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": 2.8475825688840812,
+      "learning_rate": 1.7421602787456448e-07,
+      "logits/chosen": -11.5,
+      "logits/rejected": -11.25,
+      "logps/chosen": -452.0,
+      "logps/rejected": -400.0,
+      "loss": 1.0,
+      "rewards/accuracies": 0.0,
+      "rewards/chosen": 0.0,
+      "rewards/margins": 0.0,
+      "rewards/rejected": 0.0,
+      "step": 1
+    },
+    {
+      "epoch": 0.010465724751439037,
+      "grad_norm": 2.397890719633437,
+      "learning_rate": 1.7421602787456445e-06,
+      "logits/chosen": -10.8125,
+      "logits/rejected": -10.875,
+      "logps/chosen": -300.0,
+      "logps/rejected": -266.0,
+      "loss": 0.9999,
+      "rewards/accuracies": 0.2083333283662796,
+      "rewards/chosen": -0.000377655029296875,
+      "rewards/margins": 0.0002574920654296875,
+      "rewards/rejected": -0.00063323974609375,
+      "step": 10
+    },
+    {
+      "epoch": 0.020931449502878074,
+      "grad_norm": 2.752402510886471,
+      "learning_rate": 3.484320557491289e-06,
+      "logits/chosen": -10.625,
+      "logits/rejected": -10.4375,
+      "logps/chosen": -328.0,
+      "logps/rejected": -312.0,
+      "loss": 0.9974,
+      "rewards/accuracies": 0.3187499940395355,
+      "rewards/chosen": 0.0025634765625,
+      "rewards/margins": 0.00146484375,
+      "rewards/rejected": 0.0010833740234375,
+      "step": 20
+    },
+    {
+      "epoch": 0.03139717425431711,
+      "grad_norm": 2.868324561366987,
+      "learning_rate": 5.226480836236934e-06,
+      "logits/chosen": -11.125,
+      "logits/rejected": -10.8125,
+      "logps/chosen": -318.0,
+      "logps/rejected": -312.0,
+      "loss": 0.9903,
+      "rewards/accuracies": 0.518750011920929,
+      "rewards/chosen": 0.003204345703125,
+      "rewards/margins": 0.005950927734375,
+      "rewards/rejected": -0.00274658203125,
+      "step": 30
+    },
+    {
+      "epoch": 0.04186289900575615,
+      "grad_norm": 2.8602768180285367,
+      "learning_rate": 6.968641114982578e-06,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -11.4375,
+      "logps/chosen": -342.0,
+      "logps/rejected": -316.0,
+      "loss": 0.9729,
+      "rewards/accuracies": 0.5562499761581421,
+      "rewards/chosen": -0.039794921875,
+      "rewards/margins": 0.02197265625,
+      "rewards/rejected": -0.0615234375,
+      "step": 40
+    },
+    {
+      "epoch": 0.052328623757195186,
+      "grad_norm": 3.1117827837725907,
+      "learning_rate": 8.710801393728225e-06,
+      "logits/chosen": -12.1875,
+      "logits/rejected": -11.875,
+      "logps/chosen": -336.0,
+      "logps/rejected": -290.0,
+      "loss": 0.9404,
+      "rewards/accuracies": 0.59375,
+      "rewards/chosen": -0.1943359375,
+      "rewards/margins": 0.05224609375,
+      "rewards/rejected": -0.2470703125,
+      "step": 50
+    },
+    {
+      "epoch": 0.06279434850863422,
+      "grad_norm": 3.9961805526252037,
+      "learning_rate": 1.0452961672473868e-05,
+      "logits/chosen": -12.0625,
+      "logits/rejected": -11.9375,
+      "logps/chosen": -412.0,
+      "logps/rejected": -380.0,
+      "loss": 0.8658,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.55078125,
+      "rewards/margins": 0.1259765625,
+      "rewards/rejected": -0.67578125,
+      "step": 60
+    },
+    {
+      "epoch": 0.07326007326007326,
+      "grad_norm": 5.214104482008826,
+      "learning_rate": 1.2195121951219513e-05,
+      "logits/chosen": -12.625,
+      "logits/rejected": -12.25,
+      "logps/chosen": -398.0,
+      "logps/rejected": -410.0,
+      "loss": 0.8331,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -0.9453125,
+      "rewards/margins": 0.294921875,
+      "rewards/rejected": -1.2421875,
+      "step": 70
+    },
+    {
+      "epoch": 0.0837257980115123,
+      "grad_norm": 3.6871376901657342,
+      "learning_rate": 1.3937282229965156e-05,
+      "logits/chosen": -13.375,
+      "logits/rejected": -13.5,
+      "logps/chosen": -424.0,
+      "logps/rejected": -400.0,
+      "loss": 0.8398,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -0.94140625,
+      "rewards/margins": 0.23828125,
+      "rewards/rejected": -1.1796875,
+      "step": 80
+    },
+    {
+      "epoch": 0.09419152276295134,
+      "grad_norm": 4.301763451013989,
+      "learning_rate": 1.56794425087108e-05,
+      "logits/chosen": -14.0,
+      "logits/rejected": -14.0,
+      "logps/chosen": -390.0,
+      "logps/rejected": -412.0,
+      "loss": 0.8041,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -1.0,
+      "rewards/margins": 0.1572265625,
+      "rewards/rejected": -1.15625,
+      "step": 90
+    },
+    {
+      "epoch": 0.10465724751439037,
+      "grad_norm": 3.910293569174661,
+      "learning_rate": 1.742160278745645e-05,
+      "logits/chosen": -13.8125,
+      "logits/rejected": -13.9375,
+      "logps/chosen": -432.0,
+      "logps/rejected": -426.0,
+      "loss": 0.7634,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -0.8125,
+      "rewards/margins": 0.34765625,
+      "rewards/rejected": -1.1640625,
+      "step": 100
+    },
+    {
+      "epoch": 0.10465724751439037,
+      "eval_logits/chosen": -13.0625,
+      "eval_logits/rejected": -12.9375,
+      "eval_logps/chosen": -392.0,
+      "eval_logps/rejected": -392.0,
+      "eval_loss": 0.7878281474113464,
+      "eval_rewards/accuracies": 0.640625,
+      "eval_rewards/chosen": -0.74609375,
+      "eval_rewards/margins": 0.283203125,
+      "eval_rewards/rejected": -1.03125,
+      "eval_runtime": 47.4243,
+      "eval_samples_per_second": 42.173,
+      "eval_steps_per_second": 0.675,
+      "step": 100
+    },
+    {
+      "epoch": 0.1151229722658294,
+      "grad_norm": 3.8169853140059726,
+      "learning_rate": 1.9163763066202093e-05,
+      "logits/chosen": -11.75,
+      "logits/rejected": -11.6875,
+      "logps/chosen": -378.0,
+      "logps/rejected": -388.0,
+      "loss": 0.8063,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.9140625,
+      "rewards/margins": 0.2578125,
+      "rewards/rejected": -1.171875,
+      "step": 110
+    },
+    {
+      "epoch": 0.12558869701726844,
+      "grad_norm": 4.514449625747369,
+      "learning_rate": 2.0905923344947736e-05,
+      "logits/chosen": -11.5625,
+      "logits/rejected": -11.0625,
+      "logps/chosen": -406.0,
+      "logps/rejected": -452.0,
+      "loss": 0.7375,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -0.9453125,
+      "rewards/margins": 0.353515625,
+      "rewards/rejected": -1.296875,
+      "step": 120
+    },
+    {
+      "epoch": 0.1360544217687075,
+      "grad_norm": 5.143567348752745,
+      "learning_rate": 2.264808362369338e-05,
+      "logits/chosen": -10.5625,
+      "logits/rejected": -10.0,
+      "logps/chosen": -480.0,
+      "logps/rejected": -486.0,
+      "loss": 0.7467,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -1.4609375,
+      "rewards/margins": 0.51953125,
+      "rewards/rejected": -1.984375,
+      "step": 130
+    },
+    {
+      "epoch": 0.14652014652014653,
+      "grad_norm": 5.664010794345305,
+      "learning_rate": 2.4390243902439026e-05,
+      "logits/chosen": -10.0625,
+      "logits/rejected": -9.875,
+      "logps/chosen": -496.0,
+      "logps/rejected": -516.0,
+      "loss": 0.7136,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -1.5703125,
+      "rewards/margins": 0.44921875,
+      "rewards/rejected": -2.015625,
+      "step": 140
+    },
+    {
+      "epoch": 0.15698587127158556,
+      "grad_norm": 5.498825774292673,
+      "learning_rate": 2.6132404181184672e-05,
+      "logits/chosen": -10.125,
+      "logits/rejected": -9.5625,
+      "logps/chosen": -436.0,
+      "logps/rejected": -416.0,
+      "loss": 0.8017,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -0.98828125,
+      "rewards/margins": 0.33984375,
+      "rewards/rejected": -1.328125,
+      "step": 150
+    },
+    {
+      "epoch": 0.1674515960230246,
+      "grad_norm": 5.022479531835636,
+      "learning_rate": 2.7874564459930312e-05,
+      "logits/chosen": -9.625,
+      "logits/rejected": -8.9375,
+      "logps/chosen": -492.0,
+      "logps/rejected": -458.0,
+      "loss": 0.7688,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -1.390625,
+      "rewards/margins": 0.33984375,
+      "rewards/rejected": -1.734375,
+      "step": 160
+    },
+    {
+      "epoch": 0.17791732077446362,
+      "grad_norm": 3.5885700337780744,
+      "learning_rate": 2.9616724738675962e-05,
+      "logits/chosen": -10.0,
+      "logits/rejected": -9.4375,
+      "logps/chosen": -470.0,
+      "logps/rejected": -474.0,
+      "loss": 0.7764,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -1.5078125,
+      "rewards/margins": 0.498046875,
+      "rewards/rejected": -2.0,
+      "step": 170
+    },
+    {
+      "epoch": 0.18838304552590268,
+      "grad_norm": 4.131766893109537,
+      "learning_rate": 3.13588850174216e-05,
+      "logits/chosen": -12.0,
+      "logits/rejected": -11.6875,
+      "logps/chosen": -430.0,
+      "logps/rejected": -460.0,
+      "loss": 0.7533,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -1.4375,
+      "rewards/margins": 0.421875,
+      "rewards/rejected": -1.859375,
+      "step": 180
+    },
+    {
+      "epoch": 0.1988487702773417,
+      "grad_norm": 4.6130325259501666,
+      "learning_rate": 3.310104529616725e-05,
+      "logits/chosen": -13.0625,
+      "logits/rejected": -12.8125,
+      "logps/chosen": -466.0,
+      "logps/rejected": -516.0,
+      "loss": 0.721,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -1.8671875,
+      "rewards/margins": 0.53515625,
+      "rewards/rejected": -2.40625,
+      "step": 190
+    },
+    {
+      "epoch": 0.20931449502878074,
+      "grad_norm": 3.9778280958613235,
+      "learning_rate": 3.48432055749129e-05,
+      "logits/chosen": -12.9375,
+      "logits/rejected": -12.6875,
+      "logps/chosen": -478.0,
+      "logps/rejected": -444.0,
+      "loss": 0.7498,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -1.15625,
+      "rewards/margins": 0.4296875,
+      "rewards/rejected": -1.5859375,
+      "step": 200
+    },
+    {
+      "epoch": 0.20931449502878074,
+      "eval_logits/chosen": -12.75,
+      "eval_logits/rejected": -12.4375,
+      "eval_logps/chosen": -436.0,
+      "eval_logps/rejected": -444.0,
+      "eval_loss": 0.7467656135559082,
+      "eval_rewards/accuracies": 0.671875,
+      "eval_rewards/chosen": -1.171875,
+      "eval_rewards/margins": 0.380859375,
+      "eval_rewards/rejected": -1.5546875,
+      "eval_runtime": 46.6523,
+      "eval_samples_per_second": 42.87,
+      "eval_steps_per_second": 0.686,
+      "step": 200
+    },
+    {
+      "epoch": 0.21978021978021978,
+      "grad_norm": 3.906338719361765,
+      "learning_rate": 3.6585365853658535e-05,
+      "logits/chosen": -13.1875,
+      "logits/rejected": -12.9375,
+      "logps/chosen": -512.0,
+      "logps/rejected": -532.0,
+      "loss": 0.758,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -1.8125,
+      "rewards/margins": 0.39453125,
+      "rewards/rejected": -2.203125,
+      "step": 210
+    },
+    {
+      "epoch": 0.2302459445316588,
+      "grad_norm": 3.9597077486155623,
+      "learning_rate": 3.8327526132404185e-05,
+      "logits/chosen": -14.375,
+      "logits/rejected": -14.1875,
+      "logps/chosen": -464.0,
+      "logps/rejected": -422.0,
+      "loss": 0.7521,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -1.28125,
+      "rewards/margins": 0.36328125,
+      "rewards/rejected": -1.640625,
+      "step": 220
+    },
+    {
+      "epoch": 0.24071166928309787,
+      "grad_norm": 6.511183839435885,
+      "learning_rate": 4.006968641114983e-05,
+      "logits/chosen": -14.3125,
+      "logits/rejected": -14.0,
+      "logps/chosen": -472.0,
+      "logps/rejected": -486.0,
+      "loss": 0.7275,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -1.3828125,
+      "rewards/margins": 0.43359375,
+      "rewards/rejected": -1.8125,
+      "step": 230
+    },
+    {
+      "epoch": 0.25117739403453687,
+      "grad_norm": 8.89679330754679,
+      "learning_rate": 4.181184668989547e-05,
+      "logits/chosen": -12.6875,
+      "logits/rejected": -12.25,
+      "logps/chosen": -476.0,
+      "logps/rejected": -502.0,
+      "loss": 0.7198,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -1.6328125,
+      "rewards/margins": 0.5,
+      "rewards/rejected": -2.140625,
+      "step": 240
+    },
+    {
+      "epoch": 0.2616431187859759,
+      "grad_norm": 4.560940999024058,
+      "learning_rate": 4.3554006968641115e-05,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -11.1875,
+      "logps/chosen": -552.0,
+      "logps/rejected": -568.0,
+      "loss": 0.6682,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -2.265625,
+      "rewards/margins": 0.5234375,
+      "rewards/rejected": -2.796875,
+      "step": 250
+    },
+    {
+      "epoch": 0.272108843537415,
+      "grad_norm": 4.579924517823466,
+      "learning_rate": 4.529616724738676e-05,
+      "logits/chosen": -12.25,
+      "logits/rejected": -12.0625,
+      "logps/chosen": -648.0,
+      "logps/rejected": -656.0,
+      "loss": 0.7925,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -3.09375,
+      "rewards/margins": 0.322265625,
+      "rewards/rejected": -3.421875,
+      "step": 260
+    },
+    {
+      "epoch": 0.282574568288854,
+      "grad_norm": 4.094411278908984,
+      "learning_rate": 4.703832752613241e-05,
+      "logits/chosen": -13.0,
+      "logits/rejected": -12.8125,
+      "logps/chosen": -536.0,
+      "logps/rejected": -500.0,
+      "loss": 0.7556,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -1.796875,
+      "rewards/margins": 0.328125,
+      "rewards/rejected": -2.125,
+      "step": 270
+    },
+    {
+      "epoch": 0.29304029304029305,
+      "grad_norm": 3.6640419519147036,
+      "learning_rate": 4.878048780487805e-05,
+      "logits/chosen": -11.9375,
+      "logits/rejected": -11.4375,
+      "logps/chosen": -510.0,
+      "logps/rejected": -528.0,
+      "loss": 0.7331,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -1.6328125,
+      "rewards/margins": 0.45703125,
+      "rewards/rejected": -2.09375,
+      "step": 280
+    },
+    {
+      "epoch": 0.3035060177917321,
+      "grad_norm": 4.432208202540975,
+      "learning_rate": 4.999983293481418e-05,
+      "logits/chosen": -11.6875,
+      "logits/rejected": -11.3125,
+      "logps/chosen": -508.0,
+      "logps/rejected": -498.0,
+      "loss": 0.7537,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -1.8671875,
+      "rewards/margins": 0.44921875,
+      "rewards/rejected": -2.3125,
+      "step": 290
+    },
+    {
+      "epoch": 0.3139717425431711,
+      "grad_norm": 3.079924620893602,
+      "learning_rate": 4.9996862949181595e-05,
+      "logits/chosen": -12.8125,
+      "logits/rejected": -12.625,
+      "logps/chosen": -532.0,
+      "logps/rejected": -552.0,
+      "loss": 0.8142,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -2.15625,
+      "rewards/margins": 0.390625,
+      "rewards/rejected": -2.546875,
+      "step": 300
+    },
+    {
+      "epoch": 0.3139717425431711,
+      "eval_logits/chosen": -14.8125,
+      "eval_logits/rejected": -14.75,
+      "eval_logps/chosen": -504.0,
+      "eval_logps/rejected": -516.0,
+      "eval_loss": 0.7466250061988831,
+      "eval_rewards/accuracies": 0.69140625,
+      "eval_rewards/chosen": -1.859375,
+      "eval_rewards/margins": 0.4140625,
+      "eval_rewards/rejected": -2.28125,
+      "eval_runtime": 46.6549,
+      "eval_samples_per_second": 42.868,
+      "eval_steps_per_second": 0.686,
+      "step": 300
+    },
+    {
+      "epoch": 0.32443746729461015,
+      "grad_norm": 4.414085329471303,
+      "learning_rate": 4.9990180911526896e-05,
+      "logits/chosen": -15.0,
+      "logits/rejected": -14.9375,
+      "logps/chosen": -540.0,
+      "logps/rejected": -502.0,
+      "loss": 0.7366,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -1.9765625,
+      "rewards/margins": 0.5234375,
+      "rewards/rejected": -2.5,
+      "step": 310
+    },
+    {
+      "epoch": 0.3349031920460492,
+      "grad_norm": 3.1753848654603054,
+      "learning_rate": 4.997978781413745e-05,
+      "logits/chosen": -13.5625,
+      "logits/rejected": -13.3125,
+      "logps/chosen": -502.0,
+      "logps/rejected": -536.0,
+      "loss": 0.7602,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -2.09375,
+      "rewards/margins": 0.470703125,
+      "rewards/rejected": -2.5625,
+      "step": 320
+    },
+    {
+      "epoch": 0.3453689167974882,
+      "grad_norm": 3.6939699881978627,
+      "learning_rate": 4.996568520039565e-05,
+      "logits/chosen": -13.75,
+      "logits/rejected": -13.6875,
+      "logps/chosen": -490.0,
+      "logps/rejected": -486.0,
+      "loss": 0.7346,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -1.8984375,
+      "rewards/margins": 0.328125,
+      "rewards/rejected": -2.234375,
+      "step": 330
+    },
+    {
+      "epoch": 0.35583464154892724,
+      "grad_norm": 3.826110277593326,
+      "learning_rate": 4.994787516454973e-05,
+      "logits/chosen": -13.1875,
+      "logits/rejected": -12.875,
+      "logps/chosen": -552.0,
+      "logps/rejected": -548.0,
+      "loss": 0.7699,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -2.40625,
+      "rewards/margins": 0.41015625,
+      "rewards/rejected": -2.8125,
+      "step": 340
+    },
+    {
+      "epoch": 0.3663003663003663,
+      "grad_norm": 3.373611070367837,
+      "learning_rate": 4.992636035140273e-05,
+      "logits/chosen": -13.0,
+      "logits/rejected": -12.8125,
+      "logps/chosen": -568.0,
+      "logps/rejected": -560.0,
+      "loss": 0.7368,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -2.21875,
+      "rewards/margins": 0.5234375,
+      "rewards/rejected": -2.75,
+      "step": 350
