{ "best_metric": null, "best_model_checkpoint": null, "epoch": 2.986111111111111, "eval_steps": 500, "global_step": 21500, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.06944444444444445, "grad_norm": 32.85024642944336, "learning_rate": 4.8842592592592595e-05, "logits/chosen": 3.556478261947632, "logits/rejected": 3.293552875518799, "logps/chosen": -527.9028930664062, "logps/rejected": -488.8791198730469, "loss": 0.7475, "rewards/accuracies": 0.6389999985694885, "rewards/chosen": -0.9193629026412964, "rewards/margins": 0.9515601992607117, "rewards/rejected": -1.8709231615066528, "step": 500 }, { "epoch": 0.1388888888888889, "grad_norm": 0.4692399501800537, "learning_rate": 4.768518518518519e-05, "logits/chosen": 3.758127450942993, "logits/rejected": 3.583296298980713, "logps/chosen": -556.1846923828125, "logps/rejected": -531.4818115234375, "loss": 0.7412, "rewards/accuracies": 0.7310000061988831, "rewards/chosen": -2.024296760559082, "rewards/margins": 2.4693143367767334, "rewards/rejected": -4.4936113357543945, "step": 1000 }, { "epoch": 0.20833333333333334, "grad_norm": 0.2115844041109085, "learning_rate": 4.652777777777778e-05, "logits/chosen": 3.4201669692993164, "logits/rejected": 3.196681261062622, "logps/chosen": -562.4039916992188, "logps/rejected": -539.3616943359375, "loss": 0.7542, "rewards/accuracies": 0.7089999914169312, "rewards/chosen": -3.269911289215088, "rewards/margins": 2.7116143703460693, "rewards/rejected": -5.981525897979736, "step": 1500 }, { "epoch": 0.2777777777777778, "grad_norm": 10.3626070022583, "learning_rate": 4.5370370370370374e-05, "logits/chosen": 3.5292723178863525, "logits/rejected": 3.344435453414917, "logps/chosen": -551.2387084960938, "logps/rejected": -541.3450927734375, "loss": 0.8561, "rewards/accuracies": 0.7049999833106995, "rewards/chosen": -3.477266311645508, "rewards/margins": 2.634814977645874, "rewards/rejected": -6.112081050872803, "step": 2000 }, { "epoch": 0.3472222222222222, "grad_norm": 31.721025466918945, "learning_rate": 4.4212962962962966e-05, "logits/chosen": 3.8410401344299316, "logits/rejected": 3.6350772380828857, "logps/chosen": -562.904296875, "logps/rejected": -537.2197265625, "loss": 0.8085, "rewards/accuracies": 0.7149999737739563, "rewards/chosen": -2.0515942573547363, "rewards/margins": 2.863039255142212, "rewards/rejected": -4.914632797241211, "step": 2500 }, { "epoch": 0.4166666666666667, "grad_norm": 27.109020233154297, "learning_rate": 4.305555555555556e-05, "logits/chosen": 3.5856244564056396, "logits/rejected": 3.3958754539489746, "logps/chosen": -562.5556030273438, "logps/rejected": -554.8358764648438, "loss": 0.8662, "rewards/accuracies": 0.7269999980926514, "rewards/chosen": -3.2988736629486084, "rewards/margins": 3.144113779067993, "rewards/rejected": -6.442987442016602, "step": 3000 }, { "epoch": 0.4861111111111111, "grad_norm": 20.984128952026367, "learning_rate": 4.1898148148148145e-05, "logits/chosen": 3.5219292640686035, "logits/rejected": 3.2919979095458984, "logps/chosen": -556.874267578125, "logps/rejected": -546.91552734375, "loss": 0.7505, "rewards/accuracies": 0.7360000014305115, "rewards/chosen": -3.3442256450653076, "rewards/margins": 