{ "best_metric": null, "best_model_checkpoint": null, "epoch": 3.0, "eval_steps": 500, "global_step": 939, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.003194888178913738, "grad_norm": 23.348649558806766, "learning_rate": 5.3191489361702125e-09, "logits/chosen": -1.515625, "logits/rejected": -1.640625, "logps/chosen": -251.0, "logps/rejected": -282.0, "loss": 0.6914, "rewards/accuracies": 0.0, "rewards/chosen": 0.0, "rewards/margins": 0.0, "rewards/rejected": 0.0, "step": 1 }, { "epoch": 0.03194888178913738, "grad_norm": 25.566113326100368, "learning_rate": 5.3191489361702123e-08, "logits/chosen": -1.4765625, "logits/rejected": -1.5078125, "logps/chosen": -242.0, "logps/rejected": -264.0, "loss": 0.6916, "rewards/accuracies": 0.1527777761220932, "rewards/chosen": 0.0069580078125, "rewards/margins": 0.00836181640625, "rewards/rejected": -0.0013885498046875, "step": 10 }, { "epoch": 0.06389776357827476, "grad_norm": 28.018336198539554, "learning_rate": 1.0638297872340425e-07, "logits/chosen": -1.484375, "logits/rejected": -1.5234375, "logps/chosen": -235.0, "logps/rejected": -270.0, "loss": 0.6909, "rewards/accuracies": 0.2874999940395355, "rewards/chosen": 0.00250244140625, "rewards/margins": 0.0125732421875, "rewards/rejected": -0.010009765625, "step": 20 }, { "epoch": 0.09584664536741214, "grad_norm": 23.40378981395217, "learning_rate": 1.5957446808510638e-07, "logits/chosen": -1.4375, "logits/rejected": -1.5234375, "logps/chosen": -238.0, "logps/rejected": -258.0, "loss": 0.6869, "rewards/accuracies": 0.36250001192092896, "rewards/chosen": 0.003753662109375, "rewards/margins": 0.0250244140625, "rewards/rejected": -0.021240234375, "step": 30 }, { "epoch": 0.12779552715654952, "grad_norm": 23.492853372798173, "learning_rate": 2.127659574468085e-07, "logits/chosen": -1.53125, "logits/rejected": -1.546875, "logps/chosen": -238.0, "logps/rejected": -260.0, "loss": 0.6656, "rewards/accuracies": 0.574999988079071, "rewards/chosen": -1.2218952178955078e-05, "rewards/margins": 0.06982421875, "rewards/rejected": -0.0703125, "step": 40 }, { "epoch": 0.1597444089456869, "grad_norm": 22.007025309048732, "learning_rate": 2.659574468085106e-07, "logits/chosen": -1.5, "logits/rejected": -1.5390625, "logps/chosen": -242.0, "logps/rejected": -258.0, "loss": 0.6634, "rewards/accuracies": 0.550000011920929, "rewards/chosen": -0.0224609375, "rewards/margins": 0.041259765625, "rewards/rejected": -0.06396484375, "step": 50 }, { "epoch": 0.19169329073482427, "grad_norm": 24.40454844081895, "learning_rate": 3.1914893617021275e-07, "logits/chosen": -1.5, "logits/rejected": -1.5, "logps/chosen": -240.0, "logps/rejected": -262.0, "loss": 0.6357, "rewards/accuracies": 0.6875, "rewards/chosen": 0.076171875, "rewards/margins": 0.181640625, "rewards/rejected": -0.10498046875, "step": 60 }, { "epoch": 0.22364217252396165, "grad_norm": 22.502898330526865, "learning_rate": 3.7234042553191484e-07, "logits/chosen": -1.5078125, "logits/rejected": -1.578125, "logps/chosen": -237.0, "logps/rejected": -268.0, "loss": 0.6089, "rewards/accuracies": 0.699999988079071, "rewards/chosen": 0.1455078125, "rewards/margins": 0.2109375, "rewards/rejected": -0.06494140625, "step": 70 }, { "epoch": 0.25559105431309903, "grad_norm": 26.18333762460293, "learning_rate": 4.25531914893617e-07, "logits/chosen": -1.5078125, "logits/rejected": -1.53125, "logps/chosen": -237.0, "logps/rejected": -262.0, "loss": 0.6312, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": 0.1875, "rewards/margins": 0.240234375, "rewards/rejected": -0.052734375, "step": 80 }, { "epoch": 0.28753993610223644, "grad_norm": 20.53332137935602, "learning_rate": 4.787234042553192e-07, "logits/chosen": -1.5390625, "logits/rejected": -1.46875, "logps/chosen": -236.0, "logps/rejected": -258.0, "loss": 0.6029, "rewards/accuracies": 0.6875, "rewards/chosen": 0.3046875, "rewards/margins": 0.3125, "rewards/rejected": -0.00872802734375, "step": 90 }, { "epoch": 0.3194888178913738, "grad_norm": 19.45254874610382, "learning_rate": 4.964497041420119e-07, "logits/chosen": -1.5390625, "logits/rejected": -1.4921875, "logps/chosen": -238.0, "logps/rejected": -252.0, "loss": 0.5962, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": 0.37109375, "rewards/margins": 0.306640625, "rewards/rejected": 0.0625, "step": 100 }, { "epoch": 0.3514376996805112, "grad_norm": 18.9229075549186, "learning_rate": 4.905325443786982e-07, "logits/chosen": -1.5078125, "logits/rejected": -1.5546875, "logps/chosen": -234.0, "logps/rejected": -262.0, "loss": 0.574, "rewards/accuracies": 0.6625000238418579, "rewards/chosen": 0.470703125, "rewards/margins": 