{ "best_metric": null, "best_model_checkpoint": null, "epoch": 1.3066557778685177, "eval_steps": 500, "global_step": 400, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.016333197223356473, "grad_norm": 1.1593419313430786, "learning_rate": 4.9996340229869763e-05, "logits/chosen": -1.0346699953079224, "logits/rejected": -2.6695446968078613, "logps/chosen": -1.101169466972351, "logps/rejected": -1.0235556364059448, "loss": 1.178, "odds_ratio_loss": 0.768240213394165, "rewards/accuracies": 0.3499999940395355, "rewards/chosen": -0.11011694371700287, "rewards/margins": -0.007761377841234207, "rewards/rejected": -0.10235557705163956, "sft_loss": 1.101169466972351, "step": 5 }, { "epoch": 0.032666394446712946, "grad_norm": 1.1402696371078491, "learning_rate": 4.9985361990992455e-05, "logits/chosen": -0.8193758726119995, "logits/rejected": -2.842833995819092, "logps/chosen": -1.0217273235321045, "logps/rejected": -1.0100481510162354, "loss": 1.0927, "odds_ratio_loss": 0.7092453837394714, "rewards/accuracies": 0.5249999761581421, "rewards/chosen": -0.10217274725437164, "rewards/margins": -0.0011679243762046099, "rewards/rejected": -0.10100481659173965, "sft_loss": 1.0217273235321045, "step": 10 }, { "epoch": 0.04899959167006942, "grad_norm": 1.3554140329360962, "learning_rate": 4.996706849759453e-05, "logits/chosen": -0.7696830034255981, "logits/rejected": -2.703421115875244, "logps/chosen": -1.0294106006622314, "logps/rejected": -1.0599125623703003, "loss": 1.0993, "odds_ratio_loss": 0.6984036564826965, "rewards/accuracies": 0.5, "rewards/chosen": -0.10294107347726822, "rewards/margins": 0.003050185739994049, "rewards/rejected": -0.10599125921726227, "sft_loss": 1.0294106006622314, "step": 15 }, { "epoch": 0.06533278889342589, "grad_norm": 1.406853199005127, "learning_rate": 4.9941465105674435e-05, "logits/chosen": -0.698488712310791, "logits/rejected": -2.808516263961792, "logps/chosen": -1.0651928186416626, "logps/rejected": -0.996777355670929, "loss": 1.1402, "odds_ratio_loss": 0.7504132390022278, "rewards/accuracies": 0.44999998807907104, "rewards/chosen": -0.10651928186416626, "rewards/margins": -0.006841552909463644, "rewards/rejected": -0.09967772662639618, "sft_loss": 1.0651928186416626, "step": 20 }, { "epoch": 0.08166598611678236, "grad_norm": 1.238513469696045, "learning_rate": 4.99085593114345e-05, "logits/chosen": -0.6306881904602051, "logits/rejected": -2.698362350463867, "logps/chosen": -1.0242538452148438, "logps/rejected": -1.0320734977722168, "loss": 1.0937, "odds_ratio_loss": 0.6945221424102783, "rewards/accuracies": 0.5249999761581421, "rewards/chosen": -0.10242538154125214, "rewards/margins": 0.0007819652673788369, "rewards/rejected": -0.10320734977722168, "sft_loss": 1.0242538452148438, "step": 25 }, { "epoch": 0.09799918334013884, "grad_norm": 1.9588699340820312, "learning_rate": 4.986836074908616e-05, "logits/chosen": -0.5534936189651489, "logits/rejected": -2.678889036178589, "logps/chosen": -1.1003000736236572, "logps/rejected": -1.0425399541854858, "loss": 1.1745, "odds_ratio_loss": 0.7416313290596008, "rewards/accuracies": 0.5, "rewards/chosen": -0.11003003269433975, "rewards/margins": -0.005776026751846075, "rewards/rejected": -0.10425399243831635, "sft_loss": 1.1003000736236572, "step": 30 }, { "epoch": 0.1143323805634953, "grad_norm": 1.1857821941375732, "learning_rate": 