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": 1.582245517402759,
+      "learning_rate": 1.7421602787456448e-07,
+      "logits/chosen": -11.375,
+      "logits/rejected": -11.6875,
+      "logps/chosen": -466.0,
+      "logps/rejected": -608.0,
+      "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": 1.5263738363102872,
+      "learning_rate": 1.7421602787456445e-06,
+      "logits/chosen": -11.3125,
+      "logits/rejected": -10.9375,
+      "logps/chosen": -438.0,
+      "logps/rejected": -548.0,
+      "loss": 0.6921,
+      "rewards/accuracies": 0.1458333283662796,
+      "rewards/chosen": 0.00107574462890625,
+      "rewards/margins": -0.000644683837890625,
+      "rewards/rejected": 0.0017242431640625,
+      "step": 10
+    },
+    {
+      "epoch": 0.020942408376963352,
+      "grad_norm": 1.5021681553716235,
+      "learning_rate": 3.484320557491289e-06,
+      "logits/chosen": -11.125,
+      "logits/rejected": -11.0,
+      "logps/chosen": -404.0,
+      "logps/rejected": -456.0,
+      "loss": 0.6904,
+      "rewards/accuracies": 0.29374998807907104,
+      "rewards/chosen": -0.00091552734375,
+      "rewards/margins": 0.005828857421875,
+      "rewards/rejected": -0.006744384765625,
+      "step": 20
+    },
+    {
+      "epoch": 0.031413612565445025,
+      "grad_norm": 1.5002050170675545,
+      "learning_rate": 5.226480836236934e-06,
+      "logits/chosen": -11.5,
+      "logits/rejected": -11.625,
+      "logps/chosen": -422.0,
+      "logps/rejected": -466.0,
+      "loss": 0.6876,
+      "rewards/accuracies": 0.4749999940395355,
+      "rewards/chosen": -0.0206298828125,
+      "rewards/margins": 0.00933837890625,
+      "rewards/rejected": -0.030029296875,
+      "step": 30
+    },
+    {
+      "epoch": 0.041884816753926704,
+      "grad_norm": 1.8590919500425669,
+      "learning_rate": 6.968641114982578e-06,
+      "logits/chosen": -12.1875,
+      "logits/rejected": -11.9375,
+      "logps/chosen": -442.0,
+      "logps/rejected": -536.0,
+      "loss": 0.6716,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": -0.1240234375,
+      "rewards/margins": 0.05615234375,
+      "rewards/rejected": -0.1796875,
+      "step": 40
+    },
+    {
+      "epoch": 0.05235602094240838,
+      "grad_norm": 2.0828530277262693,
+      "learning_rate": 8.710801393728225e-06,
+      "logits/chosen": -12.25,
+      "logits/rejected": -12.1875,
+      "logps/chosen": -450.0,
+      "logps/rejected": -592.0,
+      "loss": 0.6404,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.3203125,
+      "rewards/margins": 0.16796875,
+      "rewards/rejected": -0.48828125,
+      "step": 50
+    },
+    {
+      "epoch": 0.06282722513089005,
+      "grad_norm": 2.150163508377863,
+      "learning_rate": 1.0452961672473868e-05,
+      "logits/chosen": -11.125,
+      "logits/rejected": -10.9375,
+      "logps/chosen": -396.0,
+      "logps/rejected": -644.0,
+      "loss": 0.603,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -0.443359375,
+      "rewards/margins": 0.369140625,
+      "rewards/rejected": -0.8125,
+      "step": 60
+    },
+    {
+      "epoch": 0.07329842931937172,
+      "grad_norm": 2.259195204037021,
+      "learning_rate": 1.2195121951219513e-05,
+      "logits/chosen": -11.5,
+      "logits/rejected": -11.3125,
+      "logps/chosen": -516.0,
+      "logps/rejected": -628.0,
+      "loss": 0.6005,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -0.81640625,
+      "rewards/margins": 0.279296875,
+      "rewards/rejected": -1.09375,
+      "step": 70
+    },
+    {
+      "epoch": 0.08376963350785341,
+      "grad_norm": 2.0904376715861623,
+      "learning_rate": 1.3937282229965156e-05,
+      "logits/chosen": -11.875,
+      "logits/rejected": -11.875,
+      "logps/chosen": -490.0,
+      "logps/rejected": -716.0,
+      "loss": 0.5793,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.95703125,
+      "rewards/margins": 0.75390625,
+      "rewards/rejected": -1.7109375,
+      "step": 80
+    },
+    {
+      "epoch": 0.09424083769633508,
+      "grad_norm": 2.2647507456401836,
+      "learning_rate": 1.56794425087108e-05,
+      "logits/chosen": -11.9375,
+      "logits/rejected": -11.9375,
+      "logps/chosen": -496.0,
+      "logps/rejected": -636.0,
+      "loss": 0.565,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -1.3359375,
+      "rewards/margins": 0.53125,
+      "rewards/rejected": -1.8671875,
+      "step": 90
+    },
+    {
+      "epoch": 0.10471204188481675,
+      "grad_norm": 3.2515110482990646,
+      "learning_rate": 1.742160278745645e-05,
+      "logits/chosen": -12.375,
+      "logits/rejected": -12.25,
+      "logps/chosen": -640.0,
+      "logps/rejected": -788.0,
+      "loss": 0.5942,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -1.65625,
+      "rewards/margins": 0.63671875,
+      "rewards/rejected": -2.296875,
+      "step": 100
+    },
+    {
+      "epoch": 0.10471204188481675,
+      "eval_logits/chosen": -11.4375,
+      "eval_logits/rejected": -11.125,
+      "eval_logps/chosen": -454.0,
+      "eval_logps/rejected": -446.0,
+      "eval_loss": 0.677984356880188,
+      "eval_rewards/accuracies": 0.60546875,
+      "eval_rewards/chosen": -1.359375,
+      "eval_rewards/margins": 0.205078125,
+      "eval_rewards/rejected": -1.5625,
+      "eval_runtime": 54.7843,
+      "eval_samples_per_second": 36.507,
+      "eval_steps_per_second": 0.584,
+      "step": 100
+    },
+    {
+      "epoch": 0.11518324607329843,
+      "grad_norm": 3.044731501422064,
+      "learning_rate": 1.9163763066202093e-05,
+      "logits/chosen": -11.0,
+      "logits/rejected": -10.9375,
+      "logps/chosen": -552.0,
+      "logps/rejected": -736.0,
+      "loss": 0.566,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -1.3125,
+      "rewards/margins": 0.5859375,
+      "rewards/rejected": -1.8984375,
+      "step": 110
+    },
+    {
+      "epoch": 0.1256544502617801,
+      "grad_norm": 3.3312722337425034,
+      "learning_rate": 2.0905923344947736e-05,
+      "logits/chosen": -11.5,
+      "logits/rejected": -11.375,
+      "logps/chosen": -528.0,
+      "logps/rejected": -680.0,
+      "loss": 0.5536,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": -1.578125,
+      "rewards/margins": 0.6328125,
+      "rewards/rejected": -2.21875,
+      "step": 120
+    },
+    {
+      "epoch": 0.13612565445026178,
+      "grad_norm": 2.321069691360544,
+      "learning_rate": 2.264808362369338e-05,
+      "logits/chosen": -11.875,
+      "logits/rejected": -11.75,
+      "logps/chosen": -600.0,
+      "logps/rejected": -736.0,
+      "loss": 0.6059,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -1.8046875,
+      "rewards/margins": 0.474609375,
+      "rewards/rejected": -2.28125,
+      "step": 130
+    },
+    {
+      "epoch": 0.14659685863874344,
+      "grad_norm": 3.07675020541296,
+      "learning_rate": 2.4390243902439026e-05,
+      "logits/chosen": -13.5,
+      "logits/rejected": -13.0,
+      "logps/chosen": -576.0,
+      "logps/rejected": -800.0,
+      "loss": 0.5517,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -1.296875,
+      "rewards/margins": 0.58203125,
+      "rewards/rejected": -1.875,
+      "step": 140
+    },
+    {
+      "epoch": 0.15706806282722513,
+      "grad_norm": 2.6654975683704993,
+      "learning_rate": 2.6132404181184672e-05,
+      "logits/chosen": -12.375,
+      "logits/rejected": -12.25,
+      "logps/chosen": -592.0,
+      "logps/rejected": -772.0,
+      "loss": 0.5593,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -1.625,
+      "rewards/margins": 0.71484375,
+      "rewards/rejected": -2.34375,
+      "step": 150
+    },
+    {
+      "epoch": 0.16753926701570682,
+      "grad_norm": 2.40614357482195,
+      "learning_rate": 2.7874564459930312e-05,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -11.1875,
+      "logps/chosen": -552.0,
+      "logps/rejected": -728.0,
+      "loss": 0.5387,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -1.5546875,
+      "rewards/margins": 0.66015625,
+      "rewards/rejected": -2.21875,
+      "step": 160
+    },
+    {
+      "epoch": 0.17801047120418848,
+      "grad_norm": 2.4356542812615305,
+      "learning_rate": 2.9616724738675962e-05,
+      "logits/chosen": -8.375,
+      "logits/rejected": -8.4375,
+      "logps/chosen": -680.0,
+      "logps/rejected": -876.0,
+      "loss": 0.5229,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -2.59375,
+      "rewards/margins": 0.83203125,
+      "rewards/rejected": -3.421875,
+      "step": 170
+    },
+    {
+      "epoch": 0.18848167539267016,
+      "grad_norm": 1.9778410054097122,
+      "learning_rate": 3.13588850174216e-05,
+      "logits/chosen": -8.5,
+      "logits/rejected": -8.3125,
+      "logps/chosen": -760.0,
+      "logps/rejected": -900.0,
+      "loss": 0.5352,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -3.15625,
+      "rewards/margins": 0.8359375,
+      "rewards/rejected": -4.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.19895287958115182,
+      "grad_norm": 2.2609862761572064,
+      "learning_rate": 3.310104529616725e-05,
+      "logits/chosen": -10.75,
+      "logits/rejected": -10.75,
+      "logps/chosen": -660.0,
+      "logps/rejected": -864.0,
+      "loss": 0.5454,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -2.671875,
+      "rewards/margins": 0.72265625,
+      "rewards/rejected": -3.390625,
+      "step": 190
+    },
+    {
+      "epoch": 0.2094240837696335,
+      "grad_norm": 2.3177692878659264,
+      "learning_rate": 3.48432055749129e-05,
+      "logits/chosen": -12.25,
+      "logits/rejected": -12.125,
+      "logps/chosen": -660.0,
+      "logps/rejected": -820.0,
+      "loss": 0.561,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -1.953125,
+      "rewards/margins": 0.68359375,
+      "rewards/rejected": -2.640625,
+      "step": 200
+    },
+    {
+      "epoch": 0.2094240837696335,
+      "eval_logits/chosen": -10.25,
+      "eval_logits/rejected": -9.75,
+      "eval_logps/chosen": -540.0,
+      "eval_logps/rejected": -540.0,
+      "eval_loss": 0.6944218873977661,
+      "eval_rewards/accuracies": 0.60546875,
+      "eval_rewards/chosen": -2.21875,
+      "eval_rewards/margins": 0.294921875,
+      "eval_rewards/rejected": -2.515625,
+      "eval_runtime": 46.6306,
+      "eval_samples_per_second": 42.89,
+      "eval_steps_per_second": 0.686,
+      "step": 200
+    },
+    {
+      "epoch": 0.2198952879581152,
+      "grad_norm": 2.9921467716825108,
+      "learning_rate": 3.6585365853658535e-05,
+      "logits/chosen": -10.875,
+      "logits/rejected": -10.8125,
+      "logps/chosen": -624.0,
+      "logps/rejected": -748.0,
+      "loss": 0.5526,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -2.5625,
+      "rewards/margins": 0.56640625,
+      "rewards/rejected": -3.140625,
+      "step": 210
+    },
+    {
+      "epoch": 0.23036649214659685,
+      "grad_norm": 2.534227548628948,
+      "learning_rate": 3.8327526132404185e-05,
+      "logits/chosen": -13.1875,
+      "logits/rejected": -12.75,
+      "logps/chosen": -636.0,
+      "logps/rejected": -776.0,
+      "loss": 0.5526,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -2.15625,
+      "rewards/margins": 0.76171875,
+      "rewards/rejected": -2.921875,
+      "step": 220
+    },
+    {
+      "epoch": 0.24083769633507854,
+      "grad_norm": 2.538407636798547,
+      "learning_rate": 4.006968641114983e-05,
+      "logits/chosen": -12.3125,
+      "logits/rejected": -12.125,
+      "logps/chosen": -696.0,
+      "logps/rejected": -856.0,
+      "loss": 0.5366,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -3.078125,
+      "rewards/margins": 0.6875,
+      "rewards/rejected": -3.765625,
+      "step": 230
+    },
+    {
+      "epoch": 0.2513089005235602,
+      "grad_norm": 2.7177700431099385,
+      "learning_rate": 4.181184668989547e-05,
+      "logits/chosen": -11.9375,
+      "logits/rejected": -11.6875,
+      "logps/chosen": -720.0,
+      "logps/rejected": -872.0,
+      "loss": 0.5854,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -3.453125,
+      "rewards/margins": 0.7890625,
+      "rewards/rejected": -4.25,
+      "step": 240
+    },
+    {
+      "epoch": 0.2617801047120419,
+      "grad_norm": 3.3750564834129504,
+      "learning_rate": 4.3554006968641115e-05,
+      "logits/chosen": -12.375,
+      "logits/rejected": -12.0625,
+      "logps/chosen": -704.0,
+      "logps/rejected": -868.0,
+      "loss": 0.5305,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -2.484375,
+      "rewards/margins": 0.85546875,
+      "rewards/rejected": -3.34375,
+      "step": 250
+    },
+    {
+      "epoch": 0.27225130890052357,
+      "grad_norm": 2.3187384311598565,
+      "learning_rate": 4.529616724738676e-05,
+      "logits/chosen": -10.4375,
+      "logits/rejected": -10.125,
+      "logps/chosen": -768.0,
+      "logps/rejected": -1020.0,
+      "loss": 0.5257,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": -3.90625,
+      "rewards/margins": 1.1953125,
+      "rewards/rejected": -5.09375,
+      "step": 260
+    },
+    {
+      "epoch": 0.28272251308900526,
+      "grad_norm": 1.959143521875051,
+      "learning_rate": 4.703832752613241e-05,
+      "logits/chosen": -11.75,
+      "logits/rejected": -11.5625,
+      "logps/chosen": -732.0,
+      "logps/rejected": -888.0,
+      "loss": 0.548,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -3.515625,
+      "rewards/margins": 0.83203125,
+      "rewards/rejected": -4.34375,
+      "step": 270
+    },
+    {
+      "epoch": 0.2931937172774869,
+      "grad_norm": 2.548392657608926,
+      "learning_rate": 4.878048780487805e-05,
+      "logits/chosen": -12.0,
+      "logits/rejected": -11.625,
+      "logps/chosen": -632.0,
+      "logps/rejected": -800.0,
+      "loss": 0.5448,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -2.640625,
+      "rewards/margins": 0.69921875,
+      "rewards/rejected": -3.328125,
+      "step": 280
+    },
+    {
+      "epoch": 0.3036649214659686,
+      "grad_norm": 2.620820857416364,
+      "learning_rate": 4.999983293481418e-05,
+      "logits/chosen": -14.375,
+      "logits/rejected": -14.25,
+      "logps/chosen": -708.0,
+      "logps/rejected": -920.0,
+      "loss": 0.5595,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -3.125,
+      "rewards/margins": 0.9140625,
+      "rewards/rejected": -4.03125,
+      "step": 290
+    },
+    {
+      "epoch": 0.31413612565445026,
+      "grad_norm": 2.4811851838998034,
+      "learning_rate": 4.9996862949181595e-05,
+      "logits/chosen": -9.875,
+      "logits/rejected": -9.5625,
+      "logps/chosen": -840.0,
+      "logps/rejected": -992.0,
+      "loss": 0.5506,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -4.28125,
+      "rewards/margins": 0.6953125,
+      "rewards/rejected": -5.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.31413612565445026,
+      "eval_logits/chosen": -8.3125,
+      "eval_logits/rejected": -7.5,
+      "eval_logps/chosen": -764.0,
+      "eval_logps/rejected": -772.0,
+      "eval_loss": 0.7181406021118164,
+      "eval_rewards/accuracies": 0.607421875,
+      "eval_rewards/chosen": -4.4375,
+      "eval_rewards/margins": 0.400390625,
+      "eval_rewards/rejected": -4.84375,
+      "eval_runtime": 46.6454,
+      "eval_samples_per_second": 42.877,
+      "eval_steps_per_second": 0.686,
+      "step": 300
+    },
+    {
+      "epoch": 0.32460732984293195,
+      "grad_norm": 2.728167916044458,
+      "learning_rate": 4.9990180911526896e-05,
+      "logits/chosen": -10.5,
+      "logits/rejected": -10.0625,
+      "logps/chosen": -680.0,
+      "logps/rejected": -844.0,
+      "loss": 0.5617,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -3.046875,
+      "rewards/margins": 0.53515625,
+      "rewards/rejected": -3.59375,
+      "step": 310
+    },
+    {
+      "epoch": 0.33507853403141363,
+      "grad_norm": 3.0150065333219325,
+      "learning_rate": 4.997978781413745e-05,
+      "logits/chosen": -9.3125,
+      "logits/rejected": -9.125,
+      "logps/chosen": -888.0,
+      "logps/rejected": -1064.0,
+      "loss": 0.5592,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -4.4375,
+      "rewards/margins": 1.09375,
+      "rewards/rejected": -5.53125,
+      "step": 320
+    },
+    {
+      "epoch": 0.34554973821989526,
+      "grad_norm": 2.5804615362734467,
+      "learning_rate": 4.996568520039565e-05,
+      "logits/chosen": -10.1875,
+      "logits/rejected": -10.125,
+      "logps/chosen": -740.0,
+      "logps/rejected": -948.0,
+      "loss": 0.5393,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -3.234375,
+      "rewards/margins": 0.94140625,
+      "rewards/rejected": -4.1875,
+      "step": 330
+    },
+    {
+      "epoch": 0.35602094240837695,
+      "grad_norm": 2.9949010065088353,
+      "learning_rate": 4.994787516454973e-05,
+      "logits/chosen": -11.0,
+      "logits/rejected": -10.5,
+      "logps/chosen": -716.0,
+      "logps/rejected": -872.0,
+      "loss": 0.525,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -3.125,
+      "rewards/margins": 0.87890625,
+      "rewards/rejected": -4.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.36649214659685864,
+      "grad_norm": 3.6024122625196804,
+      "learning_rate": 4.992636035140273e-05,
+      "logits/chosen": -10.875,
+      "logits/rejected": -10.5625,
+      "logps/chosen": -700.0,
+      "logps/rejected": -928.0,
+      "loss": 0.534,
+      "rewards/accuracies": 0.78125,
+      "rewards/chosen": -3.28125,
+      "rewards/margins": 1.0,
+      "rewards/rejected": -4.28125,
+      "step": 350
+    },
+    {
+      "epoch": 0.3769633507853403,
+      "grad_norm": 2.650227753877416,
+      "learning_rate": 4.99011439559198e-05,
+      "logits/chosen": -12.1875,
+      "logits/rejected": -11.625,
+      "logps/chosen": -780.0,