0.4375, "rewards/rejected": 0.03369140625, "step": 110 }, { "epoch": 0.38338658146964855, "grad_norm": 20.49193024317671, "learning_rate": 4.846153846153846e-07, "logits/chosen": -1.4609375, "logits/rejected": -1.53125, "logps/chosen": -233.0, "logps/rejected": -262.0, "loss": 0.5877, "rewards/accuracies": 0.737500011920929, "rewards/chosen": 0.45703125, "rewards/margins": 0.431640625, "rewards/rejected": 0.0262451171875, "step": 120 }, { "epoch": 0.41533546325878595, "grad_norm": 25.979970922508116, "learning_rate": 4.78698224852071e-07, "logits/chosen": -1.5, "logits/rejected": -1.546875, "logps/chosen": -231.0, "logps/rejected": -264.0, "loss": 0.5812, "rewards/accuracies": 0.762499988079071, "rewards/chosen": 0.54296875, "rewards/margins": 0.5390625, "rewards/rejected": 0.002532958984375, "step": 130 }, { "epoch": 0.4472843450479233, "grad_norm": 23.03890136178023, "learning_rate": 4.727810650887574e-07, "logits/chosen": -1.5390625, "logits/rejected": -1.5390625, "logps/chosen": -232.0, "logps/rejected": -264.0, "loss": 0.587, "rewards/accuracies": 0.6875, "rewards/chosen": 0.54296875, "rewards/margins": 0.54296875, "rewards/rejected": -0.00138092041015625, "step": 140 }, { "epoch": 0.4792332268370607, "grad_norm": 19.933395132964325, "learning_rate": 4.668639053254438e-07, "logits/chosen": -1.5625, "logits/rejected": -1.59375, "logps/chosen": -228.0, "logps/rejected": -258.0, "loss": 0.5603, "rewards/accuracies": 0.699999988079071, "rewards/chosen": 0.60546875, "rewards/margins": 0.5078125, "rewards/rejected": 0.09765625, "step": 150 }, { "epoch": 0.5111821086261981, "grad_norm": 22.802957126994585, "learning_rate": 4.6094674556213014e-07, "logits/chosen": -1.5546875, "logits/rejected": -1.484375, "logps/chosen": -235.0, "logps/rejected": -258.0, "loss": 0.5875, "rewards/accuracies": 0.5874999761581421, "rewards/chosen": 0.4296875, "rewards/margins": 0.4140625, "rewards/rejected": 0.0150146484375, "step": 160 }, { "epoch": 0.5431309904153354, "grad_norm": 19.859260131030616, "learning_rate": 4.5502958579881655e-07, "logits/chosen": -1.4921875, "logits/rejected": -1.4765625, "logps/chosen": -236.0, "logps/rejected": -262.0, "loss": 0.5631, "rewards/accuracies": 0.7124999761581421, "rewards/chosen": 0.40625, "rewards/margins": 0.4609375, "rewards/rejected": -0.052734375, "step": 170 }, { "epoch": 0.5750798722044729, "grad_norm": 20.463633881655507, "learning_rate": 4.491124260355029e-07, "logits/chosen": -1.5, "logits/rejected": -1.546875, "logps/chosen": -240.0, "logps/rejected": -260.0, "loss": 0.5983, "rewards/accuracies": 0.5375000238418579, "rewards/chosen": 0.32421875, "rewards/margins": 0.205078125, "rewards/rejected": 0.119140625, "step": 180 }, { "epoch": 0.6070287539936102, "grad_norm": 29.020367112295766, "learning_rate": 4.4319526627218936e-07, "logits/chosen": -1.4375, "logits/rejected": -1.515625, "logps/chosen": -230.0, "logps/rejected": -266.0, "loss": 0.5451, "rewards/accuracies": 0.7124999761581421, "rewards/chosen": 0.46484375, "rewards/margins": 0.51171875, "rewards/rejected": -0.0478515625, "step": 190 }, { "epoch": 0.6389776357827476, "grad_norm": 21.133863882283315, "learning_rate": 4.372781065088757e-07, "logits/chosen": -1.609375, "logits/rejected": -1.5703125, "logps/chosen": -238.0, "logps/rejected": -266.0, "loss": 0.5422, "rewards/accuracies": 0.6625000238418579, "rewards/chosen": 0.46484375, "rewards/margins": 0.5546875, "rewards/rejected": -0.091796875, "step": 200 }, { "epoch": 0.670926517571885, "grad_norm": 22.488542820957655, "learning_rate": 4.313609467455621e-07, "logits/chosen": -1.5390625, "logits/rejected": -1.546875, "logps/chosen": -237.0, "logps/rejected": -260.0, "loss": 0.543, "rewards/accuracies": 0.637499988079071, "rewards/chosen": 0.5546875, "rewards/margins": 0.49609375, "rewards/rejected": 0.057373046875, "step": 210 }, { "epoch": 0.7028753993610224, "grad_norm": 18.647250006144112, "learning_rate": 4.2544378698224847e-07, "logits/chosen": -1.546875, "logits/rejected": -1.5859375, "logps/chosen": -236.0, "logps/rejected": -252.0, "loss": 0.5369, "rewards/accuracies": 0.6625000238418579, "rewards/chosen": 0.62109375, "rewards/margins": 0.55859375, "rewards/rejected": 0.061279296875, "step": 220 }, { "epoch": 0.7348242811501597, "grad_norm": 20.851449008764483, "learning_rate": 4.195266272189349e-07, "logits/chosen": -1.578125, "logits/rejected": -1.65625, "logps/chosen": -232.0, "logps/rejected": -260.0, "loss": 0.5197, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": 0.69921875, "rewards/margins": 0.6640625, "rewards/rejected": 0.037353515625, "step": 230 }, { "epoch": 