diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,4795 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 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": 0.0010465724751439038,
+      "grad_norm": 1.423838489515402,
+      "learning_rate": 1.7421602787456448e-07,
+      "logits/chosen": -11.5,
+      "logits/rejected": -11.25,
+      "logps/chosen": -452.0,
+      "logps/rejected": -400.0,
+      "loss": 0.6914,
+      "rewards/accuracies": 0.0,
+      "rewards/chosen": 0.0,
+      "rewards/margins": 0.0,
+      "rewards/rejected": 0.0,
+      "step": 1
+    },
+    {
+      "epoch": 0.010465724751439037,
+      "grad_norm": 1.1995822539078944,
+      "learning_rate": 1.7421602787456445e-06,
+      "logits/chosen": -10.8125,
+      "logits/rejected": -10.875,
+      "logps/chosen": -300.0,
+      "logps/rejected": -266.0,
+      "loss": 0.6916,
+      "rewards/accuracies": 0.2291666716337204,
+      "rewards/chosen": -0.0003261566162109375,
+      "rewards/margins": -0.000705718994140625,
+      "rewards/rejected": 0.0003833770751953125,
+      "step": 10
+    },
+    {
+      "epoch": 0.020931449502878074,
+      "grad_norm": 1.3700703223261517,
+      "learning_rate": 3.484320557491289e-06,
+      "logits/chosen": -10.625,
+      "logits/rejected": -10.4375,
+      "logps/chosen": -328.0,
+      "logps/rejected": -312.0,
+      "loss": 0.6911,
+      "rewards/accuracies": 0.26249998807907104,
+      "rewards/chosen": 0.00154876708984375,
+      "rewards/margins": 0.000244140625,
+      "rewards/rejected": 0.00130462646484375,
+      "step": 20
+    },
+    {
+      "epoch": 0.03139717425431711,
+      "grad_norm": 1.4131432036851739,
+      "learning_rate": 5.226480836236934e-06,
+      "logits/chosen": -11.125,
+      "logits/rejected": -10.8125,
+      "logps/chosen": -318.0,
+      "logps/rejected": -312.0,
+      "loss": 0.688,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.0029144287109375,
+      "rewards/margins": 0.005767822265625,
+      "rewards/rejected": -0.0028533935546875,
+      "step": 30
+    },
+    {
+      "epoch": 0.04186289900575615,
+      "grad_norm": 1.4043578434938189,
+      "learning_rate": 6.968641114982578e-06,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -11.4375,
+      "logps/chosen": -342.0,
+      "logps/rejected": -316.0,
+      "loss": 0.6805,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.03955078125,
+      "rewards/margins": 0.02099609375,
+      "rewards/rejected": -0.060791015625,
+      "step": 40
+    },
+    {
+      "epoch": 0.052328623757195186,
+      "grad_norm": 1.450688162063522,
+      "learning_rate": 8.710801393728225e-06,
+      "logits/chosen": -12.125,
+      "logits/rejected": -11.8125,
+      "logps/chosen": -334.0,
+      "logps/rejected": -288.0,
+      "loss": 0.6697,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -0.1708984375,
+      "rewards/margins": 0.052001953125,
+      "rewards/rejected": -0.22265625,
+      "step": 50
+    },
+    {
+      "epoch": 0.06279434850863422,
+      "grad_norm": 1.9906017630847435,
+      "learning_rate": 1.0452961672473868e-05,
+      "logits/chosen": -12.125,
+      "logits/rejected": -11.9375,
+      "logps/chosen": -398.0,
+      "logps/rejected": -366.0,
+      "loss": 0.6486,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": -0.423828125,
+      "rewards/margins": 0.10791015625,
+      "rewards/rejected": -0.53125,
+      "step": 60
+    },
+    {
+      "epoch": 0.07326007326007326,
+      "grad_norm": 1.9154659821149622,
+      "learning_rate": 1.2195121951219513e-05,
+      "logits/chosen": -13.125,
+      "logits/rejected": -12.75,
+      "logps/chosen": -382.0,
+      "logps/rejected": -388.0,
+      "loss": 0.6397,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -0.79296875,
+      "rewards/margins": 0.232421875,
+      "rewards/rejected": -1.0234375,
+      "step": 70
+    },
+    {
+      "epoch": 0.0837257980115123,
+      "grad_norm": 1.8694878846616096,
+      "learning_rate": 1.3937282229965156e-05,
+      "logits/chosen": -13.375,
+      "logits/rejected": -13.5625,
+      "logps/chosen": -416.0,
+      "logps/rejected": -390.0,
+      "loss": 0.6388,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.8515625,
+      "rewards/margins": 0.21484375,
+      "rewards/rejected": -1.0703125,
+      "step": 80
+    },
+    {
+      "epoch": 0.09419152276295134,
+      "grad_norm": 1.7383009694538945,
+      "learning_rate": 1.56794425087108e-05,
+      "logits/chosen": -11.25,
+      "logits/rejected": -11.1875,
+      "logps/chosen": -362.0,
+      "logps/rejected": -382.0,
+      "loss": 0.6213,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": -0.71875,
+      "rewards/margins": 0.13671875,
+      "rewards/rejected": -0.85546875,
+      "step": 90
+    },
+    {
+      "epoch": 0.10465724751439037,
+      "grad_norm": 1.6390041797429404,
+      "learning_rate": 1.742160278745645e-05,
+      "logits/chosen": -11.1875,
+      "logits/rejected": -10.9375,
+      "logps/chosen": -454.0,
+      "logps/rejected": -438.0,
+      "loss": 0.6194,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -1.0234375,
+      "rewards/margins": 0.26171875,
+      "rewards/rejected": -1.28125,
+      "step": 100
+    },
+    {
+      "epoch": 0.10465724751439037,
+      "eval_logits/chosen": -11.0,
+      "eval_logits/rejected": -10.75,
+      "eval_logps/chosen": -406.0,
+      "eval_logps/rejected": -406.0,
+      "eval_loss": 0.6170937418937683,
+      "eval_rewards/accuracies": 0.67578125,
+      "eval_rewards/chosen": -0.875,
+      "eval_rewards/margins": 0.30078125,
+      "eval_rewards/rejected": -1.1796875,
+      "eval_runtime": 47.6881,
+      "eval_samples_per_second": 41.939,
+      "eval_steps_per_second": 0.671,
+      "step": 100
+    },
+    {
+      "epoch": 0.1151229722658294,
+      "grad_norm": 1.7519249948990894,
+      "learning_rate": 1.9163763066202093e-05,
+      "logits/chosen": -10.5625,
+      "logits/rejected": -10.375,
+      "logps/chosen": -362.0,
+      "logps/rejected": -374.0,
+      "loss": 0.6232,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.765625,
+      "rewards/margins": 0.255859375,
+      "rewards/rejected": -1.0234375,
+      "step": 110
+    },
+    {
+      "epoch": 0.12558869701726844,
+      "grad_norm": 2.285831332008614,
+      "learning_rate": 2.0905923344947736e-05,
+      "logits/chosen": -10.9375,
+      "logits/rejected": -10.5,
+      "logps/chosen": -402.0,
+      "logps/rejected": -444.0,
+      "loss": 0.6,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -0.89453125,
+      "rewards/margins": 0.328125,
+      "rewards/rejected": -1.2265625,
+      "step": 120
+    },
+    {
+      "epoch": 0.1360544217687075,
+      "grad_norm": 2.0847139937135464,
+      "learning_rate": 2.264808362369338e-05,
+      "logits/chosen": -11.1875,
+      "logits/rejected": -10.8125,
+      "logps/chosen": -454.0,
+      "logps/rejected": -458.0,
+      "loss": 0.6063,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -1.1953125,
+      "rewards/margins": 0.49609375,
+      "rewards/rejected": -1.6953125,
+      "step": 130
+    },
+    {
+      "epoch": 0.14652014652014653,
+      "grad_norm": 1.9493609348131078,
+      "learning_rate": 2.4390243902439026e-05,
+      "logits/chosen": -11.25,
+      "logits/rejected": -11.1875,
+      "logps/chosen": -448.0,
+      "logps/rejected": -458.0,
+      "loss": 0.5947,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -1.09375,
+      "rewards/margins": 0.3515625,
+      "rewards/rejected": -1.4453125,
+      "step": 140
+    },
+    {
+      "epoch": 0.15698587127158556,
+      "grad_norm": 1.838551251902883,
+      "learning_rate": 2.6132404181184672e-05,
+      "logits/chosen": -11.9375,
+      "logits/rejected": -11.5625,
+      "logps/chosen": -442.0,
+      "logps/rejected": -422.0,
+      "loss": 0.6101,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -1.046875,
+      "rewards/margins": 0.3359375,
+      "rewards/rejected": -1.3828125,
+      "step": 150
+    },
+    {
+      "epoch": 0.1674515960230246,
+      "grad_norm": 2.6259432978315527,
+      "learning_rate": 2.7874564459930312e-05,
+      "logits/chosen": -11.5625,
+      "logits/rejected": -11.0625,
+      "logps/chosen": -504.0,
+      "logps/rejected": -474.0,
+      "loss": 0.6084,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -1.5234375,
+      "rewards/margins": 0.3671875,
+      "rewards/rejected": -1.890625,
+      "step": 160
+    },
+    {
+      "epoch": 0.17791732077446362,
+      "grad_norm": 1.768983771164199,
+      "learning_rate": 2.9616724738675962e-05,
+      "logits/chosen": -10.4375,
+      "logits/rejected": -9.8125,
+      "logps/chosen": -504.0,
+      "logps/rejected": -516.0,
+      "loss": 0.6033,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -1.8515625,
+      "rewards/margins": 0.58203125,
+      "rewards/rejected": -2.421875,
+      "step": 170
+    },
+    {
+      "epoch": 0.18838304552590268,
+      "grad_norm": 2.066435208086371,
+      "learning_rate": 3.13588850174216e-05,
+      "logits/chosen": -10.5625,
+      "logits/rejected": -10.125,
+      "logps/chosen": -416.0,
+      "logps/rejected": -452.0,
+      "loss": 0.5945,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -1.3046875,
+      "rewards/margins": 0.4609375,
+      "rewards/rejected": -1.765625,
+      "step": 180
+    },
+    {
+      "epoch": 0.1988487702773417,
+      "grad_norm": 2.2222290554926505,
+      "learning_rate": 3.310104529616725e-05,
+      "logits/chosen": -10.375,
+      "logits/rejected": -9.9375,
+      "logps/chosen": -456.0,
+      "logps/rejected": -502.0,
+      "loss": 0.5841,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -1.765625,
+      "rewards/margins": 0.46875,
+      "rewards/rejected": -2.234375,
+      "step": 190
+    },
+    {
+      "epoch": 0.20931449502878074,
+      "grad_norm": 2.0153105172188015,
+      "learning_rate": 3.48432055749129e-05,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -11.0,
+      "logps/chosen": -528.0,
+      "logps/rejected": -496.0,
+      "loss": 0.5947,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -1.6640625,
+      "rewards/margins": 0.439453125,
+      "rewards/rejected": -2.109375,
+      "step": 200
+    },
+    {
+      "epoch": 0.20931449502878074,
+      "eval_logits/chosen": -11.75,
+      "eval_logits/rejected": -11.3125,
+      "eval_logps/chosen": -464.0,
+      "eval_logps/rejected": -472.0,
+      "eval_loss": 0.6037968993186951,
+      "eval_rewards/accuracies": 0.66796875,
+      "eval_rewards/chosen": -1.453125,
+      "eval_rewards/margins": 0.384765625,
+      "eval_rewards/rejected": -1.8359375,
+      "eval_runtime": 46.7909,
+      "eval_samples_per_second": 42.743,
+      "eval_steps_per_second": 0.684,
+      "step": 200
+    },
+    {
+      "epoch": 0.21978021978021978,
+      "grad_norm": 1.9311015283386832,
+      "learning_rate": 3.6585365853658535e-05,
+      "logits/chosen": -11.1875,
+      "logits/rejected": -10.875,
+      "logps/chosen": -520.0,
+      "logps/rejected": -520.0,
+      "loss": 0.6345,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -1.8515625,
+      "rewards/margins": 0.236328125,
+      "rewards/rejected": -2.09375,
+      "step": 210
+    },
+    {
+      "epoch": 0.2302459445316588,
+      "grad_norm": 2.1381850595397593,
+      "learning_rate": 3.8327526132404185e-05,
+      "logits/chosen": -10.375,
+      "logits/rejected": -9.75,
+      "logps/chosen": -520.0,
+      "logps/rejected": -480.0,
+      "loss": 0.598,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -1.8203125,
+      "rewards/margins": 0.41015625,
+      "rewards/rejected": -2.234375,
+      "step": 220
+    },
+    {
+      "epoch": 0.24071166928309787,
+      "grad_norm": 1.8436301576547345,
+      "learning_rate": 4.006968641114983e-05,
+      "logits/chosen": -10.125,
+      "logits/rejected": -9.625,
+      "logps/chosen": -504.0,
+      "logps/rejected": -520.0,
+      "loss": 0.5939,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -1.6953125,
+      "rewards/margins": 0.45703125,
+      "rewards/rejected": -2.15625,
+      "step": 230
+    },
+    {
+      "epoch": 0.25117739403453687,
+      "grad_norm": 2.1090000749792583,
+      "learning_rate": 4.181184668989547e-05,
+      "logits/chosen": -9.0,
+      "logits/rejected": -8.5625,
+      "logps/chosen": -524.0,
+      "logps/rejected": -544.0,
+      "loss": 0.5842,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -2.125,
+      "rewards/margins": 0.43359375,
+      "rewards/rejected": -2.5625,
+      "step": 240
+    },
+    {
+      "epoch": 0.2616431187859759,
+      "grad_norm": 2.1080648541903577,
+      "learning_rate": 4.3554006968641115e-05,
+      "logits/chosen": -9.6875,
+      "logits/rejected": -9.3125,
+      "logps/chosen": -556.0,
+      "logps/rejected": -572.0,
+      "loss": 0.5688,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -2.328125,
+      "rewards/margins": 0.5234375,
+      "rewards/rejected": -2.84375,
+      "step": 250
+    },
+    {
+      "epoch": 0.272108843537415,
+      "grad_norm": 4.450060780179574,
+      "learning_rate": 4.529616724738676e-05,
+      "logits/chosen": -11.0,
+      "logits/rejected": -10.6875,
+      "logps/chosen": -588.0,
+      "logps/rejected": -600.0,
+      "loss": 0.6062,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -2.515625,
+      "rewards/margins": 0.361328125,
+      "rewards/rejected": -2.875,
+      "step": 260
+    },
+    {
+      "epoch": 0.282574568288854,
+      "grad_norm": 2.1361509347925356,
+      "learning_rate": 4.703832752613241e-05,
+      "logits/chosen": -11.0625,
+      "logits/rejected": -10.8125,
+      "logps/chosen": -532.0,
+      "logps/rejected": -510.0,
+      "loss": 0.5841,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -1.7890625,
+      "rewards/margins": 0.4453125,
+      "rewards/rejected": -2.234375,
+      "step": 270
+    },
+    {
+      "epoch": 0.29304029304029305,
+      "grad_norm": 1.6603692666594028,
+      "learning_rate": 4.878048780487805e-05,
+      "logits/chosen": -11.5625,
+      "logits/rejected": -11.125,
+      "logps/chosen": -532.0,
+      "logps/rejected": -560.0,
+      "loss": 0.6069,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": -1.8515625,
+      "rewards/margins": 0.5390625,
+      "rewards/rejected": -2.390625,
+      "step": 280
+    },
+    {
+      "epoch": 0.3035060177917321,
+      "grad_norm": 3.338939927835405,
+      "learning_rate": 4.999983293481418e-05,
+      "logits/chosen": -13.875,
+      "logits/rejected": -13.75,
+      "logps/chosen": -512.0,
+      "logps/rejected": -486.0,
+      "loss": 0.6259,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -1.890625,
+      "rewards/margins": 0.3046875,
+      "rewards/rejected": -2.203125,
+      "step": 290
+    },
+    {
+      "epoch": 0.3139717425431711,
+      "grad_norm": 1.8286208937617223,
+      "learning_rate": 4.9996862949181595e-05,
+      "logits/chosen": -13.1875,
+      "logits/rejected": -13.0,
+      "logps/chosen": -556.0,
+      "logps/rejected": -572.0,
+      "loss": 0.6583,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": -2.390625,
+      "rewards/margins": 0.34375,
+      "rewards/rejected": -2.734375,
+      "step": 300
+    },
+    {
+      "epoch": 0.3139717425431711,
+      "eval_logits/chosen": -13.5,
+      "eval_logits/rejected": -13.1875,
+      "eval_logps/chosen": -544.0,
+      "eval_logps/rejected": -560.0,
+      "eval_loss": 0.600656270980835,
+      "eval_rewards/accuracies": 0.67578125,
+      "eval_rewards/chosen": -2.234375,
+      "eval_rewards/margins": 0.494140625,
+      "eval_rewards/rejected": -2.734375,
+      "eval_runtime": 46.7732,
+      "eval_samples_per_second": 42.759,
+      "eval_steps_per_second": 0.684,
+      "step": 300
+    },
+    {
+      "epoch": 0.32443746729461015,
+      "grad_norm": 2.152901324234895,
+      "learning_rate": 4.9990180911526896e-05,
+      "logits/chosen": -14.5625,
+      "logits/rejected": -14.3125,
+      "logps/chosen": -528.0,
+      "logps/rejected": -484.0,
+      "loss": 0.6148,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -1.8671875,
+      "rewards/margins": 0.45703125,
+      "rewards/rejected": -2.328125,
+      "step": 310
+    },
+    {
+      "epoch": 0.3349031920460492,
+      "grad_norm": 1.452134069826545,
+      "learning_rate": 4.997978781413745e-05,
+      "logits/chosen": -12.75,
+      "logits/rejected": -12.4375,
+      "logps/chosen": -482.0,
+      "logps/rejected": -516.0,
+      "loss": 0.6112,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -1.90625,
+      "rewards/margins": 0.44140625,
+      "rewards/rejected": -2.34375,
+      "step": 320
+    },
+    {
+      "epoch": 0.3453689167974882,
+      "grad_norm": 1.8074237494925378,
+      "learning_rate": 4.996568520039565e-05,
+      "logits/chosen": -12.5625,
+      "logits/rejected": -12.1875,
+      "logps/chosen": -520.0,
+      "logps/rejected": -528.0,
+      "loss": 0.6076,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -2.203125,
+      "rewards/margins": 0.431640625,
+      "rewards/rejected": -2.625,
+      "step": 330
+    },
+    {
+      "epoch": 0.35583464154892724,
+      "grad_norm": 1.8519693854023662,
+      "learning_rate": 4.994787516454973e-05,
+      "logits/chosen": -12.5625,
+      "logits/rejected": -12.125,
+      "logps/chosen": -520.0,
+      "logps/rejected": -516.0,
+      "loss": 0.5887,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -2.078125,
+      "rewards/margins": 0.390625,
+      "rewards/rejected": -2.46875,
+      "step": 340
+    },
+    {
+      "epoch": 0.3663003663003663,
+      "grad_norm": 2.140471411373253,
+      "learning_rate": 4.992636035140273e-05,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -11.125,
+      "logps/chosen": -560.0,
+      "logps/rejected": -536.0,