4.982088118802922e-05, "logits/chosen": -0.6129628419876099, "logits/rejected": -2.876364231109619, "logps/chosen": -0.9628928303718567, "logps/rejected": -0.952060341835022, "loss": 1.0336, "odds_ratio_loss": 0.7068274617195129, "rewards/accuracies": 0.42500001192092896, "rewards/chosen": -0.09628929197788239, "rewards/margins": -0.001083258306607604, "rewards/rejected": -0.09520602226257324, "sft_loss": 0.9628928303718567, "step": 35 }, { "epoch": 0.13066557778685178, "grad_norm": 1.33924400806427, "learning_rate": 4.976613452940604e-05, "logits/chosen": -0.7091814279556274, "logits/rejected": -2.6595664024353027, "logps/chosen": -0.89775550365448, "logps/rejected": -0.935952365398407, "loss": 0.9642, "odds_ratio_loss": 0.6644185781478882, "rewards/accuracies": 0.625, "rewards/chosen": -0.08977555483579636, "rewards/margins": 0.003819683101028204, "rewards/rejected": -0.0935952290892601, "sft_loss": 0.89775550365448, "step": 40 }, { "epoch": 0.14699877501020825, "grad_norm": 1.136667013168335, "learning_rate": 4.9704136802031485e-05, "logits/chosen": -0.7233263850212097, "logits/rejected": -2.6673388481140137, "logps/chosen": -0.9036429524421692, "logps/rejected": -0.9419530630111694, "loss": 0.9699, "odds_ratio_loss": 0.662254273891449, "rewards/accuracies": 0.625, "rewards/chosen": -0.09036429226398468, "rewards/margins": 0.0038310133386403322, "rewards/rejected": -0.09419531375169754, "sft_loss": 0.9036429524421692, "step": 45 }, { "epoch": 0.16333197223356472, "grad_norm": 1.4290399551391602, "learning_rate": 4.9634906157700036e-05, "logits/chosen": -0.6113058924674988, "logits/rejected": -2.705373764038086, "logps/chosen": -0.9550830125808716, "logps/rejected": -0.9817631840705872, "loss": 1.0223, "odds_ratio_loss": 0.6725824475288391, "rewards/accuracies": 0.574999988079071, "rewards/chosen": -0.09550830721855164, "rewards/margins": 0.002668013097718358, "rewards/rejected": -0.09817632287740707, "sft_loss": 0.9550830125808716, "step": 50 }, { "epoch": 0.17966516945692118, "grad_norm": 1.4945589303970337, "learning_rate": 4.955846286587122e-05, "logits/chosen": -0.647710919380188, "logits/rejected": -2.7208316326141357, "logps/chosen": -0.9462896585464478, "logps/rejected": -0.9507497549057007, "loss": 1.0148, "odds_ratio_loss": 0.6852728128433228, "rewards/accuracies": 0.550000011920929, "rewards/chosen": -0.0946289673447609, "rewards/margins": 0.0004460157942958176, "rewards/rejected": -0.09507497400045395, "sft_loss": 0.9462896585464478, "step": 55 }, { "epoch": 0.19599836668027767, "grad_norm": 1.3553156852722168, "learning_rate": 4.9474829307735115e-05, "logits/chosen": -0.6279743909835815, "logits/rejected": -2.6854496002197266, "logps/chosen": -0.8896121978759766, "logps/rejected": -0.9993942975997925, "loss": 0.9509, "odds_ratio_loss": 0.6129180192947388, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -0.08896122127771378, "rewards/margins": 0.010978218168020248, "rewards/rejected": -0.09993942826986313, "sft_loss": 0.8896121978759766, "step": 60 }, { "epoch": 0.21233156390363414, "grad_norm": 1.4247411489486694, "learning_rate": 4.9384029969659573e-05, "logits/chosen": -0.698381781578064, "logits/rejected": -2.7967753410339355, "logps/chosen": -0.9455701112747192, "logps/rejected": -0.9564310908317566, "loss": 1.0147, "odds_ratio_loss": 0.6911824941635132, "rewards/accuracies": 