+      "logps/rejected": -1032.0,
+      "loss": 0.5353,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": -3.71875,
+      "rewards/margins": 1.1640625,
+      "rewards/rejected": -4.875,
+      "step": 360
+    },
+    {
+      "epoch": 0.387434554973822,
+      "grad_norm": 2.4010085612350345,
+      "learning_rate": 4.987222972275369e-05,
+      "logits/chosen": -8.5625,
+      "logits/rejected": -8.0625,
+      "logps/chosen": -908.0,
+      "logps/rejected": -1096.0,
+      "loss": 0.5455,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": -4.96875,
+      "rewards/margins": 1.03125,
+      "rewards/rejected": -6.0,
+      "step": 370
+    },
+    {
+      "epoch": 0.39790575916230364,
+      "grad_norm": 2.034008157585721,
+      "learning_rate": 4.983962194568872e-05,
+      "logits/chosen": -9.9375,
+      "logits/rejected": -9.5,
+      "logps/chosen": -864.0,
+      "logps/rejected": -1128.0,
+      "loss": 0.5344,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -4.4375,
+      "rewards/margins": 1.078125,
+      "rewards/rejected": -5.53125,
+      "step": 380
+    },
+    {
+      "epoch": 0.4083769633507853,
+      "grad_norm": 2.4174156914986704,
+      "learning_rate": 4.9803325467003084e-05,
+      "logits/chosen": -10.875,
+      "logits/rejected": -10.3125,
+      "logps/chosen": -832.0,
+      "logps/rejected": -1088.0,
+      "loss": 0.5098,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -4.03125,
+      "rewards/margins": 1.171875,
+      "rewards/rejected": -5.21875,
+      "step": 390
+    },
+    {
+      "epoch": 0.418848167539267,
+      "grad_norm": 3.3476915694916918,
+      "learning_rate": 4.9763345676749816e-05,
+      "logits/chosen": -10.25,
+      "logits/rejected": -9.875,
+      "logps/chosen": -928.0,
+      "logps/rejected": -1208.0,
+      "loss": 0.5233,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -5.25,
+      "rewards/margins": 1.4453125,
+      "rewards/rejected": -6.6875,
+      "step": 400
+    },
+    {
+      "epoch": 0.418848167539267,
+      "eval_logits/chosen": -13.625,
+      "eval_logits/rejected": -12.875,
+      "eval_logps/chosen": -720.0,
+      "eval_logps/rejected": -736.0,
+      "eval_loss": 0.8159219026565552,
+      "eval_rewards/accuracies": 0.583984375,
+      "eval_rewards/chosen": -4.03125,
+      "eval_rewards/margins": 0.466796875,
+      "eval_rewards/rejected": -4.5,
+      "eval_runtime": 46.6586,
+      "eval_samples_per_second": 42.865,
+      "eval_steps_per_second": 0.686,
+      "step": 400
+    },
+    {
+      "epoch": 0.4293193717277487,
+      "grad_norm": 2.586399973896129,
+      "learning_rate": 4.9719688511956374e-05,
+      "logits/chosen": -13.9375,
+      "logits/rejected": -13.375,
+      "logps/chosen": -816.0,
+      "logps/rejected": -1088.0,
+      "loss": 0.5349,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": -3.921875,
+      "rewards/margins": 1.3203125,
+      "rewards/rejected": -5.25,
+      "step": 410
+    },
+    {
+      "epoch": 0.4397905759162304,
+      "grad_norm": 2.9277764977659055,
+      "learning_rate": 4.967236045574294e-05,
+      "logits/chosen": -10.6875,
+      "logits/rejected": -10.0,
+      "logps/chosen": -888.0,
+      "logps/rejected": -1104.0,
+      "loss": 0.5166,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -4.75,
+      "rewards/margins": 1.1875,
+      "rewards/rejected": -5.9375,
+      "step": 420
+    },
+    {
+      "epoch": 0.450261780104712,
+      "grad_norm": 2.1729239398577644,
+      "learning_rate": 4.96213685363597e-05,
+      "logits/chosen": -9.8125,
+      "logits/rejected": -9.4375,
+      "logps/chosen": -856.0,
+      "logps/rejected": -1016.0,
+      "loss": 0.5763,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -4.53125,
+      "rewards/margins": 0.86328125,
+      "rewards/rejected": -5.40625,
+      "step": 430
+    },
+    {
+      "epoch": 0.4607329842931937,
+      "grad_norm": 2.7623880256483684,
+      "learning_rate": 4.956672032614315e-05,
+      "logits/chosen": -12.0,
+      "logits/rejected": -11.625,
+      "logps/chosen": -760.0,
+      "logps/rejected": -936.0,
+      "loss": 0.5593,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -3.4375,
+      "rewards/margins": 0.7578125,
+      "rewards/rejected": -4.1875,
+      "step": 440
+    },
+    {
+      "epoch": 0.4712041884816754,
+      "grad_norm": 2.8537544022057633,
+      "learning_rate": 4.950842394039157e-05,
+      "logits/chosen": -11.8125,
+      "logits/rejected": -11.3125,
+      "logps/chosen": -904.0,
+      "logps/rejected": -1120.0,
+      "loss": 0.5358,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -5.1875,
+      "rewards/margins": 1.171875,
+      "rewards/rejected": -6.34375,
+      "step": 450
+    },
+    {
+      "epoch": 0.4816753926701571,
+      "grad_norm": 2.491837555416849,
+      "learning_rate": 4.944648803615994e-05,
+      "logits/chosen": -13.8125,
+      "logits/rejected": -13.1875,
+      "logps/chosen": -820.0,
+      "logps/rejected": -1016.0,
+      "loss": 0.5427,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -4.03125,
+      "rewards/margins": 1.1796875,
+      "rewards/rejected": -5.21875,
+      "step": 460
+    },
+    {
+      "epoch": 0.49214659685863876,
+      "grad_norm": 3.1316085810060597,
+      "learning_rate": 4.938092181097433e-05,
+      "logits/chosen": -14.25,
+      "logits/rejected": -13.9375,
+      "logps/chosen": -800.0,
+      "logps/rejected": -976.0,
+      "loss": 0.4985,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": -3.984375,
+      "rewards/margins": 1.0,
+      "rewards/rejected": -5.0,
+      "step": 470
+    },
+    {
+      "epoch": 0.5026178010471204,
+      "grad_norm": 4.236106193310983,
+      "learning_rate": 4.931173500146607e-05,
+      "logits/chosen": -13.25,
+      "logits/rejected": -12.75,
+      "logps/chosen": -868.0,
+      "logps/rejected": -1144.0,
+      "loss": 0.5287,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -4.8125,
+      "rewards/margins": 1.4765625,
+      "rewards/rejected": -6.28125,
+      "step": 480
+    },
+    {
+      "epoch": 0.5130890052356021,
+      "grad_norm": 2.644398046825569,
+      "learning_rate": 4.923893788192582e-05,
+      "logits/chosen": -13.75,
+      "logits/rejected": -13.25,
+      "logps/chosen": -760.0,
+      "logps/rejected": -928.0,
+      "loss": 0.5147,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -3.375,
+      "rewards/margins": 1.015625,
+      "rewards/rejected": -4.40625,
+      "step": 490
+    },
+    {
+      "epoch": 0.5235602094240838,
+      "grad_norm": 4.059056197201719,
+      "learning_rate": 4.9162541262777906e-05,
+      "logits/chosen": -11.25,
+      "logits/rejected": -10.5625,
+      "logps/chosen": -780.0,
+      "logps/rejected": -984.0,
+      "loss": 0.5164,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -3.578125,
+      "rewards/margins": 0.9609375,
+      "rewards/rejected": -4.53125,
+      "step": 500
+    },
+    {
+      "epoch": 0.5235602094240838,
+      "eval_logits/chosen": -8.3125,
+      "eval_logits/rejected": -6.65625,
+      "eval_logps/chosen": -636.0,
+      "eval_logps/rejected": -672.0,
+      "eval_loss": 0.7724218964576721,
+      "eval_rewards/accuracies": 0.626953125,
+      "eval_rewards/chosen": -3.15625,
+      "eval_rewards/margins": 0.66015625,
+      "eval_rewards/rejected": -3.828125,
+      "eval_runtime": 46.6427,
+      "eval_samples_per_second": 42.879,
+      "eval_steps_per_second": 0.686,
+      "step": 500
+    },
+    {
+      "epoch": 0.5340314136125655,
+      "grad_norm": 2.662031273800223,
+      "learning_rate": 4.90825564889749e-05,
+      "logits/chosen": -11.3125,
+      "logits/rejected": -10.6875,
+      "logps/chosen": -712.0,
+      "logps/rejected": -888.0,
+      "loss": 0.5222,
+      "rewards/accuracies": 0.78125,
+      "rewards/chosen": -2.75,
+      "rewards/margins": 0.9765625,
+      "rewards/rejected": -3.71875,
+      "step": 510
+    },
+    {
+      "epoch": 0.5445026178010471,
+      "grad_norm": 3.226497544027977,
+      "learning_rate": 4.899899543831288e-05,
+      "logits/chosen": -9.875,
+      "logits/rejected": -8.875,
+      "logps/chosen": -804.0,
+      "logps/rejected": -1048.0,
+      "loss": 0.4952,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -3.609375,
+      "rewards/margins": 1.46875,
+      "rewards/rejected": -5.0625,
+      "step": 520
+    },
+    {
+      "epoch": 0.5549738219895288,
+      "grad_norm": 3.154187459868018,
+      "learning_rate": 4.891187051966765e-05,
+      "logits/chosen": -9.125,
+      "logits/rejected": -8.5625,
+      "logps/chosen": -812.0,
+      "logps/rejected": -940.0,
+      "loss": 0.5335,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -3.828125,
+      "rewards/margins": 0.8125,
+      "rewards/rejected": -4.625,
+      "step": 530
+    },
+    {
+      "epoch": 0.5654450261780105,
+      "grad_norm": 3.170157111167069,
+      "learning_rate": 4.882119467115191e-05,
+      "logits/chosen": -10.0,
+      "logits/rejected": -9.625,
+      "logps/chosen": -748.0,
+      "logps/rejected": -972.0,
+      "loss": 0.5404,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -3.453125,
+      "rewards/margins": 1.0078125,
+      "rewards/rejected": -4.46875,
+      "step": 540
+    },
+    {
+      "epoch": 0.5759162303664922,
+      "grad_norm": 2.634840823127897,
+      "learning_rate": 4.872698135819402e-05,
+      "logits/chosen": -10.125,
+      "logits/rejected": -9.5625,
+      "logps/chosen": -812.0,
+      "logps/rejected": -1048.0,
+      "loss": 0.5301,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -3.96875,
+      "rewards/margins": 1.03125,
+      "rewards/rejected": -5.0,
+      "step": 550
+    },
+    {
+      "epoch": 0.5863874345549738,
+      "grad_norm": 2.8169975227892716,
+      "learning_rate": 4.862924457153831e-05,
+      "logits/chosen": -9.0625,
+      "logits/rejected": -8.625,
+      "logps/chosen": -784.0,
+      "logps/rejected": -1032.0,
+      "loss": 0.4821,
+      "rewards/accuracies": 0.768750011920929,
+      "rewards/chosen": -4.40625,
+      "rewards/margins": 1.1796875,
+      "rewards/rejected": -5.59375,
+      "step": 560
+    },
+    {
+      "epoch": 0.5968586387434555,
+      "grad_norm": 2.404754553912183,
+      "learning_rate": 4.8527998825167484e-05,
+      "logits/chosen": -7.8125,
+      "logits/rejected": -7.6875,
+      "logps/chosen": -924.0,
+      "logps/rejected": -1144.0,
+      "loss": 0.4674,
+      "rewards/accuracies": 0.768750011920929,
+      "rewards/chosen": -5.375,
+      "rewards/margins": 1.421875,
+      "rewards/rejected": -6.8125,
+      "step": 570
+    },
+    {
+      "epoch": 0.6073298429319371,
+      "grad_norm": 2.496733491642304,
+      "learning_rate": 4.842325915414728e-05,
+      "logits/chosen": -10.0,
+      "logits/rejected": -9.375,
+      "logps/chosen": -832.0,
+      "logps/rejected": -1112.0,
+      "loss": 0.5142,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": -4.5625,
+      "rewards/margins": 1.6484375,
+      "rewards/rejected": -6.21875,
+      "step": 580
+    },
+    {
+      "epoch": 0.6178010471204188,
+      "grad_norm": 3.282918671224393,
+      "learning_rate": 4.831504111239373e-05,
+      "logits/chosen": -10.625,
+      "logits/rejected": -9.75,
+      "logps/chosen": -908.0,
+      "logps/rejected": -1208.0,
+      "loss": 0.4821,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": -4.875,
+      "rewards/margins": 1.75,
+      "rewards/rejected": -6.625,
+      "step": 590
+    },
+    {
+      "epoch": 0.6282722513089005,
+      "grad_norm": 2.4131390450477173,
+      "learning_rate": 4.8203360770363425e-05,
+      "logits/chosen": -7.65625,
+      "logits/rejected": -7.25,
+      "logps/chosen": -1056.0,
+      "logps/rejected": -1320.0,
+      "loss": 0.5142,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -6.34375,
+      "rewards/margins": 1.7734375,
+      "rewards/rejected": -8.125,
+      "step": 600
+    },
+    {
+      "epoch": 0.6282722513089005,
+      "eval_logits/chosen": -7.40625,
+      "eval_logits/rejected": -6.4375,
+      "eval_logps/chosen": -852.0,
+      "eval_logps/rejected": -860.0,
+      "eval_loss": 0.7633437514305115,
+      "eval_rewards/accuracies": 0.564453125,
+      "eval_rewards/chosen": -5.34375,
+      "eval_rewards/margins": 0.376953125,
+      "eval_rewards/rejected": -5.71875,
+      "eval_runtime": 46.6524,
+      "eval_samples_per_second": 42.87,
+      "eval_steps_per_second": 0.686,
+      "step": 600
+    },
+    {
+      "epoch": 0.6387434554973822,
+      "grad_norm": 2.8929559033137595,
+      "learning_rate": 4.808823471266702e-05,
+      "logits/chosen": -9.4375,
+      "logits/rejected": -9.25,
+      "logps/chosen": -864.0,
+      "logps/rejected": -1080.0,
+      "loss": 0.5153,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -4.59375,
+      "rewards/margins": 1.171875,
+      "rewards/rejected": -5.78125,
+      "step": 610
+    },
+    {
+      "epoch": 0.6492146596858639,
+      "grad_norm": 2.4910936225080227,
+      "learning_rate": 4.796968003560639e-05,
+      "logits/chosen": -10.875,
+      "logits/rejected": -10.375,
+      "logps/chosen": -864.0,
+      "logps/rejected": -1184.0,
+      "loss": 0.5141,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": -4.5625,
+      "rewards/margins": 1.71875,
+      "rewards/rejected": -6.28125,
+      "step": 620
+    },
+    {
+      "epoch": 0.6596858638743456,
+      "grad_norm": 2.2899945014197467,
+      "learning_rate": 4.7847714344635864e-05,
+      "logits/chosen": -11.5625,
+      "logits/rejected": -11.0,
+      "logps/chosen": -900.0,
+      "logps/rejected": -1088.0,
+      "loss": 0.5145,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -5.09375,
+      "rewards/margins": 1.203125,
+      "rewards/rejected": -6.28125,
+      "step": 630
+    },
+    {
+      "epoch": 0.6701570680628273,
+      "grad_norm": 2.1500746204108623,
+      "learning_rate": 4.772235575174776e-05,
+      "logits/chosen": -11.125,
+      "logits/rejected": -10.3125,
+      "logps/chosen": -1024.0,
+      "logps/rejected": -1304.0,
+      "loss": 0.4773,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -5.84375,
+      "rewards/margins": 1.4140625,
+      "rewards/rejected": -7.25,
+      "step": 640
+    },
+    {
+      "epoch": 0.680628272251309,
+      "grad_norm": 2.633705064835583,
+      "learning_rate": 4.7593622872782754e-05,
+      "logits/chosen": -10.25,
+      "logits/rejected": -9.6875,
+      "logps/chosen": -908.0,
+      "logps/rejected": -1112.0,
+      "loss": 0.4948,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -5.28125,
+      "rewards/margins": 0.94140625,
+      "rewards/rejected": -6.21875,
+      "step": 650
+    },
+    {
+      "epoch": 0.6910994764397905,
+      "grad_norm": 2.2951995186399277,
+      "learning_rate": 4.746153482466542e-05,
+      "logits/chosen": -8.875,
+      "logits/rejected": -8.1875,
+      "logps/chosen": -980.0,
+      "logps/rejected": -1152.0,
+      "loss": 0.5016,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -5.84375,
+      "rewards/margins": 1.0859375,
+      "rewards/rejected": -6.9375,
+      "step": 660
+    },
+    {
+      "epoch": 0.7015706806282722,
+      "grad_norm": 2.1689085624054703,
+      "learning_rate": 4.732611122256537e-05,
+      "logits/chosen": -10.3125,
+      "logits/rejected": -9.875,
+      "logps/chosen": -960.0,
+      "logps/rejected": -1160.0,
+      "loss": 0.4773,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -5.28125,
+      "rewards/margins": 1.2734375,
+      "rewards/rejected": -6.53125,
+      "step": 670
+    },
+    {
+      "epoch": 0.7120418848167539,
+      "grad_norm": 2.237995573105497,
+      "learning_rate": 4.718737217698435e-05,
+      "logits/chosen": -10.75,
+      "logits/rejected": -10.5,
+      "logps/chosen": -824.0,
+      "logps/rejected": -1016.0,
+      "loss": 0.5256,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -4.46875,
+      "rewards/margins": 1.0078125,
+      "rewards/rejected": -5.46875,
+      "step": 680
+    },
+    {
+      "epoch": 0.7225130890052356,
+      "grad_norm": 2.8369807600016763,
+      "learning_rate": 4.704533829076982e-05,
+      "logits/chosen": -11.75,
+      "logits/rejected": -10.8125,
+      "logps/chosen": -816.0,
+      "logps/rejected": -1160.0,
+      "loss": 0.4752,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": -4.15625,
+      "rewards/margins": 1.6796875,
+      "rewards/rejected": -5.84375,
+      "step": 690
+    },
+    {
+      "epoch": 0.7329842931937173,
+      "grad_norm": 2.832657180557495,
+      "learning_rate": 4.6900030656055455e-05,
+      "logits/chosen": -10.1875,
+      "logits/rejected": -9.8125,
+      "logps/chosen": -904.0,
+      "logps/rejected": -1112.0,
+      "loss": 0.4846,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -4.875,
+      "rewards/margins": 1.1640625,
+      "rewards/rejected": -6.03125,
+      "step": 700
+    },
+    {
+      "epoch": 0.7329842931937173,
+      "eval_logits/chosen": -8.625,
+      "eval_logits/rejected": -7.25,
+      "eval_logps/chosen": -888.0,
+      "eval_logps/rejected": -924.0,
+      "eval_loss": 0.813796877861023,
+      "eval_rewards/accuracies": 0.599609375,
+      "eval_rewards/chosen": -5.6875,
+      "eval_rewards/margins": 0.6640625,
+      "eval_rewards/rejected": -6.34375,
+      "eval_runtime": 46.6535,
+      "eval_samples_per_second": 42.869,
+      "eval_steps_per_second": 0.686,
+      "step": 700
+    },
+    {
+      "epoch": 0.743455497382199,
+      "grad_norm": 3.38180386213725,
+      "learning_rate": 4.6751470851128906e-05,
+      "logits/chosen": -9.75,
+      "logits/rejected": -8.75,
+      "logps/chosen": -952.0,
+      "logps/rejected": -1192.0,
+      "loss": 0.4763,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": -5.15625,
+      "rewards/margins": 1.5,