3.2203891277313232, "rewards/rejected": -6.564614295959473, "step": 3500 }, { "epoch": 0.5555555555555556, "grad_norm": 0.33463263511657715, "learning_rate": 4.074074074074074e-05, "logits/chosen": 3.6497881412506104, "logits/rejected": 3.432859182357788, "logps/chosen": -567.9624633789062, "logps/rejected": -559.707275390625, "loss": 0.7974, "rewards/accuracies": 0.7229999899864197, "rewards/chosen": -2.9807636737823486, "rewards/margins": 3.382077693939209, "rewards/rejected": -6.3628411293029785, "step": 4000 }, { "epoch": 0.625, "grad_norm": 9.498452186584473, "learning_rate": 3.958333333333333e-05, "logits/chosen": 3.4272067546844482, "logits/rejected": 3.1543126106262207, "logps/chosen": -555.7978515625, "logps/rejected": -537.4580688476562, "loss": 0.7591, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -2.860079765319824, "rewards/margins": 3.3233377933502197, "rewards/rejected": -6.183417320251465, "step": 4500 }, { "epoch": 0.6944444444444444, "grad_norm": 8.980354309082031, "learning_rate": 3.8425925925925924e-05, "logits/chosen": 3.2811481952667236, "logits/rejected": 3.031383514404297, "logps/chosen": -546.9360961914062, "logps/rejected": -524.9453735351562, "loss": 0.7517, "rewards/accuracies": 0.7360000014305115, "rewards/chosen": -2.686689853668213, "rewards/margins": 3.413407564163208, "rewards/rejected": -6.100097179412842, "step": 5000 }, { "epoch": 0.7638888888888888, "grad_norm": 0.3623337149620056, "learning_rate": 3.726851851851852e-05, "logits/chosen": 3.3945746421813965, "logits/rejected": 3.1526079177856445, "logps/chosen": -557.0965576171875, "logps/rejected": -544.6822509765625, "loss": 0.8129, "rewards/accuracies": 0.722000002861023, "rewards/chosen": -2.9042410850524902, "rewards/margins": 3.3669819831848145, "rewards/rejected": -6.271223545074463, "step": 5500 }, { "epoch": 0.8333333333333334, "grad_norm": 11.856613159179688, "learning_rate": 3.611111111111111e-05, "logits/chosen": 3.3205320835113525, "logits/rejected": 3.075706720352173, "logps/chosen": -565.4915161132812, "logps/rejected": -553.1411743164062, "loss": 0.7355, "rewards/accuracies": 0.7110000252723694, "rewards/chosen": -3.6243250370025635, "rewards/margins": 3.4897994995117188, "rewards/rejected": -7.114124774932861, "step": 6000 }, { "epoch": 0.9027777777777778, "grad_norm": 13.970710754394531, "learning_rate": 3.49537037037037e-05, "logits/chosen": 3.340500831604004, "logits/rejected": 3.0907979011535645, "logps/chosen": -562.9676513671875, "logps/rejected": -546.9876708984375, "loss": 0.7527, "rewards/accuracies": 0.7289999723434448, "rewards/chosen": -4.343613147735596, "rewards/margins": 3.5196118354797363, "rewards/rejected": -7.86322546005249, "step": 6500 }, { "epoch": 0.9722222222222222, "grad_norm": 6.55237340927124, "learning_rate": 3.3796296296296295e-05, "logits/chosen": 3.494124174118042, "logits/rejected": 3.220623016357422, "logps/chosen": -560.8843383789062, "logps/rejected": -545.5639038085938, "loss": 0.7805, "rewards/accuracies": 0.7170000076293945, "rewards/chosen": -2.7601752281188965, "rewards/margins": 3.549665927886963, "rewards/rejected": -6.309841632843018, "step": 7000 }, { "epoch": 1.0416666666666667, "grad_norm": 3.183572769165039, "learning_rate": 