updata beta=0.01 version

.ipynb_checkpoints/trainer_state-checkpoint.json

@@ -1,1441 +1,721 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.9994767137624281,
+  "epoch": 1.0,
   "eval_steps": 500,
-  "global_step": 955,
+  "global_step": 478,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.010465724751439037,
-      "grad_norm": 7.894938945770264,
-      "learning_rate": 2.0833333333333333e-07,
-      "logits/chosen": -0.3671875,
-      "logits/rejected": -0.2177734375,
-      "logps/chosen": -264.0,
-      "logps/rejected": -280.0,
-      "loss": 0.6932,
-      "rewards/accuracies": 0.21875,
-      "rewards/chosen": 0.000873565673828125,
-      "rewards/margins": 0.003082275390625,
-      "rewards/rejected": -0.0022125244140625,
+      "epoch": 0.02092050209205021,
+      "grad_norm": 1.1945245265960693,
+      "learning_rate": 4.1666666666666667e-07,
+      "logits/chosen": -0.30859375,
+      "logits/rejected": -0.2890625,
+      "logps/chosen": -312.0,
+      "logps/rejected": -288.0,
+      "loss": 0.692,
+      "rewards/accuracies": 0.26249998807907104,
+      "rewards/chosen": 0.0007476806640625,
+      "rewards/margins": 0.0005035400390625,
+      "rewards/rejected": 0.000240325927734375,
       "step": 10
     },
     {
-      "epoch": 0.020931449502878074,
-      "grad_norm": 8.288711547851562,
-      "learning_rate": 4.1666666666666667e-07,
-      "logits/chosen": -0.373046875,
-      "logits/rejected": -0.4296875,
-      "logps/chosen": -358.0,
-      "logps/rejected": -296.0,
-      "loss": 0.6947,
-      "rewards/accuracies": 0.2874999940395355,
-      "rewards/chosen": -0.0015106201171875,
-      "rewards/margins": -0.0001773834228515625,
-      "rewards/rejected": -0.00133514404296875,
+      "epoch": 0.04184100418410042,
+      "grad_norm": 1.1291463375091553,
+      "learning_rate": 8.333333333333333e-07,
+      "logits/chosen": -0.330078125,
+      "logits/rejected": -0.294921875,
+      "logps/chosen": -302.0,
+      "logps/rejected": -268.0,
+      "loss": 0.6921,
+      "rewards/accuracies": 0.3531250059604645,
+      "rewards/chosen": -0.00098419189453125,
+      "rewards/margins": 0.000614166259765625,
+      "rewards/rejected": -0.00160980224609375,
       "step": 20
     },
     {
-      "epoch": 0.03139717425431711,
-      "grad_norm": 8.539895057678223,
-      "learning_rate": 6.249999999999999e-07,
-      "logits/chosen": -0.38671875,
-      "logits/rejected": -0.45703125,
-      "logps/chosen": -286.0,
-      "logps/rejected": -249.0,
-      "loss": 0.6912,
-      "rewards/accuracies": 0.3125,
-      "rewards/chosen": -0.000537872314453125,
-      "rewards/margins": -0.00101470947265625,
-      "rewards/rejected": 0.000484466552734375,
+      "epoch": 0.06276150627615062,
+      "grad_norm": 1.1083910465240479,
+      "learning_rate": 9.995691082675907e-07,
+      "logits/chosen": -0.22265625,
+      "logits/rejected": -0.2314453125,
+      "logps/chosen": -338.0,
+      "logps/rejected": -306.0,
+      "loss": 0.6914,
+      "rewards/accuracies": 0.46875,
+      "rewards/chosen": -0.0006103515625,
+      "rewards/margins": 0.005645751953125,
+      "rewards/rejected": -0.006256103515625,
       "step": 30
     },
     {
-      "epoch": 0.04186289900575615,
-      "grad_norm": 7.561408996582031,
-      "learning_rate": 8.333333333333333e-07,
-      "logits/chosen": -0.30859375,
-      "logits/rejected": -0.2080078125,
-      "logps/chosen": -312.0,
-      "logps/rejected": -306.0,
-      "loss": 0.689,
-      "rewards/accuracies": 0.4625000059604645,
-      "rewards/chosen": 0.01068115234375,
-      "rewards/margins": 0.0240478515625,
-      "rewards/rejected": -0.01336669921875,
+      "epoch": 0.08368200836820083,
+      "grad_norm": 1.1081091165542603,
+      "learning_rate": 9.969385700404345e-07,
+      "logits/chosen": -0.35546875,
+      "logits/rejected": -0.34765625,
+      "logps/chosen": -328.0,
+      "logps/rejected": -300.0,
+      "loss": 0.6882,
+      "rewards/accuracies": 0.5718749761581421,
+      "rewards/chosen": 0.003509521484375,
+      "rewards/margins": 0.011474609375,
+      "rewards/rejected": -0.0079345703125,
       "step": 40
     },
     {
-      "epoch": 0.052328623757195186,
-      "grad_norm": 7.783715724945068,
-      "learning_rate": 9.999880027023293e-07,
-      "logits/chosen": -0.3671875,
-      "logits/rejected": -0.38671875,
-      "logps/chosen": -302.0,
-      "logps/rejected": -268.0,
-      "loss": 0.6772,
-      "rewards/accuracies": 0.625,
-      "rewards/chosen": 0.0257568359375,
-      "rewards/margins": 0.048828125,
-      "rewards/rejected": -0.02294921875,
+      "epoch": 0.10460251046025104,
+      "grad_norm": 1.1411256790161133,
+      "learning_rate": 9.91929453572245e-07,
+      "logits/chosen": -0.259765625,
+      "logits/rejected": -0.279296875,
+      "logps/chosen": -342.0,
+      "logps/rejected": -284.0,
+      "loss": 0.6843,
+      "rewards/accuracies": 0.640625,
+      "rewards/chosen": 0.0086669921875,
+      "rewards/margins": 0.02197265625,
+      "rewards/rejected": -0.0133056640625,
       "step": 50
     },
     {
-      "epoch": 0.06279434850863422,
-      "grad_norm": 6.941147327423096,
-      "learning_rate": 9.995681577335256e-07,
-      "logits/chosen": -0.34375,
-      "logits/rejected": -0.220703125,
-      "logps/chosen": -296.0,
-      "logps/rejected": -284.0,
-      "loss": 0.6664,
-      "rewards/accuracies": 0.5874999761581421,
-      "rewards/chosen": 0.032470703125,
-      "rewards/margins": 0.059326171875,
-      "rewards/rejected": -0.0267333984375,
+      "epoch": 0.12552301255230125,
+      "grad_norm": 1.1798174381256104,
+      "learning_rate": 9.845657348152955e-07,
+      "logits/chosen": -0.33203125,
+      "logits/rejected": -0.28515625,
+      "logps/chosen": -330.0,
+      "logps/rejected": -272.0,
+      "loss": 0.6789,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.0038909912109375,
+      "rewards/margins": 0.033203125,
+      "rewards/rejected": -0.029296875,
       "step": 60
     },
     {
-      "epoch": 0.07326007326007326,
-      "grad_norm": 7.589351654052734,
-      "learning_rate": 9.985490234976131e-07,
-      "logits/chosen": -0.40625,
-      "logits/rejected": -0.478515625,
-      "logps/chosen": -316.0,
-      "logps/rejected": -250.0,
-      "loss": 0.6589,
-      "rewards/accuracies": 0.606249988079071,
-      "rewards/chosen": 0.056884765625,
-      "rewards/margins": 0.10205078125,
-      "rewards/rejected": -0.044921875,
+      "epoch": 0.14644351464435146,
+      "grad_norm": 1.3184058666229248,
+      "learning_rate": 9.748826599393632e-07,
+      "logits/chosen": -0.396484375,
+      "logits/rejected": -0.3046875,
+      "logps/chosen": -282.0,
+      "logps/rejected": -268.0,
+      "loss": 0.675,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -0.000690460205078125,
+      "rewards/margins": 0.043212890625,
+      "rewards/rejected": -0.0439453125,
       "step": 70
     },
     {
-      "epoch": 0.0837257980115123,
-      "grad_norm": 7.692485809326172,
-      "learning_rate": 9.969318225629239e-07,
-      "logits/chosen": -0.4375,
-      "logits/rejected": -0.47265625,
-      "logps/chosen": -362.0,
-      "logps/rejected": -322.0,
-      "loss": 0.651,
-      "rewards/accuracies": 0.65625,
-      "rewards/chosen": 0.05908203125,
-      "rewards/margins": 0.15234375,
-      "rewards/rejected": -0.0927734375,
+      "epoch": 0.16736401673640167,
+      "grad_norm": 1.1700127124786377,
+      "learning_rate": 9.629265766272291e-07,
+      "logits/chosen": -0.49609375,
+      "logits/rejected": -0.404296875,
+      "logps/chosen": -302.0,
+      "logps/rejected": -290.0,
+      "loss": 0.6704,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": -0.005828857421875,
+      "rewards/margins": 0.03955078125,
+      "rewards/rejected": -0.04541015625,
       "step": 80
     },
     {
-      "epoch": 0.09419152276295134,
-      "grad_norm": 7.012711524963379,
-      "learning_rate": 9.947184949473476e-07,
-      "logits/chosen": -0.365234375,
-      "logits/rejected": -0.3203125,
-      "logps/chosen": -334.0,
-      "logps/rejected": -284.0,
-      "loss": 0.6364,
-      "rewards/accuracies": 0.637499988079071,
-      "rewards/chosen": 0.05908203125,
-      "rewards/margins": 0.169921875,
-      "rewards/rejected": -0.1103515625,
+      "epoch": 0.18828451882845187,
+      "grad_norm": 1.171791672706604,
+      "learning_rate": 9.487547122331964e-07,
+      "logits/chosen": -0.4609375,
+      "logits/rejected": -0.412109375,
+      "logps/chosen": -314.0,
+      "logps/rejected": -308.0,
+      "loss": 0.6674,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.014404296875,
+      "rewards/margins": 0.05859375,
+      "rewards/rejected": -0.0732421875,
       "step": 90
     },
     {
-      "epoch": 0.10465724751439037,
-      "grad_norm": 7.267467498779297,
-      "learning_rate": 9.919116957910565e-07,
-      "logits/chosen": -0.546875,
-      "logits/rejected": -0.50390625,
-      "logps/chosen": -330.0,
-      "logps/rejected": -280.0,
-      "loss": 0.6289,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": 0.07275390625,
-      "rewards/margins": 0.26953125,
-      "rewards/rejected": -0.1962890625,
+      "epoch": 0.20920502092050208,
+      "grad_norm": 1.2314943075180054,
+      "learning_rate": 9.324348998664548e-07,
+      "logits/chosen": -0.453125,
+      "logits/rejected": -0.41796875,
+      "logps/chosen": -312.0,
+      "logps/rejected": -294.0,
+      "loss": 0.6568,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.01043701171875,
+      "rewards/margins": 0.083984375,
+      "rewards/rejected": -0.0947265625,
       "step": 100
     },
     {
-      "epoch": 0.1151229722658294,
-      "grad_norm": 7.708714485168457,
-      "learning_rate": 9.88514792171362e-07,
-      "logits/chosen": -0.453125,
-      "logits/rejected": -0.390625,
-      "logps/chosen": -306.0,
-      "logps/rejected": -298.0,
-      "loss": 0.6172,
-      "rewards/accuracies": 0.6312500238418579,
-      "rewards/chosen": 0.0302734375,
-      "rewards/margins": 0.1875,
-      "rewards/rejected": -0.1572265625,
+      "epoch": 0.2301255230125523,
+      "grad_norm": 1.1103886365890503,
+      "learning_rate": 9.140452537103941e-07,
+      "logits/chosen": -0.55859375,
+      "logits/rejected": -0.5625,
+      "logps/chosen": -330.0,
+      "logps/rejected": -300.0,
+      "loss": 0.6515,
+      "rewards/accuracies": 0.640625,
+      "rewards/chosen": -0.045166015625,
+      "rewards/margins": 0.0888671875,
+      "rewards/rejected": -0.1337890625,
       "step": 110
     },
     {
-      "epoch": 0.12558869701726844,
-      "grad_norm": 7.422197341918945,
-      "learning_rate": 9.845318590635185e-07,
-      "logits/chosen": -0.4765625,
-      "logits/rejected": -0.44140625,
-      "logps/chosen": -314.0,
-      "logps/rejected": -256.0,
-      "loss": 0.6065,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.03759765625,
-      "rewards/margins": 0.349609375,
-      "rewards/rejected": -0.3125,
+      "epoch": 0.2510460251046025,
+      "grad_norm": 1.180451512336731,
+      "learning_rate": 8.936737951319275e-07,
+      "logits/chosen": -0.609375,
+      "logits/rejected": -0.5625,
+      "logps/chosen": -342.0,
+      "logps/rejected": -316.0,
+      "loss": 0.6445,
+      "rewards/accuracies": 0.703125,
+      "rewards/chosen": -0.06494140625,
+      "rewards/margins": 0.1513671875,
+      "rewards/rejected": -0.2158203125,
       "step": 120
     },
     {
-      "epoch": 0.1360544217687075,
-      "grad_norm": 7.663195610046387,
-      "learning_rate": 9.799676744523238e-07,
-      "logits/chosen": -0.578125,
-      "logits/rejected": -0.51953125,
-      "logps/chosen": -328.0,
-      "logps/rejected": -278.0,
-      "loss": 0.6167,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": -0.00689697265625,
-      "rewards/margins": 0.32421875,
-      "rewards/rejected": -0.33203125,
+      "epoch": 0.2719665271966527,
+      "grad_norm": 1.2995760440826416,
+      "learning_rate": 8.714180313704489e-07,
+      "logits/chosen": -0.7109375,
+      "logits/rejected": -0.578125,
+      "logps/chosen": -340.0,
+      "logps/rejected": -328.0,
+      "loss": 0.6395,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.111328125,
+      "rewards/margins": 0.1591796875,
+      "rewards/rejected": -0.26953125,
       "step": 130
     },
     {
-      "epoch": 0.14652014652014653,
-      "grad_norm": 7.429771900177002,
-      "learning_rate": 9.748277136003789e-07,
-      "logits/chosen": -0.5,
-      "logits/rejected": -0.296875,
-      "logps/chosen": -306.0,
-      "logps/rejected": -296.0,
-      "loss": 0.6256,
-      "rewards/accuracies": 0.637499988079071,
-      "rewards/chosen": -0.07666015625,
-      "rewards/margins": 0.1962890625,
-      "rewards/rejected": -0.2734375,
+      "epoch": 0.2928870292887029,
+      "grad_norm": 1.2812036275863647,
+      "learning_rate": 8.473844888230064e-07,
+      "logits/chosen": -0.6875,
+      "logits/rejected": -0.64453125,
+      "logps/chosen": -332.0,
+      "logps/rejected": -290.0,
+      "loss": 0.6381,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.1845703125,
+      "rewards/margins": 0.158203125,
+      "rewards/rejected": -0.34375,
       "step": 140
     },
     {
-      "epoch": 0.15698587127158556,
-      "grad_norm": 7.726008415222168,
-      "learning_rate": 9.691181424798824e-07,
-      "logits/chosen": -0.56640625,
-      "logits/rejected": -0.5546875,
-      "logps/chosen": -276.0,
-      "logps/rejected": -264.0,
-      "loss": 0.6199,
-      "rewards/accuracies": 0.6312500238418579,
-      "rewards/chosen": -0.07568359375,
-      "rewards/margins": 0.1953125,
-      "rewards/rejected": -0.271484375,