+      "rewards/rejected": -6.65625,
+      "step": 710
+    },
+    {
+      "epoch": 0.7539267015706806,
+      "grad_norm": 2.135094059484422,
+      "learning_rate": 4.6599680937227454e-05,
+      "logits/chosen": -11.1875,
+      "logits/rejected": -10.125,
+      "logps/chosen": -880.0,
+      "logps/rejected": -1112.0,
+      "loss": 0.4684,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": -4.5625,
+      "rewards/margins": 1.53125,
+      "rewards/rejected": -6.09375,
+      "step": 720
+    },
+    {
+      "epoch": 0.7643979057591623,
+      "grad_norm": 2.7871919268924965,
+      "learning_rate": 4.644468345526183e-05,
+      "logits/chosen": -11.375,
+      "logits/rejected": -10.625,
+      "logps/chosen": -888.0,
+      "logps/rejected": -1208.0,
+      "loss": 0.5149,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -4.75,
+      "rewards/margins": 1.703125,
+      "rewards/rejected": -6.4375,
+      "step": 730
+    },
+    {
+      "epoch": 0.774869109947644,
+      "grad_norm": 2.824019091892997,
+      "learning_rate": 4.628650142246894e-05,
+      "logits/chosen": -9.9375,
+      "logits/rejected": -9.125,
+      "logps/chosen": -880.0,
+      "logps/rejected": -1112.0,
+      "loss": 0.4964,
+      "rewards/accuracies": 0.793749988079071,
+      "rewards/chosen": -4.96875,
+      "rewards/margins": 1.3125,
+      "rewards/rejected": -6.28125,
+      "step": 740
+    },
+    {
+      "epoch": 0.7853403141361257,
+      "grad_norm": 3.1209058777021714,
+      "learning_rate": 4.6125158328993765e-05,
+      "logits/chosen": -10.0625,
+      "logits/rejected": -9.1875,
+      "logps/chosen": -848.0,
+      "logps/rejected": -1072.0,
+      "loss": 0.4976,
+      "rewards/accuracies": 0.78125,
+      "rewards/chosen": -4.59375,
+      "rewards/margins": 0.88671875,
+      "rewards/rejected": -5.5,
+      "step": 750
+    },
+    {
+      "epoch": 0.7958115183246073,
+      "grad_norm": 3.4125851321038625,
+      "learning_rate": 4.5960678134401064e-05,
+      "logits/chosen": -10.6875,
+      "logits/rejected": -9.6875,
+      "logps/chosen": -884.0,
+      "logps/rejected": -1144.0,
+      "loss": 0.4801,
+      "rewards/accuracies": 0.768750011920929,
+      "rewards/chosen": -4.6875,
+      "rewards/margins": 1.3984375,
+      "rewards/rejected": -6.0625,
+      "step": 760
+    },
+    {
+      "epoch": 0.806282722513089,
+      "grad_norm": 3.762260741004996,
+      "learning_rate": 4.579308526411734e-05,
+      "logits/chosen": -9.0,
+      "logits/rejected": -8.5,
+      "logps/chosen": -1004.0,
+      "logps/rejected": -1224.0,
+      "loss": 0.4636,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": -6.0,
+      "rewards/margins": 1.484375,
+      "rewards/rejected": -7.5,
+      "step": 770
+    },
+    {
+      "epoch": 0.8167539267015707,
+      "grad_norm": 3.9679923042468954,
+      "learning_rate": 4.562240460580367e-05,
+      "logits/chosen": -10.5,
+      "logits/rejected": -9.9375,
+      "logps/chosen": -892.0,
+      "logps/rejected": -1152.0,
+      "loss": 0.4316,
+      "rewards/accuracies": 0.8187500238418579,
+      "rewards/chosen": -4.9375,
+      "rewards/margins": 1.4375,
+      "rewards/rejected": -6.375,
+      "step": 780
+    },
+    {
+      "epoch": 0.8272251308900523,
+      "grad_norm": 2.314774282426851,
+      "learning_rate": 4.544866150565985e-05,
+      "logits/chosen": -10.1875,
+      "logits/rejected": -9.5625,
+      "logps/chosen": -972.0,
+      "logps/rejected": -1272.0,
+      "loss": 0.5085,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": -5.59375,
+      "rewards/margins": 1.6328125,
+      "rewards/rejected": -7.21875,
+      "step": 790
+    },
+    {
+      "epoch": 0.837696335078534,
+      "grad_norm": 2.6804262409159385,
+      "learning_rate": 4.527188176466051e-05,
+      "logits/chosen": -11.0,
+      "logits/rejected": -10.75,
+      "logps/chosen": -908.0,
+      "logps/rejected": -1112.0,
+      "loss": 0.475,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -4.75,
+      "rewards/margins": 1.140625,
+      "rewards/rejected": -5.90625,
+      "step": 800
+    },
+    {
+      "epoch": 0.837696335078534,
+      "eval_logits/chosen": -10.0625,
+      "eval_logits/rejected": -8.9375,
+      "eval_logps/chosen": -868.0,
+      "eval_logps/rejected": -900.0,
+      "eval_loss": 0.7353593707084656,
+      "eval_rewards/accuracies": 0.61328125,
+      "eval_rewards/chosen": -5.5,
+      "eval_rewards/margins": 0.6015625,
+      "eval_rewards/rejected": -6.09375,
+      "eval_runtime": 46.6513,
+      "eval_samples_per_second": 42.871,
+      "eval_steps_per_second": 0.686,
+      "step": 800
+    },
+    {
+      "epoch": 0.8481675392670157,
+      "grad_norm": 2.3642703551498863,
+      "learning_rate": 4.509209163472357e-05,
+      "logits/chosen": -11.25,
+      "logits/rejected": -10.3125,
+      "logps/chosen": -960.0,
+      "logps/rejected": -1296.0,
+      "loss": 0.4508,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": -5.28125,
+      "rewards/margins": 1.703125,
+      "rewards/rejected": -7.0,
+      "step": 810
+    },
+    {
+      "epoch": 0.8586387434554974,
+      "grad_norm": 1.9388731619599733,
+      "learning_rate": 4.4909317814811896e-05,
+      "logits/chosen": -10.1875,
+      "logits/rejected": -9.5,
+      "logps/chosen": -980.0,
+      "logps/rejected": -1264.0,
+      "loss": 0.4925,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -5.6875,
+      "rewards/margins": 1.5859375,
+      "rewards/rejected": -7.28125,
+      "step": 820
+    },
+    {
+      "epoch": 0.8691099476439791,
+      "grad_norm": 2.795751920306338,
+      "learning_rate": 4.472358744696845e-05,
+      "logits/chosen": -11.125,
+      "logits/rejected": -10.5625,
+      "logps/chosen": -940.0,
+      "logps/rejected": -1144.0,
+      "loss": 0.5011,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -5.375,
+      "rewards/margins": 1.328125,
+      "rewards/rejected": -6.71875,
+      "step": 830
+    },
+    {
+      "epoch": 0.8795811518324608,
+      "grad_norm": 2.4026863352022167,
+      "learning_rate": 4.4534928112285645e-05,
+      "logits/chosen": -10.3125,
+      "logits/rejected": -9.5,
+      "logps/chosen": -992.0,
+      "logps/rejected": -1224.0,
+      "loss": 0.4933,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -5.71875,
+      "rewards/margins": 1.28125,
+      "rewards/rejected": -7.0,
+      "step": 840
+    },
+    {
+      "epoch": 0.8900523560209425,
+      "grad_norm": 2.4775270195103496,
+      "learning_rate": 4.43433678268096e-05,
+      "logits/chosen": -9.125,
+      "logits/rejected": -8.4375,
+      "logps/chosen": -992.0,
+      "logps/rejected": -1216.0,
+      "loss": 0.4605,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -6.0,
+      "rewards/margins": 1.3359375,
+      "rewards/rejected": -7.3125,
+      "step": 850
+    },
+    {
+      "epoch": 0.900523560209424,
+      "grad_norm": 2.5385019657243424,
+      "learning_rate": 4.414893503737968e-05,
+      "logits/chosen": -9.5,
+      "logits/rejected": -8.375,
+      "logps/chosen": -928.0,
+      "logps/rejected": -1128.0,
+      "loss": 0.4882,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -5.5,
+      "rewards/margins": 1.359375,
+      "rewards/rejected": -6.84375,
+      "step": 860
+    },
+    {
+      "epoch": 0.9109947643979057,
+      "grad_norm": 2.641691812834352,
+      "learning_rate": 4.395165861740414e-05,
+      "logits/chosen": -10.875,
+      "logits/rejected": -10.3125,
+      "logps/chosen": -812.0,
+      "logps/rejected": -1056.0,
+      "loss": 0.4784,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": -4.34375,
+      "rewards/margins": 1.3984375,
+      "rewards/rejected": -5.75,
+      "step": 870
+    },
+    {
+      "epoch": 0.9214659685863874,
+      "grad_norm": 2.4341279971251977,
+      "learning_rate": 4.375156786257241e-05,
+      "logits/chosen": -11.0,
+      "logits/rejected": -10.5,
+      "logps/chosen": -928.0,
+      "logps/rejected": -1120.0,
+      "loss": 0.4985,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": -4.8125,
+      "rewards/margins": 1.1484375,
+      "rewards/rejected": -5.96875,
+      "step": 880
+    },
+    {
+      "epoch": 0.9319371727748691,
+      "grad_norm": 2.122364195576178,
+      "learning_rate": 4.354869248650466e-05,
+      "logits/chosen": -10.75,
+      "logits/rejected": -9.9375,
+      "logps/chosen": -864.0,
+      "logps/rejected": -1040.0,
+      "loss": 0.4762,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": -4.53125,
+      "rewards/margins": 1.140625,
+      "rewards/rejected": -5.65625,
+      "step": 890
+    },
+    {
+      "epoch": 0.9424083769633508,
+      "grad_norm": 4.027936423662177,
+      "learning_rate": 4.334306261633931e-05,
+      "logits/chosen": -7.8125,
+      "logits/rejected": -6.875,
+      "logps/chosen": -1104.0,
+      "logps/rejected": -1328.0,
+      "loss": 0.4734,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": -6.71875,
+      "rewards/margins": 1.28125,
+      "rewards/rejected": -8.0,
+      "step": 900
+    },
+    {
+      "epoch": 0.9424083769633508,
+      "eval_logits/chosen": -7.34375,
+      "eval_logits/rejected": -5.75,
+      "eval_logps/chosen": -1072.0,
+      "eval_logps/rejected": -1120.0,
+      "eval_loss": 0.8620625138282776,
+      "eval_rewards/accuracies": 0.607421875,
+      "eval_rewards/chosen": -7.5625,
+      "eval_rewards/margins": 0.73046875,
+      "eval_rewards/rejected": -8.3125,
+      "eval_runtime": 46.6481,
+      "eval_samples_per_second": 42.874,
+      "eval_steps_per_second": 0.686,
+      "step": 900
+    },
+    {
+      "epoch": 0.9528795811518325,
+      "grad_norm": 3.2745267150043333,
+      "learning_rate": 4.313470878825911e-05,
+      "logits/chosen": -10.0,
+      "logits/rejected": -9.125,
+      "logps/chosen": -1040.0,
+      "logps/rejected": -1368.0,
+      "loss": 0.4797,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": -6.3125,
+      "rewards/margins": 1.7890625,
+      "rewards/rejected": -8.125,
+      "step": 910
+    },
+    {
+      "epoch": 0.9633507853403142,
+      "grad_norm": 1.9679786295179182,
+      "learning_rate": 4.292366194295652e-05,
+      "logits/chosen": -11.0,
+      "logits/rejected": -10.1875,
+      "logps/chosen": -940.0,
+      "logps/rejected": -1144.0,
+      "loss": 0.4708,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": -5.15625,
+      "rewards/margins": 1.4921875,
+      "rewards/rejected": -6.65625,
+      "step": 920
+    },
+    {
+      "epoch": 0.9738219895287958,
+      "grad_norm": 2.467595527518976,
+      "learning_rate": 4.2709953421038964e-05,
+      "logits/chosen": -10.6875,
+      "logits/rejected": -9.9375,
+      "logps/chosen": -896.0,
+      "logps/rejected": -1128.0,
+      "loss": 0.457,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -5.09375,
+      "rewards/margins": 1.3515625,
+      "rewards/rejected": -6.4375,
+      "step": 930
+    },
+    {
+      "epoch": 0.9842931937172775,
+      "grad_norm": 2.607984317889705,
+      "learning_rate": 4.249361495837478e-05,
+      "logits/chosen": -10.1875,
+      "logits/rejected": -9.3125,
+      "logps/chosen": -1004.0,
+      "logps/rejected": -1240.0,
+      "loss": 0.4818,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": -6.0,
+      "rewards/margins": 1.421875,
+      "rewards/rejected": -7.4375,
+      "step": 940
+    },
+    {
+      "epoch": 0.9947643979057592,
+      "grad_norm": 2.6996469835601102,
+      "learning_rate": 4.227467868138035e-05,
+      "logits/chosen": -10.0625,
+      "logits/rejected": -9.3125,
+      "logps/chosen": -968.0,
+      "logps/rejected": -1248.0,
+      "loss": 0.4726,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -5.5,
+      "rewards/margins": 1.5078125,
+      "rewards/rejected": -7.0,
+      "step": 950
+    },
+    {
+      "epoch": 1.0052356020942408,
+      "grad_norm": 1.2576788188640022,
+      "learning_rate": 4.205317710224938e-05,
+      "logits/chosen": -12.0625,
+      "logits/rejected": -11.5,
+      "logps/chosen": -880.0,
+      "logps/rejected": -1192.0,
+      "loss": 0.3033,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": -4.5625,
+      "rewards/margins": 2.40625,
+      "rewards/rejected": -6.96875,
+      "step": 960
+    },
+    {
+      "epoch": 1.0157068062827226,
+      "grad_norm": 0.8604580733208288,
+      "learning_rate": 4.1829143114124733e-05,
+      "logits/chosen": -11.25,
+      "logits/rejected": -10.0,
+      "logps/chosen": -1112.0,
+      "logps/rejected": -1624.0,
+      "loss": 0.1738,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -7.0,
+      "rewards/margins": 3.953125,
+      "rewards/rejected": -10.9375,
+      "step": 970
+    },
+    {
+      "epoch": 1.0261780104712042,
+      "grad_norm": 1.174372433159922,
+      "learning_rate": 4.160260998621387e-05,
+      "logits/chosen": -12.5625,
+      "logits/rejected": -11.125,
+      "logps/chosen": -1088.0,
+      "logps/rejected": -1600.0,
+      "loss": 0.1385,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -6.40625,
+      "rewards/margins": 4.0625,
+      "rewards/rejected": -10.4375,
+      "step": 980
+    },
+    {
+      "epoch": 1.036649214659686,
+      "grad_norm": 1.300839240198588,
+      "learning_rate": 4.137361135884824e-05,
+      "logits/chosen": -9.4375,
+      "logits/rejected": -8.0625,
+      "logps/chosen": -1176.0,
+      "logps/rejected": -1720.0,
+      "loss": 0.1374,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -7.59375,
+      "rewards/margins": 4.1875,
+      "rewards/rejected": -11.8125,
+      "step": 990
+    },
+    {
+      "epoch": 1.0471204188481675,
+      "grad_norm": 2.0827152263073163,
+      "learning_rate": 4.114218123848777e-05,
+      "logits/chosen": -6.6875,
+      "logits/rejected": -5.5625,
+      "logps/chosen": -1344.0,
+      "logps/rejected": -1920.0,
+      "loss": 0.146,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -9.5,
+      "rewards/margins": 4.59375,
+      "rewards/rejected": -14.0625,
+      "step": 1000
+    },
+    {
+      "epoch": 1.0471204188481675,
+      "eval_logits/chosen": -4.40625,
+      "eval_logits/rejected": -2.703125,
+      "eval_logps/chosen": -1384.0,
+      "eval_logps/rejected": -1440.0,
+      "eval_loss": 1.1525156497955322,
+      "eval_rewards/accuracies": 0.568359375,
+      "eval_rewards/chosen": -10.6875,
+      "eval_rewards/margins": 0.84765625,
+      "eval_rewards/rejected": -11.5,
+      "eval_runtime": 46.6558,
+      "eval_samples_per_second": 42.867,
+      "eval_steps_per_second": 0.686,
+      "step": 1000
+    },
+    {
+      "epoch": 1.057591623036649,
+      "grad_norm": 3.994610680964802,
+      "learning_rate": 4.0908353992670824e-05,
+      "logits/chosen": -8.375,
+      "logits/rejected": -6.875,
+      "logps/chosen": -1128.0,
+      "logps/rejected": -1688.0,
+      "loss": 0.1548,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -7.28125,
+      "rewards/margins": 4.65625,
+      "rewards/rejected": -11.9375,
+      "step": 1010
+    },
+    {
+      "epoch": 1.068062827225131,
+      "grad_norm": 3.698765886248804,
+      "learning_rate": 4.067216434491059e-05,
+      "logits/chosen": -7.46875,
+      "logits/rejected": -6.3125,
+      "logps/chosen": -1048.0,
+      "logps/rejected": -1568.0,
+      "loss": 0.1485,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -6.40625,
+      "rewards/margins": 4.0625,
+      "rewards/rejected": -10.5,
+      "step": 1020
+    },
+    {
+      "epoch": 1.0785340314136125,
+      "grad_norm": 1.6276575619454736,
+      "learning_rate": 4.043364736953866e-05,
+      "logits/chosen": -6.4375,
+      "logits/rejected": -5.5625,
+      "logps/chosen": -1176.0,
+      "logps/rejected": -1696.0,
+      "loss": 0.164,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -7.125,
+      "rewards/margins": 4.15625,
+      "rewards/rejected": -11.3125,
+      "step": 1030
+    },
+    {
+      "epoch": 1.0890052356020943,
+      "grad_norm": 2.2553475006678574,
+      "learning_rate": 4.019283848649643e-05,
+      "logits/chosen": -9.5625,
+      "logits/rejected": -8.5,
+      "logps/chosen": -1056.0,
+      "logps/rejected": -1552.0,
+      "loss": 0.1701,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -6.40625,
+      "rewards/margins": 3.984375,
+      "rewards/rejected": -10.375,
+      "step": 1040
+    },
+    {
+      "epoch": 1.0994764397905759,
+      "grad_norm": 1.523265452666949,
+      "learning_rate": 3.9949773456075215e-05,
+      "logits/chosen": -9.4375,
+      "logits/rejected": -8.3125,
+      "logps/chosen": -1000.0,
+      "logps/rejected": -1528.0,
+      "loss": 0.1274,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -6.25,
+      "rewards/margins": 4.09375,
+      "rewards/rejected": -10.375,
+      "step": 1050
+    },
+    {
+      "epoch": 1.1099476439790577,
+      "grad_norm": 1.1666046345019474,
+      "learning_rate": 3.970448837360585e-05,
+      "logits/chosen": -9.125,
+      "logits/rejected": -7.5625,
+      "logps/chosen": -1272.0,
+      "logps/rejected": -1784.0,
+      "loss": 0.137,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -8.125,
+      "rewards/margins": 4.21875,
+      "rewards/rejected": -12.3125,
+      "step": 1060
+    },
+    {
+      "epoch": 1.1204188481675392,
+      "grad_norm": 1.7821092277761532,
+      "learning_rate": 3.945701966409846e-05,
+      "logits/chosen": -9.8125,
+      "logits/rejected": -8.3125,
+      "logps/chosen": -1088.0,
+      "logps/rejected": -1672.0,
+      "loss": 0.1317,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -6.75,
+      "rewards/margins": 4.71875,
+      "rewards/rejected": -11.4375,
+      "step": 1070
+    },
+    {
+      "epoch": 1.130890052356021,
+      "grad_norm": 3.5961877150043184,
+      "learning_rate": 3.9207404076833374e-05,
+      "logits/chosen": -9.1875,
