diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,5819 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.0,
+  "eval_steps": 500,
+  "global_step": 3822,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0005232862375719519,
+      "grad_norm": 10.351250771822766,
+      "learning_rate": 1.3054830287206266e-07,
+      "logits/chosen": -12.5625,
+      "logits/rejected": -11.6875,
+      "logps/chosen": -430.0,
+      "logps/rejected": -460.0,
+      "loss": 0.6914,
+      "rewards/accuracies": 0.0,
+      "rewards/chosen": 0.0,
+      "rewards/margins": 0.0,
+      "rewards/rejected": 0.0,
+      "step": 1
+    },
+    {
+      "epoch": 0.0052328623757195184,
+      "grad_norm": 9.844286946232794,
+      "learning_rate": 1.3054830287206267e-06,
+      "logits/chosen": -11.3125,
+      "logits/rejected": -11.3125,
+      "logps/chosen": -364.0,
+      "logps/rejected": -290.0,
+      "loss": 0.6938,
+      "rewards/accuracies": 0.2083333283662796,
+      "rewards/chosen": 0.003997802734375,
+      "rewards/margins": 0.0033111572265625,
+      "rewards/rejected": 0.000701904296875,
+      "step": 10
+    },
+    {
+      "epoch": 0.010465724751439037,
+      "grad_norm": 8.840907831144664,
+      "learning_rate": 2.6109660574412534e-06,
+      "logits/chosen": -11.0625,
+      "logits/rejected": -11.0625,
+      "logps/chosen": -264.0,
+      "logps/rejected": -256.0,
+      "loss": 0.691,
+      "rewards/accuracies": 0.2750000059604645,
+      "rewards/chosen": -0.0019989013671875,
+      "rewards/margins": 0.009521484375,
+      "rewards/rejected": -0.01153564453125,
+      "step": 20
+    },
+    {
+      "epoch": 0.015698587127158554,
+      "grad_norm": 10.083186641323694,
+      "learning_rate": 3.9164490861618806e-06,
+      "logits/chosen": -10.375,
+      "logits/rejected": -10.3125,
+      "logps/chosen": -328.0,
+      "logps/rejected": -318.0,
+      "loss": 0.6803,
+      "rewards/accuracies": 0.42500001192092896,
+      "rewards/chosen": 0.00970458984375,
+      "rewards/margins": 0.023681640625,
+      "rewards/rejected": -0.01397705078125,
+      "step": 30
+    },
+    {
+      "epoch": 0.020931449502878074,
+      "grad_norm": 9.068339095588007,
+      "learning_rate": 5.221932114882507e-06,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -11.0625,
+      "logps/chosen": -336.0,
+      "logps/rejected": -312.0,
+      "loss": 0.6719,
+      "rewards/accuracies": 0.4749999940395355,
+      "rewards/chosen": -0.032470703125,
+      "rewards/margins": 0.030517578125,
+      "rewards/rejected": -0.06298828125,
+      "step": 40
+    },
+    {
+      "epoch": 0.026164311878597593,
+      "grad_norm": 9.546615633761498,
+      "learning_rate": 6.527415143603134e-06,
+      "logits/chosen": -12.1875,
+      "logits/rejected": -12.0,
+      "logps/chosen": -316.0,
+      "logps/rejected": -314.0,
+      "loss": 0.6523,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": -0.09912109375,
+      "rewards/margins": 0.1376953125,
+      "rewards/rejected": -0.2373046875,
+      "step": 50
+    },
+    {
+      "epoch": 0.03139717425431711,
+      "grad_norm": 9.19274633645371,
+      "learning_rate": 7.832898172323761e-06,
+      "logits/chosen": -12.4375,
+      "logits/rejected": -12.125,
+      "logps/chosen": -334.0,
+      "logps/rejected": -328.0,
+      "loss": 0.6243,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -0.3203125,
+      "rewards/margins": 0.1455078125,
+      "rewards/rejected": -0.466796875,
+      "step": 60
+    },
+    {
+      "epoch": 0.03663003663003663,
+      "grad_norm": 10.83447468535562,
+      "learning_rate": 9.138381201044387e-06,
+      "logits/chosen": -13.625,
+      "logits/rejected": -13.5625,
+      "logps/chosen": -380.0,
+      "logps/rejected": -340.0,
+      "loss": 0.609,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.6796875,
+      "rewards/margins": 0.287109375,
+      "rewards/rejected": -0.96875,
+      "step": 70
+    },
+    {
+      "epoch": 0.04186289900575615,
+      "grad_norm": 9.13507791543036,
+      "learning_rate": 1.0443864229765014e-05,
+      "logits/chosen": -13.25,
+      "logits/rejected": -13.25,
+      "logps/chosen": -324.0,
+      "logps/rejected": -320.0,
+      "loss": 0.6258,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": -0.5703125,
+      "rewards/margins": 0.283203125,
+      "rewards/rejected": -0.8515625,
+      "step": 80
+    },
+    {
+      "epoch": 0.04709576138147567,
+      "grad_norm": 8.742369309742912,
+      "learning_rate": 1.174934725848564e-05,
+      "logits/chosen": -11.875,
+      "logits/rejected": -11.5,
+      "logps/chosen": -296.0,
+      "logps/rejected": -272.0,
+      "loss": 0.6439,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -0.28515625,
+      "rewards/margins": 0.208984375,
+      "rewards/rejected": -0.494140625,
+      "step": 90
+    },
+    {
+      "epoch": 0.052328623757195186,
+      "grad_norm": 11.89346927211031,
+      "learning_rate": 1.3054830287206268e-05,
+      "logits/chosen": -11.875,
+      "logits/rejected": -10.875,
+      "logps/chosen": -356.0,
+      "logps/rejected": -286.0,
+      "loss": 0.6267,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -0.380859375,
+      "rewards/margins": 0.27734375,
+      "rewards/rejected": -0.65625,
+      "step": 100
+    },
+    {
+      "epoch": 0.0575614861329147,
+      "grad_norm": 7.722831940344756,
+      "learning_rate": 1.4360313315926893e-05,
+      "logits/chosen": -11.125,
+      "logits/rejected": -10.625,
+      "logps/chosen": -320.0,
+      "logps/rejected": -292.0,
+      "loss": 0.6081,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.6328125,
+      "rewards/margins": 0.443359375,
+      "rewards/rejected": -1.078125,
+      "step": 110
+    },
+    {
+      "epoch": 0.06279434850863422,
+      "grad_norm": 10.557403681525182,
+      "learning_rate": 1.5665796344647522e-05,
+      "logits/chosen": -11.6875,
+      "logits/rejected": -11.375,
+      "logps/chosen": -422.0,
+      "logps/rejected": -376.0,
+      "loss": 0.6488,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -0.578125,
+      "rewards/margins": 0.28515625,
+      "rewards/rejected": -0.86328125,
+      "step": 120
+    },
+    {
+      "epoch": 0.06802721088435375,
+      "grad_norm": 8.52113738726324,
+      "learning_rate": 1.6971279373368146e-05,
+      "logits/chosen": -11.9375,
+      "logits/rejected": -11.125,
+      "logps/chosen": -282.0,
+      "logps/rejected": -288.0,
+      "loss": 0.6485,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.6796875,
+      "rewards/margins": 0.2890625,
+      "rewards/rejected": -0.96875,
+      "step": 130
+    },
+    {
+      "epoch": 0.07326007326007326,
+      "grad_norm": 10.818036746056942,
+      "learning_rate": 1.8276762402088773e-05,
+      "logits/chosen": -12.3125,
+      "logits/rejected": -12.125,
+      "logps/chosen": -362.0,
+      "logps/rejected": -340.0,
+      "loss": 0.5629,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -1.0703125,
+      "rewards/margins": 0.6015625,
+      "rewards/rejected": -1.671875,
+      "step": 140
+    },
+    {
+      "epoch": 0.07849293563579278,
+      "grad_norm": 9.701498673533864,
+      "learning_rate": 1.95822454308094e-05,
+      "logits/chosen": -13.125,
+      "logits/rejected": -13.0,
+      "logps/chosen": -358.0,
+      "logps/rejected": -326.0,
+      "loss": 0.6388,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -1.515625,
+      "rewards/margins": 0.48046875,
+      "rewards/rejected": -1.9921875,
+      "step": 150
+    },
+    {
+      "epoch": 0.0837257980115123,
+      "grad_norm": 9.972948367466696,
+      "learning_rate": 2.0887728459530027e-05,
+      "logits/chosen": -13.3125,
+      "logits/rejected": -13.5,
+      "logps/chosen": -366.0,
+      "logps/rejected": -324.0,
+      "loss": 0.6264,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -1.328125,
+      "rewards/margins": 0.68359375,
+      "rewards/rejected": -2.015625,
+      "step": 160
+    },
+    {
+      "epoch": 0.08895866038723181,
+      "grad_norm": 9.17511762621532,
+      "learning_rate": 2.2193211488250655e-05,
+      "logits/chosen": -13.375,
+      "logits/rejected": -13.25,
+      "logps/chosen": -338.0,
+      "logps/rejected": -346.0,
+      "loss": 0.5969,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": -1.3359375,
+      "rewards/margins": 0.1865234375,
+      "rewards/rejected": -1.5234375,
+      "step": 170
+    },
+    {
+      "epoch": 0.09419152276295134,
+      "grad_norm": 6.965621587975365,
+      "learning_rate": 2.349869451697128e-05,
+      "logits/chosen": -12.8125,
+      "logits/rejected": -12.75,
+      "logps/chosen": -302.0,
+      "logps/rejected": -314.0,
+      "loss": 0.6303,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -1.359375,
+      "rewards/margins": 0.359375,
+      "rewards/rejected": -1.71875,
+      "step": 180
+    },
+    {
+      "epoch": 0.09942438513867086,
+      "grad_norm": 9.553188606130272,
+      "learning_rate": 2.4804177545691905e-05,
+      "logits/chosen": -12.4375,
+      "logits/rejected": -12.5625,
+      "logps/chosen": -416.0,
+      "logps/rejected": -356.0,
+      "loss": 0.5885,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -1.7578125,
+      "rewards/margins": 0.62890625,
+      "rewards/rejected": -2.390625,
+      "step": 190
+    },
+    {
+      "epoch": 0.10465724751439037,
+      "grad_norm": 9.121837612837107,
+      "learning_rate": 2.6109660574412536e-05,
+      "logits/chosen": -12.8125,
+      "logits/rejected": -12.5,
+      "logps/chosen": -364.0,
+      "logps/rejected": -360.0,
+      "loss": 0.6058,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -1.859375,
+      "rewards/margins": 0.38671875,
+      "rewards/rejected": -2.25,
+      "step": 200
+    },
+    {
+      "epoch": 0.10989010989010989,
+      "grad_norm": 8.059177993813114,
+      "learning_rate": 2.741514360313316e-05,
+      "logits/chosen": -13.0,
+      "logits/rejected": -12.875,
+      "logps/chosen": -316.0,
+      "logps/rejected": -294.0,
+      "loss": 0.6417,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -1.796875,
+      "rewards/margins": 0.306640625,
+      "rewards/rejected": -2.109375,
+      "step": 210
+    },
+    {
+      "epoch": 0.1151229722658294,
+      "grad_norm": 8.500755304845391,
+      "learning_rate": 2.8720626631853787e-05,
+      "logits/chosen": -13.125,
+      "logits/rejected": -12.625,
+      "logps/chosen": -344.0,
+      "logps/rejected": -354.0,
+      "loss": 0.7723,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -2.4375,
+      "rewards/margins": 0.578125,
+      "rewards/rejected": -3.015625,
+      "step": 220
+    },
+    {
+      "epoch": 0.12035583464154893,
+      "grad_norm": 9.230475840269714,
+      "learning_rate": 3.0026109660574414e-05,
+      "logits/chosen": -12.375,
+      "logits/rejected": -12.125,
+      "logps/chosen": -386.0,
+      "logps/rejected": -422.0,
+      "loss": 0.6046,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -2.359375,
+      "rewards/margins": 0.63671875,
+      "rewards/rejected": -3.0,
+      "step": 230
+    },
+    {
+      "epoch": 0.12558869701726844,
+      "grad_norm": 9.204392189192681,
+      "learning_rate": 3.1331592689295045e-05,
+      "logits/chosen": -10.875,
+      "logits/rejected": -10.0,
+      "logps/chosen": -328.0,
+      "logps/rejected": -336.0,
+      "loss": 0.6512,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -1.9765625,
+      "rewards/margins": 0.65234375,
+      "rewards/rejected": -2.640625,
+      "step": 240
+    },
+    {
+      "epoch": 0.13082155939298795,
+      "grad_norm": 9.749349255282869,
+      "learning_rate": 3.263707571801567e-05,
+      "logits/chosen": -11.0625,
+      "logits/rejected": -10.4375,
+      "logps/chosen": -388.0,
+      "logps/rejected": -348.0,
+      "loss": 0.6508,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -2.0625,
+      "rewards/margins": 0.439453125,
+      "rewards/rejected": -2.5,
+      "step": 250
+    },
+    {
+      "epoch": 0.1360544217687075,
+      "grad_norm": 10.257472173487507,
+      "learning_rate": 3.394255874673629e-05,
+      "logits/chosen": -10.9375,
+      "logits/rejected": -10.4375,
+      "logps/chosen": -368.0,
+      "logps/rejected": -338.0,
+      "loss": 0.6354,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -2.0625,
+      "rewards/margins": 0.68359375,
+      "rewards/rejected": -2.75,
+      "step": 260
+    },
+    {
+      "epoch": 0.141287284144427,
+      "grad_norm": 6.958550084405995,
+      "learning_rate": 3.524804177545692e-05,
+      "logits/chosen": -9.875,
+      "logits/rejected": -9.5625,
+      "logps/chosen": -364.0,
+      "logps/rejected": -362.0,
+      "loss": 0.5584,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -1.8125,
+      "rewards/margins": 0.7109375,
+      "rewards/rejected": -2.515625,
+      "step": 270
+    },
+    {
+      "epoch": 0.14652014652014653,
+      "grad_norm": 9.847805644914816,
+      "learning_rate": 3.6553524804177546e-05,
+      "logits/chosen": -11.625,
+      "logits/rejected": -11.6875,
+      "logps/chosen": -410.0,
+      "logps/rejected": -382.0,
+      "loss": 0.6862,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -2.75,
+      "rewards/margins": 0.49609375,
+      "rewards/rejected": -3.25,
+      "step": 280
+    },
+    {
+      "epoch": 0.15175300889586604,
+      "grad_norm": 8.573440905919824,
+      "learning_rate": 3.7859007832898173e-05,
+      "logits/chosen": -12.875,
+      "logits/rejected": -12.4375,
+      "logps/chosen": -382.0,
+      "logps/rejected": -336.0,
+      "loss": 0.7394,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -1.9609375,
+      "rewards/margins": 0.62109375,
+      "rewards/rejected": -2.578125,
+      "step": 290
+    },
+    {
+      "epoch": 0.15698587127158556,
+      "grad_norm": 8.80725505375811,
+      "learning_rate": 3.91644908616188e-05,
+      "logits/chosen": -13.75,
+      "logits/rejected": -13.25,
+      "logps/chosen": -380.0,
+      "logps/rejected": -344.0,
+      "loss": 0.6646,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -2.3125,
+      "rewards/margins": 0.458984375,
+      "rewards/rejected": -2.765625,
+      "step": 300
+    },
+    {
+      "epoch": 0.16221873364730507,
+      "grad_norm": 8.968657211210374,
+      "learning_rate": 4.046997389033943e-05,
+      "logits/chosen": -12.25,
+      "logits/rejected": -11.9375,
+      "logps/chosen": -386.0,
+      "logps/rejected": -346.0,
+      "loss": 0.6784,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -2.4375,
+      "rewards/margins": 0.578125,
+      "rewards/rejected": -3.015625,
+      "step": 310
+    },
+    {
+      "epoch": 0.1674515960230246,
+      "grad_norm": 10.194843430881596,
+      "learning_rate": 4.1775456919060055e-05,
+      "logits/chosen": -13.625,
+      "logits/rejected": -13.0625,
+      "logps/chosen": -442.0,
+      "logps/rejected": -372.0,
+      "loss": 0.5713,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -3.5625,
+      "rewards/margins": 0.76171875,
+      "rewards/rejected": -4.3125,
+      "step": 320
+    },
+    {
+      "epoch": 0.1726844583987441,
+      "grad_norm": 10.313402119423792,
+      "learning_rate": 4.308093994778068e-05,
+      "logits/chosen": -12.6875,
+      "logits/rejected": -12.375,
+      "logps/chosen": -398.0,
+      "logps/rejected": -364.0,
+      "loss": 0.735,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -3.984375,
+      "rewards/margins": 0.9375,
+      "rewards/rejected": -4.90625,
+      "step": 330
+    },
+    {
+      "epoch": 0.17791732077446362,
+      "grad_norm": 9.173592771992633,
+      "learning_rate": 4.438642297650131e-05,
+      "logits/chosen": -11.1875,
+      "logits/rejected": -10.8125,
+      "logps/chosen": -406.0,
+      "logps/rejected": -382.0,
+      "loss": 0.6655,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -4.125,
+      "rewards/margins": 0.8515625,
+      "rewards/rejected": -4.96875,
+      "step": 340
+    },
+    {
+      "epoch": 0.18315018315018314,
+      "grad_norm": 8.34069764210929,
+      "learning_rate": 4.5691906005221936e-05,
+      "logits/chosen": -11.375,
+      "logits/rejected": -11.25,
+      "logps/chosen": -332.0,
+      "logps/rejected": -360.0,
+      "loss": 0.6806,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -4.59375,
+      "rewards/margins": 0.87109375,
+      "rewards/rejected": -5.46875,
+      "step": 350
+    },
+    {
+      "epoch": 0.18838304552590268,
+      "grad_norm": 11.451242673237575,
+      "learning_rate": 4.699738903394256e-05,
+      "logits/chosen": -12.8125,
+      "logits/rejected": -12.5,
+      "logps/chosen": -424.0,
+      "logps/rejected": -404.0,
+      "loss": 0.767,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -4.4375,
+      "rewards/margins": 0.578125,
+      "rewards/rejected": -5.03125,
+      "step": 360
+    },
+    {
+      "epoch": 0.1936159079016222,
+      "grad_norm": 8.706088083363465,
+      "learning_rate": 4.830287206266319e-05,
+      "logits/chosen": -12.875,
+      "logits/rejected": -12.6875,
+      "logps/chosen": -380.0,
+      "logps/rejected": -362.0,
+      "loss": 0.7948,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -4.4375,
+      "rewards/margins": 0.55078125,
+      "rewards/rejected": -5.0,
+      "step": 370
+    },
+    {
+      "epoch": 0.1988487702773417,
+      "grad_norm": 8.71050562327642,
+      "learning_rate": 4.960835509138381e-05,
+      "logits/chosen": -12.875,
+      "logits/rejected": -12.8125,
+      "logps/chosen": -360.0,
+      "logps/rejected": -402.0,
+      "loss": 0.7885,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -4.40625,
+      "rewards/margins": 0.79296875,
+      "rewards/rejected": -5.21875,
+      "step": 380
+    },
+    {
+      "epoch": 0.20408163265306123,
+      "grad_norm": 9.631202895027702,
+      "learning_rate": 4.9999488859837295e-05,
+      "logits/chosen": -13.375,
+      "logits/rejected": -13.1875,
+      "logps/chosen": -448.0,
+      "logps/rejected": -396.0,
+      "loss": 0.8545,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -4.25,
+      "rewards/margins": 0.93359375,
+      "rewards/rejected": -5.1875,
+      "step": 390
+    },
+    {
+      "epoch": 0.20931449502878074,
+      "grad_norm": 12.175473256479712,
+      "learning_rate": 4.999698536649904e-05,
+      "logits/chosen": -14.5,
+      "logits/rejected": -14.5,
+      "logps/chosen": -462.0,
+      "logps/rejected": -392.0,
+      "loss": 0.7412,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -4.8125,
+      "rewards/margins": 0.65234375,
+      "rewards/rejected": -5.4375,
+      "step": 400
+    },
+    {
+      "epoch": 0.21454735740450026,
+      "grad_norm": 10.472391673320981,
+      "learning_rate": 4.999239584575648e-05,
+      "logits/chosen": -14.75,
+      "logits/rejected": -14.875,
+      "logps/chosen": -446.0,
+      "logps/rejected": -438.0,
+      "loss": 0.7897,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -4.84375,
+      "rewards/margins": 0.298828125,
+      "rewards/rejected": -5.15625,
+      "step": 410
+    },
+    {
+      "epoch": 0.21978021978021978,
+      "grad_norm": 10.549150549694106,
+      "learning_rate": 4.9985720680610434e-05,
+      "logits/chosen": -14.6875,
+      "logits/rejected": -14.6875,
+      "logps/chosen": -412.0,
+      "logps/rejected": -390.0,
+      "loss": 0.7813,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": -4.625,
+      "rewards/margins": 0.2734375,
+      "rewards/rejected": -4.90625,
+      "step": 420
+    },
+    {
+      "epoch": 0.2250130821559393,
+      "grad_norm": 8.610558700109573,
+      "learning_rate": 4.997696042811118e-05,
+      "logits/chosen": -14.625,
+      "logits/rejected": -14.8125,
+      "logps/chosen": -416.0,
+      "logps/rejected": -344.0,
+      "loss": 0.7214,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": -4.375,
+      "rewards/margins": 0.59765625,
+      "rewards/rejected": -4.96875,
+      "step": 430
+    },
+    {
+      "epoch": 0.2302459445316588,
+      "grad_norm": 9.047522311209711,
+      "learning_rate": 4.996611581931193e-05,
+      "logits/chosen": -14.125,
+      "logits/rejected": -14.25,
+      "logps/chosen": -440.0,
+      "logps/rejected": -374.0,
+      "loss": 0.7039,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -4.5625,
+      "rewards/margins": 0.498046875,
+      "rewards/rejected": -5.0625,
+      "step": 440
+    },
+    {
+      "epoch": 0.23547880690737832,
+      "grad_norm": 10.207379585431303,
+      "learning_rate": 4.995318775920787e-05,
+      "logits/chosen": -13.5,
+      "logits/rejected": -13.5,
+      "logps/chosen": -384.0,
+      "logps/rejected": -386.0,
+      "loss": 0.7792,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -4.71875,
+      "rewards/margins": 0.130859375,
+      "rewards/rejected": -4.84375,
+      "step": 450
+    },
+    {
+      "epoch": 0.24071166928309787,
+      "grad_norm": 9.821006860178837,
+      "learning_rate": 4.9938177326660587e-05,
+      "logits/chosen": -13.375,
+      "logits/rejected": -13.375,
+      "logps/chosen": -478.0,
+      "logps/rejected": -436.0,
+      "loss": 0.6816,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -4.625,
+      "rewards/margins": 1.046875,
+      "rewards/rejected": -5.65625,