+      "loss": 0.6115,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -2.140625,
+      "rewards/margins": 0.353515625,
+      "rewards/rejected": -2.5,
+      "step": 350
+    },
+    {
+      "epoch": 0.37676609105180536,
+      "grad_norm": 2.0553592638995295,
+      "learning_rate": 4.99011439559198e-05,
+      "logits/chosen": -11.9375,
+      "logits/rejected": -11.5625,
+      "logps/chosen": -520.0,
+      "logps/rejected": -540.0,
+      "loss": 0.6021,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": -2.0,
+      "rewards/margins": 0.515625,
+      "rewards/rejected": -2.515625,
+      "step": 360
+    },
+    {
+      "epoch": 0.3872318158032444,
+      "grad_norm": 1.8838859177468847,
+      "learning_rate": 4.987222972275369e-05,
+      "logits/chosen": -11.375,
+      "logits/rejected": -11.125,
+      "logps/chosen": -502.0,
+      "logps/rejected": -508.0,
+      "loss": 0.5714,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -1.90625,
+      "rewards/margins": 0.435546875,
+      "rewards/rejected": -2.34375,
+      "step": 370
+    },
+    {
+      "epoch": 0.3976975405546834,
+      "grad_norm": 1.962445127175461,
+      "learning_rate": 4.983962194568872e-05,
+      "logits/chosen": -9.5625,
+      "logits/rejected": -9.1875,
+      "logps/chosen": -512.0,
+      "logps/rejected": -548.0,
+      "loss": 0.5882,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -2.140625,
+      "rewards/margins": 0.5078125,
+      "rewards/rejected": -2.640625,
+      "step": 380
+    },
+    {
+      "epoch": 0.40816326530612246,
+      "grad_norm": 1.7245889407962303,
+      "learning_rate": 4.9803325467003084e-05,
+      "logits/chosen": -9.1875,
+      "logits/rejected": -8.75,
+      "logps/chosen": -486.0,
+      "logps/rejected": -492.0,
+      "loss": 0.6186,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -1.5859375,
+      "rewards/margins": 0.30859375,
+      "rewards/rejected": -1.890625,
+      "step": 390
+    },
+    {
+      "epoch": 0.4186289900575615,
+      "grad_norm": 2.1192298359392163,
+      "learning_rate": 4.9763345676749816e-05,
+      "logits/chosen": -9.3125,
+      "logits/rejected": -8.75,
+      "logps/chosen": -510.0,
+      "logps/rejected": -532.0,
+      "loss": 0.6003,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -1.90625,
+      "rewards/margins": 0.43359375,
+      "rewards/rejected": -2.34375,
+      "step": 400
+    },
+    {
+      "epoch": 0.4186289900575615,
+      "eval_logits/chosen": -10.3125,
+      "eval_logits/rejected": -9.75,
+      "eval_logps/chosen": -502.0,
+      "eval_logps/rejected": -528.0,
+      "eval_loss": 0.589187502861023,
+      "eval_rewards/accuracies": 0.701171875,
+      "eval_rewards/chosen": -1.8359375,
+      "eval_rewards/margins": 0.55859375,
+      "eval_rewards/rejected": -2.390625,
+      "eval_runtime": 47.4574,
+      "eval_samples_per_second": 42.143,
+      "eval_steps_per_second": 0.674,
+      "step": 400
+    },
+    {
+      "epoch": 0.4290947148090005,
+      "grad_norm": 1.7950239024727874,
+      "learning_rate": 4.9719688511956374e-05,
+      "logits/chosen": -9.6875,
+      "logits/rejected": -9.1875,
+      "logps/chosen": -572.0,
+      "logps/rejected": -580.0,
+      "loss": 0.6201,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -2.484375,
+      "rewards/margins": 0.51171875,
+      "rewards/rejected": -3.0,
+      "step": 410
+    },
+    {
+      "epoch": 0.43956043956043955,
+      "grad_norm": 1.6290932612717146,
+      "learning_rate": 4.967236045574294e-05,
+      "logits/chosen": -10.125,
+      "logits/rejected": -9.375,
+      "logps/chosen": -556.0,
+      "logps/rejected": -552.0,
+      "loss": 0.6201,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -2.21875,
+      "rewards/margins": 0.4296875,
+      "rewards/rejected": -2.65625,
+      "step": 420
+    },
+    {
+      "epoch": 0.4500261643118786,
+      "grad_norm": 1.5792681116046905,
+      "learning_rate": 4.96213685363597e-05,
+      "logits/chosen": -10.6875,
+      "logits/rejected": -10.1875,
+      "logps/chosen": -490.0,
+      "logps/rejected": -506.0,
+      "loss": 0.6235,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -1.875,
+      "rewards/margins": 0.3046875,
+      "rewards/rejected": -2.171875,
+      "step": 430
+    },
+    {
+      "epoch": 0.4604918890633176,
+      "grad_norm": 2.2485287204441438,
+      "learning_rate": 4.956672032614315e-05,
+      "logits/chosen": -11.25,
+      "logits/rejected": -10.8125,
+      "logps/chosen": -528.0,
+      "logps/rejected": -502.0,
+      "loss": 0.6035,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -1.8515625,
+      "rewards/margins": 0.40625,
+      "rewards/rejected": -2.265625,
+      "step": 440
+    },
+    {
+      "epoch": 0.47095761381475665,
+      "grad_norm": 1.8850325904883005,
+      "learning_rate": 4.950842394039157e-05,
+      "logits/chosen": -9.9375,
+      "logits/rejected": -9.25,
+      "logps/chosen": -568.0,
+      "logps/rejected": -612.0,
+      "loss": 0.5959,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -2.296875,
+      "rewards/margins": 0.703125,
+      "rewards/rejected": -3.0,
+      "step": 450
+    },
+    {
+      "epoch": 0.48142333856619574,
+      "grad_norm": 1.7459882804220206,
+      "learning_rate": 4.944648803615994e-05,
+      "logits/chosen": -10.6875,
+      "logits/rejected": -10.25,
+      "logps/chosen": -520.0,
+      "logps/rejected": -536.0,
+      "loss": 0.5951,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -2.140625,
+      "rewards/margins": 0.51171875,
+      "rewards/rejected": -2.640625,
+      "step": 460
+    },
+    {
+      "epoch": 0.49188906331763477,
+      "grad_norm": 1.6726566999491639,
+      "learning_rate": 4.938092181097433e-05,
+      "logits/chosen": -11.625,
+      "logits/rejected": -11.0,
+      "logps/chosen": -516.0,
+      "logps/rejected": -524.0,
+      "loss": 0.5277,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -1.859375,
+      "rewards/margins": 0.64453125,
+      "rewards/rejected": -2.5,
+      "step": 470
+    },
+    {
+      "epoch": 0.5023547880690737,
+      "grad_norm": 2.1124929125955494,
+      "learning_rate": 4.931173500146607e-05,
+      "logits/chosen": -10.6875,
+      "logits/rejected": -9.75,
+      "logps/chosen": -544.0,
+      "logps/rejected": -592.0,
+      "loss": 0.5775,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -2.28125,
+      "rewards/margins": 0.76953125,
+      "rewards/rejected": -3.0625,
+      "step": 480
+    },
+    {
+      "epoch": 0.5128205128205128,
+      "grad_norm": 1.9248431327899287,
+      "learning_rate": 4.923893788192582e-05,
+      "logits/chosen": -9.875,
+      "logits/rejected": -8.625,
+      "logps/chosen": -600.0,
+      "logps/rejected": -612.0,
+      "loss": 0.5563,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": -2.3125,
+      "rewards/margins": 0.9296875,
+      "rewards/rejected": -3.25,
+      "step": 490
+    },
+    {
+      "epoch": 0.5232862375719518,
+      "grad_norm": 1.634547798917106,
+      "learning_rate": 4.9162541262777906e-05,
+      "logits/chosen": -10.625,
+      "logits/rejected": -10.125,
+      "logps/chosen": -510.0,
+      "logps/rejected": -516.0,
+      "loss": 0.5701,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -2.078125,
+      "rewards/margins": 0.54296875,
+      "rewards/rejected": -2.625,
+      "step": 500
+    },
+    {
+      "epoch": 0.5232862375719518,
+      "eval_logits/chosen": -11.0,
+      "eval_logits/rejected": -10.5,
+      "eval_logps/chosen": -516.0,
+      "eval_logps/rejected": -540.0,
+      "eval_loss": 0.577218770980835,
+      "eval_rewards/accuracies": 0.6875,
+      "eval_rewards/chosen": -1.96875,
+      "eval_rewards/margins": 0.5390625,
+      "eval_rewards/rejected": -2.5,
+      "eval_runtime": 46.7704,
+      "eval_samples_per_second": 42.762,
+      "eval_steps_per_second": 0.684,
+      "step": 500
+    },
+    {
+      "epoch": 0.533751962323391,
+      "grad_norm": 1.6420620280484055,
+      "learning_rate": 4.90825564889749e-05,
+      "logits/chosen": -10.625,
+      "logits/rejected": -10.0625,
+      "logps/chosen": -544.0,
+      "logps/rejected": -572.0,
+      "loss": 0.57,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -2.078125,
+      "rewards/margins": 0.6796875,
+      "rewards/rejected": -2.765625,
+      "step": 510
+    },
+    {
+      "epoch": 0.54421768707483,
+      "grad_norm": 2.002572369961546,
+      "learning_rate": 4.899899543831288e-05,
+      "logits/chosen": -7.78125,
+      "logits/rejected": -7.125,
+      "logps/chosen": -620.0,
+      "logps/rejected": -636.0,
+      "loss": 0.5863,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -2.96875,
+      "rewards/margins": 0.51953125,
+      "rewards/rejected": -3.5,
+      "step": 520
+    },
+    {
+      "epoch": 0.554683411826269,
+      "grad_norm": 1.6511215262986447,
+      "learning_rate": 4.891187051966765e-05,
+      "logits/chosen": -10.0,
+      "logits/rejected": -9.375,
+      "logps/chosen": -516.0,
+      "logps/rejected": -540.0,
+      "loss": 0.6075,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -1.96875,
+      "rewards/margins": 0.56640625,
+      "rewards/rejected": -2.53125,
+      "step": 530
+    },
+    {
+      "epoch": 0.565149136577708,
+      "grad_norm": 1.7449800492439143,
+      "learning_rate": 4.882119467115191e-05,
+      "logits/chosen": -10.3125,
+      "logits/rejected": -9.375,
+      "logps/chosen": -516.0,
+      "logps/rejected": -508.0,
+      "loss": 0.5848,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -1.6640625,
+      "rewards/margins": 0.46875,
+      "rewards/rejected": -2.125,
+      "step": 540
+    },
+    {
+      "epoch": 0.5756148613291471,
+      "grad_norm": 1.3839716354458609,
+      "learning_rate": 4.872698135819402e-05,
+      "logits/chosen": -9.1875,
+      "logits/rejected": -8.375,
+      "logps/chosen": -556.0,
+      "logps/rejected": -544.0,
+      "loss": 0.6022,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -2.40625,
+      "rewards/margins": 0.51953125,
+      "rewards/rejected": -2.921875,
+      "step": 550
+    },
+    {
+      "epoch": 0.5860805860805861,
+      "grad_norm": 1.8151605373491886,
+      "learning_rate": 4.862924457153831e-05,
+      "logits/chosen": -9.5625,
+      "logits/rejected": -8.6875,
+      "logps/chosen": -532.0,
+      "logps/rejected": -556.0,
+      "loss": 0.5948,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -2.015625,
+      "rewards/margins": 0.6328125,
+      "rewards/rejected": -2.65625,
+      "step": 560
+    },
+    {
+      "epoch": 0.5965463108320251,
+      "grad_norm": 1.513503501669786,
+      "learning_rate": 4.8527998825167484e-05,
+      "logits/chosen": -9.8125,
+      "logits/rejected": -9.0625,
+      "logps/chosen": -560.0,
+      "logps/rejected": -576.0,
+      "loss": 0.5666,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -2.296875,
+      "rewards/margins": 0.478515625,
+      "rewards/rejected": -2.78125,
+      "step": 570
+    },
+    {
+      "epoch": 0.6070120355834642,
+      "grad_norm": 1.9418969743992534,
+      "learning_rate": 4.842325915414728e-05,
+      "logits/chosen": -7.75,
+      "logits/rejected": -7.0,
+      "logps/chosen": -636.0,
+      "logps/rejected": -632.0,
+      "loss": 0.5829,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -3.40625,
+      "rewards/margins": 0.5390625,
+      "rewards/rejected": -3.9375,
+      "step": 580
+    },
+    {
+      "epoch": 0.6174777603349032,
+      "grad_norm": 1.6420700932006496,
+      "learning_rate": 4.831504111239373e-05,
+      "logits/chosen": -9.875,
+      "logits/rejected": -8.875,
+      "logps/chosen": -628.0,
+      "logps/rejected": -664.0,
+      "loss": 0.5526,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -3.03125,
+      "rewards/margins": 0.64453125,
+      "rewards/rejected": -3.671875,
+      "step": 590
+    },
+    {
+      "epoch": 0.6279434850863422,
+      "grad_norm": 1.5933072439575753,
+      "learning_rate": 4.8203360770363425e-05,
+      "logits/chosen": -10.75,
+      "logits/rejected": -10.0,
+      "logps/chosen": -564.0,
+      "logps/rejected": -600.0,
+      "loss": 0.55,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -2.515625,
+      "rewards/margins": 0.62890625,
+      "rewards/rejected": -3.15625,
+      "step": 600
+    },
+    {
+      "epoch": 0.6279434850863422,
+      "eval_logits/chosen": -10.4375,
+      "eval_logits/rejected": -9.5625,
+      "eval_logps/chosen": -588.0,
+      "eval_logps/rejected": -632.0,
+      "eval_loss": 0.5670937299728394,
+      "eval_rewards/accuracies": 0.712890625,
+      "eval_rewards/chosen": -2.6875,
+      "eval_rewards/margins": 0.7265625,
+      "eval_rewards/rejected": -3.421875,
+      "eval_runtime": 46.7684,
+      "eval_samples_per_second": 42.764,
+      "eval_steps_per_second": 0.684,
+      "step": 600
+    },
+    {
+      "epoch": 0.6384092098377813,
+      "grad_norm": 1.7588690739714166,
+      "learning_rate": 4.808823471266702e-05,
+      "logits/chosen": -10.4375,
+      "logits/rejected": -9.125,
+      "logps/chosen": -628.0,
+      "logps/rejected": -624.0,
+      "loss": 0.5906,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -2.671875,
+      "rewards/margins": 0.765625,
+      "rewards/rejected": -3.4375,
+      "step": 610
+    },
+    {
+      "epoch": 0.6488749345892203,
+      "grad_norm": 1.8598149067547787,
+      "learning_rate": 4.796968003560639e-05,
+      "logits/chosen": -8.9375,
+      "logits/rejected": -8.0625,
+      "logps/chosen": -556.0,
+      "logps/rejected": -576.0,
+      "loss": 0.5665,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -2.375,
+      "rewards/margins": 0.7265625,
+      "rewards/rejected": -3.109375,
+      "step": 620
+    },
+    {
+      "epoch": 0.6593406593406593,
+      "grad_norm": 1.6894526811370767,
+      "learning_rate": 4.7847714344635864e-05,
+      "logits/chosen": -8.9375,
+      "logits/rejected": -7.90625,
+      "logps/chosen": -564.0,
+      "logps/rejected": -580.0,
+      "loss": 0.5489,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": -2.453125,
+      "rewards/margins": 0.72265625,
+      "rewards/rejected": -3.171875,
+      "step": 630
+    },
+    {
+      "epoch": 0.6698063840920984,
+      "grad_norm": 2.3643105347238493,
+      "learning_rate": 4.772235575174776e-05,
+      "logits/chosen": -8.5625,
+      "logits/rejected": -7.5625,
+      "logps/chosen": -592.0,
+      "logps/rejected": -608.0,
+      "loss": 0.5785,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -2.78125,
+      "rewards/margins": 0.6875,
+      "rewards/rejected": -3.46875,
+      "step": 640
+    },
+    {
+      "epoch": 0.6802721088435374,
+      "grad_norm": 1.5633903270957807,
+      "learning_rate": 4.7593622872782754e-05,
+      "logits/chosen": -9.75,
+      "logits/rejected": -9.125,
+      "logps/chosen": -528.0,
+      "logps/rejected": -584.0,
+      "loss": 0.5661,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -2.21875,
+      "rewards/margins": 0.54296875,
+      "rewards/rejected": -2.765625,
+      "step": 650
+    },
+    {
+      "epoch": 0.6907378335949764,
+      "grad_norm": 1.5295325214260047,
+      "learning_rate": 4.746153482466542e-05,
+      "logits/chosen": -7.375,
+      "logits/rejected": -6.21875,
+      "logps/chosen": -640.0,
+      "logps/rejected": -672.0,
+      "loss": 0.5572,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": -3.03125,
+      "rewards/margins": 0.78125,
+      "rewards/rejected": -3.8125,
+      "step": 660
+    },
+    {
+      "epoch": 0.7012035583464155,
+      "grad_norm": 1.597206430655417,
+      "learning_rate": 4.732611122256537e-05,
+      "logits/chosen": -7.84375,
+      "logits/rejected": -7.0625,
+      "logps/chosen": -680.0,
+      "logps/rejected": -720.0,
+      "loss": 0.5445,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -3.34375,
+      "rewards/margins": 0.625,
+      "rewards/rejected": -3.984375,
+      "step": 670
+    },
+    {
+      "epoch": 0.7116692830978545,
+      "grad_norm": 1.9360061992487616,
+      "learning_rate": 4.718737217698435e-05,
+      "logits/chosen": -10.75,
+      "logits/rejected": -10.0625,
+      "logps/chosen": -588.0,
+      "logps/rejected": -608.0,
+      "loss": 0.5946,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -2.203125,
+      "rewards/margins": 0.6015625,
+      "rewards/rejected": -2.796875,
+      "step": 680
+    },
+    {
+      "epoch": 0.7221350078492935,
+      "grad_norm": 1.5217924791636162,
+      "learning_rate": 4.704533829076982e-05,
+      "logits/chosen": -10.4375,
+      "logits/rejected": -9.125,
+      "logps/chosen": -548.0,
+      "logps/rejected": -572.0,
+      "loss": 0.5323,
+      "rewards/accuracies": 0.768750011920929,
+      "rewards/chosen": -2.28125,
+      "rewards/margins": 0.81640625,
+      "rewards/rejected": -3.09375,
+      "step": 690
+    },
+    {
+      "epoch": 0.7326007326007326,
+      "grad_norm": 2.1553389513644277,
+      "learning_rate": 4.6900030656055455e-05,
+      "logits/chosen": -10.125,
+      "logits/rejected": -9.4375,
+      "logps/chosen": -548.0,
+      "logps/rejected": -584.0,
+      "loss": 0.554,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -2.5,
+      "rewards/margins": 0.65625,
+      "rewards/rejected": -3.171875,
+      "step": 700
+    },
+    {
+      "epoch": 0.7326007326007326,
+      "eval_logits/chosen": -10.0625,
+      "eval_logits/rejected": -9.25,
+      "eval_logps/chosen": -580.0,
+      "eval_logps/rejected": -628.0,
+      "eval_loss": 0.566656231880188,
+      "eval_rewards/accuracies": 0.728515625,
+      "eval_rewards/chosen": -2.625,
+      "eval_rewards/margins": 0.75,
+      "eval_rewards/rejected": -3.375,
+      "eval_runtime": 46.768,
+      "eval_samples_per_second": 42.764,
+      "eval_steps_per_second": 0.684,
+      "step": 700
+    },