+      "logits/rejected": -7.6875,
+      "logps/chosen": -1192.0,
+      "logps/rejected": -1704.0,
+      "loss": 0.1348,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -7.375,
+      "rewards/margins": 4.34375,
+      "rewards/rejected": -11.75,
+      "step": 1080
+    },
+    {
+      "epoch": 1.1413612565445026,
+      "grad_norm": 2.039390136273228,
+      "learning_rate": 3.895567867990379e-05,
+      "logits/chosen": -8.0,
+      "logits/rejected": -6.03125,
+      "logps/chosen": -1272.0,
+      "logps/rejected": -1824.0,
+      "loss": 0.13,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.0625,
+      "rewards/margins": 4.8125,
+      "rewards/rejected": -12.875,
+      "step": 1090
+    },
+    {
+      "epoch": 1.1518324607329844,
+      "grad_norm": 1.5900754235751582,
+      "learning_rate": 3.870188085471114e-05,
+      "logits/chosen": -7.5,
+      "logits/rejected": -5.78125,
+      "logps/chosen": -1184.0,
+      "logps/rejected": -1672.0,
+      "loss": 0.1356,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -7.65625,
+      "rewards/margins": 4.34375,
+      "rewards/rejected": -12.0,
+      "step": 1100
+    },
+    {
+      "epoch": 1.1518324607329844,
+      "eval_logits/chosen": -5.53125,
+      "eval_logits/rejected": -3.6875,
+      "eval_logps/chosen": -1232.0,
+      "eval_logps/rejected": -1296.0,
+      "eval_loss": 0.9971562623977661,
+      "eval_rewards/accuracies": 0.609375,
+      "eval_rewards/chosen": -9.1875,
+      "eval_rewards/margins": 0.90625,
+      "eval_rewards/rejected": -10.0625,
+      "eval_runtime": 46.6258,
+      "eval_samples_per_second": 42.895,
+      "eval_steps_per_second": 0.686,
+      "step": 1100
+    },
+    {
+      "epoch": 1.162303664921466,
+      "grad_norm": 2.456500645589885,
+      "learning_rate": 3.844604829041395e-05,
+      "logits/chosen": -8.125,
+      "logits/rejected": -6.25,
+      "logps/chosen": -1120.0,
+      "logps/rejected": -1640.0,
+      "loss": 0.1413,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -7.53125,
+      "rewards/margins": 4.28125,
+      "rewards/rejected": -11.8125,
+      "step": 1110
+    },
+    {
+      "epoch": 1.1727748691099475,
+      "grad_norm": 2.9128742152946328,
+      "learning_rate": 3.818821897833095e-05,
+      "logits/chosen": -9.0625,
+      "logits/rejected": -7.75,
+      "logps/chosen": -1176.0,
+      "logps/rejected": -1728.0,
+      "loss": 0.1384,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -7.59375,
+      "rewards/margins": 4.625,
+      "rewards/rejected": -12.25,
+      "step": 1120
+    },
+    {
+      "epoch": 1.1832460732984293,
+      "grad_norm": 1.8525098528649633,
+      "learning_rate": 3.792843120629935e-05,
+      "logits/chosen": -9.75,
+      "logits/rejected": -7.9375,
+      "logps/chosen": -1112.0,
+      "logps/rejected": -1648.0,
+      "loss": 0.1584,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -7.125,
+      "rewards/margins": 4.3125,
+      "rewards/rejected": -11.4375,
+      "step": 1130
+    },
+    {
+      "epoch": 1.193717277486911,
+      "grad_norm": 2.4894408923481186,
+      "learning_rate": 3.766672355298907e-05,
+      "logits/chosen": -10.1875,
+      "logits/rejected": -9.3125,
+      "logps/chosen": -1096.0,
+      "logps/rejected": -1616.0,
+      "loss": 0.1484,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -7.21875,
+      "rewards/margins": 4.0625,
+      "rewards/rejected": -11.25,
+      "step": 1140
+    },
+    {
+      "epoch": 1.2041884816753927,
+      "grad_norm": 2.428943113803734,
+      "learning_rate": 3.740313488217373e-05,
+      "logits/chosen": -11.0,
+      "logits/rejected": -9.8125,
+      "logps/chosen": -1104.0,
+      "logps/rejected": -1672.0,
+      "loss": 0.1334,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -7.15625,
+      "rewards/margins": 4.71875,
+      "rewards/rejected": -11.875,
+      "step": 1150
+    },
+    {
+      "epoch": 1.2146596858638743,
+      "grad_norm": 1.6877001313641127,
+      "learning_rate": 3.7137704336959464e-05,
+      "logits/chosen": -10.375,
+      "logits/rejected": -8.8125,
+      "logps/chosen": -1160.0,
+      "logps/rejected": -1784.0,
+      "loss": 0.1161,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -7.71875,
+      "rewards/margins": 5.0,
+      "rewards/rejected": -12.75,
+      "step": 1160
+    },
+    {
+      "epoch": 1.225130890052356,
+      "grad_norm": 1.9072738397714215,
+      "learning_rate": 3.687047133397201e-05,
+      "logits/chosen": -10.25,
+      "logits/rejected": -8.6875,
+      "logps/chosen": -1256.0,
+      "logps/rejected": -1912.0,
+      "loss": 0.1201,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -8.625,
+      "rewards/margins": 5.3125,
+      "rewards/rejected": -13.9375,
+      "step": 1170
+    },
+    {
+      "epoch": 1.2356020942408377,
+      "grad_norm": 2.343319772250893,
+      "learning_rate": 3.6601475557503415e-05,
+      "logits/chosen": -9.9375,
+      "logits/rejected": -8.25,
+      "logps/chosen": -1256.0,
+      "logps/rejected": -1848.0,
+      "loss": 0.1607,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -8.375,
+      "rewards/margins": 4.84375,
+      "rewards/rejected": -13.25,
+      "step": 1180
+    },
+    {
+      "epoch": 1.2460732984293195,
+      "grad_norm": 2.1505219308500307,
+      "learning_rate": 3.633075695361882e-05,
+      "logits/chosen": -8.75,
+      "logits/rejected": -7.03125,
+      "logps/chosen": -1256.0,
+      "logps/rejected": -1880.0,
+      "loss": 0.1355,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.75,
+      "rewards/margins": 5.09375,
+      "rewards/rejected": -13.875,
+      "step": 1190
+    },
+    {
+      "epoch": 1.256544502617801,
+      "grad_norm": 1.1222338906202685,
+      "learning_rate": 3.6058355724224477e-05,
+      "logits/chosen": -10.25,
+      "logits/rejected": -8.5,
+      "logps/chosen": -1128.0,
+      "logps/rejected": -1656.0,
+      "loss": 0.1237,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -7.21875,
+      "rewards/margins": 4.25,
+      "rewards/rejected": -11.5,
+      "step": 1200
+    },
+    {
+      "epoch": 1.256544502617801,
+      "eval_logits/chosen": -8.125,
+      "eval_logits/rejected": -6.34375,
+      "eval_logps/chosen": -1168.0,
+      "eval_logps/rejected": -1216.0,
+      "eval_loss": 0.999343752861023,
+      "eval_rewards/accuracies": 0.6015625,
+      "eval_rewards/chosen": -8.5,
+      "eval_rewards/margins": 0.76171875,
+      "eval_rewards/rejected": -9.25,
+      "eval_runtime": 46.6649,
+      "eval_samples_per_second": 42.859,
+      "eval_steps_per_second": 0.686,
+      "step": 1200
+    },
+    {
+      "epoch": 1.2670157068062826,
+      "grad_norm": 2.313804590650624,
+      "learning_rate": 3.5784312321097744e-05,
+      "logits/chosen": -8.5,
+      "logits/rejected": -6.71875,
+      "logps/chosen": -1208.0,
+      "logps/rejected": -1736.0,
+      "loss": 0.1314,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -7.8125,
+      "rewards/margins": 4.53125,
+      "rewards/rejected": -12.375,
+      "step": 1210
+    },
+    {
+      "epoch": 1.2774869109947644,
+      "grad_norm": 2.9031427744429736,
+      "learning_rate": 3.550866743987994e-05,
+      "logits/chosen": -7.03125,
+      "logits/rejected": -5.21875,
+      "logps/chosen": -1296.0,
+      "logps/rejected": -1824.0,
+      "loss": 0.1504,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -8.6875,
+      "rewards/margins": 4.5,
+      "rewards/rejected": -13.1875,
+      "step": 1220
+    },
+    {
+      "epoch": 1.2879581151832462,
+      "grad_norm": 1.710155637141148,
+      "learning_rate": 3.5231462014033024e-05,
+      "logits/chosen": -8.0625,
+      "logits/rejected": -6.0,
+      "logps/chosen": -1152.0,
+      "logps/rejected": -1680.0,
+      "loss": 0.1199,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -7.65625,
+      "rewards/margins": 4.59375,
+      "rewards/rejected": -12.25,
+      "step": 1230
+    },
+    {
+      "epoch": 1.2984293193717278,
+      "grad_norm": 1.6879249777383887,
+      "learning_rate": 3.495273720876095e-05,
+      "logits/chosen": -5.90625,
+      "logits/rejected": -4.40625,
+      "logps/chosen": -1304.0,
+      "logps/rejected": -1840.0,
+      "loss": 0.1188,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -8.875,
+      "rewards/margins": 4.375,
+      "rewards/rejected": -13.25,
+      "step": 1240
+    },
+    {
+      "epoch": 1.3089005235602094,
+      "grad_norm": 1.4402514325866238,
+      "learning_rate": 3.4672534414896616e-05,
+      "logits/chosen": -8.9375,
+      "logits/rejected": -7.03125,
+      "logps/chosen": -1256.0,
+      "logps/rejected": -1752.0,
+      "loss": 0.1329,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -8.1875,
+      "rewards/margins": 4.375,
+      "rewards/rejected": -12.5625,
+      "step": 1250
+    },
+    {
+      "epoch": 1.3193717277486912,
+      "grad_norm": 2.0495408205865346,
+      "learning_rate": 3.439089524275526e-05,
+      "logits/chosen": -8.5,
+      "logits/rejected": -6.6875,
+      "logps/chosen": -1112.0,
+      "logps/rejected": -1608.0,
+      "loss": 0.1421,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -7.25,
+      "rewards/margins": 4.125,
+      "rewards/rejected": -11.375,
+      "step": 1260
+    },
+    {
+      "epoch": 1.3298429319371727,
+      "grad_norm": 1.9371809370917994,
+      "learning_rate": 3.410786151595533e-05,
+      "logits/chosen": -8.0625,
+      "logits/rejected": -6.1875,
+      "logps/chosen": -1176.0,
+      "logps/rejected": -1664.0,
+      "loss": 0.147,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -7.84375,
+      "rewards/margins": 4.25,
+      "rewards/rejected": -12.125,
+      "step": 1270
+    },
+    {
+      "epoch": 1.3403141361256545,
+      "grad_norm": 1.6265418274873171,
+      "learning_rate": 3.382347526520765e-05,
+      "logits/chosen": -7.5625,
+      "logits/rejected": -5.84375,
+      "logps/chosen": -1200.0,
+      "logps/rejected": -1712.0,
+      "loss": 0.1466,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -7.96875,
+      "rewards/margins": 3.765625,
+      "rewards/rejected": -11.75,
+      "step": 1280
+    },
+    {
+      "epoch": 1.350785340314136,
+      "grad_norm": 2.276988462536782,
+      "learning_rate": 3.353777872207381e-05,
+      "logits/chosen": -6.28125,
+      "logits/rejected": -4.78125,
+      "logps/chosen": -1224.0,
+      "logps/rejected": -1712.0,
+      "loss": 0.138,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -8.1875,
+      "rewards/margins": 4.15625,
+      "rewards/rejected": -12.375,
+      "step": 1290
+    },
+    {
+      "epoch": 1.3612565445026177,
+      "grad_norm": 1.647469970307563,
+      "learning_rate": 3.325081431269475e-05,
+      "logits/chosen": -5.28125,
+      "logits/rejected": -3.40625,
+      "logps/chosen": -1216.0,
+      "logps/rejected": -1760.0,
+      "loss": 0.1387,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -8.4375,
+      "rewards/margins": 4.40625,
+      "rewards/rejected": -12.8125,
+      "step": 1300
+    },
+    {
+      "epoch": 1.3612565445026177,
+      "eval_logits/chosen": -4.5625,
+      "eval_logits/rejected": -2.421875,
+      "eval_logps/chosen": -1408.0,
+      "eval_logps/rejected": -1488.0,
+      "eval_loss": 1.168468713760376,
+      "eval_rewards/accuracies": 0.59765625,
+      "eval_rewards/chosen": -10.875,
+      "eval_rewards/margins": 1.0625,
+      "eval_rewards/rejected": -11.9375,
+      "eval_runtime": 46.6401,
+      "eval_samples_per_second": 42.882,
+      "eval_steps_per_second": 0.686,
+      "step": 1300
+    },
+    {
+      "epoch": 1.3717277486910995,
+      "grad_norm": 2.4394086197188503,
+      "learning_rate": 3.296262465149046e-05,
+      "logits/chosen": -6.28125,
+      "logits/rejected": -4.03125,
+      "logps/chosen": -1376.0,
+      "logps/rejected": -2008.0,
+      "loss": 0.1324,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.5,
+      "rewards/margins": 5.0,
+      "rewards/rejected": -14.5,
+      "step": 1310
+    },
+    {
+      "epoch": 1.3821989528795813,
+      "grad_norm": 2.620378844366017,
+      "learning_rate": 3.2673252534831687e-05,
+      "logits/chosen": -6.0,
+      "logits/rejected": -4.09375,
+      "logps/chosen": -1376.0,
+      "logps/rejected": -1952.0,
+      "loss": 0.1556,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -9.75,
+      "rewards/margins": 4.75,
+      "rewards/rejected": -14.5,
+      "step": 1320
+    },
+    {
+      "epoch": 1.3926701570680629,
+      "grad_norm": 1.2886579832772507,
+      "learning_rate": 3.2382740934684696e-05,
+      "logits/chosen": -6.8125,
+      "logits/rejected": -4.78125,
+      "logps/chosen": -1272.0,
+      "logps/rejected": -1848.0,
+      "loss": 0.1421,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -8.6875,
+      "rewards/margins": 4.46875,
+      "rewards/rejected": -13.125,
+      "step": 1330
+    },
+    {
+      "epoch": 1.4031413612565444,
+      "grad_norm": 2.425757545679554,
+      "learning_rate": 3.209113299222982e-05,
+      "logits/chosen": -6.53125,
+      "logits/rejected": -5.34375,
+      "logps/chosen": -1144.0,
+      "logps/rejected": -1712.0,
+      "loss": 0.1424,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -7.90625,
+      "rewards/margins": 4.28125,
+      "rewards/rejected": -12.1875,
+      "step": 1340
+    },
+    {
+      "epoch": 1.4136125654450262,
+      "grad_norm": 2.402288090297736,
+      "learning_rate": 3.179847201145505e-05,
+      "logits/chosen": -5.90625,
+      "logits/rejected": -4.21875,
+      "logps/chosen": -1240.0,
+      "logps/rejected": -1768.0,
+      "loss": 0.1339,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -8.1875,
+      "rewards/margins": 4.1875,
+      "rewards/rejected": -12.375,
+      "step": 1350
+    },
+    {
+      "epoch": 1.4240837696335078,
+      "grad_norm": 3.3632267314004856,
+      "learning_rate": 3.150480145272528e-05,
+      "logits/chosen": -6.21875,
+      "logits/rejected": -4.875,
+      "logps/chosen": -1296.0,
+      "logps/rejected": -1808.0,
+      "loss": 0.1606,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -8.9375,
+      "rewards/margins": 4.1875,
+      "rewards/rejected": -13.125,
+      "step": 1360
+    },
+    {
+      "epoch": 1.4345549738219896,
+      "grad_norm": 1.2061259885036963,
+      "learning_rate": 3.121016492632848e-05,
+      "logits/chosen": -7.375,
+      "logits/rejected": -5.625,
+      "logps/chosen": -1256.0,
+      "logps/rejected": -1808.0,
+      "loss": 0.1434,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -8.5,
+      "rewards/margins": 4.5,
+      "rewards/rejected": -13.0,
+      "step": 1370
+    },
+    {
+      "epoch": 1.4450261780104712,
+      "grad_norm": 2.065880760186589,
+      "learning_rate": 3.091460618599951e-05,
+      "logits/chosen": -6.59375,
+      "logits/rejected": -4.1875,
+      "logps/chosen": -1256.0,
+      "logps/rejected": -1832.0,
+      "loss": 0.1191,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -8.5,
+      "rewards/margins": 4.53125,
+      "rewards/rejected": -13.0,
+      "step": 1380
+    },
+    {
+      "epoch": 1.4554973821989527,
+      "grad_norm": 2.5237513475830076,
+      "learning_rate": 3.061816912242265e-05,
+      "logits/chosen": -6.15625,
+      "logits/rejected": -4.65625,
+      "logps/chosen": -1304.0,
+      "logps/rejected": -1904.0,
+      "loss": 0.1252,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -8.875,
+      "rewards/margins": 4.65625,
+      "rewards/rejected": -13.5625,
+      "step": 1390
+    },
+    {
+      "epoch": 1.4659685863874345,
+      "grad_norm": 2.415003405690821,
+      "learning_rate": 3.0320897756713785e-05,
+      "logits/chosen": -6.3125,
+      "logits/rejected": -4.25,
+      "logps/chosen": -1296.0,
+      "logps/rejected": -1904.0,
+      "loss": 0.1298,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -8.75,
+      "rewards/margins": 4.8125,
+      "rewards/rejected": -13.5625,
+      "step": 1400
+    },
+    {
+      "epoch": 1.4659685863874345,
+      "eval_logits/chosen": -4.9375,
+      "eval_logits/rejected": -2.984375,
+      "eval_logps/chosen": -1368.0,
+      "eval_logps/rejected": -1440.0,
+      "eval_loss": 1.0589531660079956,
+      "eval_rewards/accuracies": 0.599609375,
+      "eval_rewards/chosen": -10.4375,
+      "eval_rewards/margins": 1.03125,
+      "eval_rewards/rejected": -11.5,
+      "eval_runtime": 46.6488,
+      "eval_samples_per_second": 42.874,
+      "eval_steps_per_second": 0.686,
+      "step": 1400
+    },
+    {
+      "epoch": 1.4764397905759163,
+      "grad_norm": 2.616377633595522,
+      "learning_rate": 3.002283623388332e-05,
+      "logits/chosen": -6.8125,
+      "logits/rejected": -5.8125,
+      "logps/chosen": -1248.0,
+      "logps/rejected": -1784.0,
+      "loss": 0.1286,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -8.6875,
+      "rewards/margins": 4.40625,
+      "rewards/rejected": -13.125,
+      "step": 1410
+    },
+    {
+      "epoch": 1.486910994764398,
+      "grad_norm": 2.704853003531505,
+      "learning_rate": 2.9724028816280507e-05,
+      "logits/chosen": -7.0625,
+      "logits/rejected": -5.96875,
+      "logps/chosen": -1280.0,
+      "logps/rejected": -1848.0,
+      "loss": 0.1211,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -8.875,
+      "rewards/margins": 4.53125,
+      "rewards/rejected": -13.4375,
+      "step": 1420
+    },
+    {
+      "epoch": 1.4973821989528795,
+      "grad_norm": 3.0357510036280972,
+      "learning_rate": 2.9424519877020524e-05,
+      "logits/chosen": -9.0625,
+      "logits/rejected": -7.59375,
+      "logps/chosen": -1328.0,
+      "logps/rejected": -1952.0,
+      "loss": 0.1298,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.8125,
+      "rewards/margins": 4.8125,
+      "rewards/rejected": -13.625,
+      "step": 1430