+      "step": 460
+    },
+    {
+      "epoch": 0.24594453165881738,
+      "grad_norm": 10.675355174224261,
+      "learning_rate": 4.99210857743081e-05,
+      "logits/chosen": -12.875,
+      "logits/rejected": -12.875,
+      "logps/chosen": -448.0,
+      "logps/rejected": -436.0,
+      "loss": 0.6973,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -5.875,
+      "rewards/margins": 0.72265625,
+      "rewards/rejected": -6.59375,
+      "step": 470
+    },
+    {
+      "epoch": 0.25117739403453687,
+      "grad_norm": 10.26484583598167,
+      "learning_rate": 4.990191452846024e-05,
+      "logits/chosen": -13.0,
+      "logits/rejected": -13.0,
+      "logps/chosen": -406.0,
+      "logps/rejected": -396.0,
+      "loss": 0.6783,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -5.34375,
+      "rewards/margins": 0.66796875,
+      "rewards/rejected": -6.0,
+      "step": 480
+    },
+    {
+      "epoch": 0.2564102564102564,
+      "grad_norm": 9.035937614807631,
+      "learning_rate": 4.988066518897971e-05,
+      "logits/chosen": -13.625,
+      "logits/rejected": -13.3125,
+      "logps/chosen": -464.0,
+      "logps/rejected": -436.0,
+      "loss": 0.6354,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -6.09375,
+      "rewards/margins": 0.921875,
+      "rewards/rejected": -7.0,
+      "step": 490
+    },
+    {
+      "epoch": 0.2616431187859759,
+      "grad_norm": 6.87488260728394,
+      "learning_rate": 4.985733952914852e-05,
+      "logits/chosen": -15.5625,
+      "logits/rejected": -15.375,
+      "logps/chosen": -452.0,
+      "logps/rejected": -438.0,
+      "loss": 0.6495,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -7.0,
+      "rewards/margins": 0.69921875,
+      "rewards/rejected": -7.71875,
+      "step": 500
+    },
+    {
+      "epoch": 0.2668759811616955,
+      "grad_norm": 9.363373210153325,
+      "learning_rate": 4.983193949552002e-05,
+      "logits/chosen": -15.625,
+      "logits/rejected": -15.9375,
+      "logps/chosen": -450.0,
+      "logps/rejected": -420.0,
+      "loss": 0.7343,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -6.5,
+      "rewards/margins": 0.458984375,
+      "rewards/rejected": -6.9375,
+      "step": 510
+    },
+    {
+      "epoch": 0.272108843537415,
+      "grad_norm": 7.7281840818721,
+      "learning_rate": 4.980446720775646e-05,
+      "logits/chosen": -14.125,
+      "logits/rejected": -14.25,
+      "logps/chosen": -464.0,
+      "logps/rejected": -474.0,
+      "loss": 0.742,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -5.375,
+      "rewards/margins": 0.58984375,
+      "rewards/rejected": -5.9375,
+      "step": 520
+    },
+    {
+      "epoch": 0.2773417059131345,
+      "grad_norm": 9.271525518013695,
+      "learning_rate": 4.9774924958452084e-05,
+      "logits/chosen": -14.5625,
+      "logits/rejected": -14.75,
+      "logps/chosen": -492.0,
+      "logps/rejected": -408.0,
+      "loss": 0.8633,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": -6.3125,
+      "rewards/margins": 0.265625,
+      "rewards/rejected": -6.59375,
+      "step": 530
+    },
+    {
+      "epoch": 0.282574568288854,
+      "grad_norm": 11.416271168800474,
+      "learning_rate": 4.974331521294186e-05,
+      "logits/chosen": -14.75,
+      "logits/rejected": -14.75,
+      "logps/chosen": -496.0,
+      "logps/rejected": -460.0,
+      "loss": 0.7933,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -7.3125,
+      "rewards/margins": 0.5625,
+      "rewards/rejected": -7.90625,
+      "step": 540
+    },
+    {
+      "epoch": 0.28780743066457354,
+      "grad_norm": 7.787620742028088,
+      "learning_rate": 4.97096406090957e-05,
+      "logits/chosen": -14.3125,
+      "logits/rejected": -14.125,
+      "logps/chosen": -492.0,
+      "logps/rejected": -468.0,
+      "loss": 0.7545,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -6.125,
+      "rewards/margins": 0.69921875,
+      "rewards/rejected": -6.8125,
+      "step": 550
+    },
+    {
+      "epoch": 0.29304029304029305,
+      "grad_norm": 8.382409522213168,
+      "learning_rate": 4.96739039570983e-05,
+      "logits/chosen": -14.25,
+      "logits/rejected": -14.125,
+      "logps/chosen": -444.0,
+      "logps/rejected": -436.0,
+      "loss": 0.7372,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -5.6875,
+      "rewards/margins": 0.609375,
+      "rewards/rejected": -6.3125,
+      "step": 560
+    },
+    {
+      "epoch": 0.29827315541601257,
+      "grad_norm": 9.779033425632107,
+      "learning_rate": 4.963610823921471e-05,
+      "logits/chosen": -14.625,
+      "logits/rejected": -14.4375,
+      "logps/chosen": -482.0,
+      "logps/rejected": -434.0,
+      "loss": 0.9479,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -7.125,
+      "rewards/margins": -0.15625,
+      "rewards/rejected": -6.96875,
+      "step": 570
+    },
+    {
+      "epoch": 0.3035060177917321,
+      "grad_norm": 9.733444132976096,
+      "learning_rate": 4.959625660954139e-05,
+      "logits/chosen": -14.5,
+      "logits/rejected": -14.625,
+      "logps/chosen": -444.0,
+      "logps/rejected": -388.0,
+      "loss": 0.744,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -6.78125,
+      "rewards/margins": 0.65234375,
+      "rewards/rejected": -7.40625,
+      "step": 580
+    },
+    {
+      "epoch": 0.3087388801674516,
+      "grad_norm": 8.051810745094972,
+      "learning_rate": 4.9554352393743045e-05,
+      "logits/chosen": -13.125,
+      "logits/rejected": -13.0625,
+      "logps/chosen": -470.0,
+      "logps/rejected": -452.0,
+      "loss": 0.9418,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -8.3125,
+      "rewards/margins": 0.64453125,
+      "rewards/rejected": -9.0,
+      "step": 590
+    },
+    {
+      "epoch": 0.3139717425431711,
+      "grad_norm": 5.903225463973512,
+      "learning_rate": 4.9510399088775047e-05,
+      "logits/chosen": -12.75,
+      "logits/rejected": -12.75,
+      "logps/chosen": -500.0,
+      "logps/rejected": -502.0,
+      "loss": 0.7086,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -8.125,
+      "rewards/margins": 0.50390625,
+      "rewards/rejected": -8.5625,
+      "step": 600
+    },
+    {
+      "epoch": 0.31920460491889063,
+      "grad_norm": 11.634410981967923,
+      "learning_rate": 4.9464400362591644e-05,
+      "logits/chosen": -12.1875,
+      "logits/rejected": -12.125,
+      "logps/chosen": -420.0,
+      "logps/rejected": -396.0,
+      "loss": 0.7854,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": -6.5625,
+      "rewards/margins": 0.50390625,
+      "rewards/rejected": -7.09375,
+      "step": 610
+    },
+    {
+      "epoch": 0.32443746729461015,
+      "grad_norm": 8.791552603599007,
+      "learning_rate": 4.941636005383986e-05,
+      "logits/chosen": -12.375,
+      "logits/rejected": -12.1875,
+      "logps/chosen": -528.0,
+      "logps/rejected": -402.0,
+      "loss": 0.8591,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -6.59375,
+      "rewards/margins": 0.86328125,
+      "rewards/rejected": -7.4375,
+      "step": 620
+    },
+    {
+      "epoch": 0.32967032967032966,
+      "grad_norm": 9.02432053656773,
+      "learning_rate": 4.936628217153914e-05,
+      "logits/chosen": -12.3125,
+      "logits/rejected": -12.3125,
+      "logps/chosen": -386.0,
+      "logps/rejected": -408.0,
+      "loss": 0.7811,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": -6.40625,
+      "rewards/margins": 0.4296875,
+      "rewards/rejected": -6.8125,
+      "step": 630
+    },
+    {
+      "epoch": 0.3349031920460492,
+      "grad_norm": 9.394934456092882,
+      "learning_rate": 4.931417089474682e-05,
+      "logits/chosen": -13.625,
+      "logits/rejected": -13.6875,
+      "logps/chosen": -462.0,
+      "logps/rejected": -438.0,
+      "loss": 0.6975,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -6.5625,
+      "rewards/margins": 0.796875,
+      "rewards/rejected": -7.375,
+      "step": 640
+    },
+    {
+      "epoch": 0.3401360544217687,
+      "grad_norm": 7.439435296736013,
+      "learning_rate": 4.926003057220935e-05,
+      "logits/chosen": -15.3125,
+      "logits/rejected": -15.4375,
+      "logps/chosen": -470.0,
+      "logps/rejected": -434.0,
+      "loss": 0.7351,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -7.0,
+      "rewards/margins": 0.609375,
+      "rewards/rejected": -7.625,
+      "step": 650
+    },
+    {
+      "epoch": 0.3453689167974882,
+      "grad_norm": 9.162980300519617,
+      "learning_rate": 4.92038657219994e-05,
+      "logits/chosen": -16.375,
+      "logits/rejected": -16.5,
+      "logps/chosen": -402.0,
+      "logps/rejected": -392.0,
+      "loss": 0.743,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -6.53125,
+      "rewards/margins": 0.671875,
+      "rewards/rejected": -7.1875,
+      "step": 660
+    },
+    {
+      "epoch": 0.35060177917320773,
+      "grad_norm": 11.704890362462061,
+      "learning_rate": 4.914568103113882e-05,
+      "logits/chosen": -15.6875,
+      "logits/rejected": -16.0,
+      "logps/chosen": -442.0,
+      "logps/rejected": -418.0,
+      "loss": 0.7136,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -6.59375,
+      "rewards/margins": 0.671875,
+      "rewards/rejected": -7.28125,
+      "step": 670
+    },
+    {
+      "epoch": 0.35583464154892724,
+      "grad_norm": 7.227855368315395,
+      "learning_rate": 4.908548135520752e-05,
+      "logits/chosen": -14.8125,
+      "logits/rejected": -14.9375,
+      "logps/chosen": -456.0,
+      "logps/rejected": -416.0,
+      "loss": 0.6655,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -6.9375,
+      "rewards/margins": 0.5390625,
+      "rewards/rejected": -7.5,
+      "step": 680
+    },
+    {
+      "epoch": 0.36106750392464676,
+      "grad_norm": 9.146985437020582,
+      "learning_rate": 4.9023271717938224e-05,
+      "logits/chosen": -14.125,
+      "logits/rejected": -14.0625,
+      "logps/chosen": -528.0,
+      "logps/rejected": -486.0,
+      "loss": 0.7974,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -9.0625,
+      "rewards/margins": 0.53515625,
+      "rewards/rejected": -9.5625,
+      "step": 690
+    },
+    {
+      "epoch": 0.3663003663003663,
+      "grad_norm": 9.278505608910281,
+      "learning_rate": 4.8959057310797286e-05,
+      "logits/chosen": -14.0625,
+      "logits/rejected": -14.3125,
+      "logps/chosen": -486.0,
+      "logps/rejected": -428.0,
+      "loss": 0.7751,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": -6.90625,
+      "rewards/margins": 0.427734375,
+      "rewards/rejected": -7.3125,
+      "step": 700
+    },
+    {
+      "epoch": 0.3715332286760858,
+      "grad_norm": 9.712228745007655,
+      "learning_rate": 4.889284349255141e-05,
+      "logits/chosen": -14.6875,
+      "logits/rejected": -14.6875,
+      "logps/chosen": -506.0,
+      "logps/rejected": -474.0,
+      "loss": 0.6715,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -7.46875,
+      "rewards/margins": 1.1328125,
+      "rewards/rejected": -8.5625,
+      "step": 710
+    },
+    {
+      "epoch": 0.37676609105180536,
+      "grad_norm": 8.521168690100787,
+      "learning_rate": 4.8824635788820475e-05,
+      "logits/chosen": -14.375,
+      "logits/rejected": -14.25,
+      "logps/chosen": -448.0,
+      "logps/rejected": -446.0,
+      "loss": 0.8019,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -7.875,
+      "rewards/margins": 0.453125,
+      "rewards/rejected": -8.3125,
+      "step": 720
+    },
+    {
+      "epoch": 0.3819989534275249,
+      "grad_norm": 11.10196898625787,
+      "learning_rate": 4.8754439891616434e-05,
+      "logits/chosen": -14.1875,
+      "logits/rejected": -14.1875,
+      "logps/chosen": -472.0,
+      "logps/rejected": -458.0,
+      "loss": 0.8141,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": -7.34375,
+      "rewards/margins": 0.60546875,
+      "rewards/rejected": -7.9375,
+      "step": 730
+    },
+    {
+      "epoch": 0.3872318158032444,
+      "grad_norm": 10.792601975369012,
+      "learning_rate": 4.8682261658868264e-05,
+      "logits/chosen": -14.25,
+      "logits/rejected": -14.375,
+      "logps/chosen": -450.0,
+      "logps/rejected": -414.0,
+      "loss": 0.6468,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -7.34375,
+      "rewards/margins": 0.8828125,
+      "rewards/rejected": -8.25,
+      "step": 740
+    },
+    {
+      "epoch": 0.3924646781789639,
+      "grad_norm": 55.736192989145714,
+      "learning_rate": 4.860810711393317e-05,
+      "logits/chosen": -14.75,
+      "logits/rejected": -14.5625,
+      "logps/chosen": -480.0,
+      "logps/rejected": -502.0,
+      "loss": 1.0953,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -8.5625,
+      "rewards/margins": 0.93359375,
+      "rewards/rejected": -9.5,
+      "step": 750
+    },
+    {
+      "epoch": 0.3976975405546834,
+      "grad_norm": 8.990314887586853,
+      "learning_rate": 4.853198244509386e-05,
+      "logits/chosen": -14.625,
+      "logits/rejected": -14.625,
+      "logps/chosen": -450.0,
+      "logps/rejected": -434.0,
+      "loss": 0.7577,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -7.78125,
+      "rewards/margins": 0.6640625,
+      "rewards/rejected": -8.4375,
+      "step": 760
+    },
+    {
+      "epoch": 0.40293040293040294,
+      "grad_norm": 7.965840471202787,
+      "learning_rate": 4.845389400504221e-05,
+      "logits/chosen": -14.375,
+      "logits/rejected": -14.25,
+      "logps/chosen": -488.0,
+      "logps/rejected": -490.0,
+      "loss": 0.7896,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": -7.5625,
+      "rewards/margins": 0.484375,
+      "rewards/rejected": -8.0625,
+      "step": 770
+    },
+    {
+      "epoch": 0.40816326530612246,
+      "grad_norm": 11.497968950154975,
+      "learning_rate": 4.837384831034905e-05,
+      "logits/chosen": -13.875,
+      "logits/rejected": -13.75,
+      "logps/chosen": -466.0,
+      "logps/rejected": -438.0,
+      "loss": 0.6823,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -7.15625,
+      "rewards/margins": 0.6875,
+      "rewards/rejected": -7.84375,
+      "step": 780
+    },
+    {
+      "epoch": 0.413396127681842,
+      "grad_norm": 9.66137517497451,
+      "learning_rate": 4.829185204092039e-05,
+      "logits/chosen": -14.5625,
+      "logits/rejected": -14.25,
+      "logps/chosen": -416.0,
+      "logps/rejected": -442.0,
+      "loss": 0.7498,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -7.0,
+      "rewards/margins": 1.15625,
+      "rewards/rejected": -8.125,
+      "step": 790
+    },
+    {
+      "epoch": 0.4186289900575615,
+      "grad_norm": 10.37574761233887,
+      "learning_rate": 4.8207912039439964e-05,
+      "logits/chosen": -15.4375,
+      "logits/rejected": -15.625,
+      "logps/chosen": -504.0,
+      "logps/rejected": -466.0,
+      "loss": 0.8479,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -6.875,
+      "rewards/margins": 0.29296875,
+      "rewards/rejected": -7.1875,
+      "step": 800
+    },
+    {
+      "epoch": 0.423861852433281,
+      "grad_norm": 10.351543599017008,
+      "learning_rate": 4.812203531079819e-05,
+      "logits/chosen": -15.25,
+      "logits/rejected": -15.375,
+      "logps/chosen": -476.0,
+      "logps/rejected": -454.0,
+      "loss": 0.7395,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -7.09375,
+      "rewards/margins": 0.75,
+      "rewards/rejected": -7.84375,
+      "step": 810
+    },
+    {
+      "epoch": 0.4290947148090005,
+      "grad_norm": 8.622095411540833,
+      "learning_rate": 4.803422902150762e-05,
+      "logits/chosen": -14.375,
+      "logits/rejected": -14.4375,
+      "logps/chosen": -494.0,
+      "logps/rejected": -456.0,
+      "loss": 0.7025,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -8.6875,
+      "rewards/margins": 0.6875,
+      "rewards/rejected": -9.375,
+      "step": 820
+    },
+    {
+      "epoch": 0.43432757718472004,
+      "grad_norm": 9.264123822708422,
+      "learning_rate": 4.794450049910487e-05,
+      "logits/chosen": -12.9375,
+      "logits/rejected": -12.9375,
+      "logps/chosen": -454.0,
+      "logps/rejected": -438.0,
+      "loss": 0.8018,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -8.125,
+      "rewards/margins": 0.369140625,
+      "rewards/rejected": -8.5,
+      "step": 830
+    },
+    {
+      "epoch": 0.43956043956043955,
+      "grad_norm": 7.9655618610185845,
+      "learning_rate": 4.785285723153915e-05,
+      "logits/chosen": -11.625,
+      "logits/rejected": -12.0,
+      "logps/chosen": -536.0,
+      "logps/rejected": -472.0,
+      "loss": 0.7406,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -7.53125,
+      "rewards/margins": 0.6953125,
+      "rewards/rejected": -8.25,
+      "step": 840
+    },
+    {
+      "epoch": 0.44479330193615907,
+      "grad_norm": 10.645509151597812,
+      "learning_rate": 4.775930686654738e-05,
+      "logits/chosen": -12.0625,
+      "logits/rejected": -12.125,
+      "logps/chosen": -470.0,
+      "logps/rejected": -440.0,
+      "loss": 0.733,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": -8.125,
+      "rewards/margins": 1.0078125,
+      "rewards/rejected": -9.125,
+      "step": 850
+    },
+    {
+      "epoch": 0.4500261643118786,
+      "grad_norm": 9.095410131374702,
+      "learning_rate": 4.7663857211015936e-05,
+      "logits/chosen": -13.0,
+      "logits/rejected": -12.625,
+      "logps/chosen": -434.0,
+      "logps/rejected": -468.0,
+      "loss": 0.7619,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -6.53125,
+      "rewards/margins": 0.5546875,
+      "rewards/rejected": -7.0625,
+      "step": 860
+    },
+    {
+      "epoch": 0.4552590266875981,
+      "grad_norm": 7.213273555139093,
+      "learning_rate": 4.756651623032922e-05,
+      "logits/chosen": -12.625,
+      "logits/rejected": -12.75,
+      "logps/chosen": -458.0,
+      "logps/rejected": -412.0,
+      "loss": 0.7308,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -6.40625,
+      "rewards/margins": 0.3359375,
+      "rewards/rejected": -6.75,
+      "step": 870
+    },
+    {
+      "epoch": 0.4604918890633176,
+      "grad_norm": 10.25034124783594,
+      "learning_rate": 4.746729204770491e-05,
+      "logits/chosen": -12.25,
+      "logits/rejected": -12.125,
+      "logps/chosen": -532.0,
+      "logps/rejected": -470.0,
+      "loss": 0.688,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -8.6875,
+      "rewards/margins": 0.9140625,
+      "rewards/rejected": -9.625,
+      "step": 880
+    },
+    {
+      "epoch": 0.46572475143903713,
+      "grad_norm": 9.164696578171599,
+      "learning_rate": 4.736619294351607e-05,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -11.25,
+      "logps/chosen": -556.0,
+      "logps/rejected": -516.0,
+      "loss": 0.7735,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -9.8125,
+      "rewards/margins": 0.68359375,
+      "rewards/rejected": -10.5,
+      "step": 890
+    },
+    {
+      "epoch": 0.47095761381475665,
+      "grad_norm": 9.176482738323408,
+      "learning_rate": 4.726322735460012e-05,
+      "logits/chosen": -11.75,
+      "logits/rejected": -11.5,
+      "logps/chosen": -476.0,
+      "logps/rejected": -510.0,
+      "loss": 0.761,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -7.875,
+      "rewards/margins": 1.3984375,
+      "rewards/rejected": -9.25,
+      "step": 900
+    },
+    {
+      "epoch": 0.47619047619047616,
+      "grad_norm": 12.033840542792827,
+      "learning_rate": 4.715840387355481e-05,
+      "logits/chosen": -12.0625,
+      "logits/rejected": -11.875,
+      "logps/chosen": -452.0,
+      "logps/rejected": -440.0,
+      "loss": 0.84,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -8.75,
+      "rewards/margins": 0.83203125,
+      "rewards/rejected": -9.5625,
+      "step": 910
+    },
+    {
+      "epoch": 0.48142333856619574,
+      "grad_norm": 9.951231450685535,
+      "learning_rate": 4.705173124802114e-05,
+      "logits/chosen": -12.125,
+      "logits/rejected": -12.0625,
+      "logps/chosen": -528.0,
+      "logps/rejected": -500.0,
+      "loss": 0.6771,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -9.4375,
+      "rewards/margins": 0.451171875,
+      "rewards/rejected": -9.875,
+      "step": 920
+    },
+    {
+      "epoch": 0.48665620094191525,
+      "grad_norm": 5.982719210494255,
+      "learning_rate": 4.694321837995337e-05,
+      "logits/chosen": -12.0625,
+      "logits/rejected": -12.125,
+      "logps/chosen": -516.0,
+      "logps/rejected": -482.0,
+      "loss": 0.6545,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -9.0625,
+      "rewards/margins": 0.9296875,
+      "rewards/rejected": -10.0,
+      "step": 930
+    },
+    {
+      "epoch": 0.49188906331763477,
+      "grad_norm": 7.732935041516918,
+      "learning_rate": 4.683287432487612e-05,
+      "logits/chosen": -12.4375,
+      "logits/rejected": -12.3125,
+      "logps/chosen": -520.0,
+      "logps/rejected": -482.0,
+      "loss": 0.6515,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -9.3125,
+      "rewards/margins": 1.125,
+      "rewards/rejected": -10.4375,
+      "step": 940
+    },
+    {
+      "epoch": 0.4971219256933543,
+      "grad_norm": 11.192514507830417,