0.5, "rewards/chosen": -0.0945570170879364, "rewards/margins": 0.0010860899928957224, "rewards/rejected": -0.09564311057329178, "sft_loss": 0.9455701112747192, "step": 65 }, { "epoch": 0.2286647611269906, "grad_norm": 1.6154508590698242, "learning_rate": 4.9286091436021015e-05, "logits/chosen": -0.595032274723053, "logits/rejected": -2.740784168243408, "logps/chosen": -0.9214372634887695, "logps/rejected": -1.0210167169570923, "loss": 0.9836, "odds_ratio_loss": 0.6212525963783264, "rewards/accuracies": 0.699999988079071, "rewards/chosen": -0.09214373677968979, "rewards/margins": 0.00995794590562582, "rewards/rejected": -0.10210166871547699, "sft_loss": 0.9214372634887695, "step": 70 }, { "epoch": 0.24499795835034707, "grad_norm": 1.8231102228164673, "learning_rate": 4.918104238142104e-05, "logits/chosen": -0.5186889171600342, "logits/rejected": -2.7576358318328857, "logps/chosen": -0.9789311289787292, "logps/rejected": -1.0169849395751953, "loss": 1.0454, "odds_ratio_loss": 0.6644267439842224, "rewards/accuracies": 0.6000000238418579, "rewards/chosen": -0.09789310395717621, "rewards/margins": 0.0038053765892982483, "rewards/rejected": -0.10169849544763565, "sft_loss": 0.9789311289787292, "step": 75 }, { "epoch": 0.26133115557370357, "grad_norm": 2.05936336517334, "learning_rate": 4.906891356229103e-05, "logits/chosen": -0.47254711389541626, "logits/rejected": -2.68280029296875, "logps/chosen": -0.9620391726493835, "logps/rejected": -1.0026928186416626, "loss": 1.0294, "odds_ratio_loss": 0.6731926202774048, "rewards/accuracies": 0.5, "rewards/chosen": -0.09620390832424164, "rewards/margins": 0.004065364599227905, "rewards/rejected": -0.10026928037405014, "sft_loss": 0.9620391726493835, "step": 80 }, { "epoch": 0.27766435279706003, "grad_norm": 1.6302626132965088, "learning_rate": 4.894973780788722e-05, "logits/chosen": -0.5315309762954712, "logits/rejected": -2.7227425575256348, "logps/chosen": -0.9423310160636902, "logps/rejected": -1.071781039237976, "loss": 1.0032, "odds_ratio_loss": 0.6090766191482544, "rewards/accuracies": 0.7749999761581421, "rewards/chosen": -0.09423311054706573, "rewards/margins": 0.012944999150931835, "rewards/rejected": -0.10717810690402985, "sft_loss": 0.9423310160636902, "step": 85 }, { "epoch": 0.2939975500204165, "grad_norm": 1.7891793251037598, "learning_rate": 4.882355001067892e-05, "logits/chosen": -0.5275685787200928, "logits/rejected": -2.6127567291259766, "logps/chosen": -0.9351356625556946, "logps/rejected": -1.0083874464035034, "loss": 0.9995, "odds_ratio_loss": 0.6435801386833191, "rewards/accuracies": 0.625, "rewards/chosen": -0.09351354837417603, "rewards/margins": 0.0073251971043646336, "rewards/rejected": -0.10083875805139542, "sft_loss": 0.9351356625556946, "step": 90 }, { "epoch": 0.31033074724377296, "grad_norm": 1.852216124534607, "learning_rate": 4.869038711613259e-05, "logits/chosen": -0.5760647058486938, "logits/rejected": -2.722364664077759, "logps/chosen": -0.9301847219467163, "logps/rejected": -1.0749822854995728, "loss": 0.9903, "odds_ratio_loss": 0.601297914981842, "rewards/accuracies": 0.75, "rewards/chosen": -0.09301847964525223, "rewards/margins": 0.014479748904705048, "rewards/rejected": -0.10749821364879608, "sft_loss": 0.9301847219467163, "step": 95 }, { "epoch": 0.32666394446712943, "grad_norm": 1.8597307205200195, "learning_rate": 