+    },
+    {
+      "epoch": 1.5078534031413613,
+      "grad_norm": 2.362095273903202,
+      "learning_rate": 2.912435389339504e-05,
+      "logits/chosen": -8.875,
+      "logits/rejected": -6.65625,
+      "logps/chosen": -1224.0,
+      "logps/rejected": -1760.0,
+      "loss": 0.1319,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -8.25,
+      "rewards/margins": 4.75,
+      "rewards/rejected": -12.9375,
+      "step": 1440
+    },
+    {
+      "epoch": 1.518324607329843,
+      "grad_norm": 2.863166032070748,
+      "learning_rate": 2.8823575440267226e-05,
+      "logits/chosen": -7.59375,
+      "logits/rejected": -6.03125,
+      "logps/chosen": -1344.0,
+      "logps/rejected": -1920.0,
+      "loss": 0.144,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.5,
+      "rewards/margins": 4.6875,
+      "rewards/rejected": -14.1875,
+      "step": 1450
+    },
+    {
+      "epoch": 1.5287958115183247,
+      "grad_norm": 2.0692666473338495,
+      "learning_rate": 2.8522229183452465e-05,
+      "logits/chosen": -7.15625,
+      "logits/rejected": -5.65625,
+      "logps/chosen": -1312.0,
+      "logps/rejected": -1824.0,
+      "loss": 0.1936,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -8.9375,
+      "rewards/margins": 4.34375,
+      "rewards/rejected": -13.3125,
+      "step": 1460
+    },
+    {
+      "epoch": 1.5392670157068062,
+      "grad_norm": 2.455707497935021,
+      "learning_rate": 2.8220359873085373e-05,
+      "logits/chosen": -7.8125,
+      "logits/rejected": -6.5625,
+      "logps/chosen": -1200.0,
+      "logps/rejected": -1720.0,
+      "loss": 0.1291,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -7.90625,
+      "rewards/margins": 4.21875,
+      "rewards/rejected": -12.125,
+      "step": 1470
+    },
+    {
+      "epoch": 1.5497382198952878,
+      "grad_norm": 2.2263010110276724,
+      "learning_rate": 2.7918012336974388e-05,
+      "logits/chosen": -6.3125,
+      "logits/rejected": -4.40625,
+      "logps/chosen": -1232.0,
+      "logps/rejected": -1712.0,
+      "loss": 0.1175,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.3125,
+      "rewards/margins": 4.46875,
+      "rewards/rejected": -12.75,
+      "step": 1480
+    },
+    {
+      "epoch": 1.5602094240837696,
+      "grad_norm": 2.464334639841183,
+      "learning_rate": 2.7615231473944815e-05,
+      "logits/chosen": -5.28125,
+      "logits/rejected": -3.9375,
+      "logps/chosen": -1304.0,
+      "logps/rejected": -1840.0,
+      "loss": 0.1282,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -9.25,
+      "rewards/margins": 4.46875,
+      "rewards/rejected": -13.75,
+      "step": 1490
+    },
+    {
+      "epoch": 1.5706806282722514,
+      "grad_norm": 2.8500507853350356,
+      "learning_rate": 2.7312062247171327e-05,
+      "logits/chosen": -6.65625,
+      "logits/rejected": -5.0625,
+      "logps/chosen": -1248.0,
+      "logps/rejected": -1792.0,
+      "loss": 0.1284,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -8.8125,
+      "rewards/margins": 4.3125,
+      "rewards/rejected": -13.125,
+      "step": 1500
+    },
+    {
+      "epoch": 1.5706806282722514,
+      "eval_logits/chosen": -3.984375,
+      "eval_logits/rejected": -1.9765625,
+      "eval_logps/chosen": -1432.0,
+      "eval_logps/rejected": -1512.0,
+      "eval_loss": 1.1093593835830688,
+      "eval_rewards/accuracies": 0.619140625,
+      "eval_rewards/chosen": -11.125,
+      "eval_rewards/margins": 1.1171875,
+      "eval_rewards/rejected": -12.25,
+      "eval_runtime": 46.6551,
+      "eval_samples_per_second": 42.868,
+      "eval_steps_per_second": 0.686,
+      "step": 1500
+    },
+    {
+      "epoch": 1.581151832460733,
+      "grad_norm": 1.8267481969777848,
+      "learning_rate": 2.7008549677500876e-05,
+      "logits/chosen": -7.375,
+      "logits/rejected": -5.59375,
+      "logps/chosen": -1400.0,
+      "logps/rejected": -1968.0,
+      "loss": 0.1052,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.625,
+      "rewards/margins": 4.6875,
+      "rewards/rejected": -14.375,
+      "step": 1510
+    },
+    {
+      "epoch": 1.5916230366492146,
+      "grad_norm": 1.4554245461129063,
+      "learning_rate": 2.670473883676709e-05,
+      "logits/chosen": -5.65625,
+      "logits/rejected": -3.96875,
+      "logps/chosen": -1416.0,
+      "logps/rejected": -2080.0,
+      "loss": 0.0955,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -10.0625,
+      "rewards/margins": 5.34375,
+      "rewards/rejected": -15.375,
+      "step": 1520
+    },
+    {
+      "epoch": 1.6020942408376964,
+      "grad_norm": 1.1347279717376941,
+      "learning_rate": 2.6400674841097074e-05,
+      "logits/chosen": -5.5625,
+      "logits/rejected": -3.78125,
+      "logps/chosen": -1528.0,
+      "logps/rejected": -2080.0,
+      "loss": 0.1202,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -11.0625,
+      "rewards/margins": 5.0625,
+      "rewards/rejected": -16.125,
+      "step": 1530
+    },
+    {
+      "epoch": 1.6125654450261782,
+      "grad_norm": 3.1201807061357383,
+      "learning_rate": 2.609640284421158e-05,
+      "logits/chosen": -6.21875,
+      "logits/rejected": -4.09375,
+      "logps/chosen": -1392.0,
+      "logps/rejected": -1960.0,
+      "loss": 0.1446,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -9.8125,
+      "rewards/margins": 4.84375,
+      "rewards/rejected": -14.6875,
+      "step": 1540
+    },
+    {
+      "epoch": 1.6230366492146597,
+      "grad_norm": 2.301582229102563,
+      "learning_rate": 2.5791968030719726e-05,
+      "logits/chosen": -7.03125,
+      "logits/rejected": -5.40625,
+      "logps/chosen": -1280.0,
+      "logps/rejected": -1856.0,
+      "loss": 0.1321,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -8.3125,
+      "rewards/margins": 4.4375,
+      "rewards/rejected": -12.8125,
+      "step": 1550
+    },
+    {
+      "epoch": 1.6335078534031413,
+      "grad_norm": 2.991400879098044,
+      "learning_rate": 2.548741560940902e-05,
+      "logits/chosen": -7.125,
+      "logits/rejected": -5.9375,
+      "logps/chosen": -1136.0,
+      "logps/rejected": -1760.0,
+      "loss": 0.1464,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -7.40625,
+      "rewards/margins": 4.46875,
+      "rewards/rejected": -11.875,
+      "step": 1560
+    },
+    {
+      "epoch": 1.6439790575916229,
+      "grad_norm": 0.9647598047441381,
+      "learning_rate": 2.5182790806531787e-05,
+      "logits/chosen": -5.5,
+      "logits/rejected": -4.1875,
+      "logps/chosen": -1248.0,
+      "logps/rejected": -1840.0,
+      "loss": 0.1271,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -8.25,
+      "rewards/margins": 4.75,
+      "rewards/rejected": -13.0,
+      "step": 1570
+    },
+    {
+      "epoch": 1.6544502617801047,
+      "grad_norm": 1.5707480530886089,
+      "learning_rate": 2.4878138859089074e-05,
+      "logits/chosen": -5.65625,
+      "logits/rejected": -4.1875,
+      "logps/chosen": -1264.0,
+      "logps/rejected": -1840.0,
+      "loss": 0.1319,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -8.4375,
+      "rewards/margins": 4.71875,
+      "rewards/rejected": -13.125,
+      "step": 1580
+    },
+    {
+      "epoch": 1.6649214659685865,
+      "grad_norm": 1.9017764943290447,
+      "learning_rate": 2.4573505008112924e-05,
+      "logits/chosen": -5.9375,
+      "logits/rejected": -4.625,
+      "logps/chosen": -1296.0,
+      "logps/rejected": -1872.0,
+      "loss": 0.1167,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -8.9375,
+      "rewards/margins": 4.71875,
+      "rewards/rejected": -13.625,
+      "step": 1590
+    },
+    {
+      "epoch": 1.675392670157068,
+      "grad_norm": 2.0162735021915306,
+      "learning_rate": 2.4268934491948032e-05,
+      "logits/chosen": -6.34375,
+      "logits/rejected": -5.15625,
+      "logps/chosen": -1304.0,
+      "logps/rejected": -1824.0,
+      "loss": 0.1084,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -8.6875,
+      "rewards/margins": 4.5,
+      "rewards/rejected": -13.1875,
+      "step": 1600
+    },
+    {
+      "epoch": 1.675392670157068,
+      "eval_logits/chosen": -4.1875,
+      "eval_logits/rejected": -2.28125,
+      "eval_logps/chosen": -1400.0,
+      "eval_logps/rejected": -1480.0,
+      "eval_loss": 1.1327344179153442,
+      "eval_rewards/accuracies": 0.623046875,
+      "eval_rewards/chosen": -10.8125,
+      "eval_rewards/margins": 1.1328125,
+      "eval_rewards/rejected": -11.9375,
+      "eval_runtime": 46.6466,
+      "eval_samples_per_second": 42.876,
+      "eval_steps_per_second": 0.686,
+      "step": 1600
+    },
+    {
+      "epoch": 1.6858638743455496,
+      "grad_norm": 2.1873137827769047,
+      "learning_rate": 2.396447253953385e-05,
+      "logits/chosen": -8.0625,
+      "logits/rejected": -6.21875,
+      "logps/chosen": -1312.0,
+      "logps/rejected": -1856.0,
+      "loss": 0.1316,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -8.75,
+      "rewards/margins": 4.71875,
+      "rewards/rejected": -13.4375,
+      "step": 1610
+    },
+    {
+      "epoch": 1.6963350785340314,
+      "grad_norm": 2.207454990900936,
+      "learning_rate": 2.3660164363687998e-05,
+      "logits/chosen": -7.71875,
+      "logits/rejected": -6.375,
+      "logps/chosen": -1320.0,
+      "logps/rejected": -1936.0,
+      "loss": 0.1212,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -8.625,
+      "rewards/margins": 4.8125,
+      "rewards/rejected": -13.4375,
+      "step": 1620
+    },
+    {
+      "epoch": 1.7068062827225132,
+      "grad_norm": 2.6039061147405396,
+      "learning_rate": 2.33560551543922e-05,
+      "logits/chosen": -7.5,
+      "logits/rejected": -5.875,
+      "logps/chosen": -1344.0,
+      "logps/rejected": -1840.0,
+      "loss": 0.132,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -8.9375,
+      "rewards/margins": 4.34375,
+      "rewards/rejected": -13.25,
+      "step": 1630
+    },
+    {
+      "epoch": 1.7172774869109948,
+      "grad_norm": 2.205201817015934,
+      "learning_rate": 2.3052190072081493e-05,
+      "logits/chosen": -7.28125,
+      "logits/rejected": -5.4375,
+      "logps/chosen": -1288.0,
+      "logps/rejected": -1816.0,
+      "loss": 0.121,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -8.6875,
+      "rewards/margins": 4.71875,
+      "rewards/rejected": -13.375,
+      "step": 1640
+    },
+    {
+      "epoch": 1.7277486910994764,
+      "grad_norm": 1.964845684748554,
+      "learning_rate": 2.2748614240937864e-05,
+      "logits/chosen": -6.125,
+      "logits/rejected": -4.3125,
+      "logps/chosen": -1272.0,
+      "logps/rejected": -1880.0,
+      "loss": 0.1177,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -8.8125,
+      "rewards/margins": 4.78125,
+      "rewards/rejected": -13.5625,
+      "step": 1650
+    },
+    {
+      "epoch": 1.738219895287958,
+      "grad_norm": 1.958350267089358,
+      "learning_rate": 2.2445372742189334e-05,
+      "logits/chosen": -6.375,
+      "logits/rejected": -4.71875,
+      "logps/chosen": -1376.0,
+      "logps/rejected": -2032.0,
+      "loss": 0.1178,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -9.5625,
+      "rewards/margins": 5.15625,
+      "rewards/rejected": -14.75,
+      "step": 1660
+    },
+    {
+      "epoch": 1.7486910994764397,
+      "grad_norm": 2.4190744561338517,
+      "learning_rate": 2.214251060741528e-05,
+      "logits/chosen": -6.40625,
+      "logits/rejected": -5.03125,
+      "logps/chosen": -1440.0,
+      "logps/rejected": -2144.0,
+      "loss": 0.1162,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -10.0625,
+      "rewards/margins": 5.8125,
+      "rewards/rejected": -15.875,
+      "step": 1670
+    },
+    {
+      "epoch": 1.7591623036649215,
+      "grad_norm": 3.1819690697281726,
+      "learning_rate": 2.1840072811859302e-05,
+      "logits/chosen": -6.09375,
+      "logits/rejected": -4.9375,
+      "logps/chosen": -1416.0,
+      "logps/rejected": -1968.0,
+      "loss": 0.147,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -9.9375,
+      "rewards/margins": 4.75,
+      "rewards/rejected": -14.6875,
+      "step": 1680
+    },
+    {
+      "epoch": 1.7696335078534031,
+      "grad_norm": 1.4767462518688441,
+      "learning_rate": 2.1538104267750287e-05,
+      "logits/chosen": -5.9375,
+      "logits/rejected": -4.75,
+      "logps/chosen": -1240.0,
+      "logps/rejected": -1768.0,
+      "loss": 0.1076,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.5,
+      "rewards/margins": 4.3125,
+      "rewards/rejected": -12.8125,
+      "step": 1690
+    },
+    {
+      "epoch": 1.7801047120418847,
+      "grad_norm": 1.6346180738709883,
+      "learning_rate": 2.1236649817632952e-05,
+      "logits/chosen": -6.28125,
+      "logits/rejected": -5.21875,
+      "logps/chosen": -1208.0,
+      "logps/rejected": -1760.0,
+      "loss": 0.1028,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -8.5,
+      "rewards/margins": 4.25,
+      "rewards/rejected": -12.75,
+      "step": 1700
+    },
+    {
+      "epoch": 1.7801047120418847,
+      "eval_logits/chosen": -3.625,
+      "eval_logits/rejected": -1.703125,
+      "eval_logps/chosen": -1424.0,
+      "eval_logps/rejected": -1512.0,
+      "eval_loss": 1.1308437585830688,
+      "eval_rewards/accuracies": 0.623046875,
+      "eval_rewards/chosen": -11.0,
+      "eval_rewards/margins": 1.234375,
+      "eval_rewards/rejected": -12.25,
+      "eval_runtime": 46.6437,
+      "eval_samples_per_second": 42.878,
+      "eval_steps_per_second": 0.686,
+      "step": 1700
+    },
+    {
+      "epoch": 1.7905759162303665,
+      "grad_norm": 2.2977809419919315,
+      "learning_rate": 2.093575422770872e-05,
+      "logits/chosen": -4.90625,
+      "logits/rejected": -3.78125,
+      "logps/chosen": -1320.0,
+      "logps/rejected": -1920.0,
+      "loss": 0.1092,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -9.4375,
+      "rewards/margins": 5.0625,
+      "rewards/rejected": -14.5,
+      "step": 1710
+    },
+    {
+      "epoch": 1.8010471204188483,
+      "grad_norm": 3.2986929081463185,
+      "learning_rate": 2.0635462181187827e-05,
+      "logits/chosen": -5.09375,
+      "logits/rejected": -3.765625,
+      "logps/chosen": -1512.0,
+      "logps/rejected": -2272.0,
+      "loss": 0.1379,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -10.9375,
+      "rewards/margins": 5.71875,
+      "rewards/rejected": -16.625,
+      "step": 1720
+    },
+    {
+      "epoch": 1.8115183246073299,
+      "grad_norm": 2.3041951640716314,
+      "learning_rate": 2.0335818271653917e-05,
+      "logits/chosen": -5.0,
+      "logits/rejected": -3.78125,
+      "logps/chosen": -1392.0,
+      "logps/rejected": -1928.0,
+      "loss": 0.1347,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.75,
+      "rewards/margins": 4.5,
+      "rewards/rejected": -14.25,
+      "step": 1730
+    },
+    {
+      "epoch": 1.8219895287958114,
+      "grad_norm": 1.882100776860651,
+      "learning_rate": 2.0036866996441815e-05,
+      "logits/chosen": -6.78125,
+      "logits/rejected": -4.75,
+      "logps/chosen": -1272.0,
+      "logps/rejected": -1816.0,
+      "loss": 0.1259,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.6875,
+      "rewards/margins": 4.6875,
+      "rewards/rejected": -13.3125,
+      "step": 1740
+    },
+    {
+      "epoch": 1.8324607329842932,
+      "grad_norm": 2.876926509381833,
+      "learning_rate": 1.9738652750029677e-05,
+      "logits/chosen": -5.75,
+      "logits/rejected": -4.21875,
+      "logps/chosen": -1336.0,
+      "logps/rejected": -1880.0,
+      "loss": 0.1126,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -9.0625,
+      "rewards/margins": 4.5625,
+      "rewards/rejected": -13.625,
+      "step": 1750
+    },
+    {
+      "epoch": 1.8429319371727748,
+      "grad_norm": 3.190643645172407,
+      "learning_rate": 1.944121981744631e-05,
+      "logits/chosen": -6.53125,
+      "logits/rejected": -5.09375,
+      "logps/chosen": -1416.0,
+      "logps/rejected": -1984.0,
+      "loss": 0.1348,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -9.5625,
+      "rewards/margins": 4.6875,
+      "rewards/rejected": -14.25,
+      "step": 1760
+    },
+    {
+      "epoch": 1.8534031413612566,
+      "grad_norm": 2.0355047401087405,
+      "learning_rate": 1.9144612367694887e-05,
+      "logits/chosen": -6.84375,
+      "logits/rejected": -5.15625,
+      "logps/chosen": -1296.0,
+      "logps/rejected": -1912.0,
+      "loss": 0.1113,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.8125,
+      "rewards/margins": 4.71875,
+      "rewards/rejected": -13.5625,
+      "step": 1770
+    },
+    {
+      "epoch": 1.8638743455497382,
+      "grad_norm": 3.0485713847419023,
+      "learning_rate": 1.8848874447193803e-05,
+      "logits/chosen": -5.6875,
+      "logits/rejected": -4.5,
+      "logps/chosen": -1288.0,
+      "logps/rejected": -1808.0,
+      "loss": 0.1335,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -9.0,
+      "rewards/margins": 4.5625,
+      "rewards/rejected": -13.5625,
+      "step": 1780
+    },
+    {
+      "epoch": 1.8743455497382198,
+      "grad_norm": 2.3203017320605106,
+      "learning_rate": 1.855404997323571e-05,
+      "logits/chosen": -6.53125,
+      "logits/rejected": -5.0625,
+      "logps/chosen": -1304.0,
+      "logps/rejected": -1896.0,
+      "loss": 0.1254,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -8.9375,
+      "rewards/margins": 4.75,
+      "rewards/rejected": -13.6875,
+      "step": 1790
+    },
+    {
+      "epoch": 1.8848167539267016,
+      "grad_norm": 1.939657098205315,
+      "learning_rate": 1.82601827274658e-05,
+      "logits/chosen": -7.40625,
+      "logits/rejected": -5.6875,
+      "logps/chosen": -1168.0,
+      "logps/rejected": -1704.0,