+      "learning_rate": 4.672070829112868e-05,
+      "logits/chosen": -12.6875,
+      "logits/rejected": -12.8125,
+      "logps/chosen": -498.0,
+      "logps/rejected": -488.0,
+      "loss": 0.6869,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -8.8125,
+      "rewards/margins": 1.0390625,
+      "rewards/rejected": -9.875,
+      "step": 950
+    },
+    {
+      "epoch": 0.5023547880690737,
+      "grad_norm": 9.893002359130774,
+      "learning_rate": 4.6606729639096606e-05,
+      "logits/chosen": -12.25,
+      "logits/rejected": -12.1875,
+      "logps/chosen": -520.0,
+      "logps/rejected": -520.0,
+      "loss": 0.6144,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -10.4375,
+      "rewards/margins": 1.484375,
+      "rewards/rejected": -11.9375,
+      "step": 960
+    },
+    {
+      "epoch": 0.5075876504447933,
+      "grad_norm": 11.075478903393348,
+      "learning_rate": 4.6490947880430515e-05,
+      "logits/chosen": -11.5,
+      "logits/rejected": -11.0,
+      "logps/chosen": -584.0,
+      "logps/rejected": -520.0,
+      "loss": 0.7253,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -10.75,
+      "rewards/margins": 1.4453125,
+      "rewards/rejected": -12.1875,
+      "step": 970
+    },
+    {
+      "epoch": 0.5128205128205128,
+      "grad_norm": 7.207931474384908,
+      "learning_rate": 4.637337267725239e-05,
+      "logits/chosen": -12.0625,
+      "logits/rejected": -11.8125,
+      "logps/chosen": -584.0,
+      "logps/rejected": -536.0,
+      "loss": 0.6751,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -10.25,
+      "rewards/margins": 1.5078125,
+      "rewards/rejected": -11.75,
+      "step": 980
+    },
+    {
+      "epoch": 0.5180533751962323,
+      "grad_norm": 10.223265502840688,
+      "learning_rate": 4.625401384134921e-05,
+      "logits/chosen": -12.75,
+      "logits/rejected": -12.6875,
+      "logps/chosen": -504.0,
+      "logps/rejected": -466.0,
+      "loss": 0.6907,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -9.5625,
+      "rewards/margins": 1.0,
+      "rewards/rejected": -10.5625,
+      "step": 990
+    },
+    {
+      "epoch": 0.5232862375719518,
+      "grad_norm": 9.366756020201832,
+      "learning_rate": 4.613288133335418e-05,
+      "logits/chosen": -12.1875,
+      "logits/rejected": -12.0625,
+      "logps/chosen": -490.0,
+      "logps/rejected": -472.0,
+      "loss": 0.7333,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -9.625,
+      "rewards/margins": 1.0,
+      "rewards/rejected": -10.625,
+      "step": 1000
+    },
+    {
+      "epoch": 0.5285190999476713,
+      "grad_norm": 9.668457676351762,
+      "learning_rate": 4.600998526191553e-05,
+      "logits/chosen": -12.6875,
+      "logits/rejected": -12.5625,
+      "logps/chosen": -540.0,
+      "logps/rejected": -544.0,
+      "loss": 0.8312,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -9.9375,
+      "rewards/margins": 0.73828125,
+      "rewards/rejected": -10.6875,
+      "step": 1010
+    },
+    {
+      "epoch": 0.533751962323391,
+      "grad_norm": 9.169851106004764,
+      "learning_rate": 4.588533588285287e-05,
+      "logits/chosen": -12.1875,
+      "logits/rejected": -12.1875,
+      "logps/chosen": -540.0,
+      "logps/rejected": -502.0,
+      "loss": 0.6978,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -10.375,
+      "rewards/margins": 1.3046875,
+      "rewards/rejected": -11.6875,
+      "step": 1020
+    },
+    {
+      "epoch": 0.5389848246991105,
+      "grad_norm": 6.76805241291535,
+      "learning_rate": 4.5758943598301354e-05,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -11.1875,
+      "logps/chosen": -556.0,
+      "logps/rejected": -492.0,
+      "loss": 0.7095,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -10.875,
+      "rewards/margins": 0.66796875,
+      "rewards/rejected": -11.5,
+      "step": 1030
+    },
+    {
+      "epoch": 0.54421768707483,
+      "grad_norm": 7.48260799269685,
+      "learning_rate": 4.5630818955843646e-05,
+      "logits/chosen": -12.0,
+      "logits/rejected": -11.8125,
+      "logps/chosen": -506.0,
+      "logps/rejected": -536.0,
+      "loss": 0.7073,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -9.9375,
+      "rewards/margins": 1.03125,
+      "rewards/rejected": -10.9375,
+      "step": 1040
+    },
+    {
+      "epoch": 0.5494505494505495,
+      "grad_norm": 9.671760718133113,
+      "learning_rate": 4.550097264762968e-05,
+      "logits/chosen": -12.625,
+      "logits/rejected": -12.625,
+      "logps/chosen": -492.0,
+      "logps/rejected": -492.0,
+      "loss": 0.8316,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": -8.875,
+      "rewards/margins": 0.71875,
+      "rewards/rejected": -9.5625,
+      "step": 1050
+    },
+    {
+      "epoch": 0.554683411826269,
+      "grad_norm": 7.232272637524744,
+      "learning_rate": 4.536941550948439e-05,
+      "logits/chosen": -13.3125,
+      "logits/rejected": -13.25,
+      "logps/chosen": -512.0,
+      "logps/rejected": -482.0,
+      "loss": 0.7443,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -9.25,
+      "rewards/margins": 0.84765625,
+      "rewards/rejected": -10.0625,
+      "step": 1060
+    },
+    {
+      "epoch": 0.5599162742019885,
+      "grad_norm": 7.573253707315908,
+      "learning_rate": 4.5236158520003444e-05,
+      "logits/chosen": -12.9375,
+      "logits/rejected": -13.0,
+      "logps/chosen": -536.0,
+      "logps/rejected": -502.0,
+      "loss": 0.6025,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -8.1875,
+      "rewards/margins": 1.0390625,
+      "rewards/rejected": -9.25,
+      "step": 1070
+    },
+    {
+      "epoch": 0.565149136577708,
+      "grad_norm": 7.888840261037634,
+      "learning_rate": 4.510121279963709e-05,
+      "logits/chosen": -12.1875,
+      "logits/rejected": -12.3125,
+      "logps/chosen": -528.0,
+      "logps/rejected": -496.0,
+      "loss": 0.7204,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -10.0625,
+      "rewards/margins": 1.03125,
+      "rewards/rejected": -11.0625,
+      "step": 1080
+    },
+    {
+      "epoch": 0.5703819989534276,
+      "grad_norm": 13.762939010968964,
+      "learning_rate": 4.4964589609762095e-05,
+      "logits/chosen": -11.9375,
+      "logits/rejected": -12.125,
+      "logps/chosen": -564.0,
+      "logps/rejected": -494.0,
+      "loss": 0.8346,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -11.4375,
+      "rewards/margins": 0.7734375,
+      "rewards/rejected": -12.1875,
+      "step": 1090
+    },
+    {
+      "epoch": 0.5756148613291471,
+      "grad_norm": 9.56155800775074,
+      "learning_rate": 4.482630035174205e-05,
+      "logits/chosen": -12.0,
+      "logits/rejected": -11.9375,
+      "logps/chosen": -512.0,
+      "logps/rejected": -492.0,
+      "loss": 0.6975,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -10.8125,
+      "rewards/margins": 0.828125,
+      "rewards/rejected": -11.625,
+      "step": 1100
+    },
+    {
+      "epoch": 0.5808477237048666,
+      "grad_norm": 9.650034622926128,
+      "learning_rate": 4.468635656597582e-05,
+      "logits/chosen": -12.0,
+      "logits/rejected": -11.9375,
+      "logps/chosen": -490.0,
+      "logps/rejected": -488.0,
+      "loss": 0.7859,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -9.75,
+      "rewards/margins": 0.77734375,
+      "rewards/rejected": -10.5,
+      "step": 1110
+    },
+    {
+      "epoch": 0.5860805860805861,
+      "grad_norm": 9.009994906294855,
+      "learning_rate": 4.454476993093454e-05,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -11.375,
+      "logps/chosen": -580.0,
+      "logps/rejected": -524.0,
+      "loss": 0.9287,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -10.375,
+      "rewards/margins": 0.58203125,
+      "rewards/rejected": -10.9375,
+      "step": 1120
+    },
+    {
+      "epoch": 0.5913134484563056,
+      "grad_norm": 10.042283135138778,
+      "learning_rate": 4.440155226218703e-05,
+      "logits/chosen": -11.375,
+      "logits/rejected": -11.1875,
+      "logps/chosen": -496.0,
+      "logps/rejected": -504.0,
+      "loss": 0.8404,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -9.25,
+      "rewards/margins": 0.66015625,
+      "rewards/rejected": -9.875,
+      "step": 1130
+    },
+    {
+      "epoch": 0.5965463108320251,
+      "grad_norm": 8.418582913270656,
+      "learning_rate": 4.425671551141376e-05,
+      "logits/chosen": -11.125,
+      "logits/rejected": -10.875,
+      "logps/chosen": -564.0,
+      "logps/rejected": -520.0,
+      "loss": 0.6583,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -10.5625,
+      "rewards/margins": 0.8671875,
+      "rewards/rejected": -11.5,
+      "step": 1140
+    },
+    {
+      "epoch": 0.6017791732077447,
+      "grad_norm": 9.416691885986982,
+      "learning_rate": 4.411027176540948e-05,
+      "logits/chosen": -10.75,
+      "logits/rejected": -10.5625,
+      "logps/chosen": -524.0,
+      "logps/rejected": -510.0,
+      "loss": 0.6848,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -11.0,
+      "rewards/margins": 1.234375,
+      "rewards/rejected": -12.1875,
+      "step": 1150
+    },
+    {
+      "epoch": 0.6070120355834642,
+      "grad_norm": 10.865786644213799,
+      "learning_rate": 4.396223324507454e-05,
+      "logits/chosen": -10.8125,
+      "logits/rejected": -10.75,
+      "logps/chosen": -524.0,
+      "logps/rejected": -454.0,
+      "loss": 0.9435,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -11.4375,
+      "rewards/margins": 0.640625,
+      "rewards/rejected": -12.125,
+      "step": 1160
+    },
+    {
+      "epoch": 0.6122448979591837,
+      "grad_norm": 12.184578875756303,
+      "learning_rate": 4.3812612304395046e-05,
+      "logits/chosen": -11.25,
+      "logits/rejected": -11.125,
+      "logps/chosen": -556.0,
+      "logps/rejected": -564.0,
+      "loss": 0.7482,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -11.375,
+      "rewards/margins": 0.76953125,
+      "rewards/rejected": -12.125,
+      "step": 1170
+    },
+    {
+      "epoch": 0.6174777603349032,
+      "grad_norm": 7.183055929786656,
+      "learning_rate": 4.366142142941195e-05,
+      "logits/chosen": -10.6875,
+      "logits/rejected": -10.375,
+      "logps/chosen": -552.0,
+      "logps/rejected": -516.0,
+      "loss": 0.711,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -11.125,
+      "rewards/margins": 0.8515625,
+      "rewards/rejected": -12.0,
+      "step": 1180
+    },
+    {
+      "epoch": 0.6227106227106227,
+      "grad_norm": 7.359506759653038,
+      "learning_rate": 4.350867323717902e-05,
+      "logits/chosen": -10.625,
+      "logits/rejected": -10.375,
+      "logps/chosen": -528.0,
+      "logps/rejected": -510.0,
+      "loss": 0.5868,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -10.6875,
+      "rewards/margins": 0.7890625,
+      "rewards/rejected": -11.4375,
+      "step": 1190
+    },
+    {
+      "epoch": 0.6279434850863422,
+      "grad_norm": 6.501759078401181,
+      "learning_rate": 4.335438047470996e-05,
+      "logits/chosen": -10.75,
+      "logits/rejected": -10.5,
+      "logps/chosen": -528.0,
+      "logps/rejected": -536.0,
+      "loss": 0.6786,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -10.625,
+      "rewards/margins": 1.4296875,
+      "rewards/rejected": -12.0625,
+      "step": 1200
+    },
+    {
+      "epoch": 0.6331763474620618,
+      "grad_norm": 7.620878974669834,
+      "learning_rate": 4.3198556017914635e-05,
+      "logits/chosen": -11.1875,
+      "logits/rejected": -10.875,
+      "logps/chosen": -588.0,
+      "logps/rejected": -524.0,
+      "loss": 0.7357,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -11.125,
+      "rewards/margins": 1.40625,
+      "rewards/rejected": -12.5625,
+      "step": 1210
+    },
+    {
+      "epoch": 0.6384092098377813,
+      "grad_norm": 9.34056834357026,
+      "learning_rate": 4.30412128705246e-05,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -11.3125,
+      "logps/chosen": -564.0,
+      "logps/rejected": -520.0,
+      "loss": 0.7723,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -10.375,
+      "rewards/margins": 1.109375,
+      "rewards/rejected": -11.4375,
+      "step": 1220
+    },
+    {
+      "epoch": 0.6436420722135008,
+      "grad_norm": 7.674870336983743,
+      "learning_rate": 4.28823641630079e-05,
+      "logits/chosen": -11.375,
+      "logits/rejected": -11.1875,
+      "logps/chosen": -568.0,
+      "logps/rejected": -516.0,
+      "loss": 0.7292,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -10.125,
+      "rewards/margins": 1.5078125,
+      "rewards/rejected": -11.625,
+      "step": 1230
+    },
+    {
+      "epoch": 0.6488749345892203,
+      "grad_norm": 7.40544549406474,
+      "learning_rate": 4.2722023151473294e-05,
+      "logits/chosen": -10.9375,
+      "logits/rejected": -10.9375,
+      "logps/chosen": -486.0,
+      "logps/rejected": -492.0,
+      "loss": 0.7212,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -10.75,
+      "rewards/margins": 1.1171875,
+      "rewards/rejected": -11.875,
+      "step": 1240
+    },
+    {
+      "epoch": 0.6541077969649398,
+      "grad_norm": 9.53123578774151,
+      "learning_rate": 4.256020321656405e-05,
+      "logits/chosen": -10.625,
+      "logits/rejected": -10.375,
+      "logps/chosen": -560.0,
+      "logps/rejected": -552.0,
+      "loss": 0.7306,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -12.3125,
+      "rewards/margins": 0.95703125,
+      "rewards/rejected": -13.25,
+      "step": 1250
+    },
+    {
+      "epoch": 0.6593406593406593,
+      "grad_norm": 6.161183952229396,
+      "learning_rate": 4.239691786234133e-05,
+      "logits/chosen": -11.0,
+      "logits/rejected": -10.9375,
+      "logps/chosen": -544.0,
+      "logps/rejected": -488.0,
+      "loss": 0.6762,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -10.9375,
+      "rewards/margins": 1.140625,
+      "rewards/rejected": -12.125,
+      "step": 1260
+    },
+    {
+      "epoch": 0.6645735217163788,
+      "grad_norm": 8.172666817720033,
+      "learning_rate": 4.223218071515721e-05,
+      "logits/chosen": -11.0,
+      "logits/rejected": -10.875,
+      "logps/chosen": -544.0,
+      "logps/rejected": -516.0,
+      "loss": 0.6903,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -11.0625,
+      "rewards/margins": 0.99609375,
+      "rewards/rejected": -12.0625,
+      "step": 1270
+    },
+    {
+      "epoch": 0.6698063840920984,
+      "grad_norm": 12.188827913123875,
+      "learning_rate": 4.206600552251756e-05,
+      "logits/chosen": -11.25,
+      "logits/rejected": -11.125,
+      "logps/chosen": -524.0,
+      "logps/rejected": -486.0,
+      "loss": 0.79,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -10.5,
+      "rewards/margins": 0.9609375,
+      "rewards/rejected": -11.4375,
+      "step": 1280
+    },
+    {
+      "epoch": 0.6750392464678179,
+      "grad_norm": 18.826047916539064,
+      "learning_rate": 4.189840615193486e-05,
+      "logits/chosen": -11.6875,
+      "logits/rejected": -11.375,
+      "logps/chosen": -498.0,
+      "logps/rejected": -520.0,
+      "loss": 0.7081,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -9.8125,
+      "rewards/margins": 0.96875,
+      "rewards/rejected": -10.8125,
+      "step": 1290
+    },
+    {
+      "epoch": 0.6802721088435374,
+      "grad_norm": 6.969800701972501,
+      "learning_rate": 4.172939658977084e-05,
+      "logits/chosen": -11.1875,
+      "logits/rejected": -11.0,
+      "logps/chosen": -528.0,
+      "logps/rejected": -544.0,
+      "loss": 0.7148,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -10.625,
+      "rewards/margins": 0.921875,
+      "rewards/rejected": -11.5625,
+      "step": 1300
+    },
+    {
+      "epoch": 0.6855049712192569,
+      "grad_norm": 14.859556755269326,
+      "learning_rate": 4.155899094006938e-05,
+      "logits/chosen": -10.9375,
+      "logits/rejected": -10.8125,
+      "logps/chosen": -564.0,
+      "logps/rejected": -528.0,
+      "loss": 0.7416,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -10.875,
+      "rewards/margins": 0.9609375,
+      "rewards/rejected": -11.875,
+      "step": 1310
+    },
+    {
+      "epoch": 0.6907378335949764,
+      "grad_norm": 6.957089873593148,
+      "learning_rate": 4.138720342337947e-05,
+      "logits/chosen": -11.125,
+      "logits/rejected": -11.0,
+      "logps/chosen": -564.0,
+      "logps/rejected": -552.0,
+      "loss": 0.6879,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -11.75,
+      "rewards/margins": 1.0859375,
+      "rewards/rejected": -12.8125,
+      "step": 1320
+    },
+    {
+      "epoch": 0.6959706959706959,
+      "grad_norm": 9.707692830662324,
+      "learning_rate": 4.121404837556851e-05,
+      "logits/chosen": -11.9375,
+      "logits/rejected": -11.5625,
+      "logps/chosen": -580.0,
+      "logps/rejected": -616.0,
+      "loss": 0.6995,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -11.5625,
+      "rewards/margins": 1.546875,
+      "rewards/rejected": -13.0625,
+      "step": 1330
+    },
+    {
+      "epoch": 0.7012035583464155,
+      "grad_norm": 8.770735737960996,
+      "learning_rate": 4.103954024662594e-05,
+      "logits/chosen": -12.375,
+      "logits/rejected": -12.4375,
+      "logps/chosen": -568.0,
+      "logps/rejected": -536.0,
+      "loss": 0.719,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": -11.25,
+      "rewards/margins": 0.640625,
+      "rewards/rejected": -11.875,
+      "step": 1340
+    },
+    {
+      "epoch": 0.706436420722135,
+      "grad_norm": 7.1593445389385515,
+      "learning_rate": 4.086369359945743e-05,
+      "logits/chosen": -12.5,
+      "logits/rejected": -12.4375,
+      "logps/chosen": -576.0,
+      "logps/rejected": -564.0,
+      "loss": 0.7039,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -10.5625,
+      "rewards/margins": 1.0546875,
+      "rewards/rejected": -11.625,
+      "step": 1350
+    },
+    {
+      "epoch": 0.7116692830978545,
+      "grad_norm": 9.243196425255588,
+      "learning_rate": 4.0686523108669496e-05,
+      "logits/chosen": -12.5625,
+      "logits/rejected": -12.625,
+      "logps/chosen": -600.0,
+      "logps/rejected": -572.0,
+      "loss": 0.8775,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -10.9375,
+      "rewards/margins": 1.125,
+      "rewards/rejected": -12.0625,
+      "step": 1360
+    },
+    {
+      "epoch": 0.716902145473574,
+      "grad_norm": 11.110076974797884,
+      "learning_rate": 4.050804355934498e-05,
+      "logits/chosen": -11.25,
+      "logits/rejected": -11.1875,
+      "logps/chosen": -568.0,
+      "logps/rejected": -524.0,
+      "loss": 0.6187,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -11.4375,
+      "rewards/margins": 0.9375,
+      "rewards/rejected": -12.375,
+      "step": 1370
+    },
+    {
+      "epoch": 0.7221350078492935,
+      "grad_norm": 6.00630186538614,
+      "learning_rate": 4.032826984580914e-05,
+      "logits/chosen": -12.4375,
+      "logits/rejected": -12.125,
+      "logps/chosen": -512.0,
+      "logps/rejected": -484.0,
+      "loss": 0.6801,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -10.3125,
+      "rewards/margins": 1.078125,
+      "rewards/rejected": -11.375,
+      "step": 1380
+    },
+    {
+      "epoch": 0.727367870225013,
+      "grad_norm": 8.977754421118522,
+      "learning_rate": 4.014721697038678e-05,
+      "logits/chosen": -11.6875,
+      "logits/rejected": -11.4375,
+      "logps/chosen": -490.0,
+      "logps/rejected": -482.0,
+      "loss": 0.6641,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -9.625,
+      "rewards/margins": 1.1015625,
+      "rewards/rejected": -10.75,
+      "step": 1390
+    },
+    {
+      "epoch": 0.7326007326007326,
+      "grad_norm": 9.068303784048501,
+      "learning_rate": 3.996490004215021e-05,
+      "logits/chosen": -11.8125,
+      "logits/rejected": -11.6875,
+      "logps/chosen": -512.0,
+      "logps/rejected": -498.0,
+      "loss": 0.6787,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -10.3125,
+      "rewards/margins": 1.125,
+      "rewards/rejected": -11.4375,
+      "step": 1400
+    },
+    {
+      "epoch": 0.7378335949764521,
+      "grad_norm": 10.091388237018714,
+      "learning_rate": 3.978133427565842e-05,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -11.125,
+      "logps/chosen": -524.0,
+      "logps/rejected": -516.0,
+      "loss": 0.7733,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -11.25,
+      "rewards/margins": 0.84375,
+      "rewards/rejected": -12.125,
+      "step": 1410
+    },
+    {
+      "epoch": 0.7430664573521716,
+      "grad_norm": 9.639237375339137,
+      "learning_rate": 3.9596534989687416e-05,
+      "logits/chosen": -12.4375,
+      "logits/rejected": -12.1875,
+      "logps/chosen": -506.0,
+      "logps/rejected": -504.0,
+      "loss": 0.7797,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -10.0,
+      "rewards/margins": 1.8671875,
+      "rewards/rejected": -11.875,
+      "step": 1420
+    },
+    {
+      "epoch": 0.7482993197278912,
+      "grad_norm": 6.806097018429323,
+      "learning_rate": 3.9410517605951824e-05,
+      "logits/chosen": -13.125,