+      "epoch": 0.3138075313807531,
+      "grad_norm": 1.8293964862823486,
+      "learning_rate": 8.216882031596096e-07,
+      "logits/chosen": -0.73046875,
+      "logits/rejected": -0.6953125,
+      "logps/chosen": -340.0,
+      "logps/rejected": -322.0,
+      "loss": 0.6355,
+      "rewards/accuracies": 0.653124988079071,
+      "rewards/chosen": -0.220703125,
+      "rewards/margins": 0.17578125,
+      "rewards/rejected": -0.396484375,
       "step": 150
     },
     {
-      "epoch": 0.1674515960230246,
-      "grad_norm": 7.5920844078063965,
-      "learning_rate": 9.628458103758402e-07,
-      "logits/chosen": -0.609375,
-      "logits/rejected": -0.52734375,
-      "logps/chosen": -330.0,
-      "logps/rejected": -312.0,
-      "loss": 0.6082,
-      "rewards/accuracies": 0.612500011920929,
-      "rewards/chosen": -0.06787109375,
-      "rewards/margins": 0.294921875,
-      "rewards/rejected": -0.36328125,
+      "epoch": 0.33472803347280333,
+      "grad_norm": 1.3217679262161255,
+      "learning_rate": 7.944521687092142e-07,
+      "logits/chosen": -0.87109375,
+      "logits/rejected": -0.8046875,
+      "logps/chosen": -344.0,
+      "logps/rejected": -320.0,
+      "loss": 0.6231,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -0.21875,
+      "rewards/margins": 0.240234375,
+      "rewards/rejected": -0.458984375,
       "step": 160
     },
     {
-      "epoch": 0.17791732077446362,
-      "grad_norm": 6.563006401062012,
-      "learning_rate": 9.560182416695637e-07,
-      "logits/chosen": -0.5625,
-      "logits/rejected": -0.59375,
-      "logps/chosen": -310.0,
-      "logps/rejected": -292.0,
-      "loss": 0.6156,
-      "rewards/accuracies": 0.65625,
-      "rewards/chosen": -0.10205078125,
-      "rewards/margins": 0.25,
-      "rewards/rejected": -0.3515625,
+      "epoch": 0.35564853556485354,
+      "grad_norm": 1.493200659751892,
+      "learning_rate": 7.658067497518772e-07,
+      "logits/chosen": -0.94140625,
+      "logits/rejected": -0.92578125,
+      "logps/chosen": -324.0,
+      "logps/rejected": -308.0,
+      "loss": 0.6278,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.193359375,
+      "rewards/margins": 0.1865234375,
+      "rewards/rejected": -0.380859375,
       "step": 170
     },
     {
-      "epoch": 0.18838304552590268,
-      "grad_norm": 7.841839790344238,
-      "learning_rate": 9.486436268123111e-07,
-      "logits/chosen": -0.5859375,
-      "logits/rejected": -0.55859375,
-      "logps/chosen": -330.0,
-      "logps/rejected": -308.0,
-      "loss": 0.6261,
-      "rewards/accuracies": 0.6187499761581421,
-      "rewards/chosen": -0.1220703125,
-      "rewards/margins": 0.283203125,
-      "rewards/rejected": -0.40625,
+      "epoch": 0.37656903765690375,
+      "grad_norm": 1.3974004983901978,
+      "learning_rate": 7.358890565349105e-07,
+      "logits/chosen": -0.90234375,
+      "logits/rejected": -0.96484375,
+      "logps/chosen": -340.0,
+      "logps/rejected": -320.0,
+      "loss": 0.6182,
+      "rewards/accuracies": 0.671875,
+      "rewards/chosen": -0.2333984375,
+      "rewards/margins": 0.216796875,
+      "rewards/rejected": -0.451171875,
       "step": 180
     },
     {
-      "epoch": 0.1988487702773417,
-      "grad_norm": 6.8911027908325195,
-      "learning_rate": 9.40730812499903e-07,
-      "logits/chosen": -0.578125,
-      "logits/rejected": -0.5625,
-      "logps/chosen": -302.0,
-      "logps/rejected": -272.0,
-      "loss": 0.6092,
+      "epoch": 0.39748953974895396,
+      "grad_norm": 1.4906997680664062,
+      "learning_rate": 7.048422889997115e-07,
+      "logits/chosen": -0.96484375,
+      "logits/rejected": -0.91796875,
+      "logps/chosen": -342.0,
+      "logps/rejected": -330.0,
+      "loss": 0.632,
       "rewards/accuracies": 0.6875,
-      "rewards/chosen": -0.08984375,
-      "rewards/margins": 0.322265625,
-      "rewards/rejected": -0.412109375,
+      "rewards/chosen": -0.2578125,
+      "rewards/margins": 0.212890625,
+      "rewards/rejected": -0.470703125,
       "step": 190
     },
     {
-      "epoch": 0.20931449502878074,
-      "grad_norm": 8.3724946975708,
-      "learning_rate": 9.322892910600958e-07,
-      "logits/chosen": -0.6171875,
-      "logits/rejected": -0.451171875,
-      "logps/chosen": -298.0,
-      "logps/rejected": -278.0,
-      "loss": 0.6,
-      "rewards/accuracies": 0.65625,
-      "rewards/chosen": -0.0247802734375,
-      "rewards/margins": 0.357421875,
-      "rewards/rejected": -0.380859375,
+      "epoch": 0.41841004184100417,
+      "grad_norm": 1.4964449405670166,
+      "learning_rate": 6.72815051360494e-07,
+      "logits/chosen": -1.03125,
+      "logits/rejected": -1.015625,
+      "logps/chosen": -324.0,
+      "logps/rejected": -314.0,
+      "loss": 0.6232,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.263671875,
+      "rewards/margins": 0.1962890625,
+      "rewards/rejected": -0.458984375,
       "step": 200
     },
     {
-      "epoch": 0.21978021978021978,
-      "grad_norm": 7.580628871917725,
-      "learning_rate": 9.233291890654476e-07,
-      "logits/chosen": -0.6953125,
-      "logits/rejected": -0.66796875,
-      "logps/chosen": -318.0,
-      "logps/rejected": -284.0,
-      "loss": 0.5919,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": -0.012451171875,
-      "rewards/margins": 0.419921875,
-      "rewards/rejected": -0.431640625,
+      "epoch": 0.4393305439330544,
+      "grad_norm": 1.7729754447937012,
+      "learning_rate": 6.399606408156687e-07,
+      "logits/chosen": -1.046875,
+      "logits/rejected": -1.0078125,
+      "logps/chosen": -316.0,
+      "logps/rejected": -316.0,
+      "loss": 0.6014,
+      "rewards/accuracies": 0.684374988079071,
+      "rewards/chosen": -0.2470703125,
+      "rewards/margins": 0.2431640625,
+      "rewards/rejected": -0.490234375,
       "step": 210
     },
     {
-      "epoch": 0.2302459445316588,
-      "grad_norm": 7.269970893859863,
-      "learning_rate": 9.138612551853332e-07,
-      "logits/chosen": -0.72265625,
-      "logits/rejected": -0.68359375,
-      "logps/chosen": -286.0,
-      "logps/rejected": -302.0,
-      "loss": 0.5974,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": -0.05224609375,
-      "rewards/margins": 0.37890625,
-      "rewards/rejected": -0.431640625,
+      "epoch": 0.4602510460251046,
+      "grad_norm": 1.4692357778549194,
+      "learning_rate": 6.064363137964225e-07,
+      "logits/chosen": -1.125,
+      "logits/rejected": -0.9765625,
+      "logps/chosen": -348.0,
+      "logps/rejected": -358.0,
+      "loss": 0.6114,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.267578125,
+      "rewards/margins": 0.30859375,
+      "rewards/rejected": -0.578125,
       "step": 220
     },
     {
-      "epoch": 0.24071166928309787,
-      "grad_norm": 8.078968048095703,
-      "learning_rate": 9.03896847291683e-07,
-      "logits/chosen": -0.6328125,
-      "logits/rejected": -0.578125,
-      "logps/chosen": -348.0,
-      "logps/rejected": -308.0,
-      "loss": 0.5879,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.003662109375,
-      "rewards/margins": 0.462890625,
-      "rewards/rejected": -0.458984375,
+      "epoch": 0.4811715481171548,
+      "grad_norm": 1.377569317817688,
+      "learning_rate": 5.724025332645793e-07,
+      "logits/chosen": -1.0625,
+      "logits/rejected": -1.0390625,
+      "logps/chosen": -350.0,
+      "logps/rejected": -344.0,
+      "loss": 0.6098,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -0.2890625,
+      "rewards/margins": 0.234375,
+      "rewards/rejected": -0.5234375,
       "step": 230
     },
     {
-      "epoch": 0.25117739403453687,
-      "grad_norm": 7.896542549133301,
-      "learning_rate": 8.934479188339137e-07,
-      "logits/chosen": -0.62890625,
-      "logits/rejected": -0.625,
-      "logps/chosen": -336.0,
-      "logps/rejected": -286.0,
-      "loss": 0.5862,
-      "rewards/accuracies": 0.668749988079071,
-      "rewards/chosen": -0.06787109375,
-      "rewards/margins": 0.36328125,
-      "rewards/rejected": -0.431640625,
+      "epoch": 0.502092050209205,
+      "grad_norm": 1.5865265130996704,
+      "learning_rate": 5.380222006625179e-07,
+      "logits/chosen": -1.1171875,
+      "logits/rejected": -1.078125,
+      "logps/chosen": -334.0,
+      "logps/rejected": -328.0,
+      "loss": 0.6252,
+      "rewards/accuracies": 0.6656249761581421,
+      "rewards/chosen": -0.26953125,
+      "rewards/margins": 0.263671875,
+      "rewards/rejected": -0.53125,
       "step": 240
     },
     {
-      "epoch": 0.2616431187859759,
-      "grad_norm": 7.647547245025635,
-      "learning_rate": 8.825270044993962e-07,
-      "logits/chosen": -0.6328125,
-      "logits/rejected": -0.443359375,
-      "logps/chosen": -372.0,
-      "logps/rejected": -336.0,
-      "loss": 0.5811,
-      "rewards/accuracies": 0.793749988079071,
-      "rewards/chosen": 0.034912109375,
-      "rewards/margins": 0.57421875,
-      "rewards/rejected": -0.5390625,
+      "epoch": 0.5230125523012552,
+      "grad_norm": 1.5621623992919922,
+      "learning_rate": 5.034598761913916e-07,
+      "logits/chosen": -1.140625,
+      "logits/rejected": -1.1015625,
+      "logps/chosen": -318.0,
+      "logps/rejected": -318.0,
+      "loss": 0.6285,
+      "rewards/accuracies": 0.671875,
+      "rewards/chosen": -0.248046875,
+      "rewards/margins": 0.2314453125,
+      "rewards/rejected": -0.478515625,
       "step": 250
     },
     {
-      "epoch": 0.272108843537415,
-      "grad_norm": 7.487204551696777,
-      "learning_rate": 8.711472051766605e-07,
-      "logits/chosen": -0.73046875,
-      "logits/rejected": -0.66015625,
-      "logps/chosen": -296.0,
-      "logps/rejected": -312.0,
-      "loss": 0.5871,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": -0.02685546875,
-      "rewards/margins": 0.421875,
-      "rewards/rejected": -0.44921875,
+      "epoch": 0.5439330543933054,
+      "grad_norm": 1.3323726654052734,
+      "learning_rate": 4.688809911497609e-07,
+      "logits/chosen": -1.125,
+      "logits/rejected": -1.109375,
+      "logps/chosen": -346.0,
+      "logps/rejected": -350.0,
+      "loss": 0.6079,
+      "rewards/accuracies": 0.7281249761581421,
+      "rewards/chosen": -0.240234375,
+      "rewards/margins": 0.30859375,
+      "rewards/rejected": -0.546875,
       "step": 260
     },
     {
-      "epoch": 0.282574568288854,
-      "grad_norm": 7.606942653656006,
-      "learning_rate": 8.593221722393789e-07,
-      "logits/chosen": -0.6875,
-      "logits/rejected": -0.62109375,
-      "logps/chosen": -298.0,
-      "logps/rejected": -280.0,
-      "loss": 0.5879,
-      "rewards/accuracies": 0.6812499761581421,
-      "rewards/chosen": -0.034912109375,
-      "rewards/margins": 0.56640625,
-      "rewards/rejected": -0.6015625,
+      "epoch": 0.5648535564853556,
+      "grad_norm": 1.4896010160446167,
+      "learning_rate": 4.344510561027498e-07,
+      "logits/chosen": -1.15625,
+      "logits/rejected": -1.15625,
+      "logps/chosen": -346.0,
+      "logps/rejected": -330.0,
+      "loss": 0.5974,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.271484375,
+      "rewards/margins": 0.310546875,
+      "rewards/rejected": -0.58203125,
       "step": 270
     },
     {
-      "epoch": 0.29304029304029305,
-      "grad_norm": 6.631251335144043,
-      "learning_rate": 8.470660911699782e-07,
-      "logits/chosen": -0.6171875,
-      "logits/rejected": -0.609375,
-      "logps/chosen": -308.0,
-      "logps/rejected": -274.0,
-      "loss": 0.5836,
-      "rewards/accuracies": 0.71875,
-      "rewards/chosen": -0.1591796875,
-      "rewards/margins": 0.486328125,
-      "rewards/rejected": -0.64453125,
+      "epoch": 0.5857740585774058,
+      "grad_norm": 1.4263241291046143,
+      "learning_rate": 4.003348686717949e-07,
+      "logits/chosen": -1.1953125,
+      "logits/rejected": -1.1171875,
+      "logps/chosen": -338.0,
+      "logps/rejected": -356.0,
+      "loss": 0.6141,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -0.27734375,
+      "rewards/margins": 0.306640625,
+      "rewards/rejected": -0.58203125,
       "step": 280
     },
     {
-      "epoch": 0.3035060177917321,
-      "grad_norm": 5.671150207519531,
-      "learning_rate": 8.343936645425276e-07,
-      "logits/chosen": -0.5390625,
-      "logits/rejected": -0.55859375,
-      "logps/chosen": -318.0,
-      "logps/rejected": -296.0,
-      "loss": 0.5744,
-      "rewards/accuracies": 0.731249988079071,
-      "rewards/chosen": -0.119140625,
-      "rewards/margins": 0.50390625,
-      "rewards/rejected": -0.62109375,
+      "epoch": 0.606694560669456,
+      "grad_norm": 1.510299563407898,
+      "learning_rate": 3.666957247368757e-07,
+      "logits/chosen": -1.109375,
+      "logits/rejected": -1.078125,
+      "logps/chosen": -338.0,
+      "logps/rejected": -322.0,
+      "loss": 0.5992,
+      "rewards/accuracies": 0.7281249761581421,
+      "rewards/chosen": -0.2451171875,
+      "rewards/margins": 0.345703125,
+      "rewards/rejected": -0.58984375,
       "step": 290
     },
     {
-      "epoch": 0.3139717425431711,
-      "grad_norm": 7.8931989669799805,
-      "learning_rate": 8.213200943853158e-07,
-      "logits/chosen": -0.8125,
-      "logits/rejected": -0.5546875,
-      "logps/chosen": -288.0,
-      "logps/rejected": -292.0,
-      "loss": 0.6098,
-      "rewards/accuracies": 0.59375,
-      "rewards/chosen": -0.1357421875,
-      "rewards/margins": 0.28515625,
-      "rewards/rejected": -0.421875,
+      "epoch": 0.6276150627615062,
+      "grad_norm": 1.8962136507034302,
+      "learning_rate": 3.3369463682677234e-07,
+      "logits/chosen": -1.1875,
