{ "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": 91.90092179843542, "learning_rate": 5.3191489361702125e-09, "logits/chosen": -3.453125, "logits/rejected": -3.421875, "logps/chosen": -278.0, "logps/rejected": -262.0, "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": 90.9674980415812, "learning_rate": 5.3191489361702123e-08, "logits/chosen": -3.640625, "logits/rejected": -3.578125, "logps/chosen": -282.0, "logps/rejected": -274.0, "loss": 0.6927, "rewards/accuracies": 0.0833333358168602, "rewards/chosen": -0.00970458984375, "rewards/margins": -0.00555419921875, "rewards/rejected": -0.004180908203125, "step": 10 }, { "epoch": 0.06389776357827476, "grad_norm": 90.29736649004343, "learning_rate": 1.0638297872340425e-07, "logits/chosen": -3.53125, "logits/rejected": -3.546875, "logps/chosen": -278.0, "logps/rejected": -272.0, "loss": 0.6902, "rewards/accuracies": 0.2750000059604645, "rewards/chosen": -0.03515625, "rewards/margins": -0.0012664794921875, "rewards/rejected": -0.033935546875, "step": 20 }, { "epoch": 0.09584664536741214, "grad_norm": 87.07720807385869, "learning_rate": 1.5957446808510638e-07, "logits/chosen": -3.625, "logits/rejected": -3.625, "logps/chosen": -278.0, "logps/rejected": -272.0, "loss": 0.6624, "rewards/accuracies": 0.4375, "rewards/chosen": -0.04248046875, "rewards/margins": 0.06640625, "rewards/rejected": -0.10888671875, "step": 30 }, { "epoch": 0.12779552715654952, "grad_norm": 78.25631914811093, "learning_rate": 2.127659574468085e-07, "logits/chosen": -3.53125, "logits/rejected": -3.4375, "logps/chosen": -284.0, "logps/rejected": -270.0, "loss": 0.6405, "rewards/accuracies": 0.625, "rewards/chosen": -0.11865234375, "rewards/margins": 0.1962890625, "rewards/rejected": -0.314453125, "step": 40 }, { "epoch": 0.1597444089456869, "grad_norm": 84.80020092762481, "learning_rate": 2.659574468085106e-07, "logits/chosen": -3.515625, "logits/rejected": -3.484375, "logps/chosen": -278.0, "logps/rejected": -274.0, "loss": 0.6362, "rewards/accuracies": 0.637499988079071, "rewards/chosen": -0.169921875, "rewards/margins": 0.21484375, "rewards/rejected": -0.384765625, "step": 50 }, { "epoch": 0.19169329073482427, "grad_norm": 73.65610191156658, "learning_rate": 3.1914893617021275e-07, "logits/chosen": -3.546875, "logits/rejected": -3.390625, "logps/chosen": -284.0, "logps/rejected": -272.0, "loss": 0.5906, "rewards/accuracies": 0.637499988079071, "rewards/chosen": -0.201171875, "rewards/margins": 0.31640625, "rewards/rejected": -0.51953125, "step": 60 }, { "epoch": 0.22364217252396165, "grad_norm": 68.32598567671492, "learning_rate": 3.7234042553191484e-07, "logits/chosen": -3.515625, "logits/rejected": -3.5, "logps/chosen": -284.0, "logps/rejected": -274.0, "loss": 0.5743, "rewards/accuracies": 0.6625000238418579, "rewards/chosen": -0.330078125, "rewards/margins": 0.412109375, "rewards/rejected": -0.7421875, "step": 70 }, { "epoch": 0.25559105431309903, "grad_norm": 87.25967925719019, "learning_rate": 4.25531914893617e-07, "logits/chosen": -3.46875, "logits/rejected": -3.5, "logps/chosen": -280.0, "logps/rejected": -272.0, "loss": 0.581, "rewards/accuracies": 0.699999988079071, "rewards/chosen": -0.35546875, "rewards/margins": 0.4375, "rewards/rejected": -0.79296875, "step": 80 }, { "epoch": 0.28753993610223644, "grad_norm": 71.78294646227785, "learning_rate": 4.787234042553192e-07, "logits/chosen": -3.53125, "logits/rejected": -3.453125, "logps/chosen": -274.0, "logps/rejected": -286.0, "loss": 0.5669, "rewards/accuracies": 0.675000011920929, "rewards/chosen": -0.314453125, "rewards/margins": 0.52734375, "rewards/rejected": -0.83984375, "step": 90 }, { "epoch": 0.3194888178913738, "grad_norm": 78.76158351857062, "learning_rate": 4.964497041420119e-07, "logits/chosen": -3.59375, "logits/rejected": -3.53125, "logps/chosen": -290.0, "logps/rejected": -270.0, "loss": 0.6233, "rewards/accuracies": 0.675000011920929, "rewards/chosen": -0.255859375, "rewards/margins": 0.59765625, "rewards/rejected": -0.85546875, "step": 100 }, { "epoch": 0.3514376996805112, "grad_norm": 80.45460798252958, "learning_rate": 4.905325443786982e-07, "logits/chosen": -3.515625, "logits/rejected": -3.609375, "logps/chosen": -288.0, "logps/rejected": -270.0, "loss": 0.5553, "rewards/accuracies": 0.699999988079071, "rewards/chosen": -0.232421875, "rewards/margins": 0.5859375, "rewards/rejected": -0.81640625, "step": 110 }, { "epoch": 0.38338658146964855, "grad_norm": 62.143803745891255, "learning_rate": 