{ "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": 32.14257471485931, "learning_rate": 5.3191489361702125e-09, "logits/chosen": -1.7265625, "logits/rejected": -1.65625, "logps/chosen": -249.0, "logps/rejected": -242.0, "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": 33.76096895998352, "learning_rate": 5.3191489361702123e-08, "logits/chosen": -1.765625, "logits/rejected": -1.765625, "logps/chosen": -240.0, "logps/rejected": -244.0, "loss": 0.6926, "rewards/accuracies": 0.2083333283662796, "rewards/chosen": 0.00555419921875, "rewards/margins": 0.0, "rewards/rejected": 0.00555419921875, "step": 10 }, { "epoch": 0.06389776357827476, "grad_norm": 31.259900823289186, "learning_rate": 1.0638297872340425e-07, "logits/chosen": -1.71875, "logits/rejected": -1.75, "logps/chosen": -237.0, "logps/rejected": -241.0, "loss": 0.6916, "rewards/accuracies": 0.26249998807907104, "rewards/chosen": 0.0, "rewards/margins": 0.006256103515625, "rewards/rejected": -0.006256103515625, "step": 20 }, { "epoch": 0.09584664536741214, "grad_norm": 32.366987695897166, "learning_rate": 1.5957446808510638e-07, "logits/chosen": -1.75, "logits/rejected": -1.75, "logps/chosen": -245.0, "logps/rejected": -246.0, "loss": 0.6946, "rewards/accuracies": 0.22499999403953552, "rewards/chosen": 6.109476089477539e-06, "rewards/margins": -0.0012359619140625, "rewards/rejected": 0.0012664794921875, "step": 30 }, { "epoch": 0.12779552715654952, "grad_norm": 30.81764382136956, "learning_rate": 2.127659574468085e-07, "logits/chosen": -1.703125, "logits/rejected": -1.734375, "logps/chosen": -243.0, "logps/rejected": -240.0, "loss": 0.6818, "rewards/accuracies": 0.4000000059604645, "rewards/chosen": -0.003753662109375, "rewards/margins": 0.032470703125, "rewards/rejected": -0.036376953125, "step": 40 }, { "epoch": 0.1597444089456869, "grad_norm": 31.214992328374773, "learning_rate": 2.659574468085106e-07, "logits/chosen": -1.7109375, "logits/rejected": -1.7265625, "logps/chosen": -242.0, "logps/rejected": -243.0, "loss": 0.673, "rewards/accuracies": 0.4375, "rewards/chosen": -0.0400390625, "rewards/margins": 0.033935546875, "rewards/rejected": -0.07373046875, "step": 50 }, { "epoch": 0.19169329073482427, "grad_norm": 28.88257684001896, "learning_rate": 3.1914893617021275e-07, "logits/chosen": -1.7265625, "logits/rejected": -1.75, "logps/chosen": -240.0, "logps/rejected": -243.0, "loss": 0.6296, "rewards/accuracies": 0.574999988079071, "rewards/chosen": -0.0625, "rewards/margins": 0.111328125, "rewards/rejected": -0.173828125, "step": 60 }, { "epoch": 0.22364217252396165, "grad_norm": 30.584634735099684, "learning_rate": 3.7234042553191484e-07, "logits/chosen": -1.7109375, "logits/rejected": -1.71875, "logps/chosen": -243.0, "logps/rejected": -250.0, "loss": 0.6439, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": -0.080078125, "rewards/margins": 0.19140625, "rewards/rejected": -0.271484375, "step": 70 }, { "epoch": 0.25559105431309903, "grad_norm": 29.751500863088314, "learning_rate": 4.25531914893617e-07, "logits/chosen": -1.703125, "logits/rejected": -1.75, "logps/chosen": -240.0, "logps/rejected": -251.0, "loss": 0.6266, "rewards/accuracies": 0.6000000238418579, "rewards/chosen": -0.2060546875, "rewards/margins": 0.2099609375, "rewards/rejected": -0.41796875, "step": 80 }, { "epoch": 0.28753993610223644, "grad_norm": 32.267981389042255, "learning_rate": 4.787234042553192e-07, "logits/chosen": -1.7109375, "logits/rejected": -1.765625, "logps/chosen": -245.0, "logps/rejected": -247.0, "loss": 0.6458, "rewards/accuracies": 0.574999988079071, "rewards/chosen": -0.349609375, "rewards/margins": 0.2060546875, "rewards/rejected": -0.5546875, "step": 90 }, { "epoch": 0.3194888178913738, "grad_norm": 31.916118030442767, "learning_rate": 4.964497041420119e-07, "logits/chosen": -1.6953125, "logits/rejected": -1.7109375, "logps/chosen": -246.0, "logps/rejected": -243.0, "loss": 0.6301, "rewards/accuracies": 0.5375000238418579, "rewards/chosen": -0.29296875, "rewards/margins": 0.1767578125, "rewards/rejected": -0.470703125, "step": 100 }, { "epoch": 0.3514376996805112, "grad_norm": 30.563502769937987, "learning_rate": 4.905325443786982e-07, "logits/chosen": -1.6796875, "logits/rejected": -1.734375, "logps/chosen": -247.0, "logps/rejected": -250.0, "loss": 0.6481, "rewards/accuracies": 0.6625000238418579, "rewards/chosen": -0.23046875, "rewards/margins": 0.2421875, "rewards/rejected": -0.47265625, "step": 110 }, { "epoch": 0.38338658146964855, "grad_norm": 26.847013088556583, "learning_rate": 