0.7667731629392971, "grad_norm": 20.259258580839976, "learning_rate": 4.1360946745562133e-07, "logits/chosen": -1.6015625, "logits/rejected": -1.609375, "logps/chosen": -234.0, "logps/rejected": -270.0, "loss": 0.5115, "rewards/accuracies": 0.762499988079071, "rewards/chosen": 0.6015625, "rewards/margins": 0.74609375, "rewards/rejected": -0.1435546875, "step": 240 }, { "epoch": 0.7987220447284346, "grad_norm": 19.17384724341236, "learning_rate": 4.076923076923077e-07, "logits/chosen": -1.5859375, "logits/rejected": -1.5859375, "logps/chosen": -236.0, "logps/rejected": -260.0, "loss": 0.5275, "rewards/accuracies": 0.7875000238418579, "rewards/chosen": 0.56640625, "rewards/margins": 0.71875, "rewards/rejected": -0.154296875, "step": 250 }, { "epoch": 0.8306709265175719, "grad_norm": 18.509443199741717, "learning_rate": 4.017751479289941e-07, "logits/chosen": -1.5234375, "logits/rejected": -1.59375, "logps/chosen": -227.0, "logps/rejected": -262.0, "loss": 0.506, "rewards/accuracies": 0.800000011920929, "rewards/chosen": 0.5625, "rewards/margins": 0.6796875, "rewards/rejected": -0.1162109375, "step": 260 }, { "epoch": 0.8626198083067093, "grad_norm": 17.20437875077693, "learning_rate": 3.9585798816568044e-07, "logits/chosen": -1.5625, "logits/rejected": -1.609375, "logps/chosen": -225.0, "logps/rejected": -268.0, "loss": 0.4924, "rewards/accuracies": 0.737500011920929, "rewards/chosen": 0.75, "rewards/margins": 0.88671875, "rewards/rejected": -0.138671875, "step": 270 }, { "epoch": 0.8945686900958466, "grad_norm": 17.883187237285338, "learning_rate": 3.8994082840236685e-07, "logits/chosen": -1.5390625, "logits/rejected": -1.578125, "logps/chosen": -236.0, "logps/rejected": -264.0, "loss": 0.5164, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": 0.578125, "rewards/margins": 0.7109375, "rewards/rejected": -0.1318359375, "step": 280 }, { "epoch": 0.9265175718849841, "grad_norm": 21.48369555840101, "learning_rate": 3.840236686390532e-07, "logits/chosen": -1.5390625, "logits/rejected": -1.5625, "logps/chosen": -233.0, "logps/rejected": -256.0, "loss": 0.5068, "rewards/accuracies": 0.800000011920929, "rewards/chosen": 0.515625, "rewards/margins": 0.78125, "rewards/rejected": -0.267578125, "step": 290 }, { "epoch": 0.9584664536741214, "grad_norm": 19.429652131898607, "learning_rate": 3.7810650887573966e-07, "logits/chosen": -1.5859375, "logits/rejected": -1.546875, "logps/chosen": -241.0, "logps/rejected": -266.0, "loss": 0.5271, "rewards/accuracies": 0.75, "rewards/chosen": 0.466796875, "rewards/margins": 0.68359375, "rewards/rejected": -0.216796875, "step": 300 }, { "epoch": 0.9904153354632588, "grad_norm": 20.959288044820543, "learning_rate": 3.72189349112426e-07, "logits/chosen": -1.5390625, "logits/rejected": -1.578125, "logps/chosen": -234.0, "logps/rejected": -260.0, "loss": 0.511, "rewards/accuracies": 0.75, "rewards/chosen": 0.51953125, "rewards/margins": 0.734375, "rewards/rejected": -0.216796875, "step": 310 }, { "epoch": 1.0, "eval_logits/chosen": -1.5625, "eval_logits/rejected": -1.5390625, "eval_logps/chosen": -240.0, "eval_logps/rejected": -255.0, "eval_loss": 0.592968761920929, "eval_rewards/accuracies": 0.6428571343421936, "eval_rewards/chosen": 0.384765625, "eval_rewards/margins": 0.53125, "eval_rewards/rejected": -0.14453125, "eval_runtime": 12.2023, "eval_samples_per_second": 16.39, "eval_steps_per_second": 0.574, "step": 313 }, { "epoch": 1.0223642172523961, "grad_norm": 15.216502023078874, "learning_rate": 3.662721893491124e-07, "logits/chosen": -1.578125, "logits/rejected": -1.59375, "logps/chosen": -230.0, "logps/rejected": -262.0, "loss": 0.4383, "rewards/accuracies": 0.8500000238418579, "rewards/chosen": 0.66015625, "rewards/margins": 1.0546875, "rewards/rejected": -0.392578125, "step": 320 }, { "epoch": 1.0543130990415335, "grad_norm": 16.390892234984523, "learning_rate": 3.603550295857988e-07, "logits/chosen": -1.4921875, "logits/rejected": -1.59375, "logps/chosen": -232.0, "logps/rejected": -276.0, "loss": 0.3885, "rewards/accuracies": 0.9375, "rewards/chosen": 0.6328125, "rewards/margins": 1.2578125, "rewards/rejected": -0.625, "step": 330 }, { "epoch": 1.0862619808306708, "grad_norm": 14.29774575781621, "learning_rate": 3.544378698224852e-07, "logits/chosen": -1.5390625, "logits/rejected": -1.5546875, "logps/chosen": -228.0, "logps/rejected": -268.0, "loss": 0.3978, "rewards/accuracies": 0.875, "rewards/chosen": 0.63671875, "rewards/margins": 1.2265625, "rewards/rejected": -0.58984375, "step": 340 }, { "epoch": 1.1182108626198084, "grad_norm": 16.80859709996563, "learning_rate": 