+    {
+      "epoch": 0.7430664573521716,
+      "grad_norm": 1.9797449060676071,
+      "learning_rate": 4.6751470851128906e-05,
+      "logits/chosen": -9.9375,
+      "logits/rejected": -8.875,
+      "logps/chosen": -564.0,
+      "logps/rejected": -596.0,
+      "loss": 0.5913,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -2.625,
+      "rewards/margins": 0.6484375,
+      "rewards/rejected": -3.28125,
+      "step": 710
+    },
+    {
+      "epoch": 0.7535321821036107,
+      "grad_norm": 1.9049411439351829,
+      "learning_rate": 4.6599680937227454e-05,
+      "logits/chosen": -10.0625,
+      "logits/rejected": -9.125,
+      "logps/chosen": -604.0,
+      "logps/rejected": -612.0,
+      "loss": 0.5628,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -2.625,
+      "rewards/margins": 0.67578125,
+      "rewards/rejected": -3.296875,
+      "step": 720
+    },
+    {
+      "epoch": 0.7639979068550498,
+      "grad_norm": 2.088445590448695,
+      "learning_rate": 4.644468345526183e-05,
+      "logits/chosen": -8.375,
+      "logits/rejected": -7.5625,
+      "logps/chosen": -640.0,
+      "logps/rejected": -644.0,
+      "loss": 0.5915,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": -3.265625,
+      "rewards/margins": 0.494140625,
+      "rewards/rejected": -3.765625,
+      "step": 730
+    },
+    {
+      "epoch": 0.7744636316064888,
+      "grad_norm": 3.063296982655387,
+      "learning_rate": 4.628650142246894e-05,
+      "logits/chosen": -8.25,
+      "logits/rejected": -7.5625,
+      "logps/chosen": -596.0,
+      "logps/rejected": -632.0,
+      "loss": 0.5501,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -3.015625,
+      "rewards/margins": 0.63671875,
+      "rewards/rejected": -3.65625,
+      "step": 740
+    },
+    {
+      "epoch": 0.7849293563579278,
+      "grad_norm": 1.4309276630623318,
+      "learning_rate": 4.6125158328993765e-05,
+      "logits/chosen": -8.75,
+      "logits/rejected": -7.90625,
+      "logps/chosen": -560.0,
+      "logps/rejected": -628.0,
+      "loss": 0.5307,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -2.78125,
+      "rewards/margins": 0.73828125,
+      "rewards/rejected": -3.515625,
+      "step": 750
+    },
+    {
+      "epoch": 0.7953950811093669,
+      "grad_norm": 1.8220212478027,
+      "learning_rate": 4.5960678134401064e-05,
+      "logits/chosen": -9.25,
+      "logits/rejected": -8.0,
+      "logps/chosen": -552.0,
+      "logps/rejected": -612.0,
+      "loss": 0.5676,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -2.671875,
+      "rewards/margins": 0.734375,
+      "rewards/rejected": -3.40625,
+      "step": 760
+    },
+    {
+      "epoch": 0.8058608058608059,
+      "grad_norm": 1.5790495731119136,
+      "learning_rate": 4.579308526411734e-05,
+      "logits/chosen": -9.8125,
+      "logits/rejected": -8.9375,
+      "logps/chosen": -624.0,
+      "logps/rejected": -700.0,
+      "loss": 0.5425,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -2.734375,
+      "rewards/margins": 0.625,
+      "rewards/rejected": -3.359375,
+      "step": 770
+    },
+    {
+      "epoch": 0.8163265306122449,
+      "grad_norm": 2.0367017463968646,
+      "learning_rate": 4.562240460580367e-05,
+      "logits/chosen": -9.4375,
+      "logits/rejected": -8.25,
+      "logps/chosen": -668.0,
+      "logps/rejected": -680.0,
+      "loss": 0.5663,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -3.0625,
+      "rewards/margins": 0.74609375,
+      "rewards/rejected": -3.8125,
+      "step": 780
+    },
+    {
+      "epoch": 0.826792255363684,
+      "grad_norm": 1.889717374300922,
+      "learning_rate": 4.544866150565985e-05,
+      "logits/chosen": -9.375,
+      "logits/rejected": -8.0625,
+      "logps/chosen": -636.0,
+      "logps/rejected": -644.0,
+      "loss": 0.5957,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -3.046875,
+      "rewards/margins": 0.625,
+      "rewards/rejected": -3.671875,
+      "step": 790
+    },
+    {
+      "epoch": 0.837257980115123,
+      "grad_norm": 1.660412852646225,
+      "learning_rate": 4.527188176466051e-05,
+      "logits/chosen": -10.0625,
+      "logits/rejected": -9.125,
+      "logps/chosen": -572.0,
+      "logps/rejected": -624.0,
+      "loss": 0.5478,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -2.609375,
+      "rewards/margins": 0.67578125,
+      "rewards/rejected": -3.28125,
+      "step": 800
+    },
+    {
+      "epoch": 0.837257980115123,
+      "eval_logits/chosen": -9.875,
+      "eval_logits/rejected": -8.9375,
+      "eval_logps/chosen": -592.0,
+      "eval_logps/rejected": -628.0,
+      "eval_loss": 0.5699218511581421,
+      "eval_rewards/accuracies": 0.70703125,
+      "eval_rewards/chosen": -2.71875,
+      "eval_rewards/margins": 0.66015625,
+      "eval_rewards/rejected": -3.390625,
+      "eval_runtime": 46.8116,
+      "eval_samples_per_second": 42.724,
+      "eval_steps_per_second": 0.684,
+      "step": 800
+    },
+    {
+      "epoch": 0.847723704866562,
+      "grad_norm": 1.8931681136948715,
+      "learning_rate": 4.509209163472357e-05,
+      "logits/chosen": -9.375,
+      "logits/rejected": -8.4375,
+      "logps/chosen": -636.0,
+      "logps/rejected": -632.0,
+      "loss": 0.5798,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -2.96875,
+      "rewards/margins": 0.57421875,
+      "rewards/rejected": -3.546875,
+      "step": 810
+    },
+    {
+      "epoch": 0.858189429618001,
+      "grad_norm": 1.8786697897903046,
+      "learning_rate": 4.4909317814811896e-05,
+      "logits/chosen": -9.125,
+      "logits/rejected": -8.1875,
+      "logps/chosen": -656.0,
+      "logps/rejected": -692.0,
+      "loss": 0.5807,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -3.015625,
+      "rewards/margins": 0.68359375,
+      "rewards/rejected": -3.703125,
+      "step": 820
+    },
+    {
+      "epoch": 0.8686551543694401,
+      "grad_norm": 1.4883496802875167,
+      "learning_rate": 4.472358744696845e-05,
+      "logits/chosen": -10.75,
+      "logits/rejected": -9.9375,
+      "logps/chosen": -556.0,
+      "logps/rejected": -604.0,
+      "loss": 0.5718,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -2.5625,
+      "rewards/margins": 0.6640625,
+      "rewards/rejected": -3.21875,
+      "step": 830
+    },
+    {
+      "epoch": 0.8791208791208791,
+      "grad_norm": 1.9532117165652396,
+      "learning_rate": 4.4534928112285645e-05,
+      "logits/chosen": -10.375,
+      "logits/rejected": -8.75,
+      "logps/chosen": -568.0,
+      "logps/rejected": -624.0,
+      "loss": 0.5266,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -2.421875,
+      "rewards/margins": 0.84765625,
+      "rewards/rejected": -3.265625,
+      "step": 840
+    },
+    {
+      "epoch": 0.8895866038723181,
+      "grad_norm": 2.5191105236573863,
+      "learning_rate": 4.43433678268096e-05,
+      "logits/chosen": -7.59375,
+      "logits/rejected": -5.75,
+      "logps/chosen": -616.0,
+      "logps/rejected": -636.0,
+      "loss": 0.5352,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -2.875,
+      "rewards/margins": 0.89453125,
+      "rewards/rejected": -3.765625,
+      "step": 850
+    },
+    {
+      "epoch": 0.9000523286237572,
+      "grad_norm": 1.7511375796991135,
+      "learning_rate": 4.414893503737968e-05,
+      "logits/chosen": -7.625,
+      "logits/rejected": -6.25,
+      "logps/chosen": -684.0,
+      "logps/rejected": -696.0,
+      "loss": 0.5746,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -3.5,
+      "rewards/margins": 0.58984375,
+      "rewards/rejected": -4.09375,
+      "step": 860
+    },
+    {
+      "epoch": 0.9105180533751962,
+      "grad_norm": 1.6576281921470557,
+      "learning_rate": 4.395165861740414e-05,
+      "logits/chosen": -10.0625,
+      "logits/rejected": -9.1875,
+      "logps/chosen": -572.0,
+      "logps/rejected": -588.0,
+      "loss": 0.5837,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -2.53125,
+      "rewards/margins": 0.5703125,
+      "rewards/rejected": -3.09375,
+      "step": 870
+    },
+    {
+      "epoch": 0.9209837781266352,
+      "grad_norm": 1.7836434206135487,
+      "learning_rate": 4.375156786257241e-05,
+      "logits/chosen": -9.5,
+      "logits/rejected": -8.6875,
+      "logps/chosen": -628.0,
+      "logps/rejected": -680.0,
+      "loss": 0.5563,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -2.796875,
+      "rewards/margins": 0.6328125,
+      "rewards/rejected": -3.4375,
+      "step": 880
+    },
+    {
+      "epoch": 0.9314495028780743,
+      "grad_norm": 1.368517931775755,
+      "learning_rate": 4.354869248650466e-05,
+      "logits/chosen": -10.0,
+      "logits/rejected": -9.0,
+      "logps/chosen": -640.0,
+      "logps/rejected": -684.0,
+      "loss": 0.59,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -2.96875,
+      "rewards/margins": 0.72265625,
+      "rewards/rejected": -3.6875,
+      "step": 890
+    },
+    {
+      "epoch": 0.9419152276295133,
+      "grad_norm": 2.055267172154011,
+      "learning_rate": 4.334306261633931e-05,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -10.8125,
+      "logps/chosen": -604.0,
+      "logps/rejected": -652.0,
+      "loss": 0.5759,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -2.734375,
+      "rewards/margins": 0.5625,
+      "rewards/rejected": -3.296875,
+      "step": 900
+    },
+    {
+      "epoch": 0.9419152276295133,
+      "eval_logits/chosen": -11.1875,
+      "eval_logits/rejected": -10.25,
+      "eval_logps/chosen": -592.0,
+      "eval_logps/rejected": -632.0,
+      "eval_loss": 0.5660468935966492,
+      "eval_rewards/accuracies": 0.708984375,
+      "eval_rewards/chosen": -2.75,
+      "eval_rewards/margins": 0.69140625,
+      "eval_rewards/rejected": -3.4375,
+      "eval_runtime": 46.766,
+      "eval_samples_per_second": 42.766,
+      "eval_steps_per_second": 0.684,
+      "step": 900
+    },
+    {
+      "epoch": 0.9523809523809523,
+      "grad_norm": 1.7833383755046561,
+      "learning_rate": 4.313470878825911e-05,
+      "logits/chosen": -10.625,
+      "logits/rejected": -9.6875,
+      "logps/chosen": -510.0,
+      "logps/rejected": -556.0,
+      "loss": 0.5839,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -2.515625,
+      "rewards/margins": 0.63671875,
+      "rewards/rejected": -3.15625,
+      "step": 910
+    },
+    {
+      "epoch": 0.9628466771323915,
+      "grad_norm": 1.4763726780981958,
+      "learning_rate": 4.292366194295652e-05,
+      "logits/chosen": -10.25,
+      "logits/rejected": -9.5,
+      "logps/chosen": -612.0,
+      "logps/rejected": -620.0,
+      "loss": 0.5648,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -2.859375,
+      "rewards/margins": 0.58203125,
+      "rewards/rejected": -3.4375,
+      "step": 920
+    },
+    {
+      "epoch": 0.9733124018838305,
+      "grad_norm": 1.508415596628491,
+      "learning_rate": 4.2709953421038964e-05,
+      "logits/chosen": -9.1875,
+      "logits/rejected": -7.84375,
+      "logps/chosen": -676.0,
+      "logps/rejected": -672.0,
+      "loss": 0.5621,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -3.15625,
+      "rewards/margins": 0.64453125,
+      "rewards/rejected": -3.796875,
+      "step": 930
+    },
+    {
+      "epoch": 0.9837781266352695,
+      "grad_norm": 1.656384197278812,
+      "learning_rate": 4.249361495837478e-05,
+      "logits/chosen": -10.125,
+      "logits/rejected": -9.0,
+      "logps/chosen": -580.0,
+      "logps/rejected": -612.0,
+      "loss": 0.5309,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -2.6875,
+      "rewards/margins": 0.625,
+      "rewards/rejected": -3.3125,
+      "step": 940
+    },
+    {
+      "epoch": 0.9942438513867086,
+      "grad_norm": 1.8388851212373483,
+      "learning_rate": 4.227467868138035e-05,
+      "logits/chosen": -9.6875,
+      "logits/rejected": -8.875,
+      "logps/chosen": -628.0,
+      "logps/rejected": -672.0,
+      "loss": 0.5646,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -2.75,
+      "rewards/margins": 0.6875,
+      "rewards/rejected": -3.453125,
+      "step": 950
+    },
+    {
+      "epoch": 1.0047095761381475,
+      "grad_norm": 1.3610958015107766,
+      "learning_rate": 4.205317710224938e-05,
+      "logits/chosen": -10.0,
+      "logits/rejected": -8.9375,
+      "logps/chosen": -576.0,
+      "logps/rejected": -636.0,
+      "loss": 0.4379,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -2.546875,
+      "rewards/margins": 1.03125,
+      "rewards/rejected": -3.578125,
+      "step": 960
+    },
+    {
+      "epoch": 1.0151753008895865,
+      "grad_norm": 1.6553196811885342,
+      "learning_rate": 4.1829143114124733e-05,
+      "logits/chosen": -8.3125,
+      "logits/rejected": -6.03125,
+      "logps/chosen": -656.0,
+      "logps/rejected": -840.0,
+      "loss": 0.2537,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -2.921875,
+      "rewards/margins": 2.390625,
+      "rewards/rejected": -5.3125,
+      "step": 970
+    },
+    {
+      "epoch": 1.0256410256410255,
+      "grad_norm": 1.528713027426756,
+      "learning_rate": 4.160260998621387e-05,
+      "logits/chosen": -6.84375,
+      "logits/rejected": -3.65625,
+      "logps/chosen": -648.0,
+      "logps/rejected": -852.0,
+      "loss": 0.2268,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -3.4375,
+      "rewards/margins": 2.5625,
+      "rewards/rejected": -6.0,
+      "step": 980
+    },
+    {
+      "epoch": 1.0361067503924646,
+      "grad_norm": 2.0017510841126147,
+      "learning_rate": 4.137361135884824e-05,
+      "logits/chosen": -8.8125,
+      "logits/rejected": -5.78125,
+      "logps/chosen": -668.0,
+      "logps/rejected": -916.0,
+      "loss": 0.2104,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -3.265625,
+      "rewards/margins": 3.0625,
+      "rewards/rejected": -6.3125,
+      "step": 990
+    },
+    {
+      "epoch": 1.0465724751439036,
+      "grad_norm": 1.2577165774080123,
+      "learning_rate": 4.114218123848777e-05,
+      "logits/chosen": -9.75,
+      "logits/rejected": -7.1875,
+      "logps/chosen": -620.0,
+      "logps/rejected": -824.0,
+      "loss": 0.2284,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -2.671875,
+      "rewards/margins": 2.515625,
+      "rewards/rejected": -5.1875,
+      "step": 1000
+    },
+    {
+      "epoch": 1.0465724751439036,
+      "eval_logits/chosen": -8.8125,
+      "eval_logits/rejected": -6.8125,
+      "eval_logps/chosen": -656.0,
+      "eval_logps/rejected": -744.0,
+      "eval_loss": 0.5897031426429749,
+      "eval_rewards/accuracies": 0.73046875,
+      "eval_rewards/chosen": -3.375,
+      "eval_rewards/margins": 1.1796875,
+      "eval_rewards/rejected": -4.5625,
+      "eval_runtime": 46.7522,
+      "eval_samples_per_second": 42.779,
+      "eval_steps_per_second": 0.684,
+      "step": 1000
+    },
+    {
+      "epoch": 1.0570381998953426,
+      "grad_norm": 2.2723568730610686,
+      "learning_rate": 4.0908353992670824e-05,
+      "logits/chosen": -8.875,
+      "logits/rejected": -5.9375,
+      "logps/chosen": -592.0,
+      "logps/rejected": -792.0,
+      "loss": 0.2241,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -2.703125,
+      "rewards/margins": 2.65625,
+      "rewards/rejected": -5.34375,
+      "step": 1010
+    },
+    {
+      "epoch": 1.0675039246467817,
+      "grad_norm": 1.3993565848906937,
+      "learning_rate": 4.067216434491059e-05,
+      "logits/chosen": -8.6875,
+      "logits/rejected": -6.3125,
+      "logps/chosen": -676.0,
+      "logps/rejected": -920.0,
+      "loss": 0.1988,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -3.25,
+      "rewards/margins": 2.78125,
+      "rewards/rejected": -6.03125,
+      "step": 1020
+    },
+    {
+      "epoch": 1.077969649398221,
+      "grad_norm": 1.8501834739843528,
+      "learning_rate": 4.043364736953866e-05,
+      "logits/chosen": -9.8125,
+      "logits/rejected": -6.96875,
+      "logps/chosen": -648.0,
+      "logps/rejected": -900.0,
+      "loss": 0.192,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -3.203125,
+      "rewards/margins": 2.96875,
+      "rewards/rejected": -6.1875,
+      "step": 1030
+    },
+    {
+      "epoch": 1.08843537414966,
+      "grad_norm": 1.7012399768827051,
+      "learning_rate": 4.019283848649643e-05,
+      "logits/chosen": -9.5,
+      "logits/rejected": -6.21875,
+      "logps/chosen": -600.0,
+      "logps/rejected": -820.0,
+      "loss": 0.2425,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -2.9375,
+      "rewards/margins": 2.625,
+      "rewards/rejected": -5.5625,
+      "step": 1040
+    },
+    {
+      "epoch": 1.098901098901099,
+      "grad_norm": 1.464493641142556,
+      "learning_rate": 3.9949773456075215e-05,
+      "logits/chosen": -10.625,
+      "logits/rejected": -8.3125,
+      "logps/chosen": -608.0,
+      "logps/rejected": -856.0,
+      "loss": 0.1745,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -3.015625,
+      "rewards/margins": 2.875,
+      "rewards/rejected": -5.875,
+      "step": 1050
+    },
+    {
+      "epoch": 1.109366823652538,
+      "grad_norm": 2.341730327621805,
+      "learning_rate": 3.970448837360585e-05,
+      "logits/chosen": -9.875,
+      "logits/rejected": -7.125,
+      "logps/chosen": -696.0,
+      "logps/rejected": -932.0,
+      "loss": 0.2117,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -3.75,
+      "rewards/margins": 2.828125,
+      "rewards/rejected": -6.5625,
+      "step": 1060
+    },
+    {
+      "epoch": 1.119832548403977,
+      "grad_norm": 3.0007550014049187,
+      "learning_rate": 3.945701966409846e-05,
+      "logits/chosen": -10.9375,
+      "logits/rejected": -8.25,
+      "logps/chosen": -632.0,
+      "logps/rejected": -864.0,
+      "loss": 0.2126,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -3.0625,