+      "logits/rejected": -12.9375,
+      "logps/chosen": -532.0,
+      "logps/rejected": -488.0,
+      "loss": 0.6814,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -10.6875,
+      "rewards/margins": 0.486328125,
+      "rewards/rejected": -11.1875,
+      "step": 1430
+    },
+    {
+      "epoch": 0.7535321821036107,
+      "grad_norm": 9.16956662251956,
+      "learning_rate": 3.922329764781793e-05,
+      "logits/chosen": -12.375,
+      "logits/rejected": -12.125,
+      "logps/chosen": -588.0,
+      "logps/rejected": -544.0,
+      "loss": 0.7421,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -11.0,
+      "rewards/margins": 1.1015625,
+      "rewards/rejected": -12.125,
+      "step": 1440
+    },
+    {
+      "epoch": 0.7587650444793302,
+      "grad_norm": 7.958218718062474,
+      "learning_rate": 3.903489073900828e-05,
+      "logits/chosen": -11.5625,
+      "logits/rejected": -11.5,
+      "logps/chosen": -552.0,
+      "logps/rejected": -532.0,
+      "loss": 0.7735,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -11.5625,
+      "rewards/margins": 0.8203125,
+      "rewards/rejected": -12.375,
+      "step": 1450
+    },
+    {
+      "epoch": 0.7639979068550498,
+      "grad_norm": 8.365628327112017,
+      "learning_rate": 3.884531260229778e-05,
+      "logits/chosen": -11.75,
+      "logits/rejected": -11.625,
+      "logps/chosen": -536.0,
+      "logps/rejected": -490.0,
+      "loss": 0.7042,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": -11.1875,
+      "rewards/margins": 0.5390625,
+      "rewards/rejected": -11.75,
+      "step": 1460
+    },
+    {
+      "epoch": 0.7692307692307693,
+      "grad_norm": 7.681754910048477,
+      "learning_rate": 3.8654579058201704e-05,
+      "logits/chosen": -11.5625,
+      "logits/rejected": -11.3125,
+      "logps/chosen": -516.0,
+      "logps/rejected": -492.0,
+      "loss": 0.6215,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -10.3125,
+      "rewards/margins": 1.109375,
+      "rewards/rejected": -11.4375,
+      "step": 1470
+    },
+    {
+      "epoch": 0.7744636316064888,
+      "grad_norm": 8.699313776529005,
+      "learning_rate": 3.8462706023655404e-05,
+      "logits/chosen": -10.9375,
+      "logits/rejected": -11.0,
+      "logps/chosen": -516.0,
+      "logps/rejected": -520.0,
+      "loss": 0.7719,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -11.5625,
+      "rewards/margins": 0.89453125,
+      "rewards/rejected": -12.4375,
+      "step": 1480
+    },
+    {
+      "epoch": 0.7796964939822083,
+      "grad_norm": 10.20603489800159,
+      "learning_rate": 3.8269709510686005e-05,
+      "logits/chosen": -10.9375,
+      "logits/rejected": -10.8125,
+      "logps/chosen": -536.0,
+      "logps/rejected": -520.0,
+      "loss": 0.7257,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -11.1875,
+      "rewards/margins": 0.7265625,
+      "rewards/rejected": -11.9375,
+      "step": 1490
+    },
+    {
+      "epoch": 0.7849293563579278,
+      "grad_norm": 6.371725891794927,
+      "learning_rate": 3.807560562507624e-05,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -11.0,
+      "logps/chosen": -478.0,
+      "logps/rejected": -516.0,
+      "loss": 0.6052,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -10.8125,
+      "rewards/margins": 1.390625,
+      "rewards/rejected": -12.1875,
+      "step": 1500
+    },
+    {
+      "epoch": 0.7901622187336473,
+      "grad_norm": 8.044514431743275,
+      "learning_rate": 3.7880410565020366e-05,
+      "logits/chosen": -11.9375,
+      "logits/rejected": -11.625,
+      "logps/chosen": -532.0,
+      "logps/rejected": -552.0,
+      "loss": 0.722,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": -11.4375,
+      "rewards/margins": 0.9765625,
+      "rewards/rejected": -12.4375,
+      "step": 1510
+    },
+    {
+      "epoch": 0.7953950811093669,
+      "grad_norm": 6.3419982155290855,
+      "learning_rate": 3.76841406197724e-05,
+      "logits/chosen": -12.25,
+      "logits/rejected": -12.125,
+      "logps/chosen": -468.0,
+      "logps/rejected": -460.0,
+      "loss": 0.693,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -9.6875,
+      "rewards/margins": 1.0390625,
+      "rewards/rejected": -10.6875,
+      "step": 1520
+    },
+    {
+      "epoch": 0.8006279434850864,
+      "grad_norm": 13.822774811604797,
+      "learning_rate": 3.748681216828678e-05,
+      "logits/chosen": -12.1875,
+      "logits/rejected": -11.75,
+      "logps/chosen": -572.0,
+      "logps/rejected": -644.0,
+      "loss": 0.7042,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -11.4375,
+      "rewards/margins": 0.76953125,
+      "rewards/rejected": -12.25,
+      "step": 1530
+    },
+    {
+      "epoch": 0.8058608058608059,
+      "grad_norm": 8.295919264278089,
+      "learning_rate": 3.728844167785151e-05,
+      "logits/chosen": -11.875,
+      "logits/rejected": -11.625,
+      "logps/chosen": -584.0,
+      "logps/rejected": -572.0,
+      "loss": 0.7166,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -11.1875,
+      "rewards/margins": 0.8359375,
+      "rewards/rejected": -12.0,
+      "step": 1540
+    },
+    {
+      "epoch": 0.8110936682365254,
+      "grad_norm": 8.520816598792539,
+      "learning_rate": 3.7089045702713976e-05,
+      "logits/chosen": -12.0625,
+      "logits/rejected": -12.25,
+      "logps/chosen": -584.0,
+      "logps/rejected": -510.0,
+      "loss": 0.7012,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -10.3125,
+      "rewards/margins": 1.0390625,
+      "rewards/rejected": -11.375,
+      "step": 1550
+    },
+    {
+      "epoch": 0.8163265306122449,
+      "grad_norm": 8.142573842126938,
+      "learning_rate": 3.6888640882699425e-05,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -11.3125,
+      "logps/chosen": -584.0,
+      "logps/rejected": -568.0,
+      "loss": 0.7397,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -11.625,
+      "rewards/margins": 0.89453125,
+      "rewards/rejected": -12.5,
+      "step": 1560
+    },
+    {
+      "epoch": 0.8215593929879644,
+      "grad_norm": 7.481987186912928,
+      "learning_rate": 3.668724394182239e-05,
+      "logits/chosen": -11.375,
+      "logits/rejected": -10.9375,
+      "logps/chosen": -528.0,
+      "logps/rejected": -524.0,
+      "loss": 0.7039,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -11.9375,
+      "rewards/margins": 1.1484375,
+      "rewards/rejected": -13.0625,
+      "step": 1570
+    },
+    {
+      "epoch": 0.826792255363684,
+      "grad_norm": 8.042620411635573,
+      "learning_rate": 3.648487168689104e-05,
+      "logits/chosen": -11.5625,
+      "logits/rejected": -11.375,
+      "logps/chosen": -604.0,
+      "logps/rejected": -536.0,
+      "loss": 0.7544,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -11.25,
+      "rewards/margins": 0.69140625,
+      "rewards/rejected": -11.9375,
+      "step": 1580
+    },
+    {
+      "epoch": 0.8320251177394035,
+      "grad_norm": 6.330392212337275,
+      "learning_rate": 3.628154100610463e-05,
+      "logits/chosen": -11.875,
+      "logits/rejected": -11.5,
+      "logps/chosen": -528.0,
+      "logps/rejected": -496.0,
+      "loss": 0.7201,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -9.9375,
+      "rewards/margins": 1.0625,
+      "rewards/rejected": -11.0,
+      "step": 1590
+    },
+    {
+      "epoch": 0.837257980115123,
+      "grad_norm": 9.495665962948788,
+      "learning_rate": 3.607726886764415e-05,
+      "logits/chosen": -12.0625,
+      "logits/rejected": -11.875,
+      "logps/chosen": -506.0,
+      "logps/rejected": -544.0,
+      "loss": 0.7175,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -10.3125,
+      "rewards/margins": 0.96484375,
+      "rewards/rejected": -11.25,
+      "step": 1600
+    },
+    {
+      "epoch": 0.8424908424908425,
+      "grad_norm": 8.88992804946847,
+      "learning_rate": 3.5872072318256375e-05,
+      "logits/chosen": -11.9375,
+      "logits/rejected": -11.75,
+      "logps/chosen": -596.0,
+      "logps/rejected": -552.0,
+      "loss": 0.7063,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -10.75,
+      "rewards/margins": 1.25,
+      "rewards/rejected": -12.0,
+      "step": 1610
+    },
+    {
+      "epoch": 0.847723704866562,
+      "grad_norm": 6.549366426211227,
+      "learning_rate": 3.566596848183117e-05,
+      "logits/chosen": -11.375,
+      "logits/rejected": -11.3125,
+      "logps/chosen": -528.0,
+      "logps/rejected": -488.0,
+      "loss": 0.7635,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -11.4375,
+      "rewards/margins": 0.74609375,
+      "rewards/rejected": -12.1875,
+      "step": 1620
+    },
+    {
+      "epoch": 0.8529565672422815,
+      "grad_norm": 8.965681557356943,
+      "learning_rate": 3.54589745579726e-05,
+      "logits/chosen": -12.0,
+      "logits/rejected": -11.5625,
+      "logps/chosen": -560.0,
+      "logps/rejected": -564.0,
+      "loss": 0.7312,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -11.125,
+      "rewards/margins": 0.96875,
+      "rewards/rejected": -12.0625,
+      "step": 1630
+    },
+    {
+      "epoch": 0.858189429618001,
+      "grad_norm": 7.597359211830405,
+      "learning_rate": 3.5251107820563565e-05,
+      "logits/chosen": -12.4375,
+      "logits/rejected": -12.5625,
+      "logps/chosen": -580.0,
+      "logps/rejected": -540.0,
+      "loss": 0.7307,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -9.875,
+      "rewards/margins": 0.94140625,
+      "rewards/rejected": -10.8125,
+      "step": 1640
+    },
+    {
+      "epoch": 0.8634222919937206,
+      "grad_norm": 16.306018432981723,
+      "learning_rate": 3.504238561632424e-05,
+      "logits/chosen": -12.75,
+      "logits/rejected": -12.6875,
+      "logps/chosen": -512.0,
+      "logps/rejected": -516.0,
+      "loss": 0.788,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -9.8125,
+      "rewards/margins": 0.9375,
+      "rewards/rejected": -10.75,
+      "step": 1650
+    },
+    {
+      "epoch": 0.8686551543694401,
+      "grad_norm": 6.41824790740516,
+      "learning_rate": 3.483282536336451e-05,
+      "logits/chosen": -12.75,
+      "logits/rejected": -12.6875,
+      "logps/chosen": -468.0,
+      "logps/rejected": -464.0,
+      "loss": 0.7286,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -9.125,
+      "rewards/margins": 1.0546875,
+      "rewards/rejected": -10.125,
+      "step": 1660
+    },
+    {
+      "epoch": 0.8738880167451596,
+      "grad_norm": 8.997202419771773,
+      "learning_rate": 3.46224445497304e-05,
+      "logits/chosen": -11.8125,
+      "logits/rejected": -11.8125,
+      "logps/chosen": -552.0,
+      "logps/rejected": -532.0,
+      "loss": 0.6613,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": -10.1875,
+      "rewards/margins": 1.53125,
+      "rewards/rejected": -11.75,
+      "step": 1670
+    },
+    {
+      "epoch": 0.8791208791208791,
+      "grad_norm": 8.646781069565186,
+      "learning_rate": 3.441126073194468e-05,
+      "logits/chosen": -12.1875,
+      "logits/rejected": -11.75,
+      "logps/chosen": -504.0,
+      "logps/rejected": -532.0,
+      "loss": 0.6581,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -9.9375,
+      "rewards/margins": 1.546875,
+      "rewards/rejected": -11.5,
+      "step": 1680
+    },
+    {
+      "epoch": 0.8843537414965986,
+      "grad_norm": 7.113786319184486,
+      "learning_rate": 3.4199291533541735e-05,
+      "logits/chosen": -11.625,
+      "logits/rejected": -11.5,
+      "logps/chosen": -520.0,
+      "logps/rejected": -494.0,
+      "loss": 0.6574,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -10.4375,
+      "rewards/margins": 0.8671875,
+      "rewards/rejected": -11.3125,
+      "step": 1690
+    },
+    {
+      "epoch": 0.8895866038723181,
+      "grad_norm": 9.146215241563787,
+      "learning_rate": 3.398655464359687e-05,
+      "logits/chosen": -11.9375,
+      "logits/rejected": -11.8125,
+      "logps/chosen": -604.0,
+      "logps/rejected": -484.0,
+      "loss": 1.2266,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -12.8125,
+      "rewards/margins": -1.1484375,
+      "rewards/rejected": -11.6875,
+      "step": 1700
+    },
+    {
+      "epoch": 0.8948194662480377,
+      "grad_norm": 7.301197069927249,
+      "learning_rate": 3.377306781525015e-05,
+      "logits/chosen": -12.0,
+      "logits/rejected": -11.875,
+      "logps/chosen": -560.0,
+      "logps/rejected": -540.0,
+      "loss": 0.7462,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -10.9375,
+      "rewards/margins": 1.0234375,
+      "rewards/rejected": -11.9375,
+      "step": 1710
+    },
+    {
+      "epoch": 0.9000523286237572,
+      "grad_norm": 7.026924797144021,
+      "learning_rate": 3.3558848864224876e-05,
+      "logits/chosen": -12.25,
+      "logits/rejected": -12.1875,
+      "logps/chosen": -540.0,
+      "logps/rejected": -506.0,
+      "loss": 0.6618,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -10.4375,
+      "rewards/margins": 0.85546875,
+      "rewards/rejected": -11.3125,
+      "step": 1720
+    },
+    {
+      "epoch": 0.9052851909994767,
+      "grad_norm": 11.941983652096013,
+      "learning_rate": 3.334391566734082e-05,
+      "logits/chosen": -12.0,
+      "logits/rejected": -12.0,
+      "logps/chosen": -492.0,
+      "logps/rejected": -494.0,
+      "loss": 0.7589,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -9.375,
+      "rewards/margins": 1.1484375,
+      "rewards/rejected": -10.5,
+      "step": 1730
+    },
+    {
+      "epoch": 0.9105180533751962,
+      "grad_norm": 8.065484818366723,
+      "learning_rate": 3.3128286161022394e-05,
+      "logits/chosen": -11.6875,
+      "logits/rejected": -11.75,
+      "logps/chosen": -552.0,
+      "logps/rejected": -510.0,
+      "loss": 0.7712,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -10.5,
+      "rewards/margins": 0.984375,
+      "rewards/rejected": -11.4375,
+      "step": 1740
+    },
+    {
+      "epoch": 0.9157509157509157,
+      "grad_norm": 8.045324511262809,
+      "learning_rate": 3.2911978339801855e-05,
+      "logits/chosen": -11.9375,
+      "logits/rejected": -12.0,
+      "logps/chosen": -584.0,
+      "logps/rejected": -568.0,
+      "loss": 0.7152,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -10.375,
+      "rewards/margins": 1.2890625,
+      "rewards/rejected": -11.6875,
+      "step": 1750
+    },
+    {
+      "epoch": 0.9209837781266352,
+      "grad_norm": 10.25223828631299,
+      "learning_rate": 3.269501025481763e-05,
+      "logits/chosen": -12.375,
+      "logits/rejected": -12.3125,
+      "logps/chosen": -528.0,
+      "logps/rejected": -560.0,
+      "loss": 0.6604,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -10.125,
+      "rewards/margins": 0.89453125,
+      "rewards/rejected": -11.0,
+      "step": 1760
+    },
+    {
+      "epoch": 0.9262166405023547,
+      "grad_norm": 6.658643984408585,
+      "learning_rate": 3.2477400012307885e-05,
+      "logits/chosen": -12.25,
+      "logits/rejected": -12.25,
+      "logps/chosen": -540.0,
+      "logps/rejected": -540.0,
+      "loss": 0.7548,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -9.75,
+      "rewards/margins": 1.0,
+      "rewards/rejected": -10.75,
+      "step": 1770
+    },
+    {
+      "epoch": 0.9314495028780743,
+      "grad_norm": 7.583843351290834,
+      "learning_rate": 3.2259165772099644e-05,
+      "logits/chosen": -12.8125,
+      "logits/rejected": -12.9375,
+      "logps/chosen": -540.0,
+      "logps/rejected": -528.0,
+      "loss": 0.6979,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -9.75,
+      "rewards/margins": 1.1171875,
+      "rewards/rejected": -10.875,
+      "step": 1780
+    },
+    {
+      "epoch": 0.9366823652537938,
+      "grad_norm": 6.695330181972035,
+      "learning_rate": 3.204032574609318e-05,
+      "logits/chosen": -12.25,
+      "logits/rejected": -12.375,
+      "logps/chosen": -576.0,
+      "logps/rejected": -568.0,
+      "loss": 0.7396,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -10.5,
+      "rewards/margins": 1.046875,
+      "rewards/rejected": -11.5,
+      "step": 1790
+    },
+    {
+      "epoch": 0.9419152276295133,
+      "grad_norm": 7.41109532240638,
+      "learning_rate": 3.1820898196742335e-05,
+      "logits/chosen": -12.0,
+      "logits/rejected": -12.0,
+      "logps/chosen": -520.0,
+      "logps/rejected": -548.0,
+      "loss": 0.7853,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -11.0625,
+      "rewards/margins": 0.8203125,
+      "rewards/rejected": -11.875,
+      "step": 1800
+    },
+    {
+      "epoch": 0.9471480900052328,
+      "grad_norm": 7.998757902945492,
+      "learning_rate": 3.160090143553049e-05,
+      "logits/chosen": -11.875,
+      "logits/rejected": -11.625,
+      "logps/chosen": -476.0,
+      "logps/rejected": -476.0,
+      "loss": 0.6803,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -10.375,
+      "rewards/margins": 1.015625,
+      "rewards/rejected": -11.4375,
+      "step": 1810
+    },
+    {
+      "epoch": 0.9523809523809523,
+      "grad_norm": 7.080219482241533,
+      "learning_rate": 3.1380353821442354e-05,
+      "logits/chosen": -12.0,
+      "logits/rejected": -11.8125,
+      "logps/chosen": -448.0,
+      "logps/rejected": -446.0,
+      "loss": 0.7265,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -9.75,
+      "rewards/margins": 0.76171875,
+      "rewards/rejected": -10.5,
+      "step": 1820
+    },
+    {
+      "epoch": 0.957613814756672,
+      "grad_norm": 10.03679929204167,
+      "learning_rate": 3.1159273759431964e-05,
+      "logits/chosen": -11.5625,
+      "logits/rejected": -11.5,
+      "logps/chosen": -544.0,
+      "logps/rejected": -516.0,
+      "loss": 0.6867,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -11.125,
+      "rewards/margins": 1.0703125,
+      "rewards/rejected": -12.25,
+      "step": 1830
+    },
+    {
+      "epoch": 0.9628466771323915,
+      "grad_norm": 6.830881800921508,
+      "learning_rate": 3.0937679698886786e-05,
+      "logits/chosen": -12.0625,
+      "logits/rejected": -11.875,
+      "logps/chosen": -560.0,
+      "logps/rejected": -540.0,
+      "loss": 0.5938,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -11.125,
+      "rewards/margins": 1.3125,
+      "rewards/rejected": -12.4375,
+      "step": 1840
+    },
+    {
+      "epoch": 0.968079539508111,
+      "grad_norm": 8.529607259118398,
+      "learning_rate": 3.071559013208801e-05,
+      "logits/chosen": -11.75,
+      "logits/rejected": -11.5625,
+      "logps/chosen": -628.0,
+      "logps/rejected": -568.0,
+      "loss": 0.7147,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -11.4375,
+      "rewards/margins": 1.2109375,
+      "rewards/rejected": -12.625,
+      "step": 1850
+    },
+    {
+      "epoch": 0.9733124018838305,
+      "grad_norm": 6.915342277139341,
+      "learning_rate": 3.0493023592667446e-05,
+      "logits/chosen": -11.4375,
+      "logits/rejected": -11.25,
+      "logps/chosen": -540.0,
+      "logps/rejected": -500.0,
+      "loss": 0.7208,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -10.625,
+      "rewards/margins": 0.76953125,
+      "rewards/rejected": -11.375,
+      "step": 1860
+    },
+    {
+      "epoch": 0.97854526425955,
+      "grad_norm": 9.778948710217877,
+      "learning_rate": 3.0269998654060788e-05,
+      "logits/chosen": -12.125,
+      "logits/rejected": -12.0,
+      "logps/chosen": -502.0,
+      "logps/rejected": -462.0,
+      "loss": 0.6835,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -9.75,
+      "rewards/margins": 0.7421875,
+      "rewards/rejected": -10.5,
+      "step": 1870
+    },
+    {
+      "epoch": 0.9837781266352695,
+      "grad_norm": 5.655349340296246,
+      "learning_rate": 3.0046533927957677e-05,
+      "logits/chosen": -11.5625,
+      "logits/rejected": -11.3125,
+      "logps/chosen": -520.0,
+      "logps/rejected": -540.0,
+      "loss": 0.6201,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -10.25,
+      "rewards/margins": 1.0,
+      "rewards/rejected": -11.25,
+      "step": 1880
+    },
+    {
+      "epoch": 0.989010989010989,
+      "grad_norm": 10.130846611986492,
+      "learning_rate": 2.9822648062748536e-05,
+      "logits/chosen": -12.375,
+      "logits/rejected": -12.375,
+      "logps/chosen": -612.0,
+      "logps/rejected": -600.0,
+      "loss": 0.679,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -10.5625,
+      "rewards/margins": 1.125,
+      "rewards/rejected": -11.6875,
+      "step": 1890
+    },
+    {
+      "epoch": 0.9942438513867086,
+      "grad_norm": 7.921271665865128,
+      "learning_rate": 2.959835974196836e-05,
+      "logits/chosen": -11.6875,
+      "logits/rejected": -11.5625,
+      "logps/chosen": -516.0,
+      "logps/rejected": -520.0,
+      "loss": 0.6907,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -10.375,
+      "rewards/margins": 1.0625,
+      "rewards/rejected": -11.4375,
+      "step": 1900
+    },
+    {
+      "epoch": 0.9994767137624281,
+      "grad_norm": 8.904276645319571,
+      "learning_rate": 2.9373687682737484e-05,
+      "logits/chosen": -12.0625,
+      "logits/rejected": -12.0,
+      "logps/chosen": -588.0,
+      "logps/rejected": -544.0,