4.846153846153846e-07, "logits/chosen": -3.609375, "logits/rejected": -3.46875, "logps/chosen": -282.0, "logps/rejected": -270.0, "loss": 0.557, "rewards/accuracies": 0.7124999761581421, "rewards/chosen": -0.1884765625, "rewards/margins": 0.78515625, "rewards/rejected": -0.9765625, "step": 120 }, { "epoch": 0.41533546325878595, "grad_norm": 69.53537157369784, "learning_rate": 4.78698224852071e-07, "logits/chosen": -3.6875, "logits/rejected": -3.5625, "logps/chosen": -286.0, "logps/rejected": -284.0, "loss": 0.5435, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -0.265625, "rewards/margins": 0.8359375, "rewards/rejected": -1.1015625, "step": 130 }, { "epoch": 0.4472843450479233, "grad_norm": 68.50218683121396, "learning_rate": 4.727810650887574e-07, "logits/chosen": -3.484375, "logits/rejected": -3.5, "logps/chosen": -290.0, "logps/rejected": -280.0, "loss": 0.5412, "rewards/accuracies": 0.699999988079071, "rewards/chosen": -0.546875, "rewards/margins": 0.99609375, "rewards/rejected": -1.546875, "step": 140 }, { "epoch": 0.4792332268370607, "grad_norm": 53.728622573163506, "learning_rate": 4.668639053254438e-07, "logits/chosen": -3.59375, "logits/rejected": -3.609375, "logps/chosen": -282.0, "logps/rejected": -272.0, "loss": 0.5295, "rewards/accuracies": 0.7875000238418579, "rewards/chosen": -0.2470703125, "rewards/margins": 1.0859375, "rewards/rejected": -1.328125, "step": 150 }, { "epoch": 0.5111821086261981, "grad_norm": 73.68063510244593, "learning_rate": 4.6094674556213014e-07, "logits/chosen": -3.5, "logits/rejected": -3.625, "logps/chosen": -290.0, "logps/rejected": -272.0, "loss": 0.5547, "rewards/accuracies": 0.699999988079071, "rewards/chosen": -0.251953125, "rewards/margins": 0.71875, "rewards/rejected": -0.97265625, "step": 160 }, { "epoch": 0.5431309904153354, "grad_norm": 61.23321025324786, "learning_rate": 4.5502958579881655e-07, "logits/chosen": -3.625, "logits/rejected": -3.5625, "logps/chosen": -290.0, "logps/rejected": -274.0, "loss": 0.5391, "rewards/accuracies": 0.7875000238418579, "rewards/chosen": -0.376953125, "rewards/margins": 1.09375, "rewards/rejected": -1.46875, "step": 170 }, { "epoch": 0.5750798722044729, "grad_norm": 70.65019095299606, "learning_rate": 4.491124260355029e-07, "logits/chosen": -3.578125, "logits/rejected": -3.484375, "logps/chosen": -294.0, "logps/rejected": -280.0, "loss": 0.5284, "rewards/accuracies": 0.7875000238418579, "rewards/chosen": -0.216796875, "rewards/margins": 1.2421875, "rewards/rejected": -1.453125, "step": 180 }, { "epoch": 0.6070287539936102, "grad_norm": 78.65972402327554, "learning_rate": 4.4319526627218936e-07, "logits/chosen": -3.578125, "logits/rejected": -3.515625, "logps/chosen": -278.0, "logps/rejected": -284.0, "loss": 0.582, "rewards/accuracies": 0.7124999761581421, "rewards/chosen": -0.302734375, "rewards/margins": 1.046875, "rewards/rejected": -1.3515625, "step": 190 }, { "epoch": 0.6389776357827476, "grad_norm": 58.77306801176155, "learning_rate": 4.372781065088757e-07, "logits/chosen": -3.703125, "logits/rejected": -3.5625, "logps/chosen": -282.0, "logps/rejected": -280.0, "loss": 0.5452, "rewards/accuracies": 0.699999988079071, "rewards/chosen": -0.3203125, "rewards/margins": 0.953125, "rewards/rejected": -1.2734375, "step": 200 }, { "epoch": 0.670926517571885, "grad_norm": 71.97911960871565, "learning_rate": 4.313609467455621e-07, "logits/chosen": -3.609375, "logits/rejected": -3.609375, "logps/chosen": -274.0, "logps/rejected": -286.0, "loss": 0.5594, "rewards/accuracies": 0.7749999761581421, "rewards/chosen": -0.11376953125, "rewards/margins": 1.3203125, "rewards/rejected": -1.4375, "step": 210 }, { "epoch": 0.7028753993610224, "grad_norm": 72.67644981014261, "learning_rate": 4.2544378698224847e-07, "logits/chosen": -3.640625, "logits/rejected": -3.515625, "logps/chosen": -278.0, "logps/rejected": -276.0, "loss": 0.5427, "rewards/accuracies": 0.7124999761581421, "rewards/chosen": -0.31640625, "rewards/margins": 0.9296875, "rewards/rejected": -1.25, "step": 220 }, { "epoch": 0.7348242811501597, "grad_norm": 77.77118766206728, "learning_rate": 4.195266272189349e-07, "logits/chosen": -3.609375, "logits/rejected": -3.5625, "logps/chosen": -280.0, "logps/rejected": -278.0, "loss": 0.526, "rewards/accuracies": 0.675000011920929, "rewards/chosen": -0.267578125, "rewards/margins": 0.95703125, "rewards/rejected": -1.2265625, "step": 230 }, { "epoch": 0.7667731629392971, "grad_norm": 