+    },
+    {
+      "epoch": 0.37676609105180536,
+      "grad_norm": 3.1696240096213386,
+      "learning_rate": 4.99011439559198e-05,
+      "logits/chosen": -13.0625,
+      "logits/rejected": -12.5625,
+      "logps/chosen": -532.0,
+      "logps/rejected": -544.0,
+      "loss": 0.7123,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -2.09375,
+      "rewards/margins": 0.45703125,
+      "rewards/rejected": -2.546875,
+      "step": 360
+    },
+    {
+      "epoch": 0.3872318158032444,
+      "grad_norm": 3.705761144844104,
+      "learning_rate": 4.987222972275369e-05,
+      "logits/chosen": -12.4375,
+      "logits/rejected": -12.25,
+      "logps/chosen": -544.0,
+      "logps/rejected": -548.0,
+      "loss": 0.7155,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -2.328125,
+      "rewards/margins": 0.431640625,
+      "rewards/rejected": -2.75,
+      "step": 370
+    },
+    {
+      "epoch": 0.3976975405546834,
+      "grad_norm": 3.368674068863961,
+      "learning_rate": 4.983962194568872e-05,
+      "logits/chosen": -13.625,
+      "logits/rejected": -13.3125,
+      "logps/chosen": -520.0,
+      "logps/rejected": -544.0,
+      "loss": 0.7376,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -2.203125,
+      "rewards/margins": 0.3984375,
+      "rewards/rejected": -2.59375,
+      "step": 380
+    },
+    {
+      "epoch": 0.40816326530612246,
+      "grad_norm": 3.447195402745154,
+      "learning_rate": 4.9803325467003084e-05,
+      "logits/chosen": -14.1875,
+      "logits/rejected": -14.0,
+      "logps/chosen": -486.0,
+      "logps/rejected": -504.0,
+      "loss": 0.7664,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -1.59375,
+      "rewards/margins": 0.400390625,
+      "rewards/rejected": -2.0,
+      "step": 390
+    },
+    {
+      "epoch": 0.4186289900575615,
+      "grad_norm": 4.748043956213455,
+      "learning_rate": 4.9763345676749816e-05,
+      "logits/chosen": -14.125,
+      "logits/rejected": -13.875,
+      "logps/chosen": -536.0,
+      "logps/rejected": -564.0,
+      "loss": 0.7764,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -2.1875,
+      "rewards/margins": 0.4765625,
+      "rewards/rejected": -2.65625,
+      "step": 400
+    },
+    {
+      "epoch": 0.4186289900575615,
+      "eval_logits/chosen": -14.5625,
+      "eval_logits/rejected": -14.4375,
+      "eval_logps/chosen": -516.0,
+      "eval_logps/rejected": -528.0,
+      "eval_loss": 0.7499062418937683,
+      "eval_rewards/accuracies": 0.669921875,
+      "eval_rewards/chosen": -1.96875,
+      "eval_rewards/margins": 0.431640625,
+      "eval_rewards/rejected": -2.40625,
+      "eval_runtime": 46.6373,
+      "eval_samples_per_second": 42.884,
+      "eval_steps_per_second": 0.686,
+      "step": 400
+    },
+    {
+      "epoch": 0.4290947148090005,
+      "grad_norm": 3.3031814132745687,
+      "learning_rate": 4.9719688511956374e-05,
+      "logits/chosen": -13.4375,
+      "logits/rejected": -13.25,
+      "logps/chosen": -564.0,
+      "logps/rejected": -552.0,
+      "loss": 0.7842,
+      "rewards/accuracies": 0.581250011920929,
+      "rewards/chosen": -2.390625,
+      "rewards/margins": 0.34765625,
+      "rewards/rejected": -2.734375,
+      "step": 410
+    },
+    {
+      "epoch": 0.43956043956043955,
+      "grad_norm": 2.8593482287450147,
+      "learning_rate": 4.967236045574294e-05,
+      "logits/chosen": -13.1875,
+      "logits/rejected": -12.9375,
+      "logps/chosen": -568.0,
+      "logps/rejected": -560.0,
+      "loss": 0.7626,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -2.328125,
+      "rewards/margins": 0.431640625,
+      "rewards/rejected": -2.75,
+      "step": 420
+    },
+    {
+      "epoch": 0.4500261643118786,
+      "grad_norm": 3.091399616806401,
+      "learning_rate": 4.96213685363597e-05,
+      "logits/chosen": -14.0,
+      "logits/rejected": -13.75,
+      "logps/chosen": -532.0,
+      "logps/rejected": -564.0,
+      "loss": 0.7392,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -2.296875,
+      "rewards/margins": 0.458984375,
+      "rewards/rejected": -2.75,
+      "step": 430
+    },
+    {
+      "epoch": 0.4604918890633176,
+      "grad_norm": 3.831642848727746,
+      "learning_rate": 4.956672032614315e-05,
+      "logits/chosen": -15.375,
+      "logits/rejected": -15.25,
+      "logps/chosen": -576.0,
+      "logps/rejected": -548.0,
+      "loss": 0.7265,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -2.34375,
+      "rewards/margins": 0.375,
+      "rewards/rejected": -2.71875,
+      "step": 440
+    },
+    {
+      "epoch": 0.47095761381475665,
+      "grad_norm": 2.9932508899527352,
+      "learning_rate": 4.950842394039157e-05,
+      "logits/chosen": -15.3125,
+      "logits/rejected": -15.0625,
+      "logps/chosen": -556.0,
+      "logps/rejected": -596.0,
+      "loss": 0.7151,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": -2.15625,
+      "rewards/margins": 0.6484375,
+      "rewards/rejected": -2.8125,
+      "step": 450
+    },
+    {
+      "epoch": 0.48142333856619574,
+      "grad_norm": 3.250235478086617,
+      "learning_rate": 4.944648803615994e-05,
+      "logits/chosen": -13.875,
+      "logits/rejected": -13.375,
+      "logps/chosen": -524.0,
+      "logps/rejected": -544.0,
+      "loss": 0.7213,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -2.1875,
+      "rewards/margins": 0.53125,
+      "rewards/rejected": -2.71875,
+      "step": 460
+    },
+    {
+      "epoch": 0.49188906331763477,
+      "grad_norm": 3.040265502991071,
+      "learning_rate": 4.938092181097433e-05,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -11.0625,
+      "logps/chosen": -532.0,
+      "logps/rejected": -536.0,
+      "loss": 0.6462,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -2.015625,
+      "rewards/margins": 0.5859375,
+      "rewards/rejected": -2.609375,
+      "step": 470
+    },
+    {
+      "epoch": 0.5023547880690737,
+      "grad_norm": 3.273062248095332,
+      "learning_rate": 4.931173500146607e-05,
+      "logits/chosen": -12.3125,
+      "logits/rejected": -11.75,
+      "logps/chosen": -548.0,
+      "logps/rejected": -576.0,
+      "loss": 0.6937,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -2.328125,
+      "rewards/margins": 0.609375,
+      "rewards/rejected": -2.921875,
+      "step": 480
+    },
+    {
+      "epoch": 0.5128205128205128,
+      "grad_norm": 3.439882213435931,
+      "learning_rate": 4.923893788192582e-05,
+      "logits/chosen": -9.3125,
+      "logits/rejected": -7.5,
+      "logps/chosen": -680.0,
+      "logps/rejected": -700.0,
+      "loss": 0.6598,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -3.09375,
+      "rewards/margins": 1.046875,
+      "rewards/rejected": -4.15625,
+      "step": 490
+    },
+    {
+      "epoch": 0.5232862375719518,
+      "grad_norm": 3.7499997647243375,
+      "learning_rate": 4.9162541262777906e-05,
+      "logits/chosen": -10.875,
+      "logits/rejected": -10.375,
+      "logps/chosen": -552.0,
+      "logps/rejected": -548.0,
+      "loss": 0.731,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -2.484375,
+      "rewards/margins": 0.44140625,
+      "rewards/rejected": -2.921875,
+      "step": 500
+    },
+    {
+      "epoch": 0.5232862375719518,
+      "eval_logits/chosen": -11.0,
+      "eval_logits/rejected": -10.5,
+      "eval_logps/chosen": -560.0,
+      "eval_logps/rejected": -576.0,
+      "eval_loss": 0.7239687442779541,
+      "eval_rewards/accuracies": 0.69140625,
+      "eval_rewards/chosen": -2.421875,
+      "eval_rewards/margins": 0.4375,
+      "eval_rewards/rejected": -2.859375,
+      "eval_runtime": 46.6378,
+      "eval_samples_per_second": 42.884,
+      "eval_steps_per_second": 0.686,
+      "step": 500
+    },
+    {
+      "epoch": 0.533751962323391,
+      "grad_norm": 2.9130806322627953,
+      "learning_rate": 4.90825564889749e-05,
+      "logits/chosen": -10.75,
+      "logits/rejected": -10.25,
+      "logps/chosen": -580.0,
+      "logps/rejected": -600.0,
+      "loss": 0.7211,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -2.4375,
+      "rewards/margins": 0.5859375,
+      "rewards/rejected": -3.015625,
+      "step": 510
+    },
+    {
+      "epoch": 0.54421768707483,
+      "grad_norm": 3.6896279610655496,
+      "learning_rate": 4.899899543831288e-05,
+      "logits/chosen": -9.4375,
+      "logits/rejected": -9.0,
+      "logps/chosen": -592.0,
+      "logps/rejected": -604.0,
+      "loss": 0.7049,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -2.703125,
+      "rewards/margins": 0.47265625,
+      "rewards/rejected": -3.171875,
+      "step": 520
+    },
+    {
+      "epoch": 0.554683411826269,
+      "grad_norm": 3.4728058007543847,
+      "learning_rate": 4.891187051966765e-05,
+      "logits/chosen": -13.25,
+      "logits/rejected": -13.0625,
+      "logps/chosen": -500.0,
+      "logps/rejected": -520.0,
+      "loss": 0.7397,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -1.8046875,
+      "rewards/margins": 0.5,
+      "rewards/rejected": -2.296875,
+      "step": 530
+    },
+    {
+      "epoch": 0.565149136577708,
+      "grad_norm": 3.3052043677247167,
+      "learning_rate": 4.882119467115191e-05,
+      "logits/chosen": -13.375,
+      "logits/rejected": -12.9375,
+      "logps/chosen": -524.0,
+      "logps/rejected": -528.0,
+      "loss": 0.7365,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -1.7578125,
+      "rewards/margins": 0.578125,
+      "rewards/rejected": -2.328125,
+      "step": 540
+    },
+    {
+      "epoch": 0.5756148613291471,
+      "grad_norm": 3.194823747443967,
+      "learning_rate": 4.872698135819402e-05,
+      "logits/chosen": -13.0,
+      "logits/rejected": -12.375,
+      "logps/chosen": -540.0,
+      "logps/rejected": -528.0,
+      "loss": 0.7289,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -2.265625,
+      "rewards/margins": 0.5,
+      "rewards/rejected": -2.765625,
+      "step": 550
+    },
+    {
+      "epoch": 0.5860805860805861,
+      "grad_norm": 3.5866537006371475,
+      "learning_rate": 4.862924457153831e-05,
+      "logits/chosen": -13.0,
+      "logits/rejected": -12.125,
+      "logps/chosen": -548.0,
+      "logps/rejected": -568.0,
+      "loss": 0.7281,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -2.171875,
+      "rewards/margins": 0.609375,
+      "rewards/rejected": -2.78125,
+      "step": 560
+    },
+    {
+      "epoch": 0.5965463108320251,
+      "grad_norm": 3.382375979724736,
+      "learning_rate": 4.8527998825167484e-05,
+      "logits/chosen": -13.375,
+      "logits/rejected": -12.375,
+      "logps/chosen": -564.0,
+      "logps/rejected": -576.0,
+      "loss": 0.6935,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -2.328125,
+      "rewards/margins": 0.44921875,
+      "rewards/rejected": -2.78125,
+      "step": 570
+    },
+    {
+      "epoch": 0.6070120355834642,
+      "grad_norm": 3.600303579271625,
+      "learning_rate": 4.842325915414728e-05,
+      "logits/chosen": -12.0625,
+      "logits/rejected": -11.25,
+      "logps/chosen": -556.0,
+      "logps/rejected": -556.0,
+      "loss": 0.7107,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -2.5625,
+      "rewards/margins": 0.60546875,
+      "rewards/rejected": -3.171875,
+      "step": 580
+    },
+    {
+      "epoch": 0.6174777603349032,
+      "grad_norm": 3.433506974104068,
+      "learning_rate": 4.831504111239373e-05,
+      "logits/chosen": -12.0,
+      "logits/rejected": -11.125,
+      "logps/chosen": -616.0,
+      "logps/rejected": -644.0,
+      "loss": 0.674,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -2.90625,
+      "rewards/margins": 0.5703125,
+      "rewards/rejected": -3.46875,
+      "step": 590
+    },
+    {
+      "epoch": 0.6279434850863422,
+      "grad_norm": 3.322477709818826,
+      "learning_rate": 4.8203360770363425e-05,
+      "logits/chosen": -10.5,
+      "logits/rejected": -9.6875,
+      "logps/chosen": -652.0,
+      "logps/rejected": -692.0,
+      "loss": 0.665,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -3.390625,
+      "rewards/margins": 0.67578125,
+      "rewards/rejected": -4.0625,
+      "step": 600
+    },
+    {
+      "epoch": 0.6279434850863422,
+      "eval_logits/chosen": -10.75,
+      "eval_logits/rejected": -10.0625,
+      "eval_logps/chosen": -660.0,
+      "eval_logps/rejected": -696.0,
+      "eval_loss": 0.7044687271118164,
+      "eval_rewards/accuracies": 0.697265625,
+      "eval_rewards/chosen": -3.40625,
+      "eval_rewards/margins": 0.66796875,
+      "eval_rewards/rejected": -4.0625,
+      "eval_runtime": 46.6452,
+      "eval_samples_per_second": 42.877,
+      "eval_steps_per_second": 0.686,
+      "step": 600
+    },
+    {
+      "epoch": 0.6384092098377813,
+      "grad_norm": 2.926085485760497,
+      "learning_rate": 4.808823471266702e-05,
+      "logits/chosen": -11.5625,
+      "logits/rejected": -10.75,
+      "logps/chosen": -696.0,
+      "logps/rejected": -676.0,
+      "loss": 0.7366,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -3.359375,
+      "rewards/margins": 0.609375,
+      "rewards/rejected": -3.96875,
+      "step": 610
+    },
+    {
+      "epoch": 0.6488749345892203,
+      "grad_norm": 3.106571795002711,
+      "learning_rate": 4.796968003560639e-05,
+      "logits/chosen": -12.4375,
+      "logits/rejected": -11.9375,
+      "logps/chosen": -572.0,
+      "logps/rejected": -588.0,
+      "loss": 0.6823,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -2.578125,
+      "rewards/margins": 0.62890625,
+      "rewards/rejected": -3.203125,
+      "step": 620
+    },
+    {
+      "epoch": 0.6593406593406593,
+      "grad_norm": 2.8203312037989754,
+      "learning_rate": 4.7847714344635864e-05,
+      "logits/chosen": -13.0,
+      "logits/rejected": -12.6875,
+      "logps/chosen": -556.0,
+      "logps/rejected": -564.0,
+      "loss": 0.6764,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -2.390625,
+      "rewards/margins": 0.59375,
+      "rewards/rejected": -2.984375,
+      "step": 630
+    },
+    {
+      "epoch": 0.6698063840920984,
+      "grad_norm": 4.107968975168374,
+      "learning_rate": 4.772235575174776e-05,
+      "logits/chosen": -12.6875,
+      "logits/rejected": -12.4375,
+      "logps/chosen": -608.0,
+      "logps/rejected": -616.0,
+      "loss": 0.6929,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -2.9375,
+      "rewards/margins": 0.60546875,
+      "rewards/rejected": -3.53125,
+      "step": 640
+    },
+    {
+      "epoch": 0.6802721088435374,
+      "grad_norm": 3.290238614703423,
+      "learning_rate": 4.7593622872782754e-05,
+      "logits/chosen": -12.75,
+      "logits/rejected": -12.3125,
+      "logps/chosen": -580.0,
+      "logps/rejected": -632.0,
+      "loss": 0.7006,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": -2.71875,
+      "rewards/margins": 0.5546875,
+      "rewards/rejected": -3.265625,
+      "step": 650
+    },
+    {
+      "epoch": 0.6907378335949764,
+      "grad_norm": 3.093707244283786,
+      "learning_rate": 4.746153482466542e-05,
+      "logits/chosen": -13.125,
+      "logits/rejected": -12.5,
+      "logps/chosen": -608.0,
+      "logps/rejected": -632.0,
+      "loss": 0.6775,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -2.71875,
+      "rewards/margins": 0.671875,
+      "rewards/rejected": -3.390625,
+      "step": 660
+    },
+    {
+      "epoch": 0.7012035583464155,
+      "grad_norm": 3.095200157675764,
+      "learning_rate": 4.732611122256537e-05,
+      "logits/chosen": -12.3125,
+      "logits/rejected": -11.6875,
+      "logps/chosen": -692.0,
+      "logps/rejected": -732.0,
+      "loss": 0.6501,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -3.484375,
+      "rewards/margins": 0.59375,
+      "rewards/rejected": -4.0625,
+      "step": 670
+    },
+    {
+      "epoch": 0.7116692830978545,
+      "grad_norm": 3.5935034838953723,
+      "learning_rate": 4.718737217698435e-05,
+      "logits/chosen": -12.375,
+      "logits/rejected": -11.9375,
+      "logps/chosen": -724.0,
+      "logps/rejected": -744.0,
+      "loss": 0.7336,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -3.546875,
+      "rewards/margins": 0.58203125,
+      "rewards/rejected": -4.125,
+      "step": 680
+    },
+    {
+      "epoch": 0.7221350078492935,
+      "grad_norm": 3.7536259852972926,
+      "learning_rate": 4.704533829076982e-05,
+      "logits/chosen": -13.4375,
+      "logits/rejected": -12.6875,
+      "logps/chosen": -620.0,
+      "logps/rejected": -640.0,
+      "loss": 0.6088,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -3.0,
+      "rewards/margins": 0.73828125,
+      "rewards/rejected": -3.75,
+      "step": 690
+    },
+    {
+      "epoch": 0.7326007326007326,
+      "grad_norm": 3.8449275592217815,
+      "learning_rate": 4.6900030656055455e-05,
+      "logits/chosen": -13.5625,
+      "logits/rejected": -13.125,
+      "logps/chosen": -568.0,
+      "logps/rejected": -596.0,
+      "loss": 0.6806,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -2.671875,
+      "rewards/margins": 0.609375,
+      "rewards/rejected": -3.28125,
+      "step": 700
+    },
+    {
+      "epoch": 0.7326007326007326,
+      "eval_logits/chosen": -13.875,
+      "eval_logits/rejected": -13.4375,
+      "eval_logps/chosen": -568.0,
+      "eval_logps/rejected": -604.0,
+      "eval_loss": 0.6912265419960022,
+      "eval_rewards/accuracies": 0.70703125,
+      "eval_rewards/chosen": -2.515625,
+      "eval_rewards/margins": 0.65234375,
+      "eval_rewards/rejected": -3.15625,
+      "eval_runtime": 46.6429,
+      "eval_samples_per_second": 42.879,
+      "eval_steps_per_second": 0.686,
+      "step": 700
+    },
+    {
+      "epoch": 0.7430664573521716,
+      "grad_norm": 3.230580134741238,
+      "learning_rate": 4.6751470851128906e-05,
