch": 0.7329842931937173,
+      "eval_logits/chosen": -5.375,
+      "eval_logits/rejected": -3.9375,
+      "eval_logps/chosen": -632.0,
+      "eval_logps/rejected": -644.0,
+      "eval_loss": 0.715093731880188,
+      "eval_rewards/accuracies": 0.611328125,
+      "eval_rewards/chosen": -3.125,
+      "eval_rewards/margins": 0.419921875,
+      "eval_rewards/rejected": -3.546875,
+      "eval_runtime": 46.611,
+      "eval_samples_per_second": 42.908,
+      "eval_steps_per_second": 0.687,
+      "step": 700
+    },
+    {
+      "epoch": 0.743455497382199,
+      "grad_norm": 2.0534132879099145,
+      "learning_rate": 4.6751470851128906e-05,
+      "logits/chosen": -5.65625,
+      "logits/rejected": -5.375,
+      "logps/chosen": -780.0,
+      "logps/rejected": -844.0,
+      "loss": 0.6313,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -3.40625,
+      "rewards/margins": 0.46875,
+      "rewards/rejected": -3.875,
+      "step": 710
+    },
+    {
+      "epoch": 0.7539267015706806,
+      "grad_norm": 2.242191803721493,
+      "learning_rate": 4.6599680937227454e-05,
+      "logits/chosen": -3.6875,
+      "logits/rejected": -3.375,
+      "logps/chosen": -740.0,
+      "logps/rejected": -808.0,
+      "loss": 0.6236,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -3.109375,
+      "rewards/margins": 0.4296875,
+      "rewards/rejected": -3.53125,
+      "step": 720
+    },
+    {
+      "epoch": 0.7643979057591623,
+      "grad_norm": 1.8901824495921005,
+      "learning_rate": 4.644468345526183e-05,
+      "logits/chosen": -6.875,
+      "logits/rejected": -6.75,
+      "logps/chosen": -696.0,
+      "logps/rejected": -812.0,
+      "loss": 0.6099,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -2.765625,
+      "rewards/margins": 0.474609375,
+      "rewards/rejected": -3.234375,
+      "step": 730
+    },
+    {
+      "epoch": 0.774869109947644,
+      "grad_norm": 2.128218267564539,
+      "learning_rate": 4.628650142246894e-05,
+      "logits/chosen": -6.1875,
+      "logits/rejected": -5.96875,
+      "logps/chosen": -760.0,
+      "logps/rejected": -884.0,
+      "loss": 0.5299,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -3.71875,
+      "rewards/margins": 0.72265625,
+      "rewards/rejected": -4.4375,
+      "step": 740
+    },
+    {
+      "epoch": 0.7853403141361257,
+      "grad_norm": 2.6090557362251148,
+      "learning_rate": 4.6125158328993765e-05,
+      "logits/chosen": -6.71875,
+      "logits/rejected": -6.25,
+      "logps/chosen": -836.0,
+      "logps/rejected": -912.0,
+      "loss": 0.5944,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -4.4375,
+      "rewards/margins": 0.51171875,
+      "rewards/rejected": -4.9375,
+      "step": 750
+    },
+    {
+      "epoch": 0.7958115183246073,
+      "grad_norm": 1.8603642338209474,
+      "learning_rate": 4.5960678134401064e-05,
+      "logits/chosen": -8.6875,
+      "logits/rejected": -8.1875,
+      "logps/chosen": -716.0,
+      "logps/rejected": -852.0,
+      "loss": 0.5892,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -3.015625,
+      "rewards/margins": 0.62109375,
+      "rewards/rejected": -3.640625,
+      "step": 760
+    },
+    {
+      "epoch": 0.806282722513089,
+      "grad_norm": 2.661740316149045,
+      "learning_rate": 4.579308526411734e-05,
+      "logits/chosen": -6.625,
+      "logits/rejected": -6.1875,
+      "logps/chosen": -696.0,
+      "logps/rejected": -796.0,
+      "loss": 0.5629,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -2.921875,
+      "rewards/margins": 0.63671875,
+      "rewards/rejected": -3.546875,
+      "step": 770
+    },
+    {
+      "epoch": 0.8167539267015707,
+      "grad_norm": 2.9631561494807523,
+      "learning_rate": 4.562240460580367e-05,
+      "logits/chosen": -7.59375,
+      "logits/rejected": -7.53125,
+      "logps/chosen": -728.0,
+      "logps/rejected": -852.0,
+      "loss": 0.589,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -3.25,
+      "rewards/margins": 0.59765625,
+      "rewards/rejected": -3.84375,
+      "step": 780
+    },
+    {
+      "epoch": 0.8272251308900523,
+      "grad_norm": 2.5386052960457937,
+      "learning_rate": 4.544866150565985e-05,
+      "logits/chosen": -8.375,
+      "logits/rejected": -8.125,
+      "logps/chosen": -764.0,
+      "logps/rejected": -844.0,
+      "loss": 0.6097,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -3.59375,
+      "rewards/margins": 0.453125,
+      "rewards/rejected": -4.03125,
+      "step": 790
+    },
+    {
+      "epoch": 0.837696335078534,
+      "grad_norm": 2.193524691271299,
+      "learning_rate": 4.527188176466051e-05,
+      "logits/chosen": -8.375,
+      "logits/rejected": -8.25,
+      "logps/chosen": -808.0,
+      "logps/rejected": -920.0,
+      "loss": 0.5988,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -3.671875,
+      "rewards/margins": 0.71484375,
+      "rewards/rejected": -4.375,
+      "step": 800
+    },
+    {
+      "epoch": 0.837696335078534,
+      "eval_logits/chosen": -5.875,
+      "eval_logits/rejected": -4.5625,
+      "eval_logps/chosen": -700.0,
+      "eval_logps/rejected": -712.0,
+      "eval_loss": 0.7464374899864197,
+      "eval_rewards/accuracies": 0.59375,
+      "eval_rewards/chosen": -3.828125,
+      "eval_rewards/margins": 0.40234375,
+      "eval_rewards/rejected": -4.21875,
+      "eval_runtime": 46.6117,
+      "eval_samples_per_second": 42.908,
+      "eval_steps_per_second": 0.687,
+      "step": 800
+    },
+    {
+      "epoch": 0.8481675392670157,
+      "grad_norm": 2.4892091180809897,
+      "learning_rate": 4.509209163472357e-05,
+      "logits/chosen": -7.15625,
+      "logits/rejected": -6.375,
+      "logps/chosen": -800.0,
+      "logps/rejected": -904.0,
+      "loss": 0.5429,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -3.671875,
+      "rewards/margins": 0.62109375,
+      "rewards/rejected": -4.3125,
+      "step": 810
+    },
+    {
+      "epoch": 0.8586387434554974,
+      "grad_norm": 1.940775888408088,
+      "learning_rate": 4.4909317814811896e-05,
+      "logits/chosen": -6.09375,
+      "logits/rejected": -5.75,
+      "logps/chosen": -768.0,
+      "logps/rejected": -860.0,
+      "loss": 0.6083,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -3.59375,
+      "rewards/margins": 0.5625,
+      "rewards/rejected": -4.15625,
+      "step": 820
+    },
+    {
+      "epoch": 0.8691099476439791,
+      "grad_norm": 2.011724797746395,
+      "learning_rate": 4.472358744696845e-05,
+      "logits/chosen": -6.5,
+      "logits/rejected": -6.5,
+      "logps/chosen": -736.0,
+      "logps/rejected": -844.0,
+      "loss": 0.5941,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -3.25,
+      "rewards/margins": 0.58984375,
+      "rewards/rejected": -3.84375,
+      "step": 830
+    },
+    {
+      "epoch": 0.8795811518324608,
+      "grad_norm": 2.313368352938319,
+      "learning_rate": 4.4534928112285645e-05,
+      "logits/chosen": -6.0,
+      "logits/rejected": -5.875,
+      "logps/chosen": -804.0,
+      "logps/rejected": -880.0,
+      "loss": 0.5926,
+      "rewards/accuracies": 0.59375,
+      "rewards/chosen": -3.96875,
+      "rewards/margins": 0.365234375,
+      "rewards/rejected": -4.34375,
+      "step": 840
+    },
+    {
+      "epoch": 0.8900523560209425,
+      "grad_norm": 2.17605642795436,
+      "learning_rate": 4.43433678268096e-05,
+      "logits/chosen": -6.53125,
+      "logits/rejected": -6.3125,
+      "logps/chosen": -752.0,
+      "logps/rejected": -824.0,
+      "loss": 0.5663,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -3.625,
+      "rewards/margins": 0.4765625,
+      "rewards/rejected": -4.09375,
+      "step": 850
+    },
+    {
+      "epoch": 0.900523560209424,
+      "grad_norm": 2.343309821088261,
+      "learning_rate": 4.414893503737968e-05,
+      "logits/chosen": -6.75,
+      "logits/rejected": -6.40625,
+      "logps/chosen": -752.0,
+      "logps/rejected": -864.0,
+      "loss": 0.58,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -3.796875,
+      "rewards/margins": 0.65234375,
+      "rewards/rejected": -4.46875,
+      "step": 860
+    },
+    {
+      "epoch": 0.9109947643979057,
+      "grad_norm": 2.829044072808736,
+      "learning_rate": 4.395165861740414e-05,
+      "logits/chosen": -7.71875,
+      "logits/rejected": -7.5625,
+      "logps/chosen": -784.0,
+      "logps/rejected": -920.0,
+      "loss": 0.5762,
+      "rewards/accuracies": 0.768750011920929,
+      "rewards/chosen": -4.03125,
+      "rewards/margins": 0.8671875,
+      "rewards/rejected": -4.90625,
+      "step": 870
+    },
+    {
+      "epoch": 0.9214659685863874,
+      "grad_norm": 2.0242284696809034,
+      "learning_rate": 4.375156786257241e-05,
+      "logits/chosen": -9.8125,
+      "logits/rejected": -9.4375,
+      "logps/chosen": -824.0,
+      "logps/rejected": -924.0,
+      "loss": 0.6051,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": -3.75,
+      "rewards/margins": 0.52734375,
+      "rewards/rejected": -4.28125,
+      "step": 880
+    },
+    {
+      "epoch": 0.9319371727748691,
+      "grad_norm": 1.8953208601739784,
+      "learning_rate": 4.354869248650466e-05,
+      "logits/chosen": -10.5625,
+      "logits/rejected": -10.125,
+      "logps/chosen": -680.0,
+      "logps/rejected": -836.0,
+      "loss": 0.559,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -2.609375,
+      "rewards/margins": 0.6015625,
+      "rewards/rejected": -3.203125,
+      "step": 890
+    },
+    {
+      "epoch": 0.9424083769633508,
+      "grad_norm": 3.1082072428029326,
+      "learning_rate": 4.334306261633931e-05,
+      "logits/chosen": -6.28125,
+      "logits/rejected": -5.96875,
+      "logps/chosen": -840.0,
+      "logps/rejected": -972.0,
+      "loss": 0.5824,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -3.96875,
+      "rewards/margins": 0.68359375,
+      "rewards/rejected": -4.65625,
+      "step": 900
+    },
+    {
+      "epoch": 0.9424083769633508,
+      "eval_logits/chosen": -4.4375,
+      "eval_logits/rejected": -3.0,
+      "eval_logps/chosen": -788.0,
+      "eval_logps/rejected": -804.0,
+      "eval_loss": 0.7908124923706055,
+      "eval_rewards/accuracies": 0.58984375,
+      "eval_rewards/chosen": -4.6875,
+      "eval_rewards/margins": 0.482421875,
+      "eval_rewards/rejected": -5.15625,
+      "eval_runtime": 46.6174,
+      "eval_samples_per_second": 42.902,
+      "eval_steps_per_second": 0.686,
+      "step": 900
+    },
+    {
+      "epoch": 0.9528795811518325,
+      "grad_norm": 2.5784500810319635,
+      "learning_rate": 4.313470878825911e-05,
+      "logits/chosen": -7.375,
+      "logits/rejected": -6.9375,
+      "logps/chosen": -796.0,
+      "logps/rejected": -884.0,
+      "loss": 0.5947,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -3.875,
+      "rewards/margins": 0.6953125,
+      "rewards/rejected": -4.5625,
+      "step": 910
+    },
+    {
+      "epoch": 0.9633507853403142,
+      "grad_norm": 2.1918237202377786,
+      "learning_rate": 4.292366194295652e-05,
+      "logits/chosen": -7.75,
+      "logits/rejected": -7.0,
+      "logps/chosen": -772.0,
+      "logps/rejected": -844.0,
+      "loss": 0.5602,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -3.421875,
+      "rewards/margins": 0.66015625,
+      "rewards/rejected": -4.09375,
+      "step": 920
+    },
+    {
+      "epoch": 0.9738219895287958,
+      "grad_norm": 2.8358661707793114,
+      "learning_rate": 4.2709953421038964e-05,
+      "logits/chosen": -6.59375,
+      "logits/rejected": -6.3125,
+      "logps/chosen": -704.0,
+      "logps/rejected": -832.0,
+      "loss": 0.5916,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -3.140625,
+      "rewards/margins": 0.625,
+      "rewards/rejected": -3.765625,
+      "step": 930
+    },
+    {
+      "epoch": 0.9842931937172775,
+      "grad_norm": 2.1835135293595167,
+      "learning_rate": 4.249361495837478e-05,
+      "logits/chosen": -7.625,
+      "logits/rejected": -6.9375,
+      "logps/chosen": -664.0,
+      "logps/rejected": -800.0,
+      "loss": 0.5889,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -2.734375,
+      "rewards/margins": 0.58203125,
+      "rewards/rejected": -3.3125,
+      "step": 940
+    },
+    {
+      "epoch": 0.9947643979057592,
+      "grad_norm": 2.0717097929499078,
+      "learning_rate": 4.227467868138035e-05,
+      "logits/chosen": -6.15625,
+      "logits/rejected": -5.71875,
+      "logps/chosen": -704.0,
+      "logps/rejected": -868.0,
+      "loss": 0.5589,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -2.828125,
+      "rewards/margins": 0.78515625,
+      "rewards/rejected": -3.609375,
+      "step": 950
+    },
+    {
+      "epoch": 1.0052356020942408,
+      "grad_norm": 1.1852083896686272,
+      "learning_rate": 4.205317710224938e-05,
+      "logits/chosen": -7.6875,
+      "logits/rejected": -6.8125,
+      "logps/chosen": -692.0,
+      "logps/rejected": -852.0,
+      "loss": 0.4257,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": -2.625,
+      "rewards/margins": 1.4609375,
+      "rewards/rejected": -4.09375,
+      "step": 960
+    },
+    {
+      "epoch": 1.0157068062827226,
+      "grad_norm": 1.3651146435938215,
+      "learning_rate": 4.1829143114124733e-05,
+      "logits/chosen": -5.71875,
+      "logits/rejected": -5.15625,
+      "logps/chosen": -796.0,
+      "logps/rejected": -1136.0,
+      "loss": 0.2274,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -3.859375,