+      "rewards/margins": 2.828125,
+      "rewards/rejected": -5.875,
+      "step": 1070
+    },
+    {
+      "epoch": 1.130298273155416,
+      "grad_norm": 2.3757813202535085,
+      "learning_rate": 3.9207404076833374e-05,
+      "logits/chosen": -10.375,
+      "logits/rejected": -8.5625,
+      "logps/chosen": -652.0,
+      "logps/rejected": -940.0,
+      "loss": 0.2015,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -3.484375,
+      "rewards/margins": 3.078125,
+      "rewards/rejected": -6.5625,
+      "step": 1080
+    },
+    {
+      "epoch": 1.1407639979068551,
+      "grad_norm": 1.965074857620583,
+      "learning_rate": 3.895567867990379e-05,
+      "logits/chosen": -8.8125,
+      "logits/rejected": -5.875,
+      "logps/chosen": -704.0,
+      "logps/rejected": -1012.0,
+      "loss": 0.187,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -3.84375,
+      "rewards/margins": 3.34375,
+      "rewards/rejected": -7.1875,
+      "step": 1090
+    },
+    {
+      "epoch": 1.1512297226582942,
+      "grad_norm": 2.1682834626567344,
+      "learning_rate": 3.870188085471114e-05,
+      "logits/chosen": -9.625,
+      "logits/rejected": -7.5,
+      "logps/chosen": -608.0,
+      "logps/rejected": -848.0,
+      "loss": 0.1919,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -2.75,
+      "rewards/margins": 2.796875,
+      "rewards/rejected": -5.5625,
+      "step": 1100
+    },
+    {
+      "epoch": 1.1512297226582942,
+      "eval_logits/chosen": -10.125,
+      "eval_logits/rejected": -8.375,
+      "eval_logps/chosen": -696.0,
+      "eval_logps/rejected": -784.0,
+      "eval_loss": 0.5994374752044678,
+      "eval_rewards/accuracies": 0.7265625,
+      "eval_rewards/chosen": -3.765625,
+      "eval_rewards/margins": 1.1796875,
+      "eval_rewards/rejected": -4.9375,
+      "eval_runtime": 46.7605,
+      "eval_samples_per_second": 42.771,
+      "eval_steps_per_second": 0.684,
+      "step": 1100
+    },
+    {
+      "epoch": 1.1616954474097332,
+      "grad_norm": 1.755722577601338,
+      "learning_rate": 3.844604829041395e-05,
+      "logits/chosen": -8.9375,
+      "logits/rejected": -5.9375,
+      "logps/chosen": -696.0,
+      "logps/rejected": -980.0,
+      "loss": 0.144,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -3.78125,
+      "rewards/margins": 3.265625,
+      "rewards/rejected": -7.0625,
+      "step": 1110
+    },
+    {
+      "epoch": 1.1721611721611722,
+      "grad_norm": 2.5646881572830775,
+      "learning_rate": 3.818821897833095e-05,
+      "logits/chosen": -8.75,
+      "logits/rejected": -6.15625,
+      "logps/chosen": -724.0,
+      "logps/rejected": -1004.0,
+      "loss": 0.2213,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -3.9375,
+      "rewards/margins": 3.140625,
+      "rewards/rejected": -7.09375,
+      "step": 1120
+    },
+    {
+      "epoch": 1.1826268969126112,
+      "grad_norm": 1.2353811146301794,
+      "learning_rate": 3.792843120629935e-05,
+      "logits/chosen": -10.125,
+      "logits/rejected": -7.75,
+      "logps/chosen": -692.0,
+      "logps/rejected": -964.0,
+      "loss": 0.1566,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -3.5625,
+      "rewards/margins": 3.0,
+      "rewards/rejected": -6.5625,
+      "step": 1130
+    },
+    {
+      "epoch": 1.1930926216640503,
+      "grad_norm": 1.5891026452889025,
+      "learning_rate": 3.766672355298907e-05,
+      "logits/chosen": -10.25,
+      "logits/rejected": -8.125,
+      "logps/chosen": -736.0,
+      "logps/rejected": -1032.0,
+      "loss": 0.1813,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -3.765625,
+      "rewards/margins": 3.46875,
+      "rewards/rejected": -7.25,
+      "step": 1140
+    },
+    {
+      "epoch": 1.2035583464154893,
+      "grad_norm": 2.2205622689643354,
+      "learning_rate": 3.740313488217373e-05,
+      "logits/chosen": -10.125,
+      "logits/rejected": -8.25,
+      "logps/chosen": -684.0,
+      "logps/rejected": -1004.0,
+      "loss": 0.1735,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -3.65625,
+      "rewards/margins": 3.296875,
+      "rewards/rejected": -6.96875,
+      "step": 1150
+    },
+    {
+      "epoch": 1.2140240711669283,
+      "grad_norm": 2.5614191389346765,
+      "learning_rate": 3.7137704336959464e-05,
+      "logits/chosen": -9.875,
+      "logits/rejected": -7.03125,
+      "logps/chosen": -704.0,
+      "logps/rejected": -980.0,
+      "loss": 0.1873,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -3.890625,
+      "rewards/margins": 3.203125,
+      "rewards/rejected": -7.09375,
+      "step": 1160
+    },
+    {
+      "epoch": 1.2244897959183674,
+      "grad_norm": 1.297362868487626,
+      "learning_rate": 3.687047133397201e-05,
+      "logits/chosen": -9.6875,
+      "logits/rejected": -7.5,
+      "logps/chosen": -676.0,
+      "logps/rejected": -936.0,
+      "loss": 0.1813,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -3.640625,
+      "rewards/margins": 3.046875,
+      "rewards/rejected": -6.6875,
+      "step": 1170
+    },
+    {
+      "epoch": 1.2349555206698064,
+      "grad_norm": 2.513321845058118,
+      "learning_rate": 3.6601475557503415e-05,
+      "logits/chosen": -9.8125,
+      "logits/rejected": -7.71875,
+      "logps/chosen": -684.0,
+      "logps/rejected": -976.0,
+      "loss": 0.201,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -3.640625,
+      "rewards/margins": 3.328125,
+      "rewards/rejected": -6.96875,
+      "step": 1180
+    },
+    {
+      "epoch": 1.2454212454212454,
+      "grad_norm": 2.7327830816714074,
+      "learning_rate": 3.633075695361882e-05,
+      "logits/chosen": -7.9375,
+      "logits/rejected": -5.28125,
+      "logps/chosen": -752.0,
+      "logps/rejected": -984.0,
+      "loss": 0.1784,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -4.15625,
+      "rewards/margins": 2.859375,
+      "rewards/rejected": -7.03125,
+      "step": 1190
+    },
+    {
+      "epoch": 1.2558869701726845,
+      "grad_norm": 2.909785770206169,
+      "learning_rate": 3.6058355724224477e-05,
+      "logits/chosen": -6.4375,
+      "logits/rejected": -3.734375,
+      "logps/chosen": -752.0,
+      "logps/rejected": -996.0,
+      "loss": 0.1942,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -3.765625,
+      "rewards/margins": 2.9375,
+      "rewards/rejected": -6.6875,
+      "step": 1200
+    },
+    {
+      "epoch": 1.2558869701726845,
+      "eval_logits/chosen": -5.5,
+      "eval_logits/rejected": -3.546875,
+      "eval_logps/chosen": -768.0,
+      "eval_logps/rejected": -856.0,
+      "eval_loss": 0.605843722820282,
+      "eval_rewards/accuracies": 0.71875,
+      "eval_rewards/chosen": -4.5,
+      "eval_rewards/margins": 1.171875,
+      "eval_rewards/rejected": -5.65625,
+      "eval_runtime": 46.7815,
+      "eval_samples_per_second": 42.752,
+      "eval_steps_per_second": 0.684,
+      "step": 1200
+    },
+    {
+      "epoch": 1.2663526949241235,
+      "grad_norm": 2.1118168892270743,
+      "learning_rate": 3.5784312321097744e-05,
+      "logits/chosen": -6.65625,
+      "logits/rejected": -3.203125,
+      "logps/chosen": -780.0,
+      "logps/rejected": -1004.0,
+      "loss": 0.1922,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -3.890625,
+      "rewards/margins": 3.140625,
+      "rewards/rejected": -7.03125,
+      "step": 1210
+    },
+    {
+      "epoch": 1.2768184196755625,
+      "grad_norm": 2.1458553218502736,
+      "learning_rate": 3.550866743987994e-05,
+      "logits/chosen": -6.09375,
+      "logits/rejected": -4.0,
+      "logps/chosen": -768.0,
+      "logps/rejected": -996.0,
+      "loss": 0.2335,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -4.375,
+      "rewards/margins": 2.828125,
+      "rewards/rejected": -7.21875,
+      "step": 1220
+    },
+    {
+      "epoch": 1.2872841444270016,
+      "grad_norm": 1.9704271854351658,
+      "learning_rate": 3.5231462014033024e-05,
+      "logits/chosen": -7.625,
+      "logits/rejected": -5.625,
+      "logps/chosen": -668.0,
+      "logps/rejected": -932.0,
+      "loss": 0.191,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -3.671875,
+      "rewards/margins": 3.0625,
+      "rewards/rejected": -6.75,
+      "step": 1230
+    },
+    {
+      "epoch": 1.2977498691784406,
+      "grad_norm": 2.540989435589858,
+      "learning_rate": 3.495273720876095e-05,
+      "logits/chosen": -8.6875,
+      "logits/rejected": -6.46875,
+      "logps/chosen": -680.0,
+      "logps/rejected": -952.0,
+      "loss": 0.1834,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -3.765625,
+      "rewards/margins": 3.0,
+      "rewards/rejected": -6.78125,
+      "step": 1240
+    },
+    {
+      "epoch": 1.3082155939298796,
+      "grad_norm": 2.9945900507912007,
+      "learning_rate": 3.4672534414896616e-05,
+      "logits/chosen": -9.375,
+      "logits/rejected": -6.71875,
+      "logps/chosen": -700.0,
+      "logps/rejected": -964.0,
+      "loss": 0.182,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -3.59375,
+      "rewards/margins": 3.265625,
+      "rewards/rejected": -6.84375,
+      "step": 1250
+    },
+    {
+      "epoch": 1.3186813186813187,
+      "grad_norm": 2.32202187025829,
+      "learning_rate": 3.439089524275526e-05,
+      "logits/chosen": -8.3125,
+      "logits/rejected": -6.3125,
+      "logps/chosen": -636.0,
+      "logps/rejected": -916.0,
+      "loss": 0.1788,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -3.328125,
+      "rewards/margins": 3.0,
+      "rewards/rejected": -6.34375,
+      "step": 1260
+    },
+    {
+      "epoch": 1.3291470434327577,
+      "grad_norm": 2.3372620510687323,
+      "learning_rate": 3.410786151595533e-05,
+      "logits/chosen": -7.875,
+      "logits/rejected": -5.4375,
+      "logps/chosen": -680.0,
+      "logps/rejected": -948.0,
+      "loss": 0.2101,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -3.640625,
+      "rewards/margins": 3.03125,
+      "rewards/rejected": -6.65625,
+      "step": 1270
+    },
+    {
+      "epoch": 1.3396127681841967,
+      "grad_norm": 1.8335286313017747,
+      "learning_rate": 3.382347526520765e-05,
+      "logits/chosen": -7.0625,
+      "logits/rejected": -3.78125,
+      "logps/chosen": -708.0,
+      "logps/rejected": -984.0,
+      "loss": 0.1892,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -3.828125,
+      "rewards/margins": 3.296875,
+      "rewards/rejected": -7.125,
+      "step": 1280
+    },
+    {
+      "epoch": 1.3500784929356358,
+      "grad_norm": 2.4415333840589812,
+      "learning_rate": 3.353777872207381e-05,
+      "logits/chosen": -6.125,
+      "logits/rejected": -3.734375,
+      "logps/chosen": -708.0,
+      "logps/rejected": -984.0,
+      "loss": 0.1928,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -3.75,
+      "rewards/margins": 3.1875,
+      "rewards/rejected": -6.9375,
+      "step": 1290
+    },
+    {
+      "epoch": 1.3605442176870748,
+      "grad_norm": 1.8208450562001965,
+      "learning_rate": 3.325081431269475e-05,
+      "logits/chosen": -7.46875,
+      "logits/rejected": -5.5,
+      "logps/chosen": -684.0,
+      "logps/rejected": -948.0,
+      "loss": 0.2071,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -3.640625,
+      "rewards/margins": 2.796875,
+      "rewards/rejected": -6.4375,
+      "step": 1300
+    },
+    {
+      "epoch": 1.3605442176870748,
+      "eval_logits/chosen": -7.78125,
+      "eval_logits/rejected": -6.1875,
+      "eval_logps/chosen": -752.0,
+      "eval_logps/rejected": -836.0,
+      "eval_loss": 0.5985156297683716,
+      "eval_rewards/accuracies": 0.744140625,
+      "eval_rewards/chosen": -4.3125,
+      "eval_rewards/margins": 1.1484375,
+      "eval_rewards/rejected": -5.46875,
+      "eval_runtime": 46.7523,
+      "eval_samples_per_second": 42.779,
+      "eval_steps_per_second": 0.684,
+      "step": 1300
+    },
+    {
+      "epoch": 1.3710099424385138,
+      "grad_norm": 2.119386417844308,
+      "learning_rate": 3.296262465149046e-05,
+      "logits/chosen": -8.6875,
+      "logits/rejected": -5.96875,
+      "logps/chosen": -704.0,
+      "logps/rejected": -944.0,
+      "loss": 0.1811,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -3.703125,
+      "rewards/margins": 2.953125,
+      "rewards/rejected": -6.65625,
+      "step": 1310
+    },
+    {
+      "epoch": 1.3814756671899528,
+      "grad_norm": 3.714855719416775,
+      "learning_rate": 3.2673252534831687e-05,
+      "logits/chosen": -7.5,
+      "logits/rejected": -5.21875,
+      "logps/chosen": -680.0,
+      "logps/rejected": -984.0,
+      "loss": 0.1674,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -3.90625,
+      "rewards/margins": 3.265625,
+      "rewards/rejected": -7.15625,
+      "step": 1320
+    },
+    {
+      "epoch": 1.3919413919413919,
+      "grad_norm": 3.1256151131181107,
+      "learning_rate": 3.2382740934684696e-05,
+      "logits/chosen": -7.09375,
+      "logits/rejected": -5.46875,
+      "logps/chosen": -736.0,
+      "logps/rejected": -1064.0,
+      "loss": 0.1975,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -4.15625,
+      "rewards/margins": 3.1875,
+      "rewards/rejected": -7.34375,
+      "step": 1330
+    },
+    {
+      "epoch": 1.402407116692831,
+      "grad_norm": 1.825499015678452,
+      "learning_rate": 3.209113299222982e-05,
+      "logits/chosen": -8.25,
+      "logits/rejected": -5.5,
+      "logps/chosen": -728.0,
+      "logps/rejected": -984.0,
+      "loss": 0.1741,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -3.875,
+      "rewards/margins": 3.140625,
+      "rewards/rejected": -7.03125,
+      "step": 1340
+    },
+    {
+      "epoch": 1.41287284144427,
+      "grad_norm": 2.1812895355643556,
+      "learning_rate": 3.179847201145505e-05,
+      "logits/chosen": -7.90625,
+      "logits/rejected": -5.53125,
+      "logps/chosen": -704.0,
+      "logps/rejected": -1008.0,
+      "loss": 0.1823,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -3.875,
+      "rewards/margins": 3.3125,
+      "rewards/rejected": -7.1875,
+      "step": 1350
+    },
+    {
+      "epoch": 1.423338566195709,
+      "grad_norm": 2.0963037119534986,
+      "learning_rate": 3.150480145272528e-05,
+      "logits/chosen": -8.1875,
+      "logits/rejected": -6.125,
+      "logps/chosen": -704.0,
+      "logps/rejected": -968.0,
+      "loss": 0.1921,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -3.796875,
+      "rewards/margins": 3.125,
+      "rewards/rejected": -6.9375,
+      "step": 1360
+    },
+    {
+      "epoch": 1.433804290947148,
+      "grad_norm": 4.40648517169844,
+      "learning_rate": 3.121016492632848e-05,
+      "logits/chosen": -6.90625,
+      "logits/rejected": -4.90625,
+      "logps/chosen": -776.0,
+      "logps/rejected": -1048.0,
+      "loss": 0.2238,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -4.65625,
+      "rewards/margins": 3.046875,
+      "rewards/rejected": -7.71875,
+      "step": 1370
+    },
+    {
+      "epoch": 1.4442700156985873,
+      "grad_norm": 2.0271092896630827,
+      "learning_rate": 3.091460618599951e-05,
+      "logits/chosen": -7.4375,
+      "logits/rejected": -5.0,
+      "logps/chosen": -772.0,
+      "logps/rejected": -1048.0,
+      "loss": 0.1889,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -4.5,
+      "rewards/margins": 3.265625,
+      "rewards/rejected": -7.75,
+      "step": 1380
+    },
+    {
+      "epoch": 1.454735740450026,
+      "grad_norm": 2.5666330672896627,
+      "learning_rate": 3.061816912242265e-05,
+      "logits/chosen": -8.625,
+      "logits/rejected": -6.28125,
+      "logps/chosen": -708.0,
+      "logps/rejected": -960.0,
+      "loss": 0.2217,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -3.78125,
+      "rewards/margins": 3.171875,
+      "rewards/rejected": -6.96875,
+      "step": 1390
+    },
+    {
+      "epoch": 1.4652014652014653,
+      "grad_norm": 2.17684267833633,
+      "learning_rate": 3.0320897756713785e-05,
+      "logits/chosen": -8.75,
+      "logits/rejected": -7.59375,
+      "logps/chosen": -772.0,
+      "logps/rejected": -1056.0,
+      "loss": 0.1811,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -4.28125,
+      "rewards/margins": 2.9375,
+      "rewards/rejected": -7.21875,
+      "step": 1400
+    },
+    {
+      "epoch": 1.4652014652014653,
+      "eval_logits/chosen": -8.0,
+      "eval_logits/rejected": -6.65625,
+      "eval_logps/chosen": -856.0,
+      "eval_logps/rejected": -948.0,
+      "eval_loss": 0.6044531464576721,
+      "eval_rewards/accuracies": 0.736328125,
+      "eval_rewards/chosen": -5.375,
+      "eval_rewards/margins": 1.2109375,
+      "eval_rewards/rejected": -6.5625,
+      "eval_runtime": 47.603,
+      "eval_samples_per_second": 42.014,
+      "eval_steps_per_second": 0.672,
+      "step": 1400
+    },
+    {
+      "epoch": 1.4756671899529041,
+      "grad_norm": 2.2780141068547306,
+      "learning_rate": 3.002283623388332e-05,
+      "logits/chosen": -8.25,
+      "logits/rejected": -5.875,
+      "logps/chosen": -764.0,
+      "logps/rejected": -1080.0,
+      "loss": 0.185,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -4.53125,
+      "rewards/margins": 3.375,
+      "rewards/rejected": -7.90625,
+      "step": 1410
+    },
+    {
+      "epoch": 1.4861329147043434,
+      "grad_norm": 1.8398526580527912,
+      "learning_rate": 2.9724028816280507e-05,
+      "logits/chosen": -7.46875,
+      "logits/rejected": -5.75,
+      "logps/chosen": -784.0,
+      "logps/rejected": -1048.0,
+      "loss": 0.1828,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -4.40625,
+      "rewards/margins": 3.109375,
+      "rewards/rejected": -7.5,
+      "step": 1420
+    },
+    {
+      "epoch": 1.4965986394557822,
+      "grad_norm": 2.456647708070056,
+      "learning_rate": 2.9424519877020524e-05,