3.485207100591716e-07, "logits/chosen": -1.5390625, "logits/rejected": -1.5078125, "logps/chosen": -236.0, "logps/rejected": -268.0, "loss": 0.3711, "rewards/accuracies": 0.8374999761581421, "rewards/chosen": 0.50390625, "rewards/margins": 1.0859375, "rewards/rejected": -0.58203125, "step": 350 }, { "epoch": 1.1501597444089458, "grad_norm": 17.099448590902114, "learning_rate": 3.42603550295858e-07, "logits/chosen": -1.53125, "logits/rejected": -1.5625, "logps/chosen": -228.0, "logps/rejected": -262.0, "loss": 0.3884, "rewards/accuracies": 0.800000011920929, "rewards/chosen": 0.515625, "rewards/margins": 1.0546875, "rewards/rejected": -0.54296875, "step": 360 }, { "epoch": 1.182108626198083, "grad_norm": 15.964788390337974, "learning_rate": 3.366863905325444e-07, "logits/chosen": -1.5625, "logits/rejected": -1.5625, "logps/chosen": -234.0, "logps/rejected": -266.0, "loss": 0.3964, "rewards/accuracies": 0.824999988079071, "rewards/chosen": 0.5546875, "rewards/margins": 0.97265625, "rewards/rejected": -0.416015625, "step": 370 }, { "epoch": 1.2140575079872205, "grad_norm": 17.10451165088142, "learning_rate": 3.3076923076923075e-07, "logits/chosen": -1.5546875, "logits/rejected": -1.5703125, "logps/chosen": -232.0, "logps/rejected": -260.0, "loss": 0.3918, "rewards/accuracies": 0.762499988079071, "rewards/chosen": 0.51953125, "rewards/margins": 1.015625, "rewards/rejected": -0.494140625, "step": 380 }, { "epoch": 1.2460063897763578, "grad_norm": 16.51862951602937, "learning_rate": 3.2485207100591716e-07, "logits/chosen": -1.546875, "logits/rejected": -1.6328125, "logps/chosen": -228.0, "logps/rejected": -276.0, "loss": 0.3695, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": 0.8125, "rewards/margins": 1.6953125, "rewards/rejected": -0.88671875, "step": 390 }, { "epoch": 1.2779552715654952, "grad_norm": 19.593966068797595, "learning_rate": 3.189349112426035e-07, "logits/chosen": -1.5859375, "logits/rejected": -1.625, "logps/chosen": -235.0, "logps/rejected": -276.0, "loss": 0.4045, "rewards/accuracies": 0.800000011920929, "rewards/chosen": 0.5234375, "rewards/margins": 1.125, "rewards/rejected": -0.59765625, "step": 400 }, { "epoch": 1.3099041533546325, "grad_norm": 19.821206837758886, "learning_rate": 3.130177514792899e-07, "logits/chosen": -1.546875, "logits/rejected": -1.6015625, "logps/chosen": -239.0, "logps/rejected": -266.0, "loss": 0.4309, "rewards/accuracies": 0.8500000238418579, "rewards/chosen": 0.427734375, "rewards/margins": 1.1640625, "rewards/rejected": -0.734375, "step": 410 }, { "epoch": 1.34185303514377, "grad_norm": 17.457629857186152, "learning_rate": 3.071005917159763e-07, "logits/chosen": -1.625, "logits/rejected": -1.609375, "logps/chosen": -243.0, "logps/rejected": -270.0, "loss": 0.3913, "rewards/accuracies": 0.8125, "rewards/chosen": 0.4609375, "rewards/margins": 1.1796875, "rewards/rejected": -0.72265625, "step": 420 }, { "epoch": 1.3738019169329074, "grad_norm": 16.963757931748397, "learning_rate": 3.011834319526627e-07, "logits/chosen": -1.578125, "logits/rejected": -1.609375, "logps/chosen": -235.0, "logps/rejected": -270.0, "loss": 0.3808, "rewards/accuracies": 0.8374999761581421, "rewards/chosen": 0.41796875, "rewards/margins": 1.1953125, "rewards/rejected": -0.78125, "step": 430 }, { "epoch": 1.4057507987220448, "grad_norm": 16.933512859408392, "learning_rate": 2.952662721893491e-07, "logits/chosen": -1.546875, "logits/rejected": -1.578125, "logps/chosen": -238.0, "logps/rejected": -262.0, "loss": 0.4261, "rewards/accuracies": 0.8125, "rewards/chosen": 0.40234375, "rewards/margins": 1.0, "rewards/rejected": -0.6015625, "step": 440 }, { "epoch": 1.4376996805111821, "grad_norm": 17.582118451476234, "learning_rate": 2.893491124260355e-07, "logits/chosen": -1.625, "logits/rejected": -1.6015625, "logps/chosen": -235.0, "logps/rejected": -268.0, "loss": 0.3921, "rewards/accuracies": 0.800000011920929, "rewards/chosen": 0.435546875, "rewards/margins": 1.0546875, "rewards/rejected": -0.62109375, "step": 450 }, { "epoch": 1.4696485623003195, "grad_norm": 14.599835012516651, "learning_rate": 2.834319526627219e-07, "logits/chosen": -1.5546875, "logits/rejected": -1.5859375, "logps/chosen": -234.0, "logps/rejected": -266.0, "loss": 0.357, "rewards/accuracies": 0.8125, "rewards/chosen": 0.353515625, "rewards/margins": 1.1796875, "rewards/rejected": -0.828125, "step": 460 }, { "epoch": 1.5015974440894568, "grad_norm": 17.01992762798607, "learning_rate": 2.7751479289940824e-07, "logits/chosen": -1.5625, "logits/rejected": -1.625, "logps/chosen": -238.0, "logps/rejected": -272.0, "loss": 0.3465, "rewards/accuracies": 0.875, "rewards/chosen": 