+      "logits/chosen": -13.5625,
+      "logits/rejected": -13.0,
+      "logps/chosen": -548.0,
+      "logps/rejected": -572.0,
+      "loss": 0.7354,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -2.484375,
+      "rewards/margins": 0.53125,
+      "rewards/rejected": -3.015625,
+      "step": 710
+    },
+    {
+      "epoch": 0.7535321821036107,
+      "grad_norm": 3.3056424527009165,
+      "learning_rate": 4.6599680937227454e-05,
+      "logits/chosen": -13.5625,
+      "logits/rejected": -12.9375,
+      "logps/chosen": -576.0,
+      "logps/rejected": -568.0,
+      "loss": 0.699,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -2.375,
+      "rewards/margins": 0.5078125,
+      "rewards/rejected": -2.875,
+      "step": 720
+    },
+    {
+      "epoch": 0.7639979068550498,
+      "grad_norm": 3.496301294606076,
+      "learning_rate": 4.644468345526183e-05,
+      "logits/chosen": -12.0,
+      "logits/rejected": -11.5,
+      "logps/chosen": -636.0,
+      "logps/rejected": -628.0,
+      "loss": 0.7225,
+      "rewards/accuracies": 0.59375,
+      "rewards/chosen": -3.203125,
+      "rewards/margins": 0.39453125,
+      "rewards/rejected": -3.59375,
+      "step": 730
+    },
+    {
+      "epoch": 0.7744636316064888,
+      "grad_norm": 3.3005907062778554,
+      "learning_rate": 4.628650142246894e-05,
+      "logits/chosen": -12.875,
+      "logits/rejected": -12.4375,
+      "logps/chosen": -596.0,
+      "logps/rejected": -620.0,
+      "loss": 0.6622,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -3.015625,
+      "rewards/margins": 0.546875,
+      "rewards/rejected": -3.5625,
+      "step": 740
+    },
+    {
+      "epoch": 0.7849293563579278,
+      "grad_norm": 3.524578086257254,
+      "learning_rate": 4.6125158328993765e-05,
+      "logits/chosen": -13.1875,
+      "logits/rejected": -12.5,
+      "logps/chosen": -576.0,
+      "logps/rejected": -636.0,
+      "loss": 0.631,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -2.921875,
+      "rewards/margins": 0.69140625,
+      "rewards/rejected": -3.609375,
+      "step": 750
+    },
+    {
+      "epoch": 0.7953950811093669,
+      "grad_norm": 3.62167855313147,
+      "learning_rate": 4.5960678134401064e-05,
+      "logits/chosen": -12.25,
+      "logits/rejected": -11.25,
+      "logps/chosen": -612.0,
+      "logps/rejected": -668.0,
+      "loss": 0.6737,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -3.28125,
+      "rewards/margins": 0.6796875,
+      "rewards/rejected": -3.96875,
+      "step": 760
+    },
+    {
+      "epoch": 0.8058608058608059,
+      "grad_norm": 2.8873464894931185,
+      "learning_rate": 4.579308526411734e-05,
+      "logits/chosen": -13.875,
+      "logits/rejected": -13.125,
+      "logps/chosen": -636.0,
+      "logps/rejected": -696.0,
+      "loss": 0.6542,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -2.859375,
+      "rewards/margins": 0.46875,
+      "rewards/rejected": -3.328125,
+      "step": 770
+    },
+    {
+      "epoch": 0.8163265306122449,
+      "grad_norm": 3.4139012330967025,
+      "learning_rate": 4.562240460580367e-05,
+      "logits/chosen": -13.8125,
+      "logits/rejected": -13.375,
+      "logps/chosen": -600.0,
+      "logps/rejected": -596.0,
+      "loss": 0.6885,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -2.375,
+      "rewards/margins": 0.6171875,
+      "rewards/rejected": -2.984375,
+      "step": 780
+    },
+    {
+      "epoch": 0.826792255363684,
+      "grad_norm": 3.3180761786786035,
+      "learning_rate": 4.544866150565985e-05,
+      "logits/chosen": -12.375,
+      "logits/rejected": -11.5,
+      "logps/chosen": -612.0,
+      "logps/rejected": -616.0,
+      "loss": 0.7408,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -2.796875,
+      "rewards/margins": 0.59765625,
+      "rewards/rejected": -3.390625,
+      "step": 790
+    },
+    {
+      "epoch": 0.837257980115123,
+      "grad_norm": 3.7265337457481467,
+      "learning_rate": 4.527188176466051e-05,
+      "logits/chosen": -13.5,
+      "logits/rejected": -13.0,
+      "logps/chosen": -540.0,
+      "logps/rejected": -588.0,
+      "loss": 0.6597,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": -2.25,
+      "rewards/margins": 0.65625,
+      "rewards/rejected": -2.90625,
+      "step": 800
+    },
+    {
+      "epoch": 0.837257980115123,
+      "eval_logits/chosen": -13.5,
+      "eval_logits/rejected": -13.3125,
+      "eval_logps/chosen": -548.0,
+      "eval_logps/rejected": -576.0,
+      "eval_loss": 0.7086874842643738,
+      "eval_rewards/accuracies": 0.677734375,
+      "eval_rewards/chosen": -2.296875,
+      "eval_rewards/margins": 0.56640625,
+      "eval_rewards/rejected": -2.859375,
+      "eval_runtime": 46.6675,
+      "eval_samples_per_second": 42.856,
+      "eval_steps_per_second": 0.686,
+      "step": 800
+    },
+    {
+      "epoch": 0.847723704866562,
+      "grad_norm": 3.4090033855198416,
+      "learning_rate": 4.509209163472357e-05,
+      "logits/chosen": -12.625,
+      "logits/rejected": -12.25,
+      "logps/chosen": -604.0,
+      "logps/rejected": -588.0,
+      "loss": 0.7048,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -2.625,
+      "rewards/margins": 0.482421875,
+      "rewards/rejected": -3.109375,
+      "step": 810
+    },
+    {
+      "epoch": 0.858189429618001,
+      "grad_norm": 3.666681054241992,
+      "learning_rate": 4.4909317814811896e-05,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -10.5625,
+      "logps/chosen": -624.0,
+      "logps/rejected": -652.0,
+      "loss": 0.7103,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -2.6875,
+      "rewards/margins": 0.6328125,
+      "rewards/rejected": -3.3125,
+      "step": 820
+    },
+    {
+      "epoch": 0.8686551543694401,
+      "grad_norm": 3.295178039577545,
+      "learning_rate": 4.472358744696845e-05,
+      "logits/chosen": -12.375,
+      "logits/rejected": -11.75,
+      "logps/chosen": -548.0,
+      "logps/rejected": -584.0,
+      "loss": 0.7062,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -2.46875,
+      "rewards/margins": 0.57421875,
+      "rewards/rejected": -3.03125,
+      "step": 830
+    },
+    {
+      "epoch": 0.8791208791208791,
+      "grad_norm": 3.7609472561769777,
+      "learning_rate": 4.4534928112285645e-05,
+      "logits/chosen": -11.0,
+      "logits/rejected": -9.5,
+      "logps/chosen": -588.0,
+      "logps/rejected": -636.0,
+      "loss": 0.6548,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -2.625,
+      "rewards/margins": 0.765625,
+      "rewards/rejected": -3.390625,
+      "step": 840
+    },
+    {
+      "epoch": 0.8895866038723181,
+      "grad_norm": 3.094401058074156,
+      "learning_rate": 4.43433678268096e-05,
+      "logits/chosen": -11.5,
+      "logits/rejected": -10.375,
+      "logps/chosen": -596.0,
+      "logps/rejected": -596.0,
+      "loss": 0.6278,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": -2.6875,
+      "rewards/margins": 0.69140625,
+      "rewards/rejected": -3.390625,
+      "step": 850
+    },
+    {
+      "epoch": 0.9000523286237572,
+      "grad_norm": 3.324021946705048,
+      "learning_rate": 4.414893503737968e-05,
+      "logits/chosen": -12.8125,
+      "logits/rejected": -12.125,
+      "logps/chosen": -636.0,
+      "logps/rejected": -640.0,
+      "loss": 0.6659,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -3.015625,
+      "rewards/margins": 0.5078125,
+      "rewards/rejected": -3.515625,
+      "step": 860
+    },
+    {
+      "epoch": 0.9105180533751962,
+      "grad_norm": 3.134320071444541,
+      "learning_rate": 4.395165861740414e-05,
+      "logits/chosen": -13.9375,
+      "logits/rejected": -13.4375,
+      "logps/chosen": -596.0,
+      "logps/rejected": -616.0,
+      "loss": 0.7148,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -2.765625,
+      "rewards/margins": 0.62109375,
+      "rewards/rejected": -3.390625,
+      "step": 870
+    },
+    {
+      "epoch": 0.9209837781266352,
+      "grad_norm": 2.954867612691979,
+      "learning_rate": 4.375156786257241e-05,
+      "logits/chosen": -13.0625,
+      "logits/rejected": -12.5625,
+      "logps/chosen": -636.0,
+      "logps/rejected": -684.0,
+      "loss": 0.651,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -2.875,
+      "rewards/margins": 0.59375,
+      "rewards/rejected": -3.453125,
+      "step": 880
+    },
+    {
+      "epoch": 0.9314495028780743,
+      "grad_norm": 3.340131928554516,
+      "learning_rate": 4.354869248650466e-05,
+      "logits/chosen": -13.9375,
+      "logits/rejected": -13.3125,
+      "logps/chosen": -600.0,
+      "logps/rejected": -628.0,
+      "loss": 0.7107,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -2.5625,
+      "rewards/margins": 0.53515625,
+      "rewards/rejected": -3.09375,
+      "step": 890
+    },
+    {
+      "epoch": 0.9419152276295133,
+      "grad_norm": 3.140671798997394,
+      "learning_rate": 4.334306261633931e-05,
+      "logits/chosen": -14.0,
+      "logits/rejected": -13.375,
+      "logps/chosen": -588.0,
+      "logps/rejected": -640.0,
+      "loss": 0.7325,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -2.59375,
+      "rewards/margins": 0.5703125,
+      "rewards/rejected": -3.15625,
+      "step": 900
+    },
+    {
+      "epoch": 0.9419152276295133,
+      "eval_logits/chosen": -14.0,
+      "eval_logits/rejected": -13.25,
+      "eval_logps/chosen": -588.0,
+      "eval_logps/rejected": -624.0,
+      "eval_loss": 0.683843731880188,
+      "eval_rewards/accuracies": 0.708984375,
+      "eval_rewards/chosen": -2.6875,
+      "eval_rewards/margins": 0.66015625,
+      "eval_rewards/rejected": -3.359375,
+      "eval_runtime": 46.6383,
+      "eval_samples_per_second": 42.883,
+      "eval_steps_per_second": 0.686,
+      "step": 900
+    },
+    {
+      "epoch": 0.9523809523809523,
+      "grad_norm": 3.4112808781248933,
+      "learning_rate": 4.313470878825911e-05,
+      "logits/chosen": -14.1875,
+      "logits/rejected": -13.375,
+      "logps/chosen": -508.0,
+      "logps/rejected": -544.0,
+      "loss": 0.7055,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -2.5,
+      "rewards/margins": 0.5390625,
+      "rewards/rejected": -3.046875,
+      "step": 910
+    },
+    {
+      "epoch": 0.9628466771323915,
+      "grad_norm": 3.2294008942871257,
+      "learning_rate": 4.292366194295652e-05,
+      "logits/chosen": -13.75,
+      "logits/rejected": -13.25,
+      "logps/chosen": -608.0,
+      "logps/rejected": -616.0,
+      "loss": 0.6836,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -2.8125,
+      "rewards/margins": 0.55078125,
+      "rewards/rejected": -3.359375,
+      "step": 920
+    },
+    {
+      "epoch": 0.9733124018838305,
+      "grad_norm": 2.929802333614972,
+      "learning_rate": 4.2709953421038964e-05,
+      "logits/chosen": -11.875,
+      "logits/rejected": -10.75,
+      "logps/chosen": -672.0,
+      "logps/rejected": -664.0,
+      "loss": 0.6836,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -3.109375,
+      "rewards/margins": 0.6015625,
+      "rewards/rejected": -3.703125,
+      "step": 930
+    },
+    {
+      "epoch": 0.9837781266352695,
+      "grad_norm": 2.854922011225652,
+      "learning_rate": 4.249361495837478e-05,
+      "logits/chosen": -12.5625,
+      "logits/rejected": -11.625,
+      "logps/chosen": -600.0,
+      "logps/rejected": -628.0,
+      "loss": 0.6493,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -2.90625,
+      "rewards/margins": 0.56640625,
+      "rewards/rejected": -3.46875,
+      "step": 940
+    },
+    {
+      "epoch": 0.9942438513867086,
+      "grad_norm": 3.339557200561039,
+      "learning_rate": 4.227467868138035e-05,
+      "logits/chosen": -12.4375,
+      "logits/rejected": -11.8125,
+      "logps/chosen": -652.0,
+      "logps/rejected": -680.0,
+      "loss": 0.6846,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -2.984375,
+      "rewards/margins": 0.5546875,
+      "rewards/rejected": -3.53125,
+      "step": 950
+    },
+    {
+      "epoch": 1.0047095761381475,
+      "grad_norm": 2.249444368233039,
+      "learning_rate": 4.205317710224938e-05,
+      "logits/chosen": -13.0,
+      "logits/rejected": -12.375,
+      "logps/chosen": -608.0,
+      "logps/rejected": -648.0,
+      "loss": 0.5015,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -2.828125,
+      "rewards/margins": 0.8359375,
+      "rewards/rejected": -3.671875,
+      "step": 960
+    },
+    {
+      "epoch": 1.0151753008895865,
+      "grad_norm": 3.338063906335034,
+      "learning_rate": 4.1829143114124733e-05,
+      "logits/chosen": -11.9375,
+      "logits/rejected": -10.6875,
+      "logps/chosen": -632.0,
+      "logps/rejected": -764.0,
+      "loss": 0.2712,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -2.71875,
+      "rewards/margins": 1.828125,
+      "rewards/rejected": -4.53125,
+      "step": 970
+    },
+    {
+      "epoch": 1.0256410256410255,
+      "grad_norm": 2.0935097134119265,
+      "learning_rate": 4.160260998621387e-05,
+      "logits/chosen": -11.125,
+      "logits/rejected": -9.3125,
+      "logps/chosen": -604.0,
+      "logps/rejected": -712.0,
+      "loss": 0.2276,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -3.0,
+      "rewards/margins": 1.5859375,
+      "rewards/rejected": -4.59375,
+      "step": 980
+    },
+    {
+      "epoch": 1.0361067503924646,
+      "grad_norm": 2.8551198573038503,
+      "learning_rate": 4.137361135884824e-05,
+      "logits/chosen": -12.0625,
+      "logits/rejected": -10.125,
+      "logps/chosen": -648.0,
+      "logps/rejected": -804.0,
+      "loss": 0.2201,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -3.0625,
+      "rewards/margins": 2.140625,
+      "rewards/rejected": -5.21875,
+      "step": 990
+    },
+    {
+      "epoch": 1.0465724751439036,
+      "grad_norm": 2.921929337257899,
+      "learning_rate": 4.114218123848777e-05,
+      "logits/chosen": -12.375,
+      "logits/rejected": -11.0,
+      "logps/chosen": -640.0,
+      "logps/rejected": -780.0,
+      "loss": 0.2677,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -2.859375,
+      "rewards/margins": 1.859375,
+      "rewards/rejected": -4.71875,
+      "step": 1000
+    },
+    {
+      "epoch": 1.0465724751439036,
+      "eval_logits/chosen": -12.1875,
+      "eval_logits/rejected": -11.0625,
+      "eval_logps/chosen": -640.0,
+      "eval_logps/rejected": -688.0,
+      "eval_loss": 0.6726171970367432,
+      "eval_rewards/accuracies": 0.70703125,
+      "eval_rewards/chosen": -3.234375,
+      "eval_rewards/margins": 0.7734375,
+      "eval_rewards/rejected": -4.0,
+      "eval_runtime": 46.6324,
+      "eval_samples_per_second": 42.889,
+      "eval_steps_per_second": 0.686,
+      "step": 1000
+    },
+    {
+      "epoch": 1.0570381998953426,
+      "grad_norm": 2.3721806180289886,
+      "learning_rate": 4.0908353992670824e-05,
+      "logits/chosen": -11.625,
+      "logits/rejected": -10.125,
+      "logps/chosen": -588.0,
+      "logps/rejected": -720.0,
+      "loss": 0.1972,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -2.703125,
+      "rewards/margins": 1.9296875,
+      "rewards/rejected": -4.625,
+      "step": 1010
+    },
+    {
+      "epoch": 1.0675039246467817,
+      "grad_norm": 3.0526619456616735,
+      "learning_rate": 4.067216434491059e-05,
+      "logits/chosen": -10.9375,
+      "logits/rejected": -9.625,
+      "logps/chosen": -628.0,
+      "logps/rejected": -804.0,
+      "loss": 0.2113,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -2.78125,
+      "rewards/margins": 2.09375,
+      "rewards/rejected": -4.875,
+      "step": 1020
+    },
+    {
+      "epoch": 1.077969649398221,
+      "grad_norm": 2.4202319591961508,
+      "learning_rate": 4.043364736953866e-05,
+      "logits/chosen": -11.5,
+      "logits/rejected": -9.9375,
+      "logps/chosen": -628.0,
+      "logps/rejected": -784.0,
+      "loss": 0.2354,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -3.03125,
+      "rewards/margins": 1.9765625,
+      "rewards/rejected": -5.0,
+      "step": 1030
+    },
+    {
+      "epoch": 1.08843537414966,
+      "grad_norm": 3.354020735754128,
+      "learning_rate": 4.019283848649643e-05,
+      "logits/chosen": -10.1875,
+      "logits/rejected": -8.25,
+      "logps/chosen": -584.0,
+      "logps/rejected": -728.0,
+      "loss": 0.2573,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -2.8125,
+      "rewards/margins": 1.8125,
+      "rewards/rejected": -4.625,
+      "step": 1040
+    },
+    {
+      "epoch": 1.098901098901099,
+      "grad_norm": 3.494681699406123,
+      "learning_rate": 3.9949773456075215e-05,
+      "logits/chosen": -12.125,
+      "logits/rejected": -11.0,
+      "logps/chosen": -544.0,
+      "logps/rejected": -700.0,
+      "loss": 0.1675,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -2.390625,
+      "rewards/margins": 1.921875,
+      "rewards/rejected": -4.3125,
+      "step": 1050
+    },
+    {
+      "epoch": 1.109366823652538,
+      "grad_norm": 2.697501360257953,
+      "learning_rate": 3.970448837360585e-05,
+      "logits/chosen": -10.875,
+      "logits/rejected": -9.125,
+      "logps/chosen": -604.0,
+      "logps/rejected": -756.0,
+      "loss": 0.2123,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -2.8125,
+      "rewards/margins": 1.9921875,
+      "rewards/rejected": -4.8125,
+      "step": 1060
+    },
+    {
+      "epoch": 1.119832548403977,
+      "grad_norm": 3.1170992535274937,
+      "learning_rate": 3.945701966409846e-05,
+      "logits/chosen": -12.0,
+      "logits/rejected": -10.5,
+      "logps/chosen": -588.0,
+      "logps/rejected": -740.0,
+      "loss": 0.2244,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -2.640625,
+      "rewards/margins": 2.03125,
+      "rewards/rejected": -4.65625,