+      "logits/chosen": -8.625,
+      "logits/rejected": -5.9375,
+      "logps/chosen": -776.0,
+      "logps/rejected": -1104.0,
+      "loss": 0.171,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -4.3125,
+      "rewards/margins": 3.65625,
+      "rewards/rejected": -7.96875,
+      "step": 1430
+    },
+    {
+      "epoch": 1.5070643642072215,
+      "grad_norm": 1.352002355713728,
+      "learning_rate": 2.912435389339504e-05,
+      "logits/chosen": -8.375,
+      "logits/rejected": -5.6875,
+      "logps/chosen": -772.0,
+      "logps/rejected": -1088.0,
+      "loss": 0.1832,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -4.28125,
+      "rewards/margins": 3.5625,
+      "rewards/rejected": -7.84375,
+      "step": 1440
+    },
+    {
+      "epoch": 1.5175300889586603,
+      "grad_norm": 2.448099861439372,
+      "learning_rate": 2.8823575440267226e-05,
+      "logits/chosen": -7.59375,
+      "logits/rejected": -5.59375,
+      "logps/chosen": -872.0,
+      "logps/rejected": -1160.0,
+      "loss": 0.1629,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -5.0,
+      "rewards/margins": 3.484375,
+      "rewards/rejected": -8.5,
+      "step": 1450
+    },
+    {
+      "epoch": 1.5279958137100995,
+      "grad_norm": 2.7978328404357438,
+      "learning_rate": 2.8522229183452465e-05,
+      "logits/chosen": -7.59375,
+      "logits/rejected": -5.25,
+      "logps/chosen": -792.0,
+      "logps/rejected": -1120.0,
+      "loss": 0.15,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -4.65625,
+      "rewards/margins": 3.390625,
+      "rewards/rejected": -8.0625,
+      "step": 1460
+    },
+    {
+      "epoch": 1.5384615384615383,
+      "grad_norm": 2.444939431372612,
+      "learning_rate": 2.8220359873085373e-05,
+      "logits/chosen": -8.8125,
+      "logits/rejected": -6.59375,
+      "logps/chosen": -760.0,
+      "logps/rejected": -1128.0,
+      "loss": 0.1819,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -4.0,
+      "rewards/margins": 3.859375,
+      "rewards/rejected": -7.875,
+      "step": 1470
+    },
+    {
+      "epoch": 1.5489272632129776,
+      "grad_norm": 2.097638268565495,
+      "learning_rate": 2.7918012336974388e-05,
+      "logits/chosen": -8.0,
+      "logits/rejected": -4.6875,
+      "logps/chosen": -720.0,
+      "logps/rejected": -984.0,
+      "loss": 0.2026,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -4.1875,
+      "rewards/margins": 3.25,
+      "rewards/rejected": -7.4375,
+      "step": 1480
+    },
+    {
+      "epoch": 1.5593929879644164,
+      "grad_norm": 1.8995097464923203,
+      "learning_rate": 2.7615231473944815e-05,
+      "logits/chosen": -8.875,
+      "logits/rejected": -6.3125,
+      "logps/chosen": -744.0,
+      "logps/rejected": -996.0,
+      "loss": 0.1759,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -4.03125,
+      "rewards/margins": 3.0,
+      "rewards/rejected": -7.03125,
+      "step": 1490
+    },
+    {
+      "epoch": 1.5698587127158556,
+      "grad_norm": 1.7416531968989217,
+      "learning_rate": 2.7312062247171327e-05,
+      "logits/chosen": -9.125,
+      "logits/rejected": -6.9375,
+      "logps/chosen": -732.0,
+      "logps/rejected": -1048.0,
+      "loss": 0.1715,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -3.921875,
+      "rewards/margins": 3.328125,
+      "rewards/rejected": -7.25,
+      "step": 1500
+    },
+    {
+      "epoch": 1.5698587127158556,
+      "eval_logits/chosen": -8.6875,
+      "eval_logits/rejected": -7.1875,
+      "eval_logps/chosen": -792.0,
+      "eval_logps/rejected": -892.0,
+      "eval_loss": 0.6053906083106995,
+      "eval_rewards/accuracies": 0.73828125,
+      "eval_rewards/chosen": -4.71875,
+      "eval_rewards/margins": 1.3046875,
+      "eval_rewards/rejected": -6.03125,
+      "eval_runtime": 46.7517,
+      "eval_samples_per_second": 42.779,
+      "eval_steps_per_second": 0.684,
+      "step": 1500
+    },
+    {
+      "epoch": 1.5803244374672945,
+      "grad_norm": 2.1508965822891044,
+      "learning_rate": 2.7008549677500876e-05,
+      "logits/chosen": -8.625,
+      "logits/rejected": -6.5,
+      "logps/chosen": -744.0,
+      "logps/rejected": -1040.0,
+      "loss": 0.1773,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -3.9375,
+      "rewards/margins": 3.3125,
+      "rewards/rejected": -7.25,
+      "step": 1510
+    },
+    {
+      "epoch": 1.5907901622187337,
+      "grad_norm": 2.8701318459101106,
+      "learning_rate": 2.670473883676709e-05,
+      "logits/chosen": -8.9375,
+      "logits/rejected": -6.625,
+      "logps/chosen": -764.0,
+      "logps/rejected": -1048.0,
+      "loss": 0.1983,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -4.25,
+      "rewards/margins": 3.21875,
+      "rewards/rejected": -7.46875,
+      "step": 1520
+    },
+    {
+      "epoch": 1.6012558869701727,
+      "grad_norm": 1.0768288261890753,
+      "learning_rate": 2.6400674841097074e-05,
+      "logits/chosen": -8.9375,
+      "logits/rejected": -6.3125,
+      "logps/chosen": -744.0,
+      "logps/rejected": -996.0,
+      "loss": 0.147,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -4.09375,
+      "rewards/margins": 3.265625,
+      "rewards/rejected": -7.375,
+      "step": 1530
+    },
+    {
+      "epoch": 1.6117216117216118,
+      "grad_norm": 1.6991251754173282,
+      "learning_rate": 2.609640284421158e-05,
+      "logits/chosen": -9.625,
+      "logits/rejected": -7.3125,
+      "logps/chosen": -748.0,
+      "logps/rejected": -1008.0,
+      "loss": 0.1853,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -3.96875,
+      "rewards/margins": 3.1875,
+      "rewards/rejected": -7.15625,
+      "step": 1540
+    },
+    {
+      "epoch": 1.6221873364730508,
+      "grad_norm": 2.147434475827148,
+      "learning_rate": 2.5791968030719726e-05,
+      "logits/chosen": -9.5625,
+      "logits/rejected": -7.3125,
+      "logps/chosen": -748.0,
+      "logps/rejected": -1072.0,
+      "loss": 0.2114,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -3.953125,
+      "rewards/margins": 3.453125,
+      "rewards/rejected": -7.40625,
+      "step": 1550
+    },
+    {
+      "epoch": 1.6326530612244898,
+      "grad_norm": 1.9299308583721462,
+      "learning_rate": 2.548741560940902e-05,
+      "logits/chosen": -9.5625,
+      "logits/rejected": -7.28125,
+      "logps/chosen": -720.0,
+      "logps/rejected": -1024.0,
+      "loss": 0.1987,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -3.78125,
+      "rewards/margins": 3.53125,
+      "rewards/rejected": -7.3125,
+      "step": 1560
+    },
+    {
+      "epoch": 1.6431187859759289,
+      "grad_norm": 1.855284289246685,
+      "learning_rate": 2.5182790806531787e-05,
+      "logits/chosen": -9.125,
+      "logits/rejected": -6.84375,
+      "logps/chosen": -700.0,
+      "logps/rejected": -980.0,
+      "loss": 0.1612,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -3.921875,
+      "rewards/margins": 3.234375,
+      "rewards/rejected": -7.15625,
+      "step": 1570
+    },
+    {
+      "epoch": 1.653584510727368,
+      "grad_norm": 1.429633474316079,
+      "learning_rate": 2.4878138859089074e-05,
+      "logits/chosen": -8.8125,
+      "logits/rejected": -6.0625,
+      "logps/chosen": -748.0,
+      "logps/rejected": -1040.0,
+      "loss": 0.1676,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -4.1875,
+      "rewards/margins": 3.5,
+      "rewards/rejected": -7.6875,
+      "step": 1580
+    },
+    {
+      "epoch": 1.664050235478807,
+      "grad_norm": 2.545821523816558,
+      "learning_rate": 2.4573505008112924e-05,
+      "logits/chosen": -8.875,
+      "logits/rejected": -5.9375,
+      "logps/chosen": -780.0,
+      "logps/rejected": -1096.0,
+      "loss": 0.1715,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -4.5625,
+      "rewards/margins": 3.5625,
+      "rewards/rejected": -8.125,
+      "step": 1590
+    },
+    {
+      "epoch": 1.674515960230246,
+      "grad_norm": 2.5270498584967616,
+      "learning_rate": 2.4268934491948032e-05,
+      "logits/chosen": -9.4375,
+      "logits/rejected": -7.15625,
+      "logps/chosen": -756.0,
+      "logps/rejected": -1024.0,
+      "loss": 0.186,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -4.125,
+      "rewards/margins": 3.28125,
+      "rewards/rejected": -7.40625,
+      "step": 1600
+    },
+    {
+      "epoch": 1.674515960230246,
+      "eval_logits/chosen": -9.6875,
+      "eval_logits/rejected": -8.3125,
+      "eval_logps/chosen": -768.0,
+      "eval_logps/rejected": -860.0,
+      "eval_loss": 0.6277499794960022,
+      "eval_rewards/accuracies": 0.728515625,
+      "eval_rewards/chosen": -4.46875,
+      "eval_rewards/margins": 1.234375,
+      "eval_rewards/rejected": -5.71875,
+      "eval_runtime": 46.7661,
+      "eval_samples_per_second": 42.766,
+      "eval_steps_per_second": 0.684,
+      "step": 1600
+    },
+    {
+      "epoch": 1.684981684981685,
+      "grad_norm": 1.6933421692029196,
+      "learning_rate": 2.396447253953385e-05,
+      "logits/chosen": -10.125,
+      "logits/rejected": -8.0,
+      "logps/chosen": -724.0,
+      "logps/rejected": -1016.0,
+      "loss": 0.197,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -3.609375,
+      "rewards/margins": 3.3125,
+      "rewards/rejected": -6.90625,
+      "step": 1610
+    },
+    {
+      "epoch": 1.695447409733124,
+      "grad_norm": 1.7688994422254782,
+      "learning_rate": 2.3660164363687998e-05,
+      "logits/chosen": -9.5,
+      "logits/rejected": -7.03125,
+      "logps/chosen": -680.0,
+      "logps/rejected": -944.0,
+      "loss": 0.1887,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -3.546875,
+      "rewards/margins": 3.15625,
+      "rewards/rejected": -6.71875,
+      "step": 1620
+    },
+    {
+      "epoch": 1.705913134484563,
+      "grad_norm": 2.4558518767011948,
+      "learning_rate": 2.33560551543922e-05,
+      "logits/chosen": -8.625,
+      "logits/rejected": -6.78125,
+      "logps/chosen": -700.0,
+      "logps/rejected": -992.0,
+      "loss": 0.164,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -4.0,
+      "rewards/margins": 3.140625,
+      "rewards/rejected": -7.125,
+      "step": 1630
+    },
+    {
+      "epoch": 1.716378859236002,
+      "grad_norm": 2.980911099905336,
+      "learning_rate": 2.3052190072081493e-05,
+      "logits/chosen": -8.25,
+      "logits/rejected": -6.34375,
+      "logps/chosen": -816.0,
+      "logps/rejected": -1176.0,
+      "loss": 0.1775,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -4.53125,
+      "rewards/margins": 3.8125,
+      "rewards/rejected": -8.375,
+      "step": 1640
+    },
+    {
+      "epoch": 1.7268445839874411,
+      "grad_norm": 2.3029692498244594,
+      "learning_rate": 2.2748614240937864e-05,
+      "logits/chosen": -8.25,
+      "logits/rejected": -6.15625,
+      "logps/chosen": -772.0,
+      "logps/rejected": -1088.0,
+      "loss": 0.158,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -4.34375,
+      "rewards/margins": 3.59375,
+      "rewards/rejected": -7.9375,
+      "step": 1650
+    },
+    {
+      "epoch": 1.7373103087388801,
+      "grad_norm": 2.479814689365366,
+      "learning_rate": 2.2445372742189334e-05,
+      "logits/chosen": -7.0625,
+      "logits/rejected": -5.0625,
+      "logps/chosen": -768.0,
+      "logps/rejected": -1104.0,
+      "loss": 0.172,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -4.5,
+      "rewards/margins": 3.65625,
+      "rewards/rejected": -8.125,
+      "step": 1660
+    },
+    {
+      "epoch": 1.7477760334903192,
+      "grad_norm": 1.9106632863826012,
+      "learning_rate": 2.214251060741528e-05,
+      "logits/chosen": -7.1875,
+      "logits/rejected": -3.828125,
+      "logps/chosen": -808.0,
+      "logps/rejected": -1128.0,
+      "loss": 0.1515,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -4.75,
+      "rewards/margins": 3.65625,
+      "rewards/rejected": -8.375,
+      "step": 1670
+    },
+    {
+      "epoch": 1.7582417582417582,
+      "grad_norm": 2.1491643492294066,
+      "learning_rate": 2.1840072811859302e-05,
+      "logits/chosen": -6.625,
+      "logits/rejected": -4.03125,
+      "logps/chosen": -744.0,
+      "logps/rejected": -1048.0,
+      "loss": 0.1892,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -4.5625,
+      "rewards/margins": 3.328125,
+      "rewards/rejected": -7.90625,
+      "step": 1680
+    },
+    {
+      "epoch": 1.7687074829931972,
+      "grad_norm": 2.3110621431882623,
+      "learning_rate": 2.1538104267750287e-05,
+      "logits/chosen": -8.625,
+      "logits/rejected": -5.21875,
+      "logps/chosen": -796.0,
+      "logps/rejected": -1048.0,
+      "loss": 0.1622,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -4.4375,
+      "rewards/margins": 3.28125,
+      "rewards/rejected": -7.75,
+      "step": 1690
+    },
+    {
+      "epoch": 1.7791732077446363,
+      "grad_norm": 3.227877653343518,
+      "learning_rate": 2.1236649817632952e-05,
+      "logits/chosen": -7.4375,
+      "logits/rejected": -4.90625,
+      "logps/chosen": -768.0,
+      "logps/rejected": -1080.0,
+      "loss": 0.1763,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -4.5,
+      "rewards/margins": 3.390625,
+      "rewards/rejected": -7.90625,
+      "step": 1700
+    },
+    {
+      "epoch": 1.7791732077446363,
+      "eval_logits/chosen": -7.4375,
+      "eval_logits/rejected": -5.53125,
+      "eval_logps/chosen": -840.0,
+      "eval_logps/rejected": -952.0,
+      "eval_loss": 0.6385781168937683,
+      "eval_rewards/accuracies": 0.724609375,
+      "eval_rewards/chosen": -5.21875,
+      "eval_rewards/margins": 1.40625,
+      "eval_rewards/rejected": -6.625,
+      "eval_runtime": 46.7561,
+      "eval_samples_per_second": 42.775,
+      "eval_steps_per_second": 0.684,
+      "step": 1700
+    },
+    {
+      "epoch": 1.7896389324960753,
+      "grad_norm": 2.324602342413771,
+      "learning_rate": 2.093575422770872e-05,
+      "logits/chosen": -8.0625,
+      "logits/rejected": -5.1875,
+      "logps/chosen": -808.0,
+      "logps/rejected": -1056.0,
+      "loss": 0.2185,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -4.40625,
+      "rewards/margins": 3.265625,
+      "rewards/rejected": -7.6875,
+      "step": 1710
+    },
+    {
+      "epoch": 1.8001046572475143,
+      "grad_norm": 1.7703073301586785,
+      "learning_rate": 2.0635462181187827e-05,
+      "logits/chosen": -8.625,
+      "logits/rejected": -6.78125,
+      "logps/chosen": -740.0,
+      "logps/rejected": -1000.0,
+      "loss": 0.1939,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -4.0,
+      "rewards/margins": 3.046875,
+      "rewards/rejected": -7.0625,
+      "step": 1720
+    },
+    {
+      "epoch": 1.8105703819989536,
+      "grad_norm": 1.8750882746661335,
+      "learning_rate": 2.0335818271653917e-05,
+      "logits/chosen": -8.625,
+      "logits/rejected": -6.1875,
+      "logps/chosen": -740.0,
+      "logps/rejected": -1004.0,
+      "loss": 0.1626,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -4.03125,
+      "rewards/margins": 3.125,
+      "rewards/rejected": -7.15625,
+      "step": 1730
+    },
+    {
+      "epoch": 1.8210361067503924,
+      "grad_norm": 3.202095266498565,
+      "learning_rate": 2.0036866996441815e-05,
+      "logits/chosen": -7.0,
+      "logits/rejected": -4.03125,
+      "logps/chosen": -840.0,
+      "logps/rejected": -1120.0,
+      "loss": 0.2127,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -4.875,
+      "rewards/margins": 3.40625,
+      "rewards/rejected": -8.3125,
+      "step": 1740
+    },
+    {
+      "epoch": 1.8315018315018317,
+      "grad_norm": 3.4082077672389093,
+      "learning_rate": 1.9738652750029677e-05,
+      "logits/chosen": -7.6875,
+      "logits/rejected": -4.40625,
+      "logps/chosen": -812.0,
+      "logps/rejected": -1056.0,
+      "loss": 0.1718,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -4.65625,
+      "rewards/margins": 3.15625,
+      "rewards/rejected": -7.8125,
+      "step": 1750
+    },
+    {
+      "epoch": 1.8419675562532705,
+      "grad_norm": 2.136036556319782,
+      "learning_rate": 1.944121981744631e-05,
+      "logits/chosen": -7.65625,
+      "logits/rejected": -5.78125,
+      "logps/chosen": -792.0,
+      "logps/rejected": -1080.0,
+      "loss": 0.1934,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -4.8125,
+      "rewards/margins": 3.125,
+      "rewards/rejected": -7.90625,
+      "step": 1760
+    },
+    {
+      "epoch": 1.8524332810047097,
+      "grad_norm": 2.1783109645897847,
+      "learning_rate": 1.9144612367694887e-05,
+      "logits/chosen": -8.125,
+      "logits/rejected": -6.125,
+      "logps/chosen": -768.0,
+      "logps/rejected": -1056.0,
+      "loss": 0.1604,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -4.125,
+      "rewards/margins": 3.484375,
+      "rewards/rejected": -7.625,
+      "step": 1770
+    },
+    {
+      "epoch": 1.8628990057561485,
+      "grad_norm": 2.3455579838846083,
+      "learning_rate": 1.8848874447193803e-05,
+      "logits/chosen": -8.625,
+      "logits/rejected": -6.21875,
+      "logps/chosen": -788.0,
+      "logps/rejected": -1104.0,
+      "loss": 0.176,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -4.1875,
+      "rewards/margins": 3.53125,
+      "rewards/rejected": -7.71875,
+      "step": 1780
+    },
+    {
+      "epoch": 1.8733647305075878,
+      "grad_norm": 2.8425668375108333,
+      "learning_rate": 1.855404997323571e-05,
+      "logits/chosen": -8.375,
+      "logits/rejected": -5.90625,
+      "logps/chosen": -728.0,
+      "logps/rejected": -1012.0,
+      "loss": 0.1624,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -4.0,
+      "rewards/margins": 3.28125,
+      "rewards/rejected": -7.28125,
+      "step": 1790
+    },
+    {
+      "epoch": 1.8838304552590266,