4.855028811189496e-05, "logits/chosen": -0.5871164202690125, "logits/rejected": -2.6969285011291504, "logps/chosen": -1.0701930522918701, "logps/rejected": -1.0556358098983765, "loss": 1.1427, "odds_ratio_loss": 0.7247555255889893, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -0.10701930522918701, "rewards/margins": -0.0014557272661477327, "rewards/rejected": -0.10556356608867645, "sft_loss": 1.0701930522918701, "step": 100 }, { "epoch": 0.3429971416904859, "grad_norm": 1.6869758367538452, "learning_rate": 4.8403294016378095e-05, "logits/chosen": -0.4466473460197449, "logits/rejected": -2.501394510269165, "logps/chosen": -0.880301833152771, "logps/rejected": -1.0003767013549805, "loss": 0.941, "odds_ratio_loss": 0.60676509141922, "rewards/accuracies": 0.675000011920929, "rewards/chosen": -0.08803018927574158, "rewards/margins": 0.012007484212517738, "rewards/rejected": -0.10003767162561417, "sft_loss": 0.880301833152771, "step": 105 }, { "epoch": 0.35933033891384236, "grad_norm": 1.8963792324066162, "learning_rate": 4.8249447866750025e-05, "logits/chosen": -0.41199809312820435, "logits/rejected": -2.594975471496582, "logps/chosen": -0.8528003692626953, "logps/rejected": -0.9772283434867859, "loss": 0.9128, "odds_ratio_loss": 0.6004671454429626, "rewards/accuracies": 0.7749999761581421, "rewards/chosen": -0.08528003096580505, "rewards/margins": 0.012442806735634804, "rewards/rejected": -0.0977228432893753, "sft_loss": 0.8528003692626953, "step": 110 }, { "epoch": 0.3756635361371989, "grad_norm": 2.6958725452423096, "learning_rate": 4.808879470633417e-05, "logits/chosen": -0.43985629081726074, "logits/rejected": -2.6999452114105225, "logps/chosen": -0.9486901164054871, "logps/rejected": -1.0987169742584229, "loss": 1.0076, "odds_ratio_loss": 0.5889819264411926, "rewards/accuracies": 0.8500000238418579, "rewards/chosen": -0.09486902505159378, "rewards/margins": 0.015002680011093616, "rewards/rejected": -0.10987170040607452, "sft_loss": 0.9486901164054871, "step": 115 }, { "epoch": 0.39199673336055535, "grad_norm": 1.6014004945755005, "learning_rate": 4.792138157142158e-05, "logits/chosen": -0.43066906929016113, "logits/rejected": -2.634669065475464, "logps/chosen": -0.8806023597717285, "logps/rejected": -1.0933107137680054, "loss": 0.9352, "odds_ratio_loss": 0.5455727577209473, "rewards/accuracies": 0.824999988079071, "rewards/chosen": -0.08806024491786957, "rewards/margins": 0.021270830184221268, "rewards/rejected": -0.10933107137680054, "sft_loss": 0.8806023597717285, "step": 120 }, { "epoch": 0.4083299305839118, "grad_norm": 3.4042723178863525, "learning_rate": 4.7747257477499474e-05, "logits/chosen": -0.4240710735321045, "logits/rejected": -2.664668560028076, "logps/chosen": -0.8993316888809204, "logps/rejected": -1.0890860557556152, "loss": 0.9553, "odds_ratio_loss": 0.5601183772087097, "rewards/accuracies": 0.824999988079071, "rewards/chosen": -0.08993317186832428, "rewards/margins": 0.018975432962179184, "rewards/rejected": -0.10890860855579376, "sft_loss": 0.8993316888809204, "step": 125 }, { "epoch": 0.4246631278072683, "grad_norm": 1.724076509475708, "learning_rate": 4.75664734049005e-05, "logits/chosen": -0.3910972476005554, "logits/rejected": -2.571296215057373, "logps/chosen": -0.9113162755966187, "logps/rejected": -1.094262719154358, "loss": 0.9682, "odds_ratio_loss": 0.5683911442756653, "rewards/accuracies": 