+      "loss": 0.1227,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -7.84375,
+      "rewards/margins": 4.59375,
+      "rewards/rejected": -12.4375,
+      "step": 1800
+    },
+    {
+      "epoch": 1.8848167539267016,
+      "eval_logits/chosen": -4.375,
+      "eval_logits/rejected": -2.515625,
+      "eval_logps/chosen": -1264.0,
+      "eval_logps/rejected": -1336.0,
+      "eval_loss": 1.0561875104904175,
+      "eval_rewards/accuracies": 0.611328125,
+      "eval_rewards/chosen": -9.5,
+      "eval_rewards/margins": 0.99609375,
+      "eval_rewards/rejected": -10.5,
+      "eval_runtime": 46.6378,
+      "eval_samples_per_second": 42.884,
+      "eval_steps_per_second": 0.686,
+      "step": 1800
+    },
+    {
+      "epoch": 1.8952879581151834,
+      "grad_norm": 3.307246733433114,
+      "learning_rate": 1.7967316349380213e-05,
+      "logits/chosen": -6.1875,
+      "logits/rejected": -4.90625,
+      "logps/chosen": -1208.0,
+      "logps/rejected": -1752.0,
+      "loss": 0.1175,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -8.25,
+      "rewards/margins": 4.46875,
+      "rewards/rejected": -12.75,
+      "step": 1810
+    },
+    {
+      "epoch": 1.905759162303665,
+      "grad_norm": 2.556689093178943,
+      "learning_rate": 1.7675494329845516e-05,
+      "logits/chosen": -5.09375,
+      "logits/rejected": -3.453125,
+      "logps/chosen": -1304.0,
+      "logps/rejected": -1872.0,
+      "loss": 0.1214,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -9.0625,
+      "rewards/margins": 4.78125,
+      "rewards/rejected": -13.8125,
+      "step": 1820
+    },
+    {
+      "epoch": 1.9162303664921465,
+      "grad_norm": 1.5700886379294252,
+      "learning_rate": 1.738476000464026e-05,
+      "logits/chosen": -4.46875,
+      "logits/rejected": -3.09375,
+      "logps/chosen": -1360.0,
+      "logps/rejected": -2000.0,
+      "loss": 0.1272,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -9.5625,
+      "rewards/margins": 5.1875,
+      "rewards/rejected": -14.75,
+      "step": 1830
+    },
+    {
+      "epoch": 1.9267015706806283,
+      "grad_norm": 1.641614792050765,
+      "learning_rate": 1.7095156548019646e-05,
+      "logits/chosen": -5.21875,
+      "logits/rejected": -4.09375,
+      "logps/chosen": -1312.0,
+      "logps/rejected": -1992.0,
+      "loss": 0.0992,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -9.25,
+      "rewards/margins": 5.21875,
+      "rewards/rejected": -14.5,
+      "step": 1840
+    },
+    {
+      "epoch": 1.93717277486911,
+      "grad_norm": 2.2020116638544858,
+      "learning_rate": 1.6806726966304064e-05,
+      "logits/chosen": -6.15625,
+      "logits/rejected": -4.75,
+      "logps/chosen": -1312.0,
+      "logps/rejected": -1880.0,
+      "loss": 0.1283,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -8.75,
+      "rewards/margins": 4.875,
+      "rewards/rejected": -13.625,
+      "step": 1850
+    },
+    {
+      "epoch": 1.9476439790575917,
+      "grad_norm": 3.6806777570651836,
+      "learning_rate": 1.6519514091492625e-05,
+      "logits/chosen": -6.1875,
+      "logits/rejected": -4.65625,
+      "logps/chosen": -1240.0,
+      "logps/rejected": -1784.0,
+      "loss": 0.1485,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -8.6875,
+      "rewards/margins": 4.4375,
+      "rewards/rejected": -13.125,
+      "step": 1860
+    },
+    {
+      "epoch": 1.9581151832460733,
+      "grad_norm": 2.4683894236598647,
+      "learning_rate": 1.6233560574902627e-05,
+      "logits/chosen": -6.21875,
+      "logits/rejected": -4.875,
+      "logps/chosen": -1208.0,
+      "logps/rejected": -1792.0,
+      "loss": 0.1229,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -8.5,
+      "rewards/margins": 4.71875,
+      "rewards/rejected": -13.1875,
+      "step": 1870
+    },
+    {
+      "epoch": 1.9685863874345548,
+      "grad_norm": 1.8432384006404061,
+      "learning_rate": 1.5948908880835753e-05,
+      "logits/chosen": -6.1875,
+      "logits/rejected": -4.4375,
+      "logps/chosen": -1240.0,
+      "logps/rejected": -1792.0,
+      "loss": 0.1205,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -8.375,
+      "rewards/margins": 4.875,
+      "rewards/rejected": -13.25,
+      "step": 1880
+    },
+    {
+      "epoch": 1.9790575916230366,
+      "grad_norm": 2.8078625046549783,
+      "learning_rate": 1.5665601280272124e-05,
+      "logits/chosen": -6.09375,
+      "logits/rejected": -4.40625,
+      "logps/chosen": -1296.0,
+      "logps/rejected": -1776.0,
+      "loss": 0.1267,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -9.0,
+      "rewards/margins": 4.1875,
+      "rewards/rejected": -13.1875,
+      "step": 1890
+    },
+    {
+      "epoch": 1.9895287958115184,
+      "grad_norm": 2.930621737497199,
+      "learning_rate": 1.538367984459298e-05,
+      "logits/chosen": -5.09375,
+      "logits/rejected": -3.25,
+      "logps/chosen": -1280.0,
+      "logps/rejected": -1872.0,
+      "loss": 0.1119,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -9.0,
+      "rewards/margins": 5.0625,
+      "rewards/rejected": -14.0625,
+      "step": 1900
+    },
+    {
+      "epoch": 1.9895287958115184,
+      "eval_logits/chosen": -3.109375,
+      "eval_logits/rejected": -1.2734375,
+      "eval_logps/chosen": -1448.0,
+      "eval_logps/rejected": -1544.0,
+      "eval_loss": 1.1632968187332153,
+      "eval_rewards/accuracies": 0.599609375,
+      "eval_rewards/chosen": -11.3125,
+      "eval_rewards/margins": 1.234375,
+      "eval_rewards/rejected": -12.5625,
+      "eval_runtime": 46.6424,
+      "eval_samples_per_second": 42.879,
+      "eval_steps_per_second": 0.686,
+      "step": 1900
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.6684214980595502,
+      "learning_rate": 1.5103186439333133e-05,
+      "logits/chosen": -4.59375,
+      "logits/rejected": -3.15625,
+      "logps/chosen": -1272.0,
+      "logps/rejected": -1952.0,
+      "loss": 0.1129,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -8.9375,
+      "rewards/margins": 5.3125,
+      "rewards/rejected": -14.25,
+      "step": 1910
+    },
+    {
+      "epoch": 2.0104712041884816,
+      "grad_norm": 0.13112525115152174,
+      "learning_rate": 1.4824162717963829e-05,
+      "logits/chosen": -4.90625,
+      "logits/rejected": -2.140625,
+      "logps/chosen": -1224.0,
+      "logps/rejected": -1944.0,
+      "loss": 0.0193,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -8.3125,
+      "rewards/margins": 6.65625,
+      "rewards/rejected": -14.9375,
+      "step": 1920
+    },
+    {
+      "epoch": 2.020942408376963,
+      "grad_norm": 1.189860298324291,
+      "learning_rate": 1.4546650115707192e-05,
+      "logits/chosen": -4.03125,
+      "logits/rejected": -2.1875,
+      "logps/chosen": -1280.0,
+      "logps/rejected": -2096.0,
+      "loss": 0.0188,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -9.0625,
+      "rewards/margins": 7.21875,
+      "rewards/rejected": -16.25,
+      "step": 1930
+    },
+    {
+      "epoch": 2.031413612565445,
+      "grad_norm": 0.14484999474340313,
+      "learning_rate": 1.4270689843383111e-05,
+      "logits/chosen": -3.0,
+      "logits/rejected": -0.71875,
+      "logps/chosen": -1368.0,
+      "logps/rejected": -2192.0,
+      "loss": 0.0242,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -9.5625,
+      "rewards/margins": 7.28125,
+      "rewards/rejected": -16.875,
+      "step": 1940
+    },
+    {
+      "epoch": 2.0418848167539267,
+      "grad_norm": 1.6217742970691786,
+      "learning_rate": 1.3996322881289347e-05,
+      "logits/chosen": -3.453125,
+      "logits/rejected": -1.0,
+      "logps/chosen": -1384.0,
+      "logps/rejected": -2256.0,
+      "loss": 0.0259,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -10.125,
+      "rewards/margins": 7.59375,
+      "rewards/rejected": -17.75,
+      "step": 1950
+    },
+    {
+      "epoch": 2.0523560209424083,
+      "grad_norm": 0.11021242979116014,
+      "learning_rate": 1.3723589973115963e-05,
+      "logits/chosen": -2.515625,
+      "logits/rejected": -0.51171875,
+      "logps/chosen": -1424.0,
+      "logps/rejected": -2272.0,
+      "loss": 0.0179,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -10.5,
+      "rewards/margins": 7.65625,
+      "rewards/rejected": -18.125,
+      "step": 1960
+    },
+    {
+      "epoch": 2.06282722513089,
+      "grad_norm": 0.5628366758197821,
+      "learning_rate": 1.3452531619894837e-05,
+      "logits/chosen": -2.375,
+      "logits/rejected": -0.68359375,
+      "logps/chosen": -1472.0,
+      "logps/rejected": -2448.0,
+      "loss": 0.0157,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -10.875,
+      "rewards/margins": 8.1875,
+      "rewards/rejected": -19.0,
+      "step": 1970
+    },
+    {
+      "epoch": 2.073298429319372,
+      "grad_norm": 0.2900776842492875,
+      "learning_rate": 1.3183188073985245e-05,
+      "logits/chosen": -3.265625,
+      "logits/rejected": -1.1640625,
+      "logps/chosen": -1496.0,
+      "logps/rejected": -2512.0,
+      "loss": 0.0232,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -11.125,
+      "rewards/margins": 8.875,
+      "rewards/rejected": -20.0,
+      "step": 1980
+    },
+    {
+      "epoch": 2.0837696335078535,
+      "grad_norm": 0.26368985361388025,
+      "learning_rate": 1.2915599333096351e-05,
+      "logits/chosen": -3.46875,
+      "logits/rejected": -1.75,
+      "logps/chosen": -1520.0,
+      "logps/rejected": -2384.0,
+      "loss": 0.0176,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -11.0,
+      "rewards/margins": 7.8125,
+      "rewards/rejected": -18.875,
+      "step": 1990
+    },
+    {
+      "epoch": 2.094240837696335,
+      "grad_norm": 0.23935043680853396,
+      "learning_rate": 1.264980513434752e-05,
+      "logits/chosen": -3.390625,
+      "logits/rejected": -1.1640625,
+      "logps/chosen": -1504.0,
+      "logps/rejected": -2416.0,
+      "loss": 0.0166,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -10.625,
+      "rewards/margins": 8.125,
+      "rewards/rejected": -18.75,
+      "step": 2000
+    },
+    {
+      "epoch": 2.094240837696335,
+      "eval_logits/chosen": 0.054931640625,
+      "eval_logits/rejected": 1.9375,
+      "eval_logps/chosen": -1744.0,
+      "eval_logps/rejected": -1864.0,
+      "eval_loss": 1.5514531135559082,
+      "eval_rewards/accuracies": 0.583984375,
+      "eval_rewards/chosen": -14.25,
+      "eval_rewards/margins": 1.5234375,
+      "eval_rewards/rejected": -15.75,
+      "eval_runtime": 46.6515,
+      "eval_samples_per_second": 42.871,
+      "eval_steps_per_second": 0.686,
+      "step": 2000
+    },
+    {
+      "epoch": 2.1047120418848166,
+      "grad_norm": 0.18125247281437176,
+      "learning_rate": 1.2385844948367324e-05,
+      "logits/chosen": -2.828125,
+      "logits/rejected": -0.875,
+      "logps/chosen": -1512.0,
+      "logps/rejected": -2336.0,
+      "loss": 0.0341,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -11.4375,
+      "rewards/margins": 7.625,
+      "rewards/rejected": -19.0,
+      "step": 2010
+    },
+    {
+      "epoch": 2.115183246073298,
+      "grad_norm": 0.29229470864014834,
+      "learning_rate": 1.2123757973432114e-05,
+      "logits/chosen": -3.4375,
+      "logits/rejected": -1.4296875,
+      "logps/chosen": -1520.0,
+      "logps/rejected": -2336.0,
+      "loss": 0.0148,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -10.9375,
+      "rewards/margins": 7.625,
+      "rewards/rejected": -18.625,
+      "step": 2020
+    },
+    {
+      "epoch": 2.1256544502617802,
+      "grad_norm": 0.30795366200482666,
+      "learning_rate": 1.1863583129645106e-05,
+      "logits/chosen": -3.078125,
+      "logits/rejected": -1.1953125,
+      "logps/chosen": -1488.0,
+      "logps/rejected": -2448.0,
+      "loss": 0.0222,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -10.9375,
+      "rewards/margins": 8.4375,
+      "rewards/rejected": -19.375,
+      "step": 2030
+    },
+    {
+      "epoch": 2.136125654450262,
+      "grad_norm": 0.23652450702189734,
+      "learning_rate": 1.1605359053156607e-05,
+      "logits/chosen": -3.234375,
+      "logits/rejected": -1.0703125,
+      "logps/chosen": -1584.0,
+      "logps/rejected": -2560.0,
+      "loss": 0.0144,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -11.625,
+      "rewards/margins": 8.5,
+      "rewards/rejected": -20.125,
+      "step": 2040
+    },
+    {
+      "epoch": 2.1465968586387434,
+      "grad_norm": 0.30147902289963274,
+      "learning_rate": 1.1349124090426644e-05,
+      "logits/chosen": -2.9375,
+      "logits/rejected": -0.65625,
+      "logps/chosen": -1544.0,
+      "logps/rejected": -2560.0,
+      "loss": 0.0206,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -11.375,
+      "rewards/margins": 9.3125,
+      "rewards/rejected": -20.625,
+      "step": 2050
+    },
+    {
+      "epoch": 2.157068062827225,
+      "grad_norm": 0.5500534888475488,
+      "learning_rate": 1.1094916292530406e-05,
+      "logits/chosen": -2.8125,
+      "logits/rejected": -1.25,
+      "logps/chosen": -1480.0,
+      "logps/rejected": -2416.0,
+      "loss": 0.0158,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -10.625,
+      "rewards/margins": 8.375,
+      "rewards/rejected": -19.0,
+      "step": 2060
+    },
+    {
+      "epoch": 2.167539267015707,
+      "grad_norm": 0.2543616840160018,
+      "learning_rate": 1.0842773409507623e-05,
+      "logits/chosen": -3.046875,
+      "logits/rejected": -0.6796875,
+      "logps/chosen": -1432.0,
+      "logps/rejected": -2384.0,
+      "loss": 0.0189,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -10.625,
+      "rewards/margins": 8.4375,
+      "rewards/rejected": -19.125,
+      "step": 2070
+    },
+    {
+      "epoch": 2.1780104712041886,
+      "grad_norm": 1.547753200569628,
+      "learning_rate": 1.0592732884756754e-05,
+      "logits/chosen": -3.265625,
+      "logits/rejected": -1.0390625,
+      "logps/chosen": -1632.0,
+      "logps/rejected": -2704.0,
+      "loss": 0.0167,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -12.125,
+      "rewards/margins": 9.25,
+      "rewards/rejected": -21.375,
+      "step": 2080
+    },
+    {
+      "epoch": 2.18848167539267,
+      "grad_norm": 0.09366111619278192,
+      "learning_rate": 1.0344831849474507e-05,
+      "logits/chosen": -2.40625,
+      "logits/rejected": -0.12255859375,
+      "logps/chosen": -1680.0,
+      "logps/rejected": -2672.0,
+      "loss": 0.0181,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -12.625,
+      "rewards/margins": 9.0,
+      "rewards/rejected": -21.625,
+      "step": 2090
+    },
+    {
+      "epoch": 2.1989528795811517,
+      "grad_norm": 0.2173202996713778,
+      "learning_rate": 1.0099107117141879e-05,
+      "logits/chosen": -2.796875,
+      "logits/rejected": -0.6640625,
+      "logps/chosen": -1536.0,
+      "logps/rejected": -2528.0,
+      "loss": 0.0122,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -11.625,
+      "rewards/margins": 8.75,
+      "rewards/rejected": -20.375,
+      "step": 2100
+    },
+    {
+      "epoch": 2.1989528795811517,
+      "eval_logits/chosen": -0.412109375,
+      "eval_logits/rejected": 1.5703125,
+      "eval_logps/chosen": -1800.0,
+      "eval_logps/rejected": -1936.0,
+      "eval_loss": 1.669374942779541,
+      "eval_rewards/accuracies": 0.609375,
+      "eval_rewards/chosen": -14.8125,
+      "eval_rewards/margins": 1.6484375,
+      "eval_rewards/rejected": -16.5,
+      "eval_runtime": 46.6466,
+      "eval_samples_per_second": 42.876,
+      "eval_steps_per_second": 0.686,
+      "step": 2100
+    },
+    {
+      "epoch": 2.2094240837696333,
+      "grad_norm": 0.5829882131071631,
+      "learning_rate": 9.855595178057334e-06,
+      "logits/chosen": -3.421875,
+      "logits/rejected": -1.4453125,
+      "logps/chosen": -1568.0,
+      "logps/rejected": -2592.0,
+      "loss": 0.0152,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -11.4375,
+      "rewards/margins": 9.1875,
+      "rewards/rejected": -20.625,
+      "step": 2110
+    },
+    {
+      "epoch": 2.2198952879581153,
+      "grad_norm": 0.2424013055067862,
+      "learning_rate": 9.614332193917949e-06,
+      "logits/chosen": -3.53125,
+      "logits/rejected": -0.76171875,
+      "logps/chosen": -1696.0,
+      "logps/rejected": -2720.0,
+      "loss": 0.0168,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -12.5625,
+      "rewards/margins": 9.25,
+      "rewards/rejected": -21.875,
+      "step": 2120
+    },
+    {
+      "epoch": 2.230366492146597,
+      "grad_norm": 0.27028475654174455,
+      "learning_rate": 9.375353992449385e-06,
+      "logits/chosen": -3.34375,
+      "logits/rejected": -1.5078125,
+      "logps/chosen": -1640.0,
+      "logps/rejected": -2704.0,
+      "loss": 0.0127,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -12.4375,
+      "rewards/margins": 9.5,
+      "rewards/rejected": -22.0,
+      "step": 2130
+    },
+    {
+      "epoch": 2.2408376963350785,
+      "grad_norm": 0.8848690452153692,
+      "learning_rate": 9.138696062085442e-06,
+      "logits/chosen": -3.203125,
+      "logits/rejected": -0.94140625,
+      "logps/chosen": -1616.0,
+      "logps/rejected": -2608.0,
+      "loss": 0.025,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -12.1875,
+      "rewards/margins": 9.0625,
+      "rewards/rejected": -21.25,
+      "step": 2140
+    },
+    {
+      "epoch": 2.25130890052356,
+      "grad_norm": 0.4601370478678084,
+      "learning_rate": 8.904393546698007e-06,
+      "logits/chosen": -2.96875,
+      "logits/rejected": -1.234375,
+      "logps/chosen": -1608.0,
+      "logps/rejected": -2624.0,
+      "loss": 0.016,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -12.0,
+      "rewards/margins": 9.125,
+      "rewards/rejected": -21.125,
+      "step": 2150