+      "logits/rejected": -1.1484375,
+      "logps/chosen": -362.0,
+      "logps/rejected": -362.0,
+      "loss": 0.6158,
+      "rewards/accuracies": 0.7281249761581421,
+      "rewards/chosen": -0.267578125,
+      "rewards/margins": 0.32421875,
+      "rewards/rejected": -0.59375,
       "step": 300
     },
     {
-      "epoch": 0.32443746729461015,
-      "grad_norm": 7.260922908782959,
-      "learning_rate": 8.07861063944276e-07,
-      "logits/chosen": -0.63671875,
-      "logits/rejected": -0.76171875,
-      "logps/chosen": -368.0,
-      "logps/rejected": -274.0,
-      "loss": 0.5672,
-      "rewards/accuracies": 0.731249988079071,
-      "rewards/chosen": 0.0260009765625,
-      "rewards/margins": 0.6015625,
-      "rewards/rejected": -0.578125,
+      "epoch": 0.6485355648535565,
+      "grad_norm": 1.5493396520614624,
+      "learning_rate": 3.014895634385014e-07,
+      "logits/chosen": -1.109375,
+      "logits/rejected": -1.140625,
+      "logps/chosen": -352.0,
+      "logps/rejected": -340.0,
+      "loss": 0.601,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.306640625,
+      "rewards/margins": 0.2333984375,
+      "rewards/rejected": -0.5390625,
       "step": 310
     },
     {
-      "epoch": 0.3349031920460492,
-      "grad_norm": 6.909915924072266,
-      "learning_rate": 7.940327188691341e-07,
-      "logits/chosen": -0.72265625,
-      "logits/rejected": -0.7265625,
-      "logps/chosen": -306.0,
-      "logps/rejected": -282.0,
-      "loss": 0.5732,
-      "rewards/accuracies": 0.6937500238418579,
-      "rewards/chosen": -0.050537109375,
-      "rewards/margins": 0.55078125,
-      "rewards/rejected": -0.6015625,
+      "epoch": 0.6694560669456067,
+      "grad_norm": 1.5511579513549805,
+      "learning_rate": 2.7023465297476424e-07,
+      "logits/chosen": -1.15625,
+      "logits/rejected": -1.15625,
+      "logps/chosen": -354.0,
+      "logps/rejected": -358.0,
+      "loss": 0.595,
+      "rewards/accuracies": 0.690625011920929,
+      "rewards/chosen": -0.283203125,
+      "rewards/margins": 0.326171875,
+      "rewards/rejected": -0.609375,
       "step": 320
     },
     {
-      "epoch": 0.3453689167974882,
-      "grad_norm": 7.135429382324219,
-      "learning_rate": 7.798516478448514e-07,
-      "logits/chosen": -0.75390625,
-      "logits/rejected": -0.7265625,
-      "logps/chosen": -334.0,
-      "logps/rejected": -294.0,
-      "loss": 0.5739,
-      "rewards/accuracies": 0.6812499761581421,
-      "rewards/chosen": 0.031494140625,
-      "rewards/margins": 0.486328125,
-      "rewards/rejected": -0.455078125,
+      "epoch": 0.6903765690376569,
+      "grad_norm": 1.7871730327606201,
+      "learning_rate": 2.4007950591826913e-07,
+      "logits/chosen": -1.1328125,
+      "logits/rejected": -1.125,
+      "logps/chosen": -350.0,
+      "logps/rejected": -352.0,
+      "loss": 0.612,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -0.330078125,
+      "rewards/margins": 0.30078125,
+      "rewards/rejected": -0.6328125,
       "step": 330
     },
     {
-      "epoch": 0.35583464154892724,
-      "grad_norm": 6.795555591583252,
-      "learning_rate": 7.653348626915957e-07,
-      "logits/chosen": -0.73828125,
-      "logits/rejected": -0.734375,
-      "logps/chosen": -296.0,
-      "logps/rejected": -290.0,
-      "loss": 0.5676,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 0.00927734375,
-      "rewards/margins": 0.490234375,
-      "rewards/rejected": -0.482421875,
+      "epoch": 0.7112970711297071,
+      "grad_norm": 1.6344778537750244,
+      "learning_rate": 2.1116845877450805e-07,
+      "logits/chosen": -1.1953125,
+      "logits/rejected": -1.171875,
+      "logps/chosen": -330.0,
+      "logps/rejected": -316.0,
+      "loss": 0.6125,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -0.32421875,
+      "rewards/margins": 0.255859375,
+      "rewards/rejected": -0.58203125,
       "step": 340
     },
     {
-      "epoch": 0.3663003663003663,
-      "grad_norm": 6.563620567321777,
-      "learning_rate": 7.504997779571132e-07,
-      "logits/chosen": -0.63671875,
-      "logits/rejected": -0.62890625,
-      "logps/chosen": -330.0,
-      "logps/rejected": -288.0,
-      "loss": 0.5654,
-      "rewards/accuracies": 0.6187499761581421,
-      "rewards/chosen": -0.09326171875,
-      "rewards/margins": 0.31640625,
-      "rewards/rejected": -0.408203125,
+      "epoch": 0.7322175732217573,
+      "grad_norm": 1.6645337343215942,
+      "learning_rate": 1.8363989321036577e-07,
+      "logits/chosen": -1.296875,
+      "logits/rejected": -1.203125,
+      "logps/chosen": -332.0,
+      "logps/rejected": -340.0,
+      "loss": 0.6038,
+      "rewards/accuracies": 0.671875,
+      "rewards/chosen": -0.333984375,
+      "rewards/margins": 0.26953125,
+      "rewards/rejected": -0.6015625,
       "step": 350
     },
     {
-      "epoch": 0.37676609105180536,
-      "grad_norm": 8.264630317687988,
-      "learning_rate": 7.353641900259823e-07,
-      "logits/chosen": -0.75,
-      "logits/rejected": -0.83203125,
-      "logps/chosen": -340.0,
-      "logps/rejected": -288.0,
-      "loss": 0.5644,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": -0.0208740234375,
-      "rewards/margins": 0.5078125,
-      "rewards/rejected": -0.52734375,
+      "epoch": 0.7531380753138075,
+      "grad_norm": 1.5749666690826416,
+      "learning_rate": 1.5762557369534708e-07,
+      "logits/chosen": -1.15625,
+      "logits/rejected": -1.1875,
+      "logps/chosen": -322.0,
+      "logps/rejected": -348.0,
+      "loss": 0.6019,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -0.38671875,
+      "rewards/margins": 0.265625,
+      "rewards/rejected": -0.65234375,
       "step": 360
     },
     {
-      "epoch": 0.3872318158032444,
-      "grad_norm": 8.003825187683105,
-      "learning_rate": 7.199462557708097e-07,
-      "logits/chosen": -0.76171875,
-      "logits/rejected": -0.63671875,
-      "logps/chosen": -312.0,
-      "logps/rejected": -328.0,
-      "loss": 0.5807,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": -0.10888671875,
-      "rewards/margins": 0.494140625,
-      "rewards/rejected": -0.6015625,
+      "epoch": 0.7740585774058577,
+      "grad_norm": 1.582518458366394,
+      "learning_rate": 1.332500168157748e-07,
+      "logits/chosen": -1.15625,
+      "logits/rejected": -1.203125,
+      "logps/chosen": -360.0,
+      "logps/rejected": -348.0,
+      "loss": 0.6147,
+      "rewards/accuracies": 0.6656249761581421,
+      "rewards/chosen": -0.412109375,
+      "rewards/margins": 0.25390625,
+      "rewards/rejected": -0.6640625,
       "step": 370
     },
     {
-      "epoch": 0.3976975405546834,
-      "grad_norm": 6.812396049499512,
-      "learning_rate": 7.042644707709815e-07,
-      "logits/chosen": -0.66015625,
-      "logits/rejected": -0.6015625,
-      "logps/chosen": -290.0,
-      "logps/rejected": -245.0,
-      "loss": 0.5879,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": -0.03369140625,
-      "rewards/margins": 0.51171875,
-      "rewards/rejected": -0.546875,
+      "epoch": 0.7949790794979079,
+      "grad_norm": 1.690048098564148,
+      "learning_rate": 1.1062989528071681e-07,
+      "logits/chosen": -1.1953125,
+      "logits/rejected": -1.1796875,
+      "logps/chosen": -364.0,
+      "logps/rejected": -378.0,
+      "loss": 0.6108,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -0.349609375,
+      "rewards/margins": 0.2890625,
+      "rewards/rejected": -0.63671875,
       "step": 380
     },
     {
-      "epoch": 0.40816326530612246,
-      "grad_norm": 7.9978485107421875,
-      "learning_rate": 6.883376471250955e-07,
-      "logits/chosen": -0.69921875,
-      "logits/rejected": -0.734375,
-      "logps/chosen": -302.0,
-      "logps/rejected": -284.0,
-      "loss": 0.5847,
-      "rewards/accuracies": 0.6812499761581421,
-      "rewards/chosen": -0.035888671875,
-      "rewards/margins": 0.484375,
-      "rewards/rejected": -0.51953125,
+      "epoch": 0.8158995815899581,
+      "grad_norm": 1.6943994760513306,
+      "learning_rate": 8.987347947234192e-08,
+      "logits/chosen": -1.109375,
+      "logits/rejected": -1.1328125,
+      "logps/chosen": -328.0,
+      "logps/rejected": -324.0,
+      "loss": 0.6044,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -0.3203125,
+      "rewards/margins": 0.322265625,
+      "rewards/rejected": -0.64453125,
       "step": 390
     },
     {
-      "epoch": 0.4186289900575615,
-      "grad_norm": 7.784450531005859,
-      "learning_rate": 6.72184890883692e-07,
-      "logits/chosen": -0.734375,
-      "logits/rejected": -0.703125,
-      "logps/chosen": -286.0,
-      "logps/rejected": -253.0,
-      "loss": 0.5448,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": -0.034912109375,
-      "rewards/margins": 0.55078125,
-      "rewards/rejected": -0.5859375,
+      "epoch": 0.8368200836820083,
+      "grad_norm": 1.6517119407653809,
+      "learning_rate": 7.108011921370727e-08,
+      "logits/chosen": -1.1875,
+      "logits/rejected": -1.1640625,
+      "logps/chosen": -386.0,
+      "logps/rejected": -382.0,
+      "loss": 0.6041,
+      "rewards/accuracies": 0.715624988079071,
+      "rewards/chosen": -0.375,
+      "rewards/margins": 0.328125,
+      "rewards/rejected": -0.703125,
       "step": 400
     },
     {
-      "epoch": 0.4290947148090005,
-      "grad_norm": 7.214269161224365,
-      "learning_rate": 6.558255791293571e-07,
-      "logits/chosen": -0.640625,
-      "logits/rejected": -0.73046875,
-      "logps/chosen": -302.0,
-      "logps/rejected": -276.0,
-      "loss": 0.5476,
-      "rewards/accuracies": 0.6812499761581421,
-      "rewards/chosen": 0.00787353515625,
-      "rewards/margins": 0.61328125,
-      "rewards/rejected": -0.60546875,
+      "epoch": 0.8577405857740585,
+      "grad_norm": 1.6941229104995728,
+      "learning_rate": 5.433976823447262e-08,
+      "logits/chosen": -1.203125,
+      "logits/rejected": -1.1015625,
+      "logps/chosen": -340.0,
+      "logps/rejected": -352.0,
+      "loss": 0.6026,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -0.392578125,
+      "rewards/margins": 0.333984375,
+      "rewards/rejected": -0.7265625,
       "step": 410
     },
     {
-      "epoch": 0.43956043956043955,
-      "grad_norm": 8.752714157104492,
-      "learning_rate": 6.392793367316904e-07,
-      "logits/chosen": -0.73828125,
-      "logits/rejected": -0.7109375,
-      "logps/chosen": -334.0,
-      "logps/rejected": -306.0,
-      "loss": 0.5496,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.0311279296875,
-      "rewards/margins": 0.5625,
-      "rewards/rejected": -0.52734375,
+      "epoch": 0.8786610878661087,
+      "grad_norm": 1.6011600494384766,
+      "learning_rate": 3.973255361067346e-08,
+      "logits/chosen": -1.2265625,
+      "logits/rejected": -1.1875,
+      "logps/chosen": -362.0,
+      "logps/rejected": -352.0,
+      "loss": 0.5924,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -0.34375,
+      "rewards/margins": 0.2890625,
+      "rewards/rejected": -0.6328125,
       "step": 420
     },
     {
-      "epoch": 0.4500261643118786,
-      "grad_norm": 9.271011352539062,
-      "learning_rate": 6.225660128050247e-07,
-      "logits/chosen": -0.7734375,
-      "logits/rejected": -0.73828125,
-      "logps/chosen": -284.0,
-      "logps/rejected": -272.0,
-      "loss": 0.567,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": -0.05224609375,
-      "rewards/margins": 0.53515625,
-      "rewards/rejected": -0.58984375,
+      "epoch": 0.899581589958159,
+      "grad_norm": 1.7107675075531006,
+      "learning_rate": 2.732839223940914e-08,
+      "logits/chosen": -1.1484375,
+      "logits/rejected": -1.1953125,
+      "logps/chosen": -356.0,
+      "logps/rejected": -366.0,
+      "loss": 0.5962,
+      "rewards/accuracies": 0.684374988079071,
+      "rewards/chosen": -0.373046875,
+      "rewards/margins": 0.3671875,
+      "rewards/rejected": -0.7421875,
       "step": 430
     },
     {
-      "epoch": 0.4604918890633176,
-      "grad_norm": 8.128939628601074,
-      "learning_rate": 6.057056568971383e-07,
-      "logits/chosen": -0.65625,
-      "logits/rejected": -0.6953125,
-      "logps/chosen": -372.0,
-      "logps/rejected": -302.0,
-      "loss": 0.5309,
-      "rewards/accuracies": 0.7562500238418579,
-      "rewards/chosen": 0.0830078125,
-      "rewards/margins": 0.6953125,
-      "rewards/rejected": -0.609375,
+      "epoch": 0.9205020920502092,
+      "grad_norm": 1.5065439939498901,
+      "learning_rate": 1.7186656184179473e-08,
+      "logits/chosen": -1.203125,
+      "logits/rejected": -1.2109375,
+      "logps/chosen": -364.0,
+      "logps/rejected": -374.0,
+      "loss": 0.6021,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -0.361328125,
+      "rewards/margins": 0.373046875,
+      "rewards/rejected": -0.734375,
       "step": 440
     },
     {
-      "epoch": 0.47095761381475665,
-      "grad_norm": 6.812074661254883,
-      "learning_rate": 5.887184949375242e-07,
-      "logits/chosen": -0.6484375,
-      "logits/rejected": -0.7265625,
-      "logps/chosen": -318.0,
-      "logps/rejected": -296.0,
-      "loss": 0.5451,
-      "rewards/accuracies": 0.78125,
-      "rewards/chosen": -0.00384521484375,
-      "rewards/margins": 0.64453125,
-      "rewards/rejected": -0.6484375,
-      "step": 450
-    },
-    {
-      "epoch": 0.48142333856619574,
-      "grad_norm": 7.323412895202637,
-      "learning_rate": 5.716249049740689e-07,
-      "logits/chosen": -0.66015625,
-      "logits/rejected": -0.6015625,
+      "epoch": 0.9414225941422594,
+      "grad_norm": 1.560874104499817,
+      "learning_rate": 9.355888492680153e-09,