81.20199378554962, "learning_rate": 4.1360946745562133e-07, "logits/chosen": -3.59375, "logits/rejected": -3.515625, "logps/chosen": -280.0, "logps/rejected": -282.0, "loss": 0.5185, "rewards/accuracies": 0.7124999761581421, "rewards/chosen": -0.2734375, "rewards/margins": 0.984375, "rewards/rejected": -1.2578125, "step": 240 }, { "epoch": 0.7987220447284346, "grad_norm": 65.01515226981213, "learning_rate": 4.076923076923077e-07, "logits/chosen": -3.65625, "logits/rejected": -3.609375, "logps/chosen": -274.0, "logps/rejected": -282.0, "loss": 0.5235, "rewards/accuracies": 0.6625000238418579, "rewards/chosen": -0.494140625, "rewards/margins": 0.83203125, "rewards/rejected": -1.328125, "step": 250 }, { "epoch": 0.8306709265175719, "grad_norm": 66.7618546103775, "learning_rate": 4.017751479289941e-07, "logits/chosen": -3.59375, "logits/rejected": -3.625, "logps/chosen": -280.0, "logps/rejected": -278.0, "loss": 0.5166, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -0.220703125, "rewards/margins": 1.09375, "rewards/rejected": -1.3125, "step": 260 }, { "epoch": 0.8626198083067093, "grad_norm": 90.38334381915796, "learning_rate": 3.9585798816568044e-07, "logits/chosen": -3.703125, "logits/rejected": -3.578125, "logps/chosen": -284.0, "logps/rejected": -278.0, "loss": 0.5917, "rewards/accuracies": 0.6625000238418579, "rewards/chosen": -0.390625, "rewards/margins": 0.89453125, "rewards/rejected": -1.2890625, "step": 270 }, { "epoch": 0.8945686900958466, "grad_norm": 55.19009591945878, "learning_rate": 3.8994082840236685e-07, "logits/chosen": -3.578125, "logits/rejected": -3.640625, "logps/chosen": -278.0, "logps/rejected": -274.0, "loss": 0.5651, "rewards/accuracies": 0.75, "rewards/chosen": -0.265625, "rewards/margins": 1.1484375, "rewards/rejected": -1.4140625, "step": 280 }, { "epoch": 0.9265175718849841, "grad_norm": 67.38142219006318, "learning_rate": 3.840236686390532e-07, "logits/chosen": -3.5625, "logits/rejected": -3.546875, "logps/chosen": -272.0, "logps/rejected": -284.0, "loss": 0.4849, "rewards/accuracies": 0.7124999761581421, "rewards/chosen": -0.2734375, "rewards/margins": 1.2109375, "rewards/rejected": -1.484375, "step": 290 }, { "epoch": 0.9584664536741214, "grad_norm": 72.63905100297757, "learning_rate": 3.7810650887573966e-07, "logits/chosen": -3.578125, "logits/rejected": -3.59375, "logps/chosen": -288.0, "logps/rejected": -276.0, "loss": 0.5193, "rewards/accuracies": 0.800000011920929, "rewards/chosen": -0.67578125, "rewards/margins": 1.2890625, "rewards/rejected": -1.9609375, "step": 300 }, { "epoch": 0.9904153354632588, "grad_norm": 73.80738834120062, "learning_rate": 3.72189349112426e-07, "logits/chosen": -3.703125, "logits/rejected": -3.703125, "logps/chosen": -290.0, "logps/rejected": -282.0, "loss": 0.4772, "rewards/accuracies": 0.8500000238418579, "rewards/chosen": -0.640625, "rewards/margins": 1.625, "rewards/rejected": -2.265625, "step": 310 }, { "epoch": 1.0, "eval_logits/chosen": -3.625, "eval_logits/rejected": -3.671875, "eval_logps/chosen": -280.0, "eval_logps/rejected": -290.0, "eval_loss": 0.5548437237739563, "eval_rewards/accuracies": 0.6428571343421936, "eval_rewards/chosen": -0.8984375, "eval_rewards/margins": 0.6484375, "eval_rewards/rejected": -1.546875, "eval_runtime": 15.7097, "eval_samples_per_second": 12.731, "eval_steps_per_second": 0.446, "step": 313 }, { "epoch": 1.0223642172523961, "grad_norm": 25.96214150273739, "learning_rate": 3.662721893491124e-07, "logits/chosen": -3.65625, "logits/rejected": -3.578125, "logps/chosen": -272.0, "logps/rejected": -292.0, "loss": 0.2187, "rewards/accuracies": 0.925000011920929, "rewards/chosen": 0.0849609375, "rewards/margins": 2.671875, "rewards/rejected": -2.59375, "step": 320 }, { "epoch": 1.0543130990415335, "grad_norm": 37.544186657188476, "learning_rate": 3.603550295857988e-07, "logits/chosen": -3.59375, "logits/rejected": -3.625, "logps/chosen": -280.0, "logps/rejected": -292.0, "loss": 0.1177, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": 0.2109375, "rewards/margins": 3.21875, "rewards/rejected": -3.0, "step": 330 }, { "epoch": 1.0862619808306708, "grad_norm": 35.42433639921683, "learning_rate": 3.544378698224852e-07, "logits/chosen": -3.59375, "logits/rejected": -3.5, "logps/chosen": -280.0, "logps/rejected": -294.0, "loss": 0.0984, "rewards/accuracies": 0.9375, "rewards/chosen": 0.2197265625, "rewards/margins": 3.265625, "rewards/rejected": -3.046875, "step": 340 }, { "epoch": 