0.232421875, "rewards/margins": 1.1640625, "rewards/rejected": -0.93359375, "step": 470 }, { "epoch": 1.5335463258785942, "grad_norm": 16.926918987794753, "learning_rate": 2.715976331360947e-07, "logits/chosen": -1.625, "logits/rejected": -1.578125, "logps/chosen": -245.0, "logps/rejected": -266.0, "loss": 0.3938, "rewards/accuracies": 0.762499988079071, "rewards/chosen": 0.2197265625, "rewards/margins": 1.0078125, "rewards/rejected": -0.78515625, "step": 480 }, { "epoch": 1.5654952076677318, "grad_norm": 19.019339456655477, "learning_rate": 2.6568047337278106e-07, "logits/chosen": -1.6328125, "logits/rejected": -1.625, "logps/chosen": -232.0, "logps/rejected": -272.0, "loss": 0.3621, "rewards/accuracies": 0.887499988079071, "rewards/chosen": 0.369140625, "rewards/margins": 1.3125, "rewards/rejected": -0.9453125, "step": 490 }, { "epoch": 1.5974440894568689, "grad_norm": 19.071902381545915, "learning_rate": 2.5976331360946746e-07, "logits/chosen": -1.5859375, "logits/rejected": -1.6015625, "logps/chosen": -231.0, "logps/rejected": -266.0, "loss": 0.4204, "rewards/accuracies": 0.8125, "rewards/chosen": 0.314453125, "rewards/margins": 1.1796875, "rewards/rejected": -0.8671875, "step": 500 }, { "epoch": 1.6293929712460065, "grad_norm": 15.876124711628384, "learning_rate": 2.538461538461538e-07, "logits/chosen": -1.6171875, "logits/rejected": -1.609375, "logps/chosen": -236.0, "logps/rejected": -262.0, "loss": 0.3832, "rewards/accuracies": 0.8125, "rewards/chosen": 0.365234375, "rewards/margins": 1.2109375, "rewards/rejected": -0.84375, "step": 510 }, { "epoch": 1.6613418530351438, "grad_norm": 14.900626138023746, "learning_rate": 2.479289940828402e-07, "logits/chosen": -1.578125, "logits/rejected": -1.5625, "logps/chosen": -232.0, "logps/rejected": -268.0, "loss": 0.3459, "rewards/accuracies": 0.8500000238418579, "rewards/chosen": 0.458984375, "rewards/margins": 1.46875, "rewards/rejected": -1.015625, "step": 520 }, { "epoch": 1.6932907348242812, "grad_norm": 19.494063656117618, "learning_rate": 2.420118343195266e-07, "logits/chosen": -1.6015625, "logits/rejected": -1.625, "logps/chosen": -230.0, "logps/rejected": -278.0, "loss": 0.3752, "rewards/accuracies": 0.7875000238418579, "rewards/chosen": 0.3671875, "rewards/margins": 1.328125, "rewards/rejected": -0.96484375, "step": 530 }, { "epoch": 1.7252396166134185, "grad_norm": 15.766044347009432, "learning_rate": 2.36094674556213e-07, "logits/chosen": -1.59375, "logits/rejected": -1.6171875, "logps/chosen": -231.0, "logps/rejected": -270.0, "loss": 0.3902, "rewards/accuracies": 0.8500000238418579, "rewards/chosen": 0.392578125, "rewards/margins": 1.3359375, "rewards/rejected": -0.9375, "step": 540 }, { "epoch": 1.7571884984025559, "grad_norm": 17.890212932549957, "learning_rate": 2.301775147928994e-07, "logits/chosen": -1.59375, "logits/rejected": -1.6328125, "logps/chosen": -241.0, "logps/rejected": -256.0, "loss": 0.3833, "rewards/accuracies": 0.800000011920929, "rewards/chosen": 0.384765625, "rewards/margins": 0.9921875, "rewards/rejected": -0.609375, "step": 550 }, { "epoch": 1.7891373801916934, "grad_norm": 17.977480288742196, "learning_rate": 2.242603550295858e-07, "logits/chosen": -1.578125, "logits/rejected": -1.625, "logps/chosen": -239.0, "logps/rejected": -270.0, "loss": 0.3717, "rewards/accuracies": 0.8125, "rewards/chosen": 0.259765625, "rewards/margins": 1.078125, "rewards/rejected": -0.8203125, "step": 560 }, { "epoch": 1.8210862619808306, "grad_norm": 17.31063553493517, "learning_rate": 2.1834319526627217e-07, "logits/chosen": -1.6015625, "logits/rejected": -1.59375, "logps/chosen": -233.0, "logps/rejected": -280.0, "loss": 0.3695, "rewards/accuracies": 0.9125000238418579, "rewards/chosen": 0.283203125, "rewards/margins": 1.28125, "rewards/rejected": -0.99609375, "step": 570 }, { "epoch": 1.8530351437699681, "grad_norm": 21.27064447874725, "learning_rate": 2.1242603550295858e-07, "logits/chosen": -1.578125, "logits/rejected": -1.6015625, "logps/chosen": -242.0, "logps/rejected": -276.0, "loss": 0.3861, "rewards/accuracies": 0.7875000238418579, "rewards/chosen": 0.234375, "rewards/margins": 1.0546875, "rewards/rejected": -0.8203125, "step": 580 }, { "epoch": 1.8849840255591053, "grad_norm": 19.92306872555079, "learning_rate": 2.0650887573964496e-07, "logits/chosen": -1.5859375, "logits/rejected": -1.5859375, "logps/chosen": -235.0, "logps/rejected": -264.0, "loss": 0.3873, "rewards/accuracies": 0.7875000238418579, "rewards/chosen": 0.37890625, "rewards/margins": 1.1015625, "rewards/rejected": -0.72265625, "step": 590 }, { "epoch": 1.9169329073482428, "grad_norm": 22.677693548477258, "learning_rate": 