{ "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": 99.19212388021452, "learning_rate": 5.3191489361702125e-09, "logits/chosen": -3.578125, "logits/rejected": -3.46875, "logps/chosen": -247.0, "logps/rejected": -245.0, "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": 100.8587700053837, "learning_rate": 5.3191489361702123e-08, "logits/chosen": -3.609375, "logits/rejected": -3.4375, "logps/chosen": -274.0, "logps/rejected": -264.0, "loss": 0.6877, "rewards/accuracies": 0.125, "rewards/chosen": 0.002777099609375, "rewards/margins": 0.0111083984375, "rewards/rejected": -0.00836181640625, "step": 10 }, { "epoch": 0.06389776357827476, "grad_norm": 83.15951987287777, "learning_rate": 1.0638297872340425e-07, "logits/chosen": -3.609375, "logits/rejected": -3.5625, "logps/chosen": -274.0, "logps/rejected": -264.0, "loss": 0.6473, "rewards/accuracies": 0.6000000238418579, "rewards/chosen": 0.00872802734375, "rewards/margins": 0.08544921875, "rewards/rejected": -0.076171875, "step": 20 }, { "epoch": 0.09584664536741214, "grad_norm": 93.7637276950002, "learning_rate": 1.5957446808510638e-07, "logits/chosen": -3.609375, "logits/rejected": -3.515625, "logps/chosen": -284.0, "logps/rejected": -252.0, "loss": 0.5509, "rewards/accuracies": 0.7749999761581421, "rewards/chosen": 0.07275390625, "rewards/margins": 0.390625, "rewards/rejected": -0.318359375, "step": 30 }, { "epoch": 0.12779552715654952, "grad_norm": 74.20143960903192, "learning_rate": 2.127659574468085e-07, "logits/chosen": -3.46875, "logits/rejected": -3.375, "logps/chosen": -282.0, "logps/rejected": -268.0, "loss": 0.5294, "rewards/accuracies": 0.8500000238418579, "rewards/chosen": 0.251953125, "rewards/margins": 0.7578125, "rewards/rejected": -0.50390625, "step": 40 }, { "epoch": 0.1597444089456869, "grad_norm": 71.1996189930263, "learning_rate": 2.659574468085106e-07, "logits/chosen": -3.578125, "logits/rejected": -3.515625, "logps/chosen": -274.0, "logps/rejected": -260.0, "loss": 0.5293, "rewards/accuracies": 0.7749999761581421, "rewards/chosen": 0.287109375, "rewards/margins": 0.90234375, "rewards/rejected": -0.6171875, "step": 50 }, { "epoch": 0.19169329073482427, "grad_norm": 59.86423059197552, "learning_rate": 3.1914893617021275e-07, "logits/chosen": -3.5625, "logits/rejected": -3.453125, "logps/chosen": -280.0, "logps/rejected": -270.0, "loss": 0.5133, "rewards/accuracies": 0.8125, "rewards/chosen": 0.263671875, "rewards/margins": 0.94140625, "rewards/rejected": -0.67578125, "step": 60 }, { "epoch": 0.22364217252396165, "grad_norm": 61.01100752295743, "learning_rate": 3.7234042553191484e-07, "logits/chosen": -3.5, "logits/rejected": -3.421875, "logps/chosen": -272.0, "logps/rejected": -272.0, "loss": 0.5364, "rewards/accuracies": 0.7875000238418579, "rewards/chosen": 0.2412109375, "rewards/margins": 0.94140625, "rewards/rejected": -0.69921875, "step": 70 }, { "epoch": 0.25559105431309903, "grad_norm": 69.23399604165094, "learning_rate": 4.25531914893617e-07, "logits/chosen": -3.421875, "logits/rejected": -3.40625, "logps/chosen": -274.0, "logps/rejected": -274.0, "loss": 0.507, "rewards/accuracies": 0.6625000238418579, "rewards/chosen": -0.0260009765625, "rewards/margins": 0.81640625, "rewards/rejected": -0.84375, "step": 80 }, { "epoch": 0.28753993610223644, "grad_norm": 91.4754984927698, "learning_rate": 4.787234042553192e-07, "logits/chosen": -3.453125, "logits/rejected": -3.40625, "logps/chosen": -282.0, "logps/rejected": -266.0, "loss": 0.4801, "rewards/accuracies": 0.800000011920929, "rewards/chosen": 0.07421875, "rewards/margins": 0.9453125, "rewards/rejected": -0.87109375, "step": 90 }, { "epoch": 0.3194888178913738, "grad_norm": 72.3540416301487, "learning_rate": 4.964497041420119e-07, "logits/chosen": -3.421875, "logits/rejected": -3.328125, "logps/chosen": -286.0, "logps/rejected": -264.0, "loss": 0.4994, "rewards/accuracies": 0.75, "rewards/chosen": 0.0986328125, "rewards/margins": 1.09375, "rewards/rejected": -1.0, "step": 100 }, { "epoch": 0.3514376996805112, "grad_norm": 72.88717398749043, "learning_rate": 4.905325443786982e-07, "logits/chosen": -3.453125, "logits/rejected": -3.46875, "logps/chosen": -282.0, "logps/rejected": -276.0, "loss": 0.5594, "rewards/accuracies": 0.7124999761581421, "rewards/chosen": -0.0693359375, "rewards/margins": 0.84375, "rewards/rejected": -0.9140625, "step": 110 }, { "epoch": 0.38338658146964855, "grad_norm": 48.378225872153784, "learning_rate": 