1.1182108626198084, "grad_norm": 33.043206820124475, "learning_rate": 3.485207100591716e-07, "logits/chosen": -3.5625, "logits/rejected": -3.515625, "logps/chosen": -284.0, "logps/rejected": -290.0, "loss": 0.1318, "rewards/accuracies": 0.9375, "rewards/chosen": -0.35546875, "rewards/margins": 2.984375, "rewards/rejected": -3.34375, "step": 350 }, { "epoch": 1.1501597444089458, "grad_norm": 33.51039580450466, "learning_rate": 3.42603550295858e-07, "logits/chosen": -3.625, "logits/rejected": -3.5, "logps/chosen": -284.0, "logps/rejected": -306.0, "loss": 0.1046, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": -0.29296875, "rewards/margins": 3.390625, "rewards/rejected": -3.6875, "step": 360 }, { "epoch": 1.182108626198083, "grad_norm": 17.318439329466752, "learning_rate": 3.366863905325444e-07, "logits/chosen": -3.546875, "logits/rejected": -3.546875, "logps/chosen": -282.0, "logps/rejected": -300.0, "loss": 0.1291, "rewards/accuracies": 0.949999988079071, "rewards/chosen": 0.275390625, "rewards/margins": 3.796875, "rewards/rejected": -3.515625, "step": 370 }, { "epoch": 1.2140575079872205, "grad_norm": 26.01507449223922, "learning_rate": 3.3076923076923075e-07, "logits/chosen": -3.59375, "logits/rejected": -3.453125, "logps/chosen": -282.0, "logps/rejected": -310.0, "loss": 0.1081, "rewards/accuracies": 1.0, "rewards/chosen": 0.06005859375, "rewards/margins": 3.84375, "rewards/rejected": -3.78125, "step": 380 }, { "epoch": 1.2460063897763578, "grad_norm": 31.832566772837826, "learning_rate": 3.2485207100591716e-07, "logits/chosen": -3.59375, "logits/rejected": -3.484375, "logps/chosen": -280.0, "logps/rejected": -312.0, "loss": 0.1019, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.1923828125, "rewards/margins": 3.59375, "rewards/rejected": -3.78125, "step": 390 }, { "epoch": 1.2779552715654952, "grad_norm": 25.833772692336037, "learning_rate": 3.189349112426035e-07, "logits/chosen": -3.59375, "logits/rejected": -3.515625, "logps/chosen": -284.0, "logps/rejected": -302.0, "loss": 0.1152, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.412109375, "rewards/margins": 3.375, "rewards/rejected": -3.78125, "step": 400 }, { "epoch": 1.3099041533546325, "grad_norm": 41.5197594341305, "learning_rate": 3.130177514792899e-07, "logits/chosen": -3.640625, "logits/rejected": -3.546875, "logps/chosen": -278.0, "logps/rejected": -304.0, "loss": 0.1294, "rewards/accuracies": 0.9125000238418579, "rewards/chosen": -0.208984375, "rewards/margins": 3.453125, "rewards/rejected": -3.65625, "step": 410 }, { "epoch": 1.34185303514377, "grad_norm": 16.1503025747185, "learning_rate": 3.071005917159763e-07, "logits/chosen": -3.609375, "logits/rejected": -3.59375, "logps/chosen": -284.0, "logps/rejected": -300.0, "loss": 0.1114, "rewards/accuracies": 0.9375, "rewards/chosen": -0.103515625, "rewards/margins": 3.328125, "rewards/rejected": -3.4375, "step": 420 }, { "epoch": 1.3738019169329074, "grad_norm": 20.84892907348037, "learning_rate": 3.011834319526627e-07, "logits/chosen": -3.65625, "logits/rejected": -3.5625, "logps/chosen": -276.0, "logps/rejected": -302.0, "loss": 0.0938, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": 0.11083984375, "rewards/margins": 3.75, "rewards/rejected": -3.65625, "step": 430 }, { "epoch": 1.4057507987220448, "grad_norm": 23.2310306282084, "learning_rate": 2.952662721893491e-07, "logits/chosen": -3.640625, "logits/rejected": -3.515625, "logps/chosen": -270.0, "logps/rejected": -320.0, "loss": 0.1134, "rewards/accuracies": 0.949999988079071, "rewards/chosen": 0.0003910064697265625, "rewards/margins": 3.78125, "rewards/rejected": -3.78125, "step": 440 }, { "epoch": 1.4376996805111821, "grad_norm": 34.93249416861729, "learning_rate": 2.893491124260355e-07, "logits/chosen": -3.625, "logits/rejected": -3.46875, "logps/chosen": -278.0, "logps/rejected": -312.0, "loss": 0.0767, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": 0.1826171875, "rewards/margins": 4.375, "rewards/rejected": -4.1875, "step": 450 }, { "epoch": 1.4696485623003195, "grad_norm": 42.742774206384155, "learning_rate": 2.834319526627219e-07, "logits/chosen": -3.53125, "logits/rejected": -3.5625, "logps/chosen": -278.0, "logps/rejected": -300.0, "loss": 0.1054, "rewards/accuracies": 0.949999988079071, "rewards/chosen": 0.3203125, "rewards/margins": 3.921875, "rewards/rejected": -3.59375, "step": 460 }, { "epoch": 1.5015974440894568, "grad_norm": 31.13201306033402, "learning_rate": 