+      "logits/chosen": -1.1328125,
+      "logits/rejected": -1.1015625,
       "logps/chosen": -340.0,
-      "logps/rejected": -328.0,
-      "loss": 0.5538,
-      "rewards/accuracies": 0.7562500238418579,
-      "rewards/chosen": 0.01373291015625,
-      "rewards/margins": 0.72265625,
-      "rewards/rejected": -0.7109375,
-      "step": 460
-    },
-    {
-      "epoch": 0.49188906331763477,
-      "grad_norm": 6.901383876800537,
-      "learning_rate": 5.544453927272492e-07,
-      "logits/chosen": -0.69140625,
-      "logits/rejected": -0.76171875,
-      "logps/chosen": -326.0,
-      "logps/rejected": -316.0,
-      "loss": 0.5626,
-      "rewards/accuracies": 0.731249988079071,
-      "rewards/chosen": -0.1630859375,
-      "rewards/margins": 0.55078125,
-      "rewards/rejected": -0.71484375,
-      "step": 470
-    },
-    {
-      "epoch": 0.5023547880690737,
-      "grad_norm": 10.153675079345703,
-      "learning_rate": 5.372005669911693e-07,
-      "logits/chosen": -0.64453125,
-      "logits/rejected": -0.60546875,
-      "logps/chosen": -326.0,
-      "logps/rejected": -274.0,
-      "loss": 0.586,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": -0.1884765625,
-      "rewards/margins": 0.486328125,
-      "rewards/rejected": -0.67578125,
-      "step": 480
-    },
-    {
-      "epoch": 0.5128205128205128,
-      "grad_norm": 6.661205768585205,
-      "learning_rate": 5.199111149109497e-07,
-      "logits/chosen": -0.78125,
-      "logits/rejected": -0.6328125,
-      "logps/chosen": -304.0,
-      "logps/rejected": -278.0,
-      "loss": 0.5948,
-      "rewards/accuracies": 0.75,
-      "rewards/chosen": 0.003631591796875,
-      "rewards/margins": 0.65625,
-      "rewards/rejected": -0.65234375,
-      "step": 490
-    },
-    {
-      "epoch": 0.5232862375719518,
-      "grad_norm": 7.033756256103516,
-      "learning_rate": 5.025977771661266e-07,
-      "logits/chosen": -0.68359375,
-      "logits/rejected": -0.69921875,
-      "logps/chosen": -292.0,
-      "logps/rejected": -304.0,
-      "loss": 0.5819,
-      "rewards/accuracies": 0.606249988079071,
-      "rewards/chosen": -0.185546875,
-      "rewards/margins": 0.37890625,
-      "rewards/rejected": -0.5625,
-      "step": 500
-    },
-    {
-      "epoch": 0.533751962323391,
-      "grad_norm": 7.984467506408691,
-      "learning_rate": 4.852813230898279e-07,
-      "logits/chosen": -0.63671875,
-      "logits/rejected": -0.640625,
-      "logps/chosen": -350.0,
-      "logps/rejected": -302.0,
-      "loss": 0.6104,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.01904296875,
-      "rewards/margins": 0.62109375,
-      "rewards/rejected": -0.6015625,
-      "step": 510
-    },
-    {
-      "epoch": 0.54421768707483,
-      "grad_norm": 6.375154495239258,
-      "learning_rate": 4.679825257535794e-07,
-      "logits/chosen": -0.68359375,
-      "logits/rejected": -0.74609375,
-      "logps/chosen": -338.0,
-      "logps/rejected": -322.0,
-      "loss": 0.5167,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.039794921875,
-      "rewards/margins": 0.69140625,
-      "rewards/rejected": -0.65234375,
-      "step": 520
-    },
-    {
-      "epoch": 0.554683411826269,
-      "grad_norm": 6.771985054016113,
-      "learning_rate": 4.507221370476223e-07,
-      "logits/chosen": -0.7421875,
-      "logits/rejected": -0.703125,
-      "logps/chosen": -312.0,
-      "logps/rejected": -320.0,
-      "loss": 0.547,
-      "rewards/accuracies": 0.6937500238418579,
-      "rewards/chosen": 0.064453125,
-      "rewards/margins": 0.6796875,
-      "rewards/rejected": -0.61328125,
-      "step": 530
-    },
-    {
-      "epoch": 0.565149136577708,
-      "grad_norm": 7.244710922241211,
-      "learning_rate": 4.3352086278664377e-07,
-      "logits/chosen": -0.76171875,
-      "logits/rejected": -0.75,
-      "logps/chosen": -302.0,
-      "logps/rejected": -266.0,
-      "loss": 0.5402,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": -0.07177734375,
-      "rewards/margins": 0.625,
+      "logps/rejected": -346.0,
+      "loss": 0.5957,
+      "rewards/accuracies": 0.721875011920929,
+      "rewards/chosen": -0.337890625,
+      "rewards/margins": 0.357421875,
       "rewards/rejected": -0.6953125,
-      "step": 540
-    },
-    {
-      "epoch": 0.5756148613291471,
-      "grad_norm": 8.408710479736328,
-      "learning_rate": 4.1639933787077854e-07,
-      "logits/chosen": -0.6796875,
-      "logits/rejected": -0.58984375,
-      "logps/chosen": -306.0,
-      "logps/rejected": -294.0,
-      "loss": 0.552,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.060302734375,
-      "rewards/margins": 0.6015625,
-      "rewards/rejected": -0.54296875,
-      "step": 550
-    },
-    {
-      "epoch": 0.5860805860805861,
-      "grad_norm": 7.500967979431152,
-      "learning_rate": 3.9937810153168016e-07,
-      "logits/chosen": -0.6171875,
-      "logits/rejected": -0.71484375,
-      "logps/chosen": -352.0,
-      "logps/rejected": -320.0,
-      "loss": 0.5782,
-      "rewards/accuracies": 0.65625,
-      "rewards/chosen": -0.08154296875,
-      "rewards/margins": 0.466796875,
-      "rewards/rejected": -0.546875,
-      "step": 560
-    },
-    {
-      "epoch": 0.5965463108320251,
-      "grad_norm": 6.775027751922607,
-      "learning_rate": 3.8247757269335957e-07,
-      "logits/chosen": -0.75390625,
-      "logits/rejected": -0.72265625,
-      "logps/chosen": -314.0,
-      "logps/rejected": -292.0,
-      "loss": 0.5366,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.0106201171875,
-      "rewards/margins": 0.6328125,
-      "rewards/rejected": -0.625,
-      "step": 570
-    },
-    {
-      "epoch": 0.6070120355834642,
-      "grad_norm": 7.576064586639404,
-      "learning_rate": 3.657180254773445e-07,
-      "logits/chosen": -0.6640625,
-      "logits/rejected": -0.58203125,
-      "logps/chosen": -310.0,
-      "logps/rejected": -264.0,
-      "loss": 0.5501,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": -0.0281982421875,
-      "rewards/margins": 0.70703125,
-      "rewards/rejected": -0.734375,
-      "step": 580
-    },
-    {
-      "epoch": 0.6174777603349032,
-      "grad_norm": 7.297247409820557,
-      "learning_rate": 3.4911956488154694e-07,
-      "logits/chosen": -0.70703125,
-      "logits/rejected": -0.625,
-      "logps/chosen": -334.0,
-      "logps/rejected": -298.0,
-      "loss": 0.5798,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": -0.038330078125,
-      "rewards/margins": 0.5546875,
-      "rewards/rejected": -0.59375,
-      "step": 590
-    },
-    {
-      "epoch": 0.6279434850863422,
-      "grad_norm": 9.57577133178711,
-      "learning_rate": 3.327021026620137e-07,
-      "logits/chosen": -0.640625,
-      "logits/rejected": -0.625,
-      "logps/chosen": -328.0,
-      "logps/rejected": -314.0,
-      "loss": 0.5525,
-      "rewards/accuracies": 0.7437499761581421,
-      "rewards/chosen": -0.03515625,
-      "rewards/margins": 0.62890625,
-      "rewards/rejected": -0.6640625,
-      "step": 600
+      "step": 450
     },
     {
-      "epoch": 0.6384092098377813,
-      "grad_norm": 8.11950397491455,
-      "learning_rate": 3.16485333446493e-07,
-      "logits/chosen": -0.65625,
-      "logits/rejected": -0.5390625,
-      "logps/chosen": -300.0,
+      "epoch": 0.9623430962343096,
+      "grad_norm": 1.5713697671890259,
+      "learning_rate": 3.873570847285012e-09,
+      "logits/chosen": -1.234375,
+      "logits/rejected": -1.2109375,
+      "logps/chosen": -356.0,
       "logps/rejected": -356.0,
-      "loss": 0.5299,
-      "rewards/accuracies": 0.7437499761581421,
-      "rewards/chosen": -0.0296630859375,
-      "rewards/margins": 0.66015625,
-      "rewards/rejected": -0.69140625,
-      "step": 610
-    },
-    {
-      "epoch": 0.6488749345892203,
-      "grad_norm": 8.621895790100098,
-      "learning_rate": 3.004887111084704e-07,
-      "logits/chosen": -0.7421875,
-      "logits/rejected": -0.8046875,
-      "logps/chosen": -324.0,
-      "logps/rejected": -296.0,
-      "loss": 0.554,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 0.06640625,
-      "rewards/margins": 0.640625,
-      "rewards/rejected": -0.57421875,
-      "step": 620
-    },
-    {
-      "epoch": 0.6593406593406593,
-      "grad_norm": 7.23579740524292,
-      "learning_rate": 2.8473142543001816e-07,
-      "logits/chosen": -0.73828125,
-      "logits/rejected": -0.71484375,
-      "logps/chosen": -304.0,
-      "logps/rejected": -300.0,
-      "loss": 0.554,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": -0.0615234375,
-      "rewards/margins": 0.65625,
-      "rewards/rejected": -0.71875,
-      "step": 630
-    },
-    {
-      "epoch": 0.6698063840920984,
-      "grad_norm": 7.622311115264893,
-      "learning_rate": 2.6923237908145226e-07,
-      "logits/chosen": -0.765625,
-      "logits/rejected": -0.66796875,
-      "logps/chosen": -314.0,
-      "logps/rejected": -312.0,
-      "loss": 0.5305,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.0216064453125,
-      "rewards/margins": 0.609375,
-      "rewards/rejected": -0.5859375,
-      "step": 640
-    },
-    {
-      "epoch": 0.6802721088435374,
-      "grad_norm": 7.722968578338623,
-      "learning_rate": 2.540101649454119e-07,
-      "logits/chosen": -0.57421875,
-      "logits/rejected": -0.625,
-      "logps/chosen": -318.0,
-      "logps/rejected": -304.0,
-      "loss": 0.5505,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.0615234375,
-      "rewards/margins": 0.72265625,
-      "rewards/rejected": -0.66015625,
-      "step": 650
-    },
-    {
-      "epoch": 0.6907378335949764,
-      "grad_norm": 9.817713737487793,
-      "learning_rate": 2.3908304381256603e-07,
-      "logits/chosen": -0.7578125,
-      "logits/rejected": -0.58984375,
-      "logps/chosen": -300.0,
-      "logps/rejected": -290.0,
-      "loss": 0.5692,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": -0.1904296875,
-      "rewards/margins": 0.60546875,
-      "rewards/rejected": -0.796875,
-      "step": 660
-    },
-    {
-      "epoch": 0.7012035583464155,
-      "grad_norm": 8.339869499206543,
-      "learning_rate": 2.2446892247570255e-07,
-      "logits/chosen": -0.70703125,
-      "logits/rejected": -0.72265625,
-      "logps/chosen": -306.0,
-      "logps/rejected": -300.0,
-      "loss": 0.5817,
-      "rewards/accuracies": 0.75,
-      "rewards/chosen": 0.0255126953125,
-      "rewards/margins": 0.765625,
-      "rewards/rejected": -0.7421875,
-      "step": 670
-    },
-    {
-      "epoch": 0.7116692830978545,
-      "grad_norm": 8.644556045532227,
-      "learning_rate": 2.1018533224847633e-07,
-      "logits/chosen": -0.640625,
-      "logits/rejected": -0.6015625,
-      "logps/chosen": -302.0,
-      "logps/rejected": -276.0,
-      "loss": 0.5732,
-      "rewards/accuracies": 0.706250011920929,
-      "rewards/chosen": 0.10009765625,
-      "rewards/margins": 0.69140625,
-      "rewards/rejected": -0.59375,
-      "step": 680
-    },
-    {
-      "epoch": 0.7221350078492935,
-      "grad_norm": 6.904910087585449,
-      "learning_rate": 1.9624940793459055e-07,
-      "logits/chosen": -0.78515625,
-      "logits/rejected": -0.67578125,
-      "logps/chosen": -332.0,
-      "logps/rejected": -306.0,
-      "loss": 0.5436,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": 0.07958984375,
-      "rewards/margins": 0.625,
-      "rewards/rejected": -0.546875,
-      "step": 690
-    },
-    {
-      "epoch": 0.7326007326007326,
-      "grad_norm": 8.480828285217285,
-      "learning_rate": 1.8267786727263424e-07,
-      "logits/chosen": -0.64453125,
-      "logits/rejected": -0.52734375,
-      "logps/chosen": -272.0,
-      "logps/rejected": -270.0,
-      "loss": 0.5594,
-      "rewards/accuracies": 0.643750011920929,
-      "rewards/chosen": -0.06591796875,
-      "rewards/margins": 0.333984375,
-      "rewards/rejected": -0.400390625,
-      "step": 700
-    },
-    {
-      "epoch": 0.7430664573521716,
-      "grad_norm": 9.295394897460938,
-      "learning_rate": 1.694869908812399e-07,
-      "logits/chosen": -0.625,
-      "logits/rejected": -0.73046875,
-      "logps/chosen": -298.0,
-      "logps/rejected": -278.0,
-      "loss": 0.5747,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": -0.1005859375,
-      "rewards/margins": 0.478515625,
-      "rewards/rejected": -0.578125,
-      "step": 710
-    },
-    {
-      "epoch": 0.7535321821036107,
-      "grad_norm": 7.032623767852783,
-      "learning_rate": 1.5669260272861422e-07,
-      "logits/chosen": -0.7109375,
-      "logits/rejected": -0.7578125,
-      "logps/chosen": -280.0,
-      "logps/rejected": -282.0,
-      "loss": 0.546,
-      "rewards/accuracies": 0.6937500238418579,
-      "rewards/chosen": 0.0296630859375,
-      "rewards/margins": 0.63671875,
-      "rewards/rejected": -0.60546875,
-      "step": 720
-    },
-    {
-      "epoch": 0.7639979068550498,
-      "grad_norm": 7.640635967254639,
-      "learning_rate": 1.4431005114987483e-07,
-      "logits/chosen": -0.57421875,
-      "logits/rejected": -0.68359375,
-      "logps/chosen": -336.0,
-      "logps/rejected": -300.0,
-      "loss": 0.5721,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": -0.0283203125,
-      "rewards/margins": 0.56640625,
-      "rewards/rejected": -0.59765625,
-      "step": 730
-    },
-    {
-      "epoch": 0.7744636316064888,
-      "grad_norm": 8.33640193939209,
-      "learning_rate": 1.323541904349636e-07,
-      "logits/chosen": -0.58984375,
-      "logits/rejected": -0.59765625,
-      "logps/chosen": -338.0,
-      "logps/rejected": -280.0,
-      "loss": 0.5804,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": 0.005828857421875,
