diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,5094 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.998430141287284,
+  "eval_steps": 100,
+  "global_step": 2865,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.010465724751439037,
+      "grad_norm": 0.017736490905615007,
+      "learning_rate": 1.7421602787456445e-06,
+      "logits/chosen": -10.875,
+      "logits/rejected": -10.9375,
+      "logps/chosen": -1.140625,
+      "logps/rejected": -1.171875,
+      "loss": 0.9734,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.01141357421875,
+      "rewards/margins": 0.000274658203125,
+      "rewards/rejected": -0.01171875,
+      "step": 10
+    },
+    {
+      "epoch": 0.020931449502878074,
+      "grad_norm": 0.01675627818571838,
+      "learning_rate": 3.484320557491289e-06,
+      "logits/chosen": -10.625,
+      "logits/rejected": -10.5,
+      "logps/chosen": -1.171875,
+      "logps/rejected": -1.28125,
+      "loss": 0.9734,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": -0.01171875,
+      "rewards/margins": 0.00112152099609375,
+      "rewards/rejected": -0.0128173828125,
+      "step": 20
+    },
+    {
+      "epoch": 0.03139717425431711,
+      "grad_norm": 0.011154644631322676,
+      "learning_rate": 5.226480836236934e-06,
+      "logits/chosen": -11.1875,
+      "logits/rejected": -10.875,
+      "logps/chosen": -1.0859375,
+      "logps/rejected": -1.1796875,
+      "loss": 0.9735,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.53125,
+      "rewards/chosen": -0.0108642578125,
+      "rewards/margins": 0.000934600830078125,
+      "rewards/rejected": -0.01177978515625,
+      "step": 30
+    },
+    {
+      "epoch": 0.04186289900575615,
+      "grad_norm": 0.02013327570036217,
+      "learning_rate": 6.968641114982578e-06,
+      "logits/chosen": -11.5625,
+      "logits/rejected": -11.5,
+      "logps/chosen": -1.15625,
+      "logps/rejected": -1.3046875,
+      "loss": 0.9728,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.5687500238418579,
+      "rewards/chosen": -0.0115966796875,
+      "rewards/margins": 0.00147247314453125,
+      "rewards/rejected": -0.0130615234375,
+      "step": 40
+    },
+    {
+      "epoch": 0.052328623757195186,
+      "grad_norm": 0.03184991102612421,
+      "learning_rate": 8.710801393728225e-06,
+      "logits/chosen": -12.25,
+      "logits/rejected": -11.9375,
+      "logps/chosen": -1.3515625,
+      "logps/rejected": -1.5859375,
+      "loss": 0.9726,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": -0.0135498046875,
+      "rewards/margins": 0.0023345947265625,
+      "rewards/rejected": -0.015869140625,
+      "step": 50
+    },
+    {
+      "epoch": 0.06279434850863422,
+      "grad_norm": 0.027747304802348785,
+      "learning_rate": 1.0452961672473868e-05,
+      "logits/chosen": -11.125,
+      "logits/rejected": -10.8125,
+      "logps/chosen": -1.859375,
+      "logps/rejected": -1.9453125,
+      "loss": 0.9722,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.46875,
+      "rewards/chosen": -0.0185546875,
+      "rewards/margins": 0.000827789306640625,
+      "rewards/rejected": -0.0194091796875,
+      "step": 60
+    },
+    {
+      "epoch": 0.07326007326007326,
+      "grad_norm": 0.0764170021825398,
+      "learning_rate": 1.2195121951219513e-05,
+      "logits/chosen": -10.125,
+      "logits/rejected": -9.5625,
+      "logps/chosen": -2.46875,
+      "logps/rejected": -2.75,
+      "loss": 0.9723,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.581250011920929,
+      "rewards/chosen": -0.024658203125,
+      "rewards/margins": 0.00274658203125,
+      "rewards/rejected": -0.0274658203125,
+      "step": 70
+    },
+    {
+      "epoch": 0.0837257980115123,
+      "grad_norm": 0.2535752209553448,
+      "learning_rate": 1.3937282229965156e-05,
+      "logits/chosen": -9.875,
+      "logits/rejected": -9.625,
+      "logps/chosen": -13.0,
+      "logps/rejected": -15.0625,
+      "loss": 0.9616,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": -0.1298828125,
+      "rewards/margins": 0.020751953125,
+      "rewards/rejected": -0.150390625,
+      "step": 80
+    },
+    {
+      "epoch": 0.09419152276295134,
+      "grad_norm": 0.4465796445662566,
+      "learning_rate": 1.56794425087108e-05,
+      "logits/chosen": -11.5,
+      "logits/rejected": -11.4375,
+      "logps/chosen": -34.25,
+      "logps/rejected": -36.75,
+      "loss": 0.9507,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.59375,
+      "rewards/chosen": -0.341796875,
+      "rewards/margins": 0.0262451171875,
+      "rewards/rejected": -0.3671875,
+      "step": 90
+    },
+    {
+      "epoch": 0.10465724751439037,
+      "grad_norm": 0.5455033637543892,
+      "learning_rate": 1.742160278745645e-05,
+      "logits/chosen": -9.6875,
+      "logits/rejected": -9.25,
+      "logps/chosen": -36.75,
+      "logps/rejected": -44.25,
+      "loss": 0.9346,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": -0.369140625,
+      "rewards/margins": 0.0732421875,
+      "rewards/rejected": -0.44140625,
+      "step": 100
+    },
+    {
+      "epoch": 0.10465724751439037,
+      "eval_logits/chosen": -8.5625,
+      "eval_logits/rejected": -7.96875,
+      "eval_logps/chosen": -33.25,
+      "eval_logps/rejected": -41.75,
+      "eval_loss": 0.9349374771118164,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.61328125,
+      "eval_rewards/chosen": -0.33203125,
+      "eval_rewards/margins": 0.08642578125,
+      "eval_rewards/rejected": -0.41796875,
+      "eval_runtime": 23.7809,
+      "eval_samples_per_second": 84.101,
+      "eval_steps_per_second": 1.346,
+      "step": 100
+    },
+    {
+      "epoch": 0.1151229722658294,
+      "grad_norm": 0.4381611747514917,
+      "learning_rate": 1.9163763066202093e-05,
+      "logits/chosen": -7.53125,
+      "logits/rejected": -7.125,
+      "logps/chosen": -37.5,
+      "logps/rejected": -43.0,
+      "loss": 0.9443,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": -0.375,
+      "rewards/margins": 0.05419921875,
+      "rewards/rejected": -0.4296875,
+      "step": 110
+    },
+    {
+      "epoch": 0.12558869701726844,
+      "grad_norm": 0.6355196056738687,
+      "learning_rate": 2.0905923344947736e-05,
+      "logits/chosen": -5.34375,
+      "logits/rejected": -4.625,
+      "logps/chosen": -55.75,
+      "logps/rejected": -64.5,
+      "loss": 0.9164,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -0.55859375,
+      "rewards/margins": 0.08984375,
+      "rewards/rejected": -0.6484375,
+      "step": 120
+    },
+    {
+      "epoch": 0.1360544217687075,
+      "grad_norm": 0.5775406087346924,
+      "learning_rate": 2.264808362369338e-05,
+      "logits/chosen": -5.96875,
+      "logits/rejected": -5.21875,
+      "logps/chosen": -53.25,
+      "logps/rejected": -64.0,
+      "loss": 0.9222,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": -0.53125,
+      "rewards/margins": 0.1103515625,
+      "rewards/rejected": -0.640625,
+      "step": 130
+    },
+    {
+      "epoch": 0.14652014652014653,
+      "grad_norm": 0.6204772440713339,
+      "learning_rate": 2.4390243902439026e-05,
+      "logits/chosen": -6.375,
+      "logits/rejected": -5.875,
+      "logps/chosen": -33.0,
+      "logps/rejected": -40.25,
+      "loss": 0.9333,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.330078125,
+      "rewards/margins": 0.07275390625,
+      "rewards/rejected": -0.40234375,
+      "step": 140
+    },
+    {
+      "epoch": 0.15698587127158556,
+      "grad_norm": 0.6751460503008085,
+      "learning_rate": 2.6132404181184672e-05,
+      "logits/chosen": -6.15625,
+      "logits/rejected": -5.03125,
+      "logps/chosen": -41.5,
+      "logps/rejected": -54.5,
+      "loss": 0.9147,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -0.4140625,
+      "rewards/margins": 0.130859375,
+      "rewards/rejected": -0.546875,
+      "step": 150
+    },
+    {
+      "epoch": 0.1674515960230246,
+      "grad_norm": 0.8029371797475424,
+      "learning_rate": 2.7874564459930312e-05,
+      "logits/chosen": -5.25,
+      "logits/rejected": -4.1875,
+      "logps/chosen": -55.0,
+      "logps/rejected": -71.5,
+      "loss": 0.902,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.55078125,
+      "rewards/margins": 0.1630859375,
+      "rewards/rejected": -0.71484375,
+      "step": 160
+    },
+    {
+      "epoch": 0.17791732077446362,
+      "grad_norm": 0.6782824802775677,
+      "learning_rate": 2.9616724738675962e-05,
+      "logits/chosen": -5.4375,
+      "logits/rejected": -4.5625,
+      "logps/chosen": -36.75,
+      "logps/rejected": -49.75,
+      "loss": 0.9134,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -0.3671875,
+      "rewards/margins": 0.1318359375,
+      "rewards/rejected": -0.498046875,
+      "step": 170
+    },
+    {
+      "epoch": 0.18838304552590268,
+      "grad_norm": 0.7519549715114882,
+      "learning_rate": 3.13588850174216e-05,
+      "logits/chosen": -3.6875,
+      "logits/rejected": -2.703125,
+      "logps/chosen": -51.5,
+      "logps/rejected": -68.0,
+      "loss": 0.8976,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.515625,
+      "rewards/margins": 0.166015625,
+      "rewards/rejected": -0.6796875,
+      "step": 180
+    },
+    {
+      "epoch": 0.1988487702773417,
+      "grad_norm": 0.7303572379423364,
+      "learning_rate": 3.310104529616725e-05,
+      "logits/chosen": -3.546875,
+      "logits/rejected": -2.828125,
+      "logps/chosen": -43.0,
+      "logps/rejected": -56.5,
+      "loss": 0.8937,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -0.4296875,
+      "rewards/margins": 0.134765625,
+      "rewards/rejected": -0.5625,
+      "step": 190
+    },
+    {
+      "epoch": 0.20931449502878074,
+      "grad_norm": 0.7959522206756111,
+      "learning_rate": 3.48432055749129e-05,
+      "logits/chosen": -5.375,
+      "logits/rejected": -4.0,
+      "logps/chosen": -24.5,
+      "logps/rejected": -41.5,
+      "loss": 0.9139,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.5562499761581421,
+      "rewards/chosen": -0.2451171875,
+      "rewards/margins": 0.1708984375,
+      "rewards/rejected": -0.416015625,
+      "step": 200
+    },
+    {
+      "epoch": 0.20931449502878074,
+      "eval_logits/chosen": -3.453125,
+      "eval_logits/rejected": -2.4375,
+      "eval_logps/chosen": -48.5,
+      "eval_logps/rejected": -63.5,
+      "eval_loss": 0.9068750143051147,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.626953125,
+      "eval_rewards/chosen": -0.484375,
+      "eval_rewards/margins": 0.150390625,
+      "eval_rewards/rejected": -0.63671875,
+      "eval_runtime": 23.4068,
+      "eval_samples_per_second": 85.445,
+      "eval_steps_per_second": 1.367,
+      "step": 200
+    },
+    {
+      "epoch": 0.21978021978021978,
+      "grad_norm": 0.9105927645064195,
+      "learning_rate": 3.6585365853658535e-05,
+      "logits/chosen": -4.03125,
+      "logits/rejected": -3.3125,
+      "logps/chosen": -46.75,
+      "logps/rejected": -55.5,
+      "loss": 0.9186,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.581250011920929,
+      "rewards/chosen": -0.46875,
+      "rewards/margins": 0.087890625,
+      "rewards/rejected": -0.5546875,
+      "step": 210
+    },
+    {
+      "epoch": 0.2302459445316588,
+      "grad_norm": 0.7438576174370349,
+      "learning_rate": 3.8327526132404185e-05,
+      "logits/chosen": -4.84375,
+      "logits/rejected": -3.71875,
+      "logps/chosen": -37.25,
+      "logps/rejected": -51.25,
+      "loss": 0.92,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -0.37109375,
+      "rewards/margins": 0.1416015625,
+      "rewards/rejected": -0.51171875,
+      "step": 220
+    },
+    {
+      "epoch": 0.24071166928309787,
+      "grad_norm": 0.572610296145622,
+      "learning_rate": 4.006968641114983e-05,
+      "logits/chosen": -3.953125,
+      "logits/rejected": -3.0625,
+      "logps/chosen": -43.25,
+      "logps/rejected": -53.5,
+      "loss": 0.9112,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": -0.43359375,
+      "rewards/margins": 0.10107421875,
+      "rewards/rejected": -0.53515625,
+      "step": 230
+    },
+    {
+      "epoch": 0.25117739403453687,
+      "grad_norm": 0.5058455673465353,
+      "learning_rate": 4.181184668989547e-05,
+      "logits/chosen": -4.3125,
+      "logits/rejected": -3.515625,
+      "logps/chosen": -37.25,
+      "logps/rejected": -55.0,
+      "loss": 0.8864,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.373046875,
+      "rewards/margins": 0.1767578125,
+      "rewards/rejected": -0.55078125,
+      "step": 240
+    },
+    {
+      "epoch": 0.2616431187859759,
+      "grad_norm": 0.9011201643655188,
+      "learning_rate": 4.3554006968641115e-05,
+      "logits/chosen": -4.1875,
+      "logits/rejected": -3.171875,
+      "logps/chosen": -40.75,
+      "logps/rejected": -57.5,
+      "loss": 0.8894,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -0.40625,
+      "rewards/margins": 0.169921875,
+      "rewards/rejected": -0.57421875,
+      "step": 250
+    },
+    {
+      "epoch": 0.272108843537415,
+      "grad_norm": 0.7777596853340315,
+      "learning_rate": 4.529616724738676e-05,
+      "logits/chosen": -2.859375,
+      "logits/rejected": -1.71875,
+      "logps/chosen": -50.5,
+      "logps/rejected": -62.75,
+      "loss": 0.9062,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": -0.50390625,
+      "rewards/margins": 0.12060546875,
+      "rewards/rejected": -0.625,
+      "step": 260
+    },
+    {
+      "epoch": 0.282574568288854,
+      "grad_norm": 0.5497208090426289,
+      "learning_rate": 4.703832752613241e-05,
+      "logits/chosen": -4.53125,
+      "logits/rejected": -3.5625,
+      "logps/chosen": -44.0,
+      "logps/rejected": -58.5,
+      "loss": 0.8956,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -0.439453125,
+      "rewards/margins": 0.1474609375,
+      "rewards/rejected": -0.5859375,
+      "step": 270
+    },
+    {
+      "epoch": 0.29304029304029305,
+      "grad_norm": 1.1536251719155128,
+      "learning_rate": 4.878048780487805e-05,
+      "logits/chosen": -1.7890625,
+      "logits/rejected": -0.5390625,
+      "logps/chosen": -45.0,
+      "logps/rejected": -63.0,
+      "loss": 0.9015,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -0.451171875,
+      "rewards/margins": 0.1787109375,
+      "rewards/rejected": -0.62890625,
+      "step": 280
+    },
+    {
+      "epoch": 0.3035060177917321,
+      "grad_norm": 0.9258378600859519,
+      "learning_rate": 4.999983293481418e-05,
+      "logits/chosen": -1.40625,
+      "logits/rejected": -0.43359375,
+      "logps/chosen": -70.5,
+      "logps/rejected": -89.0,
+      "loss": 0.9025,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.70703125,
+      "rewards/margins": 0.1845703125,
+      "rewards/rejected": -0.890625,
+      "step": 290
+    },
+    {
+      "epoch": 0.3139717425431711,
+      "grad_norm": 0.8920721772329635,
+      "learning_rate": 4.9996862949181595e-05,
+      "logits/chosen": -5.625,
+      "logits/rejected": -4.84375,
+      "logps/chosen": -61.5,
+      "logps/rejected": -76.0,
+      "loss": 0.907,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.61328125,
+      "rewards/margins": 0.146484375,
+      "rewards/rejected": -0.76171875,
+      "step": 300
+    },
+    {
+      "epoch": 0.3139717425431711,
+      "eval_logits/chosen": -5.1875,
+      "eval_logits/rejected": -4.0,
+      "eval_logps/chosen": -69.5,
+      "eval_logps/rejected": -83.5,
+      "eval_loss": 0.9098749756813049,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.60546875,
+      "eval_rewards/chosen": -0.69140625,
+      "eval_rewards/margins": 0.1416015625,
+      "eval_rewards/rejected": -0.8359375,
+      "eval_runtime": 23.4121,
+      "eval_samples_per_second": 85.426,
+      "eval_steps_per_second": 1.367,
+      "step": 300
+    },
+    {
+      "epoch": 0.32443746729461015,
+      "grad_norm": 1.204718449120235,
+      "learning_rate": 4.9990180911526896e-05,
+      "logits/chosen": -3.953125,
+      "logits/rejected": -2.234375,
+      "logps/chosen": -68.5,
+      "logps/rejected": -86.5,
+      "loss": 0.9027,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.581250011920929,
+      "rewards/chosen": -0.6875,
+      "rewards/margins": 0.17578125,
+      "rewards/rejected": -0.86328125,
+      "step": 310
+    },
+    {
+      "epoch": 0.3349031920460492,
+      "grad_norm": 1.2080755689546043,
+      "learning_rate": 4.997978781413745e-05,
+      "logits/chosen": -1.984375,
+      "logits/rejected": -0.98828125,
+      "logps/chosen": -42.5,
+      "logps/rejected": -56.25,
+      "loss": 0.8971,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.59375,
+      "rewards/chosen": -0.42578125,
+      "rewards/margins": 0.138671875,
+      "rewards/rejected": -0.5625,
+      "step": 320
+    },
+    {
+      "epoch": 0.3453689167974882,
+      "grad_norm": 0.7281152321945176,
+      "learning_rate": 4.996568520039565e-05,
+      "logits/chosen": 0.5859375,
+      "logits/rejected": 1.28125,
+      "logps/chosen": -84.0,
+      "logps/rejected": -98.0,
+      "loss": 0.9064,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -0.83984375,
+      "rewards/margins": 0.1396484375,
+      "rewards/rejected": -0.98046875,
+      "step": 330
+    },
+    {
+      "epoch": 0.35583464154892724,
+      "grad_norm": 0.8874198333485492,
+      "learning_rate": 4.994787516454973e-05,
+      "logits/chosen": -1.8046875,
+      "logits/rejected": -0.375,
+      "logps/chosen": -64.0,
+      "logps/rejected": -80.0,
+      "loss": 0.8975,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": -0.640625,
+      "rewards/margins": 0.1591796875,
+      "rewards/rejected": -0.80078125,
+      "step": 340
+    },
+    {
+      "epoch": 0.3663003663003663,
+      "grad_norm": 0.8568362163246603,
+      "learning_rate": 4.992636035140273e-05,
+      "logits/chosen": -3.75,
+      "logits/rejected": -2.375,
+      "logps/chosen": -57.25,
+      "logps/rejected": -70.0,
+      "loss": 0.8974,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.53125,
+      "rewards/chosen": -0.5703125,
+      "rewards/margins": 0.126953125,
+      "rewards/rejected": -0.69921875,
+      "step": 350
+    },
+    {
+      "epoch": 0.37676609105180536,
+      "grad_norm": 1.501117196658269,
+      "learning_rate": 4.99011439559198e-05,
+      "logits/chosen": -1.8125,