+      "step": 1070
+    },
+    {
+      "epoch": 1.130298273155416,
+      "grad_norm": 2.454284375017908,
+      "learning_rate": 3.9207404076833374e-05,
+      "logits/chosen": -12.8125,
+      "logits/rejected": -11.875,
+      "logps/chosen": -560.0,
+      "logps/rejected": -760.0,
+      "loss": 0.2263,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -2.578125,
+      "rewards/margins": 2.15625,
+      "rewards/rejected": -4.75,
+      "step": 1080
+    },
+    {
+      "epoch": 1.1407639979068551,
+      "grad_norm": 2.750660528766078,
+      "learning_rate": 3.895567867990379e-05,
+      "logits/chosen": -11.9375,
+      "logits/rejected": -10.0625,
+      "logps/chosen": -596.0,
+      "logps/rejected": -784.0,
+      "loss": 0.1965,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -2.765625,
+      "rewards/margins": 2.15625,
+      "rewards/rejected": -4.90625,
+      "step": 1090
+    },
+    {
+      "epoch": 1.1512297226582942,
+      "grad_norm": 3.298570424007951,
+      "learning_rate": 3.870188085471114e-05,
+      "logits/chosen": -10.75,
+      "logits/rejected": -9.4375,
+      "logps/chosen": -624.0,
+      "logps/rejected": -784.0,
+      "loss": 0.2256,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -2.890625,
+      "rewards/margins": 2.03125,
+      "rewards/rejected": -4.90625,
+      "step": 1100
+    },
+    {
+      "epoch": 1.1512297226582942,
+      "eval_logits/chosen": -11.125,
+      "eval_logits/rejected": -10.0625,
+      "eval_logps/chosen": -676.0,
+      "eval_logps/rejected": -728.0,
+      "eval_loss": 0.6991875171661377,
+      "eval_rewards/accuracies": 0.708984375,
+      "eval_rewards/chosen": -3.59375,
+      "eval_rewards/margins": 0.796875,
+      "eval_rewards/rejected": -4.375,
+      "eval_runtime": 46.6642,
+      "eval_samples_per_second": 42.859,
+      "eval_steps_per_second": 0.686,
+      "step": 1100
+    },
+    {
+      "epoch": 1.1616954474097332,
+      "grad_norm": 2.5865179410938275,
+      "learning_rate": 3.844604829041395e-05,
+      "logits/chosen": -11.5625,
+      "logits/rejected": -9.875,
+      "logps/chosen": -624.0,
+      "logps/rejected": -796.0,
+      "loss": 0.1714,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -3.0625,
+      "rewards/margins": 2.15625,
+      "rewards/rejected": -5.21875,
+      "step": 1110
+    },
+    {
+      "epoch": 1.1721611721611722,
+      "grad_norm": 4.0135826845287035,
+      "learning_rate": 3.818821897833095e-05,
+      "logits/chosen": -11.625,
+      "logits/rejected": -10.375,
+      "logps/chosen": -664.0,
+      "logps/rejected": -828.0,
+      "loss": 0.2544,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -3.34375,
+      "rewards/margins": 1.9921875,
+      "rewards/rejected": -5.34375,
+      "step": 1120
+    },
+    {
+      "epoch": 1.1826268969126112,
+      "grad_norm": 2.8680374991054305,
+      "learning_rate": 3.792843120629935e-05,
+      "logits/chosen": -11.6875,
+      "logits/rejected": -10.1875,
+      "logps/chosen": -668.0,
+      "logps/rejected": -856.0,
+      "loss": 0.1783,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -3.328125,
+      "rewards/margins": 2.15625,
+      "rewards/rejected": -5.5,
+      "step": 1130
+    },
+    {
+      "epoch": 1.1930926216640503,
+      "grad_norm": 3.3269356160633055,
+      "learning_rate": 3.766672355298907e-05,
+      "logits/chosen": -11.25,
+      "logits/rejected": -10.125,
+      "logps/chosen": -676.0,
+      "logps/rejected": -864.0,
+      "loss": 0.1935,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -3.1875,
+      "rewards/margins": 2.328125,
+      "rewards/rejected": -5.53125,
+      "step": 1140
+    },
+    {
+      "epoch": 1.2035583464154893,
+      "grad_norm": 3.3229360304743754,
+      "learning_rate": 3.740313488217373e-05,
+      "logits/chosen": -12.3125,
+      "logits/rejected": -11.375,
+      "logps/chosen": -608.0,
+      "logps/rejected": -804.0,
+      "loss": 0.2069,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -2.90625,
+      "rewards/margins": 2.09375,
+      "rewards/rejected": -5.0,
+      "step": 1150
+    },
+    {
+      "epoch": 1.2140240711669283,
+      "grad_norm": 3.234166836617256,
+      "learning_rate": 3.7137704336959464e-05,
+      "logits/chosen": -13.0,
+      "logits/rejected": -11.75,
+      "logps/chosen": -620.0,
+      "logps/rejected": -780.0,
+      "loss": 0.2251,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -3.015625,
+      "rewards/margins": 2.09375,
+      "rewards/rejected": -5.09375,
+      "step": 1160
+    },
+    {
+      "epoch": 1.2244897959183674,
+      "grad_norm": 3.508664599839402,
+      "learning_rate": 3.687047133397201e-05,
+      "logits/chosen": -12.125,
+      "logits/rejected": -10.9375,
+      "logps/chosen": -620.0,
+      "logps/rejected": -788.0,
+      "loss": 0.2163,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -3.078125,
+      "rewards/margins": 2.109375,
+      "rewards/rejected": -5.1875,
+      "step": 1170
+    },
+    {
+      "epoch": 1.2349555206698064,
+      "grad_norm": 2.6354994625505963,
+      "learning_rate": 3.6601475557503415e-05,
+      "logits/chosen": -12.8125,
+      "logits/rejected": -11.75,
+      "logps/chosen": -628.0,
+      "logps/rejected": -808.0,
+      "loss": 0.2301,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -3.078125,
+      "rewards/margins": 2.21875,
+      "rewards/rejected": -5.28125,
+      "step": 1180
+    },
+    {
+      "epoch": 1.2454212454212454,
+      "grad_norm": 3.055983812762542,
+      "learning_rate": 3.633075695361882e-05,
+      "logits/chosen": -12.875,
+      "logits/rejected": -11.5,
+      "logps/chosen": -668.0,
+      "logps/rejected": -808.0,
+      "loss": 0.1963,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -3.3125,
+      "rewards/margins": 1.9375,
+      "rewards/rejected": -5.25,
+      "step": 1190
+    },
+    {
+      "epoch": 1.2558869701726845,
+      "grad_norm": 3.7958952476907233,
+      "learning_rate": 3.6058355724224477e-05,
+      "logits/chosen": -11.6875,
+      "logits/rejected": -9.75,
+      "logps/chosen": -692.0,
+      "logps/rejected": -864.0,
+      "loss": 0.1954,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -3.171875,
+      "rewards/margins": 2.203125,
+      "rewards/rejected": -5.375,
+      "step": 1200
+    },
+    {
+      "epoch": 1.2558869701726845,
+      "eval_logits/chosen": -11.3125,
+      "eval_logits/rejected": -10.125,
+      "eval_logps/chosen": -664.0,
+      "eval_logps/rejected": -720.0,
+      "eval_loss": 0.7033437490463257,
+      "eval_rewards/accuracies": 0.705078125,
+      "eval_rewards/chosen": -3.46875,
+      "eval_rewards/margins": 0.84765625,
+      "eval_rewards/rejected": -4.3125,
+      "eval_runtime": 46.6406,
+      "eval_samples_per_second": 42.881,
+      "eval_steps_per_second": 0.686,
+      "step": 1200
+    },
+    {
+      "epoch": 1.2663526949241235,
+      "grad_norm": 3.362197698375765,
+      "learning_rate": 3.5784312321097744e-05,
+      "logits/chosen": -11.75,
+      "logits/rejected": -10.0,
+      "logps/chosen": -688.0,
+      "logps/rejected": -808.0,
+      "loss": 0.2394,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -2.96875,
+      "rewards/margins": 2.109375,
+      "rewards/rejected": -5.0625,
+      "step": 1210
+    },
+    {
+      "epoch": 1.2768184196755625,
+      "grad_norm": 3.152195421494335,
+      "learning_rate": 3.550866743987994e-05,
+      "logits/chosen": -11.75,
+      "logits/rejected": -10.625,
+      "logps/chosen": -640.0,
+      "logps/rejected": -796.0,
+      "loss": 0.2341,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -3.140625,
+      "rewards/margins": 2.0625,
+      "rewards/rejected": -5.1875,
+      "step": 1220
+    },
+    {
+      "epoch": 1.2872841444270016,
+      "grad_norm": 2.8259918243226676,
+      "learning_rate": 3.5231462014033024e-05,
+      "logits/chosen": -12.125,
+      "logits/rejected": -11.0,
+      "logps/chosen": -576.0,
+      "logps/rejected": -724.0,
+      "loss": 0.1975,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -2.71875,
+      "rewards/margins": 1.921875,
+      "rewards/rejected": -4.65625,
+      "step": 1230
+    },
+    {
+      "epoch": 1.2977498691784406,
+      "grad_norm": 3.3750890607545045,
+      "learning_rate": 3.495273720876095e-05,
+      "logits/chosen": -12.875,
+      "logits/rejected": -11.6875,
+      "logps/chosen": -576.0,
+      "logps/rejected": -736.0,
+      "loss": 0.201,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -2.71875,
+      "rewards/margins": 1.8671875,
+      "rewards/rejected": -4.59375,
+      "step": 1240
+    },
+    {
+      "epoch": 1.3082155939298796,
+      "grad_norm": 2.7468180302780283,
+      "learning_rate": 3.4672534414896616e-05,
+      "logits/chosen": -13.125,
+      "logits/rejected": -11.75,
+      "logps/chosen": -616.0,
+      "logps/rejected": -764.0,
+      "loss": 0.1897,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -2.75,
+      "rewards/margins": 2.09375,
+      "rewards/rejected": -4.84375,
+      "step": 1250
+    },
+    {
+      "epoch": 1.3186813186813187,
+      "grad_norm": 3.8087989748230924,
+      "learning_rate": 3.439089524275526e-05,
+      "logits/chosen": -12.5,
+      "logits/rejected": -11.625,
+      "logps/chosen": -592.0,
+      "logps/rejected": -756.0,
+      "loss": 0.2022,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -2.890625,
+      "rewards/margins": 1.8359375,
+      "rewards/rejected": -4.71875,
+      "step": 1260
+    },
+    {
+      "epoch": 1.3291470434327577,
+      "grad_norm": 3.354938427295044,
+      "learning_rate": 3.410786151595533e-05,
+      "logits/chosen": -12.1875,
+      "logits/rejected": -10.875,
+      "logps/chosen": -624.0,
+      "logps/rejected": -784.0,
+      "loss": 0.2084,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -3.09375,
+      "rewards/margins": 1.90625,
+      "rewards/rejected": -5.0,
+      "step": 1270
+    },
+    {
+      "epoch": 1.3396127681841967,
+      "grad_norm": 3.482933300726635,
+      "learning_rate": 3.382347526520765e-05,
+      "logits/chosen": -10.75,
+      "logits/rejected": -8.375,
+      "logps/chosen": -668.0,
+      "logps/rejected": -844.0,
+      "loss": 0.1901,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -3.421875,
+      "rewards/margins": 2.34375,
+      "rewards/rejected": -5.75,
+      "step": 1280
+    },
+    {
+      "epoch": 1.3500784929356358,
+      "grad_norm": 4.0909056283946,
+      "learning_rate": 3.353777872207381e-05,
+      "logits/chosen": -9.625,
+      "logits/rejected": -8.0,
+      "logps/chosen": -688.0,
+      "logps/rejected": -872.0,
+      "loss": 0.2414,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -3.5625,
+      "rewards/margins": 2.28125,
+      "rewards/rejected": -5.84375,
+      "step": 1290
+    },
+    {
+      "epoch": 1.3605442176870748,
+      "grad_norm": 3.4644847678898896,
+      "learning_rate": 3.325081431269475e-05,
+      "logits/chosen": -11.0,
+      "logits/rejected": -9.75,
+      "logps/chosen": -656.0,
+      "logps/rejected": -836.0,
+      "loss": 0.2289,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -3.359375,
+      "rewards/margins": 1.953125,
+      "rewards/rejected": -5.3125,
+      "step": 1300
+    },
+    {
+      "epoch": 1.3605442176870748,
+      "eval_logits/chosen": -11.0,
+      "eval_logits/rejected": -9.9375,
+      "eval_logps/chosen": -692.0,
+      "eval_logps/rejected": -740.0,
+      "eval_loss": 0.672249972820282,
+      "eval_rewards/accuracies": 0.734375,
+      "eval_rewards/chosen": -3.734375,
+      "eval_rewards/margins": 0.78515625,
+      "eval_rewards/rejected": -4.5,
+      "eval_runtime": 46.6391,
+      "eval_samples_per_second": 42.882,
+      "eval_steps_per_second": 0.686,
+      "step": 1300
+    },
+    {
+      "epoch": 1.3710099424385138,
+      "grad_norm": 2.8739000782856077,
+      "learning_rate": 3.296262465149046e-05,
+      "logits/chosen": -11.125,
+      "logits/rejected": -9.5,
+      "logps/chosen": -680.0,
+      "logps/rejected": -808.0,
+      "loss": 0.1877,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -3.453125,
+      "rewards/margins": 1.84375,
+      "rewards/rejected": -5.28125,
+      "step": 1310
+    },
+    {
+      "epoch": 1.3814756671899528,
+      "grad_norm": 5.375320762757781,
+      "learning_rate": 3.2673252534831687e-05,
+      "logits/chosen": -9.5625,
+      "logits/rejected": -8.125,
+      "logps/chosen": -652.0,
+      "logps/rejected": -824.0,
+      "loss": 0.2269,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -3.625,
+      "rewards/margins": 1.921875,
+      "rewards/rejected": -5.53125,
+      "step": 1320
+    },
+    {
+      "epoch": 1.3919413919413919,
+      "grad_norm": 3.8554227343185272,
+      "learning_rate": 3.2382740934684696e-05,
+      "logits/chosen": -10.125,
+      "logits/rejected": -9.25,
+      "logps/chosen": -676.0,
+      "logps/rejected": -896.0,
+      "loss": 0.2208,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -3.578125,
+      "rewards/margins": 2.09375,
+      "rewards/rejected": -5.6875,
+      "step": 1330
+    },
+    {
+      "epoch": 1.402407116692831,
+      "grad_norm": 3.973088551973234,
+      "learning_rate": 3.209113299222982e-05,
+      "logits/chosen": -10.6875,
+      "logits/rejected": -8.9375,
+      "logps/chosen": -696.0,
+      "logps/rejected": -848.0,
+      "loss": 0.1851,
+      "rewards/accuracies": 0.8812500238418579,
+      "rewards/chosen": -3.5625,
+      "rewards/margins": 2.09375,
+      "rewards/rejected": -5.65625,
+      "step": 1340
+    },
+    {
+      "epoch": 1.41287284144427,
+      "grad_norm": 3.502083690560275,
+      "learning_rate": 3.179847201145505e-05,
+      "logits/chosen": -10.0625,
+      "logits/rejected": -8.25,
+      "logps/chosen": -672.0,
+      "logps/rejected": -884.0,
+      "loss": 0.1914,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -3.546875,
+      "rewards/margins": 2.40625,
+      "rewards/rejected": -5.96875,
+      "step": 1350
+    },
+    {
+      "epoch": 1.423338566195709,
+      "grad_norm": 3.0146439389957855,
+      "learning_rate": 3.150480145272528e-05,
+      "logits/chosen": -10.75,
+      "logits/rejected": -9.25,
+      "logps/chosen": -660.0,
+      "logps/rejected": -816.0,
+      "loss": 0.2449,
+      "rewards/accuracies": 0.893750011920929,
+      "rewards/chosen": -3.359375,
+      "rewards/margins": 2.046875,
+      "rewards/rejected": -5.40625,
+      "step": 1360
+    },
+    {
+      "epoch": 1.433804290947148,
+      "grad_norm": 3.503945733101692,
+      "learning_rate": 3.121016492632848e-05,
+      "logits/chosen": -10.9375,
+      "logits/rejected": -9.625,
+      "logps/chosen": -664.0,
+      "logps/rejected": -824.0,
+      "loss": 0.2357,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": -3.5625,
+      "rewards/margins": 1.953125,
+      "rewards/rejected": -5.5,
+      "step": 1370
+    },
+    {
+      "epoch": 1.4442700156985873,
+      "grad_norm": 3.050347580271843,
+      "learning_rate": 3.091460618599951e-05,
+      "logits/chosen": -11.625,
+      "logits/rejected": -10.125,
+      "logps/chosen": -656.0,
+      "logps/rejected": -824.0,
+      "loss": 0.2,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -3.359375,
+      "rewards/margins": 2.125,
+      "rewards/rejected": -5.46875,
+      "step": 1380
+    },
+    {
+      "epoch": 1.454735740450026,
+      "grad_norm": 3.401592595581994,
+      "learning_rate": 3.061816912242265e-05,
+      "logits/chosen": -12.4375,
+      "logits/rejected": -11.125,
+      "logps/chosen": -624.0,
+      "logps/rejected": -784.0,
+      "loss": 0.2334,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -2.953125,
+      "rewards/margins": 2.203125,
+      "rewards/rejected": -5.15625,
+      "step": 1390
+    },
+    {
+      "epoch": 1.4652014652014653,
+      "grad_norm": 3.3469598033021315,
+      "learning_rate": 3.0320897756713785e-05,
+      "logits/chosen": -12.6875,
+      "logits/rejected": -12.0625,
+      "logps/chosen": -632.0,
+      "logps/rejected": -832.0,
+      "loss": 0.2227,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -2.890625,
+      "rewards/margins": 2.046875,
+      "rewards/rejected": -4.9375,
+      "step": 1400
+    },
+    {
+      "epoch": 1.4652014652014653,
+      "eval_logits/chosen": -12.5,
+      "eval_logits/rejected": -11.8125,
+      "eval_logps/chosen": -676.0,
+      "eval_logps/rejected": -720.0,
+      "eval_loss": 0.6924843788146973,
+      "eval_rewards/accuracies": 0.6953125,
+      "eval_rewards/chosen": -3.578125,
+      "eval_rewards/margins": 0.73828125,
+      "eval_rewards/rejected": -4.3125,
+      "eval_runtime": 46.6354,
+      "eval_samples_per_second": 42.886,
+      "eval_steps_per_second": 0.686,
+      "step": 1400
+    },
+    {
+      "epoch": 1.4756671899529041,
+      "grad_norm": 3.1499922846778667,
+      "learning_rate": 3.002283623388332e-05,
+      "logits/chosen": -12.25,
+      "logits/rejected": -10.8125,
+      "logps/chosen": -636.0,
+      "logps/rejected": -824.0,
+      "loss": 0.2007,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -3.265625,
+      "rewards/margins": 2.0625,
+      "rewards/rejected": -5.3125,
+      "step": 1410
+    },
+    {
+      "epoch": 1.4861329147043434,
+      "grad_norm": 2.657139241228349,
+      "learning_rate": 2.9724028816280507e-05,
+      "logits/chosen": -11.375,
+      "logits/rejected": -10.4375,
+      "logps/chosen": -660.0,
+      "logps/rejected": -812.0,
+      "loss": 0.2016,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -3.171875,
+      "rewards/margins": 1.9765625,
+      "rewards/rejected": -5.15625,
+      "step": 1420
+    },
+    {
+      "epoch": 1.4965986394557822,
+      "grad_norm": 3.0508473895703165,
+      "learning_rate": 2.9424519877020524e-05,