+      "loss": 0.6791,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -10.8125,
+      "rewards/margins": 0.83203125,
+      "rewards/rejected": -11.625,
+      "step": 1910
+    },
+    {
+      "epoch": 1.0,
+      "eval_logits/chosen": -12.6875,
+      "eval_logits/rejected": -12.5625,
+      "eval_logps/chosen": -536.0,
+      "eval_logps/rejected": -528.0,
+      "eval_loss": 0.7097968459129333,
+      "eval_rewards/accuracies": 0.69921875,
+      "eval_rewards/chosen": -10.8125,
+      "eval_rewards/margins": 1.1328125,
+      "eval_rewards/rejected": -11.9375,
+      "eval_runtime": 47.5241,
+      "eval_samples_per_second": 42.084,
+      "eval_steps_per_second": 0.673,
+      "step": 1911
+    },
+    {
+      "epoch": 1.0047095761381475,
+      "grad_norm": 4.728380640009898,
+      "learning_rate": 2.9148650634199674e-05,
+      "logits/chosen": -12.25,
+      "logits/rejected": -12.1875,
+      "logps/chosen": -472.0,
+      "logps/rejected": -516.0,
+      "loss": 0.2767,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": -9.5625,
+      "rewards/margins": 3.671875,
+      "rewards/rejected": -13.25,
+      "step": 1920
+    },
+    {
+      "epoch": 1.0099424385138671,
+      "grad_norm": 1.024609812555977,
+      "learning_rate": 2.892326737595751e-05,
+      "logits/chosen": -12.25,
+      "logits/rejected": -11.8125,
+      "logps/chosen": -576.0,
+      "logps/rejected": -612.0,
+      "loss": 0.1663,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.375,
+      "rewards/margins": 6.4375,
+      "rewards/rejected": -15.8125,
+      "step": 1930
+    },
+    {
+      "epoch": 1.0151753008895865,
+      "grad_norm": 3.373687792077228,
+      "learning_rate": 2.869755671650512e-05,
+      "logits/chosen": -11.5625,
+      "logits/rejected": -11.25,
+      "logps/chosen": -540.0,
+      "logps/rejected": -660.0,
+      "loss": 0.1126,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -10.0,
+      "rewards/margins": 6.8125,
+      "rewards/rejected": -16.875,
+      "step": 1940
+    },
+    {
+      "epoch": 1.0204081632653061,
+      "grad_norm": 2.368031581220674,
+      "learning_rate": 2.847153749165869e-05,
+      "logits/chosen": -9.75,
+      "logits/rejected": -9.375,
+      "logps/chosen": -524.0,
+      "logps/rejected": -596.0,
+      "loss": 0.1652,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -11.75,
+      "rewards/margins": 4.5625,
+      "rewards/rejected": -16.375,
+      "step": 1950
+    },
+    {
+      "epoch": 1.0256410256410255,
+      "grad_norm": 1.63741968428215,
+      "learning_rate": 2.8245228562984516e-05,
+      "logits/chosen": -10.8125,
+      "logits/rejected": -10.125,
+      "logps/chosen": -536.0,
+      "logps/rejected": -588.0,
+      "loss": 0.1023,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -10.75,
+      "rewards/margins": 6.4375,
+      "rewards/rejected": -17.25,
+      "step": 1960
+    },
+    {
+      "epoch": 1.0308738880167452,
+      "grad_norm": 2.551100302367526,
+      "learning_rate": 2.8018648816225025e-05,
+      "logits/chosen": -11.5,
+      "logits/rejected": -11.0,
+      "logps/chosen": -556.0,
+      "logps/rejected": -640.0,
+      "loss": 0.1294,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -7.96875,
+      "rewards/margins": 8.4375,
+      "rewards/rejected": -16.375,
+      "step": 1970
+    },
+    {
+      "epoch": 1.0361067503924646,
+      "grad_norm": 2.6749422209678073,
+      "learning_rate": 2.7791817159722726e-05,
+      "logits/chosen": -10.6875,
+      "logits/rejected": -10.375,
+      "logps/chosen": -470.0,
+      "logps/rejected": -552.0,
+      "loss": 0.1154,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -9.0625,
+      "rewards/margins": 5.46875,
+      "rewards/rejected": -14.5625,
+      "step": 1980
+    },
+    {
+      "epoch": 1.0413396127681842,
+      "grad_norm": 2.655504829474548,
+      "learning_rate": 2.756475252284229e-05,
+      "logits/chosen": -11.25,
+      "logits/rejected": -10.9375,
+      "logps/chosen": -556.0,
+      "logps/rejected": -656.0,
+      "loss": 0.1274,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -9.1875,
+      "rewards/margins": 6.71875,
+      "rewards/rejected": -15.875,
+      "step": 1990
+    },
+    {
+      "epoch": 1.0465724751439036,
+      "grad_norm": 2.5200245472186418,
+      "learning_rate": 2.7337473854390865e-05,
+      "logits/chosen": -11.5625,
+      "logits/rejected": -11.375,
+      "logps/chosen": -516.0,
+      "logps/rejected": -584.0,
+      "loss": 0.164,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.0,
+      "rewards/margins": 6.25,
+      "rewards/rejected": -15.25,
+      "step": 2000
+    },
+    {
+      "epoch": 1.0518053375196232,
+      "grad_norm": 4.499576388025018,
+      "learning_rate": 2.7110000121036793e-05,
+      "logits/chosen": -11.5625,
+      "logits/rejected": -11.25,
+      "logps/chosen": -494.0,
+      "logps/rejected": -600.0,
+      "loss": 0.0866,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.9375,
+      "rewards/margins": 7.875,
+      "rewards/rejected": -16.75,
+      "step": 2010
+    },
+    {
+      "epoch": 1.0570381998953426,
+      "grad_norm": 3.108328168380246,
+      "learning_rate": 2.688235030572679e-05,
+      "logits/chosen": -11.9375,
+      "logits/rejected": -11.75,
+      "logps/chosen": -512.0,
+      "logps/rejected": -584.0,
+      "loss": 0.1356,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -9.3125,
+      "rewards/margins": 7.0625,
+      "rewards/rejected": -16.375,
+      "step": 2020
+    },
+    {
+      "epoch": 1.0622710622710623,
+      "grad_norm": 2.7373393989725763,
+      "learning_rate": 2.6654543406101833e-05,
+      "logits/chosen": -12.125,
+      "logits/rejected": -11.6875,
+      "logps/chosen": -520.0,
+      "logps/rejected": -656.0,
+      "loss": 0.1524,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -10.125,
+      "rewards/margins": 7.28125,
+      "rewards/rejected": -17.375,
+      "step": 2030
+    },
+    {
+      "epoch": 1.0675039246467817,
+      "grad_norm": 2.6186672086656926,
+      "learning_rate": 2.6426598432911763e-05,
+      "logits/chosen": -12.875,
+      "logits/rejected": -12.5,
+      "logps/chosen": -576.0,
+      "logps/rejected": -668.0,
+      "loss": 0.1237,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -9.25,
+      "rewards/margins": 7.53125,
+      "rewards/rejected": -16.75,
+      "step": 2040
+    },
+    {
+      "epoch": 1.0727367870225013,
+      "grad_norm": 1.4275150273517287,
+      "learning_rate": 2.6198534408428804e-05,
+      "logits/chosen": -12.75,
+      "logits/rejected": -12.375,
+      "logps/chosen": -532.0,
+      "logps/rejected": -628.0,
+      "loss": 0.0866,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.5,
+      "rewards/margins": 8.4375,
+      "rewards/rejected": -17.0,
+      "step": 2050
+    },
+    {
+      "epoch": 1.077969649398221,
+      "grad_norm": 4.626435577549045,
+      "learning_rate": 2.5970370364860176e-05,
+      "logits/chosen": -12.5625,
+      "logits/rejected": -12.25,
+      "logps/chosen": -478.0,
+      "logps/rejected": -636.0,
+      "loss": 0.1145,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -9.0625,
+      "rewards/margins": 8.625,
+      "rewards/rejected": -17.625,
+      "step": 2060
+    },
+    {
+      "epoch": 1.0832025117739403,
+      "grad_norm": 21.753848528929037,
+      "learning_rate": 2.574212534275978e-05,
+      "logits/chosen": -11.75,
+      "logits/rejected": -11.3125,
+      "logps/chosen": -464.0,
+      "logps/rejected": -552.0,
+      "loss": 0.1502,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.875,
+      "rewards/margins": 6.15625,
+      "rewards/rejected": -16.0,
+      "step": 2070
+    },
+    {
+      "epoch": 1.08843537414966,
+      "grad_norm": 2.4608658219211628,
+      "learning_rate": 2.5513818389439304e-05,
+      "logits/chosen": -12.1875,
+      "logits/rejected": -11.6875,
+      "logps/chosen": -556.0,
+      "logps/rejected": -656.0,
+      "loss": 0.1992,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -10.375,
+      "rewards/margins": 7.34375,
+      "rewards/rejected": -17.75,
+      "step": 2080
+    },
+    {
+      "epoch": 1.0936682365253794,
+      "grad_norm": 2.691085082609704,
+      "learning_rate": 2.5285468557378616e-05,
+      "logits/chosen": -11.75,
+      "logits/rejected": -11.5,
+      "logps/chosen": -476.0,
+      "logps/rejected": -608.0,
+      "loss": 0.1161,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -9.625,
+      "rewards/margins": 7.5625,
+      "rewards/rejected": -17.25,
+      "step": 2090
+    },
+    {
+      "epoch": 1.098901098901099,
+      "grad_norm": 2.6635752913927853,
+      "learning_rate": 2.5057094902635918e-05,
+      "logits/chosen": -11.75,
+      "logits/rejected": -11.4375,
+      "logps/chosen": -528.0,
+      "logps/rejected": -616.0,
+      "loss": 0.0969,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -9.8125,
+      "rewards/margins": 7.4375,
+      "rewards/rejected": -17.25,
+      "step": 2100
+    },
+    {
+      "epoch": 1.1041339612768184,
+      "grad_norm": 1.736617765978122,
+      "learning_rate": 2.4828716483257418e-05,
+      "logits/chosen": -12.25,
+      "logits/rejected": -11.625,
+      "logps/chosen": -548.0,
+      "logps/rejected": -636.0,
+      "loss": 0.1133,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -9.875,
+      "rewards/margins": 8.125,
+      "rewards/rejected": -18.0,
+      "step": 2110
+    },
+    {
+      "epoch": 1.109366823652538,
+      "grad_norm": 1.226880604327556,
+      "learning_rate": 2.460035235768692e-05,
+      "logits/chosen": -12.3125,
+      "logits/rejected": -12.0,
+      "logps/chosen": -500.0,
+      "logps/rejected": -644.0,
+      "loss": 0.1001,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -10.1875,
+      "rewards/margins": 8.0,
+      "rewards/rejected": -18.25,
+      "step": 2120
+    },
+    {
+      "epoch": 1.1145996860282574,
+      "grad_norm": 5.772340271804208,
+      "learning_rate": 2.4372021583175446e-05,
+      "logits/chosen": -12.25,
+      "logits/rejected": -12.125,
+      "logps/chosen": -536.0,
+      "logps/rejected": -568.0,
+      "loss": 0.1159,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -9.875,
+      "rewards/margins": 6.0,
+      "rewards/rejected": -15.875,
+      "step": 2130
+    },
+    {
+      "epoch": 1.119832548403977,
+      "grad_norm": 4.404511520229935,
+      "learning_rate": 2.4143743214190778e-05,
+      "logits/chosen": -12.8125,
+      "logits/rejected": -12.4375,
+      "logps/chosen": -510.0,
+      "logps/rejected": -640.0,
+      "loss": 0.1329,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -9.8125,
+      "rewards/margins": 7.09375,
+      "rewards/rejected": -16.875,
+      "step": 2140
+    },
+    {
+      "epoch": 1.1250654107796965,
+      "grad_norm": 2.5920340144560723,
+      "learning_rate": 2.3915536300827414e-05,
+      "logits/chosen": -13.375,
+      "logits/rejected": -13.3125,
+      "logps/chosen": -458.0,
+      "logps/rejected": -632.0,
+      "loss": 0.1462,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -9.1875,
+      "rewards/margins": 7.28125,
+      "rewards/rejected": -16.5,
+      "step": 2150
+    },
+    {
+      "epoch": 1.130298273155416,
+      "grad_norm": 1.9315423437420907,
+      "learning_rate": 2.3687419887216825e-05,
+      "logits/chosen": -13.25,
+      "logits/rejected": -12.9375,
+      "logps/chosen": -520.0,
+      "logps/rejected": -592.0,
+      "loss": 0.1172,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -9.25,
+      "rewards/margins": 6.5625,
+      "rewards/rejected": -15.8125,
+      "step": 2160
+    },
+    {
+      "epoch": 1.1355311355311355,
+      "grad_norm": 4.9387988320363085,
+      "learning_rate": 2.345941300993812e-05,
+      "logits/chosen": -13.625,
+      "logits/rejected": -13.25,
+      "logps/chosen": -524.0,
+      "logps/rejected": -628.0,
+      "loss": 0.1262,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -9.125,
+      "rewards/margins": 7.5,
+      "rewards/rejected": -16.625,
+      "step": 2170
+    },
+    {
+      "epoch": 1.1407639979068551,
+      "grad_norm": 2.6490159969730476,
+      "learning_rate": 2.3231534696429533e-05,
+      "logits/chosen": -13.125,
+      "logits/rejected": -12.625,
+      "logps/chosen": -500.0,
+      "logps/rejected": -640.0,
+      "loss": 0.0858,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -9.8125,
+      "rewards/margins": 7.40625,
+      "rewards/rejected": -17.25,
+      "step": 2180
+    },
+    {
+      "epoch": 1.1459968602825745,
+      "grad_norm": 1.8591406460035929,
+      "learning_rate": 2.3003803963400468e-05,
+      "logits/chosen": -13.625,
+      "logits/rejected": -13.25,
+      "logps/chosen": -504.0,
+      "logps/rejected": -632.0,
+      "loss": 0.1339,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -9.625,
+      "rewards/margins": 7.1875,
+      "rewards/rejected": -16.75,
+      "step": 2190
+    },
+    {
+      "epoch": 1.1512297226582942,
+      "grad_norm": 3.440279490788567,
+      "learning_rate": 2.2776239815244543e-05,
+      "logits/chosen": -12.75,
+      "logits/rejected": -12.5625,
+      "logps/chosen": -536.0,
+      "logps/rejected": -648.0,
+      "loss": 0.1023,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -8.8125,
+      "rewards/margins": 8.8125,
+      "rewards/rejected": -17.625,
+      "step": 2200
+    },
+    {
+      "epoch": 1.1564625850340136,
+      "grad_norm": 2.3561174482239413,
+      "learning_rate": 2.2548861242453742e-05,
+      "logits/chosen": -12.75,
+      "logits/rejected": -12.5,
+      "logps/chosen": -528.0,
+      "logps/rejected": -608.0,
+      "loss": 0.0903,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.8125,
+      "rewards/margins": 7.125,
+      "rewards/rejected": -16.875,
+      "step": 2210
+    },
+    {
+      "epoch": 1.1616954474097332,
+      "grad_norm": 3.129509901044873,
+      "learning_rate": 2.2321687220033523e-05,
+      "logits/chosen": -13.125,
+      "logits/rejected": -12.6875,
+      "logps/chosen": -498.0,
+      "logps/rejected": -652.0,
+      "loss": 0.1041,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -9.5,
+      "rewards/margins": 8.8125,
+      "rewards/rejected": -18.375,
+      "step": 2220
+    },
+    {
+      "epoch": 1.1669283097854526,
+      "grad_norm": 6.073604137970551,
+      "learning_rate": 2.2094736705919368e-05,
+      "logits/chosen": -13.0625,
+      "logits/rejected": -12.8125,
+      "logps/chosen": -568.0,
+      "logps/rejected": -672.0,
+      "loss": 0.1124,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -9.75,
+      "rewards/margins": 8.25,
+      "rewards/rejected": -18.0,
+      "step": 2230
+    },
+    {
+      "epoch": 1.1721611721611722,
+      "grad_norm": 4.837786173506918,
+      "learning_rate": 2.186802863939477e-05,
+      "logits/chosen": -12.9375,
+      "logits/rejected": -12.5,
+      "logps/chosen": -502.0,
+      "logps/rejected": -664.0,
+      "loss": 0.1104,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -10.625,
+      "rewards/margins": 8.625,
+      "rewards/rejected": -19.25,
+      "step": 2240
+    },
+    {
+      "epoch": 1.1773940345368916,
+      "grad_norm": 8.171951988640048,
+      "learning_rate": 2.1641581939510667e-05,
+      "logits/chosen": -13.4375,
+      "logits/rejected": -13.1875,
+      "logps/chosen": -552.0,
+      "logps/rejected": -672.0,
+      "loss": 0.0996,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -10.5,
+      "rewards/margins": 7.1875,
+      "rewards/rejected": -17.75,
+      "step": 2250
+    },
+    {
+      "epoch": 1.1826268969126112,
+      "grad_norm": 1.7116523821577394,
+      "learning_rate": 2.1415415503506653e-05,
+      "logits/chosen": -13.4375,
+      "logits/rejected": -13.1875,
+      "logps/chosen": -548.0,
+      "logps/rejected": -676.0,
+      "loss": 0.09,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -10.5625,
+      "rewards/margins": 7.9375,
+      "rewards/rejected": -18.5,
+      "step": 2260
+    },
+    {
+      "epoch": 1.1878597592883307,
+      "grad_norm": 2.8123145023686416,
+      "learning_rate": 2.1189548205233975e-05,
+      "logits/chosen": -13.625,
+      "logits/rejected": -13.25,
+      "logps/chosen": -584.0,
+      "logps/rejected": -728.0,
+      "loss": 0.0911,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -10.375,
+      "rewards/margins": 9.3125,
+      "rewards/rejected": -19.75,
+      "step": 2270
+    },
+    {
+      "epoch": 1.1930926216640503,
+      "grad_norm": 1.3758247246151873,
+      "learning_rate": 2.0963998893580487e-05,
+      "logits/chosen": -13.375,
+      "logits/rejected": -13.125,
+      "logps/chosen": -544.0,
+      "logps/rejected": -672.0,
+      "loss": 0.1232,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -9.875,
+      "rewards/margins": 9.0625,
+      "rewards/rejected": -19.0,
+      "step": 2280
+    },
+    {
+      "epoch": 1.1983254840397697,
+      "grad_norm": 2.8815572728674486,
+      "learning_rate": 2.0738786390897696e-05,
+      "logits/chosen": -13.75,
+      "logits/rejected": -13.4375,
+      "logps/chosen": -504.0,
+      "logps/rejected": -680.0,
+      "loss": 0.0962,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -9.625,
+      "rewards/margins": 8.875,
+      "rewards/rejected": -18.625,
+      "step": 2290
+    },
+    {
+      "epoch": 1.2035583464154893,
+      "grad_norm": 2.0943865882237183,
+      "learning_rate": 2.0513929491430006e-05,
+      "logits/chosen": -14.125,
+      "logits/rejected": -13.8125,
+      "logps/chosen": -516.0,
+      "logps/rejected": -660.0,
+      "loss": 0.0908,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -9.125,
+      "rewards/margins": 8.375,
+      "rewards/rejected": -17.5,
+      "step": 2300
+    },
+    {
+      "epoch": 1.2087912087912087,
+      "grad_norm": 4.687232667728419,
+      "learning_rate": 2.028944695974633e-05,
+      "logits/chosen": -14.375,
+      "logits/rejected": -13.8125,
+      "logps/chosen": -488.0,
+      "logps/rejected": -596.0,
+      "loss": 0.1219,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -9.5625,
+      "rewards/margins": 7.28125,
+      "rewards/rejected": -16.875,
+      "step": 2310
+    },
+    {
+      "epoch": 1.2140240711669283,
+      "grad_norm": 6.733912452591703,
+      "learning_rate": 2.006535752917414e-05,
+      "logits/chosen": -14.375,
+      "logits/rejected": -14.0,
+      "logps/chosen": -536.0,
+      "logps/rejected": -628.0,
+      "loss": 0.1353,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -9.75,
+      "rewards/margins": 7.6875,
+      "rewards/rejected": -17.375,
+      "step": 2320
+    },
+    {
+      "epoch": 1.2192569335426477,
+      "grad_norm": 2.3990026267001263,
+      "learning_rate": 1.9841679900236167e-05,
+      "logits/chosen": -13.6875,
+      "logits/rejected": -13.5625,
+      "logps/chosen": -528.0,
+      "logps/rejected": -624.0,
+      "loss": 0.1295,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -9.25,
+      "rewards/margins": 7.59375,
+      "rewards/rejected": -16.875,
+      "step": 2330
+    },
+    {
+      "epoch": 1.2244897959183674,
+      "grad_norm": 1.5565379549024234,
+      "learning_rate": 1.9618432739089843e-05,
+      "logits/chosen": -13.8125,
+      "logits/rejected": -13.5625,
+      "logps/chosen": -456.0,
+      "logps/rejected": -568.0,
+      "loss": 0.1056,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -8.5,
+      "rewards/margins": 7.28125,
+      "rewards/rejected": -15.8125,
+      "step": 2340
+    },
+    {
+      "epoch": 1.2297226582940868,
+      "grad_norm": 2.385154086116954,
+      "learning_rate": 1.9395634675969525e-05,
+      "logits/chosen": -13.75,
+      "logits/rejected": -13.3125,
+      "logps/chosen": -504.0,
+      "logps/rejected": -612.0,
+      "loss": 0.1503,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -9.125,
+      "rewards/margins": 7.46875,
+      "rewards/rejected": -16.5,
+      "step": 2350
+    },
+    {
+      "epoch": 1.2349555206698064,
+      "grad_norm": 0.7231966984051394,
+      "learning_rate": 1.9173304303631848e-05,
+      "logits/chosen": -13.125,
+      "logits/rejected": -12.875,
+      "logps/chosen": -504.0,
+      "logps/rejected": -608.0,
+      "loss": 0.0951,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.8125,
+      "rewards/margins": 7.46875,
+      "rewards/rejected": -16.25,
+      "step": 2360
+    },
+    {
+      "epoch": 1.2401883830455258,
+      "grad_norm": 3.595599348043387,
+      "learning_rate": 1.8951460175804104e-05,
+      "logits/chosen": -13.625,
+      "logits/rejected": -13.3125,
+      "logps/chosen": -548.0,
+      "logps/rejected": -648.0,
+      "loss": 0.1048,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -9.8125,
+      "rewards/margins": 7.875,
+      "rewards/rejected": -17.625,
+      "step": 2370
+    },
+    {
+      "epoch": 1.2454212454212454,
+      "grad_norm": 2.0288579413034777,
+      "learning_rate": 1.87301208056359e-05,
+      "logits/chosen": -12.625,
+      "logits/rejected": -12.375,
+      "logps/chosen": -512.0,
+      "logps/rejected": -616.0,
+      "loss": 0.104,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -9.375,
+      "rewards/margins": 7.90625,