4.846153846153846e-07, "logits/chosen": -3.53125, "logits/rejected": -3.375, "logps/chosen": -280.0, "logps/rejected": -270.0, "loss": 0.4589, "rewards/accuracies": 0.8374999761581421, "rewards/chosen": 0.36328125, "rewards/margins": 1.1875, "rewards/rejected": -0.8203125, "step": 120 }, { "epoch": 0.41533546325878595, "grad_norm": 97.95130569742476, "learning_rate": 4.78698224852071e-07, "logits/chosen": -3.40625, "logits/rejected": -3.375, "logps/chosen": -278.0, "logps/rejected": -274.0, "loss": 0.4837, "rewards/accuracies": 0.762499988079071, "rewards/chosen": 0.0101318359375, "rewards/margins": 1.2421875, "rewards/rejected": -1.234375, "step": 130 }, { "epoch": 0.4472843450479233, "grad_norm": 81.43086945039742, "learning_rate": 4.727810650887574e-07, "logits/chosen": -3.375, "logits/rejected": -3.390625, "logps/chosen": -276.0, "logps/rejected": -270.0, "loss": 0.5687, "rewards/accuracies": 0.6875, "rewards/chosen": -0.10107421875, "rewards/margins": 1.0078125, "rewards/rejected": -1.109375, "step": 140 }, { "epoch": 0.4792332268370607, "grad_norm": 73.82888993395747, "learning_rate": 4.668639053254438e-07, "logits/chosen": -3.390625, "logits/rejected": -3.375, "logps/chosen": -284.0, "logps/rejected": -274.0, "loss": 0.5149, "rewards/accuracies": 0.762499988079071, "rewards/chosen": -0.072265625, "rewards/margins": 1.25, "rewards/rejected": -1.3203125, "step": 150 }, { "epoch": 0.5111821086261981, "grad_norm": 76.07986743694904, "learning_rate": 4.6094674556213014e-07, "logits/chosen": -3.453125, "logits/rejected": -3.390625, "logps/chosen": -286.0, "logps/rejected": -270.0, "loss": 0.5211, "rewards/accuracies": 0.7875000238418579, "rewards/chosen": 0.023681640625, "rewards/margins": 1.5, "rewards/rejected": -1.4765625, "step": 160 }, { "epoch": 0.5431309904153354, "grad_norm": 54.39165794817972, "learning_rate": 4.5502958579881655e-07, "logits/chosen": -3.53125, "logits/rejected": -3.421875, "logps/chosen": -278.0, "logps/rejected": -276.0, "loss": 0.4454, "rewards/accuracies": 0.762499988079071, "rewards/chosen": 0.10009765625, "rewards/margins": 1.6171875, "rewards/rejected": -1.5234375, "step": 170 }, { "epoch": 0.5750798722044729, "grad_norm": 71.47811041177121, "learning_rate": 4.491124260355029e-07, "logits/chosen": -3.484375, "logits/rejected": -3.25, "logps/chosen": -278.0, "logps/rejected": -278.0, "loss": 0.5029, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": 0.02392578125, "rewards/margins": 1.34375, "rewards/rejected": -1.3203125, "step": 180 }, { "epoch": 0.6070287539936102, "grad_norm": 63.19140078208231, "learning_rate": 4.4319526627218936e-07, "logits/chosen": -3.421875, "logits/rejected": -3.296875, "logps/chosen": -284.0, "logps/rejected": -278.0, "loss": 0.4525, "rewards/accuracies": 0.762499988079071, "rewards/chosen": 0.0172119140625, "rewards/margins": 1.703125, "rewards/rejected": -1.6875, "step": 190 }, { "epoch": 0.6389776357827476, "grad_norm": 60.16768107522096, "learning_rate": 4.372781065088757e-07, "logits/chosen": -3.34375, "logits/rejected": -3.34375, "logps/chosen": -286.0, "logps/rejected": -272.0, "loss": 0.4489, "rewards/accuracies": 0.7875000238418579, "rewards/chosen": -0.03125, "rewards/margins": 1.6015625, "rewards/rejected": -1.625, "step": 200 }, { "epoch": 0.670926517571885, "grad_norm": 53.1463114314995, "learning_rate": 4.313609467455621e-07, "logits/chosen": -3.359375, "logits/rejected": -3.34375, "logps/chosen": -284.0, "logps/rejected": -276.0, "loss": 0.5062, "rewards/accuracies": 0.800000011920929, "rewards/chosen": -0.158203125, "rewards/margins": 1.40625, "rewards/rejected": -1.5625, "step": 210 }, { "epoch": 0.7028753993610224, "grad_norm": 74.19788896651235, "learning_rate": 4.2544378698224847e-07, "logits/chosen": -3.359375, "logits/rejected": -3.3125, "logps/chosen": -288.0, "logps/rejected": -272.0, "loss": 0.5495, "rewards/accuracies": 0.7124999761581421, "rewards/chosen": -0.1962890625, "rewards/margins": 1.3671875, "rewards/rejected": -1.5625, "step": 220 }, { "epoch": 0.7348242811501597, "grad_norm": 37.264928663032826, "learning_rate": 4.195266272189349e-07, "logits/chosen": -3.453125, "logits/rejected": -3.390625, "logps/chosen": -282.0, "logps/rejected": -272.0, "loss": 0.5136, "rewards/accuracies": 0.762499988079071, "rewards/chosen": -0.12109375, "rewards/margins": 1.359375, "rewards/rejected": -1.484375, "step": 230 }, { "epoch": 0.7667731629392971, "grad_norm": 51.9838491014926, "learning_rate": 