+      "logits/chosen": -11.8125,
+      "logits/rejected": -10.1875,
+      "logps/chosen": -644.0,
+      "logps/rejected": -832.0,
+      "loss": 0.1828,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -2.984375,
+      "rewards/margins": 2.28125,
+      "rewards/rejected": -5.28125,
+      "step": 1430
+    },
+    {
+      "epoch": 1.5070643642072215,
+      "grad_norm": 2.9224211944327614,
+      "learning_rate": 2.912435389339504e-05,
+      "logits/chosen": -12.25,
+      "logits/rejected": -10.875,
+      "logps/chosen": -664.0,
+      "logps/rejected": -836.0,
+      "loss": 0.2025,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -3.1875,
+      "rewards/margins": 2.125,
+      "rewards/rejected": -5.3125,
+      "step": 1440
+    },
+    {
+      "epoch": 1.5175300889586603,
+      "grad_norm": 2.6455949582564444,
+      "learning_rate": 2.8823575440267226e-05,
+      "logits/chosen": -12.6875,
+      "logits/rejected": -11.875,
+      "logps/chosen": -684.0,
+      "logps/rejected": -856.0,
+      "loss": 0.1545,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -3.09375,
+      "rewards/margins": 2.28125,
+      "rewards/rejected": -5.375,
+      "step": 1450
+    },
+    {
+      "epoch": 1.5279958137100995,
+      "grad_norm": 4.05057601686232,
+      "learning_rate": 2.8522229183452465e-05,
+      "logits/chosen": -12.8125,
+      "logits/rejected": -11.5,
+      "logps/chosen": -652.0,
+      "logps/rejected": -876.0,
+      "loss": 0.1635,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -3.25,
+      "rewards/margins": 2.34375,
+      "rewards/rejected": -5.59375,
+      "step": 1460
+    },
+    {
+      "epoch": 1.5384615384615383,
+      "grad_norm": 4.300964610499669,
+      "learning_rate": 2.8220359873085373e-05,
+      "logits/chosen": -12.9375,
+      "logits/rejected": -11.9375,
+      "logps/chosen": -692.0,
+      "logps/rejected": -904.0,
+      "loss": 0.2211,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -3.296875,
+      "rewards/margins": 2.3125,
+      "rewards/rejected": -5.625,
+      "step": 1470
+    },
+    {
+      "epoch": 1.5489272632129776,
+      "grad_norm": 3.0144007515834237,
+      "learning_rate": 2.7918012336974388e-05,
+      "logits/chosen": -12.75,
+      "logits/rejected": -11.3125,
+      "logps/chosen": -652.0,
+      "logps/rejected": -792.0,
+      "loss": 0.2118,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": -3.515625,
+      "rewards/margins": 2.0,
+      "rewards/rejected": -5.53125,
+      "step": 1480
+    },
+    {
+      "epoch": 1.5593929879644164,
+      "grad_norm": 2.902271334761215,
+      "learning_rate": 2.7615231473944815e-05,
+      "logits/chosen": -12.5,
+      "logits/rejected": -11.125,
+      "logps/chosen": -688.0,
+      "logps/rejected": -844.0,
+      "loss": 0.1755,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -3.46875,
+      "rewards/margins": 2.09375,
+      "rewards/rejected": -5.5625,
+      "step": 1490
+    },
+    {
+      "epoch": 1.5698587127158556,
+      "grad_norm": 2.712443447719808,
+      "learning_rate": 2.7312062247171327e-05,
+      "logits/chosen": -12.0,
+      "logits/rejected": -10.75,
+      "logps/chosen": -696.0,
+      "logps/rejected": -900.0,
+      "loss": 0.1902,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -3.578125,
+      "rewards/margins": 2.234375,
+      "rewards/rejected": -5.8125,
+      "step": 1500
+    },
+    {
+      "epoch": 1.5698587127158556,
+      "eval_logits/chosen": -12.0625,
+      "eval_logits/rejected": -11.125,
+      "eval_logps/chosen": -736.0,
+      "eval_logps/rejected": -792.0,
+      "eval_loss": 0.6757500171661377,
+      "eval_rewards/accuracies": 0.71484375,
+      "eval_rewards/chosen": -4.1875,
+      "eval_rewards/margins": 0.83203125,
+      "eval_rewards/rejected": -5.03125,
+      "eval_runtime": 46.6327,
+      "eval_samples_per_second": 42.888,
+      "eval_steps_per_second": 0.686,
+      "step": 1500
+    },
+    {
+      "epoch": 1.5803244374672945,
+      "grad_norm": 3.0252423932724084,
+      "learning_rate": 2.7008549677500876e-05,
+      "logits/chosen": -11.75,
+      "logits/rejected": -10.4375,
+      "logps/chosen": -716.0,
+      "logps/rejected": -912.0,
+      "loss": 0.1932,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -3.65625,
+      "rewards/margins": 2.296875,
+      "rewards/rejected": -5.96875,
+      "step": 1510
+    },
+    {
+      "epoch": 1.5907901622187337,
+      "grad_norm": 3.6996527440433926,
+      "learning_rate": 2.670473883676709e-05,
+      "logits/chosen": -12.6875,
+      "logits/rejected": -11.25,
+      "logps/chosen": -708.0,
+      "logps/rejected": -888.0,
+      "loss": 0.2112,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -3.6875,
+      "rewards/margins": 2.140625,
+      "rewards/rejected": -5.84375,
+      "step": 1520
+    },
+    {
+      "epoch": 1.6012558869701727,
+      "grad_norm": 2.8039305555115566,
+      "learning_rate": 2.6400674841097074e-05,
+      "logits/chosen": -13.0625,
+      "logits/rejected": -11.75,
+      "logps/chosen": -660.0,
+      "logps/rejected": -792.0,
+      "loss": 0.1914,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -3.265625,
+      "rewards/margins": 2.09375,
+      "rewards/rejected": -5.375,
+      "step": 1530
+    },
+    {
+      "epoch": 1.6117216117216118,
+      "grad_norm": 2.326429350735345,
+      "learning_rate": 2.609640284421158e-05,
+      "logits/chosen": -13.25,
+      "logits/rejected": -11.8125,
+      "logps/chosen": -652.0,
+      "logps/rejected": -804.0,
+      "loss": 0.2006,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -3.015625,
+      "rewards/margins": 2.078125,
+      "rewards/rejected": -5.09375,
+      "step": 1540
+    },
+    {
+      "epoch": 1.6221873364730508,
+      "grad_norm": 3.280298090047358,
+      "learning_rate": 2.5791968030719726e-05,
+      "logits/chosen": -13.1875,
+      "logits/rejected": -11.875,
+      "logps/chosen": -660.0,
+      "logps/rejected": -844.0,
+      "loss": 0.2332,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -3.078125,
+      "rewards/margins": 2.09375,
+      "rewards/rejected": -5.15625,
+      "step": 1550
+    },
+    {
+      "epoch": 1.6326530612244898,
+      "grad_norm": 3.5431341275385453,
+      "learning_rate": 2.548741560940902e-05,
+      "logits/chosen": -13.625,
+      "logits/rejected": -12.0625,
+      "logps/chosen": -652.0,
+      "logps/rejected": -844.0,
+      "loss": 0.2169,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -3.109375,
+      "rewards/margins": 2.390625,
+      "rewards/rejected": -5.5,
+      "step": 1560
+    },
+    {
+      "epoch": 1.6431187859759289,
+      "grad_norm": 2.5493080361287066,
+      "learning_rate": 2.5182790806531787e-05,
+      "logits/chosen": -13.375,
+      "logits/rejected": -12.0,
+      "logps/chosen": -648.0,
+      "logps/rejected": -804.0,
+      "loss": 0.1766,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -3.390625,
+      "rewards/margins": 2.015625,
+      "rewards/rejected": -5.40625,
+      "step": 1570
+    },
+    {
+      "epoch": 1.653584510727368,
+      "grad_norm": 2.5746422473834305,
+      "learning_rate": 2.4878138859089074e-05,
+      "logits/chosen": -13.375,
+      "logits/rejected": -11.875,
+      "logps/chosen": -672.0,
+      "logps/rejected": -828.0,
+      "loss": 0.1892,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -3.421875,
+      "rewards/margins": 2.078125,
+      "rewards/rejected": -5.5,
+      "step": 1580
+    },
+    {
+      "epoch": 1.664050235478807,
+      "grad_norm": 3.3934858804686057,
+      "learning_rate": 2.4573505008112924e-05,
+      "logits/chosen": -13.4375,
+      "logits/rejected": -11.75,
+      "logps/chosen": -680.0,
+      "logps/rejected": -844.0,
+      "loss": 0.1827,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -3.5625,
+      "rewards/margins": 2.03125,
+      "rewards/rejected": -5.59375,
+      "step": 1590
+    },
+    {
+      "epoch": 1.674515960230246,
+      "grad_norm": 3.522556209898987,
+      "learning_rate": 2.4268934491948032e-05,
+      "logits/chosen": -13.625,
+      "logits/rejected": -12.125,
+      "logps/chosen": -684.0,
+      "logps/rejected": -844.0,
+      "loss": 0.2192,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -3.40625,
+      "rewards/margins": 2.171875,
+      "rewards/rejected": -5.59375,
+      "step": 1600
+    },
+    {
+      "epoch": 1.674515960230246,
+      "eval_logits/chosen": -13.625,
+      "eval_logits/rejected": -12.875,
+      "eval_logps/chosen": -704.0,
+      "eval_logps/rejected": -748.0,
+      "eval_loss": 0.6833124756813049,
+      "eval_rewards/accuracies": 0.71484375,
+      "eval_rewards/chosen": -3.84375,
+      "eval_rewards/margins": 0.76953125,
+      "eval_rewards/rejected": -4.625,
+      "eval_runtime": 46.6402,
+      "eval_samples_per_second": 42.881,
+      "eval_steps_per_second": 0.686,
+      "step": 1600
+    },
+    {
+      "epoch": 1.684981684981685,
+      "grad_norm": 2.518405595228139,
+      "learning_rate": 2.396447253953385e-05,
+      "logits/chosen": -13.5,
+      "logits/rejected": -12.25,
+      "logps/chosen": -696.0,
+      "logps/rejected": -880.0,
+      "loss": 0.2135,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -3.3125,
+      "rewards/margins": 2.21875,
+      "rewards/rejected": -5.53125,
+      "step": 1610
+    },
+    {
+      "epoch": 1.695447409733124,
+      "grad_norm": 3.1362711538859203,
+      "learning_rate": 2.3660164363687998e-05,
+      "logits/chosen": -13.125,
+      "logits/rejected": -11.375,
+      "logps/chosen": -632.0,
+      "logps/rejected": -816.0,
+      "loss": 0.2232,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -3.109375,
+      "rewards/margins": 2.3125,
+      "rewards/rejected": -5.4375,
+      "step": 1620
+    },
+    {
+      "epoch": 1.705913134484563,
+      "grad_norm": 2.8472199288678244,
+      "learning_rate": 2.33560551543922e-05,
+      "logits/chosen": -12.3125,
+      "logits/rejected": -10.875,
+      "logps/chosen": -656.0,
+      "logps/rejected": -840.0,
+      "loss": 0.1988,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -3.5625,
+      "rewards/margins": 2.03125,
+      "rewards/rejected": -5.59375,
+      "step": 1630
+    },
+    {
+      "epoch": 1.716378859236002,
+      "grad_norm": 3.4106187681759694,
+      "learning_rate": 2.3052190072081493e-05,
+      "logits/chosen": -12.0625,
+      "logits/rejected": -10.875,
+      "logps/chosen": -720.0,
+      "logps/rejected": -948.0,
+      "loss": 0.2107,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -3.578125,
+      "rewards/margins": 2.46875,
+      "rewards/rejected": -6.0625,
+      "step": 1640
+    },
+    {
+      "epoch": 1.7268445839874411,
+      "grad_norm": 3.09881345389625,
+      "learning_rate": 2.2748614240937864e-05,
+      "logits/chosen": -12.4375,
+      "logits/rejected": -11.3125,
+      "logps/chosen": -668.0,
+      "logps/rejected": -852.0,
+      "loss": 0.1806,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -3.296875,
+      "rewards/margins": 2.234375,
+      "rewards/rejected": -5.53125,
+      "step": 1650
+    },
+    {
+      "epoch": 1.7373103087388801,
+      "grad_norm": 3.3299664055771827,
+      "learning_rate": 2.2445372742189334e-05,
+      "logits/chosen": -12.25,
+      "logits/rejected": -11.375,
+      "logps/chosen": -660.0,
+      "logps/rejected": -848.0,
+      "loss": 0.2025,
+      "rewards/accuracies": 0.8812500238418579,
+      "rewards/chosen": -3.421875,
+      "rewards/margins": 2.171875,
+      "rewards/rejected": -5.59375,
+      "step": 1660
+    },
+    {
+      "epoch": 1.7477760334903192,
+      "grad_norm": 3.398438999013097,
+      "learning_rate": 2.214251060741528e-05,
+      "logits/chosen": -12.375,
+      "logits/rejected": -10.9375,
+      "logps/chosen": -688.0,
+      "logps/rejected": -840.0,
+      "loss": 0.1982,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -3.546875,
+      "rewards/margins": 1.9609375,
+      "rewards/rejected": -5.53125,
+      "step": 1670
+    },
+    {
+      "epoch": 1.7582417582417582,
+      "grad_norm": 3.3107133930484305,
+      "learning_rate": 2.1840072811859302e-05,
+      "logits/chosen": -11.875,
+      "logits/rejected": -10.5625,
+      "logps/chosen": -640.0,
+      "logps/rejected": -828.0,
+      "loss": 0.2076,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -3.5,
+      "rewards/margins": 2.171875,
+      "rewards/rejected": -5.6875,
+      "step": 1680
+    },
+    {
+      "epoch": 1.7687074829931972,
+      "grad_norm": 3.193494795620847,
+      "learning_rate": 2.1538104267750287e-05,
+      "logits/chosen": -12.3125,
+      "logits/rejected": -10.625,
+      "logps/chosen": -700.0,
+      "logps/rejected": -832.0,
+      "loss": 0.1857,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -3.46875,
+      "rewards/margins": 2.078125,
+      "rewards/rejected": -5.5625,
+      "step": 1690
+    },
+    {
+      "epoch": 1.7791732077446363,
+      "grad_norm": 3.7595229061921973,
+      "learning_rate": 2.1236649817632952e-05,
+      "logits/chosen": -11.8125,
+      "logits/rejected": -10.625,
+      "logps/chosen": -672.0,
+      "logps/rejected": -844.0,
+      "loss": 0.2137,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -3.5625,
+      "rewards/margins": 2.0,
+      "rewards/rejected": -5.5625,
+      "step": 1700
+    },
+    {
+      "epoch": 1.7791732077446363,
+      "eval_logits/chosen": -11.9375,
+      "eval_logits/rejected": -11.0,
+      "eval_logps/chosen": -716.0,
+      "eval_logps/rejected": -764.0,
+      "eval_loss": 0.6733828186988831,
+      "eval_rewards/accuracies": 0.720703125,
+      "eval_rewards/chosen": -3.96875,
+      "eval_rewards/margins": 0.80078125,
+      "eval_rewards/rejected": -4.78125,
+      "eval_runtime": 46.6309,
+      "eval_samples_per_second": 42.89,
+      "eval_steps_per_second": 0.686,
+      "step": 1700
+    },
+    {
+      "epoch": 1.7896389324960753,
+      "grad_norm": 2.7711190723633883,
+      "learning_rate": 2.093575422770872e-05,
+      "logits/chosen": -12.0625,
+      "logits/rejected": -10.75,
+      "logps/chosen": -704.0,
+      "logps/rejected": -828.0,
+      "loss": 0.2422,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -3.375,
+      "rewards/margins": 2.0625,
+      "rewards/rejected": -5.4375,
+      "step": 1710
+    },
+    {
+      "epoch": 1.8001046572475143,
+      "grad_norm": 2.6771039461510697,
+      "learning_rate": 2.0635462181187827e-05,
+      "logits/chosen": -12.4375,
+      "logits/rejected": -11.5625,
+      "logps/chosen": -668.0,
+      "logps/rejected": -804.0,
+      "loss": 0.1819,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -3.3125,
+      "rewards/margins": 1.8203125,
+      "rewards/rejected": -5.125,
+      "step": 1720
+    },
+    {
+      "epoch": 1.8105703819989536,
+      "grad_norm": 2.737338032426354,
+      "learning_rate": 2.0335818271653917e-05,
+      "logits/chosen": -12.375,
+      "logits/rejected": -10.9375,
+      "logps/chosen": -676.0,
+      "logps/rejected": -824.0,
+      "loss": 0.19,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -3.390625,
+      "rewards/margins": 1.984375,
+      "rewards/rejected": -5.375,
+      "step": 1730
+    },
+    {
+      "epoch": 1.8210361067503924,
+      "grad_norm": 3.213331777278463,
+      "learning_rate": 2.0036866996441815e-05,
+      "logits/chosen": -11.875,
+      "logits/rejected": -10.125,
+      "logps/chosen": -724.0,
+      "logps/rejected": -884.0,
+      "loss": 0.1965,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -3.71875,
+      "rewards/margins": 2.21875,
+      "rewards/rejected": -5.9375,
+      "step": 1740
+    },
+    {
+      "epoch": 1.8315018315018317,
+      "grad_norm": 3.3540961369157087,
+      "learning_rate": 1.9738652750029677e-05,
+      "logits/chosen": -12.5,
+      "logits/rejected": -10.5625,
+      "logps/chosen": -688.0,
+      "logps/rejected": -828.0,
+      "loss": 0.2267,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -3.4375,
+      "rewards/margins": 2.0625,
+      "rewards/rejected": -5.5,
+      "step": 1750
+    },
+    {
+      "epoch": 1.8419675562532705,
+      "grad_norm": 3.0233033456080287,
+      "learning_rate": 1.944121981744631e-05,
+      "logits/chosen": -12.5,
+      "logits/rejected": -11.375,
+      "logps/chosen": -640.0,
+      "logps/rejected": -820.0,
+      "loss": 0.2109,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -3.28125,
+      "rewards/margins": 2.046875,
+      "rewards/rejected": -5.3125,
+      "step": 1760
+    },
+    {
+      "epoch": 1.8524332810047097,
+      "grad_norm": 2.683715175109903,
+      "learning_rate": 1.9144612367694887e-05,
+      "logits/chosen": -12.25,
+      "logits/rejected": -11.125,
+      "logps/chosen": -664.0,
+      "logps/rejected": -828.0,
+      "loss": 0.1632,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -3.046875,
+      "rewards/margins": 2.25,
+      "rewards/rejected": -5.3125,
+      "step": 1770
+    },
+    {
+      "epoch": 1.8628990057561485,
+      "grad_norm": 2.7456532660541275,
+      "learning_rate": 1.8848874447193803e-05,
+      "logits/chosen": -12.125,
+      "logits/rejected": -10.625,
+      "logps/chosen": -696.0,
+      "logps/rejected": -876.0,
+      "loss": 0.2115,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -3.25,
+      "rewards/margins": 2.1875,
+      "rewards/rejected": -5.4375,
+      "step": 1780
+    },
+    {
+      "epoch": 1.8733647305075878,
+      "grad_norm": 3.6362673938722865,
+      "learning_rate": 1.855404997323571e-05,
+      "logits/chosen": -12.0,
+      "logits/rejected": -10.75,
+      "logps/chosen": -640.0,
+      "logps/rejected": -796.0,
+      "loss": 0.181,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -3.125,
+      "rewards/margins": 2.0,
+      "rewards/rejected": -5.125,