4.846153846153846e-07, "logits/chosen": -1.7109375, "logits/rejected": -1.7109375, "logps/chosen": -243.0, "logps/rejected": -251.0, "loss": 0.6152, "rewards/accuracies": 0.625, "rewards/chosen": -0.2734375, "rewards/margins": 0.306640625, "rewards/rejected": -0.578125, "step": 120 }, { "epoch": 0.41533546325878595, "grad_norm": 31.074353820130803, "learning_rate": 4.78698224852071e-07, "logits/chosen": -1.703125, "logits/rejected": -1.71875, "logps/chosen": -240.0, "logps/rejected": -260.0, "loss": 0.604, "rewards/accuracies": 0.5375000238418579, "rewards/chosen": -0.47265625, "rewards/margins": 0.291015625, "rewards/rejected": -0.76171875, "step": 130 }, { "epoch": 0.4472843450479233, "grad_norm": 32.192668717979046, "learning_rate": 4.727810650887574e-07, "logits/chosen": -1.6796875, "logits/rejected": -1.6796875, "logps/chosen": -243.0, "logps/rejected": -254.0, "loss": 0.5609, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -0.416015625, "rewards/margins": 0.5234375, "rewards/rejected": -0.9375, "step": 140 }, { "epoch": 0.4792332268370607, "grad_norm": 32.155266134791546, "learning_rate": 4.668639053254438e-07, "logits/chosen": -1.6953125, "logits/rejected": -1.671875, "logps/chosen": -246.0, "logps/rejected": -251.0, "loss": 0.5987, "rewards/accuracies": 0.637499988079071, "rewards/chosen": -0.45703125, "rewards/margins": 0.453125, "rewards/rejected": -0.91015625, "step": 150 }, { "epoch": 0.5111821086261981, "grad_norm": 30.837637394035802, "learning_rate": 4.6094674556213014e-07, "logits/chosen": -1.6875, "logits/rejected": -1.703125, "logps/chosen": -248.0, "logps/rejected": -252.0, "loss": 0.6033, "rewards/accuracies": 0.637499988079071, "rewards/chosen": -0.333984375, "rewards/margins": 0.373046875, "rewards/rejected": -0.70703125, "step": 160 }, { "epoch": 0.5431309904153354, "grad_norm": 25.59680166981086, "learning_rate": 4.5502958579881655e-07, "logits/chosen": -1.671875, "logits/rejected": -1.6796875, "logps/chosen": -238.0, "logps/rejected": -248.0, "loss": 0.5817, "rewards/accuracies": 0.6000000238418579, "rewards/chosen": -0.322265625, "rewards/margins": 0.30078125, "rewards/rejected": -0.62109375, "step": 170 }, { "epoch": 0.5750798722044729, "grad_norm": 29.648466125367857, "learning_rate": 4.491124260355029e-07, "logits/chosen": -1.6875, "logits/rejected": -1.6875, "logps/chosen": -245.0, "logps/rejected": -251.0, "loss": 0.5874, "rewards/accuracies": 0.6875, "rewards/chosen": -0.38671875, "rewards/margins": 0.50390625, "rewards/rejected": -0.890625, "step": 180 }, { "epoch": 0.6070287539936102, "grad_norm": 28.442764104750204, "learning_rate": 4.4319526627218936e-07, "logits/chosen": -1.6875, "logits/rejected": -1.6875, "logps/chosen": -246.0, "logps/rejected": -249.0, "loss": 0.5737, "rewards/accuracies": 0.699999988079071, "rewards/chosen": -0.294921875, "rewards/margins": 0.5078125, "rewards/rejected": -0.80078125, "step": 190 }, { "epoch": 0.6389776357827476, "grad_norm": 26.370679072337964, "learning_rate": 4.372781065088757e-07, "logits/chosen": -1.71875, "logits/rejected": -1.703125, "logps/chosen": -240.0, "logps/rejected": -256.0, "loss": 0.5867, "rewards/accuracies": 0.5874999761581421, "rewards/chosen": -0.3828125, "rewards/margins": 0.3671875, "rewards/rejected": -0.75, "step": 200 }, { "epoch": 0.670926517571885, "grad_norm": 24.556255560376357, "learning_rate": 4.313609467455621e-07, "logits/chosen": -1.703125, "logits/rejected": -1.6953125, "logps/chosen": -246.0, "logps/rejected": -252.0, "loss": 0.5853, "rewards/accuracies": 0.6625000238418579, "rewards/chosen": -0.349609375, "rewards/margins": 0.3671875, "rewards/rejected": -0.71875, "step": 210 }, { "epoch": 0.7028753993610224, "grad_norm": 29.98817140815093, "learning_rate": 4.2544378698224847e-07, "logits/chosen": -1.6796875, "logits/rejected": -1.71875, "logps/chosen": -243.0, "logps/rejected": -256.0, "loss": 0.5872, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -0.39453125, "rewards/margins": 0.55078125, "rewards/rejected": -0.9453125, "step": 220 }, { "epoch": 0.7348242811501597, "grad_norm": 27.478577985922634, "learning_rate": 4.195266272189349e-07, "logits/chosen": -1.671875, "logits/rejected": -1.65625, "logps/chosen": -240.0, "logps/rejected": -249.0, "loss": 0.5549, "rewards/accuracies": 0.6875, "rewards/chosen": -0.3203125, "rewards/margins": 0.50390625, "rewards/rejected": -0.82421875, "step": 230 }, { "epoch": 0.7667731629392971, "grad_norm": 26.416091326825903, "learning_rate": 4.1360946745562133e-07, "logits/chosen": -1.703125, "logits/rejected": -1.71875, "logps/chosen": -240.0, "logps/rejected": -262.0, "loss": 