2.7751479289940824e-07, "logits/chosen": -3.5625, "logits/rejected": -3.515625, "logps/chosen": -282.0, "logps/rejected": -310.0, "loss": 0.0996, "rewards/accuracies": 0.925000011920929, "rewards/chosen": -0.21484375, "rewards/margins": 3.8125, "rewards/rejected": -4.03125, "step": 470 }, { "epoch": 1.5335463258785942, "grad_norm": 38.19726342777744, "learning_rate": 2.715976331360947e-07, "logits/chosen": -3.5625, "logits/rejected": -3.515625, "logps/chosen": -290.0, "logps/rejected": -310.0, "loss": 0.1086, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.486328125, "rewards/margins": 3.71875, "rewards/rejected": -4.1875, "step": 480 }, { "epoch": 1.5654952076677318, "grad_norm": 13.612650670554704, "learning_rate": 2.6568047337278106e-07, "logits/chosen": -3.546875, "logits/rejected": -3.5625, "logps/chosen": -276.0, "logps/rejected": -310.0, "loss": 0.1746, "rewards/accuracies": 0.925000011920929, "rewards/chosen": -0.2236328125, "rewards/margins": 3.78125, "rewards/rejected": -4.0, "step": 490 }, { "epoch": 1.5974440894568689, "grad_norm": 29.188006496003524, "learning_rate": 2.5976331360946746e-07, "logits/chosen": -3.53125, "logits/rejected": -3.421875, "logps/chosen": -284.0, "logps/rejected": -304.0, "loss": 0.1413, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": -0.1494140625, "rewards/margins": 3.71875, "rewards/rejected": -3.859375, "step": 500 }, { "epoch": 1.6293929712460065, "grad_norm": 21.252126379670475, "learning_rate": 2.538461538461538e-07, "logits/chosen": -3.59375, "logits/rejected": -3.53125, "logps/chosen": -274.0, "logps/rejected": -304.0, "loss": 0.1281, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": -0.01373291015625, "rewards/margins": 3.546875, "rewards/rejected": -3.5625, "step": 510 }, { "epoch": 1.6613418530351438, "grad_norm": 68.6118608030878, "learning_rate": 2.479289940828402e-07, "logits/chosen": -3.53125, "logits/rejected": -3.46875, "logps/chosen": -276.0, "logps/rejected": -302.0, "loss": 0.1096, "rewards/accuracies": 0.949999988079071, "rewards/chosen": 0.017822265625, "rewards/margins": 4.0625, "rewards/rejected": -4.03125, "step": 520 }, { "epoch": 1.6932907348242812, "grad_norm": 47.978999536964466, "learning_rate": 2.420118343195266e-07, "logits/chosen": -3.625, "logits/rejected": -3.609375, "logps/chosen": -264.0, "logps/rejected": -308.0, "loss": 0.1447, "rewards/accuracies": 0.925000011920929, "rewards/chosen": 0.390625, "rewards/margins": 4.0625, "rewards/rejected": -3.671875, "step": 530 }, { "epoch": 1.7252396166134185, "grad_norm": 13.339031183268695, "learning_rate": 2.36094674556213e-07, "logits/chosen": -3.59375, "logits/rejected": -3.578125, "logps/chosen": -276.0, "logps/rejected": -302.0, "loss": 0.1138, "rewards/accuracies": 0.925000011920929, "rewards/chosen": 0.0257568359375, "rewards/margins": 3.578125, "rewards/rejected": -3.5625, "step": 540 }, { "epoch": 1.7571884984025559, "grad_norm": 33.6403507441883, "learning_rate": 2.301775147928994e-07, "logits/chosen": -3.546875, "logits/rejected": -3.515625, "logps/chosen": -276.0, "logps/rejected": -300.0, "loss": 0.1054, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": -0.004638671875, "rewards/margins": 3.859375, "rewards/rejected": -3.859375, "step": 550 }, { "epoch": 1.7891373801916934, "grad_norm": 41.19004038669282, "learning_rate": 2.242603550295858e-07, "logits/chosen": -3.578125, "logits/rejected": -3.484375, "logps/chosen": -286.0, "logps/rejected": -300.0, "loss": 0.1276, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": 0.3359375, "rewards/margins": 4.125, "rewards/rejected": -3.78125, "step": 560 }, { "epoch": 1.8210862619808306, "grad_norm": 49.203221246475685, "learning_rate": 2.1834319526627217e-07, "logits/chosen": -3.578125, "logits/rejected": -3.5, "logps/chosen": -280.0, "logps/rejected": -298.0, "loss": 0.1439, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": 0.12255859375, "rewards/margins": 3.796875, "rewards/rejected": -3.671875, "step": 570 }, { "epoch": 1.8530351437699681, "grad_norm": 28.35702819808351, "learning_rate": 2.1242603550295858e-07, "logits/chosen": -3.546875, "logits/rejected": -3.5625, "logps/chosen": -286.0, "logps/rejected": -304.0, "loss": 0.0957, "rewards/accuracies": 1.0, "rewards/chosen": -0.07373046875, "rewards/margins": 4.0625, "rewards/rejected": -4.125, "step": 580 }, { "epoch": 1.8849840255591053, "grad_norm": 28.51988813629046, "learning_rate": 