+      "step": 1790
+    },
+    {
+      "epoch": 1.8838304552590266,
+      "grad_norm": 3.30585879196689,
+      "learning_rate": 1.82601827274658e-05,
+      "logits/chosen": -12.25,
+      "logits/rejected": -11.0625,
+      "logps/chosen": -616.0,
+      "logps/rejected": -804.0,
+      "loss": 0.2001,
+      "rewards/accuracies": 0.9437500238418579,
+      "rewards/chosen": -3.046875,
+      "rewards/margins": 2.3125,
+      "rewards/rejected": -5.375,
+      "step": 1800
+    },
+    {
+      "epoch": 1.8838304552590266,
+      "eval_logits/chosen": -12.125,
+      "eval_logits/rejected": -11.3125,
+      "eval_logps/chosen": -692.0,
+      "eval_logps/rejected": -740.0,
+      "eval_loss": 0.6733828186988831,
+      "eval_rewards/accuracies": 0.720703125,
+      "eval_rewards/chosen": -3.734375,
+      "eval_rewards/margins": 0.76171875,
+      "eval_rewards/rejected": -4.5,
+      "eval_runtime": 46.6342,
+      "eval_samples_per_second": 42.887,
+      "eval_steps_per_second": 0.686,
+      "step": 1800
+    },
+    {
+      "epoch": 1.8942961800104658,
+      "grad_norm": 2.7354320505605134,
+      "learning_rate": 1.7967316349380213e-05,
+      "logits/chosen": -11.8125,
+      "logits/rejected": -10.625,
+      "logps/chosen": -616.0,
+      "logps/rejected": -804.0,
+      "loss": 0.2058,
+      "rewards/accuracies": 0.9312499761581421,
+      "rewards/chosen": -3.1875,
+      "rewards/margins": 2.125,
+      "rewards/rejected": -5.3125,
+      "step": 1810
+    },
+    {
+      "epoch": 1.9047619047619047,
+      "grad_norm": 3.2963373479019067,
+      "learning_rate": 1.7675494329845516e-05,
+      "logits/chosen": -11.75,
+      "logits/rejected": -10.25,
+      "logps/chosen": -656.0,
+      "logps/rejected": -828.0,
+      "loss": 0.1947,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -3.5,
+      "rewards/margins": 2.078125,
+      "rewards/rejected": -5.59375,
+      "step": 1820
+    },
+    {
+      "epoch": 1.915227629513344,
+      "grad_norm": 3.00664588029746,
+      "learning_rate": 1.738476000464026e-05,
+      "logits/chosen": -10.9375,
+      "logits/rejected": -9.875,
+      "logps/chosen": -632.0,
+      "logps/rejected": -840.0,
+      "loss": 0.1968,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -3.453125,
+      "rewards/margins": 1.96875,
+      "rewards/rejected": -5.4375,
+      "step": 1830
+    },
+    {
+      "epoch": 1.9256933542647827,
+      "grad_norm": 2.9898913463301495,
+      "learning_rate": 1.7095156548019646e-05,
+      "logits/chosen": -11.25,
+      "logits/rejected": -9.5625,
+      "logps/chosen": -696.0,
+      "logps/rejected": -840.0,
+      "loss": 0.2114,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -3.359375,
+      "rewards/margins": 2.09375,
+      "rewards/rejected": -5.4375,
+      "step": 1840
+    },
+    {
+      "epoch": 1.936159079016222,
+      "grad_norm": 2.6563839514296776,
+      "learning_rate": 1.6806726966304064e-05,
+      "logits/chosen": -11.1875,
+      "logits/rejected": -9.9375,
+      "logps/chosen": -640.0,
+      "logps/rejected": -800.0,
+      "loss": 0.2029,
+      "rewards/accuracies": 0.90625,
+      "rewards/chosen": -3.171875,
+      "rewards/margins": 1.9375,
+      "rewards/rejected": -5.125,
+      "step": 1850
+    },
+    {
+      "epoch": 1.9466248037676608,
+      "grad_norm": 3.0964270423850815,
+      "learning_rate": 1.6519514091492625e-05,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -10.0,
+      "logps/chosen": -632.0,
+      "logps/rejected": -776.0,
+      "loss": 0.1849,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -3.1875,
+      "rewards/margins": 1.984375,
+      "rewards/rejected": -5.1875,
+      "step": 1860
+    },
+    {
+      "epoch": 1.9570905285191,
+      "grad_norm": 3.1796958561857784,
+      "learning_rate": 1.6233560574902627e-05,
+      "logits/chosen": -11.625,
+      "logits/rejected": -9.9375,
+      "logps/chosen": -672.0,
+      "logps/rejected": -824.0,
+      "loss": 0.1759,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -2.953125,
+      "rewards/margins": 2.28125,
+      "rewards/rejected": -5.25,
+      "step": 1870
+    },
+    {
+      "epoch": 1.9675562532705388,
+      "grad_norm": 3.178365300446779,
+      "learning_rate": 1.5948908880835753e-05,
+      "logits/chosen": -11.3125,
+      "logits/rejected": -10.0,
+      "logps/chosen": -652.0,
+      "logps/rejected": -812.0,
+      "loss": 0.2016,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -3.203125,
+      "rewards/margins": 2.171875,
+      "rewards/rejected": -5.375,
+      "step": 1880
+    },
+    {
+      "epoch": 1.978021978021978,
+      "grad_norm": 3.0068412430556717,
+      "learning_rate": 1.5665601280272124e-05,
+      "logits/chosen": -11.625,
+      "logits/rejected": -10.0625,
+      "logps/chosen": -648.0,
+      "logps/rejected": -820.0,
+      "loss": 0.2068,
+      "rewards/accuracies": 0.8687499761581421,
+      "rewards/chosen": -3.25,
+      "rewards/margins": 2.109375,
+      "rewards/rejected": -5.375,
+      "step": 1890
+    },
+    {
+      "epoch": 1.988487702773417,
+      "grad_norm": 3.058513632314578,
+      "learning_rate": 1.538367984459298e-05,
+      "logits/chosen": -10.9375,
+      "logits/rejected": -9.5,
+      "logps/chosen": -692.0,
+      "logps/rejected": -896.0,
+      "loss": 0.1713,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -3.421875,
+      "rewards/margins": 2.4375,
+      "rewards/rejected": -5.875,
+      "step": 1900
+    },
+    {
+      "epoch": 1.988487702773417,
+      "eval_logits/chosen": -10.4375,
+      "eval_logits/rejected": -9.25,
+      "eval_logps/chosen": -712.0,
+      "eval_logps/rejected": -768.0,
+      "eval_loss": 0.6680155992507935,
+      "eval_rewards/accuracies": 0.73828125,
+      "eval_rewards/chosen": -3.9375,
+      "eval_rewards/margins": 0.87890625,
+      "eval_rewards/rejected": -4.8125,
+      "eval_runtime": 46.6471,
+      "eval_samples_per_second": 42.875,
+      "eval_steps_per_second": 0.686,
+      "step": 1900
+    },
+    {
+      "epoch": 1.9989534275248562,
+      "grad_norm": 3.7442100220748364,
+      "learning_rate": 1.5103186439333133e-05,
+      "logits/chosen": -10.8125,
+      "logits/rejected": -9.0625,
+      "logps/chosen": -688.0,
+      "logps/rejected": -860.0,
+      "loss": 0.1883,
+      "rewards/accuracies": 0.918749988079071,
+      "rewards/chosen": -3.4375,
+      "rewards/margins": 2.3125,
+      "rewards/rejected": -5.75,
+      "step": 1910
+    },
+    {
+      "epoch": 2.009419152276295,
+      "grad_norm": 1.2642871885593199,
+      "learning_rate": 1.4824162717963829e-05,
+      "logits/chosen": -10.625,
+      "logits/rejected": -8.3125,
+      "logps/chosen": -700.0,
+      "logps/rejected": -940.0,
+      "loss": 0.0372,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.328125,
+      "rewards/margins": 3.09375,
+      "rewards/rejected": -6.40625,
+      "step": 1920
+    },
+    {
+      "epoch": 2.0198848770277342,
+      "grad_norm": 0.9740537747095466,
+      "learning_rate": 1.4546650115707192e-05,
+      "logits/chosen": -9.6875,
+      "logits/rejected": -7.59375,
+      "logps/chosen": -628.0,
+      "logps/rejected": -904.0,
+      "loss": 0.0303,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.3125,
+      "rewards/margins": 3.03125,
+      "rewards/rejected": -6.34375,
+      "step": 1930
+    },
+    {
+      "epoch": 2.030350601779173,
+      "grad_norm": 1.2173165412251832,
+      "learning_rate": 1.4270689843383111e-05,
+      "logits/chosen": -10.625,
+      "logits/rejected": -8.375,
+      "logps/chosen": -620.0,
+      "logps/rejected": -884.0,
+      "loss": 0.0245,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -3.046875,
+      "rewards/margins": 3.0625,
+      "rewards/rejected": -6.125,
+      "step": 1940
+    },
+    {
+      "epoch": 2.0408163265306123,
+      "grad_norm": 1.6582312576715226,
+      "learning_rate": 1.3996322881289347e-05,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -8.8125,
+      "logps/chosen": -596.0,
+      "logps/rejected": -860.0,
+      "loss": 0.0289,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -2.84375,
+      "rewards/margins": 3.15625,
+      "rewards/rejected": -6.0,
+      "step": 1950
+    },
+    {
+      "epoch": 2.051282051282051,
+      "grad_norm": 1.6830712488975539,
+      "learning_rate": 1.3723589973115963e-05,
+      "logits/chosen": -11.0,
+      "logits/rejected": -9.4375,
+      "logps/chosen": -568.0,
+      "logps/rejected": -896.0,
+      "loss": 0.033,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -2.796875,
+      "rewards/margins": 3.265625,
+      "rewards/rejected": -6.0625,
+      "step": 1960
+    },
+    {
+      "epoch": 2.0617477760334904,
+      "grad_norm": 1.544935536436392,
+      "learning_rate": 1.3452531619894837e-05,
+      "logits/chosen": -11.625,
+      "logits/rejected": -9.3125,
+      "logps/chosen": -612.0,
+      "logps/rejected": -896.0,
+      "loss": 0.0318,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -2.8125,
+      "rewards/margins": 3.296875,
+      "rewards/rejected": -6.125,
+      "step": 1970
+    },
+    {
+      "epoch": 2.072213500784929,
+      "grad_norm": 1.4851941455915225,
+      "learning_rate": 1.3183188073985245e-05,
+      "logits/chosen": -10.625,
+      "logits/rejected": -8.0625,
+      "logps/chosen": -624.0,
+      "logps/rejected": -940.0,
+      "loss": 0.0276,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -2.9375,
+      "rewards/margins": 3.59375,
+      "rewards/rejected": -6.53125,
+      "step": 1980
+    },
+    {
+      "epoch": 2.0826792255363684,
+      "grad_norm": 1.8084566553201633,
+      "learning_rate": 1.2915599333096351e-05,
+      "logits/chosen": -10.4375,
+      "logits/rejected": -8.125,
+      "logps/chosen": -636.0,
+      "logps/rejected": -920.0,
+      "loss": 0.0336,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -3.0,
+      "rewards/margins": 3.359375,
+      "rewards/rejected": -6.375,
+      "step": 1990
+    },
+    {
+      "epoch": 2.0931449502878072,
+      "grad_norm": 1.2316994077894132,
+      "learning_rate": 1.264980513434752e-05,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -8.75,
+      "logps/chosen": -588.0,
+      "logps/rejected": -904.0,
+      "loss": 0.0184,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -2.84375,
+      "rewards/margins": 3.546875,
+      "rewards/rejected": -6.40625,
+      "step": 2000
+    },
+    {
+      "epoch": 2.0931449502878072,
+      "eval_logits/chosen": -11.0625,
+      "eval_logits/rejected": -9.875,
+      "eval_logps/chosen": -704.0,
+      "eval_logps/rejected": -768.0,
+      "eval_loss": 0.6845312714576721,
+      "eval_rewards/accuracies": 0.73046875,
+      "eval_rewards/chosen": -3.859375,
+      "eval_rewards/margins": 0.9453125,
+      "eval_rewards/rejected": -4.8125,
+      "eval_runtime": 46.6361,
+      "eval_samples_per_second": 42.885,
+      "eval_steps_per_second": 0.686,
+      "step": 2000
+    },
+    {
+      "epoch": 2.1036106750392465,
+      "grad_norm": 1.4733311286442603,
+      "learning_rate": 1.2385844948367324e-05,
+      "logits/chosen": -11.75,
+      "logits/rejected": -9.1875,
+      "logps/chosen": -656.0,
+      "logps/rejected": -964.0,
+      "loss": 0.0251,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -3.109375,
+      "rewards/margins": 3.4375,
+      "rewards/rejected": -6.53125,
+      "step": 2010
+    },
+    {
+      "epoch": 2.1140763997906853,
+      "grad_norm": 1.8091307806488266,
+      "learning_rate": 1.2123757973432114e-05,
+      "logits/chosen": -11.5625,
+      "logits/rejected": -9.125,
+      "logps/chosen": -664.0,
+      "logps/rejected": -976.0,
+      "loss": 0.0232,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -3.453125,
+      "rewards/margins": 3.390625,
+      "rewards/rejected": -6.84375,
+      "step": 2020
+    },
+    {
+      "epoch": 2.1245421245421245,
+      "grad_norm": 0.5231648635098949,
+      "learning_rate": 1.1863583129645106e-05,
+      "logits/chosen": -12.0,
+      "logits/rejected": -9.4375,
+      "logps/chosen": -640.0,
+      "logps/rejected": -920.0,
+      "loss": 0.0265,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -3.09375,
+      "rewards/margins": 3.296875,
+      "rewards/rejected": -6.375,
+      "step": 2030
+    },
+    {
+      "epoch": 2.1350078492935634,
+      "grad_norm": 1.8320656141374727,
+      "learning_rate": 1.1605359053156607e-05,
+      "logits/chosen": -11.5,
+      "logits/rejected": -9.625,
+      "logps/chosen": -632.0,
+      "logps/rejected": -960.0,
+      "loss": 0.0356,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -3.3125,
+      "rewards/margins": 3.40625,
+      "rewards/rejected": -6.71875,
+      "step": 2040
+    },
+    {
+      "epoch": 2.1454735740450026,
+      "grad_norm": 0.0,
+      "learning_rate": 1.1349124090426644e-05,
+      "logits/chosen": -10.75,
+      "logits/rejected": -8.4375,
+      "logps/chosen": -672.0,
+      "logps/rejected": -964.0,
+      "loss": 0.0306,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -3.5,
+      "rewards/margins": 3.28125,
+      "rewards/rejected": -6.78125,
+      "step": 2050
+    },
+    {
+      "epoch": 2.155939298796442,
+      "grad_norm": 1.2273836154154534,
+      "learning_rate": 1.1094916292530406e-05,
+      "logits/chosen": -11.25,
+      "logits/rejected": -8.5625,
+      "logps/chosen": -664.0,
+      "logps/rejected": -964.0,
+      "loss": 0.0265,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -3.171875,
+      "rewards/margins": 3.546875,
+      "rewards/rejected": -6.71875,
+      "step": 2060
+    },
+    {
+      "epoch": 2.1664050235478807,
+      "grad_norm": 2.572344550443592,
+      "learning_rate": 1.0842773409507623e-05,
+      "logits/chosen": -10.375,
+      "logits/rejected": -8.25,
+      "logps/chosen": -648.0,
+      "logps/rejected": -984.0,
+      "loss": 0.0238,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -3.40625,
+      "rewards/margins": 3.546875,
+      "rewards/rejected": -6.96875,
+      "step": 2070
+    },
+    {
+      "epoch": 2.17687074829932,
+      "grad_norm": 2.355005529883559,
+      "learning_rate": 1.0592732884756754e-05,
+      "logits/chosen": -11.125,
+      "logits/rejected": -8.1875,
+      "logps/chosen": -692.0,
+      "logps/rejected": -988.0,
+      "loss": 0.0255,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.484375,
+      "rewards/margins": 3.359375,
+      "rewards/rejected": -6.84375,
+      "step": 2080
+    },
+    {
+      "epoch": 2.1873364730507587,
+      "grad_norm": 2.6076325534245934,
+      "learning_rate": 1.0344831849474507e-05,
+      "logits/chosen": -11.0625,
+      "logits/rejected": -9.0,
+      "logps/chosen": -684.0,
+      "logps/rejected": -980.0,
+      "loss": 0.0335,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.34375,
+      "rewards/margins": 3.421875,
+      "rewards/rejected": -6.78125,
+      "step": 2090
+    },
+    {
+      "epoch": 2.197802197802198,
+      "grad_norm": 1.6256941616554992,
+      "learning_rate": 1.0099107117141879e-05,
+      "logits/chosen": -11.8125,
+      "logits/rejected": -9.375,
+      "logps/chosen": -644.0,
+      "logps/rejected": -936.0,
+      "loss": 0.0313,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.0625,
+      "rewards/margins": 3.375,
+      "rewards/rejected": -6.4375,
+      "step": 2100
+    },
+    {
+      "epoch": 2.197802197802198,
+      "eval_logits/chosen": -11.25,
+      "eval_logits/rejected": -10.125,
+      "eval_logps/chosen": -720.0,
+      "eval_logps/rejected": -784.0,
+      "eval_loss": 0.6798359155654907,
+      "eval_rewards/accuracies": 0.740234375,
+      "eval_rewards/chosen": -4.0,
+      "eval_rewards/margins": 0.95703125,
+      "eval_rewards/rejected": -4.96875,
+      "eval_runtime": 46.6524,
+      "eval_samples_per_second": 42.87,
+      "eval_steps_per_second": 0.686,
+      "step": 2100
+    },
+    {
+      "epoch": 2.208267922553637,
+      "grad_norm": 1.0362742987373548,
+      "learning_rate": 9.855595178057334e-06,
+      "logits/chosen": -11.25,
+      "logits/rejected": -9.125,
+      "logps/chosen": -652.0,
+      "logps/rejected": -928.0,
+      "loss": 0.0235,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -3.125,
+      "rewards/margins": 3.3125,
+      "rewards/rejected": -6.4375,
+      "step": 2110
+    },
+    {
+      "epoch": 2.218733647305076,
+      "grad_norm": 1.050345037025324,
+      "learning_rate": 9.614332193917949e-06,
+      "logits/chosen": -11.1875,
+      "logits/rejected": -9.1875,
+      "logps/chosen": -652.0,
+      "logps/rejected": -948.0,
+      "loss": 0.0397,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.265625,
+      "rewards/margins": 3.3125,
+      "rewards/rejected": -6.5625,
+      "step": 2120
+    },
+    {
+      "epoch": 2.229199372056515,
+      "grad_norm": 1.702731470501781,
+      "learning_rate": 9.375353992449385e-06,
+      "logits/chosen": -11.125,
+      "logits/rejected": -8.9375,
+      "logps/chosen": -648.0,
+      "logps/rejected": -928.0,
+      "loss": 0.0327,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -3.125,
+      "rewards/margins": 3.171875,
+      "rewards/rejected": -6.3125,
+      "step": 2130
+    },
+    {
+      "epoch": 2.239665096807954,
+      "grad_norm": 1.1146351887042307,
+      "learning_rate": 9.138696062085442e-06,
+      "logits/chosen": -11.625,
+      "logits/rejected": -9.4375,
+      "logps/chosen": -600.0,
+      "logps/rejected": -936.0,
+      "loss": 0.0212,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -2.875,
+      "rewards/margins": 3.484375,
+      "rewards/rejected": -6.375,
+      "step": 2140
+    },
+    {
+      "epoch": 2.250130821559393,
+      "grad_norm": 1.1192401524087157,
+      "learning_rate": 8.904393546698007e-06,
+      "logits/chosen": -11.8125,