+      "logits/rejected": -0.1171875,
+      "logps/chosen": -74.0,
+      "logps/rejected": -88.5,
+      "loss": 0.8955,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": -0.73828125,
+      "rewards/margins": 0.146484375,
+      "rewards/rejected": -0.88671875,
+      "step": 360
+    },
+    {
+      "epoch": 0.3872318158032444,
+      "grad_norm": 0.9002093318889324,
+      "learning_rate": 4.987222972275369e-05,
+      "logits/chosen": -3.171875,
+      "logits/rejected": -1.4453125,
+      "logps/chosen": -52.0,
+      "logps/rejected": -70.0,
+      "loss": 0.8766,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -0.51953125,
+      "rewards/margins": 0.181640625,
+      "rewards/rejected": -0.703125,
+      "step": 370
+    },
+    {
+      "epoch": 0.3976975405546834,
+      "grad_norm": 1.148898564876584,
+      "learning_rate": 4.983962194568872e-05,
+      "logits/chosen": -2.703125,
+      "logits/rejected": -1.21875,
+      "logps/chosen": -66.0,
+      "logps/rejected": -84.0,
+      "loss": 0.8927,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.66015625,
+      "rewards/margins": 0.18359375,
+      "rewards/rejected": -0.84375,
+      "step": 380
+    },
+    {
+      "epoch": 0.40816326530612246,
+      "grad_norm": 1.5188691471582978,
+      "learning_rate": 4.9803325467003084e-05,
+      "logits/chosen": -3.0,
+      "logits/rejected": -1.6953125,
+      "logps/chosen": -62.25,
+      "logps/rejected": -81.5,
+      "loss": 0.9022,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.62109375,
+      "rewards/margins": 0.1904296875,
+      "rewards/rejected": -0.8125,
+      "step": 390
+    },
+    {
+      "epoch": 0.4186289900575615,
+      "grad_norm": 1.123928046928657,
+      "learning_rate": 4.9763345676749816e-05,
+      "logits/chosen": -1.0390625,
+      "logits/rejected": 0.1728515625,
+      "logps/chosen": -89.0,
+      "logps/rejected": -109.0,
+      "loss": 0.901,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -0.890625,
+      "rewards/margins": 0.19921875,
+      "rewards/rejected": -1.09375,
+      "step": 400
+    },
+    {
+      "epoch": 0.4186289900575615,
+      "eval_logits/chosen": -1.7421875,
+      "eval_logits/rejected": 0.016357421875,
+      "eval_logps/chosen": -84.0,
+      "eval_logps/rejected": -101.0,
+      "eval_loss": 0.8957499861717224,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.6328125,
+      "eval_rewards/chosen": -0.8359375,
+      "eval_rewards/margins": 0.1748046875,
+      "eval_rewards/rejected": -1.015625,
+      "eval_runtime": 23.4041,
+      "eval_samples_per_second": 85.455,
+      "eval_steps_per_second": 1.367,
+      "step": 400
+    },
+    {
+      "epoch": 0.4290947148090005,
+      "grad_norm": 1.0154271541195723,
+      "learning_rate": 4.9719688511956374e-05,
+      "logits/chosen": -5.0,
+      "logits/rejected": -2.984375,
+      "logps/chosen": -68.0,
+      "logps/rejected": -89.5,
+      "loss": 0.901,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -0.6796875,
+      "rewards/margins": 0.21484375,
+      "rewards/rejected": -0.89453125,
+      "step": 410
+    },
+    {
+      "epoch": 0.43956043956043955,
+      "grad_norm": 0.803741341771497,
+      "learning_rate": 4.967236045574294e-05,
+      "logits/chosen": -8.1875,
+      "logits/rejected": -7.0,
+      "logps/chosen": -42.75,
+      "logps/rejected": -61.0,
+      "loss": 0.8946,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -0.427734375,
+      "rewards/margins": 0.18359375,
+      "rewards/rejected": -0.61328125,
+      "step": 420
+    },
+    {
+      "epoch": 0.4500261643118786,
+      "grad_norm": 1.2129855449263687,
+      "learning_rate": 4.96213685363597e-05,
+      "logits/chosen": -3.03125,
+      "logits/rejected": -2.015625,
+      "logps/chosen": -85.0,
+      "logps/rejected": -97.5,
+      "loss": 0.9092,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.581250011920929,
+      "rewards/chosen": -0.8515625,
+      "rewards/margins": 0.126953125,
+      "rewards/rejected": -0.9765625,
+      "step": 430
+    },
+    {
+      "epoch": 0.4604918890633176,
+      "grad_norm": 1.307884862631693,
+      "learning_rate": 4.956672032614315e-05,
+      "logits/chosen": -3.4375,
+      "logits/rejected": -2.21875,
+      "logps/chosen": -71.0,
+      "logps/rejected": -91.0,
+      "loss": 0.863,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -0.7109375,
+      "rewards/margins": 0.2001953125,
+      "rewards/rejected": -0.91015625,
+      "step": 440
+    },
+    {
+      "epoch": 0.47095761381475665,
+      "grad_norm": 1.2508530301615988,
+      "learning_rate": 4.950842394039157e-05,
+      "logits/chosen": -2.515625,
+      "logits/rejected": -1.1171875,
+      "logps/chosen": -66.0,
+      "logps/rejected": -85.5,
+      "loss": 0.9219,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -0.6640625,
+      "rewards/margins": 0.1923828125,
+      "rewards/rejected": -0.85546875,
+      "step": 450
+    },
+    {
+      "epoch": 0.48142333856619574,
+      "grad_norm": 1.058561413706407,
+      "learning_rate": 4.944648803615994e-05,
+      "logits/chosen": -2.25,
+      "logits/rejected": -0.5390625,
+      "logps/chosen": -51.0,
+      "logps/rejected": -74.0,
+      "loss": 0.8859,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.5078125,
+      "rewards/margins": 0.23046875,
+      "rewards/rejected": -0.73828125,
+      "step": 460
+    },
+    {
+      "epoch": 0.49188906331763477,
+      "grad_norm": 0.8260844320330089,
+      "learning_rate": 4.938092181097433e-05,
+      "logits/chosen": -2.484375,
+      "logits/rejected": -1.4765625,
+      "logps/chosen": -55.25,
+      "logps/rejected": -72.5,
+      "loss": 0.8732,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -0.55078125,
+      "rewards/margins": 0.173828125,
+      "rewards/rejected": -0.7265625,
+      "step": 470
+    },
+    {
+      "epoch": 0.5023547880690737,
+      "grad_norm": 0.9203996025987785,
+      "learning_rate": 4.931173500146607e-05,
+      "logits/chosen": -1.71875,
+      "logits/rejected": -0.86328125,
+      "logps/chosen": -57.0,
+      "logps/rejected": -74.5,
+      "loss": 0.888,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -0.5703125,
+      "rewards/margins": 0.1767578125,
+      "rewards/rejected": -0.74609375,
+      "step": 480
+    },
+    {
+      "epoch": 0.5128205128205128,
+      "grad_norm": 1.0083739544972192,
+      "learning_rate": 4.923893788192582e-05,
+      "logits/chosen": -2.5625,
+      "logits/rejected": -0.4375,
+      "logps/chosen": -37.0,
+      "logps/rejected": -68.5,
+      "loss": 0.8981,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -0.37109375,
+      "rewards/margins": 0.3125,
+      "rewards/rejected": -0.68359375,
+      "step": 490
+    },
+    {
+      "epoch": 0.5232862375719518,
+      "grad_norm": 0.9669438133913594,
+      "learning_rate": 4.9162541262777906e-05,
+      "logits/chosen": -0.91015625,
+      "logits/rejected": 0.25390625,
+      "logps/chosen": -56.5,
+      "logps/rejected": -80.5,
+      "loss": 0.8752,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -0.5625,
+      "rewards/margins": 0.23828125,
+      "rewards/rejected": -0.8046875,
+      "step": 500
+    },
+    {
+      "epoch": 0.5232862375719518,
+      "eval_logits/chosen": -0.5625,
+      "eval_logits/rejected": 0.85546875,
+      "eval_logps/chosen": -72.5,
+      "eval_logps/rejected": -95.5,
+      "eval_loss": 0.8768125176429749,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.658203125,
+      "eval_rewards/chosen": -0.7265625,
+      "eval_rewards/margins": 0.232421875,
+      "eval_rewards/rejected": -0.95703125,
+      "eval_runtime": 23.4035,
+      "eval_samples_per_second": 85.457,
+      "eval_steps_per_second": 1.367,
+      "step": 500
+    },
+    {
+      "epoch": 0.533751962323391,
+      "grad_norm": 1.0237334563179212,
+      "learning_rate": 4.90825564889749e-05,
+      "logits/chosen": -1.0859375,
+      "logits/rejected": 0.5390625,
+      "logps/chosen": -67.5,
+      "logps/rejected": -99.5,
+      "loss": 0.8571,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -0.67578125,
+      "rewards/margins": 0.318359375,
+      "rewards/rejected": -0.99609375,
+      "step": 510
+    },
+    {
+      "epoch": 0.54421768707483,
+      "grad_norm": 2.186534332714954,
+      "learning_rate": 4.899899543831288e-05,
+      "logits/chosen": -0.396484375,
+      "logits/rejected": 0.70703125,
+      "logps/chosen": -75.0,
+      "logps/rejected": -98.5,
+      "loss": 0.8677,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": -0.75,
+      "rewards/margins": 0.232421875,
+      "rewards/rejected": -0.984375,
+      "step": 520
+    },
+    {
+      "epoch": 0.554683411826269,
+      "grad_norm": 1.316445531240021,
+      "learning_rate": 4.891187051966765e-05,
+      "logits/chosen": 0.734375,
+      "logits/rejected": 1.7421875,
+      "logps/chosen": -88.0,
+      "logps/rejected": -109.5,
+      "loss": 0.8929,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.8828125,
+      "rewards/margins": 0.2158203125,
+      "rewards/rejected": -1.09375,
+      "step": 530
+    },
+    {
+      "epoch": 0.565149136577708,
+      "grad_norm": 1.4205071171192623,
+      "learning_rate": 4.882119467115191e-05,
+      "logits/chosen": -2.9375,
+      "logits/rejected": -0.94921875,
+      "logps/chosen": -46.0,
+      "logps/rejected": -64.5,
+      "loss": 0.8799,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": -0.4609375,
+      "rewards/margins": 0.185546875,
+      "rewards/rejected": -0.6484375,
+      "step": 540
+    },
+    {
+      "epoch": 0.5756148613291471,
+      "grad_norm": 0.9351402285622962,
+      "learning_rate": 4.872698135819402e-05,
+      "logits/chosen": 0.10888671875,
+      "logits/rejected": 1.453125,
+      "logps/chosen": -81.0,
+      "logps/rejected": -97.5,
+      "loss": 0.8967,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -0.80859375,
+      "rewards/margins": 0.166015625,
+      "rewards/rejected": -0.97265625,
+      "step": 550
+    },
+    {
+      "epoch": 0.5860805860805861,
+      "grad_norm": 0.7880712407790053,
+      "learning_rate": 4.862924457153831e-05,
+      "logits/chosen": 0.439453125,
+      "logits/rejected": 1.6015625,
+      "logps/chosen": -77.0,
+      "logps/rejected": -96.5,
+      "loss": 0.8937,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.76953125,
+      "rewards/margins": 0.197265625,
+      "rewards/rejected": -0.96875,
+      "step": 560
+    },
+    {
+      "epoch": 0.5965463108320251,
+      "grad_norm": 1.1434912483714084,
+      "learning_rate": 4.8527998825167484e-05,
+      "logits/chosen": 0.64453125,
+      "logits/rejected": 1.5859375,
+      "logps/chosen": -89.0,
+      "logps/rejected": -106.5,
+      "loss": 0.8839,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.890625,
+      "rewards/margins": 0.1748046875,
+      "rewards/rejected": -1.0625,
+      "step": 570
+    },
+    {
+      "epoch": 0.6070120355834642,
+      "grad_norm": 1.0571366375620301,
+      "learning_rate": 4.842325915414728e-05,
+      "logits/chosen": 0.66796875,
+      "logits/rejected": 2.03125,
+      "logps/chosen": -84.5,
+      "logps/rejected": -107.0,
+      "loss": 0.8741,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.84375,
+      "rewards/margins": 0.224609375,
+      "rewards/rejected": -1.0703125,
+      "step": 580
+    },
+    {
+      "epoch": 0.6174777603349032,
+      "grad_norm": 0.9507230700020272,
+      "learning_rate": 4.831504111239373e-05,
+      "logits/chosen": 1.375,
+      "logits/rejected": 2.78125,
+      "logps/chosen": -58.5,
+      "logps/rejected": -76.5,
+      "loss": 0.8818,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -0.5859375,
+      "rewards/margins": 0.1806640625,
+      "rewards/rejected": -0.765625,
+      "step": 590
+    },
+    {
+      "epoch": 0.6279434850863422,
+      "grad_norm": 1.0205907492662067,
+      "learning_rate": 4.8203360770363425e-05,
+      "logits/chosen": 2.171875,
+      "logits/rejected": 3.0,
+      "logps/chosen": -77.0,
+      "logps/rejected": -96.5,
+      "loss": 0.8808,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -0.7734375,
+      "rewards/margins": 0.1953125,
+      "rewards/rejected": -0.96484375,
+      "step": 600
+    },
+    {
+      "epoch": 0.6279434850863422,
+      "eval_logits/chosen": 2.15625,
+      "eval_logits/rejected": 3.234375,
+      "eval_logps/chosen": -86.0,
+      "eval_logps/rejected": -109.5,
+      "eval_loss": 0.8741562366485596,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.64453125,
+      "eval_rewards/chosen": -0.86328125,
+      "eval_rewards/margins": 0.2333984375,
+      "eval_rewards/rejected": -1.09375,
+      "eval_runtime": 23.4061,
+      "eval_samples_per_second": 85.448,
+      "eval_steps_per_second": 1.367,
+      "step": 600
+    },
+    {
+      "epoch": 0.6384092098377813,
+      "grad_norm": 1.5053282190514092,
+      "learning_rate": 4.808823471266702e-05,
+      "logits/chosen": 1.265625,
+      "logits/rejected": 2.671875,
+      "logps/chosen": -82.5,
+      "logps/rejected": -109.5,
+      "loss": 0.8588,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.828125,
+      "rewards/margins": 0.265625,
+      "rewards/rejected": -1.09375,
+      "step": 610
+    },
+    {
+      "epoch": 0.6488749345892203,
+      "grad_norm": 1.3189088839898355,
+      "learning_rate": 4.796968003560639e-05,
+      "logits/chosen": 0.703125,
+      "logits/rejected": 1.4375,
+      "logps/chosen": -99.5,
+      "logps/rejected": -123.5,
+      "loss": 0.8692,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -0.99609375,
+      "rewards/margins": 0.2412109375,
+      "rewards/rejected": -1.234375,
+      "step": 620
+    },
+    {
+      "epoch": 0.6593406593406593,
+      "grad_norm": 1.2972908742752567,
+      "learning_rate": 4.7847714344635864e-05,
+      "logits/chosen": -0.30859375,
+      "logits/rejected": 0.734375,
+      "logps/chosen": -102.5,
+      "logps/rejected": -130.0,
+      "loss": 0.854,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -1.03125,
+      "rewards/margins": 0.2734375,
+      "rewards/rejected": -1.296875,
+      "step": 630
+    },
+    {
+      "epoch": 0.6698063840920984,
+      "grad_norm": 1.2966851083079338,
+      "learning_rate": 4.772235575174776e-05,
+      "logits/chosen": 0.9609375,
+      "logits/rejected": 1.875,
+      "logps/chosen": -53.0,
+      "logps/rejected": -74.0,
+      "loss": 0.8961,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -0.53125,
+      "rewards/margins": 0.2109375,
+      "rewards/rejected": -0.7421875,
+      "step": 640
+    },
+    {
+      "epoch": 0.6802721088435374,
+      "grad_norm": 1.0619496123570582,
+      "learning_rate": 4.7593622872782754e-05,
+      "logits/chosen": -0.08056640625,
+      "logits/rejected": 0.22265625,
+      "logps/chosen": -38.25,
+      "logps/rejected": -57.0,
+      "loss": 0.8712,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -0.380859375,
+      "rewards/margins": 0.1884765625,
+      "rewards/rejected": -0.5703125,
+      "step": 650
+    },
+    {
+      "epoch": 0.6907378335949764,
+      "grad_norm": 1.6720705117817145,
+      "learning_rate": 4.746153482466542e-05,
+      "logits/chosen": -1.09375,
+      "logits/rejected": -0.251953125,
+      "logps/chosen": -63.0,
+      "logps/rejected": -88.0,
+      "loss": 0.8607,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -0.62890625,
+      "rewards/margins": 0.25,
+      "rewards/rejected": -0.87890625,
+      "step": 660
+    },
+    {
+      "epoch": 0.7012035583464155,
+      "grad_norm": 2.3983763394133195,
+      "learning_rate": 4.732611122256537e-05,
+      "logits/chosen": -2.140625,
+      "logits/rejected": -1.265625,
+      "logps/chosen": -69.0,
+      "logps/rejected": -92.5,
+      "loss": 0.8482,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.6875,
+      "rewards/margins": 0.236328125,
+      "rewards/rejected": -0.92578125,
+      "step": 670
+    },
+    {
+      "epoch": 0.7116692830978545,
+      "grad_norm": 1.2088831280988124,
+      "learning_rate": 4.718737217698435e-05,
+      "logits/chosen": -0.54296875,
+      "logits/rejected": 0.3515625,
+      "logps/chosen": -67.0,
+      "logps/rejected": -95.5,
+      "loss": 0.8723,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -0.671875,
+      "rewards/margins": 0.28515625,
+      "rewards/rejected": -0.95703125,
+      "step": 680
+    },
+    {
+      "epoch": 0.7221350078492935,
+      "grad_norm": 1.122558588925505,
+      "learning_rate": 4.704533829076982e-05,
+      "logits/chosen": 0.75390625,
+      "logits/rejected": 1.9609375,
+      "logps/chosen": -81.0,
+      "logps/rejected": -113.5,
+      "loss": 0.8336,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.8125,
+      "rewards/margins": 0.326171875,
+      "rewards/rejected": -1.140625,
+      "step": 690
+    },
+    {
+      "epoch": 0.7326007326007326,
+      "grad_norm": 0.998414171384168,
+      "learning_rate": 4.6900030656055455e-05,
+      "logits/chosen": -0.361328125,
+      "logits/rejected": 0.62890625,
+      "logps/chosen": -59.25,
+      "logps/rejected": -89.0,
+      "loss": 0.8277,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.59375,
+      "rewards/margins": 0.298828125,
+      "rewards/rejected": -0.89453125,
+      "step": 700
+    },
+    {
+      "epoch": 0.7326007326007326,
+      "eval_logits/chosen": -0.79296875,
+      "eval_logits/rejected": 0.349609375,
+      "eval_logps/chosen": -52.0,
+      "eval_logps/rejected": -77.5,
+      "eval_loss": 0.8679062724113464,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.64453125,
+      "eval_rewards/chosen": -0.51953125,
+      "eval_rewards/margins": 0.251953125,
+      "eval_rewards/rejected": -0.7734375,
+      "eval_runtime": 23.4117,
+      "eval_samples_per_second": 85.427,
+      "eval_steps_per_second": 1.367,
+      "step": 700
+    },
+    {
+      "epoch": 0.7430664573521716,
+      "grad_norm": 1.491756760680844,
+      "learning_rate": 4.6751470851128906e-05,
+      "logits/chosen": -1.609375,
+      "logits/rejected": -0.80859375,
+      "logps/chosen": -75.0,
+      "logps/rejected": -96.5,
+      "loss": 0.8871,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -0.75,
+      "rewards/margins": 0.2138671875,
+      "rewards/rejected": -0.96484375,
+      "step": 710
+    },
+    {
+      "epoch": 0.7535321821036107,