2.0650887573964496e-07, "logits/chosen": -3.625, "logits/rejected": -3.5625, "logps/chosen": -296.0, "logps/rejected": -302.0, "loss": 0.1232, "rewards/accuracies": 0.9375, "rewards/chosen": -0.154296875, "rewards/margins": 3.75, "rewards/rejected": -3.90625, "step": 590 }, { "epoch": 1.9169329073482428, "grad_norm": 40.515944424275894, "learning_rate": 2.0059171597633133e-07, "logits/chosen": -3.5625, "logits/rejected": -3.578125, "logps/chosen": -282.0, "logps/rejected": -304.0, "loss": 0.1354, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": 0.2060546875, "rewards/margins": 4.03125, "rewards/rejected": -3.8125, "step": 600 }, { "epoch": 1.9488817891373802, "grad_norm": 39.02302040635236, "learning_rate": 1.9467455621301774e-07, "logits/chosen": -3.59375, "logits/rejected": -3.484375, "logps/chosen": -280.0, "logps/rejected": -306.0, "loss": 0.1326, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.134765625, "rewards/margins": 3.75, "rewards/rejected": -3.875, "step": 610 }, { "epoch": 1.9808306709265175, "grad_norm": 37.23876879370177, "learning_rate": 1.8875739644970412e-07, "logits/chosen": -3.453125, "logits/rejected": -3.453125, "logps/chosen": -278.0, "logps/rejected": -310.0, "loss": 0.1162, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": 0.126953125, "rewards/margins": 3.875, "rewards/rejected": -3.75, "step": 620 }, { "epoch": 2.0, "eval_logits/chosen": -3.5, "eval_logits/rejected": -3.546875, "eval_logps/chosen": -284.0, "eval_logps/rejected": -296.0, "eval_loss": 0.5545507669448853, "eval_rewards/accuracies": 0.7142857313156128, "eval_rewards/chosen": -1.3984375, "eval_rewards/margins": 0.91796875, "eval_rewards/rejected": -2.3125, "eval_runtime": 15.6333, "eval_samples_per_second": 12.793, "eval_steps_per_second": 0.448, "step": 626 }, { "epoch": 2.012779552715655, "grad_norm": 3.7674705382997487, "learning_rate": 1.8284023668639053e-07, "logits/chosen": -3.5625, "logits/rejected": -3.5, "logps/chosen": -268.0, "logps/rejected": -308.0, "loss": 0.0725, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.474609375, "rewards/margins": 4.4375, "rewards/rejected": -3.96875, "step": 630 }, { "epoch": 2.0447284345047922, "grad_norm": 2.711628532479405, "learning_rate": 1.7692307692307693e-07, "logits/chosen": -3.5, "logits/rejected": -3.5, "logps/chosen": -280.0, "logps/rejected": -308.0, "loss": 0.0174, "rewards/accuracies": 1.0, "rewards/chosen": 0.107421875, "rewards/margins": 5.125, "rewards/rejected": -5.03125, "step": 640 }, { "epoch": 2.07667731629393, "grad_norm": 6.9237131222668875, "learning_rate": 1.710059171597633e-07, "logits/chosen": -3.59375, "logits/rejected": -3.53125, "logps/chosen": -274.0, "logps/rejected": -320.0, "loss": 0.0304, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.287109375, "rewards/margins": 5.0625, "rewards/rejected": -4.75, "step": 650 }, { "epoch": 2.108626198083067, "grad_norm": 14.032379088465959, "learning_rate": 1.650887573964497e-07, "logits/chosen": -3.578125, "logits/rejected": -3.515625, "logps/chosen": -280.0, "logps/rejected": -312.0, "loss": 0.0168, "rewards/accuracies": 1.0, "rewards/chosen": 0.30859375, "rewards/margins": 5.25, "rewards/rejected": -4.9375, "step": 660 }, { "epoch": 2.1405750798722045, "grad_norm": 3.7691372037099744, "learning_rate": 1.591715976331361e-07, "logits/chosen": -3.4375, "logits/rejected": -3.40625, "logps/chosen": -280.0, "logps/rejected": -322.0, "loss": 0.0213, "rewards/accuracies": 1.0, "rewards/chosen": 0.13671875, "rewards/margins": 5.5, "rewards/rejected": -5.375, "step": 670 }, { "epoch": 2.1725239616613417, "grad_norm": 11.520705813072723, "learning_rate": 1.5325443786982248e-07, "logits/chosen": -3.546875, "logits/rejected": -3.4375, "logps/chosen": -272.0, "logps/rejected": -320.0, "loss": 0.0237, "rewards/accuracies": 1.0, "rewards/chosen": 0.50390625, "rewards/margins": 5.5625, "rewards/rejected": -5.0625, "step": 680 }, { "epoch": 2.2044728434504792, "grad_norm": 11.168075169911546, "learning_rate": 1.4733727810650885e-07, "logits/chosen": -3.421875, "logits/rejected": -3.375, "logps/chosen": -278.0, "logps/rejected": -328.0, "loss": 0.0248, "rewards/accuracies": 1.0, "rewards/chosen": -0.10791015625, "rewards/margins": 5.34375, "rewards/rejected": -5.4375, "step": 690 }, { "epoch": 2.236421725239617, "grad_norm": 14.410493738827485, "learning_rate": 1.4142011834319526e-07, "logits/chosen": -3.421875, "logits/rejected": -3.375, "logps/chosen": -276.0, "logps/rejected": -314.0, "loss": 