0.538, "rewards/accuracies": 0.6875, "rewards/chosen": -0.32421875, "rewards/margins": 0.55078125, "rewards/rejected": -0.875, "step": 240 }, { "epoch": 0.7987220447284346, "grad_norm": 28.30204123280793, "learning_rate": 4.076923076923077e-07, "logits/chosen": -1.6796875, "logits/rejected": -1.6875, "logps/chosen": -239.0, "logps/rejected": -260.0, "loss": 0.5507, "rewards/accuracies": 0.699999988079071, "rewards/chosen": -0.38671875, "rewards/margins": 0.5546875, "rewards/rejected": -0.9453125, "step": 250 }, { "epoch": 0.8306709265175719, "grad_norm": 33.17216428250153, "learning_rate": 4.017751479289941e-07, "logits/chosen": -1.6953125, "logits/rejected": -1.7421875, "logps/chosen": -244.0, "logps/rejected": -256.0, "loss": 0.5584, "rewards/accuracies": 0.762499988079071, "rewards/chosen": -0.27734375, "rewards/margins": 0.78125, "rewards/rejected": -1.0625, "step": 260 }, { "epoch": 0.8626198083067093, "grad_norm": 27.41764842123467, "learning_rate": 3.9585798816568044e-07, "logits/chosen": -1.6640625, "logits/rejected": -1.65625, "logps/chosen": -238.0, "logps/rejected": -252.0, "loss": 0.5742, "rewards/accuracies": 0.762499988079071, "rewards/chosen": -0.24609375, "rewards/margins": 0.6015625, "rewards/rejected": -0.84765625, "step": 270 }, { "epoch": 0.8945686900958466, "grad_norm": 34.061232014315166, "learning_rate": 3.8994082840236685e-07, "logits/chosen": -1.6953125, "logits/rejected": -1.703125, "logps/chosen": -242.0, "logps/rejected": -251.0, "loss": 0.5279, "rewards/accuracies": 0.7875000238418579, "rewards/chosen": -0.3203125, "rewards/margins": 0.65234375, "rewards/rejected": -0.97265625, "step": 280 }, { "epoch": 0.9265175718849841, "grad_norm": 28.160658079870018, "learning_rate": 3.840236686390532e-07, "logits/chosen": -1.6796875, "logits/rejected": -1.6953125, "logps/chosen": -248.0, "logps/rejected": -252.0, "loss": 0.5804, "rewards/accuracies": 0.699999988079071, "rewards/chosen": -0.4296875, "rewards/margins": 0.45703125, "rewards/rejected": -0.88671875, "step": 290 }, { "epoch": 0.9584664536741214, "grad_norm": 29.60484241313221, "learning_rate": 3.7810650887573966e-07, "logits/chosen": -1.65625, "logits/rejected": -1.671875, "logps/chosen": -247.0, "logps/rejected": -255.0, "loss": 0.5635, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -0.35546875, "rewards/margins": 0.400390625, "rewards/rejected": -0.75390625, "step": 300 }, { "epoch": 0.9904153354632588, "grad_norm": 33.368421615399974, "learning_rate": 3.72189349112426e-07, "logits/chosen": -1.7109375, "logits/rejected": -1.703125, "logps/chosen": -244.0, "logps/rejected": -250.0, "loss": 0.5843, "rewards/accuracies": 0.762499988079071, "rewards/chosen": -0.2578125, "rewards/margins": 0.49609375, "rewards/rejected": -0.75390625, "step": 310 }, { "epoch": 1.0, "eval_logits/chosen": -1.6953125, "eval_logits/rejected": -1.7109375, "eval_logps/chosen": -246.0, "eval_logps/rejected": -248.0, "eval_loss": 0.6162499785423279, "eval_rewards/accuracies": 0.5714285969734192, "eval_rewards/chosen": -0.380859375, "eval_rewards/margins": 0.296875, "eval_rewards/rejected": -0.6796875, "eval_runtime": 12.6011, "eval_samples_per_second": 15.872, "eval_steps_per_second": 0.556, "step": 313 }, { "epoch": 1.0223642172523961, "grad_norm": 27.543304377009523, "learning_rate": 3.662721893491124e-07, "logits/chosen": -1.65625, "logits/rejected": -1.703125, "logps/chosen": -240.0, "logps/rejected": -258.0, "loss": 0.4082, "rewards/accuracies": 0.875, "rewards/chosen": -0.03759765625, "rewards/margins": 1.0546875, "rewards/rejected": -1.09375, "step": 320 }, { "epoch": 1.0543130990415335, "grad_norm": 27.327705966461778, "learning_rate": 3.603550295857988e-07, "logits/chosen": -1.6484375, "logits/rejected": -1.65625, "logps/chosen": -242.0, "logps/rejected": -262.0, "loss": 0.3683, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": -0.345703125, "rewards/margins": 1.125, "rewards/rejected": -1.46875, "step": 330 }, { "epoch": 1.0862619808306708, "grad_norm": 17.914371176661845, "learning_rate": 3.544378698224852e-07, "logits/chosen": -1.6484375, "logits/rejected": -1.703125, "logps/chosen": -245.0, "logps/rejected": -256.0, "loss": 0.3574, "rewards/accuracies": 0.887499988079071, "rewards/chosen": -0.421875, "rewards/margins": 1.1484375, "rewards/rejected": -1.5703125, "step": 340 }, { "epoch": 1.1182108626198084, "grad_norm": 27.18764754676586, "learning_rate": 3.485207100591716e-07, "logits/chosen": -1.703125, "logits/rejected": -1.7109375, "logps/chosen": -246.0, "logps/rejected": -262.0, "loss": 