+      "grad_norm": 1.0562759343415302,
+      "learning_rate": 4.6599680937227454e-05,
+      "logits/chosen": -1.75,
+      "logits/rejected": -0.66015625,
+      "logps/chosen": -97.5,
+      "logps/rejected": -124.5,
+      "loss": 0.8414,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.97265625,
+      "rewards/margins": 0.2734375,
+      "rewards/rejected": -1.2421875,
+      "step": 720
+    },
+    {
+      "epoch": 0.7639979068550498,
+      "grad_norm": 2.091925185661099,
+      "learning_rate": 4.644468345526183e-05,
+      "logits/chosen": -1.359375,
+      "logits/rejected": -0.427734375,
+      "logps/chosen": -104.5,
+      "logps/rejected": -125.5,
+      "loss": 0.8854,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -1.046875,
+      "rewards/margins": 0.2099609375,
+      "rewards/rejected": -1.25,
+      "step": 730
+    },
+    {
+      "epoch": 0.7744636316064888,
+      "grad_norm": 1.0568462562122483,
+      "learning_rate": 4.628650142246894e-05,
+      "logits/chosen": -0.8359375,
+      "logits/rejected": 0.095703125,
+      "logps/chosen": -87.0,
+      "logps/rejected": -114.0,
+      "loss": 0.8571,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -0.87109375,
+      "rewards/margins": 0.267578125,
+      "rewards/rejected": -1.140625,
+      "step": 740
+    },
+    {
+      "epoch": 0.7849293563579278,
+      "grad_norm": 1.2514436061929697,
+      "learning_rate": 4.6125158328993765e-05,
+      "logits/chosen": -1.1953125,
+      "logits/rejected": -0.2734375,
+      "logps/chosen": -70.5,
+      "logps/rejected": -98.0,
+      "loss": 0.8355,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -0.703125,
+      "rewards/margins": 0.27734375,
+      "rewards/rejected": -0.98046875,
+      "step": 750
+    },
+    {
+      "epoch": 0.7953950811093669,
+      "grad_norm": 1.2722228487490301,
+      "learning_rate": 4.5960678134401064e-05,
+      "logits/chosen": -0.859375,
+      "logits/rejected": -0.1142578125,
+      "logps/chosen": -97.5,
+      "logps/rejected": -123.5,
+      "loss": 0.8602,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.97265625,
+      "rewards/margins": 0.263671875,
+      "rewards/rejected": -1.234375,
+      "step": 760
+    },
+    {
+      "epoch": 0.8058608058608059,
+      "grad_norm": 1.0961720046804988,
+      "learning_rate": 4.579308526411734e-05,
+      "logits/chosen": -1.3125,
+      "logits/rejected": -1.125,
+      "logps/chosen": -71.5,
+      "logps/rejected": -91.5,
+      "loss": 0.8397,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": -0.71484375,
+      "rewards/margins": 0.2021484375,
+      "rewards/rejected": -0.9140625,
+      "step": 770
+    },
+    {
+      "epoch": 0.8163265306122449,
+      "grad_norm": 1.1466951511364993,
+      "learning_rate": 4.562240460580367e-05,
+      "logits/chosen": -2.09375,
+      "logits/rejected": -0.6015625,
+      "logps/chosen": -61.25,
+      "logps/rejected": -99.0,
+      "loss": 0.8578,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -0.61328125,
+      "rewards/margins": 0.376953125,
+      "rewards/rejected": -0.9921875,
+      "step": 780
+    },
+    {
+      "epoch": 0.826792255363684,
+      "grad_norm": 1.1025314474706398,
+      "learning_rate": 4.544866150565985e-05,
+      "logits/chosen": -0.62109375,
+      "logits/rejected": 0.546875,
+      "logps/chosen": -75.5,
+      "logps/rejected": -108.5,
+      "loss": 0.861,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -0.75390625,
+      "rewards/margins": 0.333984375,
+      "rewards/rejected": -1.0859375,
+      "step": 790
+    },
+    {
+      "epoch": 0.837257980115123,
+      "grad_norm": 1.1971574680959491,
+      "learning_rate": 4.527188176466051e-05,
+      "logits/chosen": -0.421875,
+      "logits/rejected": 0.419921875,
+      "logps/chosen": -73.0,
+      "logps/rejected": -104.0,
+      "loss": 0.8341,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.73046875,
+      "rewards/margins": 0.310546875,
+      "rewards/rejected": -1.0390625,
+      "step": 800
+    },
+    {
+      "epoch": 0.837257980115123,
+      "eval_logits/chosen": 0.21875,
+      "eval_logits/rejected": 1.3046875,
+      "eval_logps/chosen": -80.5,
+      "eval_logps/rejected": -108.5,
+      "eval_loss": 0.8503124713897705,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.66015625,
+      "eval_rewards/chosen": -0.8046875,
+      "eval_rewards/margins": 0.27734375,
+      "eval_rewards/rejected": -1.0859375,
+      "eval_runtime": 23.3985,
+      "eval_samples_per_second": 85.476,
+      "eval_steps_per_second": 1.368,
+      "step": 800
+    },
+    {
+      "epoch": 0.847723704866562,
+      "grad_norm": 1.110329403131136,
+      "learning_rate": 4.509209163472357e-05,
+      "logits/chosen": 0.6875,
+      "logits/rejected": 1.9921875,
+      "logps/chosen": -93.0,
+      "logps/rejected": -123.5,
+      "loss": 0.8576,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -0.9296875,
+      "rewards/margins": 0.3046875,
+      "rewards/rejected": -1.234375,
+      "step": 810
+    },
+    {
+      "epoch": 0.858189429618001,
+      "grad_norm": 1.2958416400538662,
+      "learning_rate": 4.4909317814811896e-05,
+      "logits/chosen": 0.73046875,
+      "logits/rejected": 1.9140625,
+      "logps/chosen": -67.5,
+      "logps/rejected": -103.5,
+      "loss": 0.8617,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.67578125,
+      "rewards/margins": 0.361328125,
+      "rewards/rejected": -1.0390625,
+      "step": 820
+    },
+    {
+      "epoch": 0.8686551543694401,
+      "grad_norm": 1.252449986246239,
+      "learning_rate": 4.472358744696845e-05,
+      "logits/chosen": 0.93359375,
+      "logits/rejected": 1.65625,
+      "logps/chosen": -71.0,
+      "logps/rejected": -102.5,
+      "loss": 0.841,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.7109375,
+      "rewards/margins": 0.3125,
+      "rewards/rejected": -1.0234375,
+      "step": 830
+    },
+    {
+      "epoch": 0.8791208791208791,
+      "grad_norm": 1.3103997800625202,
+      "learning_rate": 4.4534928112285645e-05,
+      "logits/chosen": 0.59375,
+      "logits/rejected": 1.71875,
+      "logps/chosen": -79.5,
+      "logps/rejected": -109.5,
+      "loss": 0.8292,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.79296875,
+      "rewards/margins": 0.30078125,
+      "rewards/rejected": -1.09375,
+      "step": 840
+    },
+    {
+      "epoch": 0.8895866038723181,
+      "grad_norm": 1.4551178547211796,
+      "learning_rate": 4.43433678268096e-05,
+      "logits/chosen": 1.0546875,
+      "logits/rejected": 2.015625,
+      "logps/chosen": -80.5,
+      "logps/rejected": -109.0,
+      "loss": 0.8313,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.8046875,
+      "rewards/margins": 0.287109375,
+      "rewards/rejected": -1.09375,
+      "step": 850
+    },
+    {
+      "epoch": 0.9000523286237572,
+      "grad_norm": 1.4054469181948581,
+      "learning_rate": 4.414893503737968e-05,
+      "logits/chosen": 0.69140625,
+      "logits/rejected": 2.171875,
+      "logps/chosen": -84.5,
+      "logps/rejected": -113.0,
+      "loss": 0.848,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.84765625,
+      "rewards/margins": 0.28125,
+      "rewards/rejected": -1.125,
+      "step": 860
+    },
+    {
+      "epoch": 0.9105180533751962,
+      "grad_norm": 1.052453029323795,
+      "learning_rate": 4.395165861740414e-05,
+      "logits/chosen": 1.0234375,
+      "logits/rejected": 2.359375,
+      "logps/chosen": -80.5,
+      "logps/rejected": -107.5,
+      "loss": 0.8657,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -0.80859375,
+      "rewards/margins": 0.26953125,
+      "rewards/rejected": -1.078125,
+      "step": 870
+    },
+    {
+      "epoch": 0.9209837781266352,
+      "grad_norm": 1.1261227211184723,
+      "learning_rate": 4.375156786257241e-05,
+      "logits/chosen": 0.216796875,
+      "logits/rejected": 1.40625,
+      "logps/chosen": -64.5,
+      "logps/rejected": -90.5,
+      "loss": 0.8317,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -0.640625,
+      "rewards/margins": 0.26171875,
+      "rewards/rejected": -0.90234375,
+      "step": 880
+    },
+    {
+      "epoch": 0.9314495028780743,
+      "grad_norm": 1.3242500542063984,
+      "learning_rate": 4.354869248650466e-05,
+      "logits/chosen": 0.036865234375,
+      "logits/rejected": 0.95703125,
+      "logps/chosen": -68.0,
+      "logps/rejected": -92.0,
+      "loss": 0.8647,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -0.6796875,
+      "rewards/margins": 0.240234375,
+      "rewards/rejected": -0.921875,
+      "step": 890
+    },
+    {
+      "epoch": 0.9419152276295133,
+      "grad_norm": 1.3464802768473734,
+      "learning_rate": 4.334306261633931e-05,
+      "logits/chosen": -0.390625,
+      "logits/rejected": 1.1015625,
+      "logps/chosen": -68.5,
+      "logps/rejected": -102.0,
+      "loss": 0.8333,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -0.6875,
+      "rewards/margins": 0.333984375,
+      "rewards/rejected": -1.0234375,
+      "step": 900
+    },
+    {
+      "epoch": 0.9419152276295133,
+      "eval_logits/chosen": 0.640625,
+      "eval_logits/rejected": 1.84375,
+      "eval_logps/chosen": -90.0,
+      "eval_logps/rejected": -121.5,
+      "eval_loss": 0.8453750014305115,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.666015625,
+      "eval_rewards/chosen": -0.8984375,
+      "eval_rewards/margins": 0.318359375,
+      "eval_rewards/rejected": -1.21875,
+      "eval_runtime": 23.3932,
+      "eval_samples_per_second": 85.495,
+      "eval_steps_per_second": 1.368,
+      "step": 900
+    },
+    {
+      "epoch": 0.9523809523809523,
+      "grad_norm": 1.4876700297012682,
+      "learning_rate": 4.313470878825911e-05,
+      "logits/chosen": 0.66796875,
+      "logits/rejected": 1.28125,
+      "logps/chosen": -102.0,
+      "logps/rejected": -123.5,
+      "loss": 0.8806,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -1.015625,
+      "rewards/margins": 0.216796875,
+      "rewards/rejected": -1.234375,
+      "step": 910
+    },
+    {
+      "epoch": 0.9628466771323915,
+      "grad_norm": 1.2593825981046904,
+      "learning_rate": 4.292366194295652e-05,
+      "logits/chosen": -0.80078125,
+      "logits/rejected": 0.58984375,
+      "logps/chosen": -90.5,
+      "logps/rejected": -128.0,
+      "loss": 0.8437,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -0.90234375,
+      "rewards/margins": 0.375,
+      "rewards/rejected": -1.28125,
+      "step": 920
+    },
+    {
+      "epoch": 0.9733124018838305,
+      "grad_norm": 1.7394228452269094,
+      "learning_rate": 4.2709953421038964e-05,
+      "logits/chosen": -0.37109375,
+      "logits/rejected": 1.0390625,
+      "logps/chosen": -77.0,
+      "logps/rejected": -105.5,
+      "loss": 0.8284,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -0.76953125,
+      "rewards/margins": 0.28515625,
+      "rewards/rejected": -1.0546875,
+      "step": 930
+    },
+    {
+      "epoch": 0.9837781266352695,
+      "grad_norm": 2.296358378055376,
+      "learning_rate": 4.249361495837478e-05,
+      "logits/chosen": 0.55859375,
+      "logits/rejected": 1.671875,
+      "logps/chosen": -94.0,
+      "logps/rejected": -120.0,
+      "loss": 0.8552,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -0.9375,
+      "rewards/margins": 0.26171875,
+      "rewards/rejected": -1.203125,
+      "step": 940
+    },
+    {
+      "epoch": 0.9942438513867086,
+      "grad_norm": 2.004976822305089,
+      "learning_rate": 4.227467868138035e-05,
+      "logits/chosen": 0.1611328125,
+      "logits/rejected": 1.53125,
+      "logps/chosen": -94.0,
+      "logps/rejected": -127.0,
+      "loss": 0.8506,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -0.9375,
+      "rewards/margins": 0.330078125,
+      "rewards/rejected": -1.265625,
+      "step": 950
+    },
+    {
+      "epoch": 1.0047095761381475,
+      "grad_norm": 0.995350483676304,
+      "learning_rate": 4.205317710224938e-05,
+      "logits/chosen": -1.390625,
+      "logits/rejected": 0.0751953125,
+      "logps/chosen": -60.0,
+      "logps/rejected": -94.5,
+      "loss": 0.8318,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.6015625,
+      "rewards/margins": 0.34375,
+      "rewards/rejected": -0.94140625,
+      "step": 960
+    },
+    {
+      "epoch": 1.0151753008895865,
+      "grad_norm": 1.5965512835371276,
+      "learning_rate": 4.1829143114124733e-05,
+      "logits/chosen": -0.32421875,
+      "logits/rejected": 1.171875,
+      "logps/chosen": -81.0,
+      "logps/rejected": -125.0,
+      "loss": 0.8086,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.80859375,
+      "rewards/margins": 0.439453125,
+      "rewards/rejected": -1.25,
+      "step": 970
+    },
+    {
+      "epoch": 1.0256410256410255,
+      "grad_norm": 1.1483390223959904,
+      "learning_rate": 4.160260998621387e-05,
+      "logits/chosen": -0.484375,
+      "logits/rejected": 0.96484375,
+      "logps/chosen": -92.5,
+      "logps/rejected": -132.0,
+      "loss": 0.7963,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -0.921875,
+      "rewards/margins": 0.3984375,
+      "rewards/rejected": -1.3203125,
+      "step": 980
+    },
+    {
+      "epoch": 1.0361067503924646,
+      "grad_norm": 1.3798272414702553,
+      "learning_rate": 4.137361135884824e-05,
+      "logits/chosen": -1.5078125,
+      "logits/rejected": 0.7265625,
+      "logps/chosen": -71.5,
+      "logps/rejected": -131.0,
+      "loss": 0.7662,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": -0.71484375,
+      "rewards/margins": 0.59765625,
+      "rewards/rejected": -1.3125,
+      "step": 990
+    },
+    {
+      "epoch": 1.0465724751439036,
+      "grad_norm": 1.538458329174527,
+      "learning_rate": 4.114218123848777e-05,
+      "logits/chosen": -1.0234375,
+      "logits/rejected": 0.3984375,
+      "logps/chosen": -80.5,
+      "logps/rejected": -122.5,
+      "loss": 0.8071,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -0.8046875,
+      "rewards/margins": 0.419921875,
+      "rewards/rejected": -1.21875,
+      "step": 1000
+    },
+    {
+      "epoch": 1.0465724751439036,
+      "eval_logits/chosen": 0.150390625,
+      "eval_logits/rejected": 1.3515625,
+      "eval_logps/chosen": -100.0,
+      "eval_logps/rejected": -133.0,
+      "eval_loss": 0.8440937399864197,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.669921875,
+      "eval_rewards/chosen": -1.0,
+      "eval_rewards/margins": 0.333984375,
+      "eval_rewards/rejected": -1.3359375,
+      "eval_runtime": 23.3967,
+      "eval_samples_per_second": 85.482,
+      "eval_steps_per_second": 1.368,
+      "step": 1000
+    },
+    {
+      "epoch": 1.0570381998953426,
+      "grad_norm": 1.6811244112623198,
+      "learning_rate": 4.0908353992670824e-05,
+      "logits/chosen": -0.375,
+      "logits/rejected": 1.421875,
+      "logps/chosen": -86.5,
+      "logps/rejected": -138.0,
+      "loss": 0.7767,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": -0.8671875,
+      "rewards/margins": 0.515625,
+      "rewards/rejected": -1.3828125,
+      "step": 1010
+    },
+    {
+      "epoch": 1.0675039246467817,
+      "grad_norm": 1.6583170336487807,
+      "learning_rate": 4.067216434491059e-05,
+      "logits/chosen": 0.09130859375,
+      "logits/rejected": 1.71875,
+      "logps/chosen": -100.5,
+      "logps/rejected": -147.0,
+      "loss": 0.785,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -1.0078125,
+      "rewards/margins": 0.46484375,
+      "rewards/rejected": -1.46875,
+      "step": 1020
+    },
+    {
+      "epoch": 1.077969649398221,
+      "grad_norm": 1.8616551332146642,
+      "learning_rate": 4.043364736953866e-05,
+      "logits/chosen": -0.439453125,
+      "logits/rejected": 1.15625,
+      "logps/chosen": -100.0,
+      "logps/rejected": -138.0,
+      "loss": 0.7688,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -1.0,
+      "rewards/margins": 0.384765625,
+      "rewards/rejected": -1.3828125,
+      "step": 1030
+    },
+    {
+      "epoch": 1.08843537414966,
+      "grad_norm": 1.7008034994779155,
+      "learning_rate": 4.019283848649643e-05,
+      "logits/chosen": -0.9921875,
+      "logits/rejected": 0.9296875,
+      "logps/chosen": -75.0,
+      "logps/rejected": -122.5,
+      "loss": 0.762,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -0.75,
+      "rewards/margins": 0.4765625,
+      "rewards/rejected": -1.2265625,
+      "step": 1040
+    },
+    {
+      "epoch": 1.098901098901099,
+      "grad_norm": 2.09872583052328,
+      "learning_rate": 3.9949773456075215e-05,
+      "logits/chosen": -1.0390625,
+      "logits/rejected": 0.435546875,
+      "logps/chosen": -89.5,
+      "logps/rejected": -132.0,
+      "loss": 0.7685,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -0.89453125,
+      "rewards/margins": 0.42578125,
+      "rewards/rejected": -1.3203125,
+      "step": 1050
+    },
+    {
+      "epoch": 1.109366823652538,
+      "grad_norm": 2.0606095039467864,
+      "learning_rate": 3.970448837360585e-05,
+      "logits/chosen": -1.734375,
+      "logits/rejected": 0.828125,
+      "logps/chosen": -72.5,
+      "logps/rejected": -133.0,
+      "loss": 0.769,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.78125,
+      "rewards/chosen": -0.7265625,
+      "rewards/margins": 0.609375,
+      "rewards/rejected": -1.3359375,
+      "step": 1060
+    },
+    {
+      "epoch": 1.119832548403977,
+      "grad_norm": 1.7368032948244938,
+      "learning_rate": 3.945701966409846e-05,
+      "logits/chosen": -1.3828125,
+      "logits/rejected": 0.60546875,
+      "logps/chosen": -86.0,
+      "logps/rejected": -142.0,
+      "loss": 0.7609,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -0.859375,
+      "rewards/margins": 0.56640625,
+      "rewards/rejected": -1.421875,
+      "step": 1070
+    },
+    {
+      "epoch": 1.130298273155416,
+      "grad_norm": 1.7921429539269247,
+      "learning_rate": 3.9207404076833374e-05,
+      "logits/chosen": -1.3125,
+      "logits/rejected": 0.486328125,
+      "logps/chosen": -77.0,
+      "logps/rejected": -143.0,
+      "loss": 0.733,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.793749988079071,
+      "rewards/chosen": -0.765625,
+      "rewards/margins": 0.66015625,
+      "rewards/rejected": -1.4296875,
+      "step": 1080
+    },
+    {
+      "epoch": 1.1407639979068551,