-      "rewards/margins": 0.62109375,
-      "rewards/rejected": -0.6171875,
-      "step": 740
-    },
-    {
-      "epoch": 0.7849293563579278,
-      "grad_norm": 8.349899291992188,
-      "learning_rate": 1.2083936300922237e-07,
-      "logits/chosen": -0.55859375,
-      "logits/rejected": -0.58203125,
-      "logps/chosen": -332.0,
-      "logps/rejected": -334.0,
-      "loss": 0.573,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": 0.004302978515625,
-      "rewards/margins": 0.58984375,
-      "rewards/rejected": -0.5859375,
-      "step": 750
-    },
-    {
-      "epoch": 0.7953950811093669,
-      "grad_norm": 7.29964542388916,
-      "learning_rate": 1.0977938222801004e-07,
-      "logits/chosen": -0.64453125,
-      "logits/rejected": -0.71484375,
-      "logps/chosen": -306.0,
-      "logps/rejected": -286.0,
-      "loss": 0.5763,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": 0.053955078125,
-      "rewards/margins": 0.625,
-      "rewards/rejected": -0.5703125,
-      "step": 760
-    },
-    {
-      "epoch": 0.8058608058608059,
-      "grad_norm": 6.583461761474609,
-      "learning_rate": 9.918751580599999e-08,
-      "logits/chosen": -0.59375,
-      "logits/rejected": -0.60546875,
-      "logps/chosen": -340.0,
-      "logps/rejected": -296.0,
-      "loss": 0.5757,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.02490234375,
-      "rewards/margins": 0.578125,
-      "rewards/rejected": -0.5546875,
-      "step": 770
-    },
-    {
-      "epoch": 0.8163265306122449,
-      "grad_norm": 7.065498352050781,
-      "learning_rate": 8.907646990103495e-08,
-      "logits/chosen": -0.5703125,
-      "logits/rejected": -0.59765625,
-      "logps/chosen": -282.0,
-      "logps/rejected": -270.0,
-      "loss": 0.5388,
-      "rewards/accuracies": 0.6812499761581421,
-      "rewards/chosen": -0.0341796875,
-      "rewards/margins": 0.61328125,
-      "rewards/rejected": -0.6484375,
-      "step": 780
-    },
-    {
-      "epoch": 0.826792255363684,
-      "grad_norm": 6.218137741088867,
-      "learning_rate": 7.945837387163424e-08,
-      "logits/chosen": -0.65625,
-      "logits/rejected": -0.68359375,
-      "logps/chosen": -330.0,
-      "logps/rejected": -340.0,
-      "loss": 0.5702,
-      "rewards/accuracies": 0.7437499761581421,
-      "rewards/chosen": 0.048828125,
-      "rewards/margins": 0.71875,
-      "rewards/rejected": -0.671875,
-      "step": 790
-    },
-    {
-      "epoch": 0.837257980115123,
-      "grad_norm": 8.78946304321289,
-      "learning_rate": 7.034476572643854e-08,
-      "logits/chosen": -0.5703125,
-      "logits/rejected": -0.578125,
-      "logps/chosen": -300.0,
-      "logps/rejected": -306.0,
-      "loss": 0.5605,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": -0.0103759765625,
-      "rewards/margins": 0.53515625,
-      "rewards/rejected": -0.546875,
-      "step": 800
-    },
-    {
-      "epoch": 0.847723704866562,
-      "grad_norm": 7.07961893081665,
-      "learning_rate": 6.174657828304541e-08,
-      "logits/chosen": -0.66015625,
-      "logits/rejected": -0.54296875,
-      "logps/chosen": -300.0,
-      "logps/rejected": -296.0,
-      "loss": 0.5812,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": -0.01318359375,
-      "rewards/margins": 0.515625,
-      "rewards/rejected": -0.52734375,
-      "step": 810
-    },
-    {
-      "epoch": 0.858189429618001,
-      "grad_norm": 6.761130332946777,
-      "learning_rate": 5.36741260528415e-08,
-      "logits/chosen": -0.640625,
-      "logits/rejected": -0.71484375,
-      "logps/chosen": -336.0,
-      "logps/rejected": -286.0,
-      "loss": 0.5308,
-      "rewards/accuracies": 0.731249988079071,
-      "rewards/chosen": 0.12109375,
-      "rewards/margins": 0.74609375,
-      "rewards/rejected": -0.625,
-      "step": 820
-    },
-    {
-      "epoch": 0.8686551543694401,
-      "grad_norm": 6.658000469207764,
-      "learning_rate": 4.613709286756412e-08,
-      "logits/chosen": -0.625,
-      "logits/rejected": -0.61328125,
-      "logps/chosen": -302.0,
-      "logps/rejected": -276.0,
-      "loss": 0.5411,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.053466796875,
-      "rewards/margins": 0.7421875,
-      "rewards/rejected": -0.6875,
-      "step": 830
-    },
-    {
-      "epoch": 0.8791208791208791,
-      "grad_norm": 7.955780506134033,
-      "learning_rate": 3.914452026243509e-08,
-      "logits/chosen": -0.6796875,
-      "logits/rejected": -0.65625,
-      "logps/chosen": -332.0,
-      "logps/rejected": -308.0,
-      "loss": 0.549,
-      "rewards/accuracies": 0.65625,
-      "rewards/chosen": 0.018310546875,
-      "rewards/margins": 0.4921875,
-      "rewards/rejected": -0.474609375,
-      "step": 840
-    },
-    {
-      "epoch": 0.8895866038723181,
-      "grad_norm": 8.435918807983398,
-      "learning_rate": 3.270479662980247e-08,
-      "logits/chosen": -0.5546875,
-      "logits/rejected": -0.73828125,
-      "logps/chosen": -306.0,
-      "logps/rejected": -272.0,
-      "loss": 0.5817,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.068359375,
-      "rewards/margins": 0.54296875,
-      "rewards/rejected": -0.47265625,
-      "step": 850
-    },
-    {
-      "epoch": 0.9000523286237572,
-      "grad_norm": 8.153657913208008,
-      "learning_rate": 2.6825647156302865e-08,
-      "logits/chosen": -0.6015625,
-      "logits/rejected": -0.5078125,
-      "logps/chosen": -330.0,
-      "logps/rejected": -348.0,
-      "loss": 0.5387,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": -0.0135498046875,
-      "rewards/margins": 0.62890625,
-      "rewards/rejected": -0.64453125,
-      "step": 860
-    },
-    {
-      "epoch": 0.9105180533751962,
-      "grad_norm": 8.690088272094727,
-      "learning_rate": 2.151412455561441e-08,
-      "logits/chosen": -0.62109375,
-      "logits/rejected": -0.65625,
-      "logps/chosen": -320.0,
-      "logps/rejected": -286.0,
-      "loss": 0.5855,
+      "loss": 0.6004,
       "rewards/accuracies": 0.668749988079071,
-      "rewards/chosen": -0.018310546875,
-      "rewards/margins": 0.5625,
-      "rewards/rejected": -0.578125,
-      "step": 870
-    },
-    {
-      "epoch": 0.9209837781266352,
-      "grad_norm": 7.3918070793151855,
-      "learning_rate": 1.6776600607918356e-08,
-      "logits/chosen": -0.66015625,
-      "logits/rejected": -0.71484375,
-      "logps/chosen": -304.0,
-      "logps/rejected": -266.0,
-      "loss": 0.5473,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.00152587890625,
-      "rewards/margins": 0.5859375,
-      "rewards/rejected": -0.5859375,
-      "step": 880
-    },
-    {
-      "epoch": 0.9314495028780743,
-      "grad_norm": 7.804713726043701,
-      "learning_rate": 1.2618758516218186e-08,
-      "logits/chosen": -0.6484375,
-      "logits/rejected": -0.62890625,
-      "logps/chosen": -322.0,
-      "logps/rejected": -300.0,
-      "loss": 0.5591,
-      "rewards/accuracies": 0.78125,
-      "rewards/chosen": 0.12109375,
-      "rewards/margins": 0.6796875,
-      "rewards/rejected": -0.55859375,
-      "step": 890
-    },
-    {
-      "epoch": 0.9419152276295133,
-      "grad_norm": 7.053552627563477,
-      "learning_rate": 9.045586088686496e-09,
-      "logits/chosen": -0.63671875,
-      "logits/rejected": -0.67578125,
-      "logps/chosen": -346.0,
-      "logps/rejected": -292.0,
-      "loss": 0.5594,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 0.1943359375,
-      "rewards/margins": 0.61328125,
-      "rewards/rejected": -0.41796875,
-      "step": 900
-    },
-    {
-      "epoch": 0.9523809523809523,
-      "grad_norm": 7.403259754180908,
-      "learning_rate": 6.06136975521715e-09,
-      "logits/chosen": -0.73828125,
-      "logits/rejected": -0.69921875,
-      "logps/chosen": -278.0,
-      "logps/rejected": -264.0,
-      "loss": 0.5563,
-      "rewards/accuracies": 0.7875000238418579,
-      "rewards/chosen": 0.035888671875,
-      "rewards/margins": 0.6640625,
-      "rewards/rejected": -0.62890625,
-      "step": 910
-    },
-    {
-      "epoch": 0.9628466771323915,
-      "grad_norm": 7.359611511230469,
-      "learning_rate": 3.6696894253614442e-09,
-      "logits/chosen": -0.64453125,
-      "logits/rejected": -0.69140625,
-      "logps/chosen": -322.0,
-      "logps/rejected": -302.0,
-      "loss": 0.5624,
-      "rewards/accuracies": 0.731249988079071,
-      "rewards/chosen": 0.05712890625,
-      "rewards/margins": 0.671875,
-      "rewards/rejected": -0.61328125,
-      "step": 920
-    },
-    {
-      "epoch": 0.9733124018838305,
-      "grad_norm": 7.091738700866699,
-      "learning_rate": 1.8734141938160918e-09,
-      "logits/chosen": -0.59375,
-      "logits/rejected": -0.55078125,
-      "logps/chosen": -328.0,
-      "logps/rejected": -310.0,
-      "loss": 0.5677,
-      "rewards/accuracies": 0.6812499761581421,
-      "rewards/chosen": 0.038330078125,
-      "rewards/margins": 0.62890625,
-      "rewards/rejected": -0.59375,
-      "step": 930
-    },
-    {
-      "epoch": 0.9837781266352695,
-      "grad_norm": 6.281693935394287,
-      "learning_rate": 6.746988986155999e-10,
-      "logits/chosen": -0.7109375,
-      "logits/rejected": -0.625,
-      "logps/chosen": -330.0,
-      "logps/rejected": -310.0,
-      "loss": 0.5594,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.00750732421875,
-      "rewards/margins": 0.6171875,
-      "rewards/rejected": -0.609375,
-      "step": 940
+      "rewards/chosen": -0.40234375,
+      "rewards/margins": 0.30078125,
+      "rewards/rejected": -0.703125,
+      "step": 460
     },
     {
-      "epoch": 0.9942438513867086,
-      "grad_norm": 7.982404708862305,
-      "learning_rate": 7.498153615653758e-11,
-      "logits/chosen": -0.671875,
-      "logits/rejected": -0.68359375,
-      "logps/chosen": -312.0,
-      "logps/rejected": -314.0,
-      "loss": 0.5419,
-      "rewards/accuracies": 0.6937500238418579,
-      "rewards/chosen": -0.027587890625,
-      "rewards/margins": 0.59375,
-      "rewards/rejected": -0.62109375,
-      "step": 950
+      "epoch": 0.9832635983263598,
+      "grad_norm": 1.5760387182235718,
+      "learning_rate": 7.65944160348142e-10,
+      "logits/chosen": -1.2109375,
+      "logits/rejected": -1.140625,
+      "logps/chosen": -378.0,
+      "logps/rejected": -374.0,
+      "loss": 0.6114,
+      "rewards/accuracies": 0.6656249761581421,
+      "rewards/chosen": -0.3515625,
+      "rewards/margins": 0.330078125,
+      "rewards/rejected": -0.6796875,
+      "step": 470
     }
   ],
   "logging_steps": 10,
-  "max_steps": 955,
+  "max_steps": 478,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 1,
   "save_steps": 500,
@@ -1452,7 +732,7 @@
     }
   },
   "total_flos": 0.0,
-  "train_batch_size": 8,
+  "train_batch_size": 16,
   "trial_name": null,
   "trial_params": null
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5b40d0ab4920a56132143586d57d177f26c72f8d5eaae38a6ef93aaddcf67d50
+oid sha256:2c0f689042053c417ca3387a22224f757ce18251d97cd60840d94a327ed31e6d
 size 3554214752

runs/Dec31_23-26-13_notebook-7f74bd20-8f92-45b0-96c4-687ced4ee944/events.out.tfevents.1735659091.notebook-7f74bd20-8f92-45b0-96c4-687ced4ee944.2801.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b720f387937dbac03c42d4cf26ba8221c7e222f3320a4adaa9725cc45ff97b9
+size 38690

trainer_state.json

@@ -1,721 +1,721 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.9984301412872841,
+  "epoch": 1.0,
   "eval_steps": 500,
-  "global_step": 477,
+  "global_step": 478,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.020931449502878074,
-      "grad_norm": 0.24787747859954834,
+      "epoch": 0.02092050209205021,
+      "grad_norm": 1.1945245265960693,
       "learning_rate": 4.1666666666666667e-07,
-      "logits/chosen": -0.294921875,
-      "logits/rejected": -0.2490234375,
-      "logps/chosen": -314.0,
-      "logps/rejected": -290.0,
-      "loss": 0.241,
-      "rewards/accuracies": 0.2593750059604645,
-      "rewards/chosen": -0.000606536865234375,
-      "rewards/margins": 5.888938903808594e-05,
-      "rewards/rejected": -0.000667572021484375,
+      "logits/chosen": -0.30859375,
+      "logits/rejected": -0.2890625,
+      "logps/chosen": -312.0,
+      "logps/rejected": -288.0,
+      "loss": 0.692,
+      "rewards/accuracies": 0.26249998807907104,
+      "rewards/chosen": 0.0007476806640625,
+      "rewards/margins": 0.0005035400390625,
+      "rewards/rejected": 0.000240325927734375,
       "step": 10
     },
     {
-      "epoch": 0.04186289900575615,
-      "grad_norm": 0.22992154955863953,
+      "epoch": 0.04184100418410042,
+      "grad_norm": 1.1291463375091553,
       "learning_rate": 8.333333333333333e-07,
-      "logits/chosen": -0.275390625,
-      "logits/rejected": -0.265625,
-      "logps/chosen": -300.0,
-      "logps/rejected": -280.0,
-      "loss": 0.2396,
-      "rewards/accuracies": 0.3187499940395355,
-      "rewards/chosen": -0.000583648681640625,
-      "rewards/margins": -0.000942230224609375,
-      "rewards/rejected": 0.0003604888916015625,
+      "logits/chosen": -0.330078125,
+      "logits/rejected": -0.294921875,
+      "logps/chosen": -302.0,
+      "logps/rejected": -268.0,
+      "loss": 0.6921,
+      "rewards/accuracies": 0.3531250059604645,
+      "rewards/chosen": -0.00098419189453125,
+      "rewards/margins": 0.000614166259765625,
+      "rewards/rejected": -0.00160980224609375,
       "step": 20
     },
     {
-      "epoch": 0.06279434850863422,
-      "grad_norm": 0.23145779967308044,
-      "learning_rate": 9.995672040508656e-07,
-      "logits/chosen": -0.2578125,
-      "logits/rejected": -0.2041015625,
-      "logps/chosen": -302.0,