+      "rewards/rejected": -17.25,
+      "step": 2380
+    },
+    {
+      "epoch": 1.250654107796965,
+      "grad_norm": 2.3526669326687464,
+      "learning_rate": 1.8509304664154255e-05,
+      "logits/chosen": -13.0625,
+      "logits/rejected": -12.625,
+      "logps/chosen": -604.0,
+      "logps/rejected": -716.0,
+      "loss": 0.1063,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -10.375,
+      "rewards/margins": 7.84375,
+      "rewards/rejected": -18.25,
+      "step": 2390
+    },
+    {
+      "epoch": 1.2558869701726845,
+      "grad_norm": 3.112985785125725,
+      "learning_rate": 1.8289030178722132e-05,
+      "logits/chosen": -13.0,
+      "logits/rejected": -12.5,
+      "logps/chosen": -540.0,
+      "logps/rejected": -676.0,
+      "loss": 0.1041,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -8.8125,
+      "rewards/margins": 9.625,
+      "rewards/rejected": -18.5,
+      "step": 2400
+    },
+    {
+      "epoch": 1.2611198325484039,
+      "grad_norm": 1.2581184168231971,
+      "learning_rate": 1.8069315731500666e-05,
+      "logits/chosen": -13.75,
+      "logits/rejected": -13.3125,
+      "logps/chosen": -576.0,
+      "logps/rejected": -692.0,
+      "loss": 0.1438,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.3125,
+      "rewards/margins": 8.5,
+      "rewards/rejected": -17.75,
+      "step": 2410
+    },
+    {
+      "epoch": 1.2663526949241235,
+      "grad_norm": 4.762221728293621,
+      "learning_rate": 1.7850179657915183e-05,
+      "logits/chosen": -13.25,
+      "logits/rejected": -12.875,
+      "logps/chosen": -600.0,
+      "logps/rejected": -616.0,
+      "loss": 0.1127,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -10.0,
+      "rewards/margins": 7.53125,
+      "rewards/rejected": -17.5,
+      "step": 2420
+    },
+    {
+      "epoch": 1.2715855572998431,
+      "grad_norm": 3.447175858247355,
+      "learning_rate": 1.7631640245125015e-05,
+      "logits/chosen": -12.625,
+      "logits/rejected": -12.125,
+      "logps/chosen": -548.0,
+      "logps/rejected": -624.0,
+      "loss": 0.1399,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -10.8125,
+      "rewards/margins": 7.4375,
+      "rewards/rejected": -18.25,
+      "step": 2430
+    },
+    {
+      "epoch": 1.2768184196755625,
+      "grad_norm": 4.061321817953982,
+      "learning_rate": 1.7413715730497494e-05,
+      "logits/chosen": -13.0625,
+      "logits/rejected": -12.875,
+      "logps/chosen": -516.0,
+      "logps/rejected": -668.0,
+      "loss": 0.0983,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -9.5,
+      "rewards/margins": 9.0625,
+      "rewards/rejected": -18.625,
+      "step": 2440
+    },
+    {
+      "epoch": 1.282051282051282,
+      "grad_norm": 4.489903011834727,
+      "learning_rate": 1.7196424300085978e-05,
+      "logits/chosen": -12.6875,
+      "logits/rejected": -12.375,
+      "logps/chosen": -520.0,
+      "logps/rejected": -604.0,
+      "loss": 0.0957,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -9.5625,
+      "rewards/margins": 6.75,
+      "rewards/rejected": -16.375,
+      "step": 2450
+    },
+    {
+      "epoch": 1.2872841444270016,
+      "grad_norm": 3.4346300831848433,
+      "learning_rate": 1.6979784087112188e-05,
+      "logits/chosen": -12.75,
+      "logits/rejected": -12.5,
+      "logps/chosen": -468.0,
+      "logps/rejected": -572.0,
+      "loss": 0.1555,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.4375,
+      "rewards/margins": 7.03125,
+      "rewards/rejected": -16.5,
+      "step": 2460
+    },
+    {
+      "epoch": 1.2925170068027212,
+      "grad_norm": 2.4661294584896933,
+      "learning_rate": 1.6763813170453044e-05,
+      "logits/chosen": -13.0,
+      "logits/rejected": -12.5625,
+      "logps/chosen": -474.0,
+      "logps/rejected": -608.0,
+      "loss": 0.0976,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.0,
+      "rewards/margins": 7.09375,
+      "rewards/rejected": -16.0,
+      "step": 2470
+    },
+    {
+      "epoch": 1.2977498691784406,
+      "grad_norm": 3.149223335476965,
+      "learning_rate": 1.6548529573131876e-05,
+      "logits/chosen": -13.125,
+      "logits/rejected": -12.625,
+      "logps/chosen": -510.0,
+      "logps/rejected": -608.0,
+      "loss": 0.1061,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -9.625,
+      "rewards/margins": 7.21875,
+      "rewards/rejected": -16.875,
+      "step": 2480
+    },
+    {
+      "epoch": 1.30298273155416,
+      "grad_norm": 2.862506339722175,
+      "learning_rate": 1.6333951260814413e-05,
+      "logits/chosen": -13.25,
+      "logits/rejected": -12.875,
+      "logps/chosen": -592.0,
+      "logps/rejected": -640.0,
+      "loss": 0.0693,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -9.3125,
+      "rewards/margins": 8.0625,
+      "rewards/rejected": -17.375,
+      "step": 2490
+    },
+    {
+      "epoch": 1.3082155939298796,
+      "grad_norm": 2.7248445532753016,
+      "learning_rate": 1.6120096140309572e-05,
+      "logits/chosen": -13.125,
+      "logits/rejected": -12.8125,
+      "logps/chosen": -460.0,
+      "logps/rejected": -592.0,
+      "loss": 0.1068,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.875,
+      "rewards/margins": 7.25,
+      "rewards/rejected": -16.125,
+      "step": 2500
+    },
+    {
+      "epoch": 1.3134484563055993,
+      "grad_norm": 1.4186556225169547,
+      "learning_rate": 1.5906982058075038e-05,
+      "logits/chosen": -12.6875,
+      "logits/rejected": -12.3125,
+      "logps/chosen": -478.0,
+      "logps/rejected": -656.0,
+      "loss": 0.0886,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -9.625,
+      "rewards/margins": 7.5,
+      "rewards/rejected": -17.125,
+      "step": 2510
+    },
+    {
+      "epoch": 1.3186813186813187,
+      "grad_norm": 2.9852405667865174,
+      "learning_rate": 1.569462679872801e-05,
+      "logits/chosen": -12.8125,
+      "logits/rejected": -12.5625,
+      "logps/chosen": -528.0,
+      "logps/rejected": -600.0,
+      "loss": 0.1124,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -10.125,
+      "rewards/margins": 6.9375,
+      "rewards/rejected": -17.0,
+      "step": 2520
+    },
+    {
+      "epoch": 1.323914181057038,
+      "grad_norm": 3.666318609283611,
+      "learning_rate": 1.5483048083561036e-05,
+      "logits/chosen": -13.5625,
+      "logits/rejected": -13.0,
+      "logps/chosen": -540.0,
+      "logps/rejected": -676.0,
+      "loss": 0.1064,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -9.1875,
+      "rewards/margins": 9.3125,
+      "rewards/rejected": -18.5,
+      "step": 2530
+    },
+    {
+      "epoch": 1.3291470434327577,
+      "grad_norm": 2.9887842389088215,
+      "learning_rate": 1.527226356906314e-05,
+      "logits/chosen": -12.9375,
+      "logits/rejected": -12.625,
+      "logps/chosen": -458.0,
+      "logps/rejected": -588.0,
+      "loss": 0.1584,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -9.0625,
+      "rewards/margins": 7.3125,
+      "rewards/rejected": -16.375,
+      "step": 2540
+    },
+    {
+      "epoch": 1.3343799058084773,
+      "grad_norm": 4.22908382077193,
+      "learning_rate": 1.5062290845446403e-05,
+      "logits/chosen": -12.625,
+      "logits/rejected": -11.9375,
+      "logps/chosen": -540.0,
+      "logps/rejected": -636.0,
+      "loss": 0.1073,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -10.75,
+      "rewards/margins": 7.25,
+      "rewards/rejected": -18.0,
+      "step": 2550
+    },
+    {
+      "epoch": 1.3396127681841967,
+      "grad_norm": 2.546115179849579,
+      "learning_rate": 1.4853147435177992e-05,
+      "logits/chosen": -12.625,
+      "logits/rejected": -12.0625,
+      "logps/chosen": -524.0,
+      "logps/rejected": -632.0,
+      "loss": 0.1146,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -9.375,
+      "rewards/margins": 8.75,
+      "rewards/rejected": -18.125,
+      "step": 2560
+    },
+    {
+      "epoch": 1.3448456305599163,
+      "grad_norm": 2.964138623891713,
+      "learning_rate": 1.4644850791517933e-05,
+      "logits/chosen": -12.875,
+      "logits/rejected": -12.625,
+      "logps/chosen": -528.0,
+      "logps/rejected": -640.0,
+      "loss": 0.0923,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -9.125,
+      "rewards/margins": 8.1875,
+      "rewards/rejected": -17.25,
+      "step": 2570
+    },
+    {
+      "epoch": 1.3500784929356358,
+      "grad_norm": 2.833859042991735,
+      "learning_rate": 1.4437418297062589e-05,
+      "logits/chosen": -12.1875,
+      "logits/rejected": -12.0625,
+      "logps/chosen": -504.0,
+      "logps/rejected": -624.0,
+      "loss": 0.1154,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.0,
+      "rewards/margins": 7.875,
+      "rewards/rejected": -16.875,
+      "step": 2580
+    },
+    {
+      "epoch": 1.3553113553113554,
+      "grad_norm": 4.611332803169445,
+      "learning_rate": 1.4230867262294045e-05,
+      "logits/chosen": -13.0,
+      "logits/rejected": -12.5625,
+      "logps/chosen": -532.0,
+      "logps/rejected": -700.0,
+      "loss": 0.1234,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -8.75,
+      "rewards/margins": 8.75,
+      "rewards/rejected": -17.5,
+      "step": 2590
+    },
+    {
+      "epoch": 1.3605442176870748,
+      "grad_norm": 2.585933332197896,
+      "learning_rate": 1.4025214924135616e-05,
+      "logits/chosen": -12.5,
+      "logits/rejected": -12.0,
+      "logps/chosen": -464.0,
+      "logps/rejected": -584.0,
+      "loss": 0.1002,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -9.0625,
+      "rewards/margins": 6.875,
+      "rewards/rejected": -15.9375,
+      "step": 2600
+    },
+    {
+      "epoch": 1.3657770800627944,
+      "grad_norm": 5.004771837552434,
+      "learning_rate": 1.3820478444513288e-05,
+      "logits/chosen": -12.9375,
+      "logits/rejected": -12.25,
+      "logps/chosen": -540.0,
+      "logps/rejected": -636.0,
+      "loss": 0.1361,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -9.3125,
+      "rewards/margins": 7.71875,
+      "rewards/rejected": -17.0,
+      "step": 2610
+    },
+    {
+      "epoch": 1.3710099424385138,
+      "grad_norm": 3.3721942115741483,
+      "learning_rate": 1.3616674908923585e-05,
+      "logits/chosen": -12.6875,
+      "logits/rejected": -12.375,
+      "logps/chosen": -502.0,
+      "logps/rejected": -604.0,
+      "loss": 0.1074,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.0625,
+      "rewards/margins": 7.75,
+      "rewards/rejected": -16.875,
+      "step": 2620
+    },
+    {
+      "epoch": 1.3762428048142334,
+      "grad_norm": 1.926096891230754,
+      "learning_rate": 1.3413821325007834e-05,
+      "logits/chosen": -12.375,
+      "logits/rejected": -12.125,
+      "logps/chosen": -460.0,
+      "logps/rejected": -608.0,
+      "loss": 0.0734,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -9.4375,
+      "rewards/margins": 7.25,
+      "rewards/rejected": -16.75,
+      "step": 2630
+    },
+    {
+      "epoch": 1.3814756671899528,
+      "grad_norm": 2.6811569703958438,
+      "learning_rate": 1.321193462113272e-05,
+      "logits/chosen": -12.125,
+      "logits/rejected": -11.5,
+      "logps/chosen": -500.0,
+      "logps/rejected": -584.0,
+      "loss": 0.1341,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.3125,
+      "rewards/margins": 6.65625,
+      "rewards/rejected": -16.0,
+      "step": 2640
+    },
+    {
+      "epoch": 1.3867085295656725,
+      "grad_norm": 1.5488117167812219,
+      "learning_rate": 1.3011031644977716e-05,
+      "logits/chosen": -12.5,
+      "logits/rejected": -12.0,
+      "logps/chosen": -470.0,
+      "logps/rejected": -636.0,
+      "loss": 0.1135,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.5,
+      "rewards/margins": 7.875,
+      "rewards/rejected": -17.375,
+      "step": 2650
+    },
+    {
+      "epoch": 1.3919413919413919,
+      "grad_norm": 5.7062514391226555,
+      "learning_rate": 1.2811129162129065e-05,
+      "logits/chosen": -12.375,
+      "logits/rejected": -12.125,
+      "logps/chosen": -548.0,
+      "logps/rejected": -740.0,
+      "loss": 0.0949,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -9.375,
+      "rewards/margins": 9.0625,
+      "rewards/rejected": -18.5,
+      "step": 2660
+    },
+    {
+      "epoch": 1.3971742543171115,
+      "grad_norm": 1.7643119117081274,
+      "learning_rate": 1.261224385468066e-05,
+      "logits/chosen": -12.3125,
+      "logits/rejected": -11.3125,
+      "logps/chosen": -528.0,
+      "logps/rejected": -604.0,
+      "loss": 0.0787,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -10.25,
+      "rewards/margins": 7.0,
+      "rewards/rejected": -17.25,
+      "step": 2670
+    },
+    {
+      "epoch": 1.402407116692831,
+      "grad_norm": 4.394449609840489,
+      "learning_rate": 1.2414392319841957e-05,
+      "logits/chosen": -12.3125,
+      "logits/rejected": -11.875,
+      "logps/chosen": -548.0,
+      "logps/rejected": -724.0,
+      "loss": 0.1152,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.1875,
+      "rewards/margins": 11.1875,
+      "rewards/rejected": -20.375,
+      "step": 2680
+    },
+    {
+      "epoch": 1.4076399790685505,
+      "grad_norm": 5.0757669192975685,
+      "learning_rate": 1.2217591068552894e-05,
+      "logits/chosen": -12.625,
+      "logits/rejected": -12.1875,
+      "logps/chosen": -456.0,
+      "logps/rejected": -632.0,
+      "loss": 0.0827,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -9.25,
+      "rewards/margins": 7.90625,
+      "rewards/rejected": -17.125,
+      "step": 2690
+    },
+    {
+      "epoch": 1.41287284144427,
+      "grad_norm": 1.2128787751321914,
+      "learning_rate": 1.2021856524105992e-05,
+      "logits/chosen": -13.0,
+      "logits/rejected": -12.375,
+      "logps/chosen": -556.0,
+      "logps/rejected": -652.0,
+      "loss": 0.0657,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -9.5,
+      "rewards/margins": 8.75,
+      "rewards/rejected": -18.25,
+      "step": 2700
+    },
+    {
+      "epoch": 1.4181057038199896,
+      "grad_norm": 4.371148246096117,
+      "learning_rate": 1.1827205020775881e-05,
+      "logits/chosen": -12.5,
+      "logits/rejected": -12.125,
+      "logps/chosen": -528.0,
+      "logps/rejected": -624.0,
+      "loss": 0.1347,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.25,
+      "rewards/margins": 8.0,
+      "rewards/rejected": -17.25,
+      "step": 2710
+    },
+    {
+      "epoch": 1.423338566195709,
+      "grad_norm": 1.6427572152583667,
+      "learning_rate": 1.163365280245615e-05,
+      "logits/chosen": -13.125,
+      "logits/rejected": -12.6875,
+      "logps/chosen": -480.0,
+      "logps/rejected": -592.0,
+      "loss": 0.1309,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -8.5,
+      "rewards/margins": 7.34375,
+      "rewards/rejected": -15.875,
+      "step": 2720
+    },
+    {
+      "epoch": 1.4285714285714286,
+      "grad_norm": 6.792621163838377,
+      "learning_rate": 1.1441216021303777e-05,
+      "logits/chosen": -13.125,
+      "logits/rejected": -12.6875,
+      "logps/chosen": -494.0,
+      "logps/rejected": -608.0,
+      "loss": 0.1526,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -9.1875,
+      "rewards/margins": 7.3125,
+      "rewards/rejected": -16.5,
+      "step": 2730
+    },
+    {
+      "epoch": 1.433804290947148,
+      "grad_norm": 7.768597274251902,
+      "learning_rate": 1.1249910736391203e-05,
+      "logits/chosen": -13.0,
+      "logits/rejected": -12.625,
+      "logps/chosen": -496.0,
+      "logps/rejected": -616.0,
+      "loss": 0.1477,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -9.0625,
+      "rewards/margins": 8.0,
+      "rewards/rejected": -17.125,
+      "step": 2740
+    },
+    {
+      "epoch": 1.4390371533228676,
+      "grad_norm": 1.5634511255756072,
+      "learning_rate": 1.1059752912366217e-05,
+      "logits/chosen": -13.1875,
+      "logits/rejected": -12.75,
+      "logps/chosen": -462.0,
+      "logps/rejected": -660.0,
+      "loss": 0.1114,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -8.375,
+      "rewards/margins": 8.75,
+      "rewards/rejected": -17.125,
+      "step": 2750
+    },
+    {
+      "epoch": 1.4442700156985873,
+      "grad_norm": 5.473079860013893,
+      "learning_rate": 1.0870758418119659e-05,
+      "logits/chosen": -13.1875,
+      "logits/rejected": -12.6875,
+      "logps/chosen": -540.0,
+      "logps/rejected": -584.0,
+      "loss": 0.0834,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -9.3125,
+      "rewards/margins": 7.875,
+      "rewards/rejected": -17.25,
+      "step": 2760
+    },
+    {
+      "epoch": 1.4495028780743067,
+      "grad_norm": 3.4164753814709665,
+      "learning_rate": 1.0682943025461136e-05,
+      "logits/chosen": -13.3125,
+      "logits/rejected": -13.0,
+      "logps/chosen": -544.0,
+      "logps/rejected": -628.0,
+      "loss": 0.1494,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -9.25,
+      "rewards/margins": 7.4375,
+      "rewards/rejected": -16.625,
+      "step": 2770
+    },
+    {
+      "epoch": 1.454735740450026,
+      "grad_norm": 4.0004197012792835,
+      "learning_rate": 1.049632240780288e-05,
+      "logits/chosen": -12.625,
+      "logits/rejected": -11.875,
+      "logps/chosen": -486.0,
+      "logps/rejected": -556.0,
+      "loss": 0.1041,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.1875,
+      "rewards/margins": 6.46875,
+      "rewards/rejected": -15.6875,
+      "step": 2780
+    },
+    {
+      "epoch": 1.4599686028257457,
+      "grad_norm": 2.937844502391212,
+      "learning_rate": 1.0310912138851769e-05,
+      "logits/chosen": -12.9375,
+      "logits/rejected": -12.4375,
+      "logps/chosen": -520.0,
+      "logps/rejected": -688.0,
+      "loss": 0.0971,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -9.5625,
+      "rewards/margins": 8.0625,
+      "rewards/rejected": -17.625,
+      "step": 2790
+    },
+    {
+      "epoch": 1.4652014652014653,
+      "grad_norm": 4.339889798014685,
+      "learning_rate": 1.0126727691309638e-05,
+      "logits/chosen": -13.125,
+      "logits/rejected": -12.9375,
+      "logps/chosen": -552.0,
+      "logps/rejected": -704.0,
+      "loss": 0.1164,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -9.375,
+      "rewards/margins": 8.75,
+      "rewards/rejected": -18.125,
+      "step": 2800
+    },
+    {
+      "epoch": 1.4704343275771847,
+      "grad_norm": 1.8102160895635062,
+      "learning_rate": 9.943784435582166e-06,
+      "logits/chosen": -12.9375,
+      "logits/rejected": -12.25,
+      "logps/chosen": -516.0,
+      "logps/rejected": -660.0,
+      "loss": 0.0842,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.3125,
+      "rewards/margins": 8.1875,
+      "rewards/rejected": -17.5,
+      "step": 2810
+    },
+    {
+      "epoch": 1.4756671899529041,
+      "grad_norm": 1.7389222823704746,
+      "learning_rate": 9.76209763849609e-06,
+      "logits/chosen": -12.5625,
+      "logits/rejected": -11.9375,
+      "logps/chosen": -478.0,
+      "logps/rejected": -620.0,
+      "loss": 0.0995,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -9.0,
+      "rewards/margins": 8.25,
+      "rewards/rejected": -17.25,
+      "step": 2820
+    },
+    {
+      "epoch": 1.4809000523286238,
+      "grad_norm": 1.3162427246282047,
+      "learning_rate": 9.581682462025215e-06,
+      "logits/chosen": -12.9375,
+      "logits/rejected": -12.875,
+      "logps/chosen": -494.0,
+      "logps/rejected": -624.0,
+      "loss": 0.0881,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -9.0625,
+      "rewards/margins": 7.9375,
+      "rewards/rejected": -17.0,
+      "step": 2830
+    },
+    {
+      "epoch": 1.4861329147043434,
+      "grad_norm": 2.015607418663913,
+      "learning_rate": 9.40255396202518e-06,
+      "logits/chosen": -12.75,
+      "logits/rejected": -12.5,
+      "logps/chosen": -568.0,
+      "logps/rejected": -660.0,
+      "loss": 0.1589,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.4375,
+      "rewards/margins": 7.84375,
+      "rewards/rejected": -17.25,
+      "step": 2840
+    },
+    {
+      "epoch": 1.4913657770800628,
+      "grad_norm": 1.0089278154821486,
+      "learning_rate": 9.22472708697692e-06,
+      "logits/chosen": -12.9375,
+      "logits/rejected": -12.3125,
+      "logps/chosen": -516.0,
+      "logps/rejected": -624.0,
+      "loss": 0.1199,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.9375,
+      "rewards/margins": 8.3125,
+      "rewards/rejected": -17.25,
+      "step": 2850
+    },
+    {
+      "epoch": 1.4965986394557822,
+      "grad_norm": 7.8442767175126145,
+      "learning_rate": 9.048216676739295e-06,
+      "logits/chosen": -13.0625,
+      "logits/rejected": -12.875,
+      "logps/chosen": -528.0,
+      "logps/rejected": -716.0,
+      "loss": 0.1101,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.4375,
+      "rewards/margins": 10.375,
+      "rewards/rejected": -18.875,
+      "step": 2860
+    },
+    {
+      "epoch": 1.5018315018315018,
+      "grad_norm": 0.8512762475654531,
+      "learning_rate": 8.87303746131066e-06,
+      "logits/chosen": -12.75,
+      "logits/rejected": -12.125,
+      "logps/chosen": -536.0,