0.3483, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": -0.3046875, "rewards/margins": 1.390625, "rewards/rejected": -1.6953125, "step": 350 }, { "epoch": 1.1501597444089458, "grad_norm": 16.652754238092474, "learning_rate": 3.42603550295858e-07, "logits/chosen": -1.65625, "logits/rejected": -1.671875, "logps/chosen": -245.0, "logps/rejected": -260.0, "loss": 0.3208, "rewards/accuracies": 0.9125000238418579, "rewards/chosen": -0.18359375, "rewards/margins": 1.4453125, "rewards/rejected": -1.625, "step": 360 }, { "epoch": 1.182108626198083, "grad_norm": 21.849555306547938, "learning_rate": 3.366863905325444e-07, "logits/chosen": -1.6875, "logits/rejected": -1.6875, "logps/chosen": -237.0, "logps/rejected": -248.0, "loss": 0.3809, "rewards/accuracies": 0.7875000238418579, "rewards/chosen": -0.228515625, "rewards/margins": 0.9296875, "rewards/rejected": -1.15625, "step": 370 }, { "epoch": 1.2140575079872205, "grad_norm": 22.779743163941504, "learning_rate": 3.3076923076923075e-07, "logits/chosen": -1.6796875, "logits/rejected": -1.71875, "logps/chosen": -246.0, "logps/rejected": -258.0, "loss": 0.3479, "rewards/accuracies": 0.8374999761581421, "rewards/chosen": -0.177734375, "rewards/margins": 1.1171875, "rewards/rejected": -1.296875, "step": 380 }, { "epoch": 1.2460063897763578, "grad_norm": 17.054365580798475, "learning_rate": 3.2485207100591716e-07, "logits/chosen": -1.65625, "logits/rejected": -1.7109375, "logps/chosen": -237.0, "logps/rejected": -252.0, "loss": 0.3299, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": 0.0015106201171875, "rewards/margins": 1.28125, "rewards/rejected": -1.28125, "step": 390 }, { "epoch": 1.2779552715654952, "grad_norm": 20.840224552737123, "learning_rate": 3.189349112426035e-07, "logits/chosen": -1.6953125, "logits/rejected": -1.59375, "logps/chosen": -241.0, "logps/rejected": -264.0, "loss": 0.3656, "rewards/accuracies": 0.875, "rewards/chosen": -0.2265625, "rewards/margins": 1.15625, "rewards/rejected": -1.3828125, "step": 400 }, { "epoch": 1.3099041533546325, "grad_norm": 24.15916736976353, "learning_rate": 3.130177514792899e-07, "logits/chosen": -1.6328125, "logits/rejected": -1.6796875, "logps/chosen": -243.0, "logps/rejected": -258.0, "loss": 0.3516, "rewards/accuracies": 0.824999988079071, "rewards/chosen": -0.00360107421875, "rewards/margins": 1.1328125, "rewards/rejected": -1.140625, "step": 410 }, { "epoch": 1.34185303514377, "grad_norm": 23.743509110621172, "learning_rate": 3.071005917159763e-07, "logits/chosen": -1.671875, "logits/rejected": -1.65625, "logps/chosen": -245.0, "logps/rejected": -255.0, "loss": 0.3688, "rewards/accuracies": 0.862500011920929, "rewards/chosen": -0.166015625, "rewards/margins": 1.0234375, "rewards/rejected": -1.1953125, "step": 420 }, { "epoch": 1.3738019169329074, "grad_norm": 21.0489995042185, "learning_rate": 3.011834319526627e-07, "logits/chosen": -1.671875, "logits/rejected": -1.6953125, "logps/chosen": -243.0, "logps/rejected": -256.0, "loss": 0.3246, "rewards/accuracies": 0.875, "rewards/chosen": -0.126953125, "rewards/margins": 1.2265625, "rewards/rejected": -1.359375, "step": 430 }, { "epoch": 1.4057507987220448, "grad_norm": 23.98752497138385, "learning_rate": 2.952662721893491e-07, "logits/chosen": -1.6796875, "logits/rejected": -1.625, "logps/chosen": -242.0, "logps/rejected": -264.0, "loss": 0.3287, "rewards/accuracies": 0.9375, "rewards/chosen": -0.26953125, "rewards/margins": 1.390625, "rewards/rejected": -1.6640625, "step": 440 }, { "epoch": 1.4376996805111821, "grad_norm": 22.151863656806025, "learning_rate": 2.893491124260355e-07, "logits/chosen": -1.640625, "logits/rejected": -1.6953125, "logps/chosen": -242.0, "logps/rejected": -256.0, "loss": 0.3172, "rewards/accuracies": 0.887499988079071, "rewards/chosen": -0.2470703125, "rewards/margins": 1.359375, "rewards/rejected": -1.6015625, "step": 450 }, { "epoch": 1.4696485623003195, "grad_norm": 25.183351240662166, "learning_rate": 2.834319526627219e-07, "logits/chosen": -1.671875, "logits/rejected": -1.6953125, "logps/chosen": -240.0, "logps/rejected": -260.0, "loss": 0.3275, "rewards/accuracies": 0.925000011920929, "rewards/chosen": -0.240234375, "rewards/margins": 1.4609375, "rewards/rejected": -1.6953125, "step": 460 }, { "epoch": 1.5015974440894568, "grad_norm": 23.9596209798775, "learning_rate": 2.7751479289940824e-07, "logits/chosen": -1.6953125, "logits/rejected": -1.671875, "logps/chosen": -240.0, "logps/rejected": -258.0, "loss": 0.2869, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.232421875, "rewards/margins": 