3.263888888888889e-05, "logits/chosen": 3.3911421298980713, "logits/rejected": 3.123459815979004, "logps/chosen": -554.22607421875, "logps/rejected": -557.9960327148438, "loss": 0.3866, "rewards/accuracies": 0.8610000014305115, "rewards/chosen": -2.342595100402832, "rewards/margins": 5.2819952964782715, "rewards/rejected": -7.624590873718262, "step": 7500 }, { "epoch": 1.1111111111111112, "grad_norm": 0.07451526075601578, "learning_rate": 3.148148148148148e-05, "logits/chosen": 2.9968857765197754, "logits/rejected": 2.8013899326324463, "logps/chosen": -573.6311645507812, "logps/rejected": -600.4711303710938, "loss": 0.1612, "rewards/accuracies": 0.9399999976158142, "rewards/chosen": -4.748135566711426, "rewards/margins": 7.484164714813232, "rewards/rejected": -12.2322998046875, "step": 8000 }, { "epoch": 1.1805555555555556, "grad_norm": 1.8925604820251465, "learning_rate": 3.0324074074074077e-05, "logits/chosen": 3.0607872009277344, "logits/rejected": 2.8461995124816895, "logps/chosen": -560.1910400390625, "logps/rejected": -586.6576538085938, "loss": 0.1552, "rewards/accuracies": 0.9520000219345093, "rewards/chosen": -4.0625996589660645, "rewards/margins": 7.87790584564209, "rewards/rejected": -11.940506935119629, "step": 8500 }, { "epoch": 1.25, "grad_norm": 10.571070671081543, "learning_rate": 2.916666666666667e-05, "logits/chosen": 2.885572910308838, "logits/rejected": 2.6674787998199463, "logps/chosen": -603.0391845703125, "logps/rejected": -638.549560546875, "loss": 0.1445, "rewards/accuracies": 0.9580000042915344, "rewards/chosen": -6.252330303192139, "rewards/margins": 8.686140060424805, "rewards/rejected": -14.938469886779785, "step": 9000 }, { "epoch": 1.3194444444444444, "grad_norm": 0.12200064957141876, "learning_rate": 2.8009259259259263e-05, "logits/chosen": 2.839589834213257, "logits/rejected": 2.6223013401031494, "logps/chosen": -597.833740234375, "logps/rejected": -641.7092895507812, "loss": 0.1494, "rewards/accuracies": 0.9509999752044678, "rewards/chosen": -6.904899597167969, "rewards/margins": 8.972027778625488, "rewards/rejected": -15.876925468444824, "step": 9500 }, { "epoch": 1.3888888888888888, "grad_norm": 0.003055981360375881, "learning_rate": 2.6851851851851855e-05, "logits/chosen": 3.0364255905151367, "logits/rejected": 2.80228853225708, "logps/chosen": -588.6766967773438, "logps/rejected": -626.947265625, "loss": 0.1363, "rewards/accuracies": 0.9480000138282776, "rewards/chosen": -6.176986217498779, "rewards/margins": 8.589890480041504, "rewards/rejected": -14.766877174377441, "step": 10000 }, { "epoch": 1.4583333333333333, "grad_norm": 0.5345703959465027, "learning_rate": 2.5694444444444445e-05, "logits/chosen": 2.839054584503174, "logits/rejected": 2.630932092666626, "logps/chosen": -599.6796264648438, "logps/rejected": -649.5350952148438, "loss": 0.1599, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -6.972626209259033, "rewards/margins": 8.69276237487793, "rewards/rejected": -15.665388107299805, "step": 10500 }, { "epoch": 1.5277777777777777, "grad_norm": 0.4993605613708496, "learning_rate": 2.4537037037037038e-05, "logits/chosen": 3.136254072189331, "logits/rejected": 2.8473451137542725, "logps/chosen": -594.6328735351562, "logps/rejected": -616.174560546875, "loss": 