+      "grad_norm": 3.3249208512087534,
+      "learning_rate": 1.82601827274658e-05,
+      "logits/chosen": -8.5,
+      "logits/rejected": -6.0,
+      "logps/chosen": -692.0,
+      "logps/rejected": -1004.0,
+      "loss": 0.1678,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -3.8125,
+      "rewards/margins": 3.546875,
+      "rewards/rejected": -7.375,
+      "step": 1800
+    },
+    {
+      "epoch": 1.8838304552590266,
+      "eval_logits/chosen": -8.4375,
+      "eval_logits/rejected": -6.8125,
+      "eval_logps/chosen": -776.0,
+      "eval_logps/rejected": -868.0,
+      "eval_loss": 0.6220312714576721,
+      "eval_rewards/accuracies": 0.724609375,
+      "eval_rewards/chosen": -4.5625,
+      "eval_rewards/margins": 1.2265625,
+      "eval_rewards/rejected": -5.8125,
+      "eval_runtime": 46.7428,
+      "eval_samples_per_second": 42.787,
+      "eval_steps_per_second": 0.685,
+      "step": 1800
+    },
+    {
+      "epoch": 1.8942961800104658,
+      "grad_norm": 2.912279053701135,
+      "learning_rate": 1.7967316349380213e-05,
+      "logits/chosen": -8.0,
+      "logits/rejected": -5.5625,
+      "logps/chosen": -660.0,
+      "logps/rejected": -984.0,
+      "loss": 0.189,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -3.640625,
+      "rewards/margins": 3.5,
+      "rewards/rejected": -7.125,
+      "step": 1810
+    },
+    {
+      "epoch": 1.9047619047619047,
+      "grad_norm": 3.307577894170201,
+      "learning_rate": 1.7675494329845516e-05,
+      "logits/chosen": -6.90625,
+      "logits/rejected": -4.15625,
+      "logps/chosen": -752.0,
+      "logps/rejected": -1056.0,
+      "loss": 0.1734,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -4.4375,
+      "rewards/margins": 3.375,
+      "rewards/rejected": -7.84375,
+      "step": 1820
+    },
+    {
+      "epoch": 1.915227629513344,
+      "grad_norm": 2.5178946630859746,
+      "learning_rate": 1.738476000464026e-05,
+      "logits/chosen": -6.34375,
+      "logits/rejected": -4.46875,
+      "logps/chosen": -708.0,
+      "logps/rejected": -1048.0,
+      "loss": 0.1856,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -4.21875,
+      "rewards/margins": 3.296875,
+      "rewards/rejected": -7.5,
+      "step": 1830
+    },
+    {
+      "epoch": 1.9256933542647827,
+      "grad_norm": 2.796315025094076,
+      "learning_rate": 1.7095156548019646e-05,
+      "logits/chosen": -7.53125,
+      "logits/rejected": -4.65625,
+      "logps/chosen": -768.0,
+      "logps/rejected": -1032.0,
+      "loss": 0.1912,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -4.09375,
+      "rewards/margins": 3.28125,
+      "rewards/rejected": -7.375,
+      "step": 1840
+    },
+    {
+      "epoch": 1.936159079016222,
+      "grad_norm": 2.188796890564701,
+      "learning_rate": 1.6806726966304064e-05,
+      "logits/chosen": -7.90625,
+      "logits/rejected": -5.5625,
+      "logps/chosen": -708.0,
+      "logps/rejected": -992.0,
+      "loss": 0.1816,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -3.875,
+      "rewards/margins": 3.1875,
+      "rewards/rejected": -7.0625,
+      "step": 1850
+    },
+    {
+      "epoch": 1.9466248037676608,
+      "grad_norm": 2.2841928625470493,
+      "learning_rate": 1.6519514091492625e-05,
+      "logits/chosen": -7.75,
+      "logits/rejected": -5.15625,
+      "logps/chosen": -720.0,
+      "logps/rejected": -1000.0,
+      "loss": 0.1713,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -4.09375,
+      "rewards/margins": 3.3125,
+      "rewards/rejected": -7.40625,
+      "step": 1860
+    },
+    {
+      "epoch": 1.9570905285191,
+      "grad_norm": 2.0291609215213535,
+      "learning_rate": 1.6233560574902627e-05,
+      "logits/chosen": -8.1875,
+      "logits/rejected": -5.375,
+      "logps/chosen": -776.0,
+      "logps/rejected": -1056.0,
+      "loss": 0.1445,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -4.03125,
+      "rewards/margins": 3.546875,
+      "rewards/rejected": -7.5625,
+      "step": 1870
+    },
+    {
+      "epoch": 1.9675562532705388,
+      "grad_norm": 2.671546913514888,
+      "learning_rate": 1.5948908880835753e-05,
+      "logits/chosen": -7.84375,
+      "logits/rejected": -5.5,
+      "logps/chosen": -752.0,
+      "logps/rejected": -1032.0,
+      "loss": 0.1789,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -4.21875,
+      "rewards/margins": 3.40625,
+      "rewards/rejected": -7.625,
+      "step": 1880
+    },
+    {
+      "epoch": 1.978021978021978,
+      "grad_norm": 3.000395121867142,
+      "learning_rate": 1.5665601280272124e-05,
+      "logits/chosen": -8.4375,
+      "logits/rejected": -6.0,
+      "logps/chosen": -760.0,
+      "logps/rejected": -1064.0,
+      "loss": 0.1816,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -4.34375,
+      "rewards/margins": 3.453125,
+      "rewards/rejected": -7.78125,
+      "step": 1890
+    },
+    {
+      "epoch": 1.988487702773417,
+      "grad_norm": 1.96466076407299,
+      "learning_rate": 1.538367984459298e-05,
+      "logits/chosen": -8.25,
+      "logits/rejected": -6.1875,
+      "logps/chosen": -828.0,
+      "logps/rejected": -1168.0,
+      "loss": 0.1563,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -4.78125,
+      "rewards/margins": 3.8125,
+      "rewards/rejected": -8.625,
+      "step": 1900
+    },
+    {
+      "epoch": 1.988487702773417,
+      "eval_logits/chosen": -7.875,
+      "eval_logits/rejected": -6.34375,
+      "eval_logps/chosen": -868.0,
+      "eval_logps/rejected": -976.0,
+      "eval_loss": 0.6273593902587891,
+      "eval_rewards/accuracies": 0.7265625,
+      "eval_rewards/chosen": -5.5,
+      "eval_rewards/margins": 1.3671875,
+      "eval_rewards/rejected": -6.84375,
+      "eval_runtime": 46.7597,
+      "eval_samples_per_second": 42.772,
+      "eval_steps_per_second": 0.684,
+      "step": 1900
+    },
+    {
+      "epoch": 1.9989534275248562,
+      "grad_norm": 2.2632717987306865,
+      "learning_rate": 1.5103186439333133e-05,
+      "logits/chosen": -8.5,
+      "logits/rejected": -6.0,
+      "logps/chosen": -792.0,
+      "logps/rejected": -1104.0,
+      "loss": 0.1607,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -4.46875,
+      "rewards/margins": 3.71875,
+      "rewards/rejected": -8.1875,
+      "step": 1910
+    },
+    {
+      "epoch": 2.009419152276295,
+      "grad_norm": 0.4215310299131895,
+      "learning_rate": 1.4824162717963829e-05,
+      "logits/chosen": -8.375,
+      "logits/rejected": -4.90625,
+      "logps/chosen": -804.0,
+      "logps/rejected": -1248.0,
+      "loss": 0.0334,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -4.375,
+      "rewards/margins": 5.09375,
+      "rewards/rejected": -9.5,
+      "step": 1920
+    },
+    {
+      "epoch": 2.0198848770277342,
+      "grad_norm": 0.6262303617780798,
+      "learning_rate": 1.4546650115707192e-05,
+      "logits/chosen": -7.0625,
+      "logits/rejected": -3.546875,
+      "logps/chosen": -752.0,
+      "logps/rejected": -1272.0,
+      "loss": 0.031,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -4.5625,
+      "rewards/margins": 5.4375,
+      "rewards/rejected": -10.0,
+      "step": 1930
+    },
+    {
+      "epoch": 2.030350601779173,
+      "grad_norm": 0.6905096103359027,
+      "learning_rate": 1.4270689843383111e-05,
+      "logits/chosen": -7.625,
+      "logits/rejected": -3.53125,
+      "logps/chosen": -788.0,
+      "logps/rejected": -1312.0,
+      "loss": 0.0211,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -4.75,
+      "rewards/margins": 5.59375,
+      "rewards/rejected": -10.375,
+      "step": 1940
+    },
+    {
+      "epoch": 2.0408163265306123,
+      "grad_norm": 0.8670323784257796,
+      "learning_rate": 1.3996322881289347e-05,
+      "logits/chosen": -7.3125,
+      "logits/rejected": -2.21875,
+      "logps/chosen": -812.0,
+      "logps/rejected": -1352.0,
+      "loss": 0.0241,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.0,
+      "rewards/margins": 5.9375,
+      "rewards/rejected": -10.9375,
+      "step": 1950
+    },
+    {
+      "epoch": 2.051282051282051,
+      "grad_norm": 0.5002049173122795,
+      "learning_rate": 1.3723589973115963e-05,
+      "logits/chosen": -5.75,
+      "logits/rejected": -2.796875,
+      "logps/chosen": -788.0,
+      "logps/rejected": -1416.0,
+      "loss": 0.0231,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.0,
+      "rewards/margins": 6.28125,
+      "rewards/rejected": -11.3125,
+      "step": 1960
+    },
+    {
+      "epoch": 2.0617477760334904,
+      "grad_norm": 0.35868451623944014,
+      "learning_rate": 1.3452531619894837e-05,
+      "logits/chosen": -6.0625,
+      "logits/rejected": -1.4375,
+      "logps/chosen": -868.0,
+      "logps/rejected": -1472.0,
+      "loss": 0.0205,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -5.34375,
+      "rewards/margins": 6.46875,
+      "rewards/rejected": -11.8125,
+      "step": 1970
+    },
+    {
+      "epoch": 2.072213500784929,
+      "grad_norm": 0.24164555890620235,
+      "learning_rate": 1.3183188073985245e-05,
+      "logits/chosen": -5.65625,
+      "logits/rejected": -1.5078125,
+      "logps/chosen": -888.0,
+      "logps/rejected": -1488.0,
+      "loss": 0.0249,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.59375,
+      "rewards/margins": 6.4375,
+      "rewards/rejected": -12.0625,
+      "step": 1980
+    },
+    {
+      "epoch": 2.0826792255363684,
+      "grad_norm": 0.7439884110895534,
+      "learning_rate": 1.2915599333096351e-05,
+      "logits/chosen": -5.6875,
+      "logits/rejected": -1.609375,
+      "logps/chosen": -864.0,
+      "logps/rejected": -1432.0,
+      "loss": 0.0243,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -5.28125,
+      "rewards/margins": 6.21875,
+      "rewards/rejected": -11.5,
+      "step": 1990
+    },
+    {
+      "epoch": 2.0931449502878072,
+      "grad_norm": 0.3391081996752237,
+      "learning_rate": 1.264980513434752e-05,
+      "logits/chosen": -6.9375,
+      "logits/rejected": -1.6796875,
+      "logps/chosen": -760.0,
+      "logps/rejected": -1392.0,
+      "loss": 0.0144,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -4.53125,
+      "rewards/margins": 6.71875,
+      "rewards/rejected": -11.25,
+      "step": 2000
+    },
+    {
+      "epoch": 2.0931449502878072,
+      "eval_logits/chosen": -5.5,
+      "eval_logits/rejected": -3.328125,
+      "eval_logps/chosen": -960.0,
+      "eval_logps/rejected": -1112.0,
+      "eval_loss": 0.7310937643051147,
+      "eval_rewards/accuracies": 0.73046875,
+      "eval_rewards/chosen": -6.4375,
+      "eval_rewards/margins": 1.765625,
+      "eval_rewards/rejected": -8.1875,
+      "eval_runtime": 46.7529,
+      "eval_samples_per_second": 42.778,
+      "eval_steps_per_second": 0.684,
+      "step": 2000
+    },
+    {
+      "epoch": 2.1036106750392465,
+      "grad_norm": 0.420054021749488,
+      "learning_rate": 1.2385844948367324e-05,
+      "logits/chosen": -6.9375,
+      "logits/rejected": -2.09375,
+      "logps/chosen": -816.0,
+      "logps/rejected": -1448.0,
+      "loss": 0.0236,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -4.71875,
+      "rewards/margins": 6.65625,
+      "rewards/rejected": -11.375,
+      "step": 2010
+    },
+    {
+      "epoch": 2.1140763997906853,
+      "grad_norm": 0.30034429346785213,
+      "learning_rate": 1.2123757973432114e-05,
+      "logits/chosen": -6.28125,
+      "logits/rejected": -1.609375,
+      "logps/chosen": -844.0,
+      "logps/rejected": -1488.0,
+      "loss": 0.0194,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -5.21875,
+      "rewards/margins": 6.8125,
+      "rewards/rejected": -12.0,
+      "step": 2020
+    },
+    {
+      "epoch": 2.1245421245421245,
+      "grad_norm": 0.3096690479276895,
+      "learning_rate": 1.1863583129645106e-05,
+      "logits/chosen": -6.21875,
+      "logits/rejected": -1.0,
+      "logps/chosen": -836.0,
+      "logps/rejected": -1456.0,
+      "loss": 0.0267,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -5.03125,
+      "rewards/margins": 6.75,
+      "rewards/rejected": -11.8125,
+      "step": 2030
+    },
+    {
+      "epoch": 2.1350078492935634,
+      "grad_norm": 0.48655820699418,
+      "learning_rate": 1.1605359053156607e-05,
+      "logits/chosen": -5.5,
+      "logits/rejected": -1.953125,
+      "logps/chosen": -796.0,
+      "logps/rejected": -1448.0,
+      "loss": 0.0206,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -4.9375,
+      "rewards/margins": 6.65625,
+      "rewards/rejected": -11.625,
+      "step": 2040
+    },
+    {
+      "epoch": 2.1454735740450026,
+      "grad_norm": 0.42623401019353696,
+      "learning_rate": 1.1349124090426644e-05,
+      "logits/chosen": -5.0625,
+      "logits/rejected": -0.5390625,
+      "logps/chosen": -844.0,
+      "logps/rejected": -1472.0,
+      "loss": 0.0241,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -5.25,
+      "rewards/margins": 6.625,
+      "rewards/rejected": -11.875,
+      "step": 2050
+    },
+    {
+      "epoch": 2.155939298796442,
+      "grad_norm": 0.5660310673186695,
+      "learning_rate": 1.1094916292530406e-05,
+      "logits/chosen": -5.84375,
+      "logits/rejected": -1.0390625,
+      "logps/chosen": -864.0,
+      "logps/rejected": -1504.0,
+      "loss": 0.0167,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -5.15625,
+      "rewards/margins": 6.90625,
+      "rewards/rejected": -12.0625,
+      "step": 2060
+    },
+    {
+      "epoch": 2.1664050235478807,
+      "grad_norm": 0.15008118834583475,
+      "learning_rate": 1.0842773409507623e-05,
+      "logits/chosen": -5.21875,
+      "logits/rejected": -1.21875,
+      "logps/chosen": -832.0,
+      "logps/rejected": -1512.0,
+      "loss": 0.0157,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.28125,
+      "rewards/margins": 6.96875,
+      "rewards/rejected": -12.25,
+      "step": 2070
+    },
+    {
+      "epoch": 2.17687074829932,
+      "grad_norm": 1.4003103610900558,
+      "learning_rate": 1.0592732884756754e-05,
+      "logits/chosen": -6.90625,
+      "logits/rejected": -1.953125,
+      "logps/chosen": -864.0,
+      "logps/rejected": -1512.0,
+      "loss": 0.0218,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.21875,
+      "rewards/margins": 6.84375,
+      "rewards/rejected": -12.0625,
+      "step": 2080
+    },
+    {
+      "epoch": 2.1873364730507587,
+      "grad_norm": 0.30817066192832565,
+      "learning_rate": 1.0344831849474507e-05,
+      "logits/chosen": -6.3125,
+      "logits/rejected": -1.796875,
+      "logps/chosen": -868.0,
+      "logps/rejected": -1528.0,
+      "loss": 0.0201,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.1875,
+      "rewards/margins": 7.0,
+      "rewards/rejected": -12.1875,
+      "step": 2090
+    },
+    {
+      "epoch": 2.197802197802198,
+      "grad_norm": 0.5547476415586855,
+      "learning_rate": 1.0099107117141879e-05,
+      "logits/chosen": -6.4375,
+      "logits/rejected": -1.3671875,
+      "logps/chosen": -868.0,
+      "logps/rejected": -1544.0,
+      "loss": 0.029,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.25,
+      "rewards/margins": 7.28125,
+      "rewards/rejected": -12.5625,
+      "step": 2100
+    },
+    {
+      "epoch": 2.197802197802198,
+      "eval_logits/chosen": -4.75,
+      "eval_logits/rejected": -2.375,
+      "eval_logps/chosen": -1072.0,
+      "eval_logps/rejected": -1256.0,
+      "eval_loss": 0.8194687366485596,
+      "eval_rewards/accuracies": 0.728515625,
+      "eval_rewards/chosen": -7.53125,
+      "eval_rewards/margins": 2.171875,
+      "eval_rewards/rejected": -9.6875,
+      "eval_runtime": 46.7617,
+      "eval_samples_per_second": 42.77,
+      "eval_steps_per_second": 0.684,
+      "step": 2100
+    },
+    {
+      "epoch": 2.208267922553637,
+      "grad_norm": 0.5690916434085425,
+      "learning_rate": 9.855595178057334e-06,
+      "logits/chosen": -6.53125,
+      "logits/rejected": -1.65625,
+      "logps/chosen": -896.0,
+      "logps/rejected": -1560.0,
+      "loss": 0.0173,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -5.53125,
+      "rewards/margins": 7.21875,
+      "rewards/rejected": -12.75,
+      "step": 2110
+    },
+    {
+      "epoch": 2.218733647305076,
+      "grad_norm": 0.39349368246179695,
+      "learning_rate": 9.614332193917949e-06,
+      "logits/chosen": -5.90625,
+      "logits/rejected": -1.890625,
+      "logps/chosen": -892.0,
+      "logps/rejected": -1672.0,
+      "loss": 0.0187,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.6875,
+      "rewards/margins": 8.125,
+      "rewards/rejected": -13.8125,
+      "step": 2120
+    },
+    {
+      "epoch": 2.229199372056515,
+      "grad_norm": 0.3936535873926383,
+      "learning_rate": 9.375353992449385e-06,
+      "logits/chosen": -5.9375,
+      "logits/rejected": -1.609375,
+      "logps/chosen": -856.0,
+      "logps/rejected": -1528.0,
+      "loss": 0.0225,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.21875,
+      "rewards/margins": 7.125,
+      "rewards/rejected": -12.3125,
+      "step": 2130
+    },
+    {
+      "epoch": 2.239665096807954,
+      "grad_norm": 0.4792231368085232,
+      "learning_rate": 9.138696062085442e-06,
+      "logits/chosen": -5.90625,
+      "logits/rejected": -1.984375,
+      "logps/chosen": -828.0,
+      "logps/rejected": -1544.0,
+      "loss": 0.0168,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -5.15625,
+      "rewards/margins": 7.28125,
+      "rewards/rejected": -12.4375,
+      "step": 2140
+    },
+    {
+      "epoch": 2.250130821559393,
+      "grad_norm": 0.8393949132721806,
+      "learning_rate": 8.904393546698007e-06,
+      "logits/chosen": -6.71875,
+      "logits/rejected": -1.703125,
+      "logps/chosen": -944.0,
+      "logps/rejected": -1648.0,