+      "rewards/margins": 2.65625,
+      "rewards/rejected": -6.5,
+      "step": 970
+    },
+    {
+      "epoch": 1.0261780104712042,
+      "grad_norm": 1.7473462217550924,
+      "learning_rate": 4.160260998621387e-05,
+      "logits/chosen": -5.1875,
+      "logits/rejected": -4.625,
+      "logps/chosen": -952.0,
+      "logps/rejected": -1328.0,
+      "loss": 0.1936,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -5.0,
+      "rewards/margins": 3.484375,
+      "rewards/rejected": -8.5,
+      "step": 980
+    },
+    {
+      "epoch": 1.036649214659686,
+      "grad_norm": 1.6117528157054168,
+      "learning_rate": 4.137361135884824e-05,
+      "logits/chosen": -5.8125,
+      "logits/rejected": -5.28125,
+      "logps/chosen": -808.0,
+      "logps/rejected": -1184.0,
+      "loss": 0.1981,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -3.84375,
+      "rewards/margins": 3.078125,
+      "rewards/rejected": -6.9375,
+      "step": 990
+    },
+    {
+      "epoch": 1.0471204188481675,
+      "grad_norm": 2.251485347714079,
+      "learning_rate": 4.114218123848777e-05,
+      "logits/chosen": -2.96875,
+      "logits/rejected": -2.53125,
+      "logps/chosen": -812.0,
+      "logps/rejected": -1192.0,
+      "loss": 0.1932,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -4.09375,
+      "rewards/margins": 3.125,
+      "rewards/rejected": -7.21875,
+      "step": 1000
+    },
+    {
+      "epoch": 1.0471204188481675,
+      "eval_logits/chosen": -0.765625,
+      "eval_logits/rejected": 1.03125,
+      "eval_logps/chosen": -900.0,
+      "eval_logps/rejected": -944.0,
+      "eval_loss": 1.0388749837875366,
+      "eval_rewards/accuracies": 0.609375,
+      "eval_rewards/chosen": -5.78125,
+      "eval_rewards/margins": 0.7421875,
+      "eval_rewards/rejected": -6.53125,
+      "eval_runtime": 46.6113,
+      "eval_samples_per_second": 42.908,
+      "eval_steps_per_second": 0.687,
+      "step": 1000
+    },
+    {
+      "epoch": 1.057591623036649,
+      "grad_norm": 3.3529335693201165,
+      "learning_rate": 4.0908353992670824e-05,
+      "logits/chosen": -3.8125,
+      "logits/rejected": -3.3125,
+      "logps/chosen": -840.0,
+      "logps/rejected": -1208.0,
+      "loss": 0.2266,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -4.4375,
+      "rewards/margins": 3.078125,
+      "rewards/rejected": -7.53125,
+      "step": 1010
+    },
+    {
+      "epoch": 1.068062827225131,
+      "grad_norm": 2.576030914818914,
+      "learning_rate": 4.067216434491059e-05,
+      "logits/chosen": -2.6875,
+      "logits/rejected": -2.390625,
+      "logps/chosen": -924.0,
+      "logps/rejected": -1328.0,
+      "loss": 0.18,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -5.15625,
+      "rewards/margins": 3.328125,
+      "rewards/rejected": -8.5,
+      "step": 1020
+    },
+    {
+      "epoch": 1.0785340314136125,
+      "grad_norm": 1.7701894458115268,
+      "learning_rate": 4.043364736953866e-05,
+      "logits/chosen": -3.65625,
+      "logits/rejected": -2.96875,
+      "logps/chosen": -892.0,
+      "logps/rejected": -1248.0,
+      "loss": 0.2023,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -4.21875,
+      "rewards/margins": 3.484375,
+      "rewards/rejected": -7.71875,
+      "step": 1030
+    },
+    {
+      "epoch": 1.0890052356020943,
+      "grad_norm": 2.47239752450424,
+      "learning_rate": 4.019283848649643e-05,
+      "logits/chosen": -4.6875,
+      "logits/rejected": -3.90625,
+      "logps/chosen": -868.0,
+      "logps/rejected": -1256.0,
+      "loss": 0.1896,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -4.46875,
+      "rewards/margins": 3.390625,
+      "rewards/rejected": -7.84375,
+      "step": 1040
+    },
+    {
+      "epoch": 1.0994764397905759,
+      "grad_norm": 1.9754181728087736,
+      "learning_rate": 3.9949773456075215e-05,
+      "logits/chosen": -3.15625,
+      "logits/rejected": -2.28125,
+      "logps/chosen": -772.0,
+      "logps/rejected": -1168.0,
+      "loss": 0.1877,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -3.921875,
+      "rewards/margins": 3.296875,
+      "rewards/rejected": -7.21875,
+      "step": 1050
+    },
+    {
+      "epoch": 1.1099476439790577,
+      "grad_norm": 1.342306891828322,
+      "learning_rate": 3.970448837360585e-05,
+      "logits/chosen": -2.859375,
+      "logits/rejected": -2.25,
+      "logps/chosen": -908.0,
+      "logps/rejected": -1336.0,
+      "loss": 0.1558,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -4.5625,
+      "rewards/margins": 3.421875,
+      "rewards/rejected": -7.96875,
+      "step": 1060
+    },
+    {
+      "epoch": 1.1204188481675392,
+      "grad_norm": 2.305632981776322,
+      "learning_rate": 3.945701966409846e-05,
+      "logits/chosen": -1.40625,
+      "logits/rejected": -1.3203125,
+      "logps/chosen": -1056.0,
+      "logps/rejected": -1568.0,
+      "loss": 0.1863,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -6.28125,
+      "rewards/margins": 4.25,
+      "rewards/rejected": -10.5625,
+      "step": 1070
+    },
+    {
+      "epoch": 1.130890052356021,
+      "grad_norm": 1.7438028346435068,
+      "learning_rate": 3.9207404076833374e-05,
+      "logits/chosen": -2.875,
+      "logits/rejected": -2.34375,
+      "logps/chosen": -952.0,
+      "logps/rejected": -1312.0,
+      "loss": 0.1815,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -5.03125,
+      "rewards/margins": 3.5625,
+      "rewards/rejected": -8.5625,
+      "step": 1080
+    },
+    {
+      "epoch": 1.1413612565445026,
+      "grad_norm": 2.30187983164181,
+      "learning_rate": 3.895567867990379e-05,
+      "logits/chosen": -4.125,
+      "logits/rejected": -2.734375,
+      "logps/chosen": -972.0,
+      "logps/rejected": -1368.0,
+      "loss": 0.1481,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -4.9375,
+      "rewards/margins": 3.765625,
+      "rewards/rejected": -8.6875,
+      "step": 1090
+    },
+    {
+      "epoch": 1.1518324607329844,
+      "grad_norm": 2.0189474164229373,
+      "learning_rate": 3.870188085471114e-05,
+      "logits/chosen": -2.53125,
+      "logits/rejected": -1.921875,
+      "logps/chosen": -1104.0,
+      "logps/rejected": -1472.0,
+      "loss": 0.1611,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -6.90625,
+      "rewards/margins": 3.734375,
+      "rewards/rejected": -10.625,
+      "step": 1100
+    },
+    {
+      "epoch": 1.1518324607329844,
+      "eval_logits/chosen": -0.0380859375,
+      "eval_logits/rejected": 1.71875,
+      "eval_logps/chosen": -1112.0,
+      "eval_logps/rejected": -1168.0,
+      "eval_loss": 1.1812812089920044,
+      "eval_rewards/accuracies": 0.603515625,
+      "eval_rewards/chosen": -7.9375,
+      "eval_rewards/margins": 0.83984375,
+      "eval_rewards/rejected": -8.75,
+      "eval_runtime": 46.5988,
+      "eval_samples_per_second": 42.92,
+      "eval_steps_per_second": 0.687,
+      "step": 1100
+    },
+    {
+      "epoch": 1.162303664921466,
+      "grad_norm": 1.8542452753624292,
+      "learning_rate": 3.844604829041395e-05,
+      "logits/chosen": -2.859375,
+      "logits/rejected": -2.59375,
+      "logps/chosen": -920.0,
+      "logps/rejected": -1304.0,
+      "loss": 0.1783,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -5.46875,
+      "rewards/margins": 3.5625,
+      "rewards/rejected": -9.0,
+      "step": 1110
+    },
+    {
+      "epoch": 1.1727748691099475,
+      "grad_norm": 2.9911569915291,
+      "learning_rate": 3.818821897833095e-05,
+      "logits/chosen": -1.53125,
+      "logits/rejected": -1.40625,
+      "logps/chosen": -1072.0,
+      "logps/rejected": -1512.0,
+      "loss": 0.2026,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -6.375,
+      "rewards/margins": 3.984375,
+      "rewards/rejected": -10.375,
+      "step": 1120
+    },
+    {
+      "epoch": 1.1832460732984293,
+      "grad_norm": 4.662673489220819,
+      "learning_rate": 3.792843120629935e-05,
+      "logits/chosen": -1.8984375,
+      "logits/rejected": -1.09375,
+      "logps/chosen": -928.0,
+      "logps/rejected": -1360.0,
+      "loss": 0.1992,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -5.28125,
+      "rewards/margins": 3.421875,
+      "rewards/rejected": -8.6875,
+      "step": 1130
+    },
+    {
+      "epoch": 1.193717277486911,
+      "grad_norm": 2.2104391639312015,
+      "learning_rate": 3.766672355298907e-05,
+      "logits/chosen": -3.21875,
+      "logits/rejected": -2.765625,
+      "logps/chosen": -824.0,
+      "logps/rejected": -1176.0,
+      "loss": 0.1909,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -4.46875,
+      "rewards/margins": 3.125,
+      "rewards/rejected": -7.5625,
+      "step": 1140
+    },
+    {
+      "epoch": 1.2041884816753927,
+      "grad_norm": 2.5839987109145857,
+      "learning_rate": 3.740313488217373e-05,
+      "logits/chosen": -2.390625,
+      "logits/rejected": -2.125,
+      "logps/chosen": -916.0,
+      "logps/rejected": -1400.0,
+      "loss": 0.2012,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -5.28125,
+      "rewards/margins": 3.8125,
+      "rewards/rejected": -9.125,
+      "step": 1150
+    },
+    {
+      "epoch": 1.2146596858638743,
+      "grad_norm": 2.0656106534134167,
+      "learning_rate": 3.7137704336959464e-05,
+      "logits/chosen": -1.3671875,
+      "logits/rejected": -1.15625,
+      "logps/chosen": -924.0,
+      "logps/rejected": -1280.0,
+      "loss": 0.2034,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -5.4375,
+      "rewards/margins": 3.15625,
+      "rewards/rejected": -8.625,
+      "step": 1160
+    },
+    {
+      "epoch": 1.225130890052356,
+      "grad_norm": 2.2602196896503983,
+      "learning_rate": 3.687047133397201e-05,
+      "logits/chosen": -3.609375,
+      "logits/rejected": -2.46875,
+      "logps/chosen": -836.0,
+      "logps/rejected": -1208.0,
+      "loss": 0.1654,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -4.46875,
+      "rewards/margins": 3.34375,
+      "rewards/rejected": -7.8125,
+      "step": 1170
+    },
+    {
+      "epoch": 1.2356020942408377,
+      "grad_norm": 2.1601280221641463,
+      "learning_rate": 3.6601475557503415e-05,
+      "logits/chosen": -2.890625,
+      "logits/rejected": -1.75,
+      "logps/chosen": -920.0,
+      "logps/rejected": -1296.0,
+      "loss": 0.1937,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -5.0625,
+      "rewards/margins": 3.25,
+      "rewards/rejected": -8.3125,
+      "step": 1180
+    },
+    {
+      "epoch": 1.2460732984293195,
+      "grad_norm": 1.9649252935270995,
+      "learning_rate": 3.633075695361882e-05,
+      "logits/chosen": -2.59375,
+      "logits/rejected": -1.671875,
+      "logps/chosen": -836.0,
+      "logps/rejected": -1288.0,
+      "loss": 0.1867,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -4.53125,
+      "rewards/margins": 3.90625,
+      "rewards/rejected": -8.4375,
+      "step": 1190
+    },
+    {
+      "epoch": 1.256544502617801,
+      "grad_norm": 1.514820199491948,
+      "learning_rate": 3.6058355724224477e-05,
+      "logits/chosen": -2.328125,
+      "logits/rejected": -1.65625,
+      "logps/chosen": -920.0,
+      "logps/rejected": -1352.0,
+      "loss": 0.1677,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -5.125,
+      "rewards/margins": 3.515625,
+      "rewards/rejected": -8.625,
+      "step": 1200
+    },
+    {
+      "epoch": 1.256544502617801,
+      "eval_logits/chosen": 2.828125,
+      "eval_logits/rejected": 4.71875,
+      "eval_logps/chosen": -1080.0,
+      "eval_logps/rejected": -1128.0,
+      "eval_loss": 1.0714999437332153,
+      "eval_rewards/accuracies": 0.583984375,
+      "eval_rewards/chosen": -7.59375,
+      "eval_rewards/margins": 0.76171875,
+      "eval_rewards/rejected": -8.375,
+      "eval_runtime": 46.6194,
+      "eval_samples_per_second": 42.901,
+      "eval_steps_per_second": 0.686,
+      "step": 1200
+    },
+    {
+      "epoch": 1.2670157068062826,
+      "grad_norm": 3.353847324868901,
+      "learning_rate": 3.5784312321097744e-05,
+      "logits/chosen": -0.52734375,
+      "logits/rejected": -0.06787109375,
+      "logps/chosen": -976.0,
+      "logps/rejected": -1400.0,
+      "loss": 0.2077,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -5.5625,
+      "rewards/margins": 3.828125,
+      "rewards/rejected": -9.375,
+      "step": 1210
+    },
+    {
+      "epoch": 1.2774869109947644,
+      "grad_norm": 2.4635294782854182,
+      "learning_rate": 3.550866743987994e-05,
+      "logits/chosen": -2.453125,
+      "logits/rejected": -1.7421875,
+      "logps/chosen": -952.0,
+      "logps/rejected": -1360.0,
+      "loss": 0.1919,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -4.96875,
+      "rewards/margins": 3.640625,
+      "rewards/rejected": -8.625,
+      "step": 1220
+    },
+    {
+      "epoch": 1.2879581151832462,
+      "grad_norm": 1.8113149513896831,
+      "learning_rate": 3.5231462014033024e-05,
+      "logits/chosen": -2.484375,
+      "logits/rejected": -2.1875,
+      "logps/chosen": -868.0,
+      "logps/rejected": -1256.0,
+      "loss": 0.1734,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -4.71875,
+      "rewards/margins": 3.359375,
+      "rewards/rejected": -8.0625,
+      "step": 1230
+    },
+    {
+      "epoch": 1.2984293193717278,
+      "grad_norm": 1.998164056014304,
+      "learning_rate": 3.495273720876095e-05,
+      "logits/chosen": -0.70703125,
+      "logits/rejected": -0.5625,
+      "logps/chosen": -944.0,