4.1360946745562133e-07, "logits/chosen": -3.46875, "logits/rejected": -3.4375, "logps/chosen": -280.0, "logps/rejected": -276.0, "loss": 0.4229, "rewards/accuracies": 0.800000011920929, "rewards/chosen": -0.0771484375, "rewards/margins": 1.359375, "rewards/rejected": -1.4375, "step": 240 }, { "epoch": 0.7987220447284346, "grad_norm": 66.37834170527424, "learning_rate": 4.076923076923077e-07, "logits/chosen": -3.484375, "logits/rejected": -3.46875, "logps/chosen": -282.0, "logps/rejected": -278.0, "loss": 0.491, "rewards/accuracies": 0.862500011920929, "rewards/chosen": -0.058349609375, "rewards/margins": 1.703125, "rewards/rejected": -1.765625, "step": 250 }, { "epoch": 0.8306709265175719, "grad_norm": 56.1953898588245, "learning_rate": 4.017751479289941e-07, "logits/chosen": -3.4375, "logits/rejected": -3.421875, "logps/chosen": -284.0, "logps/rejected": -280.0, "loss": 0.4617, "rewards/accuracies": 0.824999988079071, "rewards/chosen": 0.0272216796875, "rewards/margins": 1.7734375, "rewards/rejected": -1.7421875, "step": 260 }, { "epoch": 0.8626198083067093, "grad_norm": 55.24726339529526, "learning_rate": 3.9585798816568044e-07, "logits/chosen": -3.484375, "logits/rejected": -3.375, "logps/chosen": -278.0, "logps/rejected": -286.0, "loss": 0.4634, "rewards/accuracies": 0.7124999761581421, "rewards/chosen": -0.2060546875, "rewards/margins": 1.3671875, "rewards/rejected": -1.578125, "step": 270 }, { "epoch": 0.8945686900958466, "grad_norm": 52.41202133207622, "learning_rate": 3.8994082840236685e-07, "logits/chosen": -3.46875, "logits/rejected": -3.4375, "logps/chosen": -280.0, "logps/rejected": -274.0, "loss": 0.4426, "rewards/accuracies": 0.8125, "rewards/chosen": 0.1787109375, "rewards/margins": 1.8359375, "rewards/rejected": -1.65625, "step": 280 }, { "epoch": 0.9265175718849841, "grad_norm": 71.23457030042947, "learning_rate": 3.840236686390532e-07, "logits/chosen": -3.53125, "logits/rejected": -3.40625, "logps/chosen": -276.0, "logps/rejected": -280.0, "loss": 0.4684, "rewards/accuracies": 0.7749999761581421, "rewards/chosen": -0.0234375, "rewards/margins": 1.5234375, "rewards/rejected": -1.546875, "step": 290 }, { "epoch": 0.9584664536741214, "grad_norm": 65.27022274437043, "learning_rate": 3.7810650887573966e-07, "logits/chosen": -3.390625, "logits/rejected": -3.375, "logps/chosen": -282.0, "logps/rejected": -280.0, "loss": 0.5252, "rewards/accuracies": 0.800000011920929, "rewards/chosen": -0.11572265625, "rewards/margins": 1.578125, "rewards/rejected": -1.6953125, "step": 300 }, { "epoch": 0.9904153354632588, "grad_norm": 55.83163671794183, "learning_rate": 3.72189349112426e-07, "logits/chosen": -3.5625, "logits/rejected": -3.46875, "logps/chosen": -276.0, "logps/rejected": -278.0, "loss": 0.5266, "rewards/accuracies": 0.7749999761581421, "rewards/chosen": -0.041259765625, "rewards/margins": 1.5859375, "rewards/rejected": -1.625, "step": 310 }, { "epoch": 1.0, "eval_logits/chosen": -3.53125, "eval_logits/rejected": -3.40625, "eval_logps/chosen": -278.0, "eval_logps/rejected": -278.0, "eval_loss": 0.46781250834465027, "eval_rewards/accuracies": 0.6785714030265808, "eval_rewards/chosen": -0.1728515625, "eval_rewards/margins": 1.09375, "eval_rewards/rejected": -1.265625, "eval_runtime": 12.4491, "eval_samples_per_second": 16.065, "eval_steps_per_second": 0.562, "step": 313 }, { "epoch": 1.0223642172523961, "grad_norm": 23.70055727816958, "learning_rate": 3.662721893491124e-07, "logits/chosen": -3.4375, "logits/rejected": -3.375, "logps/chosen": -274.0, "logps/rejected": -284.0, "loss": 0.2261, "rewards/accuracies": 0.9125000238418579, "rewards/chosen": 0.33984375, "rewards/margins": 2.625, "rewards/rejected": -2.28125, "step": 320 }, { "epoch": 1.0543130990415335, "grad_norm": 20.6098819793211, "learning_rate": 3.603550295857988e-07, "logits/chosen": -3.484375, "logits/rejected": -3.40625, "logps/chosen": -270.0, "logps/rejected": -296.0, "loss": 0.1038, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": 0.484375, "rewards/margins": 3.328125, "rewards/rejected": -2.84375, "step": 330 }, { "epoch": 1.0862619808306708, "grad_norm": 12.129974272451467, "learning_rate": 3.544378698224852e-07, "logits/chosen": -3.46875, "logits/rejected": -3.359375, "logps/chosen": -278.0, "logps/rejected": -300.0, "loss": 0.0875, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.4296875, "rewards/margins": 3.359375, "rewards/rejected": -2.921875, "step": 340 }, { "epoch": 1.1182108626198084, "grad_norm": 19.312391737005985, "learning_rate": 3.485207100591716e-07, "logits/chosen": -3.46875, "logits/rejected": -3.3125, "logps/chosen": -270.0, "logps/rejected": -284.0, "loss": 0.1201, "rewards/accuracies": 0.9375, "rewards/chosen": 0.66015625, "rewards/margins": 3.921875, "rewards/rejected": -3.265625, "step": 350 }, { "epoch": 1.1501597444089458, "grad_norm": 16.056335220522016, "learning_rate": 3.42603550295858e-07, "logits/chosen": -3.40625, "logits/rejected": -3.421875, "logps/chosen": -266.0, "logps/rejected": -300.0, "loss": 0.0965, "rewards/accuracies": 0.949999988079071, "rewards/chosen": 0.3671875, "rewards/margins": 3.296875, "rewards/rejected": -2.921875, "step": 360 }, { "epoch": 1.182108626198083, "grad_norm": 14.173305651953171, "learning_rate": 3.366863905325444e-07, "logits/chosen": -3.390625, "logits/rejected": -3.421875, "logps/chosen": -274.0, "logps/rejected": -288.0, "loss": 0.1037, "rewards/accuracies": 0.949999988079071, "rewards/chosen": 0.72265625, "rewards/margins": 3.625, "rewards/rejected": -2.90625, "step": 370 }, { "epoch": 1.2140575079872205, "grad_norm": 28.838482988149984, "learning_rate": 3.3076923076923075e-07, "logits/chosen": -3.421875, "logits/rejected": -3.46875, "logps/chosen": -280.0, "logps/rejected": -286.0, "loss": 0.0885, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": 0.69140625, "rewards/margins": 3.796875, "rewards/rejected": -3.109375, "step": 380 }, { "epoch": 1.2460063897763578, "grad_norm": 18.796201336624538, "learning_rate": 3.2485207100591716e-07, "logits/chosen": -3.515625, "logits/rejected": -3.328125, "logps/chosen": -272.0, "logps/rejected": -298.0, "loss": 0.1175, "rewards/accuracies": 0.925000011920929, "rewards/chosen": 0.546875, "rewards/margins": 3.625, "rewards/rejected": -3.078125, "step": 390 }, { "epoch": 1.2779552715654952, "grad_norm": 12.284250452584237, "learning_rate": 3.189349112426035e-07, "logits/chosen": -3.4375, "logits/rejected": -3.4375, "logps/chosen": -274.0, "logps/rejected": -292.0, "loss": 0.0702, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.7109375, "rewards/margins": 3.859375, "rewards/rejected": -3.15625, "step": 400 }, { "epoch": 1.3099041533546325, "grad_norm": 13.495027410207776, "learning_rate": 3.130177514792899e-07, "logits/chosen": -3.484375, "logits/rejected": -3.40625, "logps/chosen": -272.0, "logps/rejected": -294.0, "loss": 0.087, "rewards/accuracies": 0.949999988079071, "rewards/chosen": 0.578125, "rewards/margins": 3.640625, "rewards/rejected": -3.0625, "step": 410 }, { "epoch": 1.34185303514377, "grad_norm": 22.992235645975622, "learning_rate": 3.071005917159763e-07, "logits/chosen": -3.5, "logits/rejected": -3.421875, "logps/chosen": -286.0, "logps/rejected": -300.0, "loss": 0.1101, "rewards/accuracies": 0.9375, "rewards/chosen": 0.232421875, "rewards/margins": 3.765625, "rewards/rejected": -3.53125, "step": 420 }, { "epoch": 1.3738019169329074, "grad_norm": 21.69400097170527, "learning_rate": 3.011834319526627e-07, "logits/chosen": -3.40625, "logits/rejected": -3.40625, "logps/chosen": -272.0, "logps/rejected": -296.0, "loss": 0.117, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": 0.47265625, "rewards/margins": 3.84375, "rewards/rejected": -3.375, "step": 430 }, { "epoch": 1.4057507987220448, "grad_norm": 18.285863744766086, "learning_rate": 2.952662721893491e-07, "logits/chosen": -3.484375, "logits/rejected": -3.359375, "logps/chosen": -280.0, "logps/rejected": -282.0, "loss": 0.1335, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": 0.51171875, "rewards/margins": 3.578125, "rewards/rejected": -3.0625, "step": 440 }, { "epoch": 1.4376996805111821, "grad_norm": 17.34600664501295, "learning_rate": 2.893491124260355e-07, "logits/chosen": -3.359375, "logits/rejected": -3.34375, "logps/chosen": -284.0, "logps/rejected": -292.0, "loss": 0.083, "rewards/accuracies": 1.0, "rewards/chosen": 0.404296875, "rewards/margins": 4.1875, "rewards/rejected": -3.78125, "step": 450 }, { "epoch": 1.4696485623003195, "grad_norm": 36.80402921857842, "learning_rate": 2.834319526627219e-07, "logits/chosen": -3.4375, "logits/rejected": -3.3125, "logps/chosen": -268.0, "logps/rejected": -302.0, "loss": 0.116, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": 0.287109375, "rewards/margins": 3.671875, "rewards/rejected": -3.390625, "step": 460 }, { "epoch": 1.5015974440894568, "grad_norm": 15.918868054649243, "learning_rate": 