0.875, "rewards/chosen": -0.09113162755966187, "rewards/margins": 0.018294643610715866, "rewards/rejected": -0.10942627489566803, "sft_loss": 0.9113162755966187, "step": 130 }, { "epoch": 0.44099632503062475, "grad_norm": 1.6363219022750854, "learning_rate": 4.7379082283876566e-05, "logits/chosen": -0.5021435022354126, "logits/rejected": -2.6572070121765137, "logps/chosen": -0.8906953930854797, "logps/rejected": -1.0344655513763428, "loss": 0.9498, "odds_ratio_loss": 0.5914202928543091, "rewards/accuracies": 0.75, "rewards/chosen": -0.08906953036785126, "rewards/margins": 0.014377007260918617, "rewards/rejected": -0.10344654321670532, "sft_loss": 0.8906953930854797, "step": 135 }, { "epoch": 0.4573295222539812, "grad_norm": 1.7812143564224243, "learning_rate": 4.7185138979101864e-05, "logits/chosen": -0.507682740688324, "logits/rejected": -2.593475580215454, "logps/chosen": -0.9167561531066895, "logps/rejected": -1.1608246564865112, "loss": 0.9698, "odds_ratio_loss": 0.5303933024406433, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.09167560935020447, "rewards/margins": 0.024406855925917625, "rewards/rejected": -0.11608246713876724, "sft_loss": 0.9167561531066895, "step": 140 }, { "epoch": 0.4736627194773377, "grad_norm": 1.823789119720459, "learning_rate": 4.6984700273609495e-05, "logits/chosen": -0.3446227014064789, "logits/rejected": -2.5670173168182373, "logps/chosen": -0.8484543561935425, "logps/rejected": -1.0493099689483643, "loss": 0.903, "odds_ratio_loss": 0.5457809567451477, "rewards/accuracies": 0.8500000238418579, "rewards/chosen": -0.08484544605016708, "rewards/margins": 0.020085560157895088, "rewards/rejected": -0.10493101179599762, "sft_loss": 0.8484543561935425, "step": 145 }, { "epoch": 0.48999591670069415, "grad_norm": 1.889367699623108, "learning_rate": 4.677782485216644e-05, "logits/chosen": -0.44083428382873535, "logits/rejected": -2.573638439178467, "logps/chosen": -0.833287239074707, "logps/rejected": -1.1050159931182861, "loss": 0.8838, "odds_ratio_loss": 0.5049694180488586, "rewards/accuracies": 0.875, "rewards/chosen": -0.0833287239074707, "rewards/margins": 0.02717287465929985, "rewards/rejected": -0.11050160229206085, "sft_loss": 0.833287239074707, "step": 150 }, { "epoch": 0.5063291139240507, "grad_norm": 1.9978969097137451, "learning_rate": 4.6564573284091765e-05, "logits/chosen": -0.37475016713142395, "logits/rejected": -2.530078887939453, "logps/chosen": -0.9055852890014648, "logps/rejected": -1.3165007829666138, "loss": 0.9498, "odds_ratio_loss": 0.4424561858177185, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.09055852890014648, "rewards/margins": 0.04109155014157295, "rewards/rejected": -0.13165007531642914, "sft_loss": 0.9055852890014648, "step": 155 }, { "epoch": 0.5226623111474071, "grad_norm": 2.2659225463867188, "learning_rate": 4.6345008005522966e-05, "logits/chosen": -0.27824097871780396, "logits/rejected": -2.1937553882598877, "logps/chosen": -0.7978348731994629, "logps/rejected": -1.7242333889007568, "loss": 0.8243, "odds_ratio_loss": 0.26432332396507263, "rewards/accuracies": 1.0, "rewards/chosen": -0.07978348433971405, "rewards/margins": 0.09263985604047775, "rewards/rejected": -0.1724233478307724, "sft_loss": 0.7978348731994629, "step": 160 }, { "epoch": 0.5389955083707636, "grad_norm": 2.497178792953491, "learning_rate": 