2.0059171597633133e-07, "logits/chosen": -1.6171875, "logits/rejected": -1.6171875, "logps/chosen": -240.0, "logps/rejected": -258.0, "loss": 0.4016, "rewards/accuracies": 0.800000011920929, "rewards/chosen": 0.2314453125, "rewards/margins": 0.984375, "rewards/rejected": -0.75390625, "step": 600 }, { "epoch": 1.9488817891373802, "grad_norm": 21.51355955572424, "learning_rate": 1.9467455621301774e-07, "logits/chosen": -1.578125, "logits/rejected": -1.671875, "logps/chosen": -234.0, "logps/rejected": -266.0, "loss": 0.375, "rewards/accuracies": 0.800000011920929, "rewards/chosen": 0.25390625, "rewards/margins": 1.0859375, "rewards/rejected": -0.83203125, "step": 610 }, { "epoch": 1.9808306709265175, "grad_norm": 14.74606499354998, "learning_rate": 1.8875739644970412e-07, "logits/chosen": -1.578125, "logits/rejected": -1.59375, "logps/chosen": -243.0, "logps/rejected": -268.0, "loss": 0.3809, "rewards/accuracies": 0.8125, "rewards/chosen": 0.03759765625, "rewards/margins": 1.1328125, "rewards/rejected": -1.09375, "step": 620 }, { "epoch": 2.0, "eval_logits/chosen": -1.6171875, "eval_logits/rejected": -1.59375, "eval_logps/chosen": -244.0, "eval_logps/rejected": -260.0, "eval_loss": 0.5935156345367432, "eval_rewards/accuracies": 0.6428571343421936, "eval_rewards/chosen": -0.0556640625, "eval_rewards/margins": 0.63671875, "eval_rewards/rejected": -0.69140625, "eval_runtime": 12.2896, "eval_samples_per_second": 16.274, "eval_steps_per_second": 0.57, "step": 626 }, { "epoch": 2.012779552715655, "grad_norm": 15.817315308499108, "learning_rate": 1.8284023668639053e-07, "logits/chosen": -1.6171875, "logits/rejected": -1.6015625, "logps/chosen": -239.0, "logps/rejected": -262.0, "loss": 0.3165, "rewards/accuracies": 0.875, "rewards/chosen": 0.330078125, "rewards/margins": 1.40625, "rewards/rejected": -1.0703125, "step": 630 }, { "epoch": 2.0447284345047922, "grad_norm": 11.365629657180074, "learning_rate": 1.7692307692307693e-07, "logits/chosen": -1.625, "logits/rejected": -1.65625, "logps/chosen": -238.0, "logps/rejected": -272.0, "loss": 0.2917, "rewards/accuracies": 0.9375, "rewards/chosen": 0.361328125, "rewards/margins": 1.4375, "rewards/rejected": -1.078125, "step": 640 }, { "epoch": 2.07667731629393, "grad_norm": 13.196571731112993, "learning_rate": 1.710059171597633e-07, "logits/chosen": -1.640625, "logits/rejected": -1.6875, "logps/chosen": -231.0, "logps/rejected": -280.0, "loss": 0.2862, "rewards/accuracies": 0.9125000238418579, "rewards/chosen": 0.328125, "rewards/margins": 1.546875, "rewards/rejected": -1.21875, "step": 650 }, { "epoch": 2.108626198083067, "grad_norm": 10.695579683512666, "learning_rate": 1.650887573964497e-07, "logits/chosen": -1.578125, "logits/rejected": -1.625, "logps/chosen": -228.0, "logps/rejected": -276.0, "loss": 0.301, "rewards/accuracies": 0.9125000238418579, "rewards/chosen": 0.392578125, "rewards/margins": 1.65625, "rewards/rejected": -1.265625, "step": 660 }, { "epoch": 2.1405750798722045, "grad_norm": 11.265653450090454, "learning_rate": 1.591715976331361e-07, "logits/chosen": -1.6171875, "logits/rejected": -1.6484375, "logps/chosen": -235.0, "logps/rejected": -272.0, "loss": 0.2557, "rewards/accuracies": 0.9375, "rewards/chosen": 0.365234375, "rewards/margins": 1.6171875, "rewards/rejected": -1.25, "step": 670 }, { "epoch": 2.1725239616613417, "grad_norm": 14.15113024471477, "learning_rate": 1.5325443786982248e-07, "logits/chosen": -1.578125, "logits/rejected": -1.6328125, "logps/chosen": -234.0, "logps/rejected": -274.0, "loss": 0.2776, "rewards/accuracies": 0.949999988079071, "rewards/chosen": 0.1884765625, "rewards/margins": 1.46875, "rewards/rejected": -1.28125, "step": 680 }, { "epoch": 2.2044728434504792, "grad_norm": 14.981353716929215, "learning_rate": 1.4733727810650885e-07, "logits/chosen": -1.6328125, "logits/rejected": -1.671875, "logps/chosen": -240.0, "logps/rejected": -278.0, "loss": 0.2842, "rewards/accuracies": 0.887499988079071, "rewards/chosen": 0.294921875, "rewards/margins": 1.4296875, "rewards/rejected": -1.140625, "step": 690 }, { "epoch": 2.236421725239617, "grad_norm": 17.85170492540105, "learning_rate": 1.4142011834319526e-07, "logits/chosen": -1.65625, "logits/rejected": -1.6875, "logps/chosen": -237.0, "logps/rejected": -274.0, "loss": 0.2841, "rewards/accuracies": 0.875, "rewards/chosen": 0.2353515625, "rewards/margins": 1.5390625, "rewards/rejected": -1.3046875, "step": 700 }, { "epoch": 2.268370607028754, "grad_norm": 14.441788785513792, "learning_rate": 1.3550295857988164e-07, "logits/chosen": -1.6328125, "logits/rejected": -1.59375, "logps/chosen": -237.0, "logps/rejected": -280.0, "loss": 