2.7751479289940824e-07, "logits/chosen": -3.375, "logits/rejected": -3.375, "logps/chosen": -272.0, "logps/rejected": -292.0, "loss": 0.0565, "rewards/accuracies": 1.0, "rewards/chosen": 0.75390625, "rewards/margins": 4.15625, "rewards/rejected": -3.40625, "step": 470 }, { "epoch": 1.5335463258785942, "grad_norm": 22.888062361116138, "learning_rate": 2.715976331360947e-07, "logits/chosen": -3.390625, "logits/rejected": -3.3125, "logps/chosen": -278.0, "logps/rejected": -292.0, "loss": 0.1266, "rewards/accuracies": 0.925000011920929, "rewards/chosen": 0.640625, "rewards/margins": 3.875, "rewards/rejected": -3.234375, "step": 480 }, { "epoch": 1.5654952076677318, "grad_norm": 30.24738555666603, "learning_rate": 2.6568047337278106e-07, "logits/chosen": -3.34375, "logits/rejected": -3.3125, "logps/chosen": -272.0, "logps/rejected": -292.0, "loss": 0.1042, "rewards/accuracies": 0.949999988079071, "rewards/chosen": 0.78125, "rewards/margins": 4.1875, "rewards/rejected": -3.40625, "step": 490 }, { "epoch": 1.5974440894568689, "grad_norm": 23.550496082239732, "learning_rate": 2.5976331360946746e-07, "logits/chosen": -3.453125, "logits/rejected": -3.421875, "logps/chosen": -278.0, "logps/rejected": -284.0, "loss": 0.1374, "rewards/accuracies": 0.949999988079071, "rewards/chosen": 0.671875, "rewards/margins": 3.6875, "rewards/rejected": -3.015625, "step": 500 }, { "epoch": 1.6293929712460065, "grad_norm": 25.63710072069056, "learning_rate": 2.538461538461538e-07, "logits/chosen": -3.328125, "logits/rejected": -3.34375, "logps/chosen": -268.0, "logps/rejected": -292.0, "loss": 0.1023, "rewards/accuracies": 0.949999988079071, "rewards/chosen": 0.5546875, "rewards/margins": 3.640625, "rewards/rejected": -3.09375, "step": 510 }, { "epoch": 1.6613418530351438, "grad_norm": 24.141726605352023, "learning_rate": 2.479289940828402e-07, "logits/chosen": -3.28125, "logits/rejected": -3.296875, "logps/chosen": -270.0, "logps/rejected": -286.0, "loss": 0.082, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.890625, "rewards/margins": 4.375, "rewards/rejected": -3.5, "step": 520 }, { "epoch": 1.6932907348242812, "grad_norm": 34.35651000503782, "learning_rate": 2.420118343195266e-07, "logits/chosen": -3.421875, "logits/rejected": -3.265625, "logps/chosen": -268.0, "logps/rejected": -294.0, "loss": 0.1029, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": 0.87109375, "rewards/margins": 4.15625, "rewards/rejected": -3.28125, "step": 530 }, { "epoch": 1.7252396166134185, "grad_norm": 44.21100968000028, "learning_rate": 2.36094674556213e-07, "logits/chosen": -3.375, "logits/rejected": -3.34375, "logps/chosen": -262.0, "logps/rejected": -292.0, "loss": 0.1529, "rewards/accuracies": 0.949999988079071, "rewards/chosen": 1.0, "rewards/margins": 4.25, "rewards/rejected": -3.25, "step": 540 }, { "epoch": 1.7571884984025559, "grad_norm": 47.981400895385406, "learning_rate": 2.301775147928994e-07, "logits/chosen": -3.375, "logits/rejected": -3.3125, "logps/chosen": -270.0, "logps/rejected": -292.0, "loss": 0.1231, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": 0.83984375, "rewards/margins": 3.6875, "rewards/rejected": -2.859375, "step": 550 }, { "epoch": 1.7891373801916934, "grad_norm": 38.43543934960892, "learning_rate": 2.242603550295858e-07, "logits/chosen": -3.4375, "logits/rejected": -3.34375, "logps/chosen": -258.0, "logps/rejected": -294.0, "loss": 0.135, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": 0.7578125, "rewards/margins": 3.65625, "rewards/rejected": -2.890625, "step": 560 }, { "epoch": 1.8210862619808306, "grad_norm": 11.780750523956426, "learning_rate": 2.1834319526627217e-07, "logits/chosen": -3.375, "logits/rejected": -3.34375, "logps/chosen": -276.0, "logps/rejected": -288.0, "loss": 0.1099, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": 0.5625, "rewards/margins": 3.859375, "rewards/rejected": -3.296875, "step": 570 }, { "epoch": 1.8530351437699681, "grad_norm": 16.56566595786949, "learning_rate": 2.1242603550295858e-07, "logits/chosen": -3.3125, "logits/rejected": -3.28125, "logps/chosen": -274.0, "logps/rejected": -294.0, "loss": 0.1071, "rewards/accuracies": 0.949999988079071, "rewards/chosen": 0.54296875, "rewards/margins": 3.859375, "rewards/rejected": -3.3125, "step": 580 }, { "epoch": 1.8849840255591053, "grad_norm": 18.132322544979896, "learning_rate": 2.0650887573964496e-07, "logits/chosen": -3.3125, "logits/rejected": -3.21875, "logps/chosen": -270.0, "logps/rejected": -304.0, "loss": 