+    },
+    {
+      "epoch": 2.261780104712042,
+      "grad_norm": 0.2803650810497035,
+      "learning_rate": 8.672481240378142e-06,
+      "logits/chosen": -2.40625,
+      "logits/rejected": -0.302734375,
+      "logps/chosen": -1520.0,
+      "logps/rejected": -2480.0,
+      "loss": 0.0238,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -11.5625,
+      "rewards/margins": 8.375,
+      "rewards/rejected": -19.875,
+      "step": 2160
+    },
+    {
+      "epoch": 2.2722513089005236,
+      "grad_norm": 0.3295121660857614,
+      "learning_rate": 8.44299358226919e-06,
+      "logits/chosen": -2.328125,
+      "logits/rejected": 0.2734375,
+      "logps/chosen": -1592.0,
+      "logps/rejected": -2544.0,
+      "loss": 0.0257,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -12.0,
+      "rewards/margins": 8.625,
+      "rewards/rejected": -20.625,
+      "step": 2170
+    },
+    {
+      "epoch": 2.282722513089005,
+      "grad_norm": 0.35860568378176194,
+      "learning_rate": 8.215964651452457e-06,
+      "logits/chosen": -2.265625,
+      "logits/rejected": 0.0341796875,
+      "logps/chosen": -1576.0,
+      "logps/rejected": -2592.0,
+      "loss": 0.0143,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -11.8125,
+      "rewards/margins": 8.8125,
+      "rewards/rejected": -20.625,
+      "step": 2180
+    },
+    {
+      "epoch": 2.2931937172774868,
+      "grad_norm": 0.15758545486146067,
+      "learning_rate": 7.991428161886502e-06,
+      "logits/chosen": -2.546875,
+      "logits/rejected": -0.43359375,
+      "logps/chosen": -1632.0,
+      "logps/rejected": -2640.0,
+      "loss": 0.013,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -12.0,
+      "rewards/margins": 9.1875,
+      "rewards/rejected": -21.25,
+      "step": 2190
+    },
+    {
+      "epoch": 2.303664921465969,
+      "grad_norm": 0.3540106891584038,
+      "learning_rate": 7.76941745740061e-06,
+      "logits/chosen": -2.578125,
+      "logits/rejected": -0.56640625,
+      "logps/chosen": -1712.0,
+      "logps/rejected": -2768.0,
+      "loss": 0.0158,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -12.9375,
+      "rewards/margins": 9.4375,
+      "rewards/rejected": -22.375,
+      "step": 2200
+    },
+    {
+      "epoch": 2.303664921465969,
+      "eval_logits/chosen": 0.64453125,
+      "eval_logits/rejected": 2.515625,
+      "eval_logps/chosen": -1904.0,
+      "eval_logps/rejected": -2040.0,
+      "eval_loss": 1.791171908378601,
+      "eval_rewards/accuracies": 0.603515625,
+      "eval_rewards/chosen": -15.8125,
+      "eval_rewards/margins": 1.6875,
+      "eval_rewards/rejected": -17.5,
+      "eval_runtime": 46.6495,
+      "eval_samples_per_second": 42.873,
+      "eval_steps_per_second": 0.686,
+      "step": 2200
+    },
+    {
+      "epoch": 2.3141361256544504,
+      "grad_norm": 0.2401128865298814,
+      "learning_rate": 7.549965506743123e-06,
+      "logits/chosen": -1.9140625,
+      "logits/rejected": -0.13671875,
+      "logps/chosen": -1624.0,
+      "logps/rejected": -2560.0,
+      "loss": 0.0231,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -12.375,
+      "rewards/margins": 8.75,
+      "rewards/rejected": -21.125,
+      "step": 2210
+    },
+    {
+      "epoch": 2.324607329842932,
+      "grad_norm": 0.0730586595349473,
+      "learning_rate": 7.333104898685631e-06,
+      "logits/chosen": -2.625,
+      "logits/rejected": -0.7890625,
+      "logps/chosen": -1560.0,
+      "logps/rejected": -2560.0,
+      "loss": 0.0183,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -11.625,
+      "rewards/margins": 8.75,
+      "rewards/rejected": -20.375,
+      "step": 2220
+    },
+    {
+      "epoch": 2.3350785340314135,
+      "grad_norm": 0.1684457172881231,
+      "learning_rate": 7.118867837183518e-06,
+      "logits/chosen": -3.265625,
+      "logits/rejected": -0.96484375,
+      "logps/chosen": -1592.0,
+      "logps/rejected": -2592.0,
+      "loss": 0.0184,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -11.1875,
+      "rewards/margins": 9.3125,
+      "rewards/rejected": -20.5,
+      "step": 2230
+    },
+    {
+      "epoch": 2.345549738219895,
+      "grad_norm": 1.0109715083282962,
+      "learning_rate": 6.9072861365936045e-06,
+      "logits/chosen": -3.34375,
+      "logits/rejected": -1.3671875,
+      "logps/chosen": -1600.0,
+      "logps/rejected": -2640.0,
+      "loss": 0.0151,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -11.375,
+      "rewards/margins": 8.9375,
+      "rewards/rejected": -20.375,
+      "step": 2240
+    },
+    {
+      "epoch": 2.356020942408377,
+      "grad_norm": 0.8886833189304637,
+      "learning_rate": 6.6983912169497005e-06,
+      "logits/chosen": -2.515625,
+      "logits/rejected": -0.3359375,
+      "logps/chosen": -1544.0,
+      "logps/rejected": -2496.0,
+      "loss": 0.0192,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -11.0625,
+      "rewards/margins": 8.75,
+      "rewards/rejected": -19.75,
+      "step": 2250
+    },
+    {
+      "epoch": 2.3664921465968587,
+      "grad_norm": 0.13708274798450046,
+      "learning_rate": 6.49221409929677e-06,
+      "logits/chosen": -1.359375,
+      "logits/rejected": 0.69921875,
+      "logps/chosen": -1512.0,
+      "logps/rejected": -2496.0,
+      "loss": 0.0113,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -11.3125,
+      "rewards/margins": 8.875,
+      "rewards/rejected": -20.25,
+      "step": 2260
+    },
+    {
+      "epoch": 2.3769633507853403,
+      "grad_norm": 1.8834904381998978,
+      "learning_rate": 6.2887854010842075e-06,
+      "logits/chosen": -1.875,
+      "logits/rejected": -0.62890625,
+      "logps/chosen": -1560.0,
+      "logps/rejected": -2576.0,
+      "loss": 0.0152,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -11.6875,
+      "rewards/margins": 9.0625,
+      "rewards/rejected": -20.75,
+      "step": 2270
+    },
+    {
+      "epoch": 2.387434554973822,
+      "grad_norm": 0.33036446598709845,
+      "learning_rate": 6.0881353316191385e-06,
+      "logits/chosen": -2.3125,
+      "logits/rejected": -0.8046875,
+      "logps/chosen": -1608.0,
+      "logps/rejected": -2544.0,
+      "loss": 0.0151,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -11.875,
+      "rewards/margins": 8.5625,
+      "rewards/rejected": -20.375,
+      "step": 2280
+    },
+    {
+      "epoch": 2.3979057591623034,
+      "grad_norm": 0.5536309572727743,
+      "learning_rate": 5.890293687580381e-06,
+      "logits/chosen": -1.8671875,
+      "logits/rejected": 0.2060546875,
+      "logps/chosen": -1616.0,
+      "logps/rejected": -2608.0,
+      "loss": 0.0174,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -12.0,
+      "rewards/margins": 9.1875,
+      "rewards/rejected": -21.125,
+      "step": 2290
+    },
+    {
+      "epoch": 2.4083769633507854,
+      "grad_norm": 0.10988173112717037,
+      "learning_rate": 5.695289848593532e-06,
+      "logits/chosen": -2.125,
+      "logits/rejected": -0.9609375,
+      "logps/chosen": -1632.0,
+      "logps/rejected": -2608.0,
+      "loss": 0.0165,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -12.25,
+      "rewards/margins": 8.9375,
+      "rewards/rejected": -21.25,
+      "step": 2300
+    },
+    {
+      "epoch": 2.4083769633507854,
+      "eval_logits/chosen": 1.1171875,
+      "eval_logits/rejected": 2.921875,
+      "eval_logps/chosen": -1896.0,
+      "eval_logps/rejected": -2024.0,
+      "eval_loss": 1.781656265258789,
+      "eval_rewards/accuracies": 0.59375,
+      "eval_rewards/chosen": -15.8125,
+      "eval_rewards/margins": 1.5703125,
+      "eval_rewards/rejected": -17.375,
+      "eval_runtime": 46.6487,
+      "eval_samples_per_second": 42.874,
+      "eval_steps_per_second": 0.686,
+      "step": 2300
+    },
+    {
+      "epoch": 2.418848167539267,
+      "grad_norm": 0.14844038313112592,
+      "learning_rate": 5.50315277286812e-06,
+      "logits/chosen": -2.359375,
+      "logits/rejected": -0.61328125,
+      "logps/chosen": -1608.0,
+      "logps/rejected": -2576.0,
+      "loss": 0.0191,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -12.375,
+      "rewards/margins": 8.9375,
+      "rewards/rejected": -21.25,
+      "step": 2310
+    },
+    {
+      "epoch": 2.4293193717277486,
+      "grad_norm": 0.43936779227361,
+      "learning_rate": 5.313910992897281e-06,
+      "logits/chosen": -1.3125,
+      "logits/rejected": 0.5234375,
+      "logps/chosen": -1544.0,
+      "logps/rejected": -2608.0,
+      "loss": 0.0154,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -11.5,
+      "rewards/margins": 9.25,
+      "rewards/rejected": -20.75,
+      "step": 2320
+    },
+    {
+      "epoch": 2.4397905759162306,
+      "grad_norm": 1.0196548089785136,
+      "learning_rate": 5.127592611220658e-06,
+      "logits/chosen": -2.40625,
+      "logits/rejected": -0.34765625,
+      "logps/chosen": -1664.0,
+      "logps/rejected": -2752.0,
+      "loss": 0.0263,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -12.375,
+      "rewards/margins": 9.6875,
+      "rewards/rejected": -22.0,
+      "step": 2330
+    },
+    {
+      "epoch": 2.450261780104712,
+      "grad_norm": 0.4218483763721506,
+      "learning_rate": 4.944225296251159e-06,
+      "logits/chosen": -2.125,
+      "logits/rejected": -0.58984375,
+      "logps/chosen": -1528.0,
+      "logps/rejected": -2544.0,
+      "loss": 0.0127,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -11.75,
+      "rewards/margins": 9.0625,
+      "rewards/rejected": -20.875,
+      "step": 2340
+    },
+    {
+      "epoch": 2.4607329842931938,
+      "grad_norm": 3.0025945375518996,
+      "learning_rate": 4.763836278166172e-06,
+      "logits/chosen": -2.078125,
+      "logits/rejected": 0.07470703125,
+      "logps/chosen": -1584.0,
+      "logps/rejected": -2480.0,
+      "loss": 0.0187,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -11.8125,
+      "rewards/margins": 8.375,
+      "rewards/rejected": -20.125,
+      "step": 2350
+    },
+    {
+      "epoch": 2.4712041884816753,
+      "grad_norm": 0.4114665112435162,
+      "learning_rate": 4.586452344863876e-06,
+      "logits/chosen": -2.46875,
+      "logits/rejected": -0.7578125,
+      "logps/chosen": -1584.0,
+      "logps/rejected": -2560.0,
+      "loss": 0.0152,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -11.75,
+      "rewards/margins": 8.8125,
+      "rewards/rejected": -20.5,
+      "step": 2360
+    },
+    {
+      "epoch": 2.481675392670157,
+      "grad_norm": 0.7002478176984285,
+      "learning_rate": 4.412099837985192e-06,
+      "logits/chosen": -2.265625,
+      "logits/rejected": -0.08154296875,
+      "logps/chosen": -1624.0,
+      "logps/rejected": -2736.0,
+      "loss": 0.014,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -12.0,
+      "rewards/margins": 9.5,
+      "rewards/rejected": -21.5,
+      "step": 2370
+    },
+    {
+      "epoch": 2.492146596858639,
+      "grad_norm": 0.2414253837920038,
+      "learning_rate": 4.240804649002089e-06,
+      "logits/chosen": -2.703125,
+      "logits/rejected": -0.921875,
+      "logps/chosen": -1472.0,
+      "logps/rejected": -2496.0,
+      "loss": 0.0241,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -11.125,
+      "rewards/margins": 9.125,
+      "rewards/rejected": -20.25,
+      "step": 2380
+    },
+    {
+      "epoch": 2.5026178010471205,
+      "grad_norm": 0.4707372790690564,
+      "learning_rate": 4.072592215372598e-06,
+      "logits/chosen": -3.125,
+      "logits/rejected": -1.1875,
+      "logps/chosen": -1656.0,
+      "logps/rejected": -2624.0,
+      "loss": 0.0132,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -12.125,
+      "rewards/margins": 8.9375,
+      "rewards/rejected": -21.125,
+      "step": 2390
+    },
+    {
+      "epoch": 2.513089005235602,
+      "grad_norm": 0.6792175752753643,
+      "learning_rate": 3.907487516763389e-06,
+      "logits/chosen": -2.171875,
+      "logits/rejected": -0.388671875,
+      "logps/chosen": -1592.0,
+      "logps/rejected": -2528.0,
+      "loss": 0.0169,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -11.9375,
+      "rewards/margins": 8.6875,
+      "rewards/rejected": -20.625,
+      "step": 2400
+    },
+    {
+      "epoch": 2.513089005235602,
+      "eval_logits/chosen": 0.68359375,
+      "eval_logits/rejected": 2.515625,
+      "eval_logps/chosen": -1848.0,
+      "eval_logps/rejected": -1976.0,
+      "eval_loss": 1.7279062271118164,
+      "eval_rewards/accuracies": 0.59765625,
+      "eval_rewards/chosen": -15.25,
+      "eval_rewards/margins": 1.5625,
+      "eval_rewards/rejected": -16.875,
+      "eval_runtime": 46.657,
+      "eval_samples_per_second": 42.866,
+      "eval_steps_per_second": 0.686,
+      "step": 2400
+    },
+    {
+      "epoch": 2.5235602094240837,
+      "grad_norm": 0.5585669264048738,
+      "learning_rate": 3.745515071340286e-06,
+      "logits/chosen": -2.328125,
+      "logits/rejected": 0.220703125,
+      "logps/chosen": -1616.0,
+      "logps/rejected": -2624.0,
+      "loss": 0.0181,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -11.9375,
+      "rewards/margins": 9.125,
+      "rewards/rejected": -21.0,
+      "step": 2410
+    },
+    {
+      "epoch": 2.5340314136125652,
+      "grad_norm": 0.07493111846576456,
+      "learning_rate": 3.586698932127236e-06,
+      "logits/chosen": -3.125,
+      "logits/rejected": -0.76953125,
+      "logps/chosen": -1504.0,
+      "logps/rejected": -2432.0,
+      "loss": 0.0074,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -11.0,
+      "rewards/margins": 8.6875,
+      "rewards/rejected": -19.625,
+      "step": 2420
+    },
+    {
+      "epoch": 2.5445026178010473,
+      "grad_norm": 0.17910523942496434,
+      "learning_rate": 3.4310626834344744e-06,
+      "logits/chosen": -3.15625,
+      "logits/rejected": -1.5234375,
+      "logps/chosen": -1600.0,
+      "logps/rejected": -2560.0,
+      "loss": 0.0162,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -11.8125,
+      "rewards/margins": 8.6875,
+      "rewards/rejected": -20.5,
+      "step": 2430
+    },
+    {
+      "epoch": 2.554973821989529,
+      "grad_norm": 0.06575843343697423,
+      "learning_rate": 3.2786294373562343e-06,
+      "logits/chosen": -2.625,
+      "logits/rejected": -0.369140625,
+      "logps/chosen": -1568.0,
+      "logps/rejected": -2496.0,
+      "loss": 0.0084,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -11.5625,
+      "rewards/margins": 8.375,
+      "rewards/rejected": -19.875,
+      "step": 2440
+    },
+    {
+      "epoch": 2.5654450261780104,
+      "grad_norm": 0.45193754215926607,
+      "learning_rate": 3.1294218303385526e-06,
+      "logits/chosen": -2.453125,
+      "logits/rejected": -0.30859375,
+      "logps/chosen": -1624.0,
+      "logps/rejected": -2592.0,
+      "loss": 0.0153,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -11.875,
+      "rewards/margins": 8.5,
+      "rewards/rejected": -20.375,
+      "step": 2450
+    },
+    {
+      "epoch": 2.5759162303664924,
+      "grad_norm": 0.8057461914654177,
+      "learning_rate": 2.9834620198177483e-06,
+      "logits/chosen": -2.25,
+      "logits/rejected": -0.890625,
+      "logps/chosen": -1608.0,
+      "logps/rejected": -2576.0,
+      "loss": 0.0173,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -12.125,
+      "rewards/margins": 8.625,
+      "rewards/rejected": -20.75,
+      "step": 2460
+    },
+    {
+      "epoch": 2.5863874345549736,
+      "grad_norm": 0.2431117081850439,
+      "learning_rate": 2.840771680930068e-06,
+      "logits/chosen": -2.40625,
+      "logits/rejected": -0.8671875,
+      "logps/chosen": -1696.0,
+      "logps/rejected": -2800.0,
+      "loss": 0.0143,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -12.625,
+      "rewards/margins": 9.5,
+      "rewards/rejected": -22.125,
+      "step": 2470
+    },
+    {
+      "epoch": 2.5968586387434556,
+      "grad_norm": 0.24821789781009196,
+      "learning_rate": 2.701372003292868e-06,
+      "logits/chosen": -1.890625,
+      "logits/rejected": -0.2490234375,
+      "logps/chosen": -1544.0,
+      "logps/rejected": -2576.0,
+      "loss": 0.0133,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -11.5625,
+      "rewards/margins": 9.125,
+      "rewards/rejected": -20.625,
+      "step": 2480
+    },
+    {
+      "epoch": 2.607329842931937,
+      "grad_norm": 0.2847839899181593,
+      "learning_rate": 2.56528368785795e-06,
+      "logits/chosen": -2.71875,
+      "logits/rejected": -0.080078125,
+      "logps/chosen": -1552.0,
+      "logps/rejected": -2496.0,
+      "loss": 0.0127,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -11.3125,
+      "rewards/margins": 9.0625,
+      "rewards/rejected": -20.375,
+      "step": 2490
+    },
+    {
+      "epoch": 2.6178010471204187,
+      "grad_norm": 1.908192432045549,
+      "learning_rate": 2.432526943837471e-06,
+      "logits/chosen": -2.296875,
+      "logits/rejected": 0.412109375,
+      "logps/chosen": -1520.0,
+      "logps/rejected": -2608.0,
+      "loss": 0.0189,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -11.3125,
+      "rewards/margins": 9.3125,
+      "rewards/rejected": -20.625,
+      "step": 2500
+    },
+    {
+      "epoch": 2.6178010471204187,
+      "eval_logits/chosen": 0.83203125,
+      "eval_logits/rejected": 2.671875,
+      "eval_logps/chosen": -1856.0,
+      "eval_logps/rejected": -1992.0,
+      "eval_loss": 1.7442500591278076,
+      "eval_rewards/accuracies": 0.6015625,
+      "eval_rewards/chosen": -15.375,
+      "eval_rewards/margins": 1.640625,
+      "eval_rewards/rejected": -17.0,
+      "eval_runtime": 46.6662,
+      "eval_samples_per_second": 42.858,
+      "eval_steps_per_second": 0.686,
+      "step": 2500
+    },
+    {
+      "epoch": 2.6282722513089007,
+      "grad_norm": 0.07655568789515692,