+      "loss": 0.0163,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -5.625,
+      "rewards/margins": 7.65625,
+      "rewards/rejected": -13.3125,
+      "step": 2150
+    },
+    {
+      "epoch": 2.260596546310832,
+      "grad_norm": 0.26081064410126065,
+      "learning_rate": 8.672481240378142e-06,
+      "logits/chosen": -5.96875,
+      "logits/rejected": -1.890625,
+      "logps/chosen": -844.0,
+      "logps/rejected": -1536.0,
+      "loss": 0.02,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.21875,
+      "rewards/margins": 7.0625,
+      "rewards/rejected": -12.25,
+      "step": 2160
+    },
+    {
+      "epoch": 2.271062271062271,
+      "grad_norm": 0.17488356202171088,
+      "learning_rate": 8.44299358226919e-06,
+      "logits/chosen": -6.875,
+      "logits/rejected": -1.609375,
+      "logps/chosen": -856.0,
+      "logps/rejected": -1488.0,
+      "loss": 0.0173,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -5.03125,
+      "rewards/margins": 7.125,
+      "rewards/rejected": -12.125,
+      "step": 2170
+    },
+    {
+      "epoch": 2.2815279958137102,
+      "grad_norm": 0.92346178698253,
+      "learning_rate": 8.215964651452457e-06,
+      "logits/chosen": -6.5,
+      "logits/rejected": -1.5,
+      "logps/chosen": -892.0,
+      "logps/rejected": -1560.0,
+      "loss": 0.0186,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.21875,
+      "rewards/margins": 7.3125,
+      "rewards/rejected": -12.5,
+      "step": 2180
+    },
+    {
+      "epoch": 2.291993720565149,
+      "grad_norm": 0.3715052023584167,
+      "learning_rate": 7.991428161886502e-06,
+      "logits/chosen": -5.84375,
+      "logits/rejected": -1.4765625,
+      "logps/chosen": -832.0,
+      "logps/rejected": -1560.0,
+      "loss": 0.0183,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.15625,
+      "rewards/margins": 7.21875,
+      "rewards/rejected": -12.375,
+      "step": 2190
+    },
+    {
+      "epoch": 2.3024594453165883,
+      "grad_norm": 0.8323952472768275,
+      "learning_rate": 7.76941745740061e-06,
+      "logits/chosen": -6.15625,
+      "logits/rejected": -0.94140625,
+      "logps/chosen": -820.0,
+      "logps/rejected": -1424.0,
+      "loss": 0.0228,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -5.0,
+      "rewards/margins": 6.71875,
+      "rewards/rejected": -11.75,
+      "step": 2200
+    },
+    {
+      "epoch": 2.3024594453165883,
+      "eval_logits/chosen": -4.375,
+      "eval_logits/rejected": -1.9296875,
+      "eval_logps/chosen": -1088.0,
+      "eval_logps/rejected": -1280.0,
+      "eval_loss": 0.8282343745231628,
+      "eval_rewards/accuracies": 0.71875,
+      "eval_rewards/chosen": -7.6875,
+      "eval_rewards/margins": 2.203125,
+      "eval_rewards/rejected": -9.875,
+      "eval_runtime": 46.7524,
+      "eval_samples_per_second": 42.779,
+      "eval_steps_per_second": 0.684,
+      "step": 2200
+    },
+    {
+      "epoch": 2.312925170068027,
+      "grad_norm": 0.9318304280718913,
+      "learning_rate": 7.549965506743123e-06,
+      "logits/chosen": -5.40625,
+      "logits/rejected": 0.0203857421875,
+      "logps/chosen": -900.0,
+      "logps/rejected": -1568.0,
+      "loss": 0.0205,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -5.75,
+      "rewards/margins": 7.3125,
+      "rewards/rejected": -13.0625,
+      "step": 2210
+    },
+    {
+      "epoch": 2.3233908948194664,
+      "grad_norm": 0.5650400203896252,
+      "learning_rate": 7.333104898685631e-06,
+      "logits/chosen": -6.71875,
+      "logits/rejected": -1.8203125,
+      "logps/chosen": -900.0,
+      "logps/rejected": -1640.0,
+      "loss": 0.0185,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -5.46875,
+      "rewards/margins": 7.8125,
+      "rewards/rejected": -13.25,
+      "step": 2220
+    },
+    {
+      "epoch": 2.333856619570905,
+      "grad_norm": 0.7952751884731586,
+      "learning_rate": 7.118867837183518e-06,
+      "logits/chosen": -6.0,
+      "logits/rejected": -1.6953125,
+      "logps/chosen": -836.0,
+      "logps/rejected": -1464.0,
+      "loss": 0.0223,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -5.125,
+      "rewards/margins": 6.6875,
+      "rewards/rejected": -11.8125,
+      "step": 2230
+    },
+    {
+      "epoch": 2.3443223443223444,
+      "grad_norm": 0.22746959316725637,
+      "learning_rate": 6.9072861365936045e-06,
+      "logits/chosen": -6.1875,
+      "logits/rejected": -1.4140625,
+      "logps/chosen": -752.0,
+      "logps/rejected": -1512.0,
+      "loss": 0.0122,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -4.5625,
+      "rewards/margins": 7.71875,
+      "rewards/rejected": -12.3125,
+      "step": 2240
+    },
+    {
+      "epoch": 2.3547880690737832,
+      "grad_norm": 0.7080654201204992,
+      "learning_rate": 6.6983912169497005e-06,
+      "logits/chosen": -6.0,
+      "logits/rejected": -0.76953125,
+      "logps/chosen": -892.0,
+      "logps/rejected": -1632.0,
+      "loss": 0.0146,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -5.5625,
+      "rewards/margins": 7.96875,
+      "rewards/rejected": -13.5,
+      "step": 2250
+    },
+    {
+      "epoch": 2.3652537938252225,
+      "grad_norm": 0.1901005601169168,
+      "learning_rate": 6.49221409929677e-06,
+      "logits/chosen": -6.375,
+      "logits/rejected": -1.875,
+      "logps/chosen": -920.0,
+      "logps/rejected": -1600.0,
+      "loss": 0.0207,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.53125,
+      "rewards/margins": 7.25,
+      "rewards/rejected": -12.75,
+      "step": 2260
+    },
+    {
+      "epoch": 2.3757195185766613,
+      "grad_norm": 0.705651333129883,
+      "learning_rate": 6.2887854010842075e-06,
+      "logits/chosen": -5.625,
+      "logits/rejected": -0.5703125,
+      "logps/chosen": -900.0,
+      "logps/rejected": -1624.0,
+      "loss": 0.0205,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.53125,
+      "rewards/margins": 7.84375,
+      "rewards/rejected": -13.375,
+      "step": 2270
+    },
+    {
+      "epoch": 2.3861852433281006,
+      "grad_norm": 0.604993862925217,
+      "learning_rate": 6.0881353316191385e-06,
+      "logits/chosen": -6.03125,
+      "logits/rejected": -1.3515625,
+      "logps/chosen": -900.0,
+      "logps/rejected": -1648.0,
+      "loss": 0.0154,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.53125,
+      "rewards/margins": 7.75,
+      "rewards/rejected": -13.3125,
+      "step": 2280
+    },
+    {
+      "epoch": 2.3966509680795394,
+      "grad_norm": 0.18009987518286957,
+      "learning_rate": 5.890293687580381e-06,
+      "logits/chosen": -5.71875,
+      "logits/rejected": -1.3671875,
+      "logps/chosen": -844.0,
+      "logps/rejected": -1528.0,
+      "loss": 0.0228,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -5.0625,
+      "rewards/margins": 7.125,
+      "rewards/rejected": -12.1875,
+      "step": 2290
+    },
+    {
+      "epoch": 2.4071166928309786,
+      "grad_norm": 0.37922983655048886,
+      "learning_rate": 5.695289848593532e-06,
+      "logits/chosen": -5.875,
+      "logits/rejected": -1.375,
+      "logps/chosen": -824.0,
+      "logps/rejected": -1520.0,
+      "loss": 0.0159,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -5.0,
+      "rewards/margins": 7.25,
+      "rewards/rejected": -12.25,
+      "step": 2300
+    },
+    {
+      "epoch": 2.4071166928309786,
+      "eval_logits/chosen": -4.46875,
+      "eval_logits/rejected": -2.0625,
+      "eval_logps/chosen": -1040.0,
+      "eval_logps/rejected": -1224.0,
+      "eval_loss": 0.8055156469345093,
+      "eval_rewards/accuracies": 0.7265625,
+      "eval_rewards/chosen": -7.21875,
+      "eval_rewards/margins": 2.15625,
+      "eval_rewards/rejected": -9.375,
+      "eval_runtime": 46.7709,
+      "eval_samples_per_second": 42.762,
+      "eval_steps_per_second": 0.684,
+      "step": 2300
+    },
+    {
+      "epoch": 2.4175824175824174,
+      "grad_norm": 0.31772836755103945,
+      "learning_rate": 5.50315277286812e-06,
+      "logits/chosen": -5.8125,
+      "logits/rejected": -1.5,
+      "logps/chosen": -812.0,
+      "logps/rejected": -1568.0,
+      "loss": 0.0143,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -5.09375,
+      "rewards/margins": 7.53125,
+      "rewards/rejected": -12.625,
+      "step": 2310
+    },
+    {
+      "epoch": 2.4280481423338567,
+      "grad_norm": 0.2154370732976629,
+      "learning_rate": 5.313910992897281e-06,
+      "logits/chosen": -6.71875,
+      "logits/rejected": -1.703125,
+      "logps/chosen": -888.0,
+      "logps/rejected": -1568.0,
+      "loss": 0.0162,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.125,
+      "rewards/margins": 7.46875,
+      "rewards/rejected": -12.625,
+      "step": 2320
+    },
+    {
+      "epoch": 2.4385138670852955,
+      "grad_norm": 0.19328593029349928,
+      "learning_rate": 5.127592611220658e-06,
+      "logits/chosen": -6.21875,
+      "logits/rejected": -1.609375,
+      "logps/chosen": -864.0,
+      "logps/rejected": -1536.0,
+      "loss": 0.0183,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -5.15625,
+      "rewards/margins": 7.375,
+      "rewards/rejected": -12.5,
+      "step": 2330
+    },
+    {
+      "epoch": 2.4489795918367347,
+      "grad_norm": 0.33423736888387207,
+      "learning_rate": 4.944225296251159e-06,
+      "logits/chosen": -7.03125,
+      "logits/rejected": -2.1875,
+      "logps/chosen": -976.0,
+      "logps/rejected": -1648.0,
+      "loss": 0.017,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -5.78125,
+      "rewards/margins": 7.40625,
+      "rewards/rejected": -13.1875,
+      "step": 2340
+    },
+    {
+      "epoch": 2.4594453165881736,
+      "grad_norm": 0.5513059307990195,
+      "learning_rate": 4.763836278166172e-06,
+      "logits/chosen": -6.53125,
+      "logits/rejected": -0.95703125,
+      "logps/chosen": -828.0,
+      "logps/rejected": -1584.0,
+      "loss": 0.0128,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.125,
+      "rewards/margins": 7.875,
+      "rewards/rejected": -13.0,
+      "step": 2350
+    },
+    {
+      "epoch": 2.469911041339613,
+      "grad_norm": 0.5699591534671058,
+      "learning_rate": 4.586452344863876e-06,
+      "logits/chosen": -6.5,
+      "logits/rejected": -1.1328125,
+      "logps/chosen": -848.0,
+      "logps/rejected": -1616.0,
+      "loss": 0.0165,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.4375,
+      "rewards/margins": 8.125,
+      "rewards/rejected": -13.625,
+      "step": 2360
+    },
+    {
+      "epoch": 2.4803767660910516,
+      "grad_norm": 0.2787151315982646,
+      "learning_rate": 4.412099837985192e-06,
+      "logits/chosen": -6.0625,
+      "logits/rejected": -1.1875,
+      "logps/chosen": -868.0,
+      "logps/rejected": -1576.0,
+      "loss": 0.0244,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.46875,
+      "rewards/margins": 7.53125,
+      "rewards/rejected": -13.0,
+      "step": 2370
+    },
+    {
+      "epoch": 2.490842490842491,
+      "grad_norm": 0.7215120757147602,
+      "learning_rate": 4.240804649002089e-06,
+      "logits/chosen": -5.4375,
+      "logits/rejected": -1.3984375,
+      "logps/chosen": -856.0,
+      "logps/rejected": -1600.0,
+      "loss": 0.0148,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.53125,
+      "rewards/margins": 7.625,
+      "rewards/rejected": -13.125,
+      "step": 2380
+    },
+    {
+      "epoch": 2.50130821559393,
+      "grad_norm": 0.617491440991175,
+      "learning_rate": 4.072592215372598e-06,
+      "logits/chosen": -5.625,
+      "logits/rejected": -0.98046875,
+      "logps/chosen": -820.0,
+      "logps/rejected": -1488.0,
+      "loss": 0.0167,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -5.09375,
+      "rewards/margins": 7.3125,
+      "rewards/rejected": -12.375,
+      "step": 2390
+    },
+    {
+      "epoch": 2.511773940345369,
+      "grad_norm": 0.46234291028839625,
+      "learning_rate": 3.907487516763389e-06,
+      "logits/chosen": -6.9375,
+      "logits/rejected": -1.421875,
+      "logps/chosen": -860.0,
+      "logps/rejected": -1528.0,
+      "loss": 0.0192,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -5.125,
+      "rewards/margins": 7.15625,
+      "rewards/rejected": -12.3125,
+      "step": 2400
+    },
+    {
+      "epoch": 2.511773940345369,
+      "eval_logits/chosen": -4.78125,
+      "eval_logits/rejected": -2.390625,
+      "eval_logps/chosen": -1016.0,
+      "eval_logps/rejected": -1200.0,
+      "eval_loss": 0.7880625128746033,
+      "eval_rewards/accuracies": 0.720703125,
+      "eval_rewards/chosen": -6.96875,
+      "eval_rewards/margins": 2.09375,
+      "eval_rewards/rejected": -9.0625,
+      "eval_runtime": 46.7811,
+      "eval_samples_per_second": 42.752,
+      "eval_steps_per_second": 0.684,
+      "step": 2400
+    },
+    {
+      "epoch": 2.5222396650968077,
+      "grad_norm": 1.6851732945613722,
+      "learning_rate": 3.745515071340286e-06,
+      "logits/chosen": -6.21875,
+      "logits/rejected": -1.9609375,
+      "logps/chosen": -860.0,
+      "logps/rejected": -1504.0,
+      "loss": 0.0159,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.03125,
+      "rewards/margins": 7.03125,
+      "rewards/rejected": -12.0625,
+      "step": 2410
+    },
+    {
+      "epoch": 2.532705389848247,
+      "grad_norm": 0.5303579238258358,
+      "learning_rate": 3.586698932127236e-06,
+      "logits/chosen": -5.6875,
+      "logits/rejected": -1.34375,
+      "logps/chosen": -788.0,
+      "logps/rejected": -1440.0,
+      "loss": 0.0225,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -4.5625,
+      "rewards/margins": 6.90625,
+      "rewards/rejected": -11.5,
+      "step": 2420
+    },
+    {
+      "epoch": 2.5431711145996863,
+      "grad_norm": 0.5941785786541246,
+      "learning_rate": 3.4310626834344744e-06,
+      "logits/chosen": -5.53125,
+      "logits/rejected": -1.328125,
+      "logps/chosen": -832.0,
+      "logps/rejected": -1496.0,
+      "loss": 0.0142,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -4.90625,
+      "rewards/margins": 7.15625,
+      "rewards/rejected": -12.0625,
+      "step": 2430
+    },
+    {
+      "epoch": 2.553636839351125,
+      "grad_norm": 0.4015441177113004,
+      "learning_rate": 3.2786294373562343e-06,
+      "logits/chosen": -5.84375,
+      "logits/rejected": -0.234375,
+      "logps/chosen": -808.0,
+      "logps/rejected": -1544.0,
+      "loss": 0.0199,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -4.875,
+      "rewards/margins": 7.65625,
+      "rewards/rejected": -12.5625,
+      "step": 2440
+    },
+    {
+      "epoch": 2.564102564102564,
+      "grad_norm": 0.45650090883481115,
+      "learning_rate": 3.1294218303385526e-06,
+      "logits/chosen": -5.3125,
+      "logits/rejected": -1.453125,
+      "logps/chosen": -876.0,
+      "logps/rejected": -1536.0,
+      "loss": 0.0208,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -5.3125,
+      "rewards/margins": 7.0,
+      "rewards/rejected": -12.3125,
+      "step": 2450
+    },
+    {
+      "epoch": 2.574568288854003,
+      "grad_norm": 0.9646041040666874,
+      "learning_rate": 2.9834620198177483e-06,
+      "logits/chosen": -5.84375,
+      "logits/rejected": -1.234375,
+      "logps/chosen": -880.0,
+      "logps/rejected": -1552.0,
+      "loss": 0.0196,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -5.15625,
+      "rewards/margins": 7.34375,
+      "rewards/rejected": -12.5,
+      "step": 2460
+    },
+    {
+      "epoch": 2.5850340136054424,
+      "grad_norm": 3.891878102698795,
+      "learning_rate": 2.840771680930068e-06,
+      "logits/chosen": -5.625,
+      "logits/rejected": -1.671875,
+      "logps/chosen": -796.0,
+      "logps/rejected": -1504.0,
+      "loss": 0.0235,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -5.125,
+      "rewards/margins": 7.09375,
+      "rewards/rejected": -12.1875,
+      "step": 2470
+    },
+    {
+      "epoch": 2.595499738356881,
+      "grad_norm": 0.421140457858363,
+      "learning_rate": 2.701372003292868e-06,
+      "logits/chosen": -4.71875,
+      "logits/rejected": -1.375,
+      "logps/chosen": -864.0,
+      "logps/rejected": -1624.0,
+      "loss": 0.0176,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.4375,
+      "rewards/margins": 7.53125,
+      "rewards/rejected": -13.0,
+      "step": 2480
+    },
+    {
+      "epoch": 2.60596546310832,
+      "grad_norm": 0.7873216889216538,
+      "learning_rate": 2.56528368785795e-06,
+      "logits/chosen": -5.6875,
+      "logits/rejected": -1.75,
+      "logps/chosen": -856.0,
+      "logps/rejected": -1640.0,
+      "loss": 0.0187,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.15625,
+      "rewards/margins": 7.90625,
+      "rewards/rejected": -13.0625,
+      "step": 2490
+    },
+    {
+      "epoch": 2.6164311878597593,
+      "grad_norm": 0.38529926833259553,
+      "learning_rate": 2.432526943837471e-06,
+      "logits/chosen": -5.8125,
+      "logits/rejected": -0.9921875,
+      "logps/chosen": -800.0,
+      "logps/rejected": -1552.0,
+      "loss": 0.0158,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -4.875,
+      "rewards/margins": 7.78125,
+      "rewards/rejected": -12.625,
+      "step": 2500
+    },
+    {
+      "epoch": 2.6164311878597593,
+      "eval_logits/chosen": -3.9375,
+      "eval_logits/rejected": -1.53125,
+      "eval_logps/chosen": -1056.0,
+      "eval_logps/rejected": -1240.0,
+      "eval_loss": 0.8027187585830688,
+      "eval_rewards/accuracies": 0.7265625,
+      "eval_rewards/chosen": -7.34375,
+      "eval_rewards/margins": 2.15625,
+      "eval_rewards/rejected": -9.5,
+      "eval_runtime": 46.775,
+      "eval_samples_per_second": 42.758,
+      "eval_steps_per_second": 0.684,
+      "step": 2500
+    },
+    {
+      "epoch": 2.6268969126111985,
+      "grad_norm": 0.4756965199166916,