-      "logps/rejected": -278.0,
-      "loss": 0.2347,
-      "rewards/accuracies": 0.43437498807907104,
-      "rewards/chosen": -0.0014495849609375,
-      "rewards/margins": 0.0023651123046875,
-      "rewards/rejected": -0.003814697265625,
+      "epoch": 0.06276150627615062,
+      "grad_norm": 1.1083910465240479,
+      "learning_rate": 9.995691082675907e-07,
+      "logits/chosen": -0.22265625,
+      "logits/rejected": -0.2314453125,
+      "logps/chosen": -338.0,
+      "logps/rejected": -306.0,
+      "loss": 0.6914,
+      "rewards/accuracies": 0.46875,
+      "rewards/chosen": -0.0006103515625,
+      "rewards/margins": 0.005645751953125,
+      "rewards/rejected": -0.006256103515625,
       "step": 30
     },
     {
-      "epoch": 0.0837257980115123,
-      "grad_norm": 0.2316477745771408,
-      "learning_rate": 9.969250527610354e-07,
-      "logits/chosen": -0.30078125,
-      "logits/rejected": -0.35546875,
-      "logps/chosen": -342.0,
-      "logps/rejected": -288.0,
-      "loss": 0.2388,
-      "rewards/accuracies": 0.5531250238418579,
-      "rewards/chosen": 0.0027923583984375,
-      "rewards/margins": 0.0098876953125,
-      "rewards/rejected": -0.007110595703125,
+      "epoch": 0.08368200836820083,
+      "grad_norm": 1.1081091165542603,
+      "learning_rate": 9.969385700404345e-07,
+      "logits/chosen": -0.35546875,
+      "logits/rejected": -0.34765625,
+      "logps/chosen": -328.0,
+      "logps/rejected": -300.0,
+      "loss": 0.6882,
+      "rewards/accuracies": 0.5718749761581421,
+      "rewards/chosen": 0.003509521484375,
+      "rewards/margins": 0.011474609375,
+      "rewards/rejected": -0.0079345703125,
       "step": 40
     },
     {
-      "epoch": 0.10465724751439037,
-      "grad_norm": 0.23977532982826233,
-      "learning_rate": 9.918938793897e-07,
-      "logits/chosen": -0.306640625,
-      "logits/rejected": -0.267578125,
-      "logps/chosen": -334.0,
-      "logps/rejected": -282.0,
-      "loss": 0.2374,
-      "rewards/accuracies": 0.606249988079071,
-      "rewards/chosen": 0.00921630859375,
-      "rewards/margins": 0.01953125,
-      "rewards/rejected": -0.0103759765625,
+      "epoch": 0.10460251046025104,
+      "grad_norm": 1.1411256790161133,
+      "learning_rate": 9.91929453572245e-07,
+      "logits/chosen": -0.259765625,
+      "logits/rejected": -0.279296875,
+      "logps/chosen": -342.0,
+      "logps/rejected": -284.0,
+      "loss": 0.6843,
+      "rewards/accuracies": 0.640625,
+      "rewards/chosen": 0.0086669921875,
+      "rewards/margins": 0.02197265625,
+      "rewards/rejected": -0.0133056640625,
       "step": 50
     },
     {
-      "epoch": 0.12558869701726844,
-      "grad_norm": 0.22875246405601501,
-      "learning_rate": 9.844978718585855e-07,
-      "logits/chosen": -0.28515625,
-      "logits/rejected": -0.24609375,
-      "logps/chosen": -312.0,
-      "logps/rejected": -276.0,
-      "loss": 0.2258,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": 0.01043701171875,
-      "rewards/margins": 0.02685546875,
-      "rewards/rejected": -0.0164794921875,
+      "epoch": 0.12552301255230125,
+      "grad_norm": 1.1798174381256104,
+      "learning_rate": 9.845657348152955e-07,
+      "logits/chosen": -0.33203125,
+      "logits/rejected": -0.28515625,
+      "logps/chosen": -330.0,
+      "logps/rejected": -272.0,
+      "loss": 0.6789,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.0038909912109375,
+      "rewards/margins": 0.033203125,
+      "rewards/rejected": -0.029296875,
       "step": 60
     },
     {
-      "epoch": 0.14652014652014653,
-      "grad_norm": 0.2097596377134323,
-      "learning_rate": 9.747725872908607e-07,
-      "logits/chosen": -0.35546875,
-      "logits/rejected": -0.2314453125,
-      "logps/chosen": -318.0,
-      "logps/rejected": -286.0,
-      "loss": 0.2224,
-      "rewards/accuracies": 0.621874988079071,
-      "rewards/chosen": 0.005340576171875,
-      "rewards/margins": 0.031982421875,
-      "rewards/rejected": -0.0267333984375,
+      "epoch": 0.14644351464435146,
+      "grad_norm": 1.3184058666229248,
+      "learning_rate": 9.748826599393632e-07,
+      "logits/chosen": -0.396484375,
+      "logits/rejected": -0.3046875,
+      "logps/chosen": -282.0,
+      "logps/rejected": -268.0,
+      "loss": 0.675,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -0.000690460205078125,
+      "rewards/margins": 0.043212890625,
+      "rewards/rejected": -0.0439453125,
       "step": 70
     },
     {
-      "epoch": 0.1674515960230246,
-      "grad_norm": 0.260216623544693,
-      "learning_rate": 9.627647810663406e-07,
-      "logits/chosen": -0.365234375,
-      "logits/rejected": -0.328125,
-      "logps/chosen": -306.0,
-      "logps/rejected": -288.0,
-      "loss": 0.2262,
-      "rewards/accuracies": 0.6031249761581421,
-      "rewards/chosen": -0.01116943359375,
-      "rewards/margins": 0.03173828125,
-      "rewards/rejected": -0.04296875,
+      "epoch": 0.16736401673640167,
+      "grad_norm": 1.1700127124786377,
+      "learning_rate": 9.629265766272291e-07,
+      "logits/chosen": -0.49609375,
+      "logits/rejected": -0.404296875,
+      "logps/chosen": -302.0,
+      "logps/rejected": -290.0,
+      "loss": 0.6704,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": -0.005828857421875,
+      "rewards/margins": 0.03955078125,
+      "rewards/rejected": -0.04541015625,
       "step": 80
     },
     {
-      "epoch": 0.18838304552590268,
-      "grad_norm": 0.26707494258880615,
-      "learning_rate": 9.48532182039832e-07,
-      "logits/chosen": -0.318359375,
-      "logits/rejected": -0.328125,
-      "logps/chosen": -322.0,
-      "logps/rejected": -300.0,
-      "loss": 0.2271,
-      "rewards/accuracies": 0.637499988079071,
-      "rewards/chosen": -0.0208740234375,
-      "rewards/margins": 0.045654296875,
-      "rewards/rejected": -0.06640625,
+      "epoch": 0.18828451882845187,
+      "grad_norm": 1.171791672706604,
+      "learning_rate": 9.487547122331964e-07,
+      "logits/chosen": -0.4609375,
+      "logits/rejected": -0.412109375,
+      "logps/chosen": -314.0,
+      "logps/rejected": -308.0,
+      "loss": 0.6674,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.014404296875,
+      "rewards/margins": 0.05859375,
+      "rewards/rejected": -0.0732421875,
       "step": 90
     },
     {
-      "epoch": 0.20931449502878074,
-      "grad_norm": 0.2734158933162689,
-      "learning_rate": 9.321432150032882e-07,
-      "logits/chosen": -0.39453125,
-      "logits/rejected": -0.2890625,
-      "logps/chosen": -302.0,
-      "logps/rejected": -276.0,
-      "loss": 0.2226,
-      "rewards/accuracies": 0.653124988079071,
-      "rewards/chosen": -0.01153564453125,
-      "rewards/margins": 0.06494140625,
-      "rewards/rejected": -0.07666015625,
+      "epoch": 0.20920502092050208,
+      "grad_norm": 1.2314943075180054,
+      "learning_rate": 9.324348998664548e-07,
+      "logits/chosen": -0.453125,
+      "logits/rejected": -0.41796875,
+      "logps/chosen": -312.0,
+      "logps/rejected": -294.0,
+      "loss": 0.6568,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.01043701171875,
+      "rewards/margins": 0.083984375,
+      "rewards/rejected": -0.0947265625,
       "step": 100
     },
     {
-      "epoch": 0.2302459445316588,
-      "grad_norm": 0.21252916753292084,
-      "learning_rate": 9.13676671726063e-07,
-      "logits/chosen": -0.53515625,
-      "logits/rejected": -0.466796875,
-      "logps/chosen": -306.0,
-      "logps/rejected": -298.0,
-      "loss": 0.2129,
-      "rewards/accuracies": 0.643750011920929,
-      "rewards/chosen": -0.0234375,
-      "rewards/margins": 0.08056640625,
-      "rewards/rejected": -0.103515625,
+      "epoch": 0.2301255230125523,
+      "grad_norm": 1.1103886365890503,
+      "learning_rate": 9.140452537103941e-07,
+      "logits/chosen": -0.55859375,
+      "logits/rejected": -0.5625,
+      "logps/chosen": -330.0,
+      "logps/rejected": -300.0,
+      "loss": 0.6515,
+      "rewards/accuracies": 0.640625,
+      "rewards/chosen": -0.045166015625,
+      "rewards/margins": 0.0888671875,
+      "rewards/rejected": -0.1337890625,
       "step": 110
     },
     {
-      "epoch": 0.25117739403453687,
-      "grad_norm": 0.22458691895008087,
-      "learning_rate": 8.932213321547768e-07,
-      "logits/chosen": -0.48046875,
-      "logits/rejected": -0.42578125,
-      "logps/chosen": -348.0,
-      "logps/rejected": -304.0,
-      "loss": 0.2124,
-      "rewards/accuracies": 0.659375011920929,
-      "rewards/chosen": -0.041015625,
-      "rewards/margins": 0.09423828125,
-      "rewards/rejected": -0.1357421875,
+      "epoch": 0.2510460251046025,
+      "grad_norm": 1.180451512336731,
+      "learning_rate": 8.936737951319275e-07,
+      "logits/chosen": -0.609375,
+      "logits/rejected": -0.5625,
+      "logps/chosen": -342.0,
+      "logps/rejected": -316.0,
+      "loss": 0.6445,
+      "rewards/accuracies": 0.703125,
+      "rewards/chosen": -0.06494140625,
+      "rewards/margins": 0.1513671875,
+      "rewards/rejected": -0.2158203125,
       "step": 120
     },
     {
-      "epoch": 0.272108843537415,
-      "grad_norm": 0.24381956458091736,
-      "learning_rate": 8.708755375939161e-07,
-      "logits/chosen": -0.578125,
-      "logits/rejected": -0.435546875,
+      "epoch": 0.2719665271966527,
+      "grad_norm": 1.2995760440826416,
+      "learning_rate": 8.714180313704489e-07,
+      "logits/chosen": -0.7109375,
+      "logits/rejected": -0.578125,
       "logps/chosen": -340.0,
-      "logps/rejected": -332.0,
-      "loss": 0.1961,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": -0.045166015625,
-      "rewards/margins": 0.11474609375,
-      "rewards/rejected": -0.16015625,
+      "logps/rejected": -328.0,
+      "loss": 0.6395,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.111328125,
+      "rewards/margins": 0.1591796875,
+      "rewards/rejected": -0.26953125,
       "step": 130
     },
     {
-      "epoch": 0.29304029304029305,
-      "grad_norm": 0.21222463250160217,
-      "learning_rate": 8.467467179191492e-07,
-      "logits/chosen": -0.56640625,
-      "logits/rejected": -0.5078125,
-      "logps/chosen": -314.0,
+      "epoch": 0.2928870292887029,
+      "grad_norm": 1.2812036275863647,
+      "learning_rate": 8.473844888230064e-07,
+      "logits/chosen": -0.6875,
+      "logits/rejected": -0.64453125,
+      "logps/chosen": -332.0,
       "logps/rejected": -290.0,
-      "loss": 0.1913,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": -0.1064453125,
-      "rewards/margins": 0.119140625,
-      "rewards/rejected": -0.2255859375,
+      "loss": 0.6381,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.1845703125,
+      "rewards/margins": 0.158203125,
+      "rewards/rejected": -0.34375,
       "step": 140
     },
     {
-      "epoch": 0.3139717425431711,
-      "grad_norm": 0.2219766229391098,
-      "learning_rate": 8.209508750963328e-07,
-      "logits/chosen": -0.58984375,
-      "logits/rejected": -0.46875,
-      "logps/chosen": -318.0,
-      "logps/rejected": -310.0,
-      "loss": 0.1809,
-      "rewards/accuracies": 0.6156250238418579,
-      "rewards/chosen": -0.15234375,
-      "rewards/margins": 0.09912109375,
-      "rewards/rejected": -0.251953125,
+      "epoch": 0.3138075313807531,
+      "grad_norm": 1.8293964862823486,
+      "learning_rate": 8.216882031596096e-07,
+      "logits/chosen": -0.73046875,
+      "logits/rejected": -0.6953125,
+      "logps/chosen": -340.0,
+      "logps/rejected": -322.0,
+      "loss": 0.6355,
+      "rewards/accuracies": 0.653124988079071,
+      "rewards/chosen": -0.220703125,
+      "rewards/margins": 0.17578125,
+      "rewards/rejected": -0.396484375,
       "step": 150
     },
     {
-      "epoch": 0.3349031920460492,
-      "grad_norm": 0.24187949299812317,
-      "learning_rate": 7.936120254892471e-07,
-      "logits/chosen": -0.6015625,
-      "logits/rejected": -0.6171875,
-      "logps/chosen": -356.0,
-      "logps/rejected": -302.0,
-      "loss": 0.181,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": -0.1611328125,
-      "rewards/margins": 0.1748046875,
-      "rewards/rejected": -0.3359375,
+      "epoch": 0.33472803347280333,
+      "grad_norm": 1.3217679262161255,
+      "learning_rate": 7.944521687092142e-07,
+      "logits/chosen": -0.87109375,
+      "logits/rejected": -0.8046875,
+      "logps/chosen": -344.0,
+      "logps/rejected": -320.0,
+      "loss": 0.6231,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -0.21875,
+      "rewards/margins": 0.240234375,
+      "rewards/rejected": -0.458984375,
       "step": 160
     },
     {
-      "epoch": 0.35583464154892724,
-      "grad_norm": 0.2268301099538803,
-      "learning_rate": 7.648616036372286e-07,
-      "logits/chosen": -0.69921875,
-      "logits/rejected": -0.65625,
-      "logps/chosen": -334.0,
-      "logps/rejected": -316.0,
-      "loss": 0.171,
-      "rewards/accuracies": 0.653124988079071,
-      "rewards/chosen": -0.1669921875,
-      "rewards/margins": 0.1484375,
-      "rewards/rejected": -0.314453125,
+      "epoch": 0.35564853556485354,
+      "grad_norm": 1.493200659751892,
+      "learning_rate": 7.658067497518772e-07,
+      "logits/chosen": -0.94140625,
+      "logits/rejected": -0.92578125,
+      "logps/chosen": -324.0,
+      "logps/rejected": -308.0,
+      "loss": 0.6278,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.193359375,
+      "rewards/margins": 0.1865234375,
+      "rewards/rejected": -0.380859375,
       "step": 170
     },
     {
-      "epoch": 0.37676609105180536,
-      "grad_norm": 0.205755814909935,
-      "learning_rate": 7.348378303691029e-07,