+      "logps/rejected": -1336.0,
+      "loss": 0.1627,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -5.21875,
+      "rewards/margins": 3.5625,
+      "rewards/rejected": -8.8125,
+      "step": 1240
+    },
+    {
+      "epoch": 1.3089005235602094,
+      "grad_norm": 2.2665285554945713,
+      "learning_rate": 3.4672534414896616e-05,
+      "logits/chosen": -2.5,
+      "logits/rejected": -1.90625,
+      "logps/chosen": -972.0,
+      "logps/rejected": -1376.0,
+      "loss": 0.1756,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -5.3125,
+      "rewards/margins": 3.671875,
+      "rewards/rejected": -9.0,
+      "step": 1250
+    },
+    {
+      "epoch": 1.3193717277486912,
+      "grad_norm": 1.3912692858982743,
+      "learning_rate": 3.439089524275526e-05,
+      "logits/chosen": -1.8984375,
+      "logits/rejected": -0.89453125,
+      "logps/chosen": -840.0,
+      "logps/rejected": -1184.0,
+      "loss": 0.1778,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -4.4375,
+      "rewards/margins": 3.0625,
+      "rewards/rejected": -7.5,
+      "step": 1260
+    },
+    {
+      "epoch": 1.3298429319371727,
+      "grad_norm": 2.8563971008682816,
+      "learning_rate": 3.410786151595533e-05,
+      "logits/chosen": -0.74609375,
+      "logits/rejected": -0.369140625,
+      "logps/chosen": -876.0,
+      "logps/rejected": -1288.0,
+      "loss": 0.174,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -4.84375,
+      "rewards/margins": 3.46875,
+      "rewards/rejected": -8.3125,
+      "step": 1270
+    },
+    {
+      "epoch": 1.3403141361256545,
+      "grad_norm": 1.9026978524789542,
+      "learning_rate": 3.382347526520765e-05,
+      "logits/chosen": -0.921875,
+      "logits/rejected": -0.1259765625,
+      "logps/chosen": -888.0,
+      "logps/rejected": -1232.0,
+      "loss": 0.1899,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -4.90625,
+      "rewards/margins": 3.171875,
+      "rewards/rejected": -8.0625,
+      "step": 1280
+    },
+    {
+      "epoch": 1.350785340314136,
+      "grad_norm": 3.089030704809133,
+      "learning_rate": 3.353777872207381e-05,
+      "logits/chosen": -1.53125,
+      "logits/rejected": -1.2734375,
+      "logps/chosen": -888.0,
+      "logps/rejected": -1344.0,
+      "loss": 0.1638,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -4.78125,
+      "rewards/margins": 3.640625,
+      "rewards/rejected": -8.4375,
+      "step": 1290
+    },
+    {
+      "epoch": 1.3612565445026177,
+      "grad_norm": 2.082162326170701,
+      "learning_rate": 3.325081431269475e-05,
+      "logits/chosen": -1.3203125,
+      "logits/rejected": -0.66796875,
+      "logps/chosen": -796.0,
+      "logps/rejected": -1176.0,
+      "loss": 0.1871,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -4.28125,
+      "rewards/margins": 3.515625,
+      "rewards/rejected": -7.78125,
+      "step": 1300
+    },
+    {
+      "epoch": 1.3612565445026177,
+      "eval_logits/chosen": -0.71875,
+      "eval_logits/rejected": 1.0703125,
+      "eval_logps/chosen": -888.0,
+      "eval_logps/rejected": -948.0,
+      "eval_loss": 0.9756406545639038,
+      "eval_rewards/accuracies": 0.58984375,
+      "eval_rewards/chosen": -5.6875,
+      "eval_rewards/margins": 0.8828125,
+      "eval_rewards/rejected": -6.59375,
+      "eval_runtime": 46.623,
+      "eval_samples_per_second": 42.897,
+      "eval_steps_per_second": 0.686,
+      "step": 1300
+    },
+    {
+      "epoch": 1.3717277486910995,
+      "grad_norm": 3.5730261203502547,
+      "learning_rate": 3.296262465149046e-05,
+      "logits/chosen": -3.03125,
+      "logits/rejected": -1.6328125,
+      "logps/chosen": -864.0,
+      "logps/rejected": -1256.0,
+      "loss": 0.1706,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -4.34375,
+      "rewards/margins": 3.296875,
+      "rewards/rejected": -7.625,
+      "step": 1310
+    },
+    {
+      "epoch": 1.3821989528795813,
+      "grad_norm": 1.7771262348371315,
+      "learning_rate": 3.2673252534831687e-05,
+      "logits/chosen": -2.875,
+      "logits/rejected": -1.703125,
+      "logps/chosen": -880.0,
+      "logps/rejected": -1296.0,
+      "loss": 0.184,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -4.71875,
+      "rewards/margins": 3.5625,
+      "rewards/rejected": -8.25,
+      "step": 1320
+    },
+    {
+      "epoch": 1.3926701570680629,
+      "grad_norm": 2.3633299357127617,
+      "learning_rate": 3.2382740934684696e-05,
+      "logits/chosen": -1.875,
+      "logits/rejected": -1.2109375,
+      "logps/chosen": -908.0,
+      "logps/rejected": -1360.0,
+      "loss": 0.2019,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -5.125,
+      "rewards/margins": 3.640625,
+      "rewards/rejected": -8.75,
+      "step": 1330
+    },
+    {
+      "epoch": 1.4031413612565444,
+      "grad_norm": 1.6496365571152773,
+      "learning_rate": 3.209113299222982e-05,
+      "logits/chosen": -1.703125,
+      "logits/rejected": -1.671875,
+      "logps/chosen": -800.0,
+      "logps/rejected": -1208.0,
+      "loss": 0.1634,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -4.53125,
+      "rewards/margins": 3.390625,
+      "rewards/rejected": -7.9375,
+      "step": 1340
+    },
+    {
+      "epoch": 1.4136125654450262,
+      "grad_norm": 3.655631589597375,
+      "learning_rate": 3.179847201145505e-05,
+      "logits/chosen": -1.578125,
+      "logits/rejected": -1.0390625,
+      "logps/chosen": -948.0,
+      "logps/rejected": -1360.0,
+      "loss": 0.1829,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -5.1875,
+      "rewards/margins": 3.609375,
+      "rewards/rejected": -8.8125,
+      "step": 1350
+    },
+    {
+      "epoch": 1.4240837696335078,
+      "grad_norm": 2.5145433899444836,
+      "learning_rate": 3.150480145272528e-05,
+      "logits/chosen": -0.91015625,
+      "logits/rejected": -0.57421875,
+      "logps/chosen": -964.0,
+      "logps/rejected": -1368.0,
+      "loss": 0.185,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -5.65625,
+      "rewards/margins": 3.4375,
+      "rewards/rejected": -9.125,
+      "step": 1360
+    },
+    {
+      "epoch": 1.4345549738219896,
+      "grad_norm": 2.165015139116437,
+      "learning_rate": 3.121016492632848e-05,
+      "logits/chosen": -2.78125,
+      "logits/rejected": -2.0625,
+      "logps/chosen": -904.0,
+      "logps/rejected": -1272.0,
+      "loss": 0.198,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -4.84375,
+      "rewards/margins": 3.296875,
+      "rewards/rejected": -8.125,
+      "step": 1370
+    },
+    {
+      "epoch": 1.4450261780104712,
+      "grad_norm": 1.727156584052044,
+      "learning_rate": 3.091460618599951e-05,
+      "logits/chosen": -1.1484375,
+      "logits/rejected": -0.53515625,
+      "logps/chosen": -904.0,
+      "logps/rejected": -1344.0,
+      "loss": 0.1712,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -5.0,
+      "rewards/margins": 3.640625,
+      "rewards/rejected": -8.625,
+      "step": 1380
+    },
+    {
+      "epoch": 1.4554973821989527,
+      "grad_norm": 2.420666683508774,
+      "learning_rate": 3.061816912242265e-05,
+      "logits/chosen": -0.9453125,
+      "logits/rejected": -0.80078125,
+      "logps/chosen": -1024.0,
+      "logps/rejected": -1464.0,
+      "loss": 0.1707,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -6.03125,
+      "rewards/margins": 3.875,
+      "rewards/rejected": -9.9375,
+      "step": 1390
+    },
+    {
+      "epoch": 1.4659685863874345,
+      "grad_norm": 2.6731362014401587,
+      "learning_rate": 3.0320897756713785e-05,
+      "logits/chosen": -1.3671875,
+      "logits/rejected": -1.0234375,
+      "logps/chosen": -956.0,
+      "logps/rejected": -1416.0,
+      "loss": 0.169,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -5.3125,
+      "rewards/margins": 3.78125,
+      "rewards/rejected": -9.0625,
+      "step": 1400
+    },
+    {
+      "epoch": 1.4659685863874345,
+      "eval_logits/chosen": 1.0234375,
+      "eval_logits/rejected": 2.828125,
+      "eval_logps/chosen": -964.0,
+      "eval_logps/rejected": -1024.0,
+      "eval_loss": 0.9983281493186951,
+      "eval_rewards/accuracies": 0.59375,
+      "eval_rewards/chosen": -6.46875,
+      "eval_rewards/margins": 0.91796875,
+      "eval_rewards/rejected": -7.375,
+      "eval_runtime": 46.6216,
+      "eval_samples_per_second": 42.899,
+      "eval_steps_per_second": 0.686,
+      "step": 1400
+    },
+    {
+      "epoch": 1.4764397905759163,
+      "grad_norm": 1.722589086172948,
+      "learning_rate": 3.002283623388332e-05,
+      "logits/chosen": -1.1328125,
+      "logits/rejected": -0.546875,
+      "logps/chosen": -888.0,
+      "logps/rejected": -1296.0,
+      "loss": 0.179,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -5.15625,
+      "rewards/margins": 3.578125,
+      "rewards/rejected": -8.75,
+      "step": 1410
+    },
+    {
+      "epoch": 1.486910994764398,
+      "grad_norm": 3.989322670041371,
+      "learning_rate": 2.9724028816280507e-05,
+      "logits/chosen": -0.65625,
+      "logits/rejected": -0.703125,
+      "logps/chosen": -908.0,
+      "logps/rejected": -1304.0,
+      "loss": 0.1965,
+      "rewards/accuracies": 0.8687499761581421,
+      "rewards/chosen": -5.15625,
+      "rewards/margins": 3.34375,
+      "rewards/rejected": -8.5,
+      "step": 1420
+    },
+    {
+      "epoch": 1.4973821989528795,
+      "grad_norm": 2.53494074885738,
+      "learning_rate": 2.9424519877020524e-05,
+      "logits/chosen": -2.984375,
+      "logits/rejected": -2.34375,
+      "logps/chosen": -968.0,
+      "logps/rejected": -1384.0,
+      "loss": 0.1626,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -5.28125,
+      "rewards/margins": 3.578125,
+      "rewards/rejected": -8.875,
+      "step": 1430
+    },
+    {
+      "epoch": 1.5078534031413613,
+      "grad_norm": 1.5842256539995752,
+      "learning_rate": 2.912435389339504e-05,
+      "logits/chosen": -1.6328125,
+      "logits/rejected": -0.1728515625,
+      "logps/chosen": -908.0,
+      "logps/rejected": -1312.0,
+      "loss": 0.1581,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -5.0,
+      "rewards/margins": 3.828125,
+      "rewards/rejected": -8.8125,
+      "step": 1440
+    },
+    {
+      "epoch": 1.518324607329843,
+      "grad_norm": 3.087495512288099,
+      "learning_rate": 2.8823575440267226e-05,
+      "logits/chosen": -0.69140625,
+      "logits/rejected": -0.0673828125,
+      "logps/chosen": -988.0,
+      "logps/rejected": -1416.0,
+      "loss": 0.1797,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -5.90625,
+      "rewards/margins": 3.625,
+      "rewards/rejected": -9.5,
+      "step": 1450
+    },
+    {
+      "epoch": 1.5287958115183247,
+      "grad_norm": 1.2791632191158724,
+      "learning_rate": 2.8522229183452465e-05,
+      "logits/chosen": -0.83203125,
+      "logits/rejected": 0.29296875,
+      "logps/chosen": -1020.0,
+      "logps/rejected": -1368.0,
+      "loss": 0.186,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -5.9375,
+      "rewards/margins": 3.515625,
+      "rewards/rejected": -9.4375,
+      "step": 1460
+    },
+    {
+      "epoch": 1.5392670157068062,
+      "grad_norm": 2.0014083739423363,
+      "learning_rate": 2.8220359873085373e-05,
+      "logits/chosen": -2.453125,
+      "logits/rejected": -2.109375,
+      "logps/chosen": -876.0,
+      "logps/rejected": -1272.0,
+      "loss": 0.1622,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -4.65625,
+      "rewards/margins": 3.765625,
+      "rewards/rejected": -8.4375,
+      "step": 1470
+    },
+    {
+      "epoch": 1.5497382198952878,
+      "grad_norm": 2.576363891463643,
+      "learning_rate": 2.7918012336974388e-05,
+      "logits/chosen": -2.4375,
+      "logits/rejected": -2.171875,
+      "logps/chosen": -916.0,
+      "logps/rejected": -1320.0,
+      "loss": 0.1624,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -5.09375,
+      "rewards/margins": 3.765625,
+      "rewards/rejected": -8.875,
+      "step": 1480
+    },
+    {
+      "epoch": 1.5602094240837696,
+      "grad_norm": 2.900637359180197,
+      "learning_rate": 2.7615231473944815e-05,
+      "logits/chosen": -1.0703125,
+      "logits/rejected": -0.201171875,
+      "logps/chosen": -932.0,
+      "logps/rejected": -1352.0,
+      "loss": 0.1597,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -5.4375,
+      "rewards/margins": 3.78125,
+      "rewards/rejected": -9.1875,
+      "step": 1490
+    },
+    {
+      "epoch": 1.5706806282722514,
+      "grad_norm": 2.4130424069258796,
+      "learning_rate": 2.7312062247171327e-05,
+      "logits/chosen": -0.62890625,
+      "logits/rejected": -0.1748046875,
+      "logps/chosen": -976.0,
+      "logps/rejected": -1400.0,
+      "loss": 0.1895,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -5.96875,
+      "rewards/margins": 3.5,
+      "rewards/rejected": -9.5,
+      "step": 1500
+    },
+    {
+      "epoch": 1.5706806282722514,
+      "eval_logits/chosen": 2.421875,
+      "eval_logits/rejected": 4.28125,
+      "eval_logps/chosen": -1040.0,
+      "eval_logps/rejected": -1096.0,
+      "eval_loss": 0.9506718516349792,
+      "eval_rewards/accuracies": 0.61328125,
+      "eval_rewards/chosen": -7.21875,
+      "eval_rewards/margins": 0.83984375,
+      "eval_rewards/rejected": -8.0625,
+      "eval_runtime": 46.6245,
+      "eval_samples_per_second": 42.896,
+      "eval_steps_per_second": 0.686,
+      "step": 1500
+    },
+    {