0.1279, "rewards/accuracies": 0.9470000267028809, "rewards/chosen": -6.083535671234131, "rewards/margins": 8.711966514587402, "rewards/rejected": -14.795503616333008, "step": 11000 }, { "epoch": 1.5972222222222223, "grad_norm": 0.0059162345714867115, "learning_rate": 2.337962962962963e-05, "logits/chosen": 3.0318968296051025, "logits/rejected": 2.7672841548919678, "logps/chosen": -604.5242919921875, "logps/rejected": -643.3104858398438, "loss": 0.1094, "rewards/accuracies": 0.9580000042915344, "rewards/chosen": -6.998290538787842, "rewards/margins": 9.226086616516113, "rewards/rejected": -16.224376678466797, "step": 11500 }, { "epoch": 1.6666666666666665, "grad_norm": 11.891778945922852, "learning_rate": 2.2222222222222223e-05, "logits/chosen": 2.9720585346221924, "logits/rejected": 2.7613561153411865, "logps/chosen": -598.2454833984375, "logps/rejected": -652.2257080078125, "loss": 0.1199, "rewards/accuracies": 0.9599999785423279, "rewards/chosen": -7.267679691314697, "rewards/margins": 9.064537048339844, "rewards/rejected": -16.33221435546875, "step": 12000 }, { "epoch": 1.7361111111111112, "grad_norm": 0.01290846522897482, "learning_rate": 2.1064814814814816e-05, "logits/chosen": 2.8649771213531494, "logits/rejected": 2.5800554752349854, "logps/chosen": -606.3182373046875, "logps/rejected": -657.0851440429688, "loss": 0.1071, "rewards/accuracies": 0.9610000252723694, "rewards/chosen": -7.935286998748779, "rewards/margins": 9.804253578186035, "rewards/rejected": -17.739540100097656, "step": 12500 }, { "epoch": 1.8055555555555556, "grad_norm": 0.04387454316020012, "learning_rate": 1.990740740740741e-05, "logits/chosen": 2.8485310077667236, "logits/rejected": 2.5972745418548584, "logps/chosen": -597.68701171875, "logps/rejected": -637.2554321289062, "loss": 0.1037, "rewards/accuracies": 0.9649999737739563, "rewards/chosen": -7.1608757972717285, "rewards/margins": 9.487983703613281, "rewards/rejected": -16.64885902404785, "step": 13000 }, { "epoch": 1.875, "grad_norm": 4.594534873962402, "learning_rate": 1.8750000000000002e-05, "logits/chosen": 2.9154677391052246, "logits/rejected": 2.624795913696289, "logps/chosen": -615.2523803710938, "logps/rejected": -660.990234375, "loss": 0.099, "rewards/accuracies": 0.9649999737739563, "rewards/chosen": -7.9369049072265625, "rewards/margins": 9.701655387878418, "rewards/rejected": -17.63856315612793, "step": 13500 }, { "epoch": 1.9444444444444444, "grad_norm": 54.135902404785156, "learning_rate": 1.7592592592592595e-05, "logits/chosen": 2.855975866317749, "logits/rejected": 2.6032371520996094, "logps/chosen": -591.5145263671875, "logps/rejected": -640.7100219726562, "loss": 0.1079, "rewards/accuracies": 0.9589999914169312, "rewards/chosen": -6.821244239807129, "rewards/margins": 9.38486385345459, "rewards/rejected": -16.20611000061035, "step": 14000 }, { "epoch": 2.013888888888889, "grad_norm": 2.3375725504592992e-05, "learning_rate": 1.6435185185185187e-05, "logits/chosen": 2.8908510208129883, "logits/rejected": 2.634883403778076, "logps/chosen": -597.045166015625, "logps/rejected": -651.947265625, "loss": 0.106, "rewards/accuracies": 0.9679999947547913, "rewards/chosen": -7.668105602264404, "rewards/margins": 