-      "logits/chosen": -0.6953125,
-      "logits/rejected": -0.6796875,
-      "logps/chosen": -358.0,
-      "logps/rejected": -316.0,
-      "loss": 0.1631,
-      "rewards/accuracies": 0.5843750238418579,
-      "rewards/chosen": -0.228515625,
-      "rewards/margins": 0.130859375,
-      "rewards/rejected": -0.359375,
+      "epoch": 0.37656903765690375,
+      "grad_norm": 1.3974004983901978,
+      "learning_rate": 7.358890565349105e-07,
+      "logits/chosen": -0.90234375,
+      "logits/rejected": -0.96484375,
+      "logps/chosen": -340.0,
+      "logps/rejected": -320.0,
+      "loss": 0.6182,
+      "rewards/accuracies": 0.671875,
+      "rewards/chosen": -0.2333984375,
+      "rewards/margins": 0.216796875,
+      "rewards/rejected": -0.451171875,
       "step": 180
     },
     {
-      "epoch": 0.3976975405546834,
-      "grad_norm": 0.27961233258247375,
-      "learning_rate": 7.036850482912839e-07,
-      "logits/chosen": -0.734375,
-      "logits/rejected": -0.6171875,
-      "logps/chosen": -326.0,
-      "logps/rejected": -318.0,
-      "loss": 0.1595,
-      "rewards/accuracies": 0.653124988079071,
-      "rewards/chosen": -0.2373046875,
-      "rewards/margins": 0.169921875,
-      "rewards/rejected": -0.40625,
+      "epoch": 0.39748953974895396,
+      "grad_norm": 1.4906997680664062,
+      "learning_rate": 7.048422889997115e-07,
+      "logits/chosen": -0.96484375,
+      "logits/rejected": -0.91796875,
+      "logps/chosen": -342.0,
+      "logps/rejected": -330.0,
+      "loss": 0.632,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.2578125,
+      "rewards/margins": 0.212890625,
+      "rewards/rejected": -0.470703125,
       "step": 190
     },
     {
-      "epoch": 0.4186289900575615,
-      "grad_norm": 0.24265378713607788,
-      "learning_rate": 6.715530278447479e-07,
-      "logits/chosen": -0.76953125,
-      "logits/rejected": -0.75390625,
+      "epoch": 0.41841004184100417,
+      "grad_norm": 1.4964449405670166,
+      "learning_rate": 6.72815051360494e-07,
+      "logits/chosen": -1.03125,
+      "logits/rejected": -1.015625,
       "logps/chosen": -324.0,
-      "logps/rejected": -302.0,
-      "loss": 0.1523,
-      "rewards/accuracies": 0.6781250238418579,
-      "rewards/chosen": -0.28515625,
-      "rewards/margins": 0.13671875,
-      "rewards/rejected": -0.421875,
+      "logps/rejected": -314.0,
+      "loss": 0.6232,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.263671875,
+      "rewards/margins": 0.1962890625,
+      "rewards/rejected": -0.458984375,
       "step": 200
     },
     {
-      "epoch": 0.43956043956043955,
-      "grad_norm": 0.28898414969444275,
-      "learning_rate": 6.385962472670953e-07,
-      "logits/chosen": -0.80078125,
-      "logits/rejected": -0.765625,
-      "logps/chosen": -344.0,
-      "logps/rejected": -328.0,
-      "loss": 0.1488,
-      "rewards/accuracies": 0.671875,
-      "rewards/chosen": -0.23828125,
-      "rewards/margins": 0.22265625,
-      "rewards/rejected": -0.4609375,
+      "epoch": 0.4393305439330544,
+      "grad_norm": 1.7729754447937012,
+      "learning_rate": 6.399606408156687e-07,
+      "logits/chosen": -1.046875,
+      "logits/rejected": -1.0078125,
+      "logps/chosen": -316.0,
+      "logps/rejected": -316.0,
+      "loss": 0.6014,
+      "rewards/accuracies": 0.684374988079071,
+      "rewards/chosen": -0.2470703125,
+      "rewards/margins": 0.2431640625,
+      "rewards/rejected": -0.490234375,
       "step": 210
     },
     {
-      "epoch": 0.4604918890633176,
-      "grad_norm": 0.22757989168167114,
-      "learning_rate": 6.049731499213605e-07,
-      "logits/chosen": -0.8828125,
-      "logits/rejected": -0.84765625,
-      "logps/chosen": -358.0,
-      "logps/rejected": -324.0,
-      "loss": 0.1542,
-      "rewards/accuracies": 0.6625000238418579,
+      "epoch": 0.4602510460251046,
+      "grad_norm": 1.4692357778549194,
+      "learning_rate": 6.064363137964225e-07,
+      "logits/chosen": -1.125,
+      "logits/rejected": -0.9765625,
+      "logps/chosen": -348.0,
+      "logps/rejected": -358.0,
+      "loss": 0.6114,
+      "rewards/accuracies": 0.7124999761581421,
       "rewards/chosen": -0.267578125,
-      "rewards/margins": 0.2080078125,
-      "rewards/rejected": -0.474609375,
+      "rewards/margins": 0.30859375,
+      "rewards/rejected": -0.578125,
       "step": 220
     },
     {
-      "epoch": 0.48142333856619574,
-      "grad_norm": 0.23223264515399933,
-      "learning_rate": 5.708453825620412e-07,
-      "logits/chosen": -0.85546875,
-      "logits/rejected": -0.8203125,
-      "logps/chosen": -362.0,
-      "logps/rejected": -354.0,
-      "loss": 0.1487,
+      "epoch": 0.4811715481171548,
+      "grad_norm": 1.377569317817688,
+      "learning_rate": 5.724025332645793e-07,
+      "logits/chosen": -1.0625,
+      "logits/rejected": -1.0390625,
+      "logps/chosen": -350.0,
+      "logps/rejected": -344.0,
+      "loss": 0.6098,
       "rewards/accuracies": 0.6937500238418579,
-      "rewards/chosen": -0.30859375,
-      "rewards/margins": 0.2294921875,
-      "rewards/rejected": -0.5390625,
+      "rewards/chosen": -0.2890625,
+      "rewards/margins": 0.234375,
+      "rewards/rejected": -0.5234375,
       "step": 230
     },
     {
-      "epoch": 0.5023547880690737,
-      "grad_norm": 0.20395652949810028,
-      "learning_rate": 5.36377018200472e-07,
-      "logits/chosen": -0.86328125,
-      "logits/rejected": -0.8203125,
-      "logps/chosen": -358.0,
-      "logps/rejected": -336.0,
-      "loss": 0.1488,
-      "rewards/accuracies": 0.6468750238418579,
-      "rewards/chosen": -0.322265625,
-      "rewards/margins": 0.1943359375,
-      "rewards/rejected": -0.515625,
+      "epoch": 0.502092050209205,
+      "grad_norm": 1.5865265130996704,
+      "learning_rate": 5.380222006625179e-07,
+      "logits/chosen": -1.1171875,
+      "logits/rejected": -1.078125,
+      "logps/chosen": -334.0,
+      "logps/rejected": -328.0,
+      "loss": 0.6252,
+      "rewards/accuracies": 0.6656249761581421,
+      "rewards/chosen": -0.26953125,
+      "rewards/margins": 0.263671875,
+      "rewards/rejected": -0.53125,
       "step": 240
     },
     {
-      "epoch": 0.5232862375719518,
-      "grad_norm": 0.2625373899936676,
-      "learning_rate": 5.017337673056971e-07,
-      "logits/chosen": -0.9296875,
-      "logits/rejected": -0.8203125,
-      "logps/chosen": -326.0,
-      "logps/rejected": -328.0,
-      "loss": 0.1493,
-      "rewards/accuracies": 0.6343749761581421,
-      "rewards/chosen": -0.28515625,
-      "rewards/margins": 0.1787109375,
-      "rewards/rejected": -0.46484375,
+      "epoch": 0.5230125523012552,
+      "grad_norm": 1.5621623992919922,
+      "learning_rate": 5.034598761913916e-07,
+      "logits/chosen": -1.140625,
+      "logits/rejected": -1.1015625,
+      "logps/chosen": -318.0,
+      "logps/rejected": -318.0,
+      "loss": 0.6285,
+      "rewards/accuracies": 0.671875,
+      "rewards/chosen": -0.248046875,
+      "rewards/margins": 0.2314453125,
+      "rewards/rejected": -0.478515625,
       "step": 250
     },
     {
-      "epoch": 0.54421768707483,
-      "grad_norm": 0.21191805601119995,
-      "learning_rate": 4.6708218113308835e-07,
-      "logits/chosen": -0.92578125,
-      "logits/rejected": -0.8828125,
-      "logps/chosen": -374.0,
-      "logps/rejected": -352.0,
-      "loss": 0.143,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": -0.2578125,
-      "rewards/margins": 0.251953125,
-      "rewards/rejected": -0.51171875,
+      "epoch": 0.5439330543933054,
+      "grad_norm": 1.3323726654052734,
+      "learning_rate": 4.688809911497609e-07,
+      "logits/chosen": -1.125,
+      "logits/rejected": -1.109375,
+      "logps/chosen": -346.0,
+      "logps/rejected": -350.0,
+      "loss": 0.6079,
+      "rewards/accuracies": 0.7281249761581421,
+      "rewards/chosen": -0.240234375,
+      "rewards/margins": 0.30859375,
+      "rewards/rejected": -0.546875,
       "step": 260
     },
     {
-      "epoch": 0.565149136577708,
-      "grad_norm": 0.2124152034521103,
-      "learning_rate": 4.325888510107856e-07,
-      "logits/chosen": -1.0234375,
-      "logits/rejected": -0.95703125,
-      "logps/chosen": -336.0,
+      "epoch": 0.5648535564853556,
+      "grad_norm": 1.4896010160446167,
+      "learning_rate": 4.344510561027498e-07,
+      "logits/chosen": -1.15625,
+      "logits/rejected": -1.15625,
+      "logps/chosen": -346.0,
       "logps/rejected": -330.0,
-      "loss": 0.1302,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": -0.267578125,
-      "rewards/margins": 0.22265625,
-      "rewards/rejected": -0.490234375,
+      "loss": 0.5974,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.271484375,
+      "rewards/margins": 0.310546875,
+      "rewards/rejected": -0.58203125,
       "step": 270
     },
     {
-      "epoch": 0.5860805860805861,
-      "grad_norm": 0.20444093644618988,
-      "learning_rate": 3.9841960743348594e-07,
-      "logits/chosen": -0.984375,
-      "logits/rejected": -0.91796875,
-      "logps/chosen": -360.0,
-      "logps/rejected": -350.0,
-      "loss": 0.1426,
-      "rewards/accuracies": 0.6343749761581421,
-      "rewards/chosen": -0.298828125,
-      "rewards/margins": 0.2177734375,
-      "rewards/rejected": -0.515625,
+      "epoch": 0.5857740585774058,
+      "grad_norm": 1.4263241291046143,
+      "learning_rate": 4.003348686717949e-07,
+      "logits/chosen": -1.1953125,
+      "logits/rejected": -1.1171875,
+      "logps/chosen": -338.0,
+      "logps/rejected": -356.0,
+      "loss": 0.6141,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -0.27734375,
+      "rewards/margins": 0.306640625,
+      "rewards/rejected": -0.58203125,
       "step": 280
     },
     {
-      "epoch": 0.6070120355834642,
-      "grad_norm": 0.19832541048526764,
-      "learning_rate": 3.647387228140232e-07,
-      "logits/chosen": -1.015625,
-      "logits/rejected": -0.8984375,
-      "logps/chosen": -340.0,
-      "logps/rejected": -318.0,
-      "loss": 0.1374,
-      "rewards/accuracies": 0.715624988079071,
-      "rewards/chosen": -0.263671875,
-      "rewards/margins": 0.251953125,
-      "rewards/rejected": -0.515625,
+      "epoch": 0.606694560669456,
+      "grad_norm": 1.510299563407898,
+      "learning_rate": 3.666957247368757e-07,
+      "logits/chosen": -1.109375,
+      "logits/rejected": -1.078125,
+      "logps/chosen": -338.0,
+      "logps/rejected": -322.0,
+      "loss": 0.5992,
+      "rewards/accuracies": 0.7281249761581421,
+      "rewards/chosen": -0.2451171875,
+      "rewards/margins": 0.345703125,
+      "rewards/rejected": -0.58984375,
       "step": 290
     },
     {
-      "epoch": 0.6279434850863422,
-      "grad_norm": 0.3114321231842041,
-      "learning_rate": 3.317081217255969e-07,
-      "logits/chosen": -0.98046875,
-      "logits/rejected": -0.88671875,
-      "logps/chosen": -366.0,
-      "logps/rejected": -350.0,
-      "loss": 0.1419,
-      "rewards/accuracies": 0.6656249761581421,
-      "rewards/chosen": -0.322265625,
-      "rewards/margins": 0.2197265625,
-      "rewards/rejected": -0.54296875,
+      "epoch": 0.6276150627615062,
+      "grad_norm": 1.8962136507034302,
+      "learning_rate": 3.3369463682677234e-07,
+      "logits/chosen": -1.1875,
+      "logits/rejected": -1.1484375,
+      "logps/chosen": -362.0,
+      "logps/rejected": -362.0,
+      "loss": 0.6158,
+      "rewards/accuracies": 0.7281249761581421,
+      "rewards/chosen": -0.267578125,
+      "rewards/margins": 0.32421875,
+      "rewards/rejected": -0.59375,
       "step": 300
     },
     {
-      "epoch": 0.6488749345892203,
-      "grad_norm": 0.2070474773645401,
-      "learning_rate": 2.994866024315009e-07,
-      "logits/chosen": -1.0,
-      "logits/rejected": -0.921875,
-      "logps/chosen": -344.0,
-      "logps/rejected": -368.0,
-      "loss": 0.1384,
-      "rewards/accuracies": 0.6812499761581421,
-      "rewards/chosen": -0.294921875,
-      "rewards/margins": 0.2373046875,
-      "rewards/rejected": -0.53125,
+      "epoch": 0.6485355648535565,
+      "grad_norm": 1.5493396520614624,
+      "learning_rate": 3.014895634385014e-07,
+      "logits/chosen": -1.109375,
+      "logits/rejected": -1.140625,
+      "logps/chosen": -352.0,
+      "logps/rejected": -340.0,
+      "loss": 0.601,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.306640625,
+      "rewards/margins": 0.2333984375,
+      "rewards/rejected": -0.5390625,
       "step": 310
     },
     {
-      "epoch": 0.6698063840920984,
-      "grad_norm": 0.22273701429367065,
-      "learning_rate": 2.682290734449314e-07,
-      "logits/chosen": -1.046875,
-      "logits/rejected": -0.92578125,
-      "logps/chosen": -340.0,
-      "logps/rejected": -350.0,
-      "loss": 0.1346,
-      "rewards/accuracies": 0.71875,
-      "rewards/chosen": -0.2890625,
-      "rewards/margins": 0.259765625,
-      "rewards/rejected": -0.546875,
+      "epoch": 0.6694560669456067,
+      "grad_norm": 1.5511579513549805,
+      "learning_rate": 2.7023465297476424e-07,
+      "logits/chosen": -1.15625,
+      "logits/rejected": -1.15625,
+      "logps/chosen": -354.0,
+      "logps/rejected": -358.0,
+      "loss": 0.595,
+      "rewards/accuracies": 0.690625011920929,
+      "rewards/chosen": -0.283203125,
+      "rewards/margins": 0.326171875,
+      "rewards/rejected": -0.609375,
       "step": 320
     },
     {
-      "epoch": 0.6907378335949764,
-      "grad_norm": 0.21851293742656708,
-      "learning_rate": 2.3808580878919941e-07,
-      "logits/chosen": -0.94921875,
-      "logits/rejected": -0.859375,
-      "logps/chosen": -342.0,
-      "logps/rejected": -338.0,