0.0811, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": 1.0078125, "rewards/margins": 4.59375, "rewards/rejected": -3.5625, "step": 590 }, { "epoch": 1.9169329073482428, "grad_norm": 32.52764727647769, "learning_rate": 2.0059171597633133e-07, "logits/chosen": -3.359375, "logits/rejected": -3.25, "logps/chosen": -274.0, "logps/rejected": -292.0, "loss": 0.114, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": 0.6953125, "rewards/margins": 4.15625, "rewards/rejected": -3.46875, "step": 600 }, { "epoch": 1.9488817891373802, "grad_norm": 22.068866501441573, "learning_rate": 1.9467455621301774e-07, "logits/chosen": -3.3125, "logits/rejected": -3.265625, "logps/chosen": -276.0, "logps/rejected": -296.0, "loss": 0.0821, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.88671875, "rewards/margins": 4.4375, "rewards/rejected": -3.546875, "step": 610 }, { "epoch": 1.9808306709265175, "grad_norm": 41.19041486210437, "learning_rate": 1.8875739644970412e-07, "logits/chosen": -3.328125, "logits/rejected": -3.203125, "logps/chosen": -270.0, "logps/rejected": -296.0, "loss": 0.1093, "rewards/accuracies": 0.9375, "rewards/chosen": 0.59765625, "rewards/margins": 4.28125, "rewards/rejected": -3.671875, "step": 620 }, { "epoch": 2.0, "eval_logits/chosen": -3.359375, "eval_logits/rejected": -3.234375, "eval_logps/chosen": -282.0, "eval_logps/rejected": -288.0, "eval_loss": 0.48261719942092896, "eval_rewards/accuracies": 0.7142857313156128, "eval_rewards/chosen": -0.46484375, "eval_rewards/margins": 1.875, "eval_rewards/rejected": -2.34375, "eval_runtime": 12.43, "eval_samples_per_second": 16.09, "eval_steps_per_second": 0.563, "step": 626 }, { "epoch": 2.012779552715655, "grad_norm": 6.421580065788588, "learning_rate": 1.8284023668639053e-07, "logits/chosen": -3.25, "logits/rejected": -3.21875, "logps/chosen": -266.0, "logps/rejected": -298.0, "loss": 0.0314, "rewards/accuracies": 1.0, "rewards/chosen": 0.83984375, "rewards/margins": 4.84375, "rewards/rejected": -4.03125, "step": 630 }, { "epoch": 2.0447284345047922, "grad_norm": 2.3666113233491077, "learning_rate": 1.7692307692307693e-07, "logits/chosen": -3.28125, "logits/rejected": -3.203125, "logps/chosen": -270.0, "logps/rejected": -302.0, "loss": 0.0196, "rewards/accuracies": 1.0, "rewards/chosen": 0.90625, "rewards/margins": 5.0625, "rewards/rejected": -4.15625, "step": 640 }, { "epoch": 2.07667731629393, "grad_norm": 5.743645084563706, "learning_rate": 1.710059171597633e-07, "logits/chosen": -3.296875, "logits/rejected": -3.140625, "logps/chosen": -274.0, "logps/rejected": -300.0, "loss": 0.0206, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 1.03125, "rewards/margins": 5.46875, "rewards/rejected": -4.4375, "step": 650 }, { "epoch": 2.108626198083067, "grad_norm": 7.071815895487423, "learning_rate": 1.650887573964497e-07, "logits/chosen": -3.34375, "logits/rejected": -3.15625, "logps/chosen": -270.0, "logps/rejected": -320.0, "loss": 0.0175, "rewards/accuracies": 1.0, "rewards/chosen": 0.94140625, "rewards/margins": 5.71875, "rewards/rejected": -4.78125, "step": 660 }, { "epoch": 2.1405750798722045, "grad_norm": 5.150989860943889, "learning_rate": 1.591715976331361e-07, "logits/chosen": -3.3125, "logits/rejected": -3.234375, "logps/chosen": -272.0, "logps/rejected": -310.0, "loss": 0.0112, "rewards/accuracies": 1.0, "rewards/chosen": 0.72265625, "rewards/margins": 5.84375, "rewards/rejected": -5.125, "step": 670 }, { "epoch": 2.1725239616613417, "grad_norm": 2.416720212227166, "learning_rate": 1.5325443786982248e-07, "logits/chosen": -3.28125, "logits/rejected": -3.234375, "logps/chosen": -274.0, "logps/rejected": -314.0, "loss": 0.0146, "rewards/accuracies": 1.0, "rewards/chosen": 0.52734375, "rewards/margins": 6.03125, "rewards/rejected": -5.5, "step": 680 }, { "epoch": 2.2044728434504792, "grad_norm": 9.659222003032339, "learning_rate": 1.4733727810650885e-07, "logits/chosen": -3.25, "logits/rejected": -3.1875, "logps/chosen": -276.0, "logps/rejected": -312.0, "loss": 0.0188, "rewards/accuracies": 1.0, "rewards/chosen": 0.6796875, "rewards/margins": 5.9375, "rewards/rejected": -5.28125, "step": 690 }, { "epoch": 2.236421725239617, "grad_norm": 2.531990079316781, "learning_rate": 1.4142011834319526e-07, "logits/chosen": -3.21875, "logits/rejected": -3.171875, "logps/chosen": -272.0, "logps/rejected": -306.0, "loss": 0.0177, "rewards/accuracies": 