0.303, "rewards/accuracies": 0.8374999761581421, "rewards/chosen": 0.2451171875, "rewards/margins": 1.5625, "rewards/rejected": -1.3125, "step": 710 }, { "epoch": 2.3003194888178915, "grad_norm": 16.017805126479765, "learning_rate": 1.2958579881656802e-07, "logits/chosen": -1.65625, "logits/rejected": -1.640625, "logps/chosen": -240.0, "logps/rejected": -272.0, "loss": 0.3037, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": 0.267578125, "rewards/margins": 1.4609375, "rewards/rejected": -1.1875, "step": 720 }, { "epoch": 2.3322683706070286, "grad_norm": 14.374194587453172, "learning_rate": 1.2366863905325443e-07, "logits/chosen": -1.546875, "logits/rejected": -1.6015625, "logps/chosen": -234.0, "logps/rejected": -276.0, "loss": 0.2611, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": 0.2255859375, "rewards/margins": 1.6328125, "rewards/rejected": -1.40625, "step": 730 }, { "epoch": 2.364217252396166, "grad_norm": 17.298259126788636, "learning_rate": 1.1775147928994082e-07, "logits/chosen": -1.6015625, "logits/rejected": -1.625, "logps/chosen": -231.0, "logps/rejected": -276.0, "loss": 0.2663, "rewards/accuracies": 0.949999988079071, "rewards/chosen": 0.47265625, "rewards/margins": 1.7890625, "rewards/rejected": -1.3125, "step": 740 }, { "epoch": 2.3961661341853033, "grad_norm": 15.560832236353352, "learning_rate": 1.1183431952662721e-07, "logits/chosen": -1.6328125, "logits/rejected": -1.6484375, "logps/chosen": -236.0, "logps/rejected": -282.0, "loss": 0.2495, "rewards/accuracies": 0.9125000238418579, "rewards/chosen": 0.462890625, "rewards/margins": 1.8671875, "rewards/rejected": -1.40625, "step": 750 }, { "epoch": 2.428115015974441, "grad_norm": 13.505631748827204, "learning_rate": 1.059171597633136e-07, "logits/chosen": -1.625, "logits/rejected": -1.65625, "logps/chosen": -228.0, "logps/rejected": -284.0, "loss": 0.2635, "rewards/accuracies": 0.875, "rewards/chosen": 0.349609375, "rewards/margins": 1.671875, "rewards/rejected": -1.328125, "step": 760 }, { "epoch": 2.460063897763578, "grad_norm": 12.876093667551999, "learning_rate": 1e-07, "logits/chosen": -1.6171875, "logits/rejected": -1.6640625, "logps/chosen": -236.0, "logps/rejected": -276.0, "loss": 0.2579, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": 0.25390625, "rewards/margins": 1.5625, "rewards/rejected": -1.3125, "step": 770 }, { "epoch": 2.4920127795527156, "grad_norm": 15.762232240327554, "learning_rate": 9.408284023668639e-08, "logits/chosen": -1.6015625, "logits/rejected": -1.578125, "logps/chosen": -236.0, "logps/rejected": -282.0, "loss": 0.2702, "rewards/accuracies": 0.875, "rewards/chosen": 0.1640625, "rewards/margins": 1.7421875, "rewards/rejected": -1.578125, "step": 780 }, { "epoch": 2.523961661341853, "grad_norm": 14.086973297325832, "learning_rate": 8.816568047337278e-08, "logits/chosen": -1.671875, "logits/rejected": -1.7265625, "logps/chosen": -243.0, "logps/rejected": -274.0, "loss": 0.2447, "rewards/accuracies": 0.925000011920929, "rewards/chosen": 0.177734375, "rewards/margins": 1.6796875, "rewards/rejected": -1.5, "step": 790 }, { "epoch": 2.5559105431309903, "grad_norm": 15.821508873733881, "learning_rate": 8.224852071005916e-08, "logits/chosen": -1.625, "logits/rejected": -1.6484375, "logps/chosen": -237.0, "logps/rejected": -286.0, "loss": 0.2482, "rewards/accuracies": 0.9125000238418579, "rewards/chosen": 0.236328125, "rewards/margins": 1.9609375, "rewards/rejected": -1.71875, "step": 800 }, { "epoch": 2.587859424920128, "grad_norm": 12.958827992117726, "learning_rate": 7.633136094674555e-08, "logits/chosen": -1.6015625, "logits/rejected": -1.6171875, "logps/chosen": -228.0, "logps/rejected": -268.0, "loss": 0.2824, "rewards/accuracies": 0.9375, "rewards/chosen": 0.35546875, "rewards/margins": 1.5859375, "rewards/rejected": -1.2265625, "step": 810 }, { "epoch": 2.619808306709265, "grad_norm": 14.078315987923908, "learning_rate": 7.041420118343195e-08, "logits/chosen": -1.625, "logits/rejected": -1.65625, "logps/chosen": -240.0, "logps/rejected": -268.0, "loss": 0.3035, "rewards/accuracies": 0.862500011920929, "rewards/chosen": 0.224609375, "rewards/margins": 1.65625, "rewards/rejected": -1.4296875, "step": 820 }, { "epoch": 2.6517571884984026, "grad_norm": 10.88207889438703, "learning_rate": 6.449704142011835e-08, "logits/chosen": -1.59375, "logits/rejected": -1.6484375, "logps/chosen": -243.0, "logps/rejected": -278.0, "loss": 0.2711, "rewards/accuracies": 0.862500011920929, "rewards/chosen": 0.1455078125, "rewards/margins": 1.53125, "rewards/rejected": -1.3828125, "step": 830 }, { "epoch": 