-      "loss": 0.138,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": -0.3203125,
-      "rewards/margins": 0.232421875,
-      "rewards/rejected": -0.55078125,
+      "epoch": 0.6903765690376569,
+      "grad_norm": 1.7871730327606201,
+      "learning_rate": 2.4007950591826913e-07,
+      "logits/chosen": -1.1328125,
+      "logits/rejected": -1.125,
+      "logps/chosen": -350.0,
+      "logps/rejected": -352.0,
+      "loss": 0.612,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -0.330078125,
+      "rewards/margins": 0.30078125,
+      "rewards/rejected": -0.6328125,
       "step": 330
     },
     {
-      "epoch": 0.7116692830978545,
-      "grad_norm": 0.21169716119766235,
-      "learning_rate": 2.0920172553876285e-07,
-      "logits/chosen": -0.97265625,
-      "logits/rejected": -0.91015625,
-      "logps/chosen": -336.0,
-      "logps/rejected": -334.0,
-      "loss": 0.1417,
-      "rewards/accuracies": 0.6812499761581421,
-      "rewards/chosen": -0.298828125,
-      "rewards/margins": 0.27734375,
-      "rewards/rejected": -0.57421875,
+      "epoch": 0.7112970711297071,
+      "grad_norm": 1.6344778537750244,
+      "learning_rate": 2.1116845877450805e-07,
+      "logits/chosen": -1.1953125,
+      "logits/rejected": -1.171875,
+      "logps/chosen": -330.0,
+      "logps/rejected": -316.0,
+      "loss": 0.6125,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -0.32421875,
+      "rewards/margins": 0.255859375,
+      "rewards/rejected": -0.58203125,
       "step": 340
     },
     {
-      "epoch": 0.7326007326007326,
-      "grad_norm": 0.19472256302833557,
-      "learning_rate": 1.8171568711438512e-07,
-      "logits/chosen": -1.0390625,
-      "logits/rejected": -0.88671875,
-      "logps/chosen": -334.0,
-      "logps/rejected": -330.0,
-      "loss": 0.1372,
-      "rewards/accuracies": 0.640625,
-      "rewards/chosen": -0.30078125,
-      "rewards/margins": 0.1943359375,
-      "rewards/rejected": -0.49609375,
+      "epoch": 0.7322175732217573,
+      "grad_norm": 1.6645337343215942,
+      "learning_rate": 1.8363989321036577e-07,
+      "logits/chosen": -1.296875,
+      "logits/rejected": -1.203125,
+      "logps/chosen": -332.0,
+      "logps/rejected": -340.0,
+      "loss": 0.6038,
+      "rewards/accuracies": 0.671875,
+      "rewards/chosen": -0.333984375,
+      "rewards/margins": 0.26953125,
+      "rewards/rejected": -0.6015625,
       "step": 350
     },
     {
-      "epoch": 0.7535321821036107,
-      "grad_norm": 0.23481227457523346,
-      "learning_rate": 1.5575983568189998e-07,
-      "logits/chosen": -0.953125,
-      "logits/rejected": -0.98828125,
+      "epoch": 0.7531380753138075,
+      "grad_norm": 1.5749666690826416,
+      "learning_rate": 1.5762557369534708e-07,
+      "logits/chosen": -1.15625,
+      "logits/rejected": -1.1875,
       "logps/chosen": -322.0,
-      "logps/rejected": -324.0,
-      "loss": 0.1345,
+      "logps/rejected": -348.0,
+      "loss": 0.6019,
       "rewards/accuracies": 0.643750011920929,
-      "rewards/chosen": -0.330078125,
-      "rewards/margins": 0.208984375,
-      "rewards/rejected": -0.5390625,
+      "rewards/chosen": -0.38671875,
+      "rewards/margins": 0.265625,
+      "rewards/rejected": -0.65234375,
       "step": 360
     },
     {
-      "epoch": 0.7744636316064888,
-      "grad_norm": 0.20973995327949524,
-      "learning_rate": 1.3145895686415353e-07,
-      "logits/chosen": -0.92578125,
-      "logits/rejected": -0.9375,
-      "logps/chosen": -372.0,
-      "logps/rejected": -332.0,
-      "loss": 0.1378,
-      "rewards/accuracies": 0.628125011920929,
-      "rewards/chosen": -0.33203125,
-      "rewards/margins": 0.1865234375,
-      "rewards/rejected": -0.515625,
+      "epoch": 0.7740585774058577,
+      "grad_norm": 1.582518458366394,
+      "learning_rate": 1.332500168157748e-07,
+      "logits/chosen": -1.15625,
+      "logits/rejected": -1.203125,
+      "logps/chosen": -360.0,
+      "logps/rejected": -348.0,
+      "loss": 0.6147,
+      "rewards/accuracies": 0.6656249761581421,
+      "rewards/chosen": -0.412109375,
+      "rewards/margins": 0.25390625,
+      "rewards/rejected": -0.6640625,
       "step": 370
     },
     {
-      "epoch": 0.7953950811093669,
-      "grad_norm": 0.18481357395648956,
-      "learning_rate": 1.0892987982033757e-07,
-      "logits/chosen": -0.92578125,
-      "logits/rejected": -0.92578125,
-      "logps/chosen": -354.0,
-      "logps/rejected": -356.0,
-      "loss": 0.1338,
-      "rewards/accuracies": 0.6656249761581421,
-      "rewards/chosen": -0.3125,
-      "rewards/margins": 0.2392578125,
-      "rewards/rejected": -0.55078125,
+      "epoch": 0.7949790794979079,
+      "grad_norm": 1.690048098564148,
+      "learning_rate": 1.1062989528071681e-07,
+      "logits/chosen": -1.1953125,
+      "logits/rejected": -1.1796875,
+      "logps/chosen": -364.0,
+      "logps/rejected": -378.0,
+      "loss": 0.6108,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -0.349609375,
+      "rewards/margins": 0.2890625,
+      "rewards/rejected": -0.63671875,
       "step": 380
     },
     {
-      "epoch": 0.8163265306122449,
-      "grad_norm": 0.22186759114265442,
-      "learning_rate": 8.828091557690287e-08,
-      "logits/chosen": -0.90234375,
-      "logits/rejected": -0.875,
-      "logps/chosen": -346.0,
-      "logps/rejected": -328.0,
-      "loss": 0.1304,
-      "rewards/accuracies": 0.65625,
-      "rewards/chosen": -0.330078125,
-      "rewards/margins": 0.2177734375,
-      "rewards/rejected": -0.546875,
+      "epoch": 0.8158995815899581,
+      "grad_norm": 1.6943994760513306,
+      "learning_rate": 8.987347947234192e-08,
+      "logits/chosen": -1.109375,
+      "logits/rejected": -1.1328125,
+      "logps/chosen": -328.0,
+      "logps/rejected": -324.0,
+      "loss": 0.6044,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -0.3203125,
+      "rewards/margins": 0.322265625,
+      "rewards/rejected": -0.64453125,
       "step": 390
     },
     {
-      "epoch": 0.837257980115123,
-      "grad_norm": 0.2037443071603775,
-      "learning_rate": 6.961133631033511e-08,
-      "logits/chosen": -0.9296875,
-      "logits/rejected": -0.89453125,
-      "logps/chosen": -352.0,
-      "logps/rejected": -370.0,
-      "loss": 0.134,
-      "rewards/accuracies": 0.684374988079071,
-      "rewards/chosen": -0.333984375,
-      "rewards/margins": 0.240234375,
-      "rewards/rejected": -0.57421875,
+      "epoch": 0.8368200836820083,
+      "grad_norm": 1.6517119407653809,
+      "learning_rate": 7.108011921370727e-08,
+      "logits/chosen": -1.1875,
+      "logits/rejected": -1.1640625,
+      "logps/chosen": -386.0,
+      "logps/rejected": -382.0,
+      "loss": 0.6041,
+      "rewards/accuracies": 0.715624988079071,
+      "rewards/chosen": -0.375,
+      "rewards/margins": 0.328125,
+      "rewards/rejected": -0.703125,
       "step": 400
     },
     {
-      "epoch": 0.858189429618001,
-      "grad_norm": 0.22719213366508484,
-      "learning_rate": 5.301089808520048e-08,
-      "logits/chosen": -0.98046875,
-      "logits/rejected": -0.8984375,
-      "logps/chosen": -354.0,
-      "logps/rejected": -338.0,
-      "loss": 0.1353,
-      "rewards/accuracies": 0.671875,
-      "rewards/chosen": -0.314453125,
-      "rewards/margins": 0.259765625,
-      "rewards/rejected": -0.57421875,
+      "epoch": 0.8577405857740585,
+      "grad_norm": 1.6941229104995728,
+      "learning_rate": 5.433976823447262e-08,
+      "logits/chosen": -1.203125,
+      "logits/rejected": -1.1015625,
+      "logps/chosen": -340.0,
+      "logps/rejected": -352.0,
+      "loss": 0.6026,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -0.392578125,
+      "rewards/margins": 0.333984375,
+      "rewards/rejected": -0.7265625,
       "step": 410
     },
     {
-      "epoch": 0.8791208791208791,
-      "grad_norm": 0.20336544513702393,
-      "learning_rate": 3.855940934195145e-08,
-      "logits/chosen": -0.984375,
-      "logits/rejected": -0.91796875,
-      "logps/chosen": -350.0,
-      "logps/rejected": -338.0,
-      "loss": 0.133,
-      "rewards/accuracies": 0.6812499761581421,
-      "rewards/chosen": -0.294921875,
-      "rewards/margins": 0.263671875,
-      "rewards/rejected": -0.55859375,
+      "epoch": 0.8786610878661087,
+      "grad_norm": 1.6011600494384766,
+      "learning_rate": 3.973255361067346e-08,
+      "logits/chosen": -1.2265625,
+      "logits/rejected": -1.1875,
+      "logps/chosen": -362.0,
+      "logps/rejected": -352.0,
+      "loss": 0.5924,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -0.34375,
+      "rewards/margins": 0.2890625,
+      "rewards/rejected": -0.6328125,
       "step": 420
     },
     {
-      "epoch": 0.9000523286237572,
-      "grad_norm": 0.2312639355659485,
-      "learning_rate": 2.6326347209032817e-08,
-      "logits/chosen": -0.890625,
-      "logits/rejected": -0.875,
-      "logps/chosen": -354.0,
-      "logps/rejected": -358.0,
-      "loss": 0.1337,
-      "rewards/accuracies": 0.668749988079071,
-      "rewards/chosen": -0.328125,
-      "rewards/margins": 0.2451171875,
-      "rewards/rejected": -0.57421875,
+      "epoch": 0.899581589958159,
+      "grad_norm": 1.7107675075531006,
+      "learning_rate": 2.732839223940914e-08,
+      "logits/chosen": -1.1484375,
+      "logits/rejected": -1.1953125,
+      "logps/chosen": -356.0,
+      "logps/rejected": -366.0,
+      "loss": 0.5962,
+      "rewards/accuracies": 0.684374988079071,
+      "rewards/chosen": -0.373046875,
+      "rewards/margins": 0.3671875,
+      "rewards/rejected": -0.7421875,
       "step": 430
     },
     {
-      "epoch": 0.9209837781266352,
-      "grad_norm": 0.21316397190093994,
-      "learning_rate": 1.63705234839131e-08,
-      "logits/chosen": -0.96484375,
-      "logits/rejected": -0.95703125,
-      "logps/chosen": -346.0,
-      "logps/rejected": -322.0,
-      "loss": 0.1381,
-      "rewards/accuracies": 0.640625,
-      "rewards/chosen": -0.333984375,
-      "rewards/margins": 0.234375,
-      "rewards/rejected": -0.56640625,
+      "epoch": 0.9205020920502092,
+      "grad_norm": 1.5065439939498901,
+      "learning_rate": 1.7186656184179473e-08,
+      "logits/chosen": -1.203125,
+      "logits/rejected": -1.2109375,
+      "logps/chosen": -364.0,
+      "logps/rejected": -374.0,
+      "loss": 0.6021,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -0.361328125,
+      "rewards/margins": 0.373046875,
+      "rewards/rejected": -0.734375,
       "step": 440
     },
     {
-      "epoch": 0.9419152276295133,
-      "grad_norm": 0.18119490146636963,
-      "learning_rate": 8.739801888871468e-09,
-      "logits/chosen": -0.98828125,
-      "logits/rejected": -0.91796875,
-      "logps/chosen": -364.0,
-      "logps/rejected": -342.0,
-      "loss": 0.1327,
-      "rewards/accuracies": 0.703125,
-      "rewards/chosen": -0.25,
-      "rewards/margins": 0.279296875,
-      "rewards/rejected": -0.52734375,
+      "epoch": 0.9414225941422594,
+      "grad_norm": 1.560874104499817,
+      "learning_rate": 9.355888492680153e-09,
+      "logits/chosen": -1.1328125,
+      "logits/rejected": -1.1015625,
+      "logps/chosen": -340.0,
+      "logps/rejected": -346.0,
+      "loss": 0.5957,
+      "rewards/accuracies": 0.721875011920929,
+      "rewards/chosen": -0.337890625,
+      "rewards/margins": 0.357421875,
+      "rewards/rejected": -0.6953125,
       "step": 450
     },
     {
-      "epoch": 0.9628466771323915,
-      "grad_norm": 0.21448834240436554,
-      "learning_rate": 3.47086796086582e-09,
-      "logits/chosen": -1.03125,
-      "logits/rejected": -0.984375,
-      "logps/chosen": -334.0,
-      "logps/rejected": -330.0,
-      "loss": 0.1264,
-      "rewards/accuracies": 0.6656249761581421,
-      "rewards/chosen": -0.3046875,
-      "rewards/margins": 0.271484375,
-      "rewards/rejected": -0.57421875,
+      "epoch": 0.9623430962343096,
+      "grad_norm": 1.5713697671890259,
+      "learning_rate": 3.873570847285012e-09,
+      "logits/chosen": -1.234375,
+      "logits/rejected": -1.2109375,
+      "logps/chosen": -356.0,
+      "logps/rejected": -356.0,
+      "loss": 0.6004,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -0.40234375,
+      "rewards/margins": 0.30078125,
+      "rewards/rejected": -0.703125,
       "step": 460
     },
     {
-      "epoch": 0.9837781266352695,
-      "grad_norm": 0.20449356734752655,
-      "learning_rate": 5.89052681761626e-10,
-      "logits/chosen": -0.96484375,
-      "logits/rejected": -0.87109375,
-      "logps/chosen": -364.0,
-      "logps/rejected": -356.0,
-      "loss": 0.1367,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": -0.3125,
-      "rewards/margins": 0.240234375,
-      "rewards/rejected": -0.5546875,
+      "epoch": 0.9832635983263598,
+      "grad_norm": 1.5760387182235718,
+      "learning_rate": 7.65944160348142e-10,
+      "logits/chosen": -1.2109375,
+      "logits/rejected": -1.140625,
+      "logps/chosen": -378.0,
+      "logps/rejected": -374.0,
+      "loss": 0.6114,
+      "rewards/accuracies": 0.6656249761581421,
+      "rewards/chosen": -0.3515625,
+      "rewards/margins": 0.330078125,
+      "rewards/rejected": -0.6796875,
       "step": 470
     }
   ],
   "logging_steps": 10,
-  "max_steps": 477,
+  "max_steps": 478,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 1,
   "save_steps": 500,
@@ -732,7 +732,7 @@
     }
   },
   "total_flos": 0.0,
-  "train_batch_size": 8,
+  "train_batch_size": 16,
   "trial_name": null,
   "trial_params": null
 }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:38c0461a91737044430afbb4a7f3c236b5421ede367eb2b8033902f52041e7f3
+oid sha256:427ee2517e8c8887ce28026fa431a455f8c0d1977b412aca06f9d02da7f8f817
 size 7608