+      "learning_rate": 2.3031214857028446e-06,
+      "logits/chosen": -1.8203125,
+      "logits/rejected": 0.0888671875,
+      "logps/chosen": -1584.0,
+      "logps/rejected": -2592.0,
+      "loss": 0.0153,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -11.625,
+      "rewards/margins": 9.375,
+      "rewards/rejected": -21.0,
+      "step": 2510
+    },
+    {
+      "epoch": 2.6387434554973823,
+      "grad_norm": 0.1288903261198696,
+      "learning_rate": 2.177086530257139e-06,
+      "logits/chosen": -2.421875,
+      "logits/rejected": -0.0693359375,
+      "logps/chosen": -1552.0,
+      "logps/rejected": -2512.0,
+      "loss": 0.0177,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -11.625,
+      "rewards/margins": 8.9375,
+      "rewards/rejected": -20.625,
+      "step": 2520
+    },
+    {
+      "epoch": 2.649214659685864,
+      "grad_norm": 0.2777580455626625,
+      "learning_rate": 2.054440793781354e-06,
+      "logits/chosen": -2.796875,
+      "logits/rejected": -0.412109375,
+      "logps/chosen": -1528.0,
+      "logps/rejected": -2592.0,
+      "loss": 0.0198,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -11.25,
+      "rewards/margins": 9.5625,
+      "rewards/rejected": -20.75,
+      "step": 2530
+    },
+    {
+      "epoch": 2.6596858638743455,
+      "grad_norm": 0.06996163806870984,
+      "learning_rate": 1.9352024892550587e-06,
+      "logits/chosen": -1.8671875,
+      "logits/rejected": -0.34765625,
+      "logps/chosen": -1576.0,
+      "logps/rejected": -2608.0,
+      "loss": 0.0141,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -11.6875,
+      "rewards/margins": 9.3125,
+      "rewards/rejected": -21.0,
+      "step": 2540
+    },
+    {
+      "epoch": 2.670157068062827,
+      "grad_norm": 0.3134344838095471,
+      "learning_rate": 1.8193893236517379e-06,
+      "logits/chosen": -2.6875,
+      "logits/rejected": -0.330078125,
+      "logps/chosen": -1600.0,
+      "logps/rejected": -2560.0,
+      "loss": 0.0157,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -11.75,
+      "rewards/margins": 9.375,
+      "rewards/rejected": -21.125,
+      "step": 2550
+    },
+    {
+      "epoch": 2.680628272251309,
+      "grad_norm": 0.2679659535142388,
+      "learning_rate": 1.7070184953092956e-06,
+      "logits/chosen": -2.765625,
+      "logits/rejected": -0.80859375,
+      "logps/chosen": -1512.0,
+      "logps/rejected": -2496.0,
+      "loss": 0.0156,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -11.125,
+      "rewards/margins": 8.6875,
+      "rewards/rejected": -19.75,
+      "step": 2560
+    },
+    {
+      "epoch": 2.6910994764397906,
+      "grad_norm": 0.5343281926367408,
+      "learning_rate": 1.5981066913760918e-06,
+      "logits/chosen": -1.6875,
+      "logits/rejected": 0.08642578125,
+      "logps/chosen": -1448.0,
+      "logps/rejected": -2464.0,
+      "loss": 0.0104,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -10.75,
+      "rewards/margins": 8.75,
+      "rewards/rejected": -19.5,
+      "step": 2570
+    },
+    {
+      "epoch": 2.701570680628272,
+      "grad_norm": 0.7772208052418603,
+      "learning_rate": 1.492670085332884e-06,
+      "logits/chosen": -2.46875,
+      "logits/rejected": -0.70703125,
+      "logps/chosen": -1600.0,
+      "logps/rejected": -2640.0,
+      "loss": 0.0195,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -11.9375,
+      "rewards/margins": 9.0,
+      "rewards/rejected": -21.0,
+      "step": 2580
+    },
+    {
+      "epoch": 2.712041884816754,
+      "grad_norm": 0.3909040077870737,
+      "learning_rate": 1.3907243345910787e-06,
+      "logits/chosen": -2.0,
+      "logits/rejected": 0.33984375,
+      "logps/chosen": -1616.0,
+      "logps/rejected": -2624.0,
+      "loss": 0.0157,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -11.6875,
+      "rewards/margins": 9.125,
+      "rewards/rejected": -20.875,
+      "step": 2590
+    },
+    {
+      "epoch": 2.7225130890052354,
+      "grad_norm": 0.1069725799306176,
+      "learning_rate": 1.292284578167538e-06,
+      "logits/chosen": -2.609375,
+      "logits/rejected": -0.91015625,
+      "logps/chosen": -1512.0,
+      "logps/rejected": -2576.0,
+      "loss": 0.0134,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -11.25,
+      "rewards/margins": 9.125,
+      "rewards/rejected": -20.375,
+      "step": 2600
+    },
+    {
+      "epoch": 2.7225130890052354,
+      "eval_logits/chosen": 0.7109375,
+      "eval_logits/rejected": 2.546875,
+      "eval_logps/chosen": -1816.0,
+      "eval_logps/rejected": -1952.0,
+      "eval_loss": 1.68540620803833,
+      "eval_rewards/accuracies": 0.58984375,
+      "eval_rewards/chosen": -15.0,
+      "eval_rewards/margins": 1.6171875,
+      "eval_rewards/rejected": -16.625,
+      "eval_runtime": 46.6645,
+      "eval_samples_per_second": 42.859,
+      "eval_steps_per_second": 0.686,
+      "step": 2600
+    },
+    {
+      "epoch": 2.7329842931937174,
+      "grad_norm": 0.308940796134209,
+      "learning_rate": 1.1973654344364925e-06,
+      "logits/chosen": -2.4375,
+      "logits/rejected": -1.09375,
+      "logps/chosen": -1584.0,
+      "logps/rejected": -2592.0,
+      "loss": 0.0134,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -11.4375,
+      "rewards/margins": 8.875,
+      "rewards/rejected": -20.375,
+      "step": 2610
+    },
+    {
+      "epoch": 2.743455497382199,
+      "grad_norm": 0.19657020267705813,
+      "learning_rate": 1.1059809989586583e-06,
+      "logits/chosen": -1.75,
+      "logits/rejected": -0.453125,
+      "logps/chosen": -1536.0,
+      "logps/rejected": -2576.0,
+      "loss": 0.0227,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -11.3125,
+      "rewards/margins": 9.25,
+      "rewards/rejected": -20.5,
+      "step": 2620
+    },
+    {
+      "epoch": 2.7539267015706805,
+      "grad_norm": 3.03559659464108,
+      "learning_rate": 1.018144842388044e-06,
+      "logits/chosen": -1.9609375,
+      "logits/rejected": 0.484375,
+      "logps/chosen": -1584.0,
+      "logps/rejected": -2640.0,
+      "loss": 0.0191,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -11.4375,
+      "rewards/margins": 8.875,
+      "rewards/rejected": -20.25,
+      "step": 2630
+    },
+    {
+      "epoch": 2.7643979057591626,
+      "grad_norm": 0.1622247348983575,
+      "learning_rate": 9.338700084567109e-07,
+      "logits/chosen": -2.359375,
+      "logits/rejected": -0.44140625,
+      "logps/chosen": -1552.0,
+      "logps/rejected": -2640.0,
+      "loss": 0.0129,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -11.5,
+      "rewards/margins": 9.375,
+      "rewards/rejected": -20.875,
+      "step": 2640
+    },
+    {
+      "epoch": 2.774869109947644,
+      "grad_norm": 0.6038268939541903,
+      "learning_rate": 8.531690120377606e-07,
+      "logits/chosen": -2.390625,
+      "logits/rejected": -0.76171875,
+      "logps/chosen": -1704.0,
+      "logps/rejected": -2768.0,
+      "loss": 0.0141,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -12.5,
+      "rewards/margins": 9.25,
+      "rewards/rejected": -21.75,
+      "step": 2650
+    },
+    {
+      "epoch": 2.7853403141361257,
+      "grad_norm": 0.38710210344185003,
+      "learning_rate": 7.760538372868636e-07,
+      "logits/chosen": -2.78125,
+      "logits/rejected": -0.5078125,
+      "logps/chosen": -1552.0,
+      "logps/rejected": -2528.0,
+      "loss": 0.0164,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -11.5,
+      "rewards/margins": 8.9375,
+      "rewards/rejected": -20.5,
+      "step": 2660
+    },
+    {
+      "epoch": 2.7958115183246073,
+      "grad_norm": 0.182754981587112,
+      "learning_rate": 7.025359358626166e-07,
+      "logits/chosen": -2.375,
+      "logits/rejected": 0.12109375,
+      "logps/chosen": -1640.0,
+      "logps/rejected": -2624.0,
+      "loss": 0.0114,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -12.0,
+      "rewards/margins": 9.1875,
+      "rewards/rejected": -21.25,
+      "step": 2670
+    },
+    {
+      "epoch": 2.806282722513089,
+      "grad_norm": 0.4829630471064681,
+      "learning_rate": 6.326262252259463e-07,
+      "logits/chosen": -2.015625,
+      "logits/rejected": 0.50390625,
+      "logps/chosen": -1576.0,
+      "logps/rejected": -2544.0,
+      "loss": 0.0173,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -11.5,
+      "rewards/margins": 9.125,
+      "rewards/rejected": -20.625,
+      "step": 2680
+    },
+    {
+      "epoch": 2.816753926701571,
+      "grad_norm": 0.19443469229353988,
+      "learning_rate": 5.66335087018871e-07,
+      "logits/chosen": -2.921875,
+      "logits/rejected": -0.384765625,
+      "logps/chosen": -1568.0,
+      "logps/rejected": -2496.0,
+      "loss": 0.0072,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -11.3125,
+      "rewards/margins": 8.9375,
+      "rewards/rejected": -20.25,
+      "step": 2690
+    },
+    {
+      "epoch": 2.8272251308900525,
+      "grad_norm": 0.1635176444004718,
+      "learning_rate": 5.036723655228226e-07,
+      "logits/chosen": -2.453125,
+      "logits/rejected": -0.515625,
+      "logps/chosen": -1496.0,
+      "logps/rejected": -2448.0,
+      "loss": 0.0161,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -11.25,
+      "rewards/margins": 8.8125,
+      "rewards/rejected": -20.0,
+      "step": 2700
+    },
+    {
+      "epoch": 2.8272251308900525,
+      "eval_logits/chosen": 0.68359375,
+      "eval_logits/rejected": 2.53125,
+      "eval_logps/chosen": -1824.0,
+      "eval_logps/rejected": -1960.0,
+      "eval_loss": 1.6882343292236328,
+      "eval_rewards/accuracies": 0.587890625,
+      "eval_rewards/chosen": -15.0625,
+      "eval_rewards/margins": 1.625,
+      "eval_rewards/rejected": -16.625,
+      "eval_runtime": 46.6551,
+      "eval_samples_per_second": 42.868,
+      "eval_steps_per_second": 0.686,
+      "step": 2700
+    },
+    {
+      "epoch": 2.837696335078534,
+      "grad_norm": 0.2163692953663021,
+      "learning_rate": 4.446473661967432e-07,
+      "logits/chosen": -3.078125,
+      "logits/rejected": -0.8125,
+      "logps/chosen": -1544.0,
+      "logps/rejected": -2464.0,
+      "loss": 0.0145,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -11.375,
+      "rewards/margins": 8.8125,
+      "rewards/rejected": -20.125,
+      "step": 2710
+    },
+    {
+      "epoch": 2.8481675392670156,
+      "grad_norm": 0.27303054886309563,
+      "learning_rate": 3.892688542952411e-07,
+      "logits/chosen": -1.9140625,
+      "logits/rejected": 0.2060546875,
+      "logps/chosen": -1504.0,
+      "logps/rejected": -2480.0,
+      "loss": 0.0186,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -11.125,
+      "rewards/margins": 8.4375,
+      "rewards/rejected": -19.625,
+      "step": 2720
+    },
+    {
+      "epoch": 2.858638743455497,
+      "grad_norm": 1.9303658172652438,
+      "learning_rate": 3.3754505356694e-07,
+      "logits/chosen": -3.0625,
+      "logits/rejected": -1.6171875,
+      "logps/chosen": -1600.0,
+      "logps/rejected": -2608.0,
+      "loss": 0.017,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -11.8125,
+      "rewards/margins": 8.9375,
+      "rewards/rejected": -20.75,
+      "step": 2730
+    },
+    {
+      "epoch": 2.869109947643979,
+      "grad_norm": 0.3577776032079035,
+      "learning_rate": 2.8948364503322276e-07,
+      "logits/chosen": -2.234375,
+      "logits/rejected": -0.443359375,
+      "logps/chosen": -1504.0,
+      "logps/rejected": -2480.0,
+      "loss": 0.017,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -11.1875,
+      "rewards/margins": 8.875,
+      "rewards/rejected": -20.0,
+      "step": 2740
+    },
+    {
+      "epoch": 2.8795811518324608,
+      "grad_norm": 0.24443735291517754,
+      "learning_rate": 2.4509176584762706e-07,
+      "logits/chosen": -2.421875,
+      "logits/rejected": -0.18359375,
+      "logps/chosen": -1528.0,
+      "logps/rejected": -2448.0,
+      "loss": 0.0127,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -11.3125,
+      "rewards/margins": 8.4375,
+      "rewards/rejected": -19.75,
+      "step": 2750
+    },
+    {
+      "epoch": 2.8900523560209423,
+      "grad_norm": 0.4972489167452459,
+      "learning_rate": 2.0437600823595694e-07,
+      "logits/chosen": -2.703125,
+      "logits/rejected": -0.69140625,
+      "logps/chosen": -1624.0,
+      "logps/rejected": -2576.0,
+      "loss": 0.0148,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -11.75,
+      "rewards/margins": 8.5,
+      "rewards/rejected": -20.25,
+      "step": 2760
+    },
+    {
+      "epoch": 2.900523560209424,
+      "grad_norm": 0.601126967554501,
+      "learning_rate": 1.6734241851733e-07,
+      "logits/chosen": -2.9375,
+      "logits/rejected": -0.39453125,
+      "logps/chosen": -1648.0,
+      "logps/rejected": -2656.0,
+      "loss": 0.0144,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -11.875,
+      "rewards/margins": 9.625,
+      "rewards/rejected": -21.5,
+      "step": 2770
+    },
+    {
+      "epoch": 2.9109947643979055,
+      "grad_norm": 0.07989631772018826,
+      "learning_rate": 1.339964962062984e-07,
+      "logits/chosen": -2.671875,
+      "logits/rejected": -1.2890625,
+      "logps/chosen": -1528.0,
+      "logps/rejected": -2576.0,
+      "loss": 0.0093,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -11.5625,
+      "rewards/margins": 9.125,
+      "rewards/rejected": -20.75,
+      "step": 2780
+    },
+    {
+      "epoch": 2.9214659685863875,
+      "grad_norm": 0.2927355531444079,
+      "learning_rate": 1.0434319319617136e-07,
+      "logits/chosen": -1.9609375,
+      "logits/rejected": 0.06591796875,
+      "logps/chosen": -1528.0,
+      "logps/rejected": -2496.0,
+      "loss": 0.0161,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -11.375,
+      "rewards/margins": 8.9375,
+      "rewards/rejected": -20.25,
+      "step": 2790
+    },
+    {
+      "epoch": 2.931937172774869,
+      "grad_norm": 0.4611553014168555,
+      "learning_rate": 7.838691302365363e-08,
+      "logits/chosen": -2.390625,
+      "logits/rejected": -0.494140625,
+      "logps/chosen": -1520.0,
+      "logps/rejected": -2592.0,
+      "loss": 0.0165,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -11.125,
+      "rewards/margins": 9.25,
+      "rewards/rejected": -20.375,
+      "step": 2800
+    },
+    {
+      "epoch": 2.931937172774869,
+      "eval_logits/chosen": 0.64453125,
+      "eval_logits/rejected": 2.484375,
+      "eval_logps/chosen": -1824.0,
+      "eval_logps/rejected": -1952.0,
+      "eval_loss": 1.6896874904632568,
+      "eval_rewards/accuracies": 0.591796875,
+      "eval_rewards/chosen": -15.0625,
+      "eval_rewards/margins": 1.6171875,
+      "eval_rewards/rejected": -16.625,
+      "eval_runtime": 46.6465,
+      "eval_samples_per_second": 42.876,
+      "eval_steps_per_second": 0.686,
+      "step": 2800
+    },
+    {
+      "epoch": 2.9424083769633507,
+      "grad_norm": 0.8843762593296421,
+      "learning_rate": 5.613151021490459e-08,
+      "logits/chosen": -1.8984375,
+      "logits/rejected": -0.65234375,
+      "logps/chosen": -1456.0,
+      "logps/rejected": -2416.0,
+      "loss": 0.0149,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -11.125,
+      "rewards/margins": 8.375,
+      "rewards/rejected": -19.5,
+      "step": 2810
+    },
+    {
+      "epoch": 2.9528795811518327,
+      "grad_norm": 0.26602996563802084,
+      "learning_rate": 3.7580289713159965e-08,
+      "logits/chosen": -2.140625,
+      "logits/rejected": -0.4375,
+      "logps/chosen": -1616.0,
+      "logps/rejected": -2624.0,
+      "loss": 0.0157,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -11.875,
+      "rewards/margins": 8.9375,
+      "rewards/rejected": -20.875,
+      "step": 2820
+    },
+    {
+      "epoch": 2.9633507853403143,
+      "grad_norm": 0.16297411803369016,
+      "learning_rate": 2.2736006387927277e-08,
+      "logits/chosen": -2.765625,
+      "logits/rejected": -0.77734375,
+      "logps/chosen": -1576.0,
+      "logps/rejected": -2576.0,
+      "loss": 0.0115,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -11.5625,
+      "rewards/margins": 9.0625,
+      "rewards/rejected": -20.625,
+      "step": 2830
+    },
+    {
+      "epoch": 2.973821989528796,
+      "grad_norm": 0.23526111657034204,
+      "learning_rate": 1.1600864625893604e-08,
+      "logits/chosen": -2.609375,
+      "logits/rejected": -0.4921875,
+      "logps/chosen": -1544.0,
+      "logps/rejected": -2528.0,
+      "loss": 0.0117,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -11.25,
+      "rewards/margins": 9.0,
+      "rewards/rejected": -20.25,
+      "step": 2840
+    },
+    {
+      "epoch": 2.9842931937172774,
+      "grad_norm": 0.22464297127570537,
+      "learning_rate": 4.176518003573549e-09,
+      "logits/chosen": -2.453125,
+      "logits/rejected": -0.609375,
+      "logps/chosen": -1576.0,
+      "logps/rejected": -2560.0,
+      "loss": 0.019,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -11.75,
+      "rewards/margins": 8.625,
+      "rewards/rejected": -20.375,
+      "step": 2850
+    },
+    {
+      "epoch": 2.994764397905759,
+      "grad_norm": 0.5260568319393822,
+      "learning_rate": 4.640690417528903e-10,
+      "logits/chosen": -2.90625,
+      "logits/rejected": -0.890625,
+      "logps/chosen": -1528.0,
+      "logps/rejected": -2464.0,
+      "loss": 0.0162,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -11.5,
+      "rewards/margins": 8.75,
+      "rewards/rejected": -20.25,
+      "step": 2860
+    },
+    {
+      "epoch": 3.0,
+      "step": 2865,
+      "total_flos": 0.0,
+      "train_loss": 0.22519320912178067,
+      "train_runtime": 12281.0925,
+      "train_samples_per_second": 14.93,
+      "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
+}