1.46875, "rewards/rejected": -1.6953125, "step": 470 }, { "epoch": 1.5335463258785942, "grad_norm": 20.884264225876592, "learning_rate": 2.715976331360947e-07, "logits/chosen": -1.6875, "logits/rejected": -1.65625, "logps/chosen": -238.0, "logps/rejected": -262.0, "loss": 0.3432, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": -0.1142578125, "rewards/margins": 1.5546875, "rewards/rejected": -1.671875, "step": 480 }, { "epoch": 1.5654952076677318, "grad_norm": 27.424882575382, "learning_rate": 2.6568047337278106e-07, "logits/chosen": -1.6015625, "logits/rejected": -1.65625, "logps/chosen": -250.0, "logps/rejected": -264.0, "loss": 0.3153, "rewards/accuracies": 0.887499988079071, "rewards/chosen": -0.5625, "rewards/margins": 1.4296875, "rewards/rejected": -1.984375, "step": 490 }, { "epoch": 1.5974440894568689, "grad_norm": 25.335220219602046, "learning_rate": 2.5976331360946746e-07, "logits/chosen": -1.6875, "logits/rejected": -1.65625, "logps/chosen": -244.0, "logps/rejected": -260.0, "loss": 0.3238, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": -0.279296875, "rewards/margins": 1.5859375, "rewards/rejected": -1.8671875, "step": 500 }, { "epoch": 1.6293929712460065, "grad_norm": 26.281054947397205, "learning_rate": 2.538461538461538e-07, "logits/chosen": -1.640625, "logits/rejected": -1.6875, "logps/chosen": -243.0, "logps/rejected": -255.0, "loss": 0.3293, "rewards/accuracies": 0.8374999761581421, "rewards/chosen": -0.1962890625, "rewards/margins": 1.40625, "rewards/rejected": -1.6015625, "step": 510 }, { "epoch": 1.6613418530351438, "grad_norm": 22.709993123443564, "learning_rate": 2.479289940828402e-07, "logits/chosen": -1.6796875, "logits/rejected": -1.671875, "logps/chosen": -246.0, "logps/rejected": -262.0, "loss": 0.306, "rewards/accuracies": 0.875, "rewards/chosen": -0.1865234375, "rewards/margins": 1.4609375, "rewards/rejected": -1.6484375, "step": 520 }, { "epoch": 1.6932907348242812, "grad_norm": 24.667533534060208, "learning_rate": 2.420118343195266e-07, "logits/chosen": -1.6328125, "logits/rejected": -1.6875, "logps/chosen": -239.0, "logps/rejected": -260.0, "loss": 0.3221, "rewards/accuracies": 0.862500011920929, "rewards/chosen": -0.0152587890625, "rewards/margins": 1.421875, "rewards/rejected": -1.4375, "step": 530 }, { "epoch": 1.7252396166134185, "grad_norm": 16.738356385012416, "learning_rate": 2.36094674556213e-07, "logits/chosen": -1.6328125, "logits/rejected": -1.6875, "logps/chosen": -240.0, "logps/rejected": -260.0, "loss": 0.3271, "rewards/accuracies": 0.8374999761581421, "rewards/chosen": -0.125, "rewards/margins": 1.609375, "rewards/rejected": -1.734375, "step": 540 }, { "epoch": 1.7571884984025559, "grad_norm": 19.302611241291174, "learning_rate": 2.301775147928994e-07, "logits/chosen": -1.6484375, "logits/rejected": -1.65625, "logps/chosen": -248.0, "logps/rejected": -268.0, "loss": 0.318, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": -0.1337890625, "rewards/margins": 1.546875, "rewards/rejected": -1.6796875, "step": 550 }, { "epoch": 1.7891373801916934, "grad_norm": 17.493675951115446, "learning_rate": 2.242603550295858e-07, "logits/chosen": -1.65625, "logits/rejected": -1.65625, "logps/chosen": -240.0, "logps/rejected": -266.0, "loss": 0.2807, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": -0.1884765625, "rewards/margins": 1.546875, "rewards/rejected": -1.7421875, "step": 560 }, { "epoch": 1.8210862619808306, "grad_norm": 22.291882579304676, "learning_rate": 2.1834319526627217e-07, "logits/chosen": -1.65625, "logits/rejected": -1.671875, "logps/chosen": -245.0, "logps/rejected": -262.0, "loss": 0.2885, "rewards/accuracies": 0.9125000238418579, "rewards/chosen": -0.1240234375, "rewards/margins": 1.7265625, "rewards/rejected": -1.8515625, "step": 570 }, { "epoch": 1.8530351437699681, "grad_norm": 26.161100115274262, "learning_rate": 2.1242603550295858e-07, "logits/chosen": -1.640625, "logits/rejected": -1.6328125, "logps/chosen": -246.0, "logps/rejected": -268.0, "loss": 0.3145, "rewards/accuracies": 0.887499988079071, "rewards/chosen": -0.287109375, "rewards/margins": 1.71875, "rewards/rejected": -2.0, "step": 580 }, { "epoch": 1.8849840255591053, "grad_norm": 28.81776072776743, "learning_rate": 2.0650887573964496e-07, "logits/chosen": -1.6484375, "logits/rejected": -1.6484375, "logps/chosen": -240.0, "logps/rejected": -260.0, "loss": 0.3249, "rewards/accuracies": 0.862500011920929, "rewards/chosen": -0.0234375, "rewards/margins": 1.5078125, "rewards/rejected": -1.53125, "step": 