+      "learning_rate": 2.3031214857028446e-06,
+      "logits/chosen": -5.9375,
+      "logits/rejected": -1.0,
+      "logps/chosen": -880.0,
+      "logps/rejected": -1536.0,
+      "loss": 0.0197,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.28125,
+      "rewards/margins": 7.28125,
+      "rewards/rejected": -12.5625,
+      "step": 2510
+    },
+    {
+      "epoch": 2.6373626373626373,
+      "grad_norm": 0.1942815704536251,
+      "learning_rate": 2.177086530257139e-06,
+      "logits/chosen": -5.4375,
+      "logits/rejected": -0.59765625,
+      "logps/chosen": -852.0,
+      "logps/rejected": -1592.0,
+      "loss": 0.0223,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.28125,
+      "rewards/margins": 7.6875,
+      "rewards/rejected": -13.0,
+      "step": 2520
+    },
+    {
+      "epoch": 2.647828362114076,
+      "grad_norm": 0.40053766542842284,
+      "learning_rate": 2.054440793781354e-06,
+      "logits/chosen": -5.6875,
+      "logits/rejected": -1.5703125,
+      "logps/chosen": -820.0,
+      "logps/rejected": -1536.0,
+      "loss": 0.0167,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.0625,
+      "rewards/margins": 7.375,
+      "rewards/rejected": -12.4375,
+      "step": 2530
+    },
+    {
+      "epoch": 2.6582940868655154,
+      "grad_norm": 0.13153977646937196,
+      "learning_rate": 1.9352024892550587e-06,
+      "logits/chosen": -5.53125,
+      "logits/rejected": -0.388671875,
+      "logps/chosen": -848.0,
+      "logps/rejected": -1552.0,
+      "loss": 0.0164,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.15625,
+      "rewards/margins": 7.53125,
+      "rewards/rejected": -12.6875,
+      "step": 2540
+    },
+    {
+      "epoch": 2.6687598116169546,
+      "grad_norm": 0.3005166665112543,
+      "learning_rate": 1.8193893236517379e-06,
+      "logits/chosen": -5.5625,
+      "logits/rejected": -0.9140625,
+      "logps/chosen": -872.0,
+      "logps/rejected": -1528.0,
+      "loss": 0.0155,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.375,
+      "rewards/margins": 6.96875,
+      "rewards/rejected": -12.375,
+      "step": 2550
+    },
+    {
+      "epoch": 2.6792255363683934,
+      "grad_norm": 0.6457895710128586,
+      "learning_rate": 1.7070184953092956e-06,
+      "logits/chosen": -5.3125,
+      "logits/rejected": -0.67578125,
+      "logps/chosen": -860.0,
+      "logps/rejected": -1600.0,
+      "loss": 0.0163,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -5.34375,
+      "rewards/margins": 7.65625,
+      "rewards/rejected": -13.0,
+      "step": 2560
+    },
+    {
+      "epoch": 2.6896912611198327,
+      "grad_norm": 0.18363505100898042,
+      "learning_rate": 1.5981066913760918e-06,
+      "logits/chosen": -5.28125,
+      "logits/rejected": -0.55078125,
+      "logps/chosen": -884.0,
+      "logps/rejected": -1584.0,
+      "loss": 0.0121,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.3125,
+      "rewards/margins": 7.5,
+      "rewards/rejected": -12.8125,
+      "step": 2570
+    },
+    {
+      "epoch": 2.7001569858712715,
+      "grad_norm": 0.07860222158878338,
+      "learning_rate": 1.492670085332884e-06,
+      "logits/chosen": -5.25,
+      "logits/rejected": -0.5078125,
+      "logps/chosen": -860.0,
+      "logps/rejected": -1544.0,
+      "loss": 0.0152,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.46875,
+      "rewards/margins": 7.375,
+      "rewards/rejected": -12.8125,
+      "step": 2580
+    },
+    {
+      "epoch": 2.7106227106227108,
+      "grad_norm": 2.2403149453708333,
+      "learning_rate": 1.3907243345910787e-06,
+      "logits/chosen": -4.28125,
+      "logits/rejected": -0.23046875,
+      "logps/chosen": -836.0,
+      "logps/rejected": -1520.0,
+      "loss": 0.0212,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -5.34375,
+      "rewards/margins": 7.125,
+      "rewards/rejected": -12.4375,
+      "step": 2590
+    },
+    {
+      "epoch": 2.7210884353741496,
+      "grad_norm": 0.38315791022404566,
+      "learning_rate": 1.292284578167538e-06,
+      "logits/chosen": -5.40625,
+      "logits/rejected": -1.15625,
+      "logps/chosen": -860.0,
+      "logps/rejected": -1568.0,
+      "loss": 0.0193,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.375,
+      "rewards/margins": 7.4375,
+      "rewards/rejected": -12.8125,
+      "step": 2600
+    },
+    {
+      "epoch": 2.7210884353741496,
+      "eval_logits/chosen": -3.59375,
+      "eval_logits/rejected": -1.1796875,
+      "eval_logps/chosen": -1080.0,
+      "eval_logps/rejected": -1280.0,
+      "eval_loss": 0.8204843997955322,
+      "eval_rewards/accuracies": 0.73828125,
+      "eval_rewards/chosen": -7.625,
+      "eval_rewards/margins": 2.25,
+      "eval_rewards/rejected": -9.875,
+      "eval_runtime": 46.7669,
+      "eval_samples_per_second": 42.765,
+      "eval_steps_per_second": 0.684,
+      "step": 2600
+    },
+    {
+      "epoch": 2.731554160125589,
+      "grad_norm": 0.198101401399119,
+      "learning_rate": 1.1973654344364925e-06,
+      "logits/chosen": -5.0,
+      "logits/rejected": -0.734375,
+      "logps/chosen": -812.0,
+      "logps/rejected": -1536.0,
+      "loss": 0.026,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.125,
+      "rewards/margins": 7.375,
+      "rewards/rejected": -12.5,
+      "step": 2610
+    },
+    {
+      "epoch": 2.7420198848770276,
+      "grad_norm": 0.23137614204349777,
+      "learning_rate": 1.1059809989586583e-06,
+      "logits/chosen": -4.875,
+      "logits/rejected": -0.162109375,
+      "logps/chosen": -832.0,
+      "logps/rejected": -1552.0,
+      "loss": 0.0143,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -5.3125,
+      "rewards/margins": 7.46875,
+      "rewards/rejected": -12.75,
+      "step": 2620
+    },
+    {
+      "epoch": 2.752485609628467,
+      "grad_norm": 0.4737825889335588,
+      "learning_rate": 1.018144842388044e-06,
+      "logits/chosen": -4.59375,
+      "logits/rejected": -0.58984375,
+      "logps/chosen": -816.0,
+      "logps/rejected": -1568.0,
+      "loss": 0.0204,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.15625,
+      "rewards/margins": 7.65625,
+      "rewards/rejected": -12.8125,
+      "step": 2630
+    },
+    {
+      "epoch": 2.7629513343799057,
+      "grad_norm": 0.5155703466680202,
+      "learning_rate": 9.338700084567109e-07,
+      "logits/chosen": -5.5625,
+      "logits/rejected": -0.419921875,
+      "logps/chosen": -864.0,
+      "logps/rejected": -1552.0,
+      "loss": 0.0185,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.25,
+      "rewards/margins": 7.40625,
+      "rewards/rejected": -12.6875,
+      "step": 2640
+    },
+    {
+      "epoch": 2.773417059131345,
+      "grad_norm": 0.5012521170037756,
+      "learning_rate": 8.531690120377606e-07,
+      "logits/chosen": -5.40625,
+      "logits/rejected": -0.73828125,
+      "logps/chosen": -848.0,
+      "logps/rejected": -1504.0,
+      "loss": 0.0134,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.21875,
+      "rewards/margins": 6.9375,
+      "rewards/rejected": -12.1875,
+      "step": 2650
+    },
+    {
+      "epoch": 2.7838827838827838,
+      "grad_norm": 0.6242923251770285,
+      "learning_rate": 7.760538372868636e-07,
+      "logits/chosen": -5.625,
+      "logits/rejected": -0.55859375,
+      "logps/chosen": -820.0,
+      "logps/rejected": -1488.0,
+      "loss": 0.0206,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -5.15625,
+      "rewards/margins": 7.09375,
+      "rewards/rejected": -12.25,
+      "step": 2660
+    },
+    {
+      "epoch": 2.794348508634223,
+      "grad_norm": 1.6249684618909024,
+      "learning_rate": 7.025359358626166e-07,
+      "logits/chosen": -5.0625,
+      "logits/rejected": 0.240234375,
+      "logps/chosen": -820.0,
+      "logps/rejected": -1440.0,
+      "loss": 0.021,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -5.09375,
+      "rewards/margins": 6.84375,
+      "rewards/rejected": -11.9375,
+      "step": 2670
+    },
+    {
+      "epoch": 2.804814233385662,
+      "grad_norm": 0.6179710722592769,
+      "learning_rate": 6.326262252259463e-07,
+      "logits/chosen": -5.40625,
+      "logits/rejected": -0.90234375,
+      "logps/chosen": -904.0,
+      "logps/rejected": -1608.0,
+      "loss": 0.0135,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.4375,
+      "rewards/margins": 7.59375,
+      "rewards/rejected": -13.0625,
+      "step": 2680
+    },
+    {
+      "epoch": 2.815279958137101,
+      "grad_norm": 0.2650415046088335,
+      "learning_rate": 5.66335087018871e-07,
+      "logits/chosen": -4.78125,
+      "logits/rejected": -0.404296875,
+      "logps/chosen": -832.0,
+      "logps/rejected": -1560.0,
+      "loss": 0.0161,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.25,
+      "rewards/margins": 7.59375,
+      "rewards/rejected": -12.875,
+      "step": 2690
+    },
+    {
+      "epoch": 2.82574568288854,
+      "grad_norm": 2.156332609092923,
+      "learning_rate": 5.036723655228226e-07,
+      "logits/chosen": -6.03125,
+      "logits/rejected": -0.6875,
+      "logps/chosen": -860.0,
+      "logps/rejected": -1544.0,
+      "loss": 0.0229,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.1875,
+      "rewards/margins": 7.25,
+      "rewards/rejected": -12.4375,
+      "step": 2700
+    },
+    {
+      "epoch": 2.82574568288854,
+      "eval_logits/chosen": -3.953125,
+      "eval_logits/rejected": -1.5390625,
+      "eval_logps/chosen": -1064.0,
+      "eval_logps/rejected": -1256.0,
+      "eval_loss": 0.8136249780654907,
+      "eval_rewards/accuracies": 0.7265625,
+      "eval_rewards/chosen": -7.4375,
+      "eval_rewards/margins": 2.21875,
+      "eval_rewards/rejected": -9.625,
+      "eval_runtime": 46.7943,
+      "eval_samples_per_second": 42.74,
+      "eval_steps_per_second": 0.684,
+      "step": 2700
+    },
+    {
+      "epoch": 2.836211407639979,
+      "grad_norm": 0.7868836589166489,
+      "learning_rate": 4.446473661967432e-07,
+      "logits/chosen": -5.84375,
+      "logits/rejected": -1.546875,
+      "logps/chosen": -852.0,
+      "logps/rejected": -1568.0,
+      "loss": 0.0134,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -4.90625,
+      "rewards/margins": 7.75,
+      "rewards/rejected": -12.6875,
+      "step": 2710
+    },
+    {
+      "epoch": 2.846677132391418,
+      "grad_norm": 0.5405466094266178,
+      "learning_rate": 3.892688542952411e-07,
+      "logits/chosen": -5.53125,
+      "logits/rejected": -0.9296875,
+      "logps/chosen": -840.0,
+      "logps/rejected": -1544.0,
+      "loss": 0.014,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.15625,
+      "rewards/margins": 7.46875,
+      "rewards/rejected": -12.625,
+      "step": 2720
+    },
+    {
+      "epoch": 2.857142857142857,
+      "grad_norm": 0.1336856184832448,
+      "learning_rate": 3.3754505356694e-07,
+      "logits/chosen": -5.71875,
+      "logits/rejected": -1.1328125,
+      "logps/chosen": -804.0,
+      "logps/rejected": -1608.0,
+      "loss": 0.0182,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -4.96875,
+      "rewards/margins": 8.0625,
+      "rewards/rejected": -13.0,
+      "step": 2730
+    },
+    {
+      "epoch": 2.867608581894296,
+      "grad_norm": 1.017087469961355,
+      "learning_rate": 2.8948364503322276e-07,
+      "logits/chosen": -5.4375,
+      "logits/rejected": -1.71875,
+      "logps/chosen": -844.0,
+      "logps/rejected": -1584.0,
+      "loss": 0.0179,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.25,
+      "rewards/margins": 7.625,
+      "rewards/rejected": -12.875,
+      "step": 2740
+    },
+    {
+      "epoch": 2.8780743066457353,
+      "grad_norm": 0.26857100101145814,
+      "learning_rate": 2.4509176584762706e-07,
+      "logits/chosen": -5.5,
+      "logits/rejected": -0.2109375,
+      "logps/chosen": -852.0,
+      "logps/rejected": -1520.0,
+      "loss": 0.0177,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.3125,
+      "rewards/margins": 7.15625,
+      "rewards/rejected": -12.5,
+      "step": 2750
+    },
+    {
+      "epoch": 2.8885400313971745,
+      "grad_norm": 0.1655015176486014,
+      "learning_rate": 2.0437600823595694e-07,
+      "logits/chosen": -5.25,
+      "logits/rejected": -0.30859375,
+      "logps/chosen": -868.0,
+      "logps/rejected": -1552.0,
+      "loss": 0.0172,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.46875,
+      "rewards/margins": 7.15625,
+      "rewards/rejected": -12.625,
+      "step": 2760
+    },
+    {
+      "epoch": 2.8990057561486133,
+      "grad_norm": 0.2948831665999949,
+      "learning_rate": 1.6734241851733e-07,
+      "logits/chosen": -5.8125,
+      "logits/rejected": -0.2373046875,
+      "logps/chosen": -872.0,
+      "logps/rejected": -1576.0,
+      "loss": 0.0176,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.09375,
+      "rewards/margins": 7.5625,
+      "rewards/rejected": -12.625,
+      "step": 2770
+    },
+    {
+      "epoch": 2.909471480900052,
+      "grad_norm": 0.15859358514709437,
+      "learning_rate": 1.339964962062984e-07,
+      "logits/chosen": -6.0,
+      "logits/rejected": -0.703125,
+      "logps/chosen": -900.0,
+      "logps/rejected": -1584.0,
+      "loss": 0.0208,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.4375,
+      "rewards/margins": 7.21875,
+      "rewards/rejected": -12.6875,
+      "step": 2780
+    },
+    {
+      "epoch": 2.9199372056514914,
+      "grad_norm": 0.2631987469531854,
+      "learning_rate": 1.0434319319617136e-07,
+      "logits/chosen": -5.65625,
+      "logits/rejected": -1.1484375,
+      "logps/chosen": -848.0,
+      "logps/rejected": -1592.0,
+      "loss": 0.0176,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.125,
+      "rewards/margins": 7.75,
+      "rewards/rejected": -12.875,
+      "step": 2790
+    },
+    {
+      "epoch": 2.9304029304029307,
+      "grad_norm": 0.23443101177561032,
+      "learning_rate": 7.838691302365363e-08,
+      "logits/chosen": -5.28125,
+      "logits/rejected": -0.70703125,
+      "logps/chosen": -808.0,
+      "logps/rejected": -1504.0,
+      "loss": 0.0213,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -4.96875,
+      "rewards/margins": 7.5,
+      "rewards/rejected": -12.4375,
+      "step": 2800
+    },
+    {
+      "epoch": 2.9304029304029307,
+      "eval_logits/chosen": -4.0,
+      "eval_logits/rejected": -1.578125,
+      "eval_logps/chosen": -1056.0,
+      "eval_logps/rejected": -1248.0,
+      "eval_loss": 0.8120781183242798,
+      "eval_rewards/accuracies": 0.728515625,
+      "eval_rewards/chosen": -7.40625,
+      "eval_rewards/margins": 2.21875,
+      "eval_rewards/rejected": -9.625,
+      "eval_runtime": 46.7489,
+      "eval_samples_per_second": 42.782,
+      "eval_steps_per_second": 0.685,
+      "step": 2800
+    },
+    {
+      "epoch": 2.9408686551543695,
+      "grad_norm": 0.9031656092188057,
+      "learning_rate": 5.613151021490459e-08,
+      "logits/chosen": -5.65625,
+      "logits/rejected": -1.03125,
+      "logps/chosen": -864.0,
+      "logps/rejected": -1640.0,
+      "loss": 0.0147,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.3125,
+      "rewards/margins": 8.0625,
+      "rewards/rejected": -13.3125,
+      "step": 2810
+    },
+    {
+      "epoch": 2.9513343799058083,
+      "grad_norm": 0.9314231306373978,
+      "learning_rate": 3.7580289713159965e-08,
+      "logits/chosen": -5.8125,
+      "logits/rejected": -0.5703125,
+      "logps/chosen": -824.0,
+      "logps/rejected": -1576.0,
+      "loss": 0.0192,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -5.125,
+      "rewards/margins": 7.78125,
+      "rewards/rejected": -12.9375,
+      "step": 2820
+    },
+    {
+      "epoch": 2.9618001046572475,
+      "grad_norm": 0.9314993504280183,
+      "learning_rate": 2.2736006387927277e-08,
+      "logits/chosen": -5.875,
+      "logits/rejected": -1.359375,
+      "logps/chosen": -876.0,
+      "logps/rejected": -1560.0,
+      "loss": 0.0201,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -5.40625,
+      "rewards/margins": 7.28125,
+      "rewards/rejected": -12.6875,
+      "step": 2830
+    },
+    {
+      "epoch": 2.9722658294086868,
+      "grad_norm": 0.3117071415935163,
+      "learning_rate": 1.1600864625893604e-08,
+      "logits/chosen": -5.8125,
+      "logits/rejected": 0.1572265625,
+      "logps/chosen": -856.0,
+      "logps/rejected": -1552.0,
+      "loss": 0.0149,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.1875,
+      "rewards/margins": 7.53125,
+      "rewards/rejected": -12.75,
+      "step": 2840
+    },
+    {
+      "epoch": 2.9827315541601256,
+      "grad_norm": 0.7392544427480813,
+      "learning_rate": 4.176518003573549e-09,
+      "logits/chosen": -5.625,
+      "logits/rejected": -1.9765625,
+      "logps/chosen": -824.0,
+      "logps/rejected": -1584.0,
+      "loss": 0.021,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -5.03125,
+      "rewards/margins": 7.8125,
+      "rewards/rejected": -12.8125,
+      "step": 2850
+    },
+    {
+      "epoch": 2.9931972789115644,
+      "grad_norm": 0.552443918332781,
+      "learning_rate": 4.640690417528903e-10,
+      "logits/chosen": -5.09375,
+      "logits/rejected": -1.2265625,
+      "logps/chosen": -848.0,
+      "logps/rejected": -1520.0,
+      "loss": 0.0145,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -5.21875,
+      "rewards/margins": 7.03125,
+      "rewards/rejected": -12.25,
+      "step": 2860
+    },
+    {
+      "epoch": 2.998430141287284,
+      "step": 2865,
+      "total_flos": 0.0,
+      "train_loss": 0.26550976372930185,
+      "train_runtime": 11897.7673,
+      "train_samples_per_second": 15.415,
+      "train_steps_per_second": 0.241
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 2865,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}