9.93225383758545, "rewards/rejected": -17.600358963012695, "step": 14500 }, { "epoch": 2.0833333333333335, "grad_norm": 1.9361701011657715, "learning_rate": 1.527777777777778e-05, "logits/chosen": 2.695993661880493, "logits/rejected": 2.460109233856201, "logps/chosen": -595.7433471679688, "logps/rejected": -674.5877685546875, "loss": 0.0023, "rewards/accuracies": 0.9990000128746033, "rewards/chosen": -7.24924373626709, "rewards/margins": 12.572339057922363, "rewards/rejected": -19.821582794189453, "step": 15000 }, { "epoch": 2.1527777777777777, "grad_norm": 0.00951045285910368, "learning_rate": 1.412037037037037e-05, "logits/chosen": 2.6940438747406006, "logits/rejected": 2.3725061416625977, "logps/chosen": -623.6839599609375, "logps/rejected": -705.4991455078125, "loss": 0.0008, "rewards/accuracies": 1.0, "rewards/chosen": -7.520936012268066, "rewards/margins": 13.330465316772461, "rewards/rejected": -20.85140037536621, "step": 15500 }, { "epoch": 2.2222222222222223, "grad_norm": 0.001970913726836443, "learning_rate": 1.2962962962962962e-05, "logits/chosen": 2.7482447624206543, "logits/rejected": 2.4920620918273926, "logps/chosen": -608.4854125976562, "logps/rejected": -686.771240234375, "loss": 0.0033, "rewards/accuracies": 0.9990000128746033, "rewards/chosen": -7.450053691864014, "rewards/margins": 13.123722076416016, "rewards/rejected": -20.573780059814453, "step": 16000 }, { "epoch": 2.2916666666666665, "grad_norm": 0.011360249482095242, "learning_rate": 1.1805555555555555e-05, "logits/chosen": 2.576791524887085, "logits/rejected": 2.2490572929382324, "logps/chosen": -606.9148559570312, "logps/rejected": -680.7444458007812, "loss": 0.0068, "rewards/accuracies": 0.996999979019165, "rewards/chosen": -7.9679388999938965, "rewards/margins": 13.592195510864258, "rewards/rejected": -21.560136795043945, "step": 16500 }, { "epoch": 2.361111111111111, "grad_norm": 0.38025176525115967, "learning_rate": 1.0648148148148148e-05, "logits/chosen": 2.553051233291626, "logits/rejected": 2.2183096408843994, "logps/chosen": -614.6328125, "logps/rejected": -694.4602661132812, "loss": 0.0034, "rewards/accuracies": 1.0, "rewards/chosen": -8.198136329650879, "rewards/margins": 13.553929328918457, "rewards/rejected": -21.752065658569336, "step": 17000 }, { "epoch": 2.4305555555555554, "grad_norm": 0.03263687342405319, "learning_rate": 9.490740740740741e-06, "logits/chosen": 2.672771692276001, "logits/rejected": 2.395473003387451, "logps/chosen": -608.4843139648438, "logps/rejected": -698.3759155273438, "loss": 0.0073, "rewards/accuracies": 0.9959999918937683, "rewards/chosen": -7.606570720672607, "rewards/margins": 13.408744812011719, "rewards/rejected": -21.01531410217285, "step": 17500 }, { "epoch": 2.5, "grad_norm": 0.004458485636860132, "learning_rate": 8.333333333333334e-06, "logits/chosen": 2.5551936626434326, "logits/rejected": 2.2830896377563477, "logps/chosen": -611.794921875, "logps/rejected": -703.889404296875, "loss": 0.0026, "rewards/accuracies": 0.9980000257492065, "rewards/chosen": -8.585135459899902, "rewards/margins": 13.675087928771973, "rewards/rejected": -22.260225296020508, "step": 18000 }, { "epoch": 2.5694444444444446, "grad_norm": 1.5490238638449227e-06, "learning_rate": 