0.0269, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.359375, "rewards/margins": 5.5625, "rewards/rejected": -5.1875, "step": 700 }, { "epoch": 2.268370607028754, "grad_norm": 14.7083036532153, "learning_rate": 1.3550295857988164e-07, "logits/chosen": -3.46875, "logits/rejected": -3.515625, "logps/chosen": -270.0, "logps/rejected": -324.0, "loss": 0.0278, "rewards/accuracies": 1.0, "rewards/chosen": 0.16796875, "rewards/margins": 5.875, "rewards/rejected": -5.6875, "step": 710 }, { "epoch": 2.3003194888178915, "grad_norm": 5.626342910377171, "learning_rate": 1.2958579881656802e-07, "logits/chosen": -3.484375, "logits/rejected": -3.390625, "logps/chosen": -276.0, "logps/rejected": -322.0, "loss": 0.0193, "rewards/accuracies": 1.0, "rewards/chosen": 0.59765625, "rewards/margins": 6.21875, "rewards/rejected": -5.625, "step": 720 }, { "epoch": 2.3322683706070286, "grad_norm": 13.550496699505759, "learning_rate": 1.2366863905325443e-07, "logits/chosen": -3.46875, "logits/rejected": -3.484375, "logps/chosen": -276.0, "logps/rejected": -316.0, "loss": 0.0171, "rewards/accuracies": 1.0, "rewards/chosen": 0.486328125, "rewards/margins": 5.90625, "rewards/rejected": -5.4375, "step": 730 }, { "epoch": 2.364217252396166, "grad_norm": 14.488022500712189, "learning_rate": 1.1775147928994082e-07, "logits/chosen": -3.46875, "logits/rejected": -3.375, "logps/chosen": -280.0, "logps/rejected": -320.0, "loss": 0.014, "rewards/accuracies": 1.0, "rewards/chosen": 0.1455078125, "rewards/margins": 5.78125, "rewards/rejected": -5.625, "step": 740 }, { "epoch": 2.3961661341853033, "grad_norm": 6.315943736821987, "learning_rate": 1.1183431952662721e-07, "logits/chosen": -3.4375, "logits/rejected": -3.375, "logps/chosen": -286.0, "logps/rejected": -320.0, "loss": 0.0148, "rewards/accuracies": 1.0, "rewards/chosen": -0.146484375, "rewards/margins": 6.0, "rewards/rejected": -6.15625, "step": 750 }, { "epoch": 2.428115015974441, "grad_norm": 11.454733824781856, "learning_rate": 1.059171597633136e-07, "logits/chosen": -3.4375, "logits/rejected": -3.390625, "logps/chosen": -274.0, "logps/rejected": -324.0, "loss": 0.0174, "rewards/accuracies": 1.0, "rewards/chosen": 0.314453125, "rewards/margins": 5.90625, "rewards/rejected": -5.59375, "step": 760 }, { "epoch": 2.460063897763578, "grad_norm": 6.0042969880614905, "learning_rate": 1e-07, "logits/chosen": -3.484375, "logits/rejected": -3.46875, "logps/chosen": -282.0, "logps/rejected": -328.0, "loss": 0.0302, "rewards/accuracies": 1.0, "rewards/chosen": 0.02001953125, "rewards/margins": 6.34375, "rewards/rejected": -6.3125, "step": 770 }, { "epoch": 2.4920127795527156, "grad_norm": 10.72173710281155, "learning_rate": 9.408284023668639e-08, "logits/chosen": -3.453125, "logits/rejected": -3.46875, "logps/chosen": -278.0, "logps/rejected": -334.0, "loss": 0.0114, "rewards/accuracies": 1.0, "rewards/chosen": 0.0026397705078125, "rewards/margins": 6.0625, "rewards/rejected": -6.0625, "step": 780 }, { "epoch": 2.523961661341853, "grad_norm": 24.493058061439847, "learning_rate": 8.816568047337278e-08, "logits/chosen": -3.40625, "logits/rejected": -3.359375, "logps/chosen": -294.0, "logps/rejected": -324.0, "loss": 0.021, "rewards/accuracies": 1.0, "rewards/chosen": 0.1005859375, "rewards/margins": 6.15625, "rewards/rejected": -6.0625, "step": 790 }, { "epoch": 2.5559105431309903, "grad_norm": 15.886645515243313, "learning_rate": 8.224852071005916e-08, "logits/chosen": -3.65625, "logits/rejected": -3.46875, "logps/chosen": -280.0, "logps/rejected": -328.0, "loss": 0.0234, "rewards/accuracies": 1.0, "rewards/chosen": -0.07861328125, "rewards/margins": 6.375, "rewards/rejected": -6.46875, "step": 800 }, { "epoch": 2.587859424920128, "grad_norm": 3.9555619209445485, "learning_rate": 7.633136094674555e-08, "logits/chosen": -3.578125, "logits/rejected": -3.53125, "logps/chosen": -276.0, "logps/rejected": -338.0, "loss": 0.0164, "rewards/accuracies": 1.0, "rewards/chosen": -0.396484375, "rewards/margins": 6.15625, "rewards/rejected": -6.5625, "step": 810 }, { "epoch": 2.619808306709265, "grad_norm": 2.6920925261494197, "learning_rate": 7.041420118343195e-08, "logits/chosen": -3.578125, "logits/rejected": -3.53125, "logps/chosen": -268.0, "logps/rejected": -322.0, "loss": 0.0187, "rewards/accuracies": 1.0, "rewards/chosen": 0.1943359375, "rewards/margins": 5.96875, "rewards/rejected": -5.78125, "step": 820 }, { "epoch": 2.6517571884984026, "grad_norm": 28.104820989043308, "learning_rate": 