+      "epoch": 1.581151832460733,
+      "grad_norm": 1.5025426097828793,
+      "learning_rate": 2.7008549677500876e-05,
+      "logits/chosen": -1.734375,
+      "logits/rejected": -0.8203125,
+      "logps/chosen": -976.0,
+      "logps/rejected": -1376.0,
+      "loss": 0.162,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -5.3125,
+      "rewards/margins": 3.578125,
+      "rewards/rejected": -8.875,
+      "step": 1510
+    },
+    {
+      "epoch": 1.5916230366492146,
+      "grad_norm": 2.8860226473413513,
+      "learning_rate": 2.670473883676709e-05,
+      "logits/chosen": -1.3671875,
+      "logits/rejected": -0.35546875,
+      "logps/chosen": -904.0,
+      "logps/rejected": -1288.0,
+      "loss": 0.146,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -4.875,
+      "rewards/margins": 3.4375,
+      "rewards/rejected": -8.3125,
+      "step": 1520
+    },
+    {
+      "epoch": 1.6020942408376964,
+      "grad_norm": 2.9754825605501387,
+      "learning_rate": 2.6400674841097074e-05,
+      "logits/chosen": -1.359375,
+      "logits/rejected": -0.12255859375,
+      "logps/chosen": -1096.0,
+      "logps/rejected": -1512.0,
+      "loss": 0.1559,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -6.75,
+      "rewards/margins": 3.765625,
+      "rewards/rejected": -10.5,
+      "step": 1530
+    },
+    {
+      "epoch": 1.6125654450261782,
+      "grad_norm": 2.4094970103909827,
+      "learning_rate": 2.609640284421158e-05,
+      "logits/chosen": -1.40625,
+      "logits/rejected": -0.84765625,
+      "logps/chosen": -1080.0,
+      "logps/rejected": -1552.0,
+      "loss": 0.1886,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -6.6875,
+      "rewards/margins": 4.0,
+      "rewards/rejected": -10.6875,
+      "step": 1540
+    },
+    {
+      "epoch": 1.6230366492146597,
+      "grad_norm": 2.6206751088456968,
+      "learning_rate": 2.5791968030719726e-05,
+      "logits/chosen": -2.90625,
+      "logits/rejected": -1.6953125,
+      "logps/chosen": -1008.0,
+      "logps/rejected": -1416.0,
+      "loss": 0.1705,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -5.5625,
+      "rewards/margins": 3.71875,
+      "rewards/rejected": -9.25,
+      "step": 1550
+    },
+    {
+      "epoch": 1.6335078534031413,
+      "grad_norm": 1.6672837501667066,
+      "learning_rate": 2.548741560940902e-05,
+      "logits/chosen": -3.078125,
+      "logits/rejected": -2.328125,
+      "logps/chosen": -884.0,
+      "logps/rejected": -1272.0,
+      "loss": 0.1728,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -4.875,
+      "rewards/margins": 3.484375,
+      "rewards/rejected": -8.375,
+      "step": 1560
+    },
+    {
+      "epoch": 1.6439790575916229,
+      "grad_norm": 2.294464578726442,
+      "learning_rate": 2.5182790806531787e-05,
+      "logits/chosen": -1.7109375,
+      "logits/rejected": -1.265625,
+      "logps/chosen": -952.0,
+      "logps/rejected": -1336.0,
+      "loss": 0.1707,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -5.34375,
+      "rewards/margins": 3.53125,
+      "rewards/rejected": -8.875,
+      "step": 1570
+    },
+    {
+      "epoch": 1.6544502617801047,
+      "grad_norm": 2.8738031128553847,
+      "learning_rate": 2.4878138859089074e-05,
+      "logits/chosen": -1.6484375,
+      "logits/rejected": -0.59375,
+      "logps/chosen": -952.0,
+      "logps/rejected": -1360.0,
+      "loss": 0.1447,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -5.46875,
+      "rewards/margins": 3.84375,
+      "rewards/rejected": -9.3125,
+      "step": 1580
+    },
+    {
+      "epoch": 1.6649214659685865,
+      "grad_norm": 2.1920222499573456,
+      "learning_rate": 2.4573505008112924e-05,
+      "logits/chosen": -1.4296875,
+      "logits/rejected": -0.44140625,
+      "logps/chosen": -1056.0,
+      "logps/rejected": -1528.0,
+      "loss": 0.1818,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -6.40625,
+      "rewards/margins": 4.125,
+      "rewards/rejected": -10.5,
+      "step": 1590
+    },
+    {
+      "epoch": 1.675392670157068,
+      "grad_norm": 2.182513378135568,
+      "learning_rate": 2.4268934491948032e-05,
+      "logits/chosen": -0.94140625,
+      "logits/rejected": -0.33203125,
+      "logps/chosen": -1104.0,
+      "logps/rejected": -1544.0,
+      "loss": 0.1599,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -6.8125,
+      "rewards/margins": 3.71875,
+      "rewards/rejected": -10.5,
+      "step": 1600
+    },
+    {
+      "epoch": 1.675392670157068,
+      "eval_logits/chosen": 2.125,
+      "eval_logits/rejected": 3.921875,
+      "eval_logps/chosen": -1128.0,
+      "eval_logps/rejected": -1184.0,
+      "eval_loss": 1.0500469207763672,
+      "eval_rewards/accuracies": 0.595703125,
+      "eval_rewards/chosen": -8.0625,
+      "eval_rewards/margins": 0.859375,
+      "eval_rewards/rejected": -8.9375,
+      "eval_runtime": 46.6192,
+      "eval_samples_per_second": 42.901,
+      "eval_steps_per_second": 0.686,
+      "step": 1600
+    },
+    {
+      "epoch": 1.6858638743455496,
+      "grad_norm": 3.22825197787464,
+      "learning_rate": 2.396447253953385e-05,
+      "logits/chosen": -2.390625,
+      "logits/rejected": -1.8515625,
+      "logps/chosen": -1080.0,
+      "logps/rejected": -1488.0,
+      "loss": 0.1534,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -6.40625,
+      "rewards/margins": 3.6875,
+      "rewards/rejected": -10.125,
+      "step": 1610
+    },
+    {
+      "epoch": 1.6963350785340314,
+      "grad_norm": 3.0758785286255446,
+      "learning_rate": 2.3660164363687998e-05,
+      "logits/chosen": -3.3125,
+      "logits/rejected": -2.875,
+      "logps/chosen": -1008.0,
+      "logps/rejected": -1424.0,
+      "loss": 0.1877,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -5.5,
+      "rewards/margins": 3.640625,
+      "rewards/rejected": -9.1875,
+      "step": 1620
+    },
+    {
+      "epoch": 1.7068062827225132,
+      "grad_norm": 2.947498395275439,
+      "learning_rate": 2.33560551543922e-05,
+      "logits/chosen": -3.1875,
+      "logits/rejected": -2.65625,
+      "logps/chosen": -940.0,
+      "logps/rejected": -1296.0,
+      "loss": 0.1705,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -4.96875,
+      "rewards/margins": 3.296875,
+      "rewards/rejected": -8.25,
+      "step": 1630
+    },
+    {
+      "epoch": 1.7172774869109948,
+      "grad_norm": 2.185922266902538,
+      "learning_rate": 2.3052190072081493e-05,
+      "logits/chosen": -2.21875,
+      "logits/rejected": -1.1484375,
+      "logps/chosen": -952.0,
+      "logps/rejected": -1368.0,
+      "loss": 0.1778,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -5.3125,
+      "rewards/margins": 3.84375,
+      "rewards/rejected": -9.1875,
+      "step": 1640
+    },
+    {
+      "epoch": 1.7277486910994764,
+      "grad_norm": 1.7776087785834662,
+      "learning_rate": 2.2748614240937864e-05,
+      "logits/chosen": -2.28125,
+      "logits/rejected": -1.7421875,
+      "logps/chosen": -912.0,
+      "logps/rejected": -1328.0,
+      "loss": 0.1501,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -5.28125,
+      "rewards/margins": 3.515625,
+      "rewards/rejected": -8.8125,
+      "step": 1650
+    },
+    {
+      "epoch": 1.738219895287958,
+      "grad_norm": 1.665175308038632,
+      "learning_rate": 2.2445372742189334e-05,
+      "logits/chosen": -2.59375,
+      "logits/rejected": -2.125,
+      "logps/chosen": -1000.0,
+      "logps/rejected": -1472.0,
+      "loss": 0.1456,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -5.8125,
+      "rewards/margins": 3.828125,
+      "rewards/rejected": -9.625,
+      "step": 1660
+    },
+    {
+      "epoch": 1.7486910994764397,
+      "grad_norm": 2.975625370155348,
+      "learning_rate": 2.214251060741528e-05,
+      "logits/chosen": -2.15625,
+      "logits/rejected": -1.28125,
+      "logps/chosen": -1064.0,
+      "logps/rejected": -1616.0,
+      "loss": 0.1619,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -6.3125,
+      "rewards/margins": 4.1875,
+      "rewards/rejected": -10.5,
+      "step": 1670
+    },
+    {
+      "epoch": 1.7591623036649215,
+      "grad_norm": 3.99258129665376,
+      "learning_rate": 2.1840072811859302e-05,
+      "logits/chosen": -2.4375,
+      "logits/rejected": -1.53125,
+      "logps/chosen": -1004.0,
+      "logps/rejected": -1384.0,
+      "loss": 0.1688,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -5.6875,
+      "rewards/margins": 3.6875,
+      "rewards/rejected": -9.375,
+      "step": 1680
+    },
+    {
+      "epoch": 1.7696335078534031,
+      "grad_norm": 2.0440376154573756,
+      "learning_rate": 2.1538104267750287e-05,
+      "logits/chosen": -1.5234375,
+      "logits/rejected": -0.97265625,
+      "logps/chosen": -952.0,
+      "logps/rejected": -1392.0,
+      "loss": 0.1505,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -5.5625,
+      "rewards/margins": 3.765625,
+      "rewards/rejected": -9.3125,
+      "step": 1690
+    },
+    {
+      "epoch": 1.7801047120418847,
+      "grad_norm": 2.393050551884817,
+      "learning_rate": 2.1236649817632952e-05,
+      "logits/chosen": -2.5,
+      "logits/rejected": -2.21875,
+      "logps/chosen": -900.0,
+      "logps/rejected": -1360.0,
+      "loss": 0.1422,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -5.34375,
+      "rewards/margins": 3.734375,
+      "rewards/rejected": -9.0625,
+      "step": 1700
+    },
+    {
+      "epoch": 1.7801047120418847,
+      "eval_logits/chosen": 0.0234375,
+      "eval_logits/rejected": 2.03125,
+      "eval_logps/chosen": -1048.0,
+      "eval_logps/rejected": -1112.0,
+      "eval_loss": 1.082124948501587,
+      "eval_rewards/accuracies": 0.5703125,
+      "eval_rewards/chosen": -7.3125,
+      "eval_rewards/margins": 0.9453125,
+      "eval_rewards/rejected": -8.25,
+      "eval_runtime": 46.6163,
+      "eval_samples_per_second": 42.903,
+      "eval_steps_per_second": 0.686,
+      "step": 1700
+    },
+    {
+      "epoch": 1.7905759162303665,
+      "grad_norm": 1.6288504669302029,
+      "learning_rate": 2.093575422770872e-05,
+      "logits/chosen": -1.1015625,
+      "logits/rejected": -0.2001953125,
+      "logps/chosen": -940.0,
+      "logps/rejected": -1344.0,
+      "loss": 0.1378,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -5.65625,
+      "rewards/margins": 3.890625,
+      "rewards/rejected": -9.5625,
+      "step": 1710
+    },
+    {
+      "epoch": 1.8010471204188483,
+      "grad_norm": 2.827264219268634,
+      "learning_rate": 2.0635462181187827e-05,
+      "logits/chosen": -2.4375,
+      "logits/rejected": -1.4609375,
+      "logps/chosen": -1032.0,
+      "logps/rejected": -1488.0,
+      "loss": 0.1503,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -6.09375,
+      "rewards/margins": 4.0,
+      "rewards/rejected": -10.0625,
+      "step": 1720
+    },
+    {
+      "epoch": 1.8115183246073299,
+      "grad_norm": 2.1631471599125986,
+      "learning_rate": 2.0335818271653917e-05,
+      "logits/chosen": -2.40625,
+      "logits/rejected": -1.7734375,
+      "logps/chosen": -948.0,
+      "logps/rejected": -1344.0,
+      "loss": 0.161,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -5.34375,
+      "rewards/margins": 3.53125,
+      "rewards/rejected": -8.875,
+      "step": 1730
+    },
+    {
+      "epoch": 1.8219895287958114,
+      "grad_norm": 2.510751089678634,
+      "learning_rate": 2.0036866996441815e-05,
+      "logits/chosen": -2.484375,
+      "logits/rejected": -1.421875,
+      "logps/chosen": -968.0,
+      "logps/rejected": -1352.0,
+      "loss": 0.1436,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -5.59375,
+      "rewards/margins": 3.875,
+      "rewards/rejected": -9.4375,
+      "step": 1740
+    },
+    {
+      "epoch": 1.8324607329842932,
+      "grad_norm": 1.783107473979916,
+      "learning_rate": 1.9738652750029677e-05,
+      "logits/chosen": -1.3515625,
+      "logits/rejected": -0.4296875,
+      "logps/chosen": -1064.0,
+      "logps/rejected": -1488.0,
+      "loss": 0.1607,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -6.40625,
+      "rewards/margins": 3.96875,
+      "rewards/rejected": -10.375,
+      "step": 1750
+    },
+    {
+      "epoch": 1.8429319371727748,
+      "grad_norm": 2.8785198771845675,
+      "learning_rate": 1.944121981744631e-05,
+      "logits/chosen": -2.640625,
+      "logits/rejected": -1.5546875,
+      "logps/chosen": -1080.0,
+      "logps/rejected": -1512.0,
+      "loss": 0.1401,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -6.1875,
+      "rewards/margins": 3.890625,
+      "rewards/rejected": -10.0625,
+      "step": 1760
+    },
+    {
+      "epoch": 1.8534031413612566,
+      "grad_norm": 2.0960457285022915,
+      "learning_rate": 1.9144612367694887e-05,
+      "logits/chosen": -1.671875,
+      "logits/rejected": -0.890625,
+      "logps/chosen": -976.0,
+      "logps/rejected": -1424.0,
+      "loss": 0.1488,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -5.78125,
+      "rewards/margins": 3.828125,
+      "rewards/rejected": -9.625,
+      "step": 1770
+    },
+    {
+      "epoch": 1.8638743455497382,
+      "grad_norm": 2.694419762263526,
+      "learning_rate": 1.8848874447193803e-05,
+      "logits/chosen": -0.6640625,
+      "logits/rejected": 0.000244140625,
+      "logps/chosen": -1012.0,
+      "logps/rejected": -1456.0,
+      "loss": 0.1638,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -6.1875,
+      "rewards/margins": 3.78125,