+      "grad_norm": 1.4176720705563766,
+      "learning_rate": 3.895567867990379e-05,
+      "logits/chosen": -1.765625,
+      "logits/rejected": 0.375,
+      "logps/chosen": -85.0,
+      "logps/rejected": -154.0,
+      "loss": 0.7498,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -0.8515625,
+      "rewards/margins": 0.69140625,
+      "rewards/rejected": -1.5390625,
+      "step": 1090
+    },
+    {
+      "epoch": 1.1512297226582942,
+      "grad_norm": 2.0691789103139286,
+      "learning_rate": 3.870188085471114e-05,
+      "logits/chosen": -1.6640625,
+      "logits/rejected": 0.04931640625,
+      "logps/chosen": -74.5,
+      "logps/rejected": -121.5,
+      "loss": 0.7845,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -0.7421875,
+      "rewards/margins": 0.46875,
+      "rewards/rejected": -1.2109375,
+      "step": 1100
+    },
+    {
+      "epoch": 1.1512297226582942,
+      "eval_logits/chosen": -1.5078125,
+      "eval_logits/rejected": 0.330078125,
+      "eval_logps/chosen": -84.5,
+      "eval_logps/rejected": -122.5,
+      "eval_loss": 0.8306562304496765,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.666015625,
+      "eval_rewards/chosen": -0.84765625,
+      "eval_rewards/margins": 0.380859375,
+      "eval_rewards/rejected": -1.2265625,
+      "eval_runtime": 23.4033,
+      "eval_samples_per_second": 85.458,
+      "eval_steps_per_second": 1.367,
+      "step": 1100
+    },
+    {
+      "epoch": 1.1616954474097332,
+      "grad_norm": 1.4716680743666684,
+      "learning_rate": 3.844604829041395e-05,
+      "logits/chosen": -1.4296875,
+      "logits/rejected": 0.72265625,
+      "logps/chosen": -85.5,
+      "logps/rejected": -147.0,
+      "loss": 0.7435,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": -0.8515625,
+      "rewards/margins": 0.61328125,
+      "rewards/rejected": -1.46875,
+      "step": 1110
+    },
+    {
+      "epoch": 1.1721611721611722,
+      "grad_norm": 2.779618449951347,
+      "learning_rate": 3.818821897833095e-05,
+      "logits/chosen": -2.734375,
+      "logits/rejected": -0.65234375,
+      "logps/chosen": -64.0,
+      "logps/rejected": -116.5,
+      "loss": 0.758,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.640625,
+      "rewards/margins": 0.5234375,
+      "rewards/rejected": -1.1640625,
+      "step": 1120
+    },
+    {
+      "epoch": 1.1826268969126112,
+      "grad_norm": 1.5644794128490174,
+      "learning_rate": 3.792843120629935e-05,
+      "logits/chosen": -3.234375,
+      "logits/rejected": -1.1328125,
+      "logps/chosen": -65.0,
+      "logps/rejected": -114.0,
+      "loss": 0.7643,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -0.65234375,
+      "rewards/margins": 0.490234375,
+      "rewards/rejected": -1.140625,
+      "step": 1130
+    },
+    {
+      "epoch": 1.1930926216640503,
+      "grad_norm": 1.637679154307981,
+      "learning_rate": 3.766672355298907e-05,
+      "logits/chosen": -1.5859375,
+      "logits/rejected": 0.1279296875,
+      "logps/chosen": -87.0,
+      "logps/rejected": -143.0,
+      "loss": 0.7611,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": -0.8671875,
+      "rewards/margins": 0.55859375,
+      "rewards/rejected": -1.4296875,
+      "step": 1140
+    },
+    {
+      "epoch": 1.2035583464154893,
+      "grad_norm": 2.2610418724389065,
+      "learning_rate": 3.740313488217373e-05,
+      "logits/chosen": -1.6953125,
+      "logits/rejected": -0.392578125,
+      "logps/chosen": -73.0,
+      "logps/rejected": -121.0,
+      "loss": 0.7579,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.73046875,
+      "rewards/margins": 0.478515625,
+      "rewards/rejected": -1.2109375,
+      "step": 1150
+    },
+    {
+      "epoch": 1.2140240711669283,
+      "grad_norm": 3.048343986458122,
+      "learning_rate": 3.7137704336959464e-05,
+      "logits/chosen": -0.96875,
+      "logits/rejected": 1.109375,
+      "logps/chosen": -88.0,
+      "logps/rejected": -146.0,
+      "loss": 0.78,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.87890625,
+      "rewards/margins": 0.58203125,
+      "rewards/rejected": -1.4609375,
+      "step": 1160
+    },
+    {
+      "epoch": 1.2244897959183674,
+      "grad_norm": 3.547482841978148,
+      "learning_rate": 3.687047133397201e-05,
+      "logits/chosen": -0.890625,
+      "logits/rejected": 0.7578125,
+      "logps/chosen": -90.0,
+      "logps/rejected": -145.0,
+      "loss": 0.7749,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -0.8984375,
+      "rewards/margins": 0.55078125,
+      "rewards/rejected": -1.4453125,
+      "step": 1170
+    },
+    {
+      "epoch": 1.2349555206698064,
+      "grad_norm": 2.313639296898369,
+      "learning_rate": 3.6601475557503415e-05,
+      "logits/chosen": -0.51171875,
+      "logits/rejected": 1.078125,
+      "logps/chosen": -100.0,
+      "logps/rejected": -159.0,
+      "loss": 0.7669,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -1.0,
+      "rewards/margins": 0.59375,
+      "rewards/rejected": -1.59375,
+      "step": 1180
+    },
+    {
+      "epoch": 1.2454212454212454,
+      "grad_norm": 2.1254126808194416,
+      "learning_rate": 3.633075695361882e-05,
+      "logits/chosen": -0.318359375,
+      "logits/rejected": 1.171875,
+      "logps/chosen": -109.5,
+      "logps/rejected": -157.0,
+      "loss": 0.7384,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -1.09375,
+      "rewards/margins": 0.474609375,
+      "rewards/rejected": -1.5625,
+      "step": 1190
+    },
+    {
+      "epoch": 1.2558869701726845,
+      "grad_norm": 2.0051911796504758,
+      "learning_rate": 3.6058355724224477e-05,
+      "logits/chosen": -1.4609375,
+      "logits/rejected": 0.28125,
+      "logps/chosen": -86.0,
+      "logps/rejected": -134.0,
+      "loss": 0.7483,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": -0.859375,
+      "rewards/margins": 0.48046875,
+      "rewards/rejected": -1.34375,
+      "step": 1200
+    },
+    {
+      "epoch": 1.2558869701726845,
+      "eval_logits/chosen": -0.416015625,
+      "eval_logits/rejected": 0.98046875,
+      "eval_logps/chosen": -94.5,
+      "eval_logps/rejected": -133.0,
+      "eval_loss": 0.8353124856948853,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.67578125,
+      "eval_rewards/chosen": -0.9453125,
+      "eval_rewards/margins": 0.380859375,
+      "eval_rewards/rejected": -1.328125,
+      "eval_runtime": 23.393,
+      "eval_samples_per_second": 85.496,
+      "eval_steps_per_second": 1.368,
+      "step": 1200
+    },
+    {
+      "epoch": 1.2663526949241235,
+      "grad_norm": 2.346021319411282,
+      "learning_rate": 3.5784312321097744e-05,
+      "logits/chosen": -1.390625,
+      "logits/rejected": 0.8046875,
+      "logps/chosen": -82.0,
+      "logps/rejected": -147.0,
+      "loss": 0.7576,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.81640625,
+      "rewards/margins": 0.6484375,
+      "rewards/rejected": -1.46875,
+      "step": 1210
+    },
+    {
+      "epoch": 1.2768184196755625,
+      "grad_norm": 1.9344029206126447,
+      "learning_rate": 3.550866743987994e-05,
+      "logits/chosen": -0.76953125,
+      "logits/rejected": 1.203125,
+      "logps/chosen": -87.5,
+      "logps/rejected": -147.0,
+      "loss": 0.7815,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -0.875,
+      "rewards/margins": 0.59375,
+      "rewards/rejected": -1.46875,
+      "step": 1220
+    },
+    {
+      "epoch": 1.2872841444270016,
+      "grad_norm": 4.4366431884203195,
+      "learning_rate": 3.5231462014033024e-05,
+      "logits/chosen": -0.18359375,
+      "logits/rejected": 1.4765625,
+      "logps/chosen": -88.0,
+      "logps/rejected": -155.0,
+      "loss": 0.7111,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -0.8828125,
+      "rewards/margins": 0.6640625,
+      "rewards/rejected": -1.546875,
+      "step": 1230
+    },
+    {
+      "epoch": 1.2977498691784406,
+      "grad_norm": 2.956627226386763,
+      "learning_rate": 3.495273720876095e-05,
+      "logits/chosen": -0.283203125,
+      "logits/rejected": 1.3046875,
+      "logps/chosen": -96.0,
+      "logps/rejected": -153.0,
+      "loss": 0.7752,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -0.9609375,
+      "rewards/margins": 0.5703125,
+      "rewards/rejected": -1.53125,
+      "step": 1240
+    },
+    {
+      "epoch": 1.3082155939298796,
+      "grad_norm": 2.270966311325831,
+      "learning_rate": 3.4672534414896616e-05,
+      "logits/chosen": -0.251953125,
+      "logits/rejected": 1.6953125,
+      "logps/chosen": -99.5,
+      "logps/rejected": -156.0,
+      "loss": 0.7312,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -0.9921875,
+      "rewards/margins": 0.5625,
+      "rewards/rejected": -1.5625,
+      "step": 1250
+    },
+    {
+      "epoch": 1.3186813186813187,
+      "grad_norm": 3.304649619384499,
+      "learning_rate": 3.439089524275526e-05,
+      "logits/chosen": 0.392578125,
+      "logits/rejected": 1.9765625,
+      "logps/chosen": -102.5,
+      "logps/rejected": -155.0,
+      "loss": 0.7594,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -1.0234375,
+      "rewards/margins": 0.52734375,
+      "rewards/rejected": -1.5546875,
+      "step": 1260
+    },
+    {
+      "epoch": 1.3291470434327577,
+      "grad_norm": 2.2814250072882656,
+      "learning_rate": 3.410786151595533e-05,
+      "logits/chosen": 0.390625,
+      "logits/rejected": 1.9921875,
+      "logps/chosen": -110.0,
+      "logps/rejected": -157.0,
+      "loss": 0.7922,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -1.1015625,
+      "rewards/margins": 0.470703125,
+      "rewards/rejected": -1.5703125,
+      "step": 1270
+    },
+    {
+      "epoch": 1.3396127681841967,
+      "grad_norm": 2.3138135502070476,
+      "learning_rate": 3.382347526520765e-05,
+      "logits/chosen": 0.31640625,
+      "logits/rejected": 2.578125,
+      "logps/chosen": -105.0,
+      "logps/rejected": -168.0,
+      "loss": 0.7339,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.768750011920929,
+      "rewards/chosen": -1.046875,
+      "rewards/margins": 0.6328125,
+      "rewards/rejected": -1.6875,
+      "step": 1280
+    },
+    {
+      "epoch": 1.3500784929356358,
+      "grad_norm": 1.7668407295880866,
+      "learning_rate": 3.353777872207381e-05,
+      "logits/chosen": 0.349609375,
+      "logits/rejected": 2.171875,
+      "logps/chosen": -108.0,
+      "logps/rejected": -168.0,
+      "loss": 0.755,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": -1.078125,
+      "rewards/margins": 0.59765625,
+      "rewards/rejected": -1.6796875,
+      "step": 1290
+    },
+    {
+      "epoch": 1.3605442176870748,
+      "grad_norm": 2.2238858170319444,
+      "learning_rate": 3.325081431269475e-05,
+      "logits/chosen": -0.78125,
+      "logits/rejected": 0.99609375,
+      "logps/chosen": -84.5,
+      "logps/rejected": -130.0,
+      "loss": 0.7802,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": -0.84375,
+      "rewards/margins": 0.45703125,
+      "rewards/rejected": -1.3046875,
+      "step": 1300
+    },
+    {
+      "epoch": 1.3605442176870748,
+      "eval_logits/chosen": -1.5859375,
+      "eval_logits/rejected": 0.341796875,
+      "eval_logps/chosen": -62.0,
+      "eval_logps/rejected": -100.5,
+      "eval_loss": 0.836312472820282,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.705078125,
+      "eval_rewards/chosen": -0.62109375,
+      "eval_rewards/margins": 0.3828125,
+      "eval_rewards/rejected": -1.0,
+      "eval_runtime": 23.6314,
+      "eval_samples_per_second": 84.633,
+      "eval_steps_per_second": 1.354,
+      "step": 1300
+    },
+    {
+      "epoch": 1.3710099424385138,
+      "grad_norm": 1.9690849737568896,
+      "learning_rate": 3.296262465149046e-05,
+      "logits/chosen": -2.140625,
+      "logits/rejected": 0.4140625,
+      "logps/chosen": -60.5,
+      "logps/rejected": -119.0,
+      "loss": 0.7729,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.60546875,
+      "rewards/margins": 0.5859375,
+      "rewards/rejected": -1.1875,
+      "step": 1310
+    },
+    {
+      "epoch": 1.3814756671899528,
+      "grad_norm": 2.3884339066598628,
+      "learning_rate": 3.2673252534831687e-05,
+      "logits/chosen": -0.859375,
+      "logits/rejected": 0.84375,
+      "logps/chosen": -88.5,
+      "logps/rejected": -146.0,
+      "loss": 0.7523,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.88671875,
+      "rewards/margins": 0.578125,
+      "rewards/rejected": -1.4609375,
+      "step": 1320
+    },
+    {
+      "epoch": 1.3919413919413919,
+      "grad_norm": 2.3514873538785204,
+      "learning_rate": 3.2382740934684696e-05,
+      "logits/chosen": -0.357421875,
+      "logits/rejected": 1.1015625,
+      "logps/chosen": -101.5,
+      "logps/rejected": -151.0,
+      "loss": 0.7358,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -1.015625,
+      "rewards/margins": 0.494140625,
+      "rewards/rejected": -1.5078125,
+      "step": 1330
+    },
+    {
+      "epoch": 1.402407116692831,
+      "grad_norm": 2.682980997199581,
+      "learning_rate": 3.209113299222982e-05,
+      "logits/chosen": -0.75,
+      "logits/rejected": 1.34375,
+      "logps/chosen": -92.0,
+      "logps/rejected": -153.0,
+      "loss": 0.7303,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": -0.91796875,
+      "rewards/margins": 0.609375,
+      "rewards/rejected": -1.5234375,
+      "step": 1340
+    },
+    {
+      "epoch": 1.41287284144427,
+      "grad_norm": 2.2441887586235123,
+      "learning_rate": 3.179847201145505e-05,
+      "logits/chosen": -1.1796875,
+      "logits/rejected": 0.671875,
+      "logps/chosen": -77.0,
+      "logps/rejected": -139.0,
+      "loss": 0.7398,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -0.7734375,
+      "rewards/margins": 0.62109375,
+      "rewards/rejected": -1.390625,
+      "step": 1350
+    },
+    {
+      "epoch": 1.423338566195709,
+      "grad_norm": 2.622246518582053,
+      "learning_rate": 3.150480145272528e-05,
+      "logits/chosen": -0.66015625,
+      "logits/rejected": 1.140625,
+      "logps/chosen": -85.5,
+      "logps/rejected": -153.0,
+      "loss": 0.7186,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -0.85546875,
+      "rewards/margins": 0.67578125,
+      "rewards/rejected": -1.53125,
+      "step": 1360
+    },
+    {
+      "epoch": 1.433804290947148,
+      "grad_norm": 1.8424019818433308,
+      "learning_rate": 3.121016492632848e-05,
+      "logits/chosen": -0.11328125,
+      "logits/rejected": 1.375,
+      "logps/chosen": -110.0,
+      "logps/rejected": -158.0,
+      "loss": 0.742,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -1.1015625,
+      "rewards/margins": 0.478515625,
+      "rewards/rejected": -1.578125,
+      "step": 1370
+    },
+    {
+      "epoch": 1.4442700156985873,
+      "grad_norm": 2.122770926458258,
+      "learning_rate": 3.091460618599951e-05,
+      "logits/chosen": -0.609375,
+      "logits/rejected": 0.94140625,
+      "logps/chosen": -90.0,
+      "logps/rejected": -143.0,
+      "loss": 0.7337,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -0.90234375,
+      "rewards/margins": 0.53515625,
+      "rewards/rejected": -1.4296875,
+      "step": 1380
+    },
+    {
+      "epoch": 1.454735740450026,
+      "grad_norm": 2.38055285633703,
+      "learning_rate": 3.061816912242265e-05,
+      "logits/chosen": -0.7265625,
+      "logits/rejected": 1.46875,
+      "logps/chosen": -99.5,
+      "logps/rejected": -156.0,
+      "loss": 0.751,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.99609375,
+      "rewards/margins": 0.5703125,
+      "rewards/rejected": -1.5625,
+      "step": 1390
+    },
+    {
+      "epoch": 1.4652014652014653,
+      "grad_norm": 2.857905249445792,
+      "learning_rate": 3.0320897756713785e-05,
+      "logits/chosen": -0.60546875,
+      "logits/rejected": 0.828125,
+      "logps/chosen": -94.5,
+      "logps/rejected": -147.0,
+      "loss": 0.7499,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -0.9453125,
+      "rewards/margins": 0.52734375,
+      "rewards/rejected": -1.46875,
+      "step": 1400
+    },
+    {
+      "epoch": 1.4652014652014653,
+      "eval_logits/chosen": -0.171875,
+      "eval_logits/rejected": 1.453125,
+      "eval_logps/chosen": -97.0,
+      "eval_logps/rejected": -141.0,
+      "eval_loss": 0.8227812647819519,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.701171875,
+      "eval_rewards/chosen": -0.97265625,
+      "eval_rewards/margins": 0.44140625,
+      "eval_rewards/rejected": -1.4140625,
+      "eval_runtime": 23.4077,
+      "eval_samples_per_second": 85.442,
+      "eval_steps_per_second": 1.367,
+      "step": 1400
+    },
+    {
+      "epoch": 1.4756671899529041,
+      "grad_norm": 2.760178688324096,
+      "learning_rate": 3.002283623388332e-05,
+      "logits/chosen": -0.423828125,
+      "logits/rejected": 1.484375,
+      "logps/chosen": -89.5,
+      "logps/rejected": -161.0,
+      "loss": 0.713,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.768750011920929,
+      "rewards/chosen": -0.89453125,
+      "rewards/margins": 0.71875,
+      "rewards/rejected": -1.6171875,
+      "step": 1410
+    },
+    {
+      "epoch": 1.4861329147043434,
+      "grad_norm": 2.333266803030169,
+      "learning_rate": 2.9724028816280507e-05,
+      "logits/chosen": 0.12451171875,
+      "logits/rejected": 1.703125,
+      "logps/chosen": -101.5,
+      "logps/rejected": -153.0,
+      "loss": 0.7393,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -1.015625,
+      "rewards/margins": 0.51171875,
+      "rewards/rejected": -1.53125,
+      "step": 1420
+    },
+    {
+      "epoch": 1.4965986394557822,
+      "grad_norm": 2.5547709281271604,
+      "learning_rate": 2.9424519877020524e-05,
+      "logits/chosen": -0.64453125,
+      "logits/rejected": 1.578125,
+      "logps/chosen": -96.5,
+      "logps/rejected": -174.0,
+      "loss": 0.6987,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.78125,
+      "rewards/chosen": -0.96484375,
+      "rewards/margins": 0.7734375,
+      "rewards/rejected": -1.734375,
+      "step": 1430
+    },
+    {
+      "epoch": 1.5070643642072215,
+      "grad_norm": 2.230798311247942,
+      "learning_rate": 2.912435389339504e-05,
+      "logits/chosen": -1.0078125,
+      "logits/rejected": 1.4140625,
+      "logps/chosen": -96.5,
+      "logps/rejected": -165.0,