590 }, { "epoch": 1.9169329073482428, "grad_norm": 19.977671114909082, "learning_rate": 2.0059171597633133e-07, "logits/chosen": -1.6171875, "logits/rejected": -1.6640625, "logps/chosen": -239.0, "logps/rejected": -254.0, "loss": 0.3448, "rewards/accuracies": 0.925000011920929, "rewards/chosen": -0.115234375, "rewards/margins": 1.421875, "rewards/rejected": -1.5390625, "step": 600 }, { "epoch": 1.9488817891373802, "grad_norm": 17.659228575887397, "learning_rate": 1.9467455621301774e-07, "logits/chosen": -1.671875, "logits/rejected": -1.640625, "logps/chosen": -247.0, "logps/rejected": -266.0, "loss": 0.3007, "rewards/accuracies": 0.9375, "rewards/chosen": -0.162109375, "rewards/margins": 1.53125, "rewards/rejected": -1.6953125, "step": 610 }, { "epoch": 1.9808306709265175, "grad_norm": 18.405238339078597, "learning_rate": 1.8875739644970412e-07, "logits/chosen": -1.65625, "logits/rejected": -1.6484375, "logps/chosen": -242.0, "logps/rejected": -260.0, "loss": 0.3216, "rewards/accuracies": 0.887499988079071, "rewards/chosen": -0.12255859375, "rewards/margins": 1.46875, "rewards/rejected": -1.5859375, "step": 620 }, { "epoch": 2.0, "eval_logits/chosen": -1.65625, "eval_logits/rejected": -1.671875, "eval_logps/chosen": -250.0, "eval_logps/rejected": -253.0, "eval_loss": 0.600781261920929, "eval_rewards/accuracies": 0.6428571343421936, "eval_rewards/chosen": -0.734375, "eval_rewards/margins": 0.44140625, "eval_rewards/rejected": -1.171875, "eval_runtime": 12.6155, "eval_samples_per_second": 15.854, "eval_steps_per_second": 0.555, "step": 626 }, { "epoch": 2.012779552715655, "grad_norm": 15.229138030548278, "learning_rate": 1.8284023668639053e-07, "logits/chosen": -1.609375, "logits/rejected": -1.6171875, "logps/chosen": -240.0, "logps/rejected": -268.0, "loss": 0.2259, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.201171875, "rewards/margins": 1.8046875, "rewards/rejected": -2.0, "step": 630 }, { "epoch": 2.0447284345047922, "grad_norm": 17.70681792433403, "learning_rate": 1.7692307692307693e-07, "logits/chosen": -1.6171875, "logits/rejected": -1.609375, "logps/chosen": -237.0, "logps/rejected": -264.0, "loss": 0.2073, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": 0.208984375, "rewards/margins": 2.203125, "rewards/rejected": -2.0, "step": 640 }, { "epoch": 2.07667731629393, "grad_norm": 13.769490197345323, "learning_rate": 1.710059171597633e-07, "logits/chosen": -1.6484375, "logits/rejected": -1.65625, "logps/chosen": -242.0, "logps/rejected": -270.0, "loss": 0.2069, "rewards/accuracies": 0.9125000238418579, "rewards/chosen": -0.0859375, "rewards/margins": 2.015625, "rewards/rejected": -2.109375, "step": 650 }, { "epoch": 2.108626198083067, "grad_norm": 10.553376765871462, "learning_rate": 1.650887573964497e-07, "logits/chosen": -1.640625, "logits/rejected": -1.65625, "logps/chosen": -238.0, "logps/rejected": -266.0, "loss": 0.188, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.06640625, "rewards/margins": 2.21875, "rewards/rejected": -2.28125, "step": 660 }, { "epoch": 2.1405750798722045, "grad_norm": 12.717503262116809, "learning_rate": 1.591715976331361e-07, "logits/chosen": -1.6640625, "logits/rejected": -1.625, "logps/chosen": -240.0, "logps/rejected": -258.0, "loss": 0.2087, "rewards/accuracies": 0.9375, "rewards/chosen": -0.05419921875, "rewards/margins": 1.828125, "rewards/rejected": -1.8828125, "step": 670 }, { "epoch": 2.1725239616613417, "grad_norm": 17.368724259557464, "learning_rate": 1.5325443786982248e-07, "logits/chosen": -1.6484375, "logits/rejected": -1.6875, "logps/chosen": -244.0, "logps/rejected": -264.0, "loss": 0.196, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.09423828125, "rewards/margins": 2.28125, "rewards/rejected": -2.375, "step": 680 }, { "epoch": 2.2044728434504792, "grad_norm": 15.997180398783524, "learning_rate": 1.4733727810650885e-07, "logits/chosen": -1.625, "logits/rejected": -1.640625, "logps/chosen": -239.0, "logps/rejected": -262.0, "loss": 0.1955, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.07763671875, "rewards/margins": 1.890625, "rewards/rejected": -1.96875, "step": 690 }, { "epoch": 2.236421725239617, "grad_norm": 16.465231501289626, "learning_rate": 1.4142011834319526e-07, "logits/chosen": -1.6171875, "logits/rejected": -1.6796875, "logps/chosen": -247.0, "logps/rejected": -264.0, "loss": 0.1818, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": 0.0087890625, "rewards/margins": 2.03125, "rewards/rejected": -2.03125, "step": 700 }, { "epoch": 2.268370607028754, "grad_norm": 