+      "logits/rejected": -9.3125,
+      "logps/chosen": -672.0,
+      "logps/rejected": -968.0,
+      "loss": 0.0266,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -2.90625,
+      "rewards/margins": 3.5625,
+      "rewards/rejected": -6.46875,
+      "step": 2150
+    },
+    {
+      "epoch": 2.260596546310832,
+      "grad_norm": 1.0577323804895846,
+      "learning_rate": 8.672481240378142e-06,
+      "logits/chosen": -11.5625,
+      "logits/rejected": -9.375,
+      "logps/chosen": -624.0,
+      "logps/rejected": -948.0,
+      "loss": 0.0378,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -3.046875,
+      "rewards/margins": 3.34375,
+      "rewards/rejected": -6.375,
+      "step": 2160
+    },
+    {
+      "epoch": 2.271062271062271,
+      "grad_norm": 2.9582583615096465,
+      "learning_rate": 8.44299358226919e-06,
+      "logits/chosen": -11.5,
+      "logits/rejected": -9.375,
+      "logps/chosen": -676.0,
+      "logps/rejected": -928.0,
+      "loss": 0.0251,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -3.203125,
+      "rewards/margins": 3.28125,
+      "rewards/rejected": -6.5,
+      "step": 2170
+    },
+    {
+      "epoch": 2.2815279958137102,
+      "grad_norm": 2.1527236207784943,
+      "learning_rate": 8.215964651452457e-06,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -9.1875,
+      "logps/chosen": -684.0,
+      "logps/rejected": -960.0,
+      "loss": 0.028,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.109375,
+      "rewards/margins": 3.375,
+      "rewards/rejected": -6.5,
+      "step": 2180
+    },
+    {
+      "epoch": 2.291993720565149,
+      "grad_norm": 1.3246928136229525,
+      "learning_rate": 7.991428161886502e-06,
+      "logits/chosen": -11.1875,
+      "logits/rejected": -9.0,
+      "logps/chosen": -624.0,
+      "logps/rejected": -964.0,
+      "loss": 0.0247,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.078125,
+      "rewards/margins": 3.34375,
+      "rewards/rejected": -6.4375,
+      "step": 2190
+    },
+    {
+      "epoch": 2.3024594453165883,
+      "grad_norm": 1.3777355813128263,
+      "learning_rate": 7.76941745740061e-06,
+      "logits/chosen": -10.875,
+      "logits/rejected": -8.5625,
+      "logps/chosen": -636.0,
+      "logps/rejected": -868.0,
+      "loss": 0.0401,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -3.15625,
+      "rewards/margins": 3.0,
+      "rewards/rejected": -6.15625,
+      "step": 2200
+    },
+    {
+      "epoch": 2.3024594453165883,
+      "eval_logits/chosen": -10.6875,
+      "eval_logits/rejected": -9.375,
+      "eval_logps/chosen": -732.0,
+      "eval_logps/rejected": -800.0,
+      "eval_loss": 0.6864687204360962,
+      "eval_rewards/accuracies": 0.736328125,
+      "eval_rewards/chosen": -4.15625,
+      "eval_rewards/margins": 0.94921875,
+      "eval_rewards/rejected": -5.09375,
+      "eval_runtime": 46.6353,
+      "eval_samples_per_second": 42.886,
+      "eval_steps_per_second": 0.686,
+      "step": 2200
+    },
+    {
+      "epoch": 2.312925170068027,
+      "grad_norm": 0.6632152914047581,
+      "learning_rate": 7.549965506743123e-06,
+      "logits/chosen": -10.75,
+      "logits/rejected": -7.9375,
+      "logps/chosen": -676.0,
+      "logps/rejected": -932.0,
+      "loss": 0.029,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -3.515625,
+      "rewards/margins": 3.171875,
+      "rewards/rejected": -6.6875,
+      "step": 2210
+    },
+    {
+      "epoch": 2.3233908948194664,
+      "grad_norm": 1.8176671432800557,
+      "learning_rate": 7.333104898685631e-06,
+      "logits/chosen": -11.3125,
+      "logits/rejected": -9.125,
+      "logps/chosen": -684.0,
+      "logps/rejected": -976.0,
+      "loss": 0.0383,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -3.296875,
+      "rewards/margins": 3.3125,
+      "rewards/rejected": -6.625,
+      "step": 2220
+    },
+    {
+      "epoch": 2.333856619570905,
+      "grad_norm": 1.3431154179510505,
+      "learning_rate": 7.118867837183518e-06,
+      "logits/chosen": -10.9375,
+      "logits/rejected": -9.125,
+      "logps/chosen": -648.0,
+      "logps/rejected": -940.0,
+      "loss": 0.0239,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -3.25,
+      "rewards/margins": 3.296875,
+      "rewards/rejected": -6.53125,
+      "step": 2230
+    },
+    {
+      "epoch": 2.3443223443223444,
+      "grad_norm": 2.467352136770471,
+      "learning_rate": 6.9072861365936045e-06,
+      "logits/chosen": -11.25,
+      "logits/rejected": -8.9375,
+      "logps/chosen": -596.0,
+      "logps/rejected": -924.0,
+      "loss": 0.0208,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.03125,
+      "rewards/margins": 3.390625,
+      "rewards/rejected": -6.4375,
+      "step": 2240
+    },
+    {
+      "epoch": 2.3547880690737832,
+      "grad_norm": 2.198113395796719,
+      "learning_rate": 6.6983912169497005e-06,
+      "logits/chosen": -11.1875,
+      "logits/rejected": -8.6875,
+      "logps/chosen": -668.0,
+      "logps/rejected": -948.0,
+      "loss": 0.0183,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -3.28125,
+      "rewards/margins": 3.375,
+      "rewards/rejected": -6.65625,
+      "step": 2250
+    },
+    {
+      "epoch": 2.3652537938252225,
+      "grad_norm": 2.1463601171006643,
+      "learning_rate": 6.49221409929677e-06,
+      "logits/chosen": -11.375,
+      "logits/rejected": -9.375,
+      "logps/chosen": -680.0,
+      "logps/rejected": -984.0,
+      "loss": 0.0323,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.078125,
+      "rewards/margins": 3.46875,
+      "rewards/rejected": -6.5625,
+      "step": 2260
+    },
+    {
+      "epoch": 2.3757195185766613,
+      "grad_norm": 1.3281795027550845,
+      "learning_rate": 6.2887854010842075e-06,
+      "logits/chosen": -10.875,
+      "logits/rejected": -8.5625,
+      "logps/chosen": -660.0,
+      "logps/rejected": -952.0,
+      "loss": 0.0239,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.125,
+      "rewards/margins": 3.5,
+      "rewards/rejected": -6.625,
+      "step": 2270
+    },
+    {
+      "epoch": 2.3861852433281006,
+      "grad_norm": 1.3165230202814568,
+      "learning_rate": 6.0881353316191385e-06,
+      "logits/chosen": -11.5,
+      "logits/rejected": -8.8125,
+      "logps/chosen": -644.0,
+      "logps/rejected": -996.0,
+      "loss": 0.0229,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -2.984375,
+      "rewards/margins": 3.765625,
+      "rewards/rejected": -6.75,
+      "step": 2280
+    },
+    {
+      "epoch": 2.3966509680795394,
+      "grad_norm": 0.0,
+      "learning_rate": 5.890293687580381e-06,
+      "logits/chosen": -10.625,
+      "logits/rejected": -8.3125,
+      "logps/chosen": -652.0,
+      "logps/rejected": -960.0,
+      "loss": 0.037,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -3.15625,
+      "rewards/margins": 3.40625,
+      "rewards/rejected": -6.5625,
+      "step": 2290
+    },
+    {
+      "epoch": 2.4071166928309786,
+      "grad_norm": 1.0986592674305082,
+      "learning_rate": 5.695289848593532e-06,
+      "logits/chosen": -10.75,
+      "logits/rejected": -8.375,
+      "logps/chosen": -632.0,
+      "logps/rejected": -976.0,
+      "loss": 0.0211,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -3.0625,
+      "rewards/margins": 3.703125,
+      "rewards/rejected": -6.78125,
+      "step": 2300
+    },
+    {
+      "epoch": 2.4071166928309786,
+      "eval_logits/chosen": -10.125,
+      "eval_logits/rejected": -8.75,
+      "eval_logps/chosen": -740.0,
+      "eval_logps/rejected": -812.0,
+      "eval_loss": 0.6873906254768372,
+      "eval_rewards/accuracies": 0.73828125,
+      "eval_rewards/chosen": -4.21875,
+      "eval_rewards/margins": 1.0078125,
+      "eval_rewards/rejected": -5.21875,
+      "eval_runtime": 46.6344,
+      "eval_samples_per_second": 42.887,
+      "eval_steps_per_second": 0.686,
+      "step": 2300
+    },
+    {
+      "epoch": 2.4175824175824174,
+      "grad_norm": 1.0095492445495,
+      "learning_rate": 5.50315277286812e-06,
+      "logits/chosen": -10.5625,
+      "logits/rejected": -8.1875,
+      "logps/chosen": -620.0,
+      "logps/rejected": -988.0,
+      "loss": 0.0233,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.1875,
+      "rewards/margins": 3.6875,
+      "rewards/rejected": -6.875,
+      "step": 2310
+    },
+    {
+      "epoch": 2.4280481423338567,
+      "grad_norm": 0.0,
+      "learning_rate": 5.313910992897281e-06,
+      "logits/chosen": -11.0625,
+      "logits/rejected": -8.5625,
+      "logps/chosen": -676.0,
+      "logps/rejected": -968.0,
+      "loss": 0.0301,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -3.046875,
+      "rewards/margins": 3.5625,
+      "rewards/rejected": -6.59375,
+      "step": 2320
+    },
+    {
+      "epoch": 2.4385138670852955,
+      "grad_norm": 0.0029568308092483434,
+      "learning_rate": 5.127592611220658e-06,
+      "logits/chosen": -10.6875,
+      "logits/rejected": -8.4375,
+      "logps/chosen": -676.0,
+      "logps/rejected": -944.0,
+      "loss": 0.02,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.234375,
+      "rewards/margins": 3.359375,
+      "rewards/rejected": -6.59375,
+      "step": 2330
+    },
+    {
+      "epoch": 2.4489795918367347,
+      "grad_norm": 2.0442430938540697,
+      "learning_rate": 4.944225296251159e-06,
+      "logits/chosen": -11.375,
+      "logits/rejected": -8.8125,
+      "logps/chosen": -716.0,
+      "logps/rejected": -1004.0,
+      "loss": 0.0222,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -3.15625,
+      "rewards/margins": 3.59375,
+      "rewards/rejected": -6.75,
+      "step": 2340
+    },
+    {
+      "epoch": 2.4594453165881736,
+      "grad_norm": 0.8259148196205537,
+      "learning_rate": 4.763836278166172e-06,
+      "logits/chosen": -11.1875,
+      "logits/rejected": -8.0,
+      "logps/chosen": -624.0,
+      "logps/rejected": -944.0,
+      "loss": 0.0135,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -3.0625,
+      "rewards/margins": 3.53125,
+      "rewards/rejected": -6.59375,
+      "step": 2350
+    },
+    {
+      "epoch": 2.469911041339613,
+      "grad_norm": 0.9009328571483026,
+      "learning_rate": 4.586452344863876e-06,
+      "logits/chosen": -11.0,
+      "logits/rejected": -8.0625,
+      "logps/chosen": -608.0,
+      "logps/rejected": -928.0,
+      "loss": 0.0248,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -3.0625,
+      "rewards/margins": 3.65625,
+      "rewards/rejected": -6.71875,
+      "step": 2360
+    },
+    {
+      "epoch": 2.4803767660910516,
+      "grad_norm": 0.9954189130008081,
+      "learning_rate": 4.412099837985192e-06,
+      "logits/chosen": -10.9375,
+      "logits/rejected": -8.3125,
+      "logps/chosen": -632.0,
+      "logps/rejected": -916.0,
+      "loss": 0.0181,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -3.078125,
+      "rewards/margins": 3.265625,
+      "rewards/rejected": -6.34375,
+      "step": 2370
+    },
+    {
+      "epoch": 2.490842490842491,
+      "grad_norm": 0.0,
+      "learning_rate": 4.240804649002089e-06,
+      "logits/chosen": -10.5,
+      "logits/rejected": -8.5,
+      "logps/chosen": -624.0,
+      "logps/rejected": -948.0,
+      "loss": 0.0201,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.21875,
+      "rewards/margins": 3.40625,
+      "rewards/rejected": -6.625,
+      "step": 2380
+    },
+    {
+      "epoch": 2.50130821559393,
+      "grad_norm": 0.7860084244488942,
+      "learning_rate": 4.072592215372598e-06,
+      "logits/chosen": -10.375,
+      "logits/rejected": -8.0625,
+      "logps/chosen": -624.0,
+      "logps/rejected": -904.0,
+      "loss": 0.0237,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -3.140625,
+      "rewards/margins": 3.34375,
+      "rewards/rejected": -6.5,
+      "step": 2390
+    },
+    {
+      "epoch": 2.511773940345369,
+      "grad_norm": 1.2623899013822348,
+      "learning_rate": 3.907487516763389e-06,
+      "logits/chosen": -11.125,
+      "logits/rejected": -8.375,
+      "logps/chosen": -672.0,
+      "logps/rejected": -968.0,
+      "loss": 0.0239,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -3.265625,
+      "rewards/margins": 3.421875,
+      "rewards/rejected": -6.6875,
+      "step": 2400
+    },
+    {
+      "epoch": 2.511773940345369,
+      "eval_logits/chosen": -10.1875,
+      "eval_logits/rejected": -8.875,
+      "eval_logps/chosen": -736.0,
+      "eval_logps/rejected": -800.0,
+      "eval_loss": 0.6858203411102295,
+      "eval_rewards/accuracies": 0.73828125,
+      "eval_rewards/chosen": -4.15625,
+      "eval_rewards/margins": 0.9765625,
+      "eval_rewards/rejected": -5.125,
+      "eval_runtime": 46.6437,
+      "eval_samples_per_second": 42.878,
+      "eval_steps_per_second": 0.686,
+      "step": 2400
+    },
+    {
+      "epoch": 2.5222396650968077,
+      "grad_norm": 2.2939797564506894,
+      "learning_rate": 3.745515071340286e-06,
+      "logits/chosen": -10.75,
+      "logits/rejected": -8.75,
+      "logps/chosen": -672.0,
+      "logps/rejected": -956.0,
+      "loss": 0.0269,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -3.15625,
+      "rewards/margins": 3.453125,
+      "rewards/rejected": -6.59375,
+      "step": 2410
+    },
+    {
+      "epoch": 2.532705389848247,
+      "grad_norm": 1.4894004335543474,
+      "learning_rate": 3.586698932127236e-06,
+      "logits/chosen": -10.375,
+      "logits/rejected": -8.0,
+      "logps/chosen": -632.0,
+      "logps/rejected": -928.0,
+      "loss": 0.032,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -3.03125,
+      "rewards/margins": 3.34375,
+      "rewards/rejected": -6.375,
+      "step": 2420
+    },
+    {
+      "epoch": 2.5431711145996863,
+      "grad_norm": 1.0198172465645305,
+      "learning_rate": 3.4310626834344744e-06,
+      "logits/chosen": -10.5,
+      "logits/rejected": -8.375,
+      "logps/chosen": -660.0,
+      "logps/rejected": -968.0,
+      "loss": 0.0116,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -3.203125,
+      "rewards/margins": 3.5625,
+      "rewards/rejected": -6.75,
+      "step": 2430
+    },
+    {
+      "epoch": 2.553636839351125,
+      "grad_norm": 0.9853090185166195,
+      "learning_rate": 3.2786294373562343e-06,
+      "logits/chosen": -10.875,
+      "logits/rejected": -7.96875,
+      "logps/chosen": -644.0,
+      "logps/rejected": -984.0,
+      "loss": 0.0234,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.265625,
+      "rewards/margins": 3.640625,
+      "rewards/rejected": -6.90625,
+      "step": 2440
+    },
+    {
+      "epoch": 2.564102564102564,
+      "grad_norm": 1.2891520821789724,
+      "learning_rate": 3.1294218303385526e-06,
+      "logits/chosen": -10.6875,
+      "logits/rejected": -8.6875,
+      "logps/chosen": -668.0,
+      "logps/rejected": -940.0,
+      "loss": 0.0286,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -3.234375,
+      "rewards/margins": 3.125,
+      "rewards/rejected": -6.375,
+      "step": 2450
+    },
+    {
+      "epoch": 2.574568288854003,
+      "grad_norm": 1.9956402445166013,
+      "learning_rate": 2.9834620198177483e-06,
+      "logits/chosen": -11.125,
+      "logits/rejected": -8.75,
+      "logps/chosen": -668.0,
+      "logps/rejected": -960.0,
+      "loss": 0.0203,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -3.03125,
+      "rewards/margins": 3.515625,
+      "rewards/rejected": -6.5625,
+      "step": 2460
+    },
+    {
+      "epoch": 2.5850340136054424,
+      "grad_norm": 1.634804367444625,
+      "learning_rate": 2.840771680930068e-06,
+      "logits/chosen": -10.6875,
+      "logits/rejected": -8.75,
+      "logps/chosen": -604.0,
+      "logps/rejected": -948.0,
+      "loss": 0.0325,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -3.1875,
+      "rewards/margins": 3.46875,
+      "rewards/rejected": -6.65625,
+      "step": 2470
+    },
+    {
+      "epoch": 2.595499738356881,
+      "grad_norm": 2.159520682770496,
+      "learning_rate": 2.701372003292868e-06,
+      "logits/chosen": -10.375,
+      "logits/rejected": -8.625,
+      "logps/chosen": -648.0,
+      "logps/rejected": -988.0,
+      "loss": 0.0304,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.265625,
+      "rewards/margins": 3.34375,
+      "rewards/rejected": -6.625,
+      "step": 2480
+    },
+    {
+      "epoch": 2.60596546310832,
+      "grad_norm": 0.8316538975233464,
+      "learning_rate": 2.56528368785795e-06,
+      "logits/chosen": -10.625,
+      "logits/rejected": -8.8125,
+      "logps/chosen": -672.0,
+      "logps/rejected": -1016.0,
+      "loss": 0.025,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -3.28125,
+      "rewards/margins": 3.484375,
+      "rewards/rejected": -6.78125,
+      "step": 2490
+    },
+    {
+      "epoch": 2.6164311878597593,
+      "grad_norm": 2.4109338527870157,
+      "learning_rate": 2.432526943837471e-06,
+      "logits/chosen": -10.6875,
+      "logits/rejected": -8.25,
+      "logps/chosen": -624.0,
+      "logps/rejected": -952.0,
+      "loss": 0.0188,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -3.125,
+      "rewards/margins": 3.515625,
+      "rewards/rejected": -6.625,
+      "step": 2500
+    },
+    {
+      "epoch": 2.6164311878597593,
+      "eval_logits/chosen": -10.125,
+      "eval_logits/rejected": -8.8125,
+      "eval_logps/chosen": -744.0,
+      "eval_logps/rejected": -816.0,
+      "eval_loss": 0.6902031302452087,
+      "eval_rewards/accuracies": 0.732421875,
+      "eval_rewards/chosen": -4.28125,
+      "eval_rewards/margins": 0.98828125,
+      "eval_rewards/rejected": -5.25,
+      "eval_runtime": 46.6387,
+      "eval_samples_per_second": 42.883,
+      "eval_steps_per_second": 0.686,
+      "step": 2500
+    },
+    {
+      "epoch": 2.6268969126111985,
+      "grad_norm": 0.42590205670170256,