+      "loss": 0.7295,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -0.96484375,
+      "rewards/margins": 0.6875,
+      "rewards/rejected": -1.65625,
+      "step": 1440
+    },
+    {
+      "epoch": 1.5175300889586603,
+      "grad_norm": 2.5463058319121075,
+      "learning_rate": 2.8823575440267226e-05,
+      "logits/chosen": -0.70703125,
+      "logits/rejected": 1.453125,
+      "logps/chosen": -102.0,
+      "logps/rejected": -161.0,
+      "loss": 0.7246,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -1.0234375,
+      "rewards/margins": 0.58984375,
+      "rewards/rejected": -1.609375,
+      "step": 1450
+    },
+    {
+      "epoch": 1.5279958137100995,
+      "grad_norm": 2.2696178543142462,
+      "learning_rate": 2.8522229183452465e-05,
+      "logits/chosen": -1.15625,
+      "logits/rejected": 1.5546875,
+      "logps/chosen": -78.0,
+      "logps/rejected": -152.0,
+      "loss": 0.7241,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": -0.78125,
+      "rewards/margins": 0.73828125,
+      "rewards/rejected": -1.5234375,
+      "step": 1460
+    },
+    {
+      "epoch": 1.5384615384615383,
+      "grad_norm": 2.4277396763770667,
+      "learning_rate": 2.8220359873085373e-05,
+      "logits/chosen": -0.06396484375,
+      "logits/rejected": 2.015625,
+      "logps/chosen": -114.0,
+      "logps/rejected": -177.0,
+      "loss": 0.7625,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -1.140625,
+      "rewards/margins": 0.6328125,
+      "rewards/rejected": -1.7734375,
+      "step": 1470
+    },
+    {
+      "epoch": 1.5489272632129776,
+      "grad_norm": 2.4193616965160536,
+      "learning_rate": 2.7918012336974388e-05,
+      "logits/chosen": 0.625,
+      "logits/rejected": 2.375,
+      "logps/chosen": -101.5,
+      "logps/rejected": -164.0,
+      "loss": 0.7303,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -1.015625,
+      "rewards/margins": 0.62890625,
+      "rewards/rejected": -1.640625,
+      "step": 1480
+    },
+    {
+      "epoch": 1.5593929879644164,
+      "grad_norm": 2.759465996192548,
+      "learning_rate": 2.7615231473944815e-05,
+      "logits/chosen": -0.65625,
+      "logits/rejected": 1.484375,
+      "logps/chosen": -88.5,
+      "logps/rejected": -155.0,
+      "loss": 0.7085,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.88671875,
+      "rewards/margins": 0.6640625,
+      "rewards/rejected": -1.546875,
+      "step": 1490
+    },
+    {
+      "epoch": 1.5698587127158556,
+      "grad_norm": 2.0440248665227148,
+      "learning_rate": 2.7312062247171327e-05,
+      "logits/chosen": -2.125,
+      "logits/rejected": 0.328125,
+      "logps/chosen": -81.5,
+      "logps/rejected": -148.0,
+      "loss": 0.6966,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.8125,
+      "rewards/margins": 0.66796875,
+      "rewards/rejected": -1.484375,
+      "step": 1500
+    },
+    {
+      "epoch": 1.5698587127158556,
+      "eval_logits/chosen": -0.330078125,
+      "eval_logits/rejected": 1.5,
+      "eval_logps/chosen": -106.0,
+      "eval_logps/rejected": -152.0,
+      "eval_loss": 0.8231250047683716,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.68359375,
+      "eval_rewards/chosen": -1.0625,
+      "eval_rewards/margins": 0.4609375,
+      "eval_rewards/rejected": -1.5234375,
+      "eval_runtime": 23.4098,
+      "eval_samples_per_second": 85.434,
+      "eval_steps_per_second": 1.367,
+      "step": 1500
+    },
+    {
+      "epoch": 1.5803244374672945,
+      "grad_norm": 2.4824081341310205,
+      "learning_rate": 2.7008549677500876e-05,
+      "logits/chosen": -1.3046875,
+      "logits/rejected": 1.0390625,
+      "logps/chosen": -88.0,
+      "logps/rejected": -166.0,
+      "loss": 0.6871,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": -0.8828125,
+      "rewards/margins": 0.77734375,
+      "rewards/rejected": -1.65625,
+      "step": 1510
+    },
+    {
+      "epoch": 1.5907901622187337,
+      "grad_norm": 2.533797477410996,
+      "learning_rate": 2.670473883676709e-05,
+      "logits/chosen": -1.4375,
+      "logits/rejected": 0.9609375,
+      "logps/chosen": -98.0,
+      "logps/rejected": -172.0,
+      "loss": 0.7064,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": -0.98046875,
+      "rewards/margins": 0.7421875,
+      "rewards/rejected": -1.71875,
+      "step": 1520
+    },
+    {
+      "epoch": 1.6012558869701727,
+      "grad_norm": 2.230305178898535,
+      "learning_rate": 2.6400674841097074e-05,
+      "logits/chosen": -1.0625,
+      "logits/rejected": 1.7734375,
+      "logps/chosen": -106.0,
+      "logps/rejected": -175.0,
+      "loss": 0.73,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.78125,
+      "rewards/chosen": -1.0546875,
+      "rewards/margins": 0.6953125,
+      "rewards/rejected": -1.75,
+      "step": 1530
+    },
+    {
+      "epoch": 1.6117216117216118,
+      "grad_norm": 2.6180685065201525,
+      "learning_rate": 2.609640284421158e-05,
+      "logits/chosen": -1.4375,
+      "logits/rejected": 1.3671875,
+      "logps/chosen": -94.0,
+      "logps/rejected": -171.0,
+      "loss": 0.7113,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -0.94140625,
+      "rewards/margins": 0.76953125,
+      "rewards/rejected": -1.7109375,
+      "step": 1540
+    },
+    {
+      "epoch": 1.6221873364730508,
+      "grad_norm": 3.645930343772631,
+      "learning_rate": 2.5791968030719726e-05,
+      "logits/chosen": -2.578125,
+      "logits/rejected": 0.51953125,
+      "logps/chosen": -82.0,
+      "logps/rejected": -159.0,
+      "loss": 0.7455,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": -0.81640625,
+      "rewards/margins": 0.7734375,
+      "rewards/rejected": -1.59375,
+      "step": 1550
+    },
+    {
+      "epoch": 1.6326530612244898,
+      "grad_norm": 2.8006248835063596,
+      "learning_rate": 2.548741560940902e-05,
+      "logits/chosen": -1.703125,
+      "logits/rejected": 1.4453125,
+      "logps/chosen": -75.0,
+      "logps/rejected": -161.0,
+      "loss": 0.7256,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.78125,
+      "rewards/chosen": -0.75,
+      "rewards/margins": 0.86328125,
+      "rewards/rejected": -1.609375,
+      "step": 1560
+    },
+    {
+      "epoch": 1.6431187859759289,
+      "grad_norm": 2.3438771682827086,
+      "learning_rate": 2.5182790806531787e-05,
+      "logits/chosen": -0.271484375,
+      "logits/rejected": 2.21875,
+      "logps/chosen": -106.5,
+      "logps/rejected": -179.0,
+      "loss": 0.7113,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": -1.0703125,
+      "rewards/margins": 0.7265625,
+      "rewards/rejected": -1.7890625,
+      "step": 1570
+    },
+    {
+      "epoch": 1.653584510727368,
+      "grad_norm": 2.0876809724125533,
+      "learning_rate": 2.4878138859089074e-05,
+      "logits/chosen": -0.1396484375,
+      "logits/rejected": 1.984375,
+      "logps/chosen": -91.5,
+      "logps/rejected": -158.0,
+      "loss": 0.7082,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -0.9140625,
+      "rewards/margins": 0.6640625,
+      "rewards/rejected": -1.578125,
+      "step": 1580
+    },
+    {
+      "epoch": 1.664050235478807,
+      "grad_norm": 2.54073526422114,
+      "learning_rate": 2.4573505008112924e-05,
+      "logits/chosen": 0.02099609375,
+      "logits/rejected": 2.421875,
+      "logps/chosen": -80.5,
+      "logps/rejected": -165.0,
+      "loss": 0.6965,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": -0.8046875,
+      "rewards/margins": 0.84765625,
+      "rewards/rejected": -1.65625,
+      "step": 1590
+    },
+    {
+      "epoch": 1.674515960230246,
+      "grad_norm": 2.2237508509252626,
+      "learning_rate": 2.4268934491948032e-05,
+      "logits/chosen": 0.109375,
+      "logits/rejected": 2.125,
+      "logps/chosen": -84.5,
+      "logps/rejected": -156.0,
+      "loss": 0.6921,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.768750011920929,
+      "rewards/chosen": -0.84765625,
+      "rewards/margins": 0.71484375,
+      "rewards/rejected": -1.5625,
+      "step": 1600
+    },
+    {
+      "epoch": 1.674515960230246,
+      "eval_logits/chosen": 0.61328125,
+      "eval_logits/rejected": 2.25,
+      "eval_logps/chosen": -107.0,
+      "eval_logps/rejected": -155.0,
+      "eval_loss": 0.8222187757492065,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.6875,
+      "eval_rewards/chosen": -1.0703125,
+      "eval_rewards/margins": 0.4765625,
+      "eval_rewards/rejected": -1.546875,
+      "eval_runtime": 23.4079,
+      "eval_samples_per_second": 85.441,
+      "eval_steps_per_second": 1.367,
+      "step": 1600
+    },
+    {
+      "epoch": 1.684981684981685,
+      "grad_norm": 2.010552597787752,
+      "learning_rate": 2.396447253953385e-05,
+      "logits/chosen": -0.34375,
+      "logits/rejected": 2.0,
+      "logps/chosen": -99.5,
+      "logps/rejected": -173.0,
+      "loss": 0.7394,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": -0.99609375,
+      "rewards/margins": 0.7265625,
+      "rewards/rejected": -1.7265625,
+      "step": 1610
+    },
+    {
+      "epoch": 1.695447409733124,
+      "grad_norm": 2.5200942468299803,
+      "learning_rate": 2.3660164363687998e-05,
+      "logits/chosen": -0.38671875,
+      "logits/rejected": 1.7109375,
+      "logps/chosen": -84.5,
+      "logps/rejected": -157.0,
+      "loss": 0.7174,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": -0.84375,
+      "rewards/margins": 0.72265625,
+      "rewards/rejected": -1.5625,
+      "step": 1620
+    },
+    {
+      "epoch": 1.705913134484563,
+      "grad_norm": 2.876092753010097,
+      "learning_rate": 2.33560551543922e-05,
+      "logits/chosen": -0.004302978515625,
+      "logits/rejected": 1.3984375,
+      "logps/chosen": -92.5,
+      "logps/rejected": -158.0,
+      "loss": 0.6984,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -0.921875,
+      "rewards/margins": 0.65625,
+      "rewards/rejected": -1.578125,
+      "step": 1630
+    },
+    {
+      "epoch": 1.716378859236002,
+      "grad_norm": 2.2335658485128316,
+      "learning_rate": 2.3052190072081493e-05,
+      "logits/chosen": -0.890625,
+      "logits/rejected": 1.125,
+      "logps/chosen": -91.0,
+      "logps/rejected": -154.0,
+      "loss": 0.7321,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -0.90625,
+      "rewards/margins": 0.6328125,
+      "rewards/rejected": -1.5390625,
+      "step": 1640
+    },
+    {
+      "epoch": 1.7268445839874411,
+      "grad_norm": 2.133850630539174,
+      "learning_rate": 2.2748614240937864e-05,
+      "logits/chosen": -1.5078125,
+      "logits/rejected": 0.984375,
+      "logps/chosen": -79.0,
+      "logps/rejected": -153.0,
+      "loss": 0.7391,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": -0.7890625,
+      "rewards/margins": 0.73828125,
+      "rewards/rejected": -1.53125,
+      "step": 1650
+    },
+    {
+      "epoch": 1.7373103087388801,
+      "grad_norm": 2.1529924557696596,
+      "learning_rate": 2.2445372742189334e-05,
+      "logits/chosen": 0.546875,
+      "logits/rejected": 2.328125,
+      "logps/chosen": -109.0,
+      "logps/rejected": -176.0,
+      "loss": 0.7035,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -1.0859375,
+      "rewards/margins": 0.66796875,
+      "rewards/rejected": -1.7578125,
+      "step": 1660
+    },
+    {
+      "epoch": 1.7477760334903192,
+      "grad_norm": 2.223074139550945,
+      "learning_rate": 2.214251060741528e-05,
+      "logits/chosen": 0.126953125,
+      "logits/rejected": 1.828125,
+      "logps/chosen": -116.0,
+      "logps/rejected": -169.0,
+      "loss": 0.7296,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -1.1640625,
+      "rewards/margins": 0.53125,
+      "rewards/rejected": -1.6953125,
+      "step": 1670
+    },
+    {
+      "epoch": 1.7582417582417582,
+      "grad_norm": 2.4048157905060794,
+      "learning_rate": 2.1840072811859302e-05,
+      "logits/chosen": -0.166015625,
+      "logits/rejected": 1.6328125,
+      "logps/chosen": -84.0,
+      "logps/rejected": -149.0,
+      "loss": 0.7293,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -0.83984375,
+      "rewards/margins": 0.64453125,
+      "rewards/rejected": -1.484375,
+      "step": 1680
+    },
+    {
+      "epoch": 1.7687074829931972,
+      "grad_norm": 2.18915947749729,
+      "learning_rate": 2.1538104267750287e-05,
+      "logits/chosen": -1.3046875,
+      "logits/rejected": 1.5859375,
+      "logps/chosen": -80.0,
+      "logps/rejected": -148.0,
+      "loss": 0.7019,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.80078125,
+      "rewards/margins": 0.6796875,
+      "rewards/rejected": -1.4765625,
+      "step": 1690
+    },
+    {
+      "epoch": 1.7791732077446363,
+      "grad_norm": 2.6990344623590556,
+      "learning_rate": 2.1236649817632952e-05,
+      "logits/chosen": -0.1787109375,
+      "logits/rejected": 1.6171875,
+      "logps/chosen": -93.5,
+      "logps/rejected": -151.0,
+      "loss": 0.7162,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -0.9375,
+      "rewards/margins": 0.57421875,
+      "rewards/rejected": -1.5078125,
+      "step": 1700
+    },
+    {
+      "epoch": 1.7791732077446363,
+      "eval_logits/chosen": 0.69921875,
+      "eval_logits/rejected": 2.46875,
+      "eval_logps/chosen": -103.0,
+      "eval_logps/rejected": -154.0,
+      "eval_loss": 0.8105937242507935,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.6953125,
+      "eval_rewards/chosen": -1.03125,
+      "eval_rewards/margins": 0.5078125,
+      "eval_rewards/rejected": -1.5390625,
+      "eval_runtime": 23.4045,
+      "eval_samples_per_second": 85.454,
+      "eval_steps_per_second": 1.367,
+      "step": 1700
+    },
+    {
+      "epoch": 1.7896389324960753,
+      "grad_norm": 3.0368064172844504,
+      "learning_rate": 2.093575422770872e-05,
+      "logits/chosen": 0.2490234375,
+      "logits/rejected": 2.421875,
+      "logps/chosen": -116.0,
+      "logps/rejected": -180.0,
+      "loss": 0.7511,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -1.1640625,
+      "rewards/margins": 0.63671875,
+      "rewards/rejected": -1.796875,
+      "step": 1710
+    },
+    {
+      "epoch": 1.8001046572475143,
+      "grad_norm": 1.9422844390661709,
+      "learning_rate": 2.0635462181187827e-05,
+      "logits/chosen": 1.109375,
+      "logits/rejected": 3.21875,
+      "logps/chosen": -109.0,
+      "logps/rejected": -184.0,
+      "loss": 0.7095,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -1.0859375,
+      "rewards/margins": 0.74609375,
+      "rewards/rejected": -1.8359375,
+      "step": 1720
+    },
+    {
+      "epoch": 1.8105703819989536,
+      "grad_norm": 2.4153235100603636,
+      "learning_rate": 2.0335818271653917e-05,
+      "logits/chosen": -0.1904296875,
+      "logits/rejected": 2.28125,
+      "logps/chosen": -79.0,
+      "logps/rejected": -154.0,
+      "loss": 0.7108,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -0.79296875,
+      "rewards/margins": 0.74609375,
+      "rewards/rejected": -1.5390625,
+      "step": 1730
+    },
+    {
+      "epoch": 1.8210361067503924,
+      "grad_norm": 2.2812788132162787,
+      "learning_rate": 2.0036866996441815e-05,
+      "logits/chosen": -0.81640625,
+      "logits/rejected": 1.71875,
+      "logps/chosen": -80.5,
+      "logps/rejected": -149.0,
+      "loss": 0.7053,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.793749988079071,
+      "rewards/chosen": -0.8046875,
+      "rewards/margins": 0.6875,
+      "rewards/rejected": -1.4921875,
+      "step": 1740
+    },
+    {
+      "epoch": 1.8315018315018317,
+      "grad_norm": 2.3835868218816687,
+      "learning_rate": 1.9738652750029677e-05,
+      "logits/chosen": 0.1923828125,
+      "logits/rejected": 2.125,
+      "logps/chosen": -107.0,
+      "logps/rejected": -176.0,
+      "loss": 0.7053,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -1.0703125,
+      "rewards/margins": 0.6875,
+      "rewards/rejected": -1.7578125,
+      "step": 1750
+    },
+    {
+      "epoch": 1.8419675562532705,
+      "grad_norm": 2.7486511515703422,
+      "learning_rate": 1.944121981744631e-05,
+      "logits/chosen": -0.2373046875,
+      "logits/rejected": 1.8203125,
+      "logps/chosen": -103.0,
+      "logps/rejected": -182.0,
+      "loss": 0.6854,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": -1.03125,
+      "rewards/margins": 0.78515625,
+      "rewards/rejected": -1.8203125,
+      "step": 1760
+    },
+    {
+      "epoch": 1.8524332810047097,
+      "grad_norm": 3.560220079391753,
+      "learning_rate": 1.9144612367694887e-05,
+      "logits/chosen": -0.69140625,
+      "logits/rejected": 1.7265625,
+      "logps/chosen": -101.0,
+      "logps/rejected": -184.0,
+      "loss": 0.6826,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.824999988079071,
+      "rewards/chosen": -1.0078125,
+      "rewards/margins": 0.828125,
+      "rewards/rejected": -1.8359375,
+      "step": 1770
+    },
+    {
+      "epoch": 1.8628990057561485,
+      "grad_norm": 3.2052182752494316,
+      "learning_rate": 1.8848874447193803e-05,
+      "logits/chosen": -1.6875,
+      "logits/rejected": 1.0234375,
+      "logps/chosen": -87.0,
+      "logps/rejected": -176.0,
+      "loss": 0.727,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8062499761581421,
+      "rewards/chosen": -0.875,
+      "rewards/margins": 0.8828125,
+      "rewards/rejected": -1.7578125,
+      "step": 1780
+    },
+    {
+      "epoch": 1.8733647305075878,
+      "grad_norm": 3.1880450555173616,
+      "learning_rate": 1.855404997323571e-05,
+      "logits/chosen": -0.455078125,
+      "logits/rejected": 1.7578125,
+      "logps/chosen": -97.5,
+      "logps/rejected": -166.0,
+      "loss": 0.7143,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -0.97265625,
+      "rewards/margins": 0.69140625,
+      "rewards/rejected": -1.6640625,
+      "step": 1790
+    },
+    {
+      "epoch": 1.8838304552590266,
+      "grad_norm": 3.161546697360618,
+      "learning_rate": 1.82601827274658e-05,
+      "logits/chosen": -0.03466796875,
+      "logits/rejected": 2.296875,
+      "logps/chosen": -108.0,
+      "logps/rejected": -174.0,
+      "loss": 0.714,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.768750011920929,
+      "rewards/chosen": -1.078125,
+      "rewards/margins": 0.66015625,
+      "rewards/rejected": -1.7421875,
+      "step": 1800
+    },
+    {
+      "epoch": 1.8838304552590266,