15.303382006736875, "learning_rate": 1.3550295857988164e-07, "logits/chosen": -1.640625, "logits/rejected": -1.671875, "logps/chosen": -240.0, "logps/rejected": -266.0, "loss": 0.196, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": -0.02001953125, "rewards/margins": 1.9375, "rewards/rejected": -1.9609375, "step": 710 }, { "epoch": 2.3003194888178915, "grad_norm": 27.787991315567947, "learning_rate": 1.2958579881656802e-07, "logits/chosen": -1.6328125, "logits/rejected": -1.609375, "logps/chosen": -247.0, "logps/rejected": -268.0, "loss": 0.2045, "rewards/accuracies": 0.925000011920929, "rewards/chosen": -0.34765625, "rewards/margins": 1.9375, "rewards/rejected": -2.28125, "step": 720 }, { "epoch": 2.3322683706070286, "grad_norm": 20.488749808987933, "learning_rate": 1.2366863905325443e-07, "logits/chosen": -1.609375, "logits/rejected": -1.6484375, "logps/chosen": -236.0, "logps/rejected": -262.0, "loss": 0.1955, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.076171875, "rewards/margins": 2.15625, "rewards/rejected": -2.078125, "step": 730 }, { "epoch": 2.364217252396166, "grad_norm": 16.34392558119839, "learning_rate": 1.1775147928994082e-07, "logits/chosen": -1.6328125, "logits/rejected": -1.6328125, "logps/chosen": -235.0, "logps/rejected": -260.0, "loss": 0.2017, "rewards/accuracies": 0.949999988079071, "rewards/chosen": 0.171875, "rewards/margins": 2.171875, "rewards/rejected": -2.0, "step": 740 }, { "epoch": 2.3961661341853033, "grad_norm": 22.986892420642484, "learning_rate": 1.1183431952662721e-07, "logits/chosen": -1.59375, "logits/rejected": -1.640625, "logps/chosen": -238.0, "logps/rejected": -262.0, "loss": 0.1846, "rewards/accuracies": 0.9125000238418579, "rewards/chosen": 0.033935546875, "rewards/margins": 2.03125, "rewards/rejected": -1.9921875, "step": 750 }, { "epoch": 2.428115015974441, "grad_norm": 22.954617926133928, "learning_rate": 1.059171597633136e-07, "logits/chosen": -1.6171875, "logits/rejected": -1.640625, "logps/chosen": -238.0, "logps/rejected": -264.0, "loss": 0.1675, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": 0.1943359375, "rewards/margins": 2.3125, "rewards/rejected": -2.125, "step": 760 }, { "epoch": 2.460063897763578, "grad_norm": 10.72220228289897, "learning_rate": 1e-07, "logits/chosen": -1.59375, "logits/rejected": -1.6015625, "logps/chosen": -239.0, "logps/rejected": -268.0, "loss": 0.1909, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": -0.083984375, "rewards/margins": 2.171875, "rewards/rejected": -2.25, "step": 770 }, { "epoch": 2.4920127795527156, "grad_norm": 22.698230184573, "learning_rate": 9.408284023668639e-08, "logits/chosen": -1.6328125, "logits/rejected": -1.6328125, "logps/chosen": -247.0, "logps/rejected": -262.0, "loss": 0.1904, "rewards/accuracies": 0.9125000238418579, "rewards/chosen": 0.044921875, "rewards/margins": 2.03125, "rewards/rejected": -1.9921875, "step": 780 }, { "epoch": 2.523961661341853, "grad_norm": 17.131418963453264, "learning_rate": 8.816568047337278e-08, "logits/chosen": -1.4921875, "logits/rejected": -1.65625, "logps/chosen": -249.0, "logps/rejected": -264.0, "loss": 0.2065, "rewards/accuracies": 0.9125000238418579, "rewards/chosen": -0.2119140625, "rewards/margins": 2.09375, "rewards/rejected": -2.296875, "step": 790 }, { "epoch": 2.5559105431309903, "grad_norm": 14.204160115025998, "learning_rate": 8.224852071005916e-08, "logits/chosen": -1.640625, "logits/rejected": -1.6328125, "logps/chosen": -247.0, "logps/rejected": -264.0, "loss": 0.1824, "rewards/accuracies": 0.9375, "rewards/chosen": -0.130859375, "rewards/margins": 1.9296875, "rewards/rejected": -2.0625, "step": 800 }, { "epoch": 2.587859424920128, "grad_norm": 16.438831527877248, "learning_rate": 7.633136094674555e-08, "logits/chosen": -1.6328125, "logits/rejected": -1.6640625, "logps/chosen": -241.0, "logps/rejected": -266.0, "loss": 0.1841, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.08154296875, "rewards/margins": 2.125, "rewards/rejected": -2.203125, "step": 810 }, { "epoch": 2.619808306709265, "grad_norm": 23.382793800196303, "learning_rate": 7.041420118343195e-08, "logits/chosen": -1.609375, "logits/rejected": -1.609375, "logps/chosen": -240.0, "logps/rejected": -266.0, "loss": 0.1805, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": -0.03271484375, "rewards/margins": 2.265625, "rewards/rejected": -2.296875, "step": 820 }, { "epoch": 2.6517571884984026, "grad_norm": 17.97702724195434, "learning_rate": 6.449704142011835e-08, "logits/chosen": -1.6640625, "logits/rejected": -1.640625, "logps/chosen": -239.0, "logps/rejected": -258.0, "loss": 