+      "learning_rate": 2.3031214857028446e-06,
+      "logits/chosen": -10.8125,
+      "logits/rejected": -8.125,
+      "logps/chosen": -676.0,
+      "logps/rejected": -964.0,
+      "loss": 0.0292,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.25,
+      "rewards/margins": 3.59375,
+      "rewards/rejected": -6.84375,
+      "step": 2510
+    },
+    {
+      "epoch": 2.6373626373626373,
+      "grad_norm": 1.9412786188072901,
+      "learning_rate": 2.177086530257139e-06,
+      "logits/chosen": -10.5,
+      "logits/rejected": -8.25,
+      "logps/chosen": -660.0,
+      "logps/rejected": -984.0,
+      "loss": 0.023,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -3.359375,
+      "rewards/margins": 3.53125,
+      "rewards/rejected": -6.90625,
+      "step": 2520
+    },
+    {
+      "epoch": 2.647828362114076,
+      "grad_norm": 1.6832727545876796,
+      "learning_rate": 2.054440793781354e-06,
+      "logits/chosen": -10.6875,
+      "logits/rejected": -8.75,
+      "logps/chosen": -628.0,
+      "logps/rejected": -960.0,
+      "loss": 0.0235,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -3.140625,
+      "rewards/margins": 3.5625,
+      "rewards/rejected": -6.6875,
+      "step": 2530
+    },
+    {
+      "epoch": 2.6582940868655154,
+      "grad_norm": 1.9016311993695802,
+      "learning_rate": 1.9352024892550587e-06,
+      "logits/chosen": -10.5625,
+      "logits/rejected": -7.78125,
+      "logps/chosen": -676.0,
+      "logps/rejected": -956.0,
+      "loss": 0.0339,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -3.40625,
+      "rewards/margins": 3.359375,
+      "rewards/rejected": -6.75,
+      "step": 2540
+    },
+    {
+      "epoch": 2.6687598116169546,
+      "grad_norm": 1.7118558276661497,
+      "learning_rate": 1.8193893236517379e-06,
+      "logits/chosen": -10.875,
+      "logits/rejected": -8.5625,
+      "logps/chosen": -660.0,
+      "logps/rejected": -960.0,
+      "loss": 0.0301,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -3.28125,
+      "rewards/margins": 3.359375,
+      "rewards/rejected": -6.625,
+      "step": 2550
+    },
+    {
+      "epoch": 2.6792255363683934,
+      "grad_norm": 2.1031818563249782,
+      "learning_rate": 1.7070184953092956e-06,
+      "logits/chosen": -10.625,
+      "logits/rejected": -8.25,
+      "logps/chosen": -672.0,
+      "logps/rejected": -1004.0,
+      "loss": 0.0225,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -3.46875,
+      "rewards/margins": 3.53125,
+      "rewards/rejected": -7.0,
+      "step": 2560
+    },
+    {
+      "epoch": 2.6896912611198327,
+      "grad_norm": 1.5770336595429248,
+      "learning_rate": 1.5981066913760918e-06,
+      "logits/chosen": -10.625,
+      "logits/rejected": -8.4375,
+      "logps/chosen": -680.0,
+      "logps/rejected": -988.0,
+      "loss": 0.018,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.25,
+      "rewards/margins": 3.578125,
+      "rewards/rejected": -6.84375,
+      "step": 2570
+    },
+    {
+      "epoch": 2.7001569858712715,
+      "grad_norm": 0.9620930926302751,
+      "learning_rate": 1.492670085332884e-06,
+      "logits/chosen": -10.625,
+      "logits/rejected": -8.25,
+      "logps/chosen": -636.0,
+      "logps/rejected": -936.0,
+      "loss": 0.0328,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.25,
+      "rewards/margins": 3.46875,
+      "rewards/rejected": -6.71875,
+      "step": 2580
+    },
+    {
+      "epoch": 2.7106227106227108,
+      "grad_norm": 2.788926317519354,
+      "learning_rate": 1.3907243345910787e-06,
+      "logits/chosen": -10.375,
+      "logits/rejected": -8.3125,
+      "logps/chosen": -640.0,
+      "logps/rejected": -936.0,
+      "loss": 0.0388,
+      "rewards/accuracies": 0.96875,
+      "rewards/chosen": -3.375,
+      "rewards/margins": 3.21875,
+      "rewards/rejected": -6.59375,
+      "step": 2590
+    },
+    {
+      "epoch": 2.7210884353741496,
+      "grad_norm": 1.0213093964974433,
+      "learning_rate": 1.292284578167538e-06,
+      "logits/chosen": -10.6875,
+      "logits/rejected": -8.625,
+      "logps/chosen": -640.0,
+      "logps/rejected": -960.0,
+      "loss": 0.0145,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.15625,
+      "rewards/margins": 3.59375,
+      "rewards/rejected": -6.75,
+      "step": 2600
+    },
+    {
+      "epoch": 2.7210884353741496,
+      "eval_logits/chosen": -10.125,
+      "eval_logits/rejected": -8.8125,
+      "eval_logps/chosen": -748.0,
+      "eval_logps/rejected": -816.0,
+      "eval_loss": 0.6873515844345093,
+      "eval_rewards/accuracies": 0.73828125,
+      "eval_rewards/chosen": -4.28125,
+      "eval_rewards/margins": 0.984375,
+      "eval_rewards/rejected": -5.28125,
+      "eval_runtime": 46.6463,
+      "eval_samples_per_second": 42.876,
+      "eval_steps_per_second": 0.686,
+      "step": 2600
+    },
+    {
+      "epoch": 2.731554160125589,
+      "grad_norm": 2.202957142588099,
+      "learning_rate": 1.1973654344364925e-06,
+      "logits/chosen": -10.5,
+      "logits/rejected": -8.4375,
+      "logps/chosen": -608.0,
+      "logps/rejected": -960.0,
+      "loss": 0.0281,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.0625,
+      "rewards/margins": 3.65625,
+      "rewards/rejected": -6.71875,
+      "step": 2610
+    },
+    {
+      "epoch": 2.7420198848770276,
+      "grad_norm": 1.0834089393686006,
+      "learning_rate": 1.1059809989586583e-06,
+      "logits/chosen": -10.625,
+      "logits/rejected": -7.90625,
+      "logps/chosen": -624.0,
+      "logps/rejected": -972.0,
+      "loss": 0.0153,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -3.234375,
+      "rewards/margins": 3.6875,
+      "rewards/rejected": -6.9375,
+      "step": 2620
+    },
+    {
+      "epoch": 2.752485609628467,
+      "grad_norm": 0.008191767531379832,
+      "learning_rate": 1.018144842388044e-06,
+      "logits/chosen": -10.375,
+      "logits/rejected": -8.4375,
+      "logps/chosen": -632.0,
+      "logps/rejected": -968.0,
+      "loss": 0.0288,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -3.3125,
+      "rewards/margins": 3.484375,
+      "rewards/rejected": -6.8125,
+      "step": 2630
+    },
+    {
+      "epoch": 2.7629513343799057,
+      "grad_norm": 1.4335746413881238,
+      "learning_rate": 9.338700084567109e-07,
+      "logits/chosen": -10.875,
+      "logits/rejected": -8.5,
+      "logps/chosen": -664.0,
+      "logps/rejected": -952.0,
+      "loss": 0.0244,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.28125,
+      "rewards/margins": 3.40625,
+      "rewards/rejected": -6.6875,
+      "step": 2640
+    },
+    {
+      "epoch": 2.773417059131345,
+      "grad_norm": 2.171309716772569,
+      "learning_rate": 8.531690120377606e-07,
+      "logits/chosen": -10.5,
+      "logits/rejected": -8.25,
+      "logps/chosen": -648.0,
+      "logps/rejected": -972.0,
+      "loss": 0.0252,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.234375,
+      "rewards/margins": 3.640625,
+      "rewards/rejected": -6.875,
+      "step": 2650
+    },
+    {
+      "epoch": 2.7838827838827838,
+      "grad_norm": 0.697164019903937,
+      "learning_rate": 7.760538372868636e-07,
+      "logits/chosen": -10.75,
+      "logits/rejected": -7.9375,
+      "logps/chosen": -628.0,
+      "logps/rejected": -936.0,
+      "loss": 0.0281,
+      "rewards/accuracies": 0.956250011920929,
+      "rewards/chosen": -3.21875,
+      "rewards/margins": 3.5,
+      "rewards/rejected": -6.71875,
+      "step": 2660
+    },
+    {
+      "epoch": 2.794348508634223,
+      "grad_norm": 1.9609473238842514,
+      "learning_rate": 7.025359358626166e-07,
+      "logits/chosen": -10.4375,
+      "logits/rejected": -7.46875,
+      "logps/chosen": -632.0,
+      "logps/rejected": -916.0,
+      "loss": 0.0328,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -3.21875,
+      "rewards/margins": 3.4375,
+      "rewards/rejected": -6.65625,
+      "step": 2670
+    },
+    {
+      "epoch": 2.804814233385662,
+      "grad_norm": 0.2620353206138499,
+      "learning_rate": 6.326262252259463e-07,
+      "logits/chosen": -10.75,
+      "logits/rejected": -8.5625,
+      "logps/chosen": -680.0,
+      "logps/rejected": -996.0,
+      "loss": 0.0146,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -3.203125,
+      "rewards/margins": 3.6875,
+      "rewards/rejected": -6.875,
+      "step": 2680
+    },
+    {
+      "epoch": 2.815279958137101,
+      "grad_norm": 0.0075468918667812215,
+      "learning_rate": 5.66335087018871e-07,
+      "logits/chosen": -10.4375,
+      "logits/rejected": -8.1875,
+      "logps/chosen": -644.0,
+      "logps/rejected": -948.0,
+      "loss": 0.0231,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -3.40625,
+      "rewards/margins": 3.3125,
+      "rewards/rejected": -6.71875,
+      "step": 2690
+    },
+    {
+      "epoch": 2.82574568288854,
+      "grad_norm": 1.7002384376147264,
+      "learning_rate": 5.036723655228226e-07,
+      "logits/chosen": -11.0625,
+      "logits/rejected": -8.25,
+      "logps/chosen": -664.0,
+      "logps/rejected": -964.0,
+      "loss": 0.0229,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -3.234375,
+      "rewards/margins": 3.4375,
+      "rewards/rejected": -6.65625,
+      "step": 2700
+    },
+    {
+      "epoch": 2.82574568288854,
+      "eval_logits/chosen": -10.0625,
+      "eval_logits/rejected": -8.75,
+      "eval_logps/chosen": -752.0,
+      "eval_logps/rejected": -824.0,
+      "eval_loss": 0.6882890462875366,
+      "eval_rewards/accuracies": 0.734375,
+      "eval_rewards/chosen": -4.34375,
+      "eval_rewards/margins": 0.9921875,
+      "eval_rewards/rejected": -5.34375,
+      "eval_runtime": 46.6487,
+      "eval_samples_per_second": 42.874,
+      "eval_steps_per_second": 0.686,
+      "step": 2700
+    },
+    {
+      "epoch": 2.836211407639979,
+      "grad_norm": 1.6738524623907045,
+      "learning_rate": 4.446473661967432e-07,
+      "logits/chosen": -10.75,
+      "logits/rejected": -8.625,
+      "logps/chosen": -672.0,
+      "logps/rejected": -992.0,
+      "loss": 0.0231,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -3.09375,
+      "rewards/margins": 3.796875,
+      "rewards/rejected": -6.90625,
+      "step": 2710
+    },
+    {
+      "epoch": 2.846677132391418,
+      "grad_norm": 1.0139935839578516,
+      "learning_rate": 3.892688542952411e-07,
+      "logits/chosen": -10.625,
+      "logits/rejected": -8.3125,
+      "logps/chosen": -644.0,
+      "logps/rejected": -960.0,
+      "loss": 0.018,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.234375,
+      "rewards/margins": 3.578125,
+      "rewards/rejected": -6.8125,
+      "step": 2720
+    },
+    {
+      "epoch": 2.857142857142857,
+      "grad_norm": 1.873245752339602,
+      "learning_rate": 3.3754505356694e-07,
+      "logits/chosen": -10.6875,
+      "logits/rejected": -8.3125,
+      "logps/chosen": -636.0,
+      "logps/rejected": -1012.0,
+      "loss": 0.0221,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -3.28125,
+      "rewards/margins": 3.765625,
+      "rewards/rejected": -7.0625,
+      "step": 2730
+    },
+    {
+      "epoch": 2.867608581894296,
+      "grad_norm": 0.0018923631905276043,
+      "learning_rate": 2.8948364503322276e-07,
+      "logits/chosen": -10.5625,
+      "logits/rejected": -8.8125,
+      "logps/chosen": -660.0,
+      "logps/rejected": -1008.0,
+      "loss": 0.0196,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.40625,
+      "rewards/margins": 3.671875,
+      "rewards/rejected": -7.0625,
+      "step": 2740
+    },
+    {
+      "epoch": 2.8780743066457353,
+      "grad_norm": 1.2279968290391152,
+      "learning_rate": 2.4509176584762706e-07,
+      "logits/chosen": -10.5625,
+      "logits/rejected": -7.75,
+      "logps/chosen": -656.0,
+      "logps/rejected": -956.0,
+      "loss": 0.0279,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -3.359375,
+      "rewards/margins": 3.484375,
+      "rewards/rejected": -6.84375,
+      "step": 2750
+    },
+    {
+      "epoch": 2.8885400313971745,
+      "grad_norm": 0.4674885634642118,
+      "learning_rate": 2.0437600823595694e-07,
+      "logits/chosen": -10.6875,
+      "logits/rejected": -7.65625,
+      "logps/chosen": -656.0,
+      "logps/rejected": -972.0,
+      "loss": 0.0241,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.359375,
+      "rewards/margins": 3.484375,
+      "rewards/rejected": -6.8125,
+      "step": 2760
+    },
+    {
+      "epoch": 2.8990057561486133,
+      "grad_norm": 1.7440636579920037,
+      "learning_rate": 1.6734241851733e-07,
+      "logits/chosen": -10.375,
+      "logits/rejected": -7.59375,
+      "logps/chosen": -696.0,
+      "logps/rejected": -992.0,
+      "loss": 0.0229,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -3.34375,
+      "rewards/margins": 3.46875,
+      "rewards/rejected": -6.8125,
+      "step": 2770
+    },
+    {
+      "epoch": 2.909471480900052,
+      "grad_norm": 0.6349021929420012,
+      "learning_rate": 1.339964962062984e-07,
+      "logits/chosen": -10.9375,
+      "logits/rejected": -8.125,
+      "logps/chosen": -688.0,
+      "logps/rejected": -1000.0,
+      "loss": 0.0327,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -3.328125,
+      "rewards/margins": 3.515625,
+      "rewards/rejected": -6.84375,
+      "step": 2780
+    },
+    {
+      "epoch": 2.9199372056514914,
+      "grad_norm": 1.25282344977746,
+      "learning_rate": 1.0434319319617136e-07,
+      "logits/chosen": -10.75,
+      "logits/rejected": -8.6875,
+      "logps/chosen": -656.0,
+      "logps/rejected": -992.0,
+      "loss": 0.0284,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -3.1875,
+      "rewards/margins": 3.703125,
+      "rewards/rejected": -6.875,
+      "step": 2790
+    },
+    {
+      "epoch": 2.9304029304029307,
+      "grad_norm": 1.2135867393274895,
+      "learning_rate": 7.838691302365363e-08,
+      "logits/chosen": -10.3125,
+      "logits/rejected": -8.0625,
+      "logps/chosen": -636.0,
+      "logps/rejected": -936.0,
+      "loss": 0.0298,
+      "rewards/accuracies": 0.981249988079071,
+      "rewards/chosen": -3.25,
+      "rewards/margins": 3.5625,
+      "rewards/rejected": -6.8125,
+      "step": 2800
+    },
+    {
+      "epoch": 2.9304029304029307,
+      "eval_logits/chosen": -10.0625,
+      "eval_logits/rejected": -8.75,
+      "eval_logps/chosen": -752.0,
+      "eval_logps/rejected": -824.0,
+      "eval_loss": 0.6883124709129333,
+      "eval_rewards/accuracies": 0.734375,
+      "eval_rewards/chosen": -4.34375,
+      "eval_rewards/margins": 0.9921875,
+      "eval_rewards/rejected": -5.34375,
+      "eval_runtime": 46.6513,
+      "eval_samples_per_second": 42.871,
+      "eval_steps_per_second": 0.686,
+      "step": 2800
+    },
+    {
+      "epoch": 2.9408686551543695,
+      "grad_norm": 1.857475565170949,
+      "learning_rate": 5.613151021490459e-08,
+      "logits/chosen": -10.8125,
+      "logits/rejected": -8.375,
+      "logps/chosen": -668.0,
+      "logps/rejected": -1004.0,
+      "loss": 0.0248,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -3.34375,
+      "rewards/margins": 3.65625,
+      "rewards/rejected": -7.0,
+      "step": 2810
+    },
+    {
+      "epoch": 2.9513343799058083,
+      "grad_norm": 2.2817412042809813,
+      "learning_rate": 3.7580289713159965e-08,
+      "logits/chosen": -10.8125,
+      "logits/rejected": -7.9375,
+      "logps/chosen": -644.0,
+      "logps/rejected": -972.0,
+      "loss": 0.0173,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -3.34375,
+      "rewards/margins": 3.546875,
+      "rewards/rejected": -6.875,
+      "step": 2820
+    },
+    {
+      "epoch": 2.9618001046572475,
+      "grad_norm": 1.352098733314749,
+      "learning_rate": 2.2736006387927277e-08,
+      "logits/chosen": -10.5625,
+      "logits/rejected": -8.5,
+      "logps/chosen": -664.0,
+      "logps/rejected": -960.0,
+      "loss": 0.0279,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -3.28125,
+      "rewards/margins": 3.390625,
+      "rewards/rejected": -6.6875,
+      "step": 2830
+    },
+    {
+      "epoch": 2.9722658294086868,
+      "grad_norm": 1.7390831298624168,
+      "learning_rate": 1.1600864625893604e-08,
+      "logits/chosen": -10.5625,
+      "logits/rejected": -7.5,
+      "logps/chosen": -680.0,
+      "logps/rejected": -956.0,
+      "loss": 0.0231,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -3.421875,
+      "rewards/margins": 3.3125,
+      "rewards/rejected": -6.71875,
+      "step": 2840
+    },
+    {
+      "epoch": 2.9827315541601256,
+      "grad_norm": 0.0,
+      "learning_rate": 4.176518003573549e-09,
+      "logits/chosen": -10.5625,
+      "logits/rejected": -9.1875,
+      "logps/chosen": -640.0,
+      "logps/rejected": -1012.0,
+      "loss": 0.021,
+      "rewards/accuracies": 0.9937499761581421,
+      "rewards/chosen": -3.1875,
+      "rewards/margins": 3.90625,
+      "rewards/rejected": -7.09375,
+      "step": 2850
+    },
+    {
+      "epoch": 2.9931972789115644,
+      "grad_norm": 1.1446044089330367,
+      "learning_rate": 4.640690417528903e-10,
+      "logits/chosen": -10.25,
+      "logits/rejected": -8.1875,
+      "logps/chosen": -668.0,
+      "logps/rejected": -980.0,
+      "loss": 0.0219,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -3.40625,
+      "rewards/margins": 3.421875,
+      "rewards/rejected": -6.8125,
+      "step": 2860
+    },
+    {
+      "epoch": 2.998430141287284,
+      "step": 2865,
+      "total_flos": 0.0,
+      "train_loss": 0.32237146187827226,
+      "train_runtime": 12289.6357,
+      "train_samples_per_second": 14.923,
+      "train_steps_per_second": 0.233
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 2865,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}