+      "eval_logits/chosen": 0.057861328125,
+      "eval_logits/rejected": 2.1875,
+      "eval_logps/chosen": -109.5,
+      "eval_logps/rejected": -162.0,
+      "eval_loss": 0.8183125257492065,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.685546875,
+      "eval_rewards/chosen": -1.09375,
+      "eval_rewards/margins": 0.53125,
+      "eval_rewards/rejected": -1.625,
+      "eval_runtime": 23.4091,
+      "eval_samples_per_second": 85.437,
+      "eval_steps_per_second": 1.367,
+      "step": 1800
+    },
+    {
+      "epoch": 1.8942961800104658,
+      "grad_norm": 4.113351074407271,
+      "learning_rate": 1.7967316349380213e-05,
+      "logits/chosen": -0.578125,
+      "logits/rejected": 2.0,
+      "logps/chosen": -99.0,
+      "logps/rejected": -180.0,
+      "loss": 0.7184,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.78125,
+      "rewards/chosen": -0.98828125,
+      "rewards/margins": 0.8046875,
+      "rewards/rejected": -1.796875,
+      "step": 1810
+    },
+    {
+      "epoch": 1.9047619047619047,
+      "grad_norm": 3.1281188822747485,
+      "learning_rate": 1.7675494329845516e-05,
+      "logits/chosen": 0.2578125,
+      "logits/rejected": 1.9765625,
+      "logps/chosen": -118.0,
+      "logps/rejected": -183.0,
+      "loss": 0.7163,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": -1.1796875,
+      "rewards/margins": 0.64453125,
+      "rewards/rejected": -1.828125,
+      "step": 1820
+    },
+    {
+      "epoch": 1.915227629513344,
+      "grad_norm": 2.771462928708798,
+      "learning_rate": 1.738476000464026e-05,
+      "logits/chosen": 0.369140625,
+      "logits/rejected": 1.828125,
+      "logps/chosen": -107.0,
+      "logps/rejected": -172.0,
+      "loss": 0.7115,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.768750011920929,
+      "rewards/chosen": -1.0703125,
+      "rewards/margins": 0.6484375,
+      "rewards/rejected": -1.71875,
+      "step": 1830
+    },
+    {
+      "epoch": 1.9256933542647827,
+      "grad_norm": 3.9799569391520007,
+      "learning_rate": 1.7095156548019646e-05,
+      "logits/chosen": -0.4296875,
+      "logits/rejected": 2.328125,
+      "logps/chosen": -104.5,
+      "logps/rejected": -182.0,
+      "loss": 0.71,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": -1.046875,
+      "rewards/margins": 0.7734375,
+      "rewards/rejected": -1.8203125,
+      "step": 1840
+    },
+    {
+      "epoch": 1.936159079016222,
+      "grad_norm": 3.6177372141385398,
+      "learning_rate": 1.6806726966304064e-05,
+      "logits/chosen": -0.08447265625,
+      "logits/rejected": 1.875,
+      "logps/chosen": -107.5,
+      "logps/rejected": -164.0,
+      "loss": 0.742,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -1.078125,
+      "rewards/margins": 0.56640625,
+      "rewards/rejected": -1.640625,
+      "step": 1850
+    },
+    {
+      "epoch": 1.9466248037676608,
+      "grad_norm": 3.1278751826458118,
+      "learning_rate": 1.6519514091492625e-05,
+      "logits/chosen": 0.1484375,
+      "logits/rejected": 2.234375,
+      "logps/chosen": -93.0,
+      "logps/rejected": -159.0,
+      "loss": 0.6965,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -0.9296875,
+      "rewards/margins": 0.6640625,
+      "rewards/rejected": -1.59375,
+      "step": 1860
+    },
+    {
+      "epoch": 1.9570905285191,
+      "grad_norm": 2.6490275232274048,
+      "learning_rate": 1.6233560574902627e-05,
+      "logits/chosen": -0.470703125,
+      "logits/rejected": 1.96875,
+      "logps/chosen": -83.0,
+      "logps/rejected": -158.0,
+      "loss": 0.6911,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.78125,
+      "rewards/chosen": -0.828125,
+      "rewards/margins": 0.75,
+      "rewards/rejected": -1.578125,
+      "step": 1870
+    },
+    {
+      "epoch": 1.9675562532705388,
+      "grad_norm": 4.8589347642001295,
+      "learning_rate": 1.5948908880835753e-05,
+      "logits/chosen": -0.39453125,
+      "logits/rejected": 1.8046875,
+      "logps/chosen": -99.0,
+      "logps/rejected": -172.0,
+      "loss": 0.6885,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.98828125,
+      "rewards/margins": 0.7265625,
+      "rewards/rejected": -1.71875,
+      "step": 1880
+    },
+    {
+      "epoch": 1.978021978021978,
+      "grad_norm": 2.3361446969502038,
+      "learning_rate": 1.5665601280272124e-05,
+      "logits/chosen": 0.146484375,
+      "logits/rejected": 2.125,
+      "logps/chosen": -103.5,
+      "logps/rejected": -174.0,
+      "loss": 0.6847,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -1.03125,
+      "rewards/margins": 0.703125,
+      "rewards/rejected": -1.734375,
+      "step": 1890
+    },
+    {
+      "epoch": 1.988487702773417,
+      "grad_norm": 3.658785181794147,
+      "learning_rate": 1.538367984459298e-05,
+      "logits/chosen": -0.28125,
+      "logits/rejected": 1.765625,
+      "logps/chosen": -93.0,
+      "logps/rejected": -174.0,
+      "loss": 0.7068,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.768750011920929,
+      "rewards/chosen": -0.93359375,
+      "rewards/margins": 0.80859375,
+      "rewards/rejected": -1.7421875,
+      "step": 1900
+    },
+    {
+      "epoch": 1.988487702773417,
+      "eval_logits/chosen": 0.318359375,
+      "eval_logits/rejected": 1.9921875,
+      "eval_logps/chosen": -97.5,
+      "eval_logps/rejected": -151.0,
+      "eval_loss": 0.8164374828338623,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.703125,
+      "eval_rewards/chosen": -0.97265625,
+      "eval_rewards/margins": 0.53515625,
+      "eval_rewards/rejected": -1.5078125,
+      "eval_runtime": 23.4107,
+      "eval_samples_per_second": 85.431,
+      "eval_steps_per_second": 1.367,
+      "step": 1900
+    },
+    {
+      "epoch": 1.9989534275248562,
+      "grad_norm": 3.302219387616319,
+      "learning_rate": 1.5103186439333133e-05,
+      "logits/chosen": 0.57421875,
+      "logits/rejected": 2.375,
+      "logps/chosen": -95.5,
+      "logps/rejected": -171.0,
+      "loss": 0.7087,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.953125,
+      "rewards/margins": 0.76171875,
+      "rewards/rejected": -1.71875,
+      "step": 1910
+    },
+    {
+      "epoch": 2.009419152276295,
+      "grad_norm": 2.130353238484687,
+      "learning_rate": 1.4824162717963829e-05,
+      "logits/chosen": -0.73046875,
+      "logits/rejected": 2.75,
+      "logps/chosen": -82.5,
+      "logps/rejected": -208.0,
+      "loss": 0.5084,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": -0.82421875,
+      "rewards/margins": 1.2578125,
+      "rewards/rejected": -2.078125,
+      "step": 1920
+    },
+    {
+      "epoch": 2.0198848770277342,
+      "grad_norm": 1.7826770722966134,
+      "learning_rate": 1.4546650115707192e-05,
+      "logits/chosen": 0.30078125,
+      "logits/rejected": 2.859375,
+      "logps/chosen": -70.0,
+      "logps/rejected": -193.0,
+      "loss": 0.5153,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -0.703125,
+      "rewards/margins": 1.234375,
+      "rewards/rejected": -1.9296875,
+      "step": 1930
+    },
+    {
+      "epoch": 2.030350601779173,
+      "grad_norm": 3.18285353991278,
+      "learning_rate": 1.4270689843383111e-05,
+      "logits/chosen": -0.2431640625,
+      "logits/rejected": 2.5,
+      "logps/chosen": -84.5,
+      "logps/rejected": -196.0,
+      "loss": 0.5181,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": -0.84375,
+      "rewards/margins": 1.1171875,
+      "rewards/rejected": -1.96875,
+      "step": 1940
+    },
+    {
+      "epoch": 2.0408163265306123,
+      "grad_norm": 1.9889361277243813,
+      "learning_rate": 1.3996322881289347e-05,
+      "logits/chosen": -1.1640625,
+      "logits/rejected": 2.015625,
+      "logps/chosen": -74.0,
+      "logps/rejected": -195.0,
+      "loss": 0.4871,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.856249988079071,
+      "rewards/chosen": -0.73828125,
+      "rewards/margins": 1.2109375,
+      "rewards/rejected": -1.953125,
+      "step": 1950
+    },
+    {
+      "epoch": 2.051282051282051,
+      "grad_norm": 2.4943767615884735,
+      "learning_rate": 1.3723589973115963e-05,
+      "logits/chosen": -1.15625,
+      "logits/rejected": 1.3828125,
+      "logps/chosen": -80.5,
+      "logps/rejected": -184.0,
+      "loss": 0.5063,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -0.8046875,
+      "rewards/margins": 1.0390625,
+      "rewards/rejected": -1.84375,
+      "step": 1960
+    },
+    {
+      "epoch": 2.0617477760334904,
+      "grad_norm": 3.137999818061474,
+      "learning_rate": 1.3452531619894837e-05,
+      "logits/chosen": -1.8046875,
+      "logits/rejected": 1.859375,
+      "logps/chosen": -73.0,
+      "logps/rejected": -198.0,
+      "loss": 0.4902,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -0.7265625,
+      "rewards/margins": 1.25,
+      "rewards/rejected": -1.9765625,
+      "step": 1970
+    },
+    {
+      "epoch": 2.072213500784929,
+      "grad_norm": 2.3802746436914126,
+      "learning_rate": 1.3183188073985245e-05,
+      "logits/chosen": -1.5,
+      "logits/rejected": 1.7578125,
+      "logps/chosen": -75.5,
+      "logps/rejected": -203.0,
+      "loss": 0.4901,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.75390625,
+      "rewards/margins": 1.2734375,
+      "rewards/rejected": -2.03125,
+      "step": 1980
+    },
+    {
+      "epoch": 2.0826792255363684,
+      "grad_norm": 3.0451052653098576,
+      "learning_rate": 1.2915599333096351e-05,
+      "logits/chosen": -1.5390625,
+      "logits/rejected": 1.6796875,
+      "logps/chosen": -77.0,
+      "logps/rejected": -193.0,
+      "loss": 0.4925,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": -0.76953125,
+      "rewards/margins": 1.1640625,
+      "rewards/rejected": -1.9296875,
+      "step": 1990
+    },
+    {
+      "epoch": 2.0931449502878072,
+      "grad_norm": 3.0594279060414213,
+      "learning_rate": 1.264980513434752e-05,
+      "logits/chosen": -1.5625,
+      "logits/rejected": 2.234375,
+      "logps/chosen": -75.0,
+      "logps/rejected": -210.0,
+      "loss": 0.4781,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -0.75,
+      "rewards/margins": 1.3515625,
+      "rewards/rejected": -2.09375,
+      "step": 2000
+    },
+    {
+      "epoch": 2.0931449502878072,
+      "eval_logits/chosen": 0.09765625,
+      "eval_logits/rejected": 1.734375,
+      "eval_logps/chosen": -119.0,
+      "eval_logps/rejected": -171.0,
+      "eval_loss": 0.8474531173706055,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.6796875,
+      "eval_rewards/chosen": -1.1875,
+      "eval_rewards/margins": 0.52734375,
+      "eval_rewards/rejected": -1.7109375,
+      "eval_runtime": 23.4111,
+      "eval_samples_per_second": 85.429,
+      "eval_steps_per_second": 1.367,
+      "step": 2000
+    },
+    {
+      "epoch": 2.1036106750392465,
+      "grad_norm": 2.401237514824198,
+      "learning_rate": 1.2385844948367324e-05,
+      "logits/chosen": -1.7265625,
+      "logits/rejected": 1.796875,
+      "logps/chosen": -85.5,
+      "logps/rejected": -216.0,
+      "loss": 0.5033,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -0.85546875,
+      "rewards/margins": 1.3046875,
+      "rewards/rejected": -2.15625,
+      "step": 2010
+    },
+    {
+      "epoch": 2.1140763997906853,
+      "grad_norm": 2.71676429164142,
+      "learning_rate": 1.2123757973432114e-05,
+      "logits/chosen": -1.9296875,
+      "logits/rejected": 1.8984375,
+      "logps/chosen": -73.0,
+      "logps/rejected": -205.0,
+      "loss": 0.4793,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8812500238418579,
+      "rewards/chosen": -0.73046875,
+      "rewards/margins": 1.3125,
+      "rewards/rejected": -2.046875,
+      "step": 2020
+    },
+    {
+      "epoch": 2.1245421245421245,
+      "grad_norm": 3.238106035905936,
+      "learning_rate": 1.1863583129645106e-05,
+      "logits/chosen": -1.9921875,
+      "logits/rejected": 1.828125,
+      "logps/chosen": -63.0,
+      "logps/rejected": -194.0,
+      "loss": 0.4788,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -0.62890625,
+      "rewards/margins": 1.3046875,
+      "rewards/rejected": -1.9375,
+      "step": 2030
+    },
+    {
+      "epoch": 2.1350078492935634,
+      "grad_norm": 3.5456261452365836,
+      "learning_rate": 1.1605359053156607e-05,
+      "logits/chosen": -1.21875,
+      "logits/rejected": 1.6953125,
+      "logps/chosen": -79.5,
+      "logps/rejected": -204.0,
+      "loss": 0.4981,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.831250011920929,
+      "rewards/chosen": -0.79296875,
+      "rewards/margins": 1.25,
+      "rewards/rejected": -2.046875,
+      "step": 2040
+    },
+    {
+      "epoch": 2.1454735740450026,
+      "grad_norm": 3.0249066526697885,
+      "learning_rate": 1.1349124090426644e-05,
+      "logits/chosen": -1.796875,
+      "logits/rejected": 1.921875,
+      "logps/chosen": -77.5,
+      "logps/rejected": -211.0,
+      "loss": 0.4918,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": -0.77734375,
+      "rewards/margins": 1.3359375,
+      "rewards/rejected": -2.109375,
+      "step": 2050
+    },
+    {
+      "epoch": 2.155939298796442,
+      "grad_norm": 2.6830837759262858,
+      "learning_rate": 1.1094916292530406e-05,
+      "logits/chosen": -2.328125,
+      "logits/rejected": 1.453125,
+      "logps/chosen": -76.5,
+      "logps/rejected": -200.0,
+      "loss": 0.4875,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8374999761581421,
+      "rewards/chosen": -0.765625,
+      "rewards/margins": 1.234375,
+      "rewards/rejected": -2.0,
+      "step": 2060
+    },
+    {
+      "epoch": 2.1664050235478807,
+      "grad_norm": 3.101489778429154,
+      "learning_rate": 1.0842773409507623e-05,
+      "logits/chosen": -1.8125,
+      "logits/rejected": 1.3203125,
+      "logps/chosen": -80.5,
+      "logps/rejected": -199.0,
+      "loss": 0.4921,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8187500238418579,
+      "rewards/chosen": -0.80859375,
+      "rewards/margins": 1.1796875,
+      "rewards/rejected": -1.984375,
+      "step": 2070
+    },
+    {
+      "epoch": 2.17687074829932,
+      "grad_norm": 4.0168805313832125,
+      "learning_rate": 1.0592732884756754e-05,
+      "logits/chosen": -2.328125,
+      "logits/rejected": 1.6328125,
+      "logps/chosen": -72.5,
+      "logps/rejected": -204.0,
+      "loss": 0.496,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8374999761581421,
+      "rewards/chosen": -0.7265625,
+      "rewards/margins": 1.3125,
+      "rewards/rejected": -2.046875,
+      "step": 2080
+    },
+    {
+      "epoch": 2.1873364730507587,
+      "grad_norm": 4.633252180091968,
+      "learning_rate": 1.0344831849474507e-05,
+      "logits/chosen": -1.953125,
+      "logits/rejected": 1.8515625,
+      "logps/chosen": -89.0,
+      "logps/rejected": -217.0,
+      "loss": 0.4791,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.890625,
+      "rewards/margins": 1.28125,
+      "rewards/rejected": -2.171875,
+      "step": 2090
+    },
+    {
+      "epoch": 2.197802197802198,
+      "grad_norm": 3.059512612131442,
+      "learning_rate": 1.0099107117141879e-05,
+      "logits/chosen": -2.65625,
+      "logits/rejected": 1.796875,
+      "logps/chosen": -55.5,
+      "logps/rejected": -206.0,
+      "loss": 0.4964,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.5546875,
+      "rewards/margins": 1.5078125,
+      "rewards/rejected": -2.0625,
+      "step": 2100
+    },
+    {
+      "epoch": 2.197802197802198,
+      "eval_logits/chosen": -0.92578125,
+      "eval_logits/rejected": 0.921875,
+      "eval_logps/chosen": -100.0,
+      "eval_logps/rejected": -155.0,
+      "eval_loss": 0.8455312252044678,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.6875,
+      "eval_rewards/chosen": -1.0,
+      "eval_rewards/margins": 0.5546875,
+      "eval_rewards/rejected": -1.5546875,
+      "eval_runtime": 23.4277,
+      "eval_samples_per_second": 85.369,
+      "eval_steps_per_second": 1.366,
+      "step": 2100
+    },
+    {
+      "epoch": 2.208267922553637,
+      "grad_norm": 3.326739733373825,
+      "learning_rate": 9.855595178057334e-06,
+      "logits/chosen": -2.359375,
+      "logits/rejected": 1.390625,
+      "logps/chosen": -73.0,
+      "logps/rejected": -214.0,
+      "loss": 0.4796,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8687499761581421,
+      "rewards/chosen": -0.7265625,
+      "rewards/margins": 1.4140625,
+      "rewards/rejected": -2.140625,
+      "step": 2110
+    },
+    {
+      "epoch": 2.218733647305076,
+      "grad_norm": 2.8887675235907446,
+      "learning_rate": 9.614332193917949e-06,
+      "logits/chosen": -1.59375,
+      "logits/rejected": 1.9375,
+      "logps/chosen": -84.5,
+      "logps/rejected": -226.0,
+      "loss": 0.4742,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -0.84375,
+      "rewards/margins": 1.4140625,
+      "rewards/rejected": -2.265625,
+      "step": 2120
+    },
+    {
+      "epoch": 2.229199372056515,
+      "grad_norm": 2.256391661151749,
+      "learning_rate": 9.375353992449385e-06,
+      "logits/chosen": -1.625,
+      "logits/rejected": 1.5390625,
+      "logps/chosen": -78.0,
+      "logps/rejected": -202.0,
+      "loss": 0.4807,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": -0.78125,
+      "rewards/margins": 1.234375,
+      "rewards/rejected": -2.015625,
+      "step": 2130
+    },
+    {
+      "epoch": 2.239665096807954,
+      "grad_norm": 2.696588408323547,
+      "learning_rate": 9.138696062085442e-06,
+      "logits/chosen": -2.0,
+      "logits/rejected": 1.8359375,
+      "logps/chosen": -80.0,
+      "logps/rejected": -211.0,
+      "loss": 0.4481,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": -0.80078125,
+      "rewards/margins": 1.3046875,
+      "rewards/rejected": -2.109375,
+      "step": 2140
+    },
+    {
+      "epoch": 2.250130821559393,
+      "grad_norm": 3.100895560312716,
+      "learning_rate": 8.904393546698007e-06,
+      "logits/chosen": -3.171875,
+      "logits/rejected": 1.6484375,
+      "logps/chosen": -62.0,
+      "logps/rejected": -217.0,
+      "loss": 0.468,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -0.62109375,
+      "rewards/margins": 1.546875,
+      "rewards/rejected": -2.171875,
+      "step": 2150
+    },
+    {
+      "epoch": 2.260596546310832,