+      "rewards/rejected": -9.9375,
+      "step": 1780
+    },
+    {
+      "epoch": 1.8743455497382198,
+      "grad_norm": 2.510244320314152,
+      "learning_rate": 1.855404997323571e-05,
+      "logits/chosen": -2.40625,
+      "logits/rejected": -1.4609375,
+      "logps/chosen": -1040.0,
+      "logps/rejected": -1488.0,
+      "loss": 0.1678,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -6.28125,
+      "rewards/margins": 3.6875,
+      "rewards/rejected": -10.0,
+      "step": 1790
+    },
+    {
+      "epoch": 1.8848167539267016,
+      "grad_norm": 1.448985234803105,
+      "learning_rate": 1.82601827274658e-05,
+      "logits/chosen": -3.15625,
+      "logits/rejected": -2.359375,
+      "logps/chosen": -916.0,
+      "logps/rejected": -1344.0,
+      "loss": 0.154,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -5.25,
+      "rewards/margins": 3.703125,
+      "rewards/rejected": -9.0,
+      "step": 1800
+    },
+    {
+      "epoch": 1.8848167539267016,
+      "eval_logits/chosen": -0.45703125,
+      "eval_logits/rejected": 1.421875,
+      "eval_logps/chosen": -984.0,
+      "eval_logps/rejected": -1048.0,
+      "eval_loss": 0.9749062657356262,
+      "eval_rewards/accuracies": 0.58984375,
+      "eval_rewards/chosen": -6.65625,
+      "eval_rewards/margins": 0.89453125,
+      "eval_rewards/rejected": -7.5625,
+      "eval_runtime": 46.6231,
+      "eval_samples_per_second": 42.897,
+      "eval_steps_per_second": 0.686,
+      "step": 1800
+    },
+    {
+      "epoch": 1.8952879581151834,
+      "grad_norm": 3.2724350221698963,
+      "learning_rate": 1.7967316349380213e-05,
+      "logits/chosen": -2.53125,
+      "logits/rejected": -1.71875,
+      "logps/chosen": -916.0,
+      "logps/rejected": -1320.0,
+      "loss": 0.1553,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -5.40625,
+      "rewards/margins": 3.5625,
+      "rewards/rejected": -8.9375,
+      "step": 1810
+    },
+    {
+      "epoch": 1.905759162303665,
+      "grad_norm": 2.4473491751937084,
+      "learning_rate": 1.7675494329845516e-05,
+      "logits/chosen": -1.265625,
+      "logits/rejected": -0.396484375,
+      "logps/chosen": -976.0,
+      "logps/rejected": -1368.0,
+      "loss": 0.1581,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -5.8125,
+      "rewards/margins": 3.515625,
+      "rewards/rejected": -9.3125,
+      "step": 1820
+    },
+    {
+      "epoch": 1.9162303664921465,
+      "grad_norm": 1.7095589904726014,
+      "learning_rate": 1.738476000464026e-05,
+      "logits/chosen": -1.078125,
+      "logits/rejected": -0.20703125,
+      "logps/chosen": -964.0,
+      "logps/rejected": -1408.0,
+      "loss": 0.1618,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -5.59375,
+      "rewards/margins": 4.09375,
+      "rewards/rejected": -9.6875,
+      "step": 1830
+    },
+    {
+      "epoch": 1.9267015706806283,
+      "grad_norm": 2.2552100205925734,
+      "learning_rate": 1.7095156548019646e-05,
+      "logits/chosen": -1.7578125,
+      "logits/rejected": -1.0546875,
+      "logps/chosen": -960.0,
+      "logps/rejected": -1456.0,
+      "loss": 0.1717,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -5.71875,
+      "rewards/margins": 4.1875,
+      "rewards/rejected": -9.875,
+      "step": 1840
+    },
+    {
+      "epoch": 1.93717277486911,
+      "grad_norm": 2.433997602340511,
+      "learning_rate": 1.6806726966304064e-05,
+      "logits/chosen": -1.4140625,
+      "logits/rejected": -0.57421875,
+      "logps/chosen": -1064.0,
+      "logps/rejected": -1504.0,
+      "loss": 0.1842,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -6.1875,
+      "rewards/margins": 4.03125,
+      "rewards/rejected": -10.1875,
+      "step": 1850
+    },
+    {
+      "epoch": 1.9476439790575917,
+      "grad_norm": 1.9326370289570818,
+      "learning_rate": 1.6519514091492625e-05,
+      "logits/chosen": -1.0234375,
+      "logits/rejected": -0.2890625,
+      "logps/chosen": -980.0,
+      "logps/rejected": -1384.0,
+      "loss": 0.1829,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -6.0,
+      "rewards/margins": 3.59375,
+      "rewards/rejected": -9.5625,
+      "step": 1860
+    },
+    {
+      "epoch": 1.9581151832460733,
+      "grad_norm": 1.7311173077371942,
+      "learning_rate": 1.6233560574902627e-05,
+      "logits/chosen": -1.5625,
+      "logits/rejected": -1.0,
+      "logps/chosen": -916.0,
+      "logps/rejected": -1312.0,
+      "loss": 0.1638,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -5.5,
+      "rewards/margins": 3.703125,
+      "rewards/rejected": -9.1875,
+      "step": 1870
+    },
+    {
+      "epoch": 1.9685863874345548,
+      "grad_norm": 2.987812700042334,
+      "learning_rate": 1.5948908880835753e-05,
+      "logits/chosen": -1.875,
+      "logits/rejected": -1.2734375,
+      "logps/chosen": -988.0,
+      "logps/rejected": -1432.0,
+      "loss": 0.1648,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -5.84375,
+      "rewards/margins": 4.0625,
+      "rewards/rejected": -9.9375,
+      "step": 1880
+    },
+    {
+      "epoch": 1.9790575916230366,
+      "grad_norm": 2.5818968193107636,
+      "learning_rate": 1.5665601280272124e-05,
+      "logits/chosen": -2.375,
+      "logits/rejected": -0.9609375,
+      "logps/chosen": -960.0,
+      "logps/rejected": -1368.0,
+      "loss": 0.1647,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -5.71875,
+      "rewards/margins": 3.8125,
+      "rewards/rejected": -9.5,
+      "step": 1890
+    },
+    {
+      "epoch": 1.9895287958115184,
+      "grad_norm": 1.9196952263911837,
+      "learning_rate": 1.538367984459298e-05,
+      "logits/chosen": -1.5078125,
+      "logits/rejected": -0.59765625,
+      "logps/chosen": -920.0,
+      "logps/rejected": -1376.0,
+      "loss": 0.1366,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -5.34375,
+      "rewards/margins": 3.921875,
+      "rewards/rejected": -9.25,
+      "step": 1900
+    },
+    {
+      "epoch": 1.9895287958115184,
+      "eval_logits/chosen": 0.5390625,
+      "eval_logits/rejected": 2.625,
+      "eval_logps/chosen": -1048.0,
+      "eval_logps/rejected": -1120.0,
+      "eval_loss": 1.005859375,
+      "eval_rewards/accuracies": 0.60546875,
+      "eval_rewards/chosen": -7.3125,
+      "eval_rewards/margins": 1.0234375,
+      "eval_rewards/rejected": -8.3125,
+      "eval_runtime": 46.6194,
+      "eval_samples_per_second": 42.901,
+      "eval_steps_per_second": 0.686,
+      "step": 1900
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 2.9094557210246195,
+      "learning_rate": 1.5103186439333133e-05,
+      "logits/chosen": -1.40625,
+      "logits/rejected": -0.84375,
+      "logps/chosen": -940.0,
+      "logps/rejected": -1400.0,
+      "loss": 0.1646,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -5.53125,
+      "rewards/margins": 3.921875,
+      "rewards/rejected": -9.4375,
+      "step": 1910
+    },
+    {
+      "epoch": 2.0104712041884816,
+      "grad_norm": 0.5745747556792017,
+      "learning_rate": 1.4824162717963829e-05,
+      "logits/chosen": -2.140625,
+      "logits/rejected": 0.6015625,
+      "logps/chosen": -880.0,
+      "logps/rejected": -1488.0,
+      "loss": 0.0305,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -4.875,
+      "rewards/margins": 5.53125,
+      "rewards/rejected": -10.375,
+      "step": 1920
+    },
+    {
+      "epoch": 2.020942408376963,
+      "grad_norm": 1.4958557767882266,
+      "learning_rate": 1.4546650115707192e-05,
+      "logits/chosen": -1.7421875,
+      "logits/rejected": -0.08642578125,
+      "logps/chosen": -952.0,
+      "logps/rejected": -1592.0,
+      "loss": 0.0268,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.78125,
+      "rewards/margins": 6.125,
+      "rewards/rejected": -11.875,
+      "step": 1930
+    },
+    {
+      "epoch": 2.031413612565445,
+      "grad_norm": 1.8785834494380491,
+      "learning_rate": 1.4270689843383111e-05,
+      "logits/chosen": 0.77734375,
+      "logits/rejected": 1.8359375,
+      "logps/chosen": -1120.0,
+      "logps/rejected": -1800.0,
+      "loss": 0.0326,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -7.09375,
+      "rewards/margins": 6.625,
+      "rewards/rejected": -13.6875,
+      "step": 1940
+    },
+    {
+      "epoch": 2.0418848167539267,
+      "grad_norm": 0.6378481099119038,
+      "learning_rate": 1.3996322881289347e-05,
+      "logits/chosen": -0.734375,
+      "logits/rejected": 1.25,
+      "logps/chosen": -1104.0,
+      "logps/rejected": -1824.0,
+      "loss": 0.026,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -7.34375,
+      "rewards/margins": 6.6875,
+      "rewards/rejected": -14.0,
+      "step": 1950
+    },
+    {
+      "epoch": 2.0523560209424083,
+      "grad_norm": 0.6024830773477339,
+      "learning_rate": 1.3723589973115963e-05,
+      "logits/chosen": -0.47265625,
+      "logits/rejected": 0.90625,
+      "logps/chosen": -1064.0,
+      "logps/rejected": -1808.0,
+      "loss": 0.0298,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -6.84375,
+      "rewards/margins": 6.6875,
+      "rewards/rejected": -13.5,
+      "step": 1960
+    },
+    {
+      "epoch": 2.06282722513089,
+      "grad_norm": 0.675999992324466,
+      "learning_rate": 1.3452531619894837e-05,
+      "logits/chosen": 0.5,
+      "logits/rejected": 2.03125,
+      "logps/chosen": -1176.0,
+      "logps/rejected": -1968.0,
+      "loss": 0.0209,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -7.75,
+      "rewards/margins": 7.125,
+      "rewards/rejected": -14.875,
+      "step": 1970
+    },
+    {
+      "epoch": 2.073298429319372,
+      "grad_norm": 1.1638112718676688,
+      "learning_rate": 1.3183188073985245e-05,
+      "logits/chosen": 0.353515625,
+      "logits/rejected": 2.140625,
+      "logps/chosen": -1200.0,
+      "logps/rejected": -2016.0,
+      "loss": 0.0267,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -8.0625,
+      "rewards/margins": 7.46875,
+      "rewards/rejected": -15.5625,
+      "step": 1980
+    },
+    {
+      "epoch": 2.0837696335078535,
+      "grad_norm": 0.18907823609229732,
+      "learning_rate": 1.2915599333096351e-05,
+      "logits/chosen": -0.154296875,
+      "logits/rejected": 1.2421875,
+      "logps/chosen": -1248.0,
+      "logps/rejected": -1968.0,
+      "loss": 0.0207,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.25,
+      "rewards/margins": 6.90625,
+      "rewards/rejected": -15.125,
+      "step": 1990
+    },
+    {
+      "epoch": 2.094240837696335,
+      "grad_norm": 0.19282696459318258,
+      "learning_rate": 1.264980513434752e-05,
+      "logits/chosen": 0.09912109375,
+      "logits/rejected": 2.0,
+      "logps/chosen": -1176.0,
+      "logps/rejected": -1952.0,
+      "loss": 0.0218,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -7.46875,
+      "rewards/margins": 6.96875,
+      "rewards/rejected": -14.4375,
+      "step": 2000
+    },
+    {
+      "epoch": 2.094240837696335,
+      "eval_logits/chosen": 3.28125,
+      "eval_logits/rejected": 5.375,
+      "eval_logps/chosen": -1328.0,
+      "eval_logps/rejected": -1400.0,
+      "eval_loss": 1.4652968645095825,
+      "eval_rewards/accuracies": 0.6015625,
+      "eval_rewards/chosen": -10.0625,
+      "eval_rewards/margins": 1.0546875,
+      "eval_rewards/rejected": -11.125,
+      "eval_runtime": 46.63,
+      "eval_samples_per_second": 42.891,
+      "eval_steps_per_second": 0.686,
+      "step": 2000
+    },
+    {
+      "epoch": 2.1047120418848166,
+      "grad_norm": 0.7904083187308059,
+      "learning_rate": 1.2385844948367324e-05,
+      "logits/chosen": 0.056396484375,
+      "logits/rejected": 1.296875,
+      "logps/chosen": -1104.0,
+      "logps/rejected": -1880.0,
+      "loss": 0.0399,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -7.34375,
+      "rewards/margins": 6.9375,
+      "rewards/rejected": -14.3125,
+      "step": 2010
+    },
+    {
+      "epoch": 2.115183246073298,
+      "grad_norm": 0.3910195015154499,
+      "learning_rate": 1.2123757973432114e-05,
+      "logits/chosen": -0.494140625,
+      "logits/rejected": 1.0546875,
+      "logps/chosen": -1168.0,
+      "logps/rejected": -1904.0,
+      "loss": 0.0228,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -7.5,
+      "rewards/margins": 6.9375,
+      "rewards/rejected": -14.4375,
+      "step": 2020
+    },
+    {
+      "epoch": 2.1256544502617802,
+      "grad_norm": 0.4488576618713909,
+      "learning_rate": 1.1863583129645106e-05,
+      "logits/chosen": -0.0966796875,
+      "logits/rejected": 1.359375,
+      "logps/chosen": -1072.0,
+      "logps/rejected": -1808.0,
+      "loss": 0.0309,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -6.8125,
+      "rewards/margins": 6.84375,
+      "rewards/rejected": -13.625,
+      "step": 2030
+    },
+    {
+      "epoch": 2.136125654450262,
+      "grad_norm": 0.37542724721263093,
+      "learning_rate": 1.1605359053156607e-05,
+      "logits/chosen": 0.030029296875,
+      "logits/rejected": 1.71875,
+      "logps/chosen": -1168.0,
+      "logps/rejected": -1920.0,
+      "loss": 0.0208,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -7.375,
+      "rewards/margins": 7.09375,
+      "rewards/rejected": -14.5,
+      "step": 2040
+    },
+    {
+      "epoch": 2.1465968586387434,
+      "grad_norm": 2.260051083820287,
+      "learning_rate": 1.1349124090426644e-05,
+      "logits/chosen": 1.0546875,
+      "logits/rejected": 2.5,
+      "logps/chosen": -1192.0,
+      "logps/rejected": -2016.0,