4.611919330113592e-05, "logits/chosen": -0.1297352910041809, "logits/rejected": -1.7525418996810913, "logps/chosen": -0.9057181477546692, "logps/rejected": -2.926887035369873, "loss": 0.9163, "odds_ratio_loss": 0.10600633919239044, "rewards/accuracies": 1.0, "rewards/chosen": -0.0905718058347702, "rewards/margins": 0.20211689174175262, "rewards/rejected": -0.2926887273788452, "sft_loss": 0.9057181477546692, "step": 165 }, { "epoch": 0.5553287055941201, "grad_norm": 1.9597887992858887, "learning_rate": 4.588719528532342e-05, "logits/chosen": -0.20320752263069153, "logits/rejected": -1.4767993688583374, "logps/chosen": -0.9343756437301636, "logps/rejected": -3.7731080055236816, "loss": 0.9406, "odds_ratio_loss": 0.062325619161129, "rewards/accuracies": 1.0, "rewards/chosen": -0.0934375673532486, "rewards/margins": 0.28387323021888733, "rewards/rejected": -0.3773108124732971, "sft_loss": 0.9343756437301636, "step": 170 }, { "epoch": 0.5716619028174765, "grad_norm": 1.9177374839782715, "learning_rate": 4.564908188283814e-05, "logits/chosen": -0.16985416412353516, "logits/rejected": -1.3144887685775757, "logps/chosen": -0.8834988474845886, "logps/rejected": -4.147261619567871, "loss": 0.8871, "odds_ratio_loss": 0.03579668328166008, "rewards/accuracies": 1.0, "rewards/chosen": -0.08834987133741379, "rewards/margins": 0.3263762891292572, "rewards/rejected": -0.414726197719574, "sft_loss": 0.8834988474845886, "step": 175 }, { "epoch": 0.587995100040833, "grad_norm": 1.9115794897079468, "learning_rate": 4.540492280890555e-05, "logits/chosen": -0.34701991081237793, "logits/rejected": -1.1767631769180298, "logps/chosen": -0.8725343942642212, "logps/rejected": -4.597090721130371, "loss": 0.8753, "odds_ratio_loss": 0.027656957507133484, "rewards/accuracies": 1.0, "rewards/chosen": -0.08725343644618988, "rewards/margins": 0.3724556267261505, "rewards/rejected": -0.4597090780735016, "sft_loss": 0.8725343942642212, "step": 180 }, { "epoch": 0.6043282972641895, "grad_norm": 1.7303317785263062, "learning_rate": 4.515478954881248e-05, "logits/chosen": -0.37866610288619995, "logits/rejected": -1.20633065700531, "logps/chosen": -0.8602287173271179, "logps/rejected": -4.4016923904418945, "loss": 0.8625, "odds_ratio_loss": 0.022269519045948982, "rewards/accuracies": 1.0, "rewards/chosen": -0.08602287620306015, "rewards/margins": 0.35414639115333557, "rewards/rejected": -0.4401692748069763, "sft_loss": 0.8602287173271179, "step": 185 }, { "epoch": 0.6206614944875459, "grad_norm": 5.377653121948242, "learning_rate": 4.4898755336977673e-05, "logits/chosen": -0.32614630460739136, "logits/rejected": -1.1425362825393677, "logps/chosen": -0.9115999937057495, "logps/rejected": -4.731467247009277, "loss": 0.9149, "odds_ratio_loss": 0.0329320952296257, "rewards/accuracies": 1.0, "rewards/chosen": -0.09115998446941376, "rewards/margins": 0.38198670744895935, "rewards/rejected": -0.4731467366218567, "sft_loss": 0.9115999937057495, "step": 190 }, { "epoch": 0.6369946917109024, "grad_norm": 1.6315754652023315, "learning_rate": 4.463689513550997e-05, "logits/chosen": -0.3550247550010681, "logits/rejected": -1.241008996963501, "logps/chosen": -0.8556908369064331, "logps/rejected": -4.392051696777344, "loss": 0.8582, "odds_ratio_loss": 0.024758270010352135, "rewards/accuracies": 1.0, "rewards/chosen": -0.08556908369064331, "rewards/margins": 0.35363608598709106, "rewards/rejected": -0.4392051696777344, "sft_loss": 