7.1759259259259266e-06, "logits/chosen": 2.552164077758789, "logits/rejected": 2.2521095275878906, "logps/chosen": -610.4937133789062, "logps/rejected": -696.794189453125, "loss": 0.0028, "rewards/accuracies": 0.9980000257492065, "rewards/chosen": -8.916465759277344, "rewards/margins": 13.4678955078125, "rewards/rejected": -22.384361267089844, "step": 18500 }, { "epoch": 2.638888888888889, "grad_norm": 0.00105581886600703, "learning_rate": 6.0185185185185185e-06, "logits/chosen": 2.5020742416381836, "logits/rejected": 2.2146670818328857, "logps/chosen": -606.8447265625, "logps/rejected": -700.9844360351562, "loss": 0.0029, "rewards/accuracies": 0.996999979019165, "rewards/chosen": -8.31242847442627, "rewards/margins": 14.17545223236084, "rewards/rejected": -22.487882614135742, "step": 19000 }, { "epoch": 2.7083333333333335, "grad_norm": 0.0019914801232516766, "learning_rate": 4.861111111111111e-06, "logits/chosen": 2.635819435119629, "logits/rejected": 2.366089344024658, "logps/chosen": -607.3903198242188, "logps/rejected": -700.4667358398438, "loss": 0.0019, "rewards/accuracies": 1.0, "rewards/chosen": -7.668735980987549, "rewards/margins": 14.018786430358887, "rewards/rejected": -21.687519073486328, "step": 19500 }, { "epoch": 2.7777777777777777, "grad_norm": 0.014177707955241203, "learning_rate": 3.7037037037037037e-06, "logits/chosen": 2.7170698642730713, "logits/rejected": 2.447775363922119, "logps/chosen": -611.0238647460938, "logps/rejected": -686.9364013671875, "loss": 0.0013, "rewards/accuracies": 1.0, "rewards/chosen": -7.709755897521973, "rewards/margins": 13.24997615814209, "rewards/rejected": -20.959733963012695, "step": 20000 }, { "epoch": 2.8472222222222223, "grad_norm": 0.00307221757248044, "learning_rate": 2.546296296296296e-06, "logits/chosen": 2.6886892318725586, "logits/rejected": 2.4070165157318115, "logps/chosen": -599.3004760742188, "logps/rejected": -683.9198608398438, "loss": 0.0027, "rewards/accuracies": 0.9980000257492065, "rewards/chosen": -7.5910773277282715, "rewards/margins": 13.034306526184082, "rewards/rejected": -20.625385284423828, "step": 20500 }, { "epoch": 2.9166666666666665, "grad_norm": 0.432113915681839, "learning_rate": 1.388888888888889e-06, "logits/chosen": 2.5857932567596436, "logits/rejected": 2.2961175441741943, "logps/chosen": -601.2279663085938, "logps/rejected": -694.5283203125, "loss": 0.0018, "rewards/accuracies": 0.9990000128746033, "rewards/chosen": -7.844883441925049, "rewards/margins": 13.75926399230957, "rewards/rejected": -21.60414695739746, "step": 21000 }, { "epoch": 2.986111111111111, "grad_norm": 0.0006206350517459214, "learning_rate": 2.3148148148148148e-07, "logits/chosen": 2.5767359733581543, "logits/rejected": 2.3402278423309326, "logps/chosen": -605.1397705078125, "logps/rejected": -705.0686645507812, "loss": 0.0058, "rewards/accuracies": 0.9959999918937683, "rewards/chosen": -8.206979751586914, "rewards/margins": 13.36180305480957, "rewards/rejected": -21.56878089904785, "step": 21500 } ], "logging_steps": 500, "max_steps": 21600, "num_input_tokens_seen": 0, "num_train_epochs": 3, "save_steps": 500, "total_flos": 0.0, "train_batch_size": 2, "trial_name": null, "trial_params": null }