+      "loss": 0.0272,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -7.84375,
+      "rewards/margins": 7.4375,
+      "rewards/rejected": -15.3125,
+      "step": 2050
+    },
+    {
+      "epoch": 2.157068062827225,
+      "grad_norm": 0.4143581561572591,
+      "learning_rate": 1.1094916292530406e-05,
+      "logits/chosen": 1.9765625,
+      "logits/rejected": 2.703125,
+      "logps/chosen": -1184.0,
+      "logps/rejected": -1928.0,
+      "loss": 0.0233,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -7.65625,
+      "rewards/margins": 7.0,
+      "rewards/rejected": -14.6875,
+      "step": 2060
+    },
+    {
+      "epoch": 2.167539267015707,
+      "grad_norm": 0.9878750929552627,
+      "learning_rate": 1.0842773409507623e-05,
+      "logits/chosen": 1.515625,
+      "logits/rejected": 2.921875,
+      "logps/chosen": -1112.0,
+      "logps/rejected": -1960.0,
+      "loss": 0.0241,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -7.3125,
+      "rewards/margins": 7.5,
+      "rewards/rejected": -14.8125,
+      "step": 2070
+    },
+    {
+      "epoch": 2.1780104712041886,
+      "grad_norm": 0.29926166514322977,
+      "learning_rate": 1.0592732884756754e-05,
+      "logits/chosen": 0.4921875,
+      "logits/rejected": 1.875,
+      "logps/chosen": -1280.0,
+      "logps/rejected": -2176.0,
+      "loss": 0.0162,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -8.5625,
+      "rewards/margins": 8.0625,
+      "rewards/rejected": -16.625,
+      "step": 2080
+    },
+    {
+      "epoch": 2.18848167539267,
+      "grad_norm": 1.3402976603622259,
+      "learning_rate": 1.0344831849474507e-05,
+      "logits/chosen": 1.1484375,
+      "logits/rejected": 2.875,
+      "logps/chosen": -1280.0,
+      "logps/rejected": -2096.0,
+      "loss": 0.0235,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.625,
+      "rewards/margins": 7.5625,
+      "rewards/rejected": -16.25,
+      "step": 2090
+    },
+    {
+      "epoch": 2.1989528795811517,
+      "grad_norm": 0.4947895336557163,
+      "learning_rate": 1.0099107117141879e-05,
+      "logits/chosen": 2.25,
+      "logits/rejected": 3.546875,
+      "logps/chosen": -1200.0,
+      "logps/rejected": -2064.0,
+      "loss": 0.0277,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.1875,
+      "rewards/margins": 7.84375,
+      "rewards/rejected": -16.0,
+      "step": 2100
+    },
+    {
+      "epoch": 2.1989528795811517,
+      "eval_logits/chosen": 4.375,
+      "eval_logits/rejected": 6.40625,
+      "eval_logps/chosen": -1440.0,
+      "eval_logps/rejected": -1528.0,
+      "eval_loss": 1.515625,
+      "eval_rewards/accuracies": 0.59765625,
+      "eval_rewards/chosen": -11.1875,
+      "eval_rewards/margins": 1.2421875,
+      "eval_rewards/rejected": -12.4375,
+      "eval_runtime": 46.6279,
+      "eval_samples_per_second": 42.893,
+      "eval_steps_per_second": 0.686,
+      "step": 2100
+    },
+    {
+      "epoch": 2.2094240837696333,
+      "grad_norm": 0.3539988047638523,
+      "learning_rate": 9.855595178057334e-06,
+      "logits/chosen": 0.734375,
+      "logits/rejected": 1.90625,
+      "logps/chosen": -1272.0,
+      "logps/rejected": -2080.0,
+      "loss": 0.0241,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -8.5,
+      "rewards/margins": 7.65625,
+      "rewards/rejected": -16.125,
+      "step": 2110
+    },
+    {
+      "epoch": 2.2198952879581153,
+      "grad_norm": 0.4958778863251541,
+      "learning_rate": 9.614332193917949e-06,
+      "logits/chosen": 1.46875,
+      "logits/rejected": 2.546875,
+      "logps/chosen": -1248.0,
+      "logps/rejected": -2080.0,
+      "loss": 0.0337,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -8.25,
+      "rewards/margins": 7.625,
+      "rewards/rejected": -15.875,
+      "step": 2120
+    },
+    {
+      "epoch": 2.230366492146597,
+      "grad_norm": 0.3398646993088154,
+      "learning_rate": 9.375353992449385e-06,
+      "logits/chosen": 0.408203125,
+      "logits/rejected": 1.515625,
+      "logps/chosen": -1216.0,
+      "logps/rejected": -2064.0,
+      "loss": 0.0184,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -8.125,
+      "rewards/margins": 7.53125,
+      "rewards/rejected": -15.6875,
+      "step": 2130
+    },
+    {
+      "epoch": 2.2408376963350785,
+      "grad_norm": 1.8555779523257858,
+      "learning_rate": 9.138696062085442e-06,
+      "logits/chosen": 0.451171875,
+      "logits/rejected": 2.125,
+      "logps/chosen": -1176.0,
+      "logps/rejected": -1968.0,
+      "loss": 0.0296,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -7.75,
+      "rewards/margins": 7.21875,
+      "rewards/rejected": -15.0,
+      "step": 2140
+    },
+    {
+      "epoch": 2.25130890052356,
+      "grad_norm": 0.2541725221711569,
+      "learning_rate": 8.904393546698007e-06,
+      "logits/chosen": 0.462890625,
+      "logits/rejected": 1.8125,
+      "logps/chosen": -1168.0,
+      "logps/rejected": -2032.0,
+      "loss": 0.0217,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -7.59375,
+      "rewards/margins": 8.3125,
+      "rewards/rejected": -15.875,
+      "step": 2150
+    },
+    {
+      "epoch": 2.261780104712042,
+      "grad_norm": 0.8203770527418772,
+      "learning_rate": 8.672481240378142e-06,
+      "logits/chosen": 0.71875,
+      "logits/rejected": 2.5625,
+      "logps/chosen": -1152.0,
+      "logps/rejected": -1936.0,
+      "loss": 0.0356,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -7.875,
+      "rewards/margins": 7.09375,
+      "rewards/rejected": -15.0,
+      "step": 2160
+    },
+    {
+      "epoch": 2.2722513089005236,
+      "grad_norm": 0.3196567652459293,
+      "learning_rate": 8.44299358226919e-06,
+      "logits/chosen": 1.2109375,
+      "logits/rejected": 2.640625,
+      "logps/chosen": -1176.0,
+      "logps/rejected": -1960.0,
+      "loss": 0.0313,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -7.8125,
+      "rewards/margins": 7.375,
+      "rewards/rejected": -15.125,
+      "step": 2170
+    },
+    {
+      "epoch": 2.282722513089005,
+      "grad_norm": 0.6572197989733776,
+      "learning_rate": 8.215964651452457e-06,
+      "logits/chosen": 1.8984375,
+      "logits/rejected": 3.296875,
+      "logps/chosen": -1264.0,
+      "logps/rejected": -2144.0,
+      "loss": 0.0195,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -8.6875,
+      "rewards/margins": 7.90625,
+      "rewards/rejected": -16.5,
+      "step": 2180
+    },
+    {
+      "epoch": 2.2931937172774868,
+      "grad_norm": 0.22666188533018228,
+      "learning_rate": 7.991428161886502e-06,
+      "logits/chosen": 1.4453125,
+      "logits/rejected": 2.5,
+      "logps/chosen": -1248.0,
+      "logps/rejected": -2048.0,
+      "loss": 0.0169,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -8.375,
+      "rewards/margins": 7.4375,
+      "rewards/rejected": -15.8125,
+      "step": 2190
+    },
+    {
+      "epoch": 2.303664921465969,
+      "grad_norm": 0.19215205857855636,
+      "learning_rate": 7.76941745740061e-06,
+      "logits/chosen": 1.3125,
+      "logits/rejected": 2.640625,
+      "logps/chosen": -1336.0,
+      "logps/rejected": -2176.0,
+      "loss": 0.0238,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -9.0625,
+      "rewards/margins": 7.90625,
+      "rewards/rejected": -17.0,
+      "step": 2200
+    },
+    {
+      "epoch": 2.303664921465969,
+      "eval_logits/chosen": 4.5,
+      "eval_logits/rejected": 6.5,
+      "eval_logps/chosen": -1448.0,
+      "eval_logps/rejected": -1544.0,
+      "eval_loss": 1.546468734741211,
+      "eval_rewards/accuracies": 0.59375,
+      "eval_rewards/chosen": -11.25,
+      "eval_rewards/margins": 1.25,
+      "eval_rewards/rejected": -12.5,
+      "eval_runtime": 46.6299,
+      "eval_samples_per_second": 42.891,
+      "eval_steps_per_second": 0.686,
+      "step": 2200
+    },
+    {
+      "epoch": 2.3141361256544504,
+      "grad_norm": 0.5475037371733472,
+      "learning_rate": 7.549965506743123e-06,
+      "logits/chosen": 1.546875,
+      "logits/rejected": 3.171875,
+      "logps/chosen": -1216.0,
+      "logps/rejected": -1984.0,
+      "loss": 0.0225,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.25,
+      "rewards/margins": 7.4375,
+      "rewards/rejected": -15.6875,
+      "step": 2210
+    },
+    {
+      "epoch": 2.324607329842932,
+      "grad_norm": 0.5713542723496158,
+      "learning_rate": 7.333104898685631e-06,
+      "logits/chosen": 1.171875,
+      "logits/rejected": 2.375,
+      "logps/chosen": -1248.0,
+      "logps/rejected": -2040.0,
+      "loss": 0.0275,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -8.5,
+      "rewards/margins": 7.59375,
+      "rewards/rejected": -16.125,
+      "step": 2220
+    },
+    {
+      "epoch": 2.3350785340314135,
+      "grad_norm": 0.3784952892838445,
+      "learning_rate": 7.118867837183518e-06,
+      "logits/chosen": 0.7890625,
+      "logits/rejected": 2.15625,
+      "logps/chosen": -1360.0,
+      "logps/rejected": -2192.0,
+      "loss": 0.0222,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.8125,
+      "rewards/margins": 8.125,
+      "rewards/rejected": -17.0,
+      "step": 2230
+    },
+    {
+      "epoch": 2.345549738219895,
+      "grad_norm": 0.18162331329271378,
+      "learning_rate": 6.9072861365936045e-06,
+      "logits/chosen": 0.87890625,
+      "logits/rejected": 1.8984375,
+      "logps/chosen": -1440.0,
+      "logps/rejected": -2288.0,
+      "loss": 0.0179,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -9.75,
+      "rewards/margins": 8.125,
+      "rewards/rejected": -17.875,
+      "step": 2240
+    },
+    {
+      "epoch": 2.356020942408377,
+      "grad_norm": 0.34146578729033666,
+      "learning_rate": 6.6983912169497005e-06,
+      "logits/chosen": 0.86328125,
+      "logits/rejected": 2.34375,
+      "logps/chosen": -1296.0,
+      "logps/rejected": -2112.0,
+      "loss": 0.0224,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.5,
+      "rewards/margins": 7.9375,
+      "rewards/rejected": -16.375,
+      "step": 2250
+    },
+    {
+      "epoch": 2.3664921465968587,
+      "grad_norm": 0.4072975181576973,
+      "learning_rate": 6.49221409929677e-06,
+      "logits/chosen": 2.421875,
+      "logits/rejected": 3.6875,
+      "logps/chosen": -1208.0,
+      "logps/rejected": -2016.0,
+      "loss": 0.0203,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -8.1875,
+      "rewards/margins": 7.6875,
+      "rewards/rejected": -15.875,
+      "step": 2260
+    },
+    {
+      "epoch": 2.3769633507853403,
+      "grad_norm": 1.5832439096351454,
+      "learning_rate": 6.2887854010842075e-06,
+      "logits/chosen": 2.4375,
+      "logits/rejected": 3.03125,
+      "logps/chosen": -1320.0,
+      "logps/rejected": -2176.0,
+      "loss": 0.0278,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -9.3125,
+      "rewards/margins": 8.0,
+      "rewards/rejected": -17.375,
+      "step": 2270
+    },
+    {
+      "epoch": 2.387434554973822,
+      "grad_norm": 1.0384212072934378,
+      "learning_rate": 6.0881353316191385e-06,
+      "logits/chosen": 1.1484375,
+      "logits/rejected": 2.625,
+      "logps/chosen": -1248.0,
+      "logps/rejected": -2080.0,
+      "loss": 0.0186,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -8.25,
+      "rewards/margins": 8.1875,
+      "rewards/rejected": -16.375,
+      "step": 2280
+    },
+    {
+      "epoch": 2.3979057591623034,
+      "grad_norm": 0.23497070200993397,
+      "learning_rate": 5.890293687580381e-06,
+      "logits/chosen": 1.671875,
+      "logits/rejected": 3.484375,
+      "logps/chosen": -1288.0,
+      "logps/rejected": -2112.0,
+      "loss": 0.0248,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -8.5625,
+      "rewards/margins": 8.25,
+      "rewards/rejected": -16.75,
+      "step": 2290
+    },
+    {
+      "epoch": 2.4083769633507854,
+      "grad_norm": 0.3119534336019471,
+      "learning_rate": 5.695289848593532e-06,
+      "logits/chosen": 1.09375,
+      "logits/rejected": 1.96875,
+      "logps/chosen": -1240.0,
+      "logps/rejected": -2032.0,
+      "loss": 0.0241,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -8.5,
+      "rewards/margins": 7.625,
+      "rewards/rejected": -16.125,
+      "step": 2300
+    },
+    {
+      "epoch": 2.4083769633507854,
+      "eval_logits/chosen": 4.875,
+      "eval_logits/rejected": 6.875,
+      "eval_logps/chosen": -1488.0,
+      "eval_logps/rejected": -1592.0,
+      "eval_loss": 1.5830390453338623,
+      "eval_rewards/accuracies": 0.59375,
+      "eval_rewards/chosen": -11.6875,
+      "eval_rewards/margins": 1.328125,
+      "eval_rewards/rejected": -13.0625,
+      "eval_runtime": 46.6327,
+      "eval_samples_per_second": 42.888,
+      "eval_steps_per_second": 0.686,
+      "step": 2300
+    },
+    {
+      "epoch": 2.418848167539267,
+      "grad_norm": 0.2041758897797465,
+      "learning_rate": 5.50315277286812e-06,
+      "logits/chosen": 1.6015625,
+      "logits/rejected": 2.609375,
+      "logps/chosen": -1232.0,
+      "logps/rejected": -2096.0,
+      "loss": 0.0213,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.5,
+      "rewards/margins": 7.9375,
+      "rewards/rejected": -16.5,
+      "step": 2310
+    },
+    {
+      "epoch": 2.4293193717277486,
+      "grad_norm": 2.1233188312029596,
+      "learning_rate": 5.313910992897281e-06,
+      "logits/chosen": 1.828125,
+      "logits/rejected": 3.296875,
+      "logps/chosen": -1216.0,
+      "logps/rejected": -2080.0,