+      "grad_norm": 4.385606909877023,
+      "learning_rate": 8.672481240378142e-06,
+      "logits/chosen": -2.171875,
+      "logits/rejected": 1.203125,
+      "logps/chosen": -68.0,
+      "logps/rejected": -199.0,
+      "loss": 0.5065,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": -0.6796875,
+      "rewards/margins": 1.3125,
+      "rewards/rejected": -1.984375,
+      "step": 2160
+    },
+    {
+      "epoch": 2.271062271062271,
+      "grad_norm": 3.2878895921872195,
+      "learning_rate": 8.44299358226919e-06,
+      "logits/chosen": -3.0625,
+      "logits/rejected": 1.7265625,
+      "logps/chosen": -66.0,
+      "logps/rejected": -206.0,
+      "loss": 0.4653,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -0.66015625,
+      "rewards/margins": 1.3984375,
+      "rewards/rejected": -2.0625,
+      "step": 2170
+    },
+    {
+      "epoch": 2.2815279958137102,
+      "grad_norm": 2.0325225334143355,
+      "learning_rate": 8.215964651452457e-06,
+      "logits/chosen": -2.28125,
+      "logits/rejected": 1.5625,
+      "logps/chosen": -74.5,
+      "logps/rejected": -208.0,
+      "loss": 0.4399,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.7421875,
+      "rewards/margins": 1.3359375,
+      "rewards/rejected": -2.078125,
+      "step": 2180
+    },
+    {
+      "epoch": 2.291993720565149,
+      "grad_norm": 3.2889604576867857,
+      "learning_rate": 7.991428161886502e-06,
+      "logits/chosen": -1.640625,
+      "logits/rejected": 1.8125,
+      "logps/chosen": -74.0,
+      "logps/rejected": -214.0,
+      "loss": 0.4477,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -0.73828125,
+      "rewards/margins": 1.40625,
+      "rewards/rejected": -2.140625,
+      "step": 2190
+    },
+    {
+      "epoch": 2.3024594453165883,
+      "grad_norm": 3.1031160416752943,
+      "learning_rate": 7.76941745740061e-06,
+      "logits/chosen": -1.671875,
+      "logits/rejected": 1.8203125,
+      "logps/chosen": -69.5,
+      "logps/rejected": -198.0,
+      "loss": 0.4723,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": -0.6953125,
+      "rewards/margins": 1.28125,
+      "rewards/rejected": -1.9765625,
+      "step": 2200
+    },
+    {
+      "epoch": 2.3024594453165883,
+      "eval_logits/chosen": -0.46484375,
+      "eval_logits/rejected": 1.296875,
+      "eval_logps/chosen": -110.0,
+      "eval_logps/rejected": -166.0,
+      "eval_loss": 0.8474687337875366,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.693359375,
+      "eval_rewards/chosen": -1.1015625,
+      "eval_rewards/margins": 0.55859375,
+      "eval_rewards/rejected": -1.65625,
+      "eval_runtime": 23.4239,
+      "eval_samples_per_second": 85.383,
+      "eval_steps_per_second": 1.366,
+      "step": 2200
+    },
+    {
+      "epoch": 2.312925170068027,
+      "grad_norm": 3.6445622578264834,
+      "learning_rate": 7.549965506743123e-06,
+      "logits/chosen": -1.7109375,
+      "logits/rejected": 2.390625,
+      "logps/chosen": -77.5,
+      "logps/rejected": -219.0,
+      "loss": 0.4724,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -0.77734375,
+      "rewards/margins": 1.40625,
+      "rewards/rejected": -2.1875,
+      "step": 2210
+    },
+    {
+      "epoch": 2.3233908948194664,
+      "grad_norm": 3.0645568250892894,
+      "learning_rate": 7.333104898685631e-06,
+      "logits/chosen": -1.7578125,
+      "logits/rejected": 1.921875,
+      "logps/chosen": -86.0,
+      "logps/rejected": -221.0,
+      "loss": 0.4512,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -0.859375,
+      "rewards/margins": 1.3515625,
+      "rewards/rejected": -2.203125,
+      "step": 2220
+    },
+    {
+      "epoch": 2.333856619570905,
+      "grad_norm": 3.1713832682733183,
+      "learning_rate": 7.118867837183518e-06,
+      "logits/chosen": -1.9609375,
+      "logits/rejected": 1.640625,
+      "logps/chosen": -71.5,
+      "logps/rejected": -196.0,
+      "loss": 0.4579,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8812500238418579,
+      "rewards/chosen": -0.71484375,
+      "rewards/margins": 1.25,
+      "rewards/rejected": -1.9609375,
+      "step": 2230
+    },
+    {
+      "epoch": 2.3443223443223444,
+      "grad_norm": 3.3771840159946067,
+      "learning_rate": 6.9072861365936045e-06,
+      "logits/chosen": -2.015625,
+      "logits/rejected": 1.515625,
+      "logps/chosen": -63.5,
+      "logps/rejected": -203.0,
+      "loss": 0.4369,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -0.63671875,
+      "rewards/margins": 1.390625,
+      "rewards/rejected": -2.03125,
+      "step": 2240
+    },
+    {
+      "epoch": 2.3547880690737832,
+      "grad_norm": 4.41039033932836,
+      "learning_rate": 6.6983912169497005e-06,
+      "logits/chosen": -2.203125,
+      "logits/rejected": 1.9765625,
+      "logps/chosen": -71.5,
+      "logps/rejected": -215.0,
+      "loss": 0.4767,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -0.71484375,
+      "rewards/margins": 1.4375,
+      "rewards/rejected": -2.15625,
+      "step": 2250
+    },
+    {
+      "epoch": 2.3652537938252225,
+      "grad_norm": 4.595871268828088,
+      "learning_rate": 6.49221409929677e-06,
+      "logits/chosen": -2.375,
+      "logits/rejected": 1.4296875,
+      "logps/chosen": -71.5,
+      "logps/rejected": -206.0,
+      "loss": 0.4767,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -0.71484375,
+      "rewards/margins": 1.34375,
+      "rewards/rejected": -2.0625,
+      "step": 2260
+    },
+    {
+      "epoch": 2.3757195185766613,
+      "grad_norm": 4.251988223976189,
+      "learning_rate": 6.2887854010842075e-06,
+      "logits/chosen": -1.390625,
+      "logits/rejected": 1.7421875,
+      "logps/chosen": -81.0,
+      "logps/rejected": -202.0,
+      "loss": 0.4776,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": -0.80859375,
+      "rewards/margins": 1.203125,
+      "rewards/rejected": -2.015625,
+      "step": 2270
+    },
+    {
+      "epoch": 2.3861852433281006,
+      "grad_norm": 3.7229333385195673,
+      "learning_rate": 6.0881353316191385e-06,
+      "logits/chosen": -1.8515625,
+      "logits/rejected": 1.828125,
+      "logps/chosen": -72.0,
+      "logps/rejected": -219.0,
+      "loss": 0.4641,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -0.71875,
+      "rewards/margins": 1.4765625,
+      "rewards/rejected": -2.1875,
+      "step": 2280
+    },
+    {
+      "epoch": 2.3966509680795394,
+      "grad_norm": 3.684182542387793,
+      "learning_rate": 5.890293687580381e-06,
+      "logits/chosen": -2.015625,
+      "logits/rejected": 1.625,
+      "logps/chosen": -71.0,
+      "logps/rejected": -212.0,
+      "loss": 0.4563,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": -0.7109375,
+      "rewards/margins": 1.4140625,
+      "rewards/rejected": -2.125,
+      "step": 2290
+    },
+    {
+      "epoch": 2.4071166928309786,
+      "grad_norm": 3.898560923074975,
+      "learning_rate": 5.695289848593532e-06,
+      "logits/chosen": -1.734375,
+      "logits/rejected": 1.8046875,
+      "logps/chosen": -75.5,
+      "logps/rejected": -217.0,
+      "loss": 0.5051,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -0.7578125,
+      "rewards/margins": 1.421875,
+      "rewards/rejected": -2.171875,
+      "step": 2300
+    },
+    {
+      "epoch": 2.4071166928309786,
+      "eval_logits/chosen": -0.2890625,
+      "eval_logits/rejected": 1.4140625,
+      "eval_logps/chosen": -113.0,
+      "eval_logps/rejected": -170.0,
+      "eval_loss": 0.8480468988418579,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.689453125,
+      "eval_rewards/chosen": -1.1328125,
+      "eval_rewards/margins": 0.56640625,
+      "eval_rewards/rejected": -1.6953125,
+      "eval_runtime": 23.4278,
+      "eval_samples_per_second": 85.369,
+      "eval_steps_per_second": 1.366,
+      "step": 2300
+    },
+    {
+      "epoch": 2.4175824175824174,
+      "grad_norm": 5.708298067968057,
+      "learning_rate": 5.50315277286812e-06,
+      "logits/chosen": -1.734375,
+      "logits/rejected": 1.7578125,
+      "logps/chosen": -72.0,
+      "logps/rejected": -216.0,
+      "loss": 0.4621,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -0.71875,
+      "rewards/margins": 1.4375,
+      "rewards/rejected": -2.15625,
+      "step": 2310
+    },
+    {
+      "epoch": 2.4280481423338567,
+      "grad_norm": 3.192863307202569,
+      "learning_rate": 5.313910992897281e-06,
+      "logits/chosen": -2.390625,
+      "logits/rejected": 1.6953125,
+      "logps/chosen": -63.0,
+      "logps/rejected": -206.0,
+      "loss": 0.4644,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8812500238418579,
+      "rewards/chosen": -0.62890625,
+      "rewards/margins": 1.4296875,
+      "rewards/rejected": -2.0625,
+      "step": 2320
+    },
+    {
+      "epoch": 2.4385138670852955,
+      "grad_norm": 2.7638361615754046,
+      "learning_rate": 5.127592611220658e-06,
+      "logits/chosen": -1.75,
+      "logits/rejected": 1.9296875,
+      "logps/chosen": -80.5,
+      "logps/rejected": -223.0,
+      "loss": 0.4483,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -0.8046875,
+      "rewards/margins": 1.421875,
+      "rewards/rejected": -2.234375,
+      "step": 2330
+    },
+    {
+      "epoch": 2.4489795918367347,
+      "grad_norm": 3.9943358680994754,
+      "learning_rate": 4.944225296251159e-06,
+      "logits/chosen": -3.140625,
+      "logits/rejected": 1.765625,
+      "logps/chosen": -66.0,
+      "logps/rejected": -220.0,
+      "loss": 0.4594,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -0.66015625,
+      "rewards/margins": 1.546875,
+      "rewards/rejected": -2.203125,
+      "step": 2340
+    },
+    {
+      "epoch": 2.4594453165881736,
+      "grad_norm": 3.7086722433636035,
+      "learning_rate": 4.763836278166172e-06,
+      "logits/chosen": -1.96875,
+      "logits/rejected": 1.703125,
+      "logps/chosen": -70.0,
+      "logps/rejected": -209.0,
+      "loss": 0.4648,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8374999761581421,
+      "rewards/chosen": -0.69921875,
+      "rewards/margins": 1.390625,
+      "rewards/rejected": -2.09375,
+      "step": 2350
+    },
+    {
+      "epoch": 2.469911041339613,
+      "grad_norm": 5.855823508238343,
+      "learning_rate": 4.586452344863876e-06,
+      "logits/chosen": -2.015625,
+      "logits/rejected": 1.4921875,
+      "logps/chosen": -73.5,
+      "logps/rejected": -208.0,
+      "loss": 0.4752,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -0.734375,
+      "rewards/margins": 1.3515625,
+      "rewards/rejected": -2.078125,
+      "step": 2360
+    },
+    {
+      "epoch": 2.4803767660910516,
+      "grad_norm": 5.089010910913349,
+      "learning_rate": 4.412099837985192e-06,
+      "logits/chosen": -1.6953125,
+      "logits/rejected": 1.703125,
+      "logps/chosen": -81.0,
+      "logps/rejected": -197.0,
+      "loss": 0.5076,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8187500238418579,
+      "rewards/chosen": -0.8125,
+      "rewards/margins": 1.15625,
+      "rewards/rejected": -1.96875,
+      "step": 2370
+    },
+    {
+      "epoch": 2.490842490842491,
+      "grad_norm": 3.275630841408743,
+      "learning_rate": 4.240804649002089e-06,
+      "logits/chosen": -0.8984375,
+      "logits/rejected": 2.28125,
+      "logps/chosen": -99.0,
+      "logps/rejected": -230.0,
+      "loss": 0.465,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.831250011920929,
+      "rewards/chosen": -0.98828125,
+      "rewards/margins": 1.3125,
+      "rewards/rejected": -2.296875,
+      "step": 2380
+    },
+    {
+      "epoch": 2.50130821559393,
+      "grad_norm": 4.188033874742694,
+      "learning_rate": 4.072592215372598e-06,
+      "logits/chosen": -1.234375,
+      "logits/rejected": 2.125,
+      "logps/chosen": -81.5,
+      "logps/rejected": -212.0,
+      "loss": 0.4626,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -0.8125,
+      "rewards/margins": 1.3125,
+      "rewards/rejected": -2.125,
+      "step": 2390
+    },
+    {
+      "epoch": 2.511773940345369,
+      "grad_norm": 5.249386517111738,
+      "learning_rate": 3.907487516763389e-06,
+      "logits/chosen": -2.15625,
+      "logits/rejected": 1.828125,
+      "logps/chosen": -79.5,
+      "logps/rejected": -213.0,
+      "loss": 0.4647,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -0.79296875,
+      "rewards/margins": 1.3359375,
+      "rewards/rejected": -2.125,
+      "step": 2400
+    },
+    {
+      "epoch": 2.511773940345369,
+      "eval_logits/chosen": -0.349609375,
+      "eval_logits/rejected": 1.453125,
+      "eval_logps/chosen": -114.0,
+      "eval_logps/rejected": -171.0,
+      "eval_loss": 0.8462968468666077,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.67578125,
+      "eval_rewards/chosen": -1.140625,
+      "eval_rewards/margins": 0.57421875,
+      "eval_rewards/rejected": -1.71875,
+      "eval_runtime": 23.4238,
+      "eval_samples_per_second": 85.383,
+      "eval_steps_per_second": 1.366,
+      "step": 2400
+    },
+    {
+      "epoch": 2.5222396650968077,
+      "grad_norm": 3.3309939101437465,
+      "learning_rate": 3.745515071340286e-06,
+      "logits/chosen": -2.265625,
+      "logits/rejected": 1.6015625,
+      "logps/chosen": -70.5,
+      "logps/rejected": -209.0,
+      "loss": 0.4647,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8687499761581421,
+      "rewards/chosen": -0.703125,
+      "rewards/margins": 1.390625,
+      "rewards/rejected": -2.09375,
+      "step": 2410
+    },
+    {
+      "epoch": 2.532705389848247,
+      "grad_norm": 3.421569347958948,
+      "learning_rate": 3.586698932127236e-06,
+      "logits/chosen": -2.03125,
+      "logits/rejected": 1.53125,
+      "logps/chosen": -66.0,
+      "logps/rejected": -211.0,
+      "loss": 0.4382,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8812500238418579,
+      "rewards/chosen": -0.66015625,
+      "rewards/margins": 1.453125,
+      "rewards/rejected": -2.109375,
+      "step": 2420
+    },
+    {
+      "epoch": 2.5431711145996863,
+      "grad_norm": 5.858944826145137,
+      "learning_rate": 3.4310626834344744e-06,
+      "logits/chosen": -1.6328125,
+      "logits/rejected": 1.9375,
+      "logps/chosen": -78.0,
+      "logps/rejected": -217.0,
+      "loss": 0.446,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -0.78125,
+      "rewards/margins": 1.390625,
+      "rewards/rejected": -2.171875,
+      "step": 2430
+    },
+    {
+      "epoch": 2.553636839351125,
+      "grad_norm": 3.7504325223428676,
+      "learning_rate": 3.2786294373562343e-06,
+      "logits/chosen": -1.5703125,
+      "logits/rejected": 2.421875,
+      "logps/chosen": -81.0,
+      "logps/rejected": -230.0,
+      "loss": 0.477,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8812500238418579,
+      "rewards/chosen": -0.8125,
+      "rewards/margins": 1.484375,
+      "rewards/rejected": -2.296875,
+      "step": 2440
+    },
+    {
+      "epoch": 2.564102564102564,
+      "grad_norm": 4.137385907568805,
+      "learning_rate": 3.1294218303385526e-06,
+      "logits/chosen": -2.546875,
+      "logits/rejected": 1.5625,
+      "logps/chosen": -70.0,
+      "logps/rejected": -212.0,
+      "loss": 0.4565,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.69921875,
+      "rewards/margins": 1.421875,
+      "rewards/rejected": -2.125,
+      "step": 2450
+    },
+    {
+      "epoch": 2.574568288854003,
+      "grad_norm": 4.699662846024779,
+      "learning_rate": 2.9834620198177483e-06,
+      "logits/chosen": -2.1875,
+      "logits/rejected": 1.8671875,
+      "logps/chosen": -73.5,
+      "logps/rejected": -214.0,
+      "loss": 0.5031,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -0.734375,
+      "rewards/margins": 1.40625,
+      "rewards/rejected": -2.140625,
+      "step": 2460
+    },
+    {
+      "epoch": 2.5850340136054424,
+      "grad_norm": 3.5347565979956204,
+      "learning_rate": 2.840771680930068e-06,
+      "logits/chosen": -2.515625,
+      "logits/rejected": 1.4765625,
+      "logps/chosen": -69.0,
+      "logps/rejected": -202.0,
+      "loss": 0.4509,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -0.69140625,
+      "rewards/margins": 1.328125,
+      "rewards/rejected": -2.015625,
+      "step": 2470
+    },
+    {
+      "epoch": 2.595499738356881,
+      "grad_norm": 4.285426514434569,
+      "learning_rate": 2.701372003292868e-06,
+      "logits/chosen": -1.90625,
+      "logits/rejected": 1.9140625,
+      "logps/chosen": -71.5,
+      "logps/rejected": -236.0,
+      "loss": 0.4508,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -0.71875,
+      "rewards/margins": 1.640625,
+      "rewards/rejected": -2.359375,
+      "step": 2480
+    },
+    {
+      "epoch": 2.60596546310832,
+      "grad_norm": 3.317876577915879,
+      "learning_rate": 2.56528368785795e-06,
+      "logits/chosen": -2.125,
+      "logits/rejected": 1.6484375,
+      "logps/chosen": -74.5,
+      "logps/rejected": -218.0,
+      "loss": 0.4659,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -0.74609375,
+      "rewards/margins": 1.4375,
+      "rewards/rejected": -2.1875,
+      "step": 2490
+    },
+    {
+      "epoch": 2.6164311878597593,
+      "grad_norm": 3.7772109894796406,
+      "learning_rate": 2.432526943837471e-06,
+      "logits/chosen": -1.6796875,
+      "logits/rejected": 1.7265625,
+      "logps/chosen": -72.5,
+      "logps/rejected": -217.0,
+      "loss": 0.4442,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -0.7265625,
+      "rewards/margins": 1.4453125,
+      "rewards/rejected": -2.171875,
+      "step": 2500
+    },
+    {
+      "epoch": 2.6164311878597593,
+      "eval_logits/chosen": -0.1435546875,
+      "eval_logits/rejected": 1.5859375,
+      "eval_logps/chosen": -123.5,
+      "eval_logps/rejected": -180.0,
+      "eval_loss": 0.8526718616485596,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.66796875,
+      "eval_rewards/chosen": -1.234375,
+      "eval_rewards/margins": 0.56640625,
+      "eval_rewards/rejected": -1.796875,
+      "eval_runtime": 23.4218,
+      "eval_samples_per_second": 85.39,
+      "eval_steps_per_second": 1.366,
+      "step": 2500
+    },
+    {
+      "epoch": 2.6268969126111985,
+      "grad_norm": 2.460082678024024,
+      "learning_rate": 2.3031214857028446e-06,
+      "logits/chosen": -2.359375,
+      "logits/rejected": 1.7109375,
+      "logps/chosen": -73.0,