1.0, "rewards/chosen": 0.96484375, "rewards/margins": 5.5625, "rewards/rejected": -4.59375, "step": 700 }, { "epoch": 2.268370607028754, "grad_norm": 4.859665778231634, "learning_rate": 1.3550295857988164e-07, "logits/chosen": -3.28125, "logits/rejected": -3.21875, "logps/chosen": -266.0, "logps/rejected": -308.0, "loss": 0.0184, "rewards/accuracies": 1.0, "rewards/chosen": 0.7421875, "rewards/margins": 5.625, "rewards/rejected": -4.875, "step": 710 }, { "epoch": 2.3003194888178915, "grad_norm": 3.246076795572837, "learning_rate": 1.2958579881656802e-07, "logits/chosen": -3.28125, "logits/rejected": -3.28125, "logps/chosen": -270.0, "logps/rejected": -308.0, "loss": 0.0161, "rewards/accuracies": 1.0, "rewards/chosen": 0.70703125, "rewards/margins": 6.0625, "rewards/rejected": -5.34375, "step": 720 }, { "epoch": 2.3322683706070286, "grad_norm": 11.323551212209285, "learning_rate": 1.2366863905325443e-07, "logits/chosen": -3.3125, "logits/rejected": -3.1875, "logps/chosen": -260.0, "logps/rejected": -300.0, "loss": 0.026, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.8203125, "rewards/margins": 5.59375, "rewards/rejected": -4.75, "step": 730 }, { "epoch": 2.364217252396166, "grad_norm": 4.32723570514713, "learning_rate": 1.1775147928994082e-07, "logits/chosen": -3.25, "logits/rejected": -3.140625, "logps/chosen": -270.0, "logps/rejected": -320.0, "loss": 0.0169, "rewards/accuracies": 1.0, "rewards/chosen": 0.5859375, "rewards/margins": 5.59375, "rewards/rejected": -5.0, "step": 740 }, { "epoch": 2.3961661341853033, "grad_norm": 7.5510847695494725, "learning_rate": 1.1183431952662721e-07, "logits/chosen": -3.328125, "logits/rejected": -3.171875, "logps/chosen": -274.0, "logps/rejected": -306.0, "loss": 0.0187, "rewards/accuracies": 1.0, "rewards/chosen": 0.625, "rewards/margins": 5.59375, "rewards/rejected": -5.0, "step": 750 }, { "epoch": 2.428115015974441, "grad_norm": 19.062627586636406, "learning_rate": 1.059171597633136e-07, "logits/chosen": -3.234375, "logits/rejected": -3.109375, "logps/chosen": -276.0, "logps/rejected": -318.0, "loss": 0.0363, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.5234375, "rewards/margins": 6.125, "rewards/rejected": -5.625, "step": 760 }, { "epoch": 2.460063897763578, "grad_norm": 8.015175182213541, "learning_rate": 1e-07, "logits/chosen": -3.125, "logits/rejected": -3.140625, "logps/chosen": -280.0, "logps/rejected": -312.0, "loss": 0.0307, "rewards/accuracies": 1.0, "rewards/chosen": 0.404296875, "rewards/margins": 6.09375, "rewards/rejected": -5.6875, "step": 770 }, { "epoch": 2.4920127795527156, "grad_norm": 5.818202304304543, "learning_rate": 9.408284023668639e-08, "logits/chosen": -3.203125, "logits/rejected": -3.125, "logps/chosen": -270.0, "logps/rejected": -330.0, "loss": 0.015, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.365234375, "rewards/margins": 6.0, "rewards/rejected": -5.65625, "step": 780 }, { "epoch": 2.523961661341853, "grad_norm": 2.5130565290456697, "learning_rate": 8.816568047337278e-08, "logits/chosen": -3.28125, "logits/rejected": -3.140625, "logps/chosen": -274.0, "logps/rejected": -324.0, "loss": 0.0171, "rewards/accuracies": 1.0, "rewards/chosen": 0.232421875, "rewards/margins": 5.875, "rewards/rejected": -5.65625, "step": 790 }, { "epoch": 2.5559105431309903, "grad_norm": 1.8203884870356968, "learning_rate": 8.224852071005916e-08, "logits/chosen": -3.1875, "logits/rejected": -3.171875, "logps/chosen": -266.0, "logps/rejected": -328.0, "loss": 0.0117, "rewards/accuracies": 1.0, "rewards/chosen": 0.5390625, "rewards/margins": 6.1875, "rewards/rejected": -5.65625, "step": 800 }, { "epoch": 2.587859424920128, "grad_norm": 4.544569959213599, "learning_rate": 7.633136094674555e-08, "logits/chosen": -3.15625, "logits/rejected": -3.15625, "logps/chosen": -270.0, "logps/rejected": -314.0, "loss": 0.0236, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": 0.7734375, "rewards/margins": 5.90625, "rewards/rejected": -5.125, "step": 810 }, { "epoch": 2.619808306709265, "grad_norm": 11.029690431094323, "learning_rate": 7.041420118343195e-08, "logits/chosen": -3.078125, "logits/rejected": -3.109375, "logps/chosen": -274.0, "logps/rejected": -308.0, "loss": 0.0199, "rewards/accuracies": 1.0, "rewards/chosen": 0.36328125, "rewards/margins": 6.0, "rewards/rejected": -5.65625, "step": 820 }, { "epoch": 2.6517571884984026, "grad_norm": 10.432324186926575, "learning_rate": 