6.449704142011835e-08, "logits/chosen": -3.546875, "logits/rejected": -3.453125, "logps/chosen": -286.0, "logps/rejected": -332.0, "loss": 0.0338, "rewards/accuracies": 1.0, "rewards/chosen": -0.0654296875, "rewards/margins": 6.34375, "rewards/rejected": -6.40625, "step": 830 }, { "epoch": 2.68370607028754, "grad_norm": 38.852024663847075, "learning_rate": 5.857988165680473e-08, "logits/chosen": -3.515625, "logits/rejected": -3.53125, "logps/chosen": -278.0, "logps/rejected": -340.0, "loss": 0.016, "rewards/accuracies": 1.0, "rewards/chosen": 0.111328125, "rewards/margins": 6.71875, "rewards/rejected": -6.59375, "step": 840 }, { "epoch": 2.7156549520766773, "grad_norm": 11.470307614538621, "learning_rate": 5.266272189349112e-08, "logits/chosen": -3.46875, "logits/rejected": -3.46875, "logps/chosen": -276.0, "logps/rejected": -334.0, "loss": 0.014, "rewards/accuracies": 1.0, "rewards/chosen": 0.26953125, "rewards/margins": 6.3125, "rewards/rejected": -6.0625, "step": 850 }, { "epoch": 2.747603833865815, "grad_norm": 13.021132027417432, "learning_rate": 4.674556213017751e-08, "logits/chosen": -3.53125, "logits/rejected": -3.453125, "logps/chosen": -278.0, "logps/rejected": -324.0, "loss": 0.0154, "rewards/accuracies": 1.0, "rewards/chosen": -0.06640625, "rewards/margins": 6.125, "rewards/rejected": -6.1875, "step": 860 }, { "epoch": 2.779552715654952, "grad_norm": 36.49663859440145, "learning_rate": 4.082840236686391e-08, "logits/chosen": -3.625, "logits/rejected": -3.5, "logps/chosen": -272.0, "logps/rejected": -326.0, "loss": 0.0161, "rewards/accuracies": 1.0, "rewards/chosen": 0.12109375, "rewards/margins": 6.4375, "rewards/rejected": -6.34375, "step": 870 }, { "epoch": 2.8115015974440896, "grad_norm": 29.264122978741245, "learning_rate": 3.4911242603550294e-08, "logits/chosen": -3.5625, "logits/rejected": -3.484375, "logps/chosen": -286.0, "logps/rejected": -324.0, "loss": 0.0243, "rewards/accuracies": 1.0, "rewards/chosen": -0.04345703125, "rewards/margins": 6.09375, "rewards/rejected": -6.15625, "step": 880 }, { "epoch": 2.8434504792332267, "grad_norm": 2.86898517845962, "learning_rate": 2.8994082840236687e-08, "logits/chosen": -3.484375, "logits/rejected": -3.4375, "logps/chosen": -276.0, "logps/rejected": -330.0, "loss": 0.0141, "rewards/accuracies": 1.0, "rewards/chosen": 0.234375, "rewards/margins": 6.34375, "rewards/rejected": -6.09375, "step": 890 }, { "epoch": 2.8753993610223643, "grad_norm": 10.07036748160809, "learning_rate": 2.3076923076923076e-08, "logits/chosen": -3.5625, "logits/rejected": -3.46875, "logps/chosen": -280.0, "logps/rejected": -328.0, "loss": 0.0145, "rewards/accuracies": 1.0, "rewards/chosen": -0.267578125, "rewards/margins": 6.0, "rewards/rejected": -6.28125, "step": 900 }, { "epoch": 2.9073482428115014, "grad_norm": 12.66492731535328, "learning_rate": 1.7159763313609465e-08, "logits/chosen": -3.53125, "logits/rejected": -3.5625, "logps/chosen": -274.0, "logps/rejected": -328.0, "loss": 0.011, "rewards/accuracies": 1.0, "rewards/chosen": 0.056884765625, "rewards/margins": 6.3125, "rewards/rejected": -6.25, "step": 910 }, { "epoch": 2.939297124600639, "grad_norm": 11.775632569946291, "learning_rate": 1.1242603550295858e-08, "logits/chosen": -3.46875, "logits/rejected": -3.375, "logps/chosen": -288.0, "logps/rejected": -334.0, "loss": 0.0128, "rewards/accuracies": 1.0, "rewards/chosen": 0.0284423828125, "rewards/margins": 6.59375, "rewards/rejected": -6.5625, "step": 920 }, { "epoch": 2.9712460063897765, "grad_norm": 22.713447013582453, "learning_rate": 5.325443786982248e-09, "logits/chosen": -3.515625, "logits/rejected": -3.5, "logps/chosen": -278.0, "logps/rejected": -328.0, "loss": 0.0164, "rewards/accuracies": 1.0, "rewards/chosen": -0.4375, "rewards/margins": 6.21875, "rewards/rejected": -6.65625, "step": 930 }, { "epoch": 3.0, "eval_logits/chosen": -3.5, "eval_logits/rejected": -3.5625, "eval_logps/chosen": -294.0, "eval_logps/rejected": -312.0, "eval_loss": 0.6788671612739563, "eval_rewards/accuracies": 0.6785714030265808, "eval_rewards/chosen": -2.4375, "eval_rewards/margins": 1.5234375, "eval_rewards/rejected": -3.953125, "eval_runtime": 18.8652, "eval_samples_per_second": 10.602, "eval_steps_per_second": 0.371, "step": 939 }, { "epoch": 3.0, "step": 939, "total_flos": 0.0, "train_loss": 0.23421857913081257, "train_runtime": 6022.7058, "train_samples_per_second": 4.975, "train_steps_per_second": 0.156 } ], "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 }