+      "logps/rejected": -672.0,
+      "loss": 0.1099,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.375,
+      "rewards/margins": 8.5625,
+      "rewards/rejected": -18.0,
+      "step": 2870
+    },
+    {
+      "epoch": 1.5070643642072215,
+      "grad_norm": 1.5593140512431443,
+      "learning_rate": 8.699204059599578e-06,
+      "logits/chosen": -12.6875,
+      "logits/rejected": -12.375,
+      "logps/chosen": -536.0,
+      "logps/rejected": -660.0,
+      "loss": 0.1107,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -9.5,
+      "rewards/margins": 8.625,
+      "rewards/rejected": -18.125,
+      "step": 2880
+    },
+    {
+      "epoch": 1.5122972265829409,
+      "grad_norm": 3.0135057825122633,
+      "learning_rate": 8.526730978204933e-06,
+      "logits/chosen": -12.5,
+      "logits/rejected": -12.3125,
+      "logps/chosen": -572.0,
+      "logps/rejected": -680.0,
+      "loss": 0.097,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -9.625,
+      "rewards/margins": 9.3125,
+      "rewards/rejected": -18.875,
+      "step": 2890
+    },
+    {
+      "epoch": 1.5175300889586603,
+      "grad_norm": 4.884151124038764,
+      "learning_rate": 8.35563261020529e-06,
+      "logits/chosen": -13.125,
+      "logits/rejected": -12.6875,
+      "logps/chosen": -548.0,
+      "logps/rejected": -728.0,
+      "loss": 0.0774,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.9375,
+      "rewards/margins": 10.75,
+      "rewards/rejected": -19.75,
+      "step": 2900
+    },
+    {
+      "epoch": 1.5227629513343799,
+      "grad_norm": 5.5540339317729295,
+      "learning_rate": 8.185923233957802e-06,
+      "logits/chosen": -12.9375,
+      "logits/rejected": -12.125,
+      "logps/chosen": -486.0,
+      "logps/rejected": -652.0,
+      "loss": 0.0961,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.1875,
+      "rewards/margins": 9.0,
+      "rewards/rejected": -18.125,
+      "step": 2910
+    },
+    {
+      "epoch": 1.5279958137100995,
+      "grad_norm": 3.6558685861938636,
+      "learning_rate": 8.017617011906618e-06,
+      "logits/chosen": -12.75,
+      "logits/rejected": -12.3125,
+      "logps/chosen": -548.0,
+      "logps/rejected": -704.0,
+      "loss": 0.08,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -9.75,
+      "rewards/margins": 7.8125,
+      "rewards/rejected": -17.5,
+      "step": 2920
+    },
+    {
+      "epoch": 1.533228676085819,
+      "grad_norm": 3.6345000092419517,
+      "learning_rate": 7.850727989401064e-06,
+      "logits/chosen": -13.0,
+      "logits/rejected": -12.75,
+      "logps/chosen": -568.0,
+      "logps/rejected": -740.0,
+      "loss": 0.0925,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.625,
+      "rewards/margins": 8.8125,
+      "rewards/rejected": -18.375,
+      "step": 2930
+    },
+    {
+      "epoch": 1.5384615384615383,
+      "grad_norm": 1.68599240926016,
+      "learning_rate": 7.685270093523534e-06,
+      "logits/chosen": -13.125,
+      "logits/rejected": -12.625,
+      "logps/chosen": -532.0,
+      "logps/rejected": -704.0,
+      "loss": 0.0917,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -9.0,
+      "rewards/margins": 10.125,
+      "rewards/rejected": -19.125,
+      "step": 2940
+    },
+    {
+      "epoch": 1.543694400837258,
+      "grad_norm": 4.5784671678935895,
+      "learning_rate": 7.521257131927212e-06,
+      "logits/chosen": -12.6875,
+      "logits/rejected": -12.375,
+      "logps/chosen": -508.0,
+      "logps/rejected": -572.0,
+      "loss": 0.1385,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -10.0,
+      "rewards/margins": 7.28125,
+      "rewards/rejected": -17.25,
+      "step": 2950
+    },
+    {
+      "epoch": 1.5489272632129776,
+      "grad_norm": 3.1832934865765763,
+      "learning_rate": 7.358702791683869e-06,
+      "logits/chosen": -13.25,
+      "logits/rejected": -12.875,
+      "logps/chosen": -478.0,
+      "logps/rejected": -600.0,
+      "loss": 0.0871,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -9.0625,
+      "rewards/margins": 8.0625,
+      "rewards/rejected": -17.125,
+      "step": 2960
+    },
+    {
+      "epoch": 1.554160125588697,
+      "grad_norm": 5.1042748038245565,
+      "learning_rate": 7.197620638141633e-06,
+      "logits/chosen": -13.3125,
+      "logits/rejected": -13.1875,
+      "logps/chosen": -494.0,
+      "logps/rejected": -616.0,
+      "loss": 0.0872,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.625,
+      "rewards/margins": 8.4375,
+      "rewards/rejected": -17.125,
+      "step": 2970
+    },
+    {
+      "epoch": 1.5593929879644164,
+      "grad_norm": 4.261194324303468,
+      "learning_rate": 7.038024113792921e-06,
+      "logits/chosen": -13.5,
+      "logits/rejected": -12.875,
+      "logps/chosen": -544.0,
+      "logps/rejected": -660.0,
+      "loss": 0.0934,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -8.75,
+      "rewards/margins": 8.75,
+      "rewards/rejected": -17.5,
+      "step": 2980
+    },
+    {
+      "epoch": 1.564625850340136,
+      "grad_norm": 5.838668015412529,
+      "learning_rate": 6.879926537152695e-06,
+      "logits/chosen": -12.9375,
+      "logits/rejected": -12.375,
+      "logps/chosen": -478.0,
+      "logps/rejected": -684.0,
+      "loss": 0.1091,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -8.625,
+      "rewards/margins": 9.75,
+      "rewards/rejected": -18.375,
+      "step": 2990
+    },
+    {
+      "epoch": 1.5698587127158556,
+      "grad_norm": 6.215751805578572,
+      "learning_rate": 6.723341101646993e-06,
+      "logits/chosen": -13.5,
+      "logits/rejected": -13.1875,
+      "logps/chosen": -544.0,
+      "logps/rejected": -680.0,
+      "loss": 0.0991,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.375,
+      "rewards/margins": 9.3125,
+      "rewards/rejected": -17.75,
+      "step": 3000
+    },
+    {
+      "epoch": 1.575091575091575,
+      "grad_norm": 6.288421625487859,
+      "learning_rate": 6.568280874511904e-06,
+      "logits/chosen": -13.375,
+      "logits/rejected": -13.0,
+      "logps/chosen": -544.0,
+      "logps/rejected": -704.0,
+      "loss": 0.0995,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.75,
+      "rewards/margins": 9.1875,
+      "rewards/rejected": -17.875,
+      "step": 3010
+    },
+    {
+      "epoch": 1.5803244374672945,
+      "grad_norm": 7.0546142777757845,
+      "learning_rate": 6.414758795703122e-06,
+      "logits/chosen": -13.375,
+      "logits/rejected": -12.9375,
+      "logps/chosen": -520.0,
+      "logps/rejected": -624.0,
+      "loss": 0.1143,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -9.125,
+      "rewards/margins": 7.59375,
+      "rewards/rejected": -16.75,
+      "step": 3020
+    },
+    {
+      "epoch": 1.585557299843014,
+      "grad_norm": 4.589578434412954,
+      "learning_rate": 6.262787676816093e-06,
+      "logits/chosen": -13.0625,
+      "logits/rejected": -12.625,
+      "logps/chosen": -552.0,
+      "logps/rejected": -648.0,
+      "loss": 0.1142,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -9.5,
+      "rewards/margins": 7.9375,
+      "rewards/rejected": -17.375,
+      "step": 3030
+    },
+    {
+      "epoch": 1.5907901622187337,
+      "grad_norm": 4.762993940765846,
+      "learning_rate": 6.112380200016832e-06,
+      "logits/chosen": -13.6875,
+      "logits/rejected": -13.1875,
+      "logps/chosen": -496.0,
+      "logps/rejected": -640.0,
+      "loss": 0.1374,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -8.625,
+      "rewards/margins": 7.875,
+      "rewards/rejected": -16.5,
+      "step": 3040
+    },
+    {
+      "epoch": 1.5960230245944533,
+      "grad_norm": 5.107564915010418,
+      "learning_rate": 5.963548916983627e-06,
+      "logits/chosen": -12.875,
+      "logits/rejected": -12.375,
+      "logps/chosen": -452.0,
+      "logps/rejected": -552.0,
+      "loss": 0.0917,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.5,
+      "rewards/margins": 7.71875,
+      "rewards/rejected": -16.25,
+      "step": 3050
+    },
+    {
+      "epoch": 1.6012558869701727,
+      "grad_norm": 2.54635356048562,
+      "learning_rate": 5.816306247859571e-06,
+      "logits/chosen": -13.3125,
+      "logits/rejected": -12.875,
+      "logps/chosen": -552.0,
+      "logps/rejected": -640.0,
+      "loss": 0.0855,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.1875,
+      "rewards/margins": 9.375,
+      "rewards/rejected": -17.5,
+      "step": 3060
+    },
+    {
+      "epoch": 1.6064887493458921,
+      "grad_norm": 3.298880234845822,
+      "learning_rate": 5.670664480216087e-06,
+      "logits/chosen": -13.25,
+      "logits/rejected": -12.75,
+      "logps/chosen": -516.0,
+      "logps/rejected": -676.0,
+      "loss": 0.0884,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.0,
+      "rewards/margins": 9.8125,
+      "rewards/rejected": -17.75,
+      "step": 3070
+    },
+    {
+      "epoch": 1.6117216117216118,
+      "grad_norm": 4.353379277791028,
+      "learning_rate": 5.526635768027489e-06,
+      "logits/chosen": -13.25,
+      "logits/rejected": -12.625,
+      "logps/chosen": -528.0,
+      "logps/rejected": -620.0,
+      "loss": 0.1106,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.5,
+      "rewards/margins": 8.875,
+      "rewards/rejected": -17.375,
+      "step": 3080
+    },
+    {
+      "epoch": 1.6169544740973314,
+      "grad_norm": 1.7350907393275037,
+      "learning_rate": 5.384232130656772e-06,
+      "logits/chosen": -12.875,
+      "logits/rejected": -12.5,
+      "logps/chosen": -536.0,
+      "logps/rejected": -704.0,
+      "loss": 0.0954,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.25,
+      "rewards/margins": 9.875,
+      "rewards/rejected": -18.125,
+      "step": 3090
+    },
+    {
+      "epoch": 1.6221873364730508,
+      "grad_norm": 1.9702905557486132,
+      "learning_rate": 5.243465451852547e-06,
+      "logits/chosen": -12.9375,
+      "logits/rejected": -12.25,
+      "logps/chosen": -512.0,
+      "logps/rejected": -660.0,
+      "loss": 0.1501,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -8.5625,
+      "rewards/margins": 8.625,
+      "rewards/rejected": -17.25,
+      "step": 3100
+    },
+    {
+      "epoch": 1.6274201988487702,
+      "grad_norm": 2.119534360879486,
+      "learning_rate": 5.104347478757313e-06,
+      "logits/chosen": -12.8125,
+      "logits/rejected": -12.25,
+      "logps/chosen": -480.0,
+      "logps/rejected": -636.0,
+      "loss": 0.1065,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -8.75,
+      "rewards/margins": 8.3125,
+      "rewards/rejected": -17.125,
+      "step": 3110
+    },
+    {
+      "epoch": 1.6326530612244898,
+      "grad_norm": 5.305348928368352,
+      "learning_rate": 4.9668898209272094e-06,
+      "logits/chosen": -13.625,
+      "logits/rejected": -13.0625,
+      "logps/chosen": -544.0,
+      "logps/rejected": -648.0,
+      "loss": 0.1039,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.0,
+      "rewards/margins": 9.6875,
+      "rewards/rejected": -17.75,
+      "step": 3120
+    },
+    {
+      "epoch": 1.6378859236002095,
+      "grad_norm": 3.865588511769586,
+      "learning_rate": 4.831103949363103e-06,
+      "logits/chosen": -13.0,
+      "logits/rejected": -12.4375,
+      "logps/chosen": -486.0,
+      "logps/rejected": -576.0,
+      "loss": 0.1176,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.1875,
+      "rewards/margins": 7.9375,
+      "rewards/rejected": -16.125,
+      "step": 3130
+    },
+    {
+      "epoch": 1.6431187859759289,
+      "grad_norm": 1.1087995031281264,
+      "learning_rate": 4.697001195553366e-06,
+      "logits/chosen": -13.3125,
+      "logits/rejected": -12.875,
+      "logps/chosen": -476.0,
+      "logps/rejected": -628.0,
+      "loss": 0.0754,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -9.0625,
+      "rewards/margins": 8.5,
+      "rewards/rejected": -17.5,
+      "step": 3140
+    },
+    {
+      "epoch": 1.6483516483516483,
+      "grad_norm": 9.607813501681845,
+      "learning_rate": 4.564592750528271e-06,
+      "logits/chosen": -13.25,
+      "logits/rejected": -12.75,
+      "logps/chosen": -506.0,
+      "logps/rejected": -584.0,
+      "loss": 0.1084,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.9375,
+      "rewards/margins": 7.5625,
+      "rewards/rejected": -16.5,
+      "step": 3150
+    },
+    {
+      "epoch": 1.653584510727368,
+      "grad_norm": 2.166356498800773,
+      "learning_rate": 4.4338896639260276e-06,
+      "logits/chosen": -13.4375,
+      "logits/rejected": -13.0,
+      "logps/chosen": -508.0,
+      "logps/rejected": -632.0,
+      "loss": 0.0882,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -8.375,
+      "rewards/margins": 8.0,
+      "rewards/rejected": -16.375,
+      "step": 3160
+    },
+    {
+      "epoch": 1.6588173731030875,
+      "grad_norm": 1.3202863612431788,
+      "learning_rate": 4.304902843070701e-06,
+      "logits/chosen": -13.375,
+      "logits/rejected": -12.75,
+      "logps/chosen": -494.0,
+      "logps/rejected": -580.0,
+      "loss": 0.0608,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -8.5,
+      "rewards/margins": 7.5625,
+      "rewards/rejected": -16.0,
+      "step": 3170
+    },
+    {
+      "epoch": 1.664050235478807,
+      "grad_norm": 6.022229933522954,
+      "learning_rate": 4.177643052062039e-06,
+      "logits/chosen": -13.375,
+      "logits/rejected": -12.75,
+      "logps/chosen": -510.0,
+      "logps/rejected": -656.0,
+      "loss": 0.0964,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -8.9375,
+      "rewards/margins": 8.0,
+      "rewards/rejected": -16.875,
+      "step": 3180
+    },
+    {
+      "epoch": 1.6692830978545263,
+      "grad_norm": 4.697489562167333,
+      "learning_rate": 4.0521209108770945e-06,
+      "logits/chosen": -13.25,
+      "logits/rejected": -12.625,
+      "logps/chosen": -506.0,
+      "logps/rejected": -640.0,
+      "loss": 0.1347,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -9.0,
+      "rewards/margins": 8.5,
+      "rewards/rejected": -17.5,
+      "step": 3190
+    },
+    {
+      "epoch": 1.674515960230246,
+      "grad_norm": 7.606755469102715,
+      "learning_rate": 3.928346894484056e-06,
+      "logits/chosen": -13.0625,
+      "logits/rejected": -12.9375,
+      "logps/chosen": -552.0,
+      "logps/rejected": -600.0,
+      "loss": 0.1338,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -9.375,
+      "rewards/margins": 6.28125,
+      "rewards/rejected": -15.625,
+      "step": 3200
+    },
+    {
+      "epoch": 1.6797488226059656,
+      "grad_norm": 6.996228868642036,
+      "learning_rate": 3.8063313319680686e-06,
+      "logits/chosen": -12.6875,
+      "logits/rejected": -12.25,
+      "logps/chosen": -544.0,
+      "logps/rejected": -680.0,
+      "loss": 0.1061,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -9.0625,
+      "rewards/margins": 8.6875,
+      "rewards/rejected": -17.75,
+      "step": 3210
+    },
+    {
+      "epoch": 1.684981684981685,
+      "grad_norm": 2.1894497750464494,
+      "learning_rate": 3.686084405669249e-06,
+      "logits/chosen": -13.4375,
+      "logits/rejected": -13.0625,
+      "logps/chosen": -540.0,
+      "logps/rejected": -712.0,
+      "loss": 0.1001,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.4375,
+      "rewards/margins": 10.25,
+      "rewards/rejected": -18.75,
+      "step": 3220
+    },
+    {
+      "epoch": 1.6902145473574044,
+      "grad_norm": 3.29711389402712,
+      "learning_rate": 3.567616150332992e-06,
+      "logits/chosen": -12.9375,
+      "logits/rejected": -12.6875,
+      "logps/chosen": -504.0,
+      "logps/rejected": -588.0,
+      "loss": 0.1136,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -8.625,
+      "rewards/margins": 7.40625,
+      "rewards/rejected": -16.0,
+      "step": 3230
+    },
+    {
+      "epoch": 1.695447409733124,
+      "grad_norm": 3.5590799179096213,
+      "learning_rate": 3.450936452272524e-06,
+      "logits/chosen": -13.25,
+      "logits/rejected": -12.8125,
+      "logps/chosen": -490.0,
+      "logps/rejected": -612.0,
+      "loss": 0.1291,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -8.4375,
+      "rewards/margins": 7.8125,
+      "rewards/rejected": -16.25,
+      "step": 3240
+    },
+    {
+      "epoch": 1.7006802721088436,
+      "grad_norm": 2.1632084987538476,
+      "learning_rate": 3.3360550485439067e-06,
+      "logits/chosen": -13.5625,
+      "logits/rejected": -13.0625,
+      "logps/chosen": -454.0,
+      "logps/rejected": -636.0,
+      "loss": 0.0754,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -7.71875,
+      "rewards/margins": 9.375,
+      "rewards/rejected": -17.125,
+      "step": 3250
+    },
+    {
+      "epoch": 1.705913134484563,
+      "grad_norm": 6.079477309011048,
+      "learning_rate": 3.222981526133434e-06,
+      "logits/chosen": -13.1875,
+      "logits/rejected": -12.9375,
+      "logps/chosen": -486.0,
+      "logps/rejected": -600.0,
+      "loss": 0.1236,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -9.0,
+      "rewards/margins": 7.5,
+      "rewards/rejected": -16.5,
+      "step": 3260
+    },
+    {
+      "epoch": 1.7111459968602825,
+      "grad_norm": 1.5499240661013363,
+      "learning_rate": 3.111725321157627e-06,
+      "logits/chosen": -13.4375,
+      "logits/rejected": -13.1875,
+      "logps/chosen": -516.0,
+      "logps/rejected": -764.0,
+      "loss": 0.0815,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -7.1875,
+      "rewards/margins": 12.0625,
+      "rewards/rejected": -19.25,
+      "step": 3270
+    },
+    {
+      "epoch": 1.716378859236002,
+      "grad_norm": 3.7623282869444963,
+      "learning_rate": 3.002295718075762e-06,
+      "logits/chosen": -13.1875,
+      "logits/rejected": -12.8125,
+      "logps/chosen": -520.0,
+      "logps/rejected": -680.0,
+      "loss": 0.0881,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.25,
+      "rewards/margins": 10.0625,
+      "rewards/rejected": -18.375,
+      "step": 3280
+    },
+    {
+      "epoch": 1.7216117216117217,
+      "grad_norm": 7.519232363887075,
+      "learning_rate": 2.8947018489150517e-06,
+      "logits/chosen": -13.25,
+      "logits/rejected": -12.625,
+      "logps/chosen": -502.0,
+      "logps/rejected": -652.0,
+      "loss": 0.0775,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.6875,
+      "rewards/margins": 8.5,
+      "rewards/rejected": -17.125,
+      "step": 3290
+    },
+    {
+      "epoch": 1.7268445839874411,
+      "grad_norm": 5.31099596157704,
+      "learning_rate": 2.7889526925085978e-06,
+      "logits/chosen": -12.75,
+      "logits/rejected": -12.8125,
+      "logps/chosen": -516.0,
+      "logps/rejected": -644.0,
+      "loss": 0.1053,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.0625,
+      "rewards/margins": 9.4375,
+      "rewards/rejected": -17.5,
+      "step": 3300
+    },
+    {
+      "epoch": 1.7320774463631605,
+      "grad_norm": 7.1196364277612165,
+      "learning_rate": 2.6850570737460916e-06,
+      "logits/chosen": -13.125,
+      "logits/rejected": -12.8125,
+      "logps/chosen": -482.0,
+      "logps/rejected": -600.0,
+      "loss": 0.1152,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -8.125,
+      "rewards/margins": 8.75,
+      "rewards/rejected": -16.875,
+      "step": 3310
+    },
+    {
+      "epoch": 1.7373103087388801,
+      "grad_norm": 2.680324074975716,
+      "learning_rate": 2.5830236628373363e-06,
+      "logits/chosen": -13.1875,
+      "logits/rejected": -13.125,
+      "logps/chosen": -502.0,
+      "logps/rejected": -676.0,
+      "loss": 0.0778,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -9.0,
+      "rewards/margins": 9.0,
+      "rewards/rejected": -18.0,
+      "step": 3320
+    },
+    {
+      "epoch": 1.7425431711145998,
+      "grad_norm": 2.2573623813902244,
+      "learning_rate": 2.482860974588755e-06,
+      "logits/chosen": -13.125,
+      "logits/rejected": -12.5,
+      "logps/chosen": -532.0,
+      "logps/rejected": -652.0,
+      "loss": 0.0792,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -9.125,
+      "rewards/margins": 8.9375,
+      "rewards/rejected": -18.125,
+      "step": 3330
+    },
+    {
+      "epoch": 1.7477760334903192,
+      "grad_norm": 2.3988175015656794,
+      "learning_rate": 2.3845773676927863e-06,
+      "logits/chosen": -13.375,
+      "logits/rejected": -12.9375,
+      "logps/chosen": -500.0,
+      "logps/rejected": -644.0,
+      "loss": 0.0737,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -9.0625,
+      "rewards/margins": 8.625,
+      "rewards/rejected": -17.75,
+      "step": 3340
+    },
+    {
+      "epoch": 1.7530088958660386,
+      "grad_norm": 2.499191138157377,
+      "learning_rate": 2.288181044030341e-06,
+      "logits/chosen": -13.3125,
+      "logits/rejected": -12.8125,
+      "logps/chosen": -454.0,
+      "logps/rejected": -592.0,
+      "loss": 0.1102,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.1875,
+      "rewards/margins": 9.0,
+      "rewards/rejected": -17.125,
+      "step": 3350
+    },
+    {
+      "epoch": 1.7582417582417582,
+      "grad_norm": 1.7297577696934656,
+      "learning_rate": 2.193680047986385e-06,
+      "logits/chosen": -12.9375,