6.449704142011835e-08, "logits/chosen": -3.15625, "logits/rejected": -2.984375, "logps/chosen": -266.0, "logps/rejected": -324.0, "loss": 0.016, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.408203125, "rewards/margins": 5.71875, "rewards/rejected": -5.3125, "step": 830 }, { "epoch": 2.68370607028754, "grad_norm": 4.274107984548081, "learning_rate": 5.857988165680473e-08, "logits/chosen": -3.1875, "logits/rejected": -3.0625, "logps/chosen": -272.0, "logps/rejected": -330.0, "loss": 0.0174, "rewards/accuracies": 1.0, "rewards/chosen": 0.15625, "rewards/margins": 6.15625, "rewards/rejected": -6.0, "step": 840 }, { "epoch": 2.7156549520766773, "grad_norm": 2.6182440295915517, "learning_rate": 5.266272189349112e-08, "logits/chosen": -3.21875, "logits/rejected": -3.15625, "logps/chosen": -268.0, "logps/rejected": -314.0, "loss": 0.0111, "rewards/accuracies": 1.0, "rewards/chosen": 0.671875, "rewards/margins": 6.21875, "rewards/rejected": -5.53125, "step": 850 }, { "epoch": 2.747603833865815, "grad_norm": 3.5668059520856636, "learning_rate": 4.674556213017751e-08, "logits/chosen": -3.078125, "logits/rejected": -3.046875, "logps/chosen": -276.0, "logps/rejected": -322.0, "loss": 0.0177, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.15234375, "rewards/margins": 6.40625, "rewards/rejected": -6.25, "step": 860 }, { "epoch": 2.779552715654952, "grad_norm": 5.155480729398528, "learning_rate": 4.082840236686391e-08, "logits/chosen": -3.203125, "logits/rejected": -3.09375, "logps/chosen": -272.0, "logps/rejected": -312.0, "loss": 0.0197, "rewards/accuracies": 1.0, "rewards/chosen": 0.671875, "rewards/margins": 5.9375, "rewards/rejected": -5.28125, "step": 870 }, { "epoch": 2.8115015974440896, "grad_norm": 7.714079734104633, "learning_rate": 3.4911242603550294e-08, "logits/chosen": -3.0625, "logits/rejected": -3.140625, "logps/chosen": -282.0, "logps/rejected": -320.0, "loss": 0.0243, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.3359375, "rewards/margins": 6.3125, "rewards/rejected": -5.96875, "step": 880 }, { "epoch": 2.8434504792332267, "grad_norm": 6.619078689284992, "learning_rate": 2.8994082840236687e-08, "logits/chosen": -3.171875, "logits/rejected": -3.078125, "logps/chosen": -268.0, "logps/rejected": -318.0, "loss": 0.0135, "rewards/accuracies": 1.0, "rewards/chosen": 0.8984375, "rewards/margins": 6.5625, "rewards/rejected": -5.6875, "step": 890 }, { "epoch": 2.8753993610223643, "grad_norm": 13.024583688697536, "learning_rate": 2.3076923076923076e-08, "logits/chosen": -3.25, "logits/rejected": -3.171875, "logps/chosen": -266.0, "logps/rejected": -318.0, "loss": 0.0232, "rewards/accuracies": 1.0, "rewards/chosen": 0.73828125, "rewards/margins": 6.59375, "rewards/rejected": -5.84375, "step": 900 }, { "epoch": 2.9073482428115014, "grad_norm": 5.308130568190089, "learning_rate": 1.7159763313609465e-08, "logits/chosen": -3.15625, "logits/rejected": -3.125, "logps/chosen": -274.0, "logps/rejected": -322.0, "loss": 0.0087, "rewards/accuracies": 1.0, "rewards/chosen": 0.43359375, "rewards/margins": 6.59375, "rewards/rejected": -6.15625, "step": 910 }, { "epoch": 2.939297124600639, "grad_norm": 25.167822205455, "learning_rate": 1.1242603550295858e-08, "logits/chosen": -3.09375, "logits/rejected": -2.96875, "logps/chosen": -286.0, "logps/rejected": -326.0, "loss": 0.0148, "rewards/accuracies": 1.0, "rewards/chosen": 0.5078125, "rewards/margins": 6.90625, "rewards/rejected": -6.375, "step": 920 }, { "epoch": 2.9712460063897765, "grad_norm": 7.429801493899376, "learning_rate": 5.325443786982248e-09, "logits/chosen": -3.15625, "logits/rejected": -3.03125, "logps/chosen": -272.0, "logps/rejected": -324.0, "loss": 0.0193, "rewards/accuracies": 1.0, "rewards/chosen": 0.61328125, "rewards/margins": 6.5, "rewards/rejected": -5.875, "step": 930 }, { "epoch": 3.0, "eval_logits/chosen": -3.25, "eval_logits/rejected": -3.09375, "eval_logps/chosen": -290.0, "eval_logps/rejected": -302.0, "eval_loss": 0.5599316358566284, "eval_rewards/accuracies": 0.7678571343421936, "eval_rewards/chosen": -1.296875, "eval_rewards/margins": 2.484375, "eval_rewards/rejected": -3.78125, "eval_runtime": 15.5752, "eval_samples_per_second": 12.841, "eval_steps_per_second": 0.449, "step": 939 }, { "epoch": 3.0, "step": 939, "total_flos": 0.0, "train_loss": 0.21089639145726213, "train_runtime": 4832.6259, "train_samples_per_second": 6.207, "train_steps_per_second": 0.194 } ], "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 }