0.1621, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": 0.004974365234375, "rewards/margins": 2.21875, "rewards/rejected": -2.21875, "step": 830 }, { "epoch": 2.68370607028754, "grad_norm": 15.72139882933752, "learning_rate": 5.857988165680473e-08, "logits/chosen": -1.640625, "logits/rejected": -1.6484375, "logps/chosen": -244.0, "logps/rejected": -268.0, "loss": 0.1928, "rewards/accuracies": 0.9375, "rewards/chosen": -0.142578125, "rewards/margins": 2.296875, "rewards/rejected": -2.4375, "step": 840 }, { "epoch": 2.7156549520766773, "grad_norm": 20.685915337750906, "learning_rate": 5.266272189349112e-08, "logits/chosen": -1.6171875, "logits/rejected": -1.6484375, "logps/chosen": -237.0, "logps/rejected": -268.0, "loss": 0.1831, "rewards/accuracies": 0.925000011920929, "rewards/chosen": 0.1875, "rewards/margins": 2.390625, "rewards/rejected": -2.203125, "step": 850 }, { "epoch": 2.747603833865815, "grad_norm": 20.038756432535152, "learning_rate": 4.674556213017751e-08, "logits/chosen": -1.609375, "logits/rejected": -1.6015625, "logps/chosen": -246.0, "logps/rejected": -268.0, "loss": 0.1985, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": -0.298828125, "rewards/margins": 2.03125, "rewards/rejected": -2.328125, "step": 860 }, { "epoch": 2.779552715654952, "grad_norm": 21.447728631568275, "learning_rate": 4.082840236686391e-08, "logits/chosen": -1.625, "logits/rejected": -1.6015625, "logps/chosen": -246.0, "logps/rejected": -266.0, "loss": 0.2021, "rewards/accuracies": 0.9125000238418579, "rewards/chosen": -0.1845703125, "rewards/margins": 1.9609375, "rewards/rejected": -2.140625, "step": 870 }, { "epoch": 2.8115015974440896, "grad_norm": 13.38540049073994, "learning_rate": 3.4911242603550294e-08, "logits/chosen": -1.640625, "logits/rejected": -1.65625, "logps/chosen": -240.0, "logps/rejected": -270.0, "loss": 0.18, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.00701904296875, "rewards/margins": 2.25, "rewards/rejected": -2.265625, "step": 880 }, { "epoch": 2.8434504792332267, "grad_norm": 17.91486249665347, "learning_rate": 2.8994082840236687e-08, "logits/chosen": -1.625, "logits/rejected": -1.6484375, "logps/chosen": -241.0, "logps/rejected": -264.0, "loss": 0.1964, "rewards/accuracies": 0.925000011920929, "rewards/chosen": 0.057373046875, "rewards/margins": 2.109375, "rewards/rejected": -2.0625, "step": 890 }, { "epoch": 2.8753993610223643, "grad_norm": 22.402213957503538, "learning_rate": 2.3076923076923076e-08, "logits/chosen": -1.625, "logits/rejected": -1.625, "logps/chosen": -240.0, "logps/rejected": -274.0, "loss": 0.1627, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": 0.1201171875, "rewards/margins": 2.578125, "rewards/rejected": -2.46875, "step": 900 }, { "epoch": 2.9073482428115014, "grad_norm": 16.63729631800027, "learning_rate": 1.7159763313609465e-08, "logits/chosen": -1.59375, "logits/rejected": -1.625, "logps/chosen": -238.0, "logps/rejected": -270.0, "loss": 0.2028, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.09765625, "rewards/margins": 2.265625, "rewards/rejected": -2.171875, "step": 910 }, { "epoch": 2.939297124600639, "grad_norm": 13.806768716470549, "learning_rate": 1.1242603550295858e-08, "logits/chosen": -1.625, "logits/rejected": -1.6171875, "logps/chosen": -244.0, "logps/rejected": -268.0, "loss": 0.1701, "rewards/accuracies": 0.9125000238418579, "rewards/chosen": -0.208984375, "rewards/margins": 2.015625, "rewards/rejected": -2.21875, "step": 920 }, { "epoch": 2.9712460063897765, "grad_norm": 12.821411876121983, "learning_rate": 5.325443786982248e-09, "logits/chosen": -1.6015625, "logits/rejected": -1.6171875, "logps/chosen": -243.0, "logps/rejected": -264.0, "loss": 0.1777, "rewards/accuracies": 0.9375, "rewards/chosen": -0.1640625, "rewards/margins": 2.171875, "rewards/rejected": -2.34375, "step": 930 }, { "epoch": 3.0, "eval_logits/chosen": -1.640625, "eval_logits/rejected": -1.6484375, "eval_logps/chosen": -251.0, "eval_logps/rejected": -255.0, "eval_loss": 0.6092968583106995, "eval_rewards/accuracies": 0.6607142686843872, "eval_rewards/chosen": -0.83984375, "eval_rewards/margins": 0.56640625, "eval_rewards/rejected": -1.40625, "eval_runtime": 15.3355, "eval_samples_per_second": 13.042, "eval_steps_per_second": 0.456, "step": 939 }, { "epoch": 3.0, "step": 939, "total_flos": 0.0, "train_loss": 0.37425009308023827, "train_runtime": 4934.2684, "train_samples_per_second": 6.071, "train_steps_per_second": 0.19 } ], "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 }