+      "loss": 0.0263,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.125,
+      "rewards/margins": 7.78125,
+      "rewards/rejected": -15.875,
+      "step": 2320
+    },
+    {
+      "epoch": 2.4397905759162306,
+      "grad_norm": 0.4388046748510395,
+      "learning_rate": 5.127592611220658e-06,
+      "logits/chosen": 0.53125,
+      "logits/rejected": 2.625,
+      "logps/chosen": -1320.0,
+      "logps/rejected": -2192.0,
+      "loss": 0.0301,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -8.8125,
+      "rewards/margins": 8.4375,
+      "rewards/rejected": -17.25,
+      "step": 2330
+    },
+    {
+      "epoch": 2.450261780104712,
+      "grad_norm": 0.47869760902004876,
+      "learning_rate": 4.944225296251159e-06,
+      "logits/chosen": 1.3125,
+      "logits/rejected": 2.578125,
+      "logps/chosen": -1216.0,
+      "logps/rejected": -2024.0,
+      "loss": 0.019,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -8.4375,
+      "rewards/margins": 7.6875,
+      "rewards/rejected": -16.125,
+      "step": 2340
+    },
+    {
+      "epoch": 2.4607329842931938,
+      "grad_norm": 1.005667498856408,
+      "learning_rate": 4.763836278166172e-06,
+      "logits/chosen": 1.3515625,
+      "logits/rejected": 3.34375,
+      "logps/chosen": -1232.0,
+      "logps/rejected": -1992.0,
+      "loss": 0.0176,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.25,
+      "rewards/margins": 7.625,
+      "rewards/rejected": -15.8125,
+      "step": 2350
+    },
+    {
+      "epoch": 2.4712041884816753,
+      "grad_norm": 0.3131019325551214,
+      "learning_rate": 4.586452344863876e-06,
+      "logits/chosen": 0.67578125,
+      "logits/rejected": 2.5625,
+      "logps/chosen": -1280.0,
+      "logps/rejected": -2160.0,
+      "loss": 0.0223,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -8.625,
+      "rewards/margins": 8.3125,
+      "rewards/rejected": -16.875,
+      "step": 2360
+    },
+    {
+      "epoch": 2.481675392670157,
+      "grad_norm": 2.0561274908618064,
+      "learning_rate": 4.412099837985192e-06,
+      "logits/chosen": 1.8984375,
+      "logits/rejected": 3.359375,
+      "logps/chosen": -1336.0,
+      "logps/rejected": -2272.0,
+      "loss": 0.0308,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -9.1875,
+      "rewards/margins": 8.3125,
+      "rewards/rejected": -17.5,
+      "step": 2370
+    },
+    {
+      "epoch": 2.492146596858639,
+      "grad_norm": 0.14000630283730817,
+      "learning_rate": 4.240804649002089e-06,
+      "logits/chosen": 1.3984375,
+      "logits/rejected": 3.0,
+      "logps/chosen": -1224.0,
+      "logps/rejected": -2048.0,
+      "loss": 0.0385,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.5,
+      "rewards/margins": 7.84375,
+      "rewards/rejected": -16.25,
+      "step": 2380
+    },
+    {
+      "epoch": 2.5026178010471205,
+      "grad_norm": 1.0643495532641605,
+      "learning_rate": 4.072592215372598e-06,
+      "logits/chosen": 0.53125,
+      "logits/rejected": 1.8125,
+      "logps/chosen": -1344.0,
+      "logps/rejected": -2192.0,
+      "loss": 0.0224,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -9.0,
+      "rewards/margins": 7.875,
+      "rewards/rejected": -16.875,
+      "step": 2390
+    },
+    {
+      "epoch": 2.513089005235602,
+      "grad_norm": 0.4533942249177286,
+      "learning_rate": 3.907487516763389e-06,
+      "logits/chosen": 1.890625,
+      "logits/rejected": 3.140625,
+      "logps/chosen": -1240.0,
+      "logps/rejected": -2064.0,
+      "loss": 0.0263,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -8.4375,
+      "rewards/margins": 7.5625,
+      "rewards/rejected": -16.0,
+      "step": 2400
+    },
+    {
+      "epoch": 2.513089005235602,
+      "eval_logits/chosen": 5.15625,
+      "eval_logits/rejected": 7.125,
+      "eval_logps/chosen": -1496.0,
+      "eval_logps/rejected": -1600.0,
+      "eval_loss": 1.551664113998413,
+      "eval_rewards/accuracies": 0.59765625,
+      "eval_rewards/chosen": -11.8125,
+      "eval_rewards/margins": 1.296875,
+      "eval_rewards/rejected": -13.0625,
+      "eval_runtime": 46.6397,
+      "eval_samples_per_second": 42.882,
+      "eval_steps_per_second": 0.686,
+      "step": 2400
+    },
+    {
+      "epoch": 2.5235602094240837,
+      "grad_norm": 1.0130242246764005,
+      "learning_rate": 3.745515071340286e-06,
+      "logits/chosen": 1.6640625,
+      "logits/rejected": 3.015625,
+      "logps/chosen": -1272.0,
+      "logps/rejected": -2128.0,
+      "loss": 0.0246,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -8.5625,
+      "rewards/margins": 7.9375,
+      "rewards/rejected": -16.5,
+      "step": 2410
+    },
+    {
+      "epoch": 2.5340314136125652,
+      "grad_norm": 0.5563637628731153,
+      "learning_rate": 3.586698932127236e-06,
+      "logits/chosen": 0.66796875,
+      "logits/rejected": 2.359375,
+      "logps/chosen": -1216.0,
+      "logps/rejected": -2048.0,
+      "loss": 0.0147,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -7.96875,
+      "rewards/margins": 8.0,
+      "rewards/rejected": -16.0,
+      "step": 2420
+    },
+    {
+      "epoch": 2.5445026178010473,
+      "grad_norm": 0.8530402940188367,
+      "learning_rate": 3.4310626834344744e-06,
+      "logits/chosen": 0.3984375,
+      "logits/rejected": 1.015625,
+      "logps/chosen": -1352.0,
+      "logps/rejected": -2192.0,
+      "loss": 0.0224,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -9.1875,
+      "rewards/margins": 7.875,
+      "rewards/rejected": -17.0,
+      "step": 2430
+    },
+    {
+      "epoch": 2.554973821989529,
+      "grad_norm": 0.9183402850934244,
+      "learning_rate": 3.2786294373562343e-06,
+      "logits/chosen": 1.140625,
+      "logits/rejected": 2.734375,
+      "logps/chosen": -1272.0,
+      "logps/rejected": -2080.0,
+      "loss": 0.0162,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.5,
+      "rewards/margins": 7.6875,
+      "rewards/rejected": -16.25,
+      "step": 2440
+    },
+    {
+      "epoch": 2.5654450261780104,
+      "grad_norm": 0.9555862800076789,
+      "learning_rate": 3.1294218303385526e-06,
+      "logits/chosen": 0.671875,
+      "logits/rejected": 2.4375,
+      "logps/chosen": -1240.0,
+      "logps/rejected": -2008.0,
+      "loss": 0.0205,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -8.0,
+      "rewards/margins": 7.75,
+      "rewards/rejected": -15.75,
+      "step": 2450
+    },
+    {
+      "epoch": 2.5759162303664924,
+      "grad_norm": 1.1487666636531495,
+      "learning_rate": 2.9834620198177483e-06,
+      "logits/chosen": 1.1796875,
+      "logits/rejected": 2.578125,
+      "logps/chosen": -1264.0,
+      "logps/rejected": -2080.0,
+      "loss": 0.03,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -8.625,
+      "rewards/margins": 7.53125,
+      "rewards/rejected": -16.125,
+      "step": 2460
+    },
+    {
+      "epoch": 2.5863874345549736,
+      "grad_norm": 3.265282786538022,
+      "learning_rate": 2.840771680930068e-06,
+      "logits/chosen": 0.640625,
+      "logits/rejected": 1.7265625,
+      "logps/chosen": -1320.0,
+      "logps/rejected": -2256.0,
+      "loss": 0.0172,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -8.875,
+      "rewards/margins": 8.375,
+      "rewards/rejected": -17.25,
+      "step": 2470
+    },
+    {
+      "epoch": 2.5968586387434556,
+      "grad_norm": 0.3622493252948582,
+      "learning_rate": 2.701372003292868e-06,
+      "logits/chosen": 1.796875,
+      "logits/rejected": 3.0,
+      "logps/chosen": -1160.0,
+      "logps/rejected": -2024.0,
+      "loss": 0.0234,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -7.75,
+      "rewards/margins": 7.875,
+      "rewards/rejected": -15.625,
+      "step": 2480
+    },
+    {
+      "epoch": 2.607329842931937,
+      "grad_norm": 0.17573875772480568,
+      "learning_rate": 2.56528368785795e-06,
+      "logits/chosen": 0.67578125,
+      "logits/rejected": 2.84375,
+      "logps/chosen": -1216.0,
+      "logps/rejected": -2048.0,
+      "loss": 0.0201,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -7.9375,
+      "rewards/margins": 8.0625,
+      "rewards/rejected": -16.0,
+      "step": 2490
+    },
+    {
+      "epoch": 2.6178010471204187,
+      "grad_norm": 0.8320037125771946,
+      "learning_rate": 2.432526943837471e-06,
+      "logits/chosen": 1.203125,
+      "logits/rejected": 3.625,
+      "logps/chosen": -1152.0,
+      "logps/rejected": -2008.0,
+      "loss": 0.018,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -7.71875,
+      "rewards/margins": 8.0625,
+      "rewards/rejected": -15.75,
+      "step": 2500
+    },
+    {
+      "epoch": 2.6178010471204187,
+      "eval_logits/chosen": 4.6875,
+      "eval_logits/rejected": 6.71875,
+      "eval_logps/chosen": -1448.0,
+      "eval_logps/rejected": -1544.0,
+      "eval_loss": 1.56640625,
+      "eval_rewards/accuracies": 0.591796875,
+      "eval_rewards/chosen": -11.3125,
+      "eval_rewards/margins": 1.2890625,
+      "eval_rewards/rejected": -12.5625,
+      "eval_runtime": 46.6325,
+      "eval_samples_per_second": 42.889,
+      "eval_steps_per_second": 0.686,
+      "step": 2500
+    },
+    {
+      "epoch": 2.6282722513089007,
+      "grad_norm": 0.1537621627325051,
+      "learning_rate": 2.3031214857028446e-06,
+      "logits/chosen": 2.203125,
+      "logits/rejected": 3.4375,
+      "logps/chosen": -1224.0,
+      "logps/rejected": -2096.0,
+      "loss": 0.0207,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -7.96875,
+      "rewards/margins": 8.3125,
+      "rewards/rejected": -16.25,
+      "step": 2510
+    },
+    {
+      "epoch": 2.6387434554973823,
+      "grad_norm": 0.5239491515021995,
+      "learning_rate": 2.177086530257139e-06,
+      "logits/chosen": 1.2578125,
+      "logits/rejected": 2.625,
+      "logps/chosen": -1216.0,
+      "logps/rejected": -2000.0,
+      "loss": 0.0263,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -8.1875,
+      "rewards/margins": 7.4375,
+      "rewards/rejected": -15.625,
+      "step": 2520
+    },
+    {
+      "epoch": 2.649214659685864,
+      "grad_norm": 0.3179245188825505,
+      "learning_rate": 2.054440793781354e-06,
+      "logits/chosen": 1.2109375,
+      "logits/rejected": 2.78125,
+      "logps/chosen": -1232.0,
+      "logps/rejected": -2064.0,
+      "loss": 0.0267,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -8.3125,
+      "rewards/margins": 7.6875,
+      "rewards/rejected": -16.0,
+      "step": 2530
+    },
+    {
+      "epoch": 2.6596858638743455,
+      "grad_norm": 0.40109831349893904,
+      "learning_rate": 1.9352024892550587e-06,
+      "logits/chosen": 1.453125,
+      "logits/rejected": 2.5625,
+      "logps/chosen": -1248.0,
+      "logps/rejected": -2160.0,
+      "loss": 0.0183,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.4375,
+      "rewards/margins": 8.4375,
+      "rewards/rejected": -16.875,
+      "step": 2540
+    },
+    {
+      "epoch": 2.670157068062827,
+      "grad_norm": 0.37839185391705854,
+      "learning_rate": 1.8193893236517379e-06,
+      "logits/chosen": 1.0703125,
+      "logits/rejected": 2.515625,
+      "logps/chosen": -1248.0,
+      "logps/rejected": -2064.0,
+      "loss": 0.0207,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -8.3125,
+      "rewards/margins": 7.9375,
+      "rewards/rejected": -16.25,
+      "step": 2550
+    },
+    {
+      "epoch": 2.680628272251309,
+      "grad_norm": 0.2533553127077687,
+      "learning_rate": 1.7070184953092956e-06,
+      "logits/chosen": 1.4921875,
+      "logits/rejected": 2.796875,
+      "logps/chosen": -1240.0,
+      "logps/rejected": -2080.0,
+      "loss": 0.0247,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -8.25,
+      "rewards/margins": 7.96875,
+      "rewards/rejected": -16.25,
+      "step": 2560
+    },
+    {
+      "epoch": 2.6910994764397906,
+      "grad_norm": 0.7639562709596049,
+      "learning_rate": 1.5981066913760918e-06,
+      "logits/chosen": 2.5625,
+      "logits/rejected": 4.3125,
+      "logps/chosen": -1160.0,
+      "logps/rejected": -1992.0,
+      "loss": 0.0158,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -7.75,
+      "rewards/margins": 7.78125,
+      "rewards/rejected": -15.5625,
+      "step": 2570
+    },
+    {
+      "epoch": 2.701570680628272,
+      "grad_norm": 0.17233256158423194,
+      "learning_rate": 1.492670085332884e-06,
+      "logits/chosen": 1.5,
+      "logits/rejected": 2.828125,
+      "logps/chosen": -1240.0,
+      "logps/rejected": -2128.0,
+      "loss": 0.0275,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -8.375,
+      "rewards/margins": 8.0,
+      "rewards/rejected": -16.375,
+      "step": 2580
+    },
+    {
+      "epoch": 2.712041884816754,
+      "grad_norm": 0.12150387193495278,
+      "learning_rate": 1.3907243345910787e-06,
+      "logits/chosen": 1.9453125,
+      "logits/rejected": 3.25,
+      "logps/chosen": -1248.0,
+      "logps/rejected": -2112.0,
+      "loss": 0.0189,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -8.0625,
+      "rewards/margins": 8.25,
+      "rewards/rejected": -16.25,
+      "step": 2590
+    },
+    {
+      "epoch": 2.7225130890052354,
+      "grad_norm": 0.3284330814884494,
+      "learning_rate": 1.292284578167538e-06,
+      "logits/chosen": 1.5859375,
+      "logits/rejected": 2.734375,
+      "logps/chosen": -1216.0,
+      "logps/rejected": -2096.0,
+      "loss": 0.0149,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -8.25,