2.68370607028754, "grad_norm": 14.648580375481389, "learning_rate": 5.857988165680473e-08, "logits/chosen": -1.6015625, "logits/rejected": -1.6484375, "logps/chosen": -241.0, "logps/rejected": -274.0, "loss": 0.2889, "rewards/accuracies": 0.862500011920929, "rewards/chosen": 0.1708984375, "rewards/margins": 1.5234375, "rewards/rejected": -1.3515625, "step": 840 }, { "epoch": 2.7156549520766773, "grad_norm": 14.952610668822093, "learning_rate": 5.266272189349112e-08, "logits/chosen": -1.6015625, "logits/rejected": -1.625, "logps/chosen": -233.0, "logps/rejected": -268.0, "loss": 0.2814, "rewards/accuracies": 0.875, "rewards/chosen": 0.30078125, "rewards/margins": 1.6484375, "rewards/rejected": -1.3515625, "step": 850 }, { "epoch": 2.747603833865815, "grad_norm": 14.079308847224443, "learning_rate": 4.674556213017751e-08, "logits/chosen": -1.6484375, "logits/rejected": -1.640625, "logps/chosen": -241.0, "logps/rejected": -272.0, "loss": 0.2493, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": 0.1552734375, "rewards/margins": 1.6875, "rewards/rejected": -1.53125, "step": 860 }, { "epoch": 2.779552715654952, "grad_norm": 12.418261084361873, "learning_rate": 4.082840236686391e-08, "logits/chosen": -1.6875, "logits/rejected": -1.6640625, "logps/chosen": -247.0, "logps/rejected": -272.0, "loss": 0.2975, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": 0.031494140625, "rewards/margins": 1.5859375, "rewards/rejected": -1.5546875, "step": 870 }, { "epoch": 2.8115015974440896, "grad_norm": 18.172327917262383, "learning_rate": 3.4911242603550294e-08, "logits/chosen": -1.5546875, "logits/rejected": -1.6171875, "logps/chosen": -238.0, "logps/rejected": -278.0, "loss": 0.2769, "rewards/accuracies": 0.925000011920929, "rewards/chosen": 0.1162109375, "rewards/margins": 1.8359375, "rewards/rejected": -1.7265625, "step": 880 }, { "epoch": 2.8434504792332267, "grad_norm": 16.82265755243597, "learning_rate": 2.8994082840236687e-08, "logits/chosen": -1.609375, "logits/rejected": -1.6796875, "logps/chosen": -238.0, "logps/rejected": -274.0, "loss": 0.2871, "rewards/accuracies": 0.9125000238418579, "rewards/chosen": 0.232421875, "rewards/margins": 1.859375, "rewards/rejected": -1.625, "step": 890 }, { "epoch": 2.8753993610223643, "grad_norm": 13.793067995533367, "learning_rate": 2.3076923076923076e-08, "logits/chosen": -1.6640625, "logits/rejected": -1.6015625, "logps/chosen": -238.0, "logps/rejected": -272.0, "loss": 0.2916, "rewards/accuracies": 0.887499988079071, "rewards/chosen": 0.18359375, "rewards/margins": 1.7265625, "rewards/rejected": -1.5390625, "step": 900 }, { "epoch": 2.9073482428115014, "grad_norm": 14.63855638742523, "learning_rate": 1.7159763313609465e-08, "logits/chosen": -1.609375, "logits/rejected": -1.625, "logps/chosen": -246.0, "logps/rejected": -278.0, "loss": 0.269, "rewards/accuracies": 0.887499988079071, "rewards/chosen": 0.12255859375, "rewards/margins": 1.640625, "rewards/rejected": -1.515625, "step": 910 }, { "epoch": 2.939297124600639, "grad_norm": 15.081828948432232, "learning_rate": 1.1242603550295858e-08, "logits/chosen": -1.6875, "logits/rejected": -1.625, "logps/chosen": -245.0, "logps/rejected": -272.0, "loss": 0.2832, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": 0.0174560546875, "rewards/margins": 1.53125, "rewards/rejected": -1.5078125, "step": 920 }, { "epoch": 2.9712460063897765, "grad_norm": 12.755168018204367, "learning_rate": 5.325443786982248e-09, "logits/chosen": -1.6484375, "logits/rejected": -1.6875, "logps/chosen": -243.0, "logps/rejected": -266.0, "loss": 0.273, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": 0.1015625, "rewards/margins": 1.4375, "rewards/rejected": -1.3359375, "step": 930 }, { "epoch": 3.0, "eval_logits/chosen": -1.640625, "eval_logits/rejected": -1.6171875, "eval_logps/chosen": -247.0, "eval_logps/rejected": -264.0, "eval_loss": 0.5956249833106995, "eval_rewards/accuracies": 0.625, "eval_rewards/chosen": -0.27734375, "eval_rewards/margins": 0.6875, "eval_rewards/rejected": -0.96484375, "eval_runtime": 14.7718, "eval_samples_per_second": 13.539, "eval_steps_per_second": 0.474, "step": 939 }, { "epoch": 3.0, "step": 939, "total_flos": 0.0, "train_loss": 0.41333865814696485, "train_runtime": 4621.2542, "train_samples_per_second": 6.49, "train_steps_per_second": 0.203 } ], "logging_steps": 10, "max_steps": 939, "num_input_tokens_seen": 0, "num_train_epochs": 3, "save_steps": 500, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": true, "should_training_stop": true }, "attributes": {} } }, "total_flos": 0.0, "train_batch_size": 8, "trial_name": null, "trial_params": null }