+      "logits/rejected": -12.75,
+      "logps/chosen": -478.0,
+      "logps/rejected": -608.0,
+      "loss": 0.1039,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -9.3125,
+      "rewards/margins": 7.375,
+      "rewards/rejected": -16.625,
+      "step": 3360
+    },
+    {
+      "epoch": 1.7634746206174778,
+      "grad_norm": 1.3844986076148504,
+      "learning_rate": 2.1010822657785673e-06,
+      "logits/chosen": -13.25,
+      "logits/rejected": -13.0625,
+      "logps/chosen": -552.0,
+      "logps/rejected": -624.0,
+      "loss": 0.0855,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -8.1875,
+      "rewards/margins": 8.625,
+      "rewards/rejected": -16.75,
+      "step": 3370
+    },
+    {
+      "epoch": 1.7687074829931972,
+      "grad_norm": 2.5009715621795063,
+      "learning_rate": 2.0103954247991525e-06,
+      "logits/chosen": -13.125,
+      "logits/rejected": -12.5625,
+      "logps/chosen": -488.0,
+      "logps/rejected": -612.0,
+      "loss": 0.09,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.625,
+      "rewards/margins": 8.6875,
+      "rewards/rejected": -17.25,
+      "step": 3380
+    },
+    {
+      "epoch": 1.7739403453689166,
+      "grad_norm": 1.8945829950156903,
+      "learning_rate": 1.9216270929701407e-06,
+      "logits/chosen": -12.9375,
+      "logits/rejected": -12.625,
+      "logps/chosen": -474.0,
+      "logps/rejected": -652.0,
+      "loss": 0.0723,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -8.75,
+      "rewards/margins": 8.0625,
+      "rewards/rejected": -16.75,
+      "step": 3390
+    },
+    {
+      "epoch": 1.7791732077446363,
+      "grad_norm": 5.069862077253051,
+      "learning_rate": 1.8347846781117201e-06,
+      "logits/chosen": -13.0625,
+      "logits/rejected": -12.625,
+      "logps/chosen": -512.0,
+      "logps/rejected": -616.0,
+      "loss": 0.1063,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.8125,
+      "rewards/margins": 9.0625,
+      "rewards/rejected": -17.875,
+      "step": 3400
+    },
+    {
+      "epoch": 1.784406070120356,
+      "grad_norm": 1.3102672583456336,
+      "learning_rate": 1.7498754273240713e-06,
+      "logits/chosen": -13.0625,
+      "logits/rejected": -12.5,
+      "logps/chosen": -516.0,
+      "logps/rejected": -656.0,
+      "loss": 0.102,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.375,
+      "rewards/margins": 9.5,
+      "rewards/rejected": -17.875,
+      "step": 3410
+    },
+    {
+      "epoch": 1.7896389324960753,
+      "grad_norm": 5.748063923557189,
+      "learning_rate": 1.6669064263826028e-06,
+      "logits/chosen": -12.875,
+      "logits/rejected": -12.625,
+      "logps/chosen": -572.0,
+      "logps/rejected": -608.0,
+      "loss": 0.1097,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.8125,
+      "rewards/margins": 8.125,
+      "rewards/rejected": -17.0,
+      "step": 3420
+    },
+    {
+      "epoch": 1.7948717948717947,
+      "grad_norm": 4.829587618568215,
+      "learning_rate": 1.5858845991466088e-06,
+      "logits/chosen": -13.0,
+      "logits/rejected": -12.875,
+      "logps/chosen": -472.0,
+      "logps/rejected": -576.0,
+      "loss": 0.1007,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -9.25,
+      "rewards/margins": 6.65625,
+      "rewards/rejected": -15.875,
+      "step": 3430
+    },
+    {
+      "epoch": 1.8001046572475143,
+      "grad_norm": 0.92994359502391,
+      "learning_rate": 1.5068167069814926e-06,
+      "logits/chosen": -12.875,
+      "logits/rejected": -12.625,
+      "logps/chosen": -560.0,
+      "logps/rejected": -692.0,
+      "loss": 0.0844,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -8.3125,
+      "rewards/margins": 9.9375,
+      "rewards/rejected": -18.25,
+      "step": 3440
+    },
+    {
+      "epoch": 1.805337519623234,
+      "grad_norm": 2.1999437018442163,
+      "learning_rate": 1.4297093481945106e-06,
+      "logits/chosen": -13.0625,
+      "logits/rejected": -12.75,
+      "logps/chosen": -506.0,
+      "logps/rejected": -608.0,
+      "loss": 0.0948,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -9.125,
+      "rewards/margins": 7.59375,
+      "rewards/rejected": -16.75,
+      "step": 3450
+    },
+    {
+      "epoch": 1.8105703819989536,
+      "grad_norm": 4.960781699983769,
+      "learning_rate": 1.3545689574841342e-06,
+      "logits/chosen": -13.125,
+      "logits/rejected": -12.5625,
+      "logps/chosen": -544.0,
+      "logps/rejected": -652.0,
+      "loss": 0.0968,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -9.0625,
+      "rewards/margins": 8.3125,
+      "rewards/rejected": -17.375,
+      "step": 3460
+    },
+    {
+      "epoch": 1.815803244374673,
+      "grad_norm": 3.66785793431449,
+      "learning_rate": 1.2814018054030623e-06,
+      "logits/chosen": -13.25,
+      "logits/rejected": -12.75,
+      "logps/chosen": -520.0,
+      "logps/rejected": -624.0,
+      "loss": 0.0943,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.875,
+      "rewards/margins": 8.8125,
+      "rewards/rejected": -17.75,
+      "step": 3470
+    },
+    {
+      "epoch": 1.8210361067503924,
+      "grad_norm": 4.803515943576416,
+      "learning_rate": 1.2102139978349497e-06,
+      "logits/chosen": -12.75,
+      "logits/rejected": -12.25,
+      "logps/chosen": -560.0,
+      "logps/rejected": -692.0,
+      "loss": 0.1218,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.625,
+      "rewards/margins": 9.0,
+      "rewards/rejected": -18.625,
+      "step": 3480
+    },
+    {
+      "epoch": 1.826268969126112,
+      "grad_norm": 1.3014751486404437,
+      "learning_rate": 1.14101147548486e-06,
+      "logits/chosen": -13.25,
+      "logits/rejected": -12.625,
+      "logps/chosen": -528.0,
+      "logps/rejected": -640.0,
+      "loss": 0.0868,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -8.5625,
+      "rewards/margins": 9.3125,
+      "rewards/rejected": -17.875,
+      "step": 3490
+    },
+    {
+      "epoch": 1.8315018315018317,
+      "grad_norm": 7.2868142451084,
+      "learning_rate": 1.0738000133834969e-06,
+      "logits/chosen": -13.4375,
+      "logits/rejected": -12.9375,
+      "logps/chosen": -532.0,
+      "logps/rejected": -616.0,
+      "loss": 0.1025,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -9.75,
+      "rewards/margins": 7.0625,
+      "rewards/rejected": -16.75,
+      "step": 3500
+    },
+    {
+      "epoch": 1.836734693877551,
+      "grad_norm": 2.182109722434071,
+      "learning_rate": 1.008585220405278e-06,
+      "logits/chosen": -13.0,
+      "logits/rejected": -12.625,
+      "logps/chosen": -462.0,
+      "logps/rejected": -612.0,
+      "loss": 0.0964,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -9.5,
+      "rewards/margins": 6.96875,
+      "rewards/rejected": -16.5,
+      "step": 3510
+    },
+    {
+      "epoch": 1.8419675562532705,
+      "grad_norm": 3.0642067374107125,
+      "learning_rate": 9.453725388002821e-07,
+      "logits/chosen": -13.25,
+      "logits/rejected": -13.0,
+      "logps/chosen": -540.0,
+      "logps/rejected": -680.0,
+      "loss": 0.1071,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -9.0,
+      "rewards/margins": 10.0,
+      "rewards/rejected": -19.0,
+      "step": 3520
+    },
+    {
+      "epoch": 1.84720041862899,
+      "grad_norm": 2.211282968905268,
+      "learning_rate": 8.841672437400528e-07,
+      "logits/chosen": -12.6875,
+      "logits/rejected": -12.4375,
+      "logps/chosen": -496.0,
+      "logps/rejected": -644.0,
+      "loss": 0.0976,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -8.875,
+      "rewards/margins": 8.25,
+      "rewards/rejected": -17.125,
+      "step": 3530
+    },
+    {
+      "epoch": 1.8524332810047097,
+      "grad_norm": 2.306667763441843,
+      "learning_rate": 8.249744428774103e-07,
+      "logits/chosen": -12.75,
+      "logits/rejected": -12.5,
+      "logps/chosen": -576.0,
+      "logps/rejected": -652.0,
+      "loss": 0.0807,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.75,
+      "rewards/margins": 9.125,
+      "rewards/rejected": -17.875,
+      "step": 3540
+    },
+    {
+      "epoch": 1.8576661433804291,
+      "grad_norm": 5.435125332604937,
+      "learning_rate": 7.677990759202086e-07,
+      "logits/chosen": -12.6875,
+      "logits/rejected": -11.9375,
+      "logps/chosen": -572.0,
+      "logps/rejected": -708.0,
+      "loss": 0.0971,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -8.875,
+      "rewards/margins": 9.375,
+      "rewards/rejected": -18.25,
+      "step": 3550
+    },
+    {
+      "epoch": 1.8628990057561485,
+      "grad_norm": 3.3348576248427326,
+      "learning_rate": 7.126459142190844e-07,
+      "logits/chosen": -12.6875,
+      "logits/rejected": -12.25,
+      "logps/chosen": -532.0,
+      "logps/rejected": -688.0,
+      "loss": 0.1057,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.125,
+      "rewards/margins": 9.0,
+      "rewards/rejected": -18.125,
+      "step": 3560
+    },
+    {
+      "epoch": 1.8681318681318682,
+      "grad_norm": 2.5808442085804644,
+      "learning_rate": 6.595195603693205e-07,
+      "logits/chosen": -12.75,
+      "logits/rejected": -12.5,
+      "logps/chosen": -488.0,
+      "logps/rejected": -604.0,
+      "loss": 0.066,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.75,
+      "rewards/margins": 7.1875,
+      "rewards/rejected": -15.9375,
+      "step": 3570
+    },
+    {
+      "epoch": 1.8733647305075878,
+      "grad_norm": 2.7672202913255366,
+      "learning_rate": 6.084244478267248e-07,
+      "logits/chosen": -12.8125,
+      "logits/rejected": -12.6875,
+      "logps/chosen": -520.0,
+      "logps/rejected": -628.0,
+      "loss": 0.1047,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -8.75,
+      "rewards/margins": 8.5,
+      "rewards/rejected": -17.25,
+      "step": 3580
+    },
+    {
+      "epoch": 1.8785975928833072,
+      "grad_norm": 2.02922718371571,
+      "learning_rate": 5.593648405376711e-07,
+      "logits/chosen": -12.9375,
+      "logits/rejected": -12.6875,
+      "logps/chosen": -524.0,
+      "logps/rejected": -660.0,
+      "loss": 0.0819,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -9.5,
+      "rewards/margins": 9.0,
+      "rewards/rejected": -18.5,
+      "step": 3590
+    },
+    {
+      "epoch": 1.8838304552590266,
+      "grad_norm": 2.046652559740739,
+      "learning_rate": 5.123448325832475e-07,
+      "logits/chosen": -13.25,
+      "logits/rejected": -12.625,
+      "logps/chosen": -472.0,
+      "logps/rejected": -580.0,
+      "loss": 0.1085,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -9.1875,
+      "rewards/margins": 7.53125,
+      "rewards/rejected": -16.75,
+      "step": 3600
+    },
+    {
+      "epoch": 1.8890633176347462,
+      "grad_norm": 5.589958818337675,
+      "learning_rate": 4.6736834783762397e-07,
+      "logits/chosen": -12.5,
+      "logits/rejected": -12.0625,
+      "logps/chosen": -484.0,
+      "logps/rejected": -612.0,
+      "loss": 0.1086,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -9.25,
+      "rewards/margins": 8.0625,
+      "rewards/rejected": -17.25,
+      "step": 3610
+    },
+    {
+      "epoch": 1.8942961800104658,
+      "grad_norm": 2.1397763501401443,
+      "learning_rate": 4.24439139640595e-07,
+      "logits/chosen": -12.75,
+      "logits/rejected": -12.375,
+      "logps/chosen": -476.0,
+      "logps/rejected": -612.0,
+      "loss": 0.1402,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -8.8125,
+      "rewards/margins": 7.75,
+      "rewards/rejected": -16.5,
+      "step": 3620
+    },
+    {
+      "epoch": 1.8995290423861853,
+      "grad_norm": 4.234807002383162,
+      "learning_rate": 3.835607904843358e-07,
+      "logits/chosen": -13.25,
+      "logits/rejected": -12.875,
+      "logps/chosen": -496.0,
+      "logps/rejected": -640.0,
+      "loss": 0.099,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -9.625,
+      "rewards/margins": 7.4375,
+      "rewards/rejected": -17.125,
+      "step": 3630
+    },
+    {
+      "epoch": 1.9047619047619047,
+      "grad_norm": 5.447862432508138,
+      "learning_rate": 3.4473671171447174e-07,
+      "logits/chosen": -12.8125,
+      "logits/rejected": -12.4375,
+      "logps/chosen": -488.0,
+      "logps/rejected": -596.0,
+      "loss": 0.0801,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.5625,
+      "rewards/margins": 8.875,
+      "rewards/rejected": -17.375,
+      "step": 3640
+    },
+    {
+      "epoch": 1.9099947671376243,
+      "grad_norm": 3.7446251130949766,
+      "learning_rate": 3.079701432453841e-07,
+      "logits/chosen": -12.6875,
+      "logits/rejected": -12.375,
+      "logps/chosen": -460.0,
+      "logps/rejected": -676.0,
+      "loss": 0.0963,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.9375,
+      "rewards/margins": 8.5625,
+      "rewards/rejected": -17.5,
+      "step": 3650
+    },
+    {
+      "epoch": 1.915227629513344,
+      "grad_norm": 3.968745889768715,
+      "learning_rate": 2.7326415328982056e-07,
+      "logits/chosen": -12.75,
+      "logits/rejected": -12.25,
+      "logps/chosen": -502.0,
+      "logps/rejected": -616.0,
+      "loss": 0.1276,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -10.125,
+      "rewards/margins": 6.875,
+      "rewards/rejected": -17.0,
+      "step": 3660
+    },
+    {
+      "epoch": 1.9204604918890633,
+      "grad_norm": 4.006875799984734,
+      "learning_rate": 2.4062163810288365e-07,
+      "logits/chosen": -13.0625,
+      "logits/rejected": -12.6875,
+      "logps/chosen": -536.0,
+      "logps/rejected": -632.0,
+      "loss": 0.0867,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.875,
+      "rewards/margins": 8.4375,
+      "rewards/rejected": -17.25,
+      "step": 3670
+    },
+    {
+      "epoch": 1.9256933542647827,
+      "grad_norm": 5.953542104090957,
+      "learning_rate": 2.100453217402959e-07,
+      "logits/chosen": -13.0625,
+      "logits/rejected": -12.5625,
+      "logps/chosen": -560.0,
+      "logps/rejected": -660.0,
+      "loss": 0.1176,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -9.625,
+      "rewards/margins": 8.125,
+      "rewards/rejected": -17.75,
+      "step": 3680
+    },
+    {
+      "epoch": 1.9309262166405023,
+      "grad_norm": 5.239490240155921,
+      "learning_rate": 1.8153775583110156e-07,
+      "logits/chosen": -13.0,
+      "logits/rejected": -12.625,
+      "logps/chosen": -458.0,
+      "logps/rejected": -604.0,
+      "loss": 0.0844,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.625,
+      "rewards/margins": 8.0,
+      "rewards/rejected": -16.625,
+      "step": 3690
+    },
+    {
+      "epoch": 1.936159079016222,
+      "grad_norm": 4.377024347096592,
+      "learning_rate": 1.5510131936472273e-07,
+      "logits/chosen": -13.3125,
+      "logits/rejected": -12.625,
+      "logps/chosen": -544.0,
+      "logps/rejected": -672.0,
+      "loss": 0.0835,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/chosen": -8.875,
+      "rewards/margins": 9.3125,
+      "rewards/rejected": -18.25,
+      "step": 3700
+    },
+    {
+      "epoch": 1.9413919413919414,
+      "grad_norm": 1.5159821172630095,
+      "learning_rate": 1.307382184924266e-07,
+      "logits/chosen": -13.375,
+      "logits/rejected": -12.8125,
+      "logps/chosen": -478.0,
+      "logps/rejected": -600.0,
+      "loss": 0.0899,
+      "rewards/accuracies": 0.9375,
+      "rewards/chosen": -9.25,
+      "rewards/margins": 7.40625,
+      "rewards/rejected": -16.625,
+      "step": 3710
+    },
+    {
+      "epoch": 1.9466248037676608,
+      "grad_norm": 4.469755228752353,
+      "learning_rate": 1.0845048634321731e-07,
+      "logits/chosen": -12.5625,
+      "logits/rejected": -12.25,
+      "logps/chosen": -516.0,
+      "logps/rejected": -588.0,
+      "loss": 0.1217,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -8.8125,
+      "rewards/margins": 7.78125,
+      "rewards/rejected": -16.625,
+      "step": 3720
+    },
+    {
+      "epoch": 1.9518576661433804,
+      "grad_norm": 3.3123314260717414,
+      "learning_rate": 8.823998285418522e-08,
+      "logits/chosen": -12.9375,
+      "logits/rejected": -12.625,
+      "logps/chosen": -544.0,
+      "logps/rejected": -648.0,
+      "loss": 0.086,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.5625,
+      "rewards/margins": 9.125,
+      "rewards/rejected": -17.75,
+      "step": 3730
+    },
+    {
+      "epoch": 1.9570905285191,
+      "grad_norm": 2.731361823113672,
+      "learning_rate": 7.010839461526752e-08,
+      "logits/chosen": -12.875,
+      "logits/rejected": -12.5,
+      "logps/chosen": -544.0,
+      "logps/rejected": -680.0,
+      "loss": 0.0895,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": -8.0,
+      "rewards/margins": 10.4375,
+      "rewards/rejected": -18.375,
+      "step": 3740
+    },
+    {
+      "epoch": 1.9623233908948194,
+      "grad_norm": 3.243299191382827,
+      "learning_rate": 5.4057234728521756e-08,
+      "logits/chosen": -12.9375,
+      "logits/rejected": -12.75,
+      "logps/chosen": -498.0,
+      "logps/rejected": -620.0,
+      "loss": 0.1018,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -8.625,
+      "rewards/margins": 8.1875,
+      "rewards/rejected": -16.75,
+      "step": 3750
+    },
+    {
+      "epoch": 1.9675562532705388,
+      "grad_norm": 5.643560413836883,
+      "learning_rate": 4.0087842681846286e-08,
+      "logits/chosen": -12.25,
+      "logits/rejected": -11.9375,
+      "logps/chosen": -528.0,
+      "logps/rejected": -604.0,
+      "loss": 0.1314,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.4375,
+      "rewards/margins": 7.6875,
+      "rewards/rejected": -17.125,
+      "step": 3760
+    },
+    {
+      "epoch": 1.9727891156462585,
+      "grad_norm": 1.3002574150876818,
+      "learning_rate": 2.820138423720309e-08,
+      "logits/chosen": -13.0,
+      "logits/rejected": -12.625,
+      "logps/chosen": -476.0,
+      "logps/rejected": -596.0,
+      "loss": 0.0943,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -9.125,
+      "rewards/margins": 7.78125,
+      "rewards/rejected": -16.875,
+      "step": 3770
+    },
+    {
+      "epoch": 1.978021978021978,
+      "grad_norm": 5.260548583290011,
+      "learning_rate": 1.839885133332053e-08,
+      "logits/chosen": -12.8125,
+      "logits/rejected": -12.25,
+      "logps/chosen": -544.0,
+      "logps/rejected": -696.0,
+      "loss": 0.1208,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": -9.125,
+      "rewards/margins": 9.875,
+      "rewards/rejected": -19.0,
+      "step": 3780
+    },
+    {
+      "epoch": 1.9832548403976975,
+      "grad_norm": 5.112452952032154,
+      "learning_rate": 1.0681062002940167e-08,
+      "logits/chosen": -13.0625,
+      "logits/rejected": -12.75,
+      "logps/chosen": -548.0,
+      "logps/rejected": -684.0,
+      "loss": 0.0699,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -8.4375,
+      "rewards/margins": 9.5,
+      "rewards/rejected": -18.0,
+      "step": 3790
+    },
+    {
+      "epoch": 1.988487702773417,
+      "grad_norm": 0.9856562812739469,
+      "learning_rate": 5.048660304524111e-09,
+      "logits/chosen": -12.875,
+      "logits/rejected": -12.4375,
+      "logps/chosen": -512.0,
+      "logps/rejected": -660.0,
+      "loss": 0.1136,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -9.5,
+      "rewards/margins": 8.8125,
+      "rewards/rejected": -18.25,
+      "step": 3800
+    },
+    {
+      "epoch": 1.9937205651491365,
+      "grad_norm": 1.084985961639873,
+      "learning_rate": 1.502116268523035e-09,
+      "logits/chosen": -13.3125,
+      "logits/rejected": -13.0625,
+      "logps/chosen": -536.0,
+      "logps/rejected": -648.0,
+      "loss": 0.1045,
+      "rewards/accuracies": 0.987500011920929,
+      "rewards/chosen": -8.75,
+      "rewards/margins": 9.0,
+      "rewards/rejected": -17.75,
+      "step": 3810
+    },
+    {
+      "epoch": 1.9989534275248562,
+      "grad_norm": 1.7332241383915472,
+      "learning_rate": 4.172585814643526e-11,
+      "logits/chosen": -13.4375,
+      "logits/rejected": -12.9375,
+      "logps/chosen": -508.0,
+      "logps/rejected": -628.0,
+      "loss": 0.0506,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": -8.5625,
+      "rewards/margins": 9.0625,
+      "rewards/rejected": -17.625,
+      "step": 3820
+    },
+    {
+      "epoch": 2.0,
+      "eval_logits/chosen": -13.3125,
+      "eval_logits/rejected": -13.0625,
+      "eval_logps/chosen": -556.0,
+      "eval_logps/rejected": -564.0,
+      "eval_loss": 0.7793359160423279,
+      "eval_rewards/accuracies": 0.72265625,
+      "eval_rewards/chosen": -11.75,
+      "eval_rewards/margins": 1.9140625,
+      "eval_rewards/rejected": -13.6875,
+      "eval_runtime": 46.77,
+      "eval_samples_per_second": 42.762,
+      "eval_steps_per_second": 0.684,
+      "step": 3822
+    },
+    {
+      "epoch": 2.0,
+      "step": 3822,
+      "total_flos": 0.0,
+      "train_loss": 0.41369995401518184,
+      "train_runtime": 7266.4225,
+      "train_samples_per_second": 16.826,
+      "train_steps_per_second": 0.526
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 3822,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}