+      "rewards/margins": 7.9375,
+      "rewards/rejected": -16.25,
+      "step": 2600
+    },
+    {
+      "epoch": 2.7225130890052354,
+      "eval_logits/chosen": 5.125,
+      "eval_logits/rejected": 7.09375,
+      "eval_logps/chosen": -1504.0,
+      "eval_logps/rejected": -1600.0,
+      "eval_loss": 1.6000781059265137,
+      "eval_rewards/accuracies": 0.5859375,
+      "eval_rewards/chosen": -11.8125,
+      "eval_rewards/margins": 1.3046875,
+      "eval_rewards/rejected": -13.125,
+      "eval_runtime": 46.6248,
+      "eval_samples_per_second": 42.896,
+      "eval_steps_per_second": 0.686,
+      "step": 2600
+    },
+    {
+      "epoch": 2.7329842931937174,
+      "grad_norm": 1.878771586699495,
+      "learning_rate": 1.1973654344364925e-06,
+      "logits/chosen": 1.3203125,
+      "logits/rejected": 2.28125,
+      "logps/chosen": -1288.0,
+      "logps/rejected": -2176.0,
+      "loss": 0.0192,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -8.5625,
+      "rewards/margins": 8.375,
+      "rewards/rejected": -17.0,
+      "step": 2610
+    },
+    {
+      "epoch": 2.743455497382199,
+      "grad_norm": 0.434697711843355,
+      "learning_rate": 1.1059809989586583e-06,
+      "logits/chosen": 2.65625,
+      "logits/rejected": 3.4375,
+      "logps/chosen": -1240.0,
+      "logps/rejected": -2128.0,
+      "loss": 0.02,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.3125,
+      "rewards/margins": 8.1875,
+      "rewards/rejected": -16.5,
+      "step": 2620
+    },
+    {
+      "epoch": 2.7539267015706805,
+      "grad_norm": 0.3403367943539808,
+      "learning_rate": 1.018144842388044e-06,
+      "logits/chosen": 2.40625,
+      "logits/rejected": 4.0625,
+      "logps/chosen": -1256.0,
+      "logps/rejected": -2176.0,
+      "loss": 0.0246,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -8.1875,
+      "rewards/margins": 8.375,
+      "rewards/rejected": -16.5,
+      "step": 2630
+    },
+    {
+      "epoch": 2.7643979057591626,
+      "grad_norm": 0.3703660142954374,
+      "learning_rate": 9.338700084567109e-07,
+      "logits/chosen": 1.3359375,
+      "logits/rejected": 2.828125,
+      "logps/chosen": -1224.0,
+      "logps/rejected": -2080.0,
+      "loss": 0.0189,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -8.125,
+      "rewards/margins": 8.25,
+      "rewards/rejected": -16.375,
+      "step": 2640
+    },
+    {
+      "epoch": 2.774869109947644,
+      "grad_norm": 0.7045147638050032,
+      "learning_rate": 8.531690120377606e-07,
+      "logits/chosen": 1.34375,
+      "logits/rejected": 2.984375,
+      "logps/chosen": -1392.0,
+      "logps/rejected": -2304.0,
+      "loss": 0.0212,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -9.25,
+      "rewards/margins": 9.0,
+      "rewards/rejected": -18.25,
+      "step": 2650
+    },
+    {
+      "epoch": 2.7853403141361257,
+      "grad_norm": 0.5362191382912724,
+      "learning_rate": 7.760538372868636e-07,
+      "logits/chosen": 1.3671875,
+      "logits/rejected": 2.703125,
+      "logps/chosen": -1264.0,
+      "logps/rejected": -2128.0,
+      "loss": 0.0225,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.5625,
+      "rewards/margins": 8.1875,
+      "rewards/rejected": -16.75,
+      "step": 2660
+    },
+    {
+      "epoch": 2.7958115183246073,
+      "grad_norm": 0.08554106919394737,
+      "learning_rate": 7.025359358626166e-07,
+      "logits/chosen": 1.265625,
+      "logits/rejected": 3.265625,
+      "logps/chosen": -1336.0,
+      "logps/rejected": -2176.0,
+      "loss": 0.0205,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.9375,
+      "rewards/margins": 8.125,
+      "rewards/rejected": -17.0,
+      "step": 2670
+    },
+    {
+      "epoch": 2.806282722513089,
+      "grad_norm": 0.4359095098488168,
+      "learning_rate": 6.326262252259463e-07,
+      "logits/chosen": 1.5703125,
+      "logits/rejected": 2.859375,
+      "logps/chosen": -1248.0,
+      "logps/rejected": -2128.0,
+      "loss": 0.0222,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.25,
+      "rewards/margins": 8.5,
+      "rewards/rejected": -16.75,
+      "step": 2680
+    },
+    {
+      "epoch": 2.816753926701571,
+      "grad_norm": 0.8703400315218335,
+      "learning_rate": 5.66335087018871e-07,
+      "logits/chosen": 1.09375,
+      "logits/rejected": 2.71875,
+      "logps/chosen": -1240.0,
+      "logps/rejected": -2080.0,
+      "loss": 0.0119,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -8.0625,
+      "rewards/margins": 8.3125,
+      "rewards/rejected": -16.375,
+      "step": 2690
+    },
+    {
+      "epoch": 2.8272251308900525,
+      "grad_norm": 0.39448533255467044,
+      "learning_rate": 5.036723655228226e-07,
+      "logits/chosen": 1.6015625,
+      "logits/rejected": 2.609375,
+      "logps/chosen": -1200.0,
+      "logps/rejected": -2040.0,
+      "loss": 0.0232,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.25,
+      "rewards/margins": 7.90625,
+      "rewards/rejected": -16.125,
+      "step": 2700
+    },
+    {
+      "epoch": 2.8272251308900525,
+      "eval_logits/chosen": 5.0625,
+      "eval_logits/rejected": 7.0625,
+      "eval_logps/chosen": -1496.0,
+      "eval_logps/rejected": -1592.0,
+      "eval_loss": 1.5903593301773071,
+      "eval_rewards/accuracies": 0.587890625,
+      "eval_rewards/chosen": -11.75,
+      "eval_rewards/margins": 1.2890625,
+      "eval_rewards/rejected": -13.0625,
+      "eval_runtime": 46.6353,
+      "eval_samples_per_second": 42.886,
+      "eval_steps_per_second": 0.686,
+      "step": 2700
+    },
+    {
+      "epoch": 2.837696335078534,
+      "grad_norm": 0.38044176677798924,
+      "learning_rate": 4.446473661967432e-07,
+      "logits/chosen": 1.6640625,
+      "logits/rejected": 3.578125,
+      "logps/chosen": -1232.0,
+      "logps/rejected": -2144.0,
+      "loss": 0.0183,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -8.25,
+      "rewards/margins": 8.625,
+      "rewards/rejected": -16.875,
+      "step": 2710
+    },
+    {
+      "epoch": 2.8481675392670156,
+      "grad_norm": 1.18983015793245,
+      "learning_rate": 3.892688542952411e-07,
+      "logits/chosen": 2.171875,
+      "logits/rejected": 3.21875,
+      "logps/chosen": -1208.0,
+      "logps/rejected": -2040.0,
+      "loss": 0.0227,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -8.125,
+      "rewards/margins": 7.71875,
+      "rewards/rejected": -15.875,
+      "step": 2720
+    },
+    {
+      "epoch": 2.858638743455497,
+      "grad_norm": 1.093710028192038,
+      "learning_rate": 3.3754505356694e-07,
+      "logits/chosen": 1.2265625,
+      "logits/rejected": 2.453125,
+      "logps/chosen": -1280.0,
+      "logps/rejected": -2176.0,
+      "loss": 0.0196,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -8.5625,
+      "rewards/margins": 8.5625,
+      "rewards/rejected": -17.125,
+      "step": 2730
+    },
+    {
+      "epoch": 2.869109947643979,
+      "grad_norm": 0.10044081104758848,
+      "learning_rate": 2.8948364503322276e-07,
+      "logits/chosen": 1.671875,
+      "logits/rejected": 3.546875,
+      "logps/chosen": -1208.0,
+      "logps/rejected": -2064.0,
+      "loss": 0.0228,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -8.125,
+      "rewards/margins": 8.3125,
+      "rewards/rejected": -16.5,
+      "step": 2740
+    },
+    {
+      "epoch": 2.8795811518324608,
+      "grad_norm": 0.2912801732088769,
+      "learning_rate": 2.4509176584762706e-07,
+      "logits/chosen": 1.8828125,
+      "logits/rejected": 3.609375,
+      "logps/chosen": -1240.0,
+      "logps/rejected": -2000.0,
+      "loss": 0.0174,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.3125,
+      "rewards/margins": 7.28125,
+      "rewards/rejected": -15.625,
+      "step": 2750
+    },
+    {
+      "epoch": 2.8900523560209423,
+      "grad_norm": 0.4821780929857923,
+      "learning_rate": 2.0437600823595694e-07,
+      "logits/chosen": 0.81640625,
+      "logits/rejected": 2.90625,
+      "logps/chosen": -1248.0,
+      "logps/rejected": -2064.0,
+      "loss": 0.0222,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.0625,
+      "rewards/margins": 7.90625,
+      "rewards/rejected": -16.0,
+      "step": 2760
+    },
+    {
+      "epoch": 2.900523560209424,
+      "grad_norm": 1.1085018152254948,
+      "learning_rate": 1.6734241851733e-07,
+      "logits/chosen": 1.390625,
+      "logits/rejected": 2.375,
+      "logps/chosen": -1352.0,
+      "logps/rejected": -2240.0,
+      "loss": 0.019,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -9.0625,
+      "rewards/margins": 8.1875,
+      "rewards/rejected": -17.25,
+      "step": 2770
+    },
+    {
+      "epoch": 2.9109947643979055,
+      "grad_norm": 0.9808201932014837,
+      "learning_rate": 1.339964962062984e-07,
+      "logits/chosen": 1.046875,
+      "logits/rejected": 1.9375,
+      "logps/chosen": -1208.0,
+      "logps/rejected": -2112.0,
+      "loss": 0.02,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.3125,
+      "rewards/margins": 8.3125,
+      "rewards/rejected": -16.625,
+      "step": 2780
+    },
+    {
+      "epoch": 2.9214659685863875,
+      "grad_norm": 0.4349330671357158,
+      "learning_rate": 1.0434319319617136e-07,
+      "logits/chosen": 1.4453125,
+      "logits/rejected": 3.296875,
+      "logps/chosen": -1232.0,
+      "logps/rejected": -2080.0,
+      "loss": 0.0227,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -8.3125,
+      "rewards/margins": 8.125,
+      "rewards/rejected": -16.5,
+      "step": 2790
+    },
+    {
+      "epoch": 2.931937172774869,
+      "grad_norm": 1.0937102902775007,
+      "learning_rate": 7.838691302365363e-08,
+      "logits/chosen": 1.375,
+      "logits/rejected": 2.953125,
+      "logps/chosen": -1232.0,
+      "logps/rejected": -2112.0,
+      "loss": 0.0165,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -8.25,
+      "rewards/margins": 8.3125,
+      "rewards/rejected": -16.5,
+      "step": 2800
+    },
+    {
+      "epoch": 2.931937172774869,
+      "eval_logits/chosen": 5.0,
+      "eval_logits/rejected": 7.0,
+      "eval_logps/chosen": -1496.0,
+      "eval_logps/rejected": -1592.0,
+      "eval_loss": 1.5947344303131104,
+      "eval_rewards/accuracies": 0.5859375,
+      "eval_rewards/chosen": -11.75,
+      "eval_rewards/margins": 1.3046875,
+      "eval_rewards/rejected": -13.0625,
+      "eval_runtime": 46.6308,
+      "eval_samples_per_second": 42.89,
+      "eval_steps_per_second": 0.686,
+      "step": 2800
+    },
+    {
+      "epoch": 2.9424083769633507,
+      "grad_norm": 0.20609942690228372,
+      "learning_rate": 5.613151021490459e-08,
+      "logits/chosen": 2.203125,
+      "logits/rejected": 3.421875,
+      "logps/chosen": -1144.0,
+      "logps/rejected": -1968.0,
+      "loss": 0.0203,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.0,
+      "rewards/margins": 7.75,
+      "rewards/rejected": -15.8125,
+      "step": 2810
+    },
+    {
+      "epoch": 2.9528795811518327,
+      "grad_norm": 0.27133829350992933,
+      "learning_rate": 3.7580289713159965e-08,
+      "logits/chosen": 2.34375,
+      "logits/rejected": 3.671875,
+      "logps/chosen": -1296.0,
+      "logps/rejected": -2288.0,
+      "loss": 0.0166,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.6875,
+      "rewards/margins": 8.9375,
+      "rewards/rejected": -17.625,
+      "step": 2820
+    },
+    {
+      "epoch": 2.9633507853403143,
+      "grad_norm": 0.7265520648943169,
+      "learning_rate": 2.2736006387927277e-08,
+      "logits/chosen": 1.53125,
+      "logits/rejected": 3.03125,
+      "logps/chosen": -1280.0,
+      "logps/rejected": -2144.0,
+      "loss": 0.0188,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -8.5625,
+      "rewards/margins": 8.25,
+      "rewards/rejected": -16.75,
+      "step": 2830
+    },
+    {
+      "epoch": 2.973821989528796,
+      "grad_norm": 1.0080298090928475,
+      "learning_rate": 1.1600864625893604e-08,
+      "logits/chosen": 1.046875,
+      "logits/rejected": 2.40625,
+      "logps/chosen": -1248.0,
+      "logps/rejected": -2128.0,
+      "loss": 0.0288,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.3125,
+      "rewards/margins": 8.375,
+      "rewards/rejected": -16.75,
+      "step": 2840
+    },
+    {
+      "epoch": 2.9842931937172774,
+      "grad_norm": 0.21249874778485772,
+      "learning_rate": 4.176518003573549e-09,
+      "logits/chosen": 1.5625,
+      "logits/rejected": 3.453125,
+      "logps/chosen": -1264.0,
+      "logps/rejected": -2112.0,
+      "loss": 0.0302,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.625,
+      "rewards/margins": 7.78125,
+      "rewards/rejected": -16.375,
+      "step": 2850
+    },
+    {
+      "epoch": 2.994764397905759,
+      "grad_norm": 0.6409701646806494,
+      "learning_rate": 4.640690417528903e-10,
+      "logits/chosen": 1.2109375,
+      "logits/rejected": 2.84375,
+      "logps/chosen": -1184.0,
+      "logps/rejected": -2040.0,
+      "loss": 0.0235,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.0,
+      "rewards/margins": 8.125,
+      "rewards/rejected": -16.125,
+      "step": 2860
+    },
+    {
+      "epoch": 3.0,
+      "step": 2865,
+      "total_flos": 0.0,
+      "train_loss": 0.2707289488735831,
+      "train_runtime": 12301.2448,
+      "train_samples_per_second": 14.906,
+      "train_steps_per_second": 0.233
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 2865,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}