+      "logps/rejected": -221.0,
+      "loss": 0.4147,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -0.73046875,
+      "rewards/margins": 1.4765625,
+      "rewards/rejected": -2.203125,
+      "step": 2510
+    },
+    {
+      "epoch": 2.6373626373626373,
+      "grad_norm": 4.188360440224396,
+      "learning_rate": 2.177086530257139e-06,
+      "logits/chosen": -1.1640625,
+      "logits/rejected": 2.203125,
+      "logps/chosen": -90.5,
+      "logps/rejected": -217.0,
+      "loss": 0.4596,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": -0.90234375,
+      "rewards/margins": 1.2734375,
+      "rewards/rejected": -2.171875,
+      "step": 2520
+    },
+    {
+      "epoch": 2.647828362114076,
+      "grad_norm": 3.0512140629719924,
+      "learning_rate": 2.054440793781354e-06,
+      "logits/chosen": -2.21875,
+      "logits/rejected": 1.5,
+      "logps/chosen": -71.5,
+      "logps/rejected": -213.0,
+      "loss": 0.431,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -0.71484375,
+      "rewards/margins": 1.421875,
+      "rewards/rejected": -2.140625,
+      "step": 2530
+    },
+    {
+      "epoch": 2.6582940868655154,
+      "grad_norm": 3.2873501678312924,
+      "learning_rate": 1.9352024892550587e-06,
+      "logits/chosen": -2.515625,
+      "logits/rejected": 1.640625,
+      "logps/chosen": -70.5,
+      "logps/rejected": -210.0,
+      "loss": 0.4891,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8687499761581421,
+      "rewards/chosen": -0.703125,
+      "rewards/margins": 1.3984375,
+      "rewards/rejected": -2.109375,
+      "step": 2540
+    },
+    {
+      "epoch": 2.6687598116169546,
+      "grad_norm": 3.9061725539574117,
+      "learning_rate": 1.8193893236517379e-06,
+      "logits/chosen": -2.65625,
+      "logits/rejected": 1.6484375,
+      "logps/chosen": -61.0,
+      "logps/rejected": -204.0,
+      "loss": 0.5015,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -0.609375,
+      "rewards/margins": 1.4296875,
+      "rewards/rejected": -2.046875,
+      "step": 2550
+    },
+    {
+      "epoch": 2.6792255363683934,
+      "grad_norm": 3.4064600921431083,
+      "learning_rate": 1.7070184953092956e-06,
+      "logits/chosen": -1.9296875,
+      "logits/rejected": 1.65625,
+      "logps/chosen": -83.0,
+      "logps/rejected": -208.0,
+      "loss": 0.4636,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -0.828125,
+      "rewards/margins": 1.25,
+      "rewards/rejected": -2.078125,
+      "step": 2560
+    },
+    {
+      "epoch": 2.6896912611198327,
+      "grad_norm": 4.343156319640629,
+      "learning_rate": 1.5981066913760918e-06,
+      "logits/chosen": -2.359375,
+      "logits/rejected": 1.8125,
+      "logps/chosen": -79.0,
+      "logps/rejected": -213.0,
+      "loss": 0.4709,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": -0.79296875,
+      "rewards/margins": 1.34375,
+      "rewards/rejected": -2.125,
+      "step": 2570
+    },
+    {
+      "epoch": 2.7001569858712715,
+      "grad_norm": 3.4585561535902714,
+      "learning_rate": 1.492670085332884e-06,
+      "logits/chosen": -2.25,
+      "logits/rejected": 1.6171875,
+      "logps/chosen": -80.5,
+      "logps/rejected": -208.0,
+      "loss": 0.4774,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8374999761581421,
+      "rewards/chosen": -0.80859375,
+      "rewards/margins": 1.2734375,
+      "rewards/rejected": -2.078125,
+      "step": 2580
+    },
+    {
+      "epoch": 2.7106227106227108,
+      "grad_norm": 4.893145910811799,
+      "learning_rate": 1.3907243345910787e-06,
+      "logits/chosen": -1.609375,
+      "logits/rejected": 1.921875,
+      "logps/chosen": -74.0,
+      "logps/rejected": -209.0,
+      "loss": 0.4698,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.856249988079071,
+      "rewards/chosen": -0.7421875,
+      "rewards/margins": 1.34375,
+      "rewards/rejected": -2.09375,
+      "step": 2590
+    },
+    {
+      "epoch": 2.7210884353741496,
+      "grad_norm": 3.212145690604128,
+      "learning_rate": 1.292284578167538e-06,
+      "logits/chosen": -2.640625,
+      "logits/rejected": 1.484375,
+      "logps/chosen": -62.25,
+      "logps/rejected": -213.0,
+      "loss": 0.4349,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.62109375,
+      "rewards/margins": 1.515625,
+      "rewards/rejected": -2.140625,
+      "step": 2600
+    },
+    {
+      "epoch": 2.7210884353741496,
+      "eval_logits/chosen": -0.58984375,
+      "eval_logits/rejected": 1.2421875,
+      "eval_logps/chosen": -112.0,
+      "eval_logps/rejected": -169.0,
+      "eval_loss": 0.8504843711853027,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.669921875,
+      "eval_rewards/chosen": -1.1171875,
+      "eval_rewards/margins": 0.57421875,
+      "eval_rewards/rejected": -1.6953125,
+      "eval_runtime": 23.4198,
+      "eval_samples_per_second": 85.398,
+      "eval_steps_per_second": 1.366,
+      "step": 2600
+    },
+    {
+      "epoch": 2.731554160125589,
+      "grad_norm": 3.7511966729270103,
+      "learning_rate": 1.1973654344364925e-06,
+      "logits/chosen": -2.4375,
+      "logits/rejected": 1.53125,
+      "logps/chosen": -68.5,
+      "logps/rejected": -210.0,
+      "loss": 0.4727,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": -0.6875,
+      "rewards/margins": 1.4140625,
+      "rewards/rejected": -2.109375,
+      "step": 2610
+    },
+    {
+      "epoch": 2.7420198848770276,
+      "grad_norm": 3.9479535826618006,
+      "learning_rate": 1.1059809989586583e-06,
+      "logits/chosen": -2.078125,
+      "logits/rejected": 1.484375,
+      "logps/chosen": -75.0,
+      "logps/rejected": -204.0,
+      "loss": 0.4553,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8687499761581421,
+      "rewards/chosen": -0.75,
+      "rewards/margins": 1.2890625,
+      "rewards/rejected": -2.03125,
+      "step": 2620
+    },
+    {
+      "epoch": 2.752485609628467,
+      "grad_norm": 4.433485553923868,
+      "learning_rate": 1.018144842388044e-06,
+      "logits/chosen": -1.7421875,
+      "logits/rejected": 1.546875,
+      "logps/chosen": -70.5,
+      "logps/rejected": -212.0,
+      "loss": 0.4735,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.856249988079071,
+      "rewards/chosen": -0.70703125,
+      "rewards/margins": 1.40625,
+      "rewards/rejected": -2.109375,
+      "step": 2630
+    },
+    {
+      "epoch": 2.7629513343799057,
+      "grad_norm": 3.0553389852591133,
+      "learning_rate": 9.338700084567109e-07,
+      "logits/chosen": -2.09375,
+      "logits/rejected": 2.15625,
+      "logps/chosen": -71.0,
+      "logps/rejected": -217.0,
+      "loss": 0.451,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": -0.7109375,
+      "rewards/margins": 1.4609375,
+      "rewards/rejected": -2.171875,
+      "step": 2640
+    },
+    {
+      "epoch": 2.773417059131345,
+      "grad_norm": 3.1328701495396727,
+      "learning_rate": 8.531690120377606e-07,
+      "logits/chosen": -2.046875,
+      "logits/rejected": 1.9140625,
+      "logps/chosen": -71.0,
+      "logps/rejected": -221.0,
+      "loss": 0.4567,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.7109375,
+      "rewards/margins": 1.5,
+      "rewards/rejected": -2.21875,
+      "step": 2650
+    },
+    {
+      "epoch": 2.7838827838827838,
+      "grad_norm": 2.489093644035697,
+      "learning_rate": 7.760538372868636e-07,
+      "logits/chosen": -2.75,
+      "logits/rejected": 1.4609375,
+      "logps/chosen": -71.5,
+      "logps/rejected": -208.0,
+      "loss": 0.474,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.831250011920929,
+      "rewards/chosen": -0.71484375,
+      "rewards/margins": 1.359375,
+      "rewards/rejected": -2.078125,
+      "step": 2660
+    },
+    {
+      "epoch": 2.794348508634223,
+      "grad_norm": 2.965479753440738,
+      "learning_rate": 7.025359358626166e-07,
+      "logits/chosen": -2.1875,
+      "logits/rejected": 1.71875,
+      "logps/chosen": -68.5,
+      "logps/rejected": -213.0,
+      "loss": 0.4816,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -0.6875,
+      "rewards/margins": 1.4453125,
+      "rewards/rejected": -2.125,
+      "step": 2670
+    },
+    {
+      "epoch": 2.804814233385662,
+      "grad_norm": 6.07837736363748,
+      "learning_rate": 6.326262252259463e-07,
+      "logits/chosen": -2.671875,
+      "logits/rejected": 1.734375,
+      "logps/chosen": -65.0,
+      "logps/rejected": -218.0,
+      "loss": 0.4423,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -0.65234375,
+      "rewards/margins": 1.5234375,
+      "rewards/rejected": -2.171875,
+      "step": 2680
+    },
+    {
+      "epoch": 2.815279958137101,
+      "grad_norm": 6.008880524138567,
+      "learning_rate": 5.66335087018871e-07,
+      "logits/chosen": -1.875,
+      "logits/rejected": 1.8515625,
+      "logps/chosen": -65.5,
+      "logps/rejected": -225.0,
+      "loss": 0.4605,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -0.65234375,
+      "rewards/margins": 1.59375,
+      "rewards/rejected": -2.25,
+      "step": 2690
+    },
+    {
+      "epoch": 2.82574568288854,
+      "grad_norm": 3.2251290674827677,
+      "learning_rate": 5.036723655228226e-07,
+      "logits/chosen": -2.75,
+      "logits/rejected": 1.484375,
+      "logps/chosen": -65.0,
+      "logps/rejected": -206.0,
+      "loss": 0.4514,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -0.65234375,
+      "rewards/margins": 1.4140625,
+      "rewards/rejected": -2.0625,
+      "step": 2700
+    },
+    {
+      "epoch": 2.82574568288854,
+      "eval_logits/chosen": -0.640625,
+      "eval_logits/rejected": 1.1953125,
+      "eval_logps/chosen": -112.0,
+      "eval_logps/rejected": -169.0,
+      "eval_loss": 0.8493437767028809,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.673828125,
+      "eval_rewards/chosen": -1.1171875,
+      "eval_rewards/margins": 0.578125,
+      "eval_rewards/rejected": -1.6953125,
+      "eval_runtime": 23.4236,
+      "eval_samples_per_second": 85.384,
+      "eval_steps_per_second": 1.366,
+      "step": 2700
+    },
+    {
+      "epoch": 2.836211407639979,
+      "grad_norm": 3.0735126078941386,
+      "learning_rate": 4.446473661967432e-07,
+      "logits/chosen": -2.390625,
+      "logits/rejected": 1.5078125,
+      "logps/chosen": -65.5,
+      "logps/rejected": -212.0,
+      "loss": 0.4473,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": -0.65625,
+      "rewards/margins": 1.4609375,
+      "rewards/rejected": -2.109375,
+      "step": 2710
+    },
+    {
+      "epoch": 2.846677132391418,
+      "grad_norm": 3.8032852423370818,
+      "learning_rate": 3.892688542952411e-07,
+      "logits/chosen": -2.171875,
+      "logits/rejected": 1.5078125,
+      "logps/chosen": -74.5,
+      "logps/rejected": -208.0,
+      "loss": 0.4627,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -0.75,
+      "rewards/margins": 1.328125,
+      "rewards/rejected": -2.078125,
+      "step": 2720
+    },
+    {
+      "epoch": 2.857142857142857,
+      "grad_norm": 2.9637865196142266,
+      "learning_rate": 3.3754505356694e-07,
+      "logits/chosen": -1.7578125,
+      "logits/rejected": 1.5859375,
+      "logps/chosen": -82.5,
+      "logps/rejected": -211.0,
+      "loss": 0.4727,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -0.82421875,
+      "rewards/margins": 1.2890625,
+      "rewards/rejected": -2.109375,
+      "step": 2730
+    },
+    {
+      "epoch": 2.867608581894296,
+      "grad_norm": 5.576458024715762,
+      "learning_rate": 2.8948364503322276e-07,
+      "logits/chosen": -2.03125,
+      "logits/rejected": 1.5703125,
+      "logps/chosen": -85.5,
+      "logps/rejected": -214.0,
+      "loss": 0.47,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8812500238418579,
+      "rewards/chosen": -0.85546875,
+      "rewards/margins": 1.28125,
+      "rewards/rejected": -2.140625,
+      "step": 2740
+    },
+    {
+      "epoch": 2.8780743066457353,
+      "grad_norm": 3.596625159237956,
+      "learning_rate": 2.4509176584762706e-07,
+      "logits/chosen": -1.9453125,
+      "logits/rejected": 2.078125,
+      "logps/chosen": -78.5,
+      "logps/rejected": -209.0,
+      "loss": 0.4307,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.856249988079071,
+      "rewards/chosen": -0.7890625,
+      "rewards/margins": 1.3046875,
+      "rewards/rejected": -2.09375,
+      "step": 2750
+    },
+    {
+      "epoch": 2.8885400313971745,
+      "grad_norm": 4.088128401232327,
+      "learning_rate": 2.0437600823595694e-07,
+      "logits/chosen": -2.390625,
+      "logits/rejected": 1.4609375,
+      "logps/chosen": -72.5,
+      "logps/rejected": -213.0,
+      "loss": 0.454,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": -0.7265625,
+      "rewards/margins": 1.40625,
+      "rewards/rejected": -2.125,
+      "step": 2760
+    },
+    {
+      "epoch": 2.8990057561486133,
+      "grad_norm": 2.470580041533123,
+      "learning_rate": 1.6734241851733e-07,
+      "logits/chosen": -2.453125,
+      "logits/rejected": 1.8046875,
+      "logps/chosen": -66.0,
+      "logps/rejected": -213.0,
+      "loss": 0.4464,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -0.6640625,
+      "rewards/margins": 1.46875,
+      "rewards/rejected": -2.125,
+      "step": 2770
+    },
+    {
+      "epoch": 2.909471480900052,
+      "grad_norm": 2.2040971822333373,
+      "learning_rate": 1.339964962062984e-07,
+      "logits/chosen": -2.84375,
+      "logits/rejected": 1.734375,
+      "logps/chosen": -76.0,
+      "logps/rejected": -215.0,
+      "loss": 0.4697,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.856249988079071,
+      "rewards/chosen": -0.7578125,
+      "rewards/margins": 1.390625,
+      "rewards/rejected": -2.15625,
+      "step": 2780
+    },
+    {
+      "epoch": 2.9199372056514914,
+      "grad_norm": 3.985477211763175,
+      "learning_rate": 1.0434319319617136e-07,
+      "logits/chosen": -2.171875,
+      "logits/rejected": 1.828125,
+      "logps/chosen": -67.5,
+      "logps/rejected": -202.0,
+      "loss": 0.4615,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8687499761581421,
+      "rewards/chosen": -0.67578125,
+      "rewards/margins": 1.3359375,
+      "rewards/rejected": -2.015625,
+      "step": 2790
+    },
+    {
+      "epoch": 2.9304029304029307,
+      "grad_norm": 3.9221514335396055,
+      "learning_rate": 7.838691302365363e-08,
+      "logits/chosen": -2.125,
+      "logits/rejected": 1.84375,
+      "logps/chosen": -69.5,
+      "logps/rejected": -209.0,
+      "loss": 0.459,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -0.6953125,
+      "rewards/margins": 1.3984375,
+      "rewards/rejected": -2.09375,
+      "step": 2800
+    },
+    {
+      "epoch": 2.9304029304029307,
+      "eval_logits/chosen": -0.578125,
+      "eval_logits/rejected": 1.2421875,
+      "eval_logps/chosen": -113.0,
+      "eval_logps/rejected": -171.0,
+      "eval_loss": 0.8496406078338623,
+      "eval_nll_loss": 0.0,
+      "eval_rewards/accuracies": 0.66796875,
+      "eval_rewards/chosen": -1.1328125,
+      "eval_rewards/margins": 0.57421875,
+      "eval_rewards/rejected": -1.703125,
+      "eval_runtime": 23.4181,
+      "eval_samples_per_second": 85.404,
+      "eval_steps_per_second": 1.366,
+      "step": 2800
+    },
+    {
+      "epoch": 2.9408686551543695,
+      "grad_norm": 3.5437259013948887,
+      "learning_rate": 5.613151021490459e-08,
+      "logits/chosen": -1.96875,
+      "logits/rejected": 1.7890625,
+      "logps/chosen": -80.0,
+      "logps/rejected": -220.0,
+      "loss": 0.4697,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -0.796875,
+      "rewards/margins": 1.3984375,
+      "rewards/rejected": -2.203125,
+      "step": 2810
+    },
+    {
+      "epoch": 2.9513343799058083,
+      "grad_norm": 8.464507978241288,
+      "learning_rate": 3.7580289713159965e-08,
+      "logits/chosen": -2.015625,
+      "logits/rejected": 1.8515625,
+      "logps/chosen": -74.0,
+      "logps/rejected": -216.0,
+      "loss": 0.4658,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -0.7421875,
+      "rewards/margins": 1.421875,
+      "rewards/rejected": -2.15625,
+      "step": 2820
+    },
+    {
+      "epoch": 2.9618001046572475,
+      "grad_norm": 4.245648455723209,
+      "learning_rate": 2.2736006387927277e-08,
+      "logits/chosen": -2.46875,
+      "logits/rejected": 1.546875,
+      "logps/chosen": -67.5,
+      "logps/rejected": -211.0,
+      "loss": 0.4569,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8687499761581421,
+      "rewards/chosen": -0.6796875,
+      "rewards/margins": 1.4296875,
+      "rewards/rejected": -2.109375,
+      "step": 2830
+    },
+    {
+      "epoch": 2.9722658294086868,
+      "grad_norm": 3.9247719496438447,
+      "learning_rate": 1.1600864625893604e-08,
+      "logits/chosen": -2.125,
+      "logits/rejected": 1.953125,
+      "logps/chosen": -72.0,
+      "logps/rejected": -219.0,
+      "loss": 0.4459,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -0.72265625,
+      "rewards/margins": 1.46875,
+      "rewards/rejected": -2.1875,
+      "step": 2840
+    },
+    {
+      "epoch": 2.9827315541601256,
+      "grad_norm": 3.7091291333967007,
+      "learning_rate": 4.176518003573549e-09,
+      "logits/chosen": -1.78125,
+      "logits/rejected": 1.3125,
+      "logps/chosen": -83.5,
+      "logps/rejected": -212.0,
+      "loss": 0.476,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": -0.83203125,
+      "rewards/margins": 1.2890625,
+      "rewards/rejected": -2.125,
+      "step": 2850
+    },
+    {
+      "epoch": 2.9931972789115644,
+      "grad_norm": 3.4248304105006686,
+      "learning_rate": 4.640690417528903e-10,
+      "logits/chosen": -1.859375,
+      "logits/rejected": 1.8984375,
+      "logps/chosen": -75.0,
+      "logps/rejected": -212.0,
+      "loss": 0.4517,
+      "nll_loss": 0.0,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": -0.74609375,
+      "rewards/margins": 1.3671875,
+      "rewards/rejected": -2.125,
+      "step": 2860
+    },
+    {
+      "epoch": 2.998430141287284,
+      "step": 2865,
+      "total_flos": 0.0,
+      "train_loss": 0.6978365646815009,
+      "train_runtime": 9531.6254,
+      "train_samples_per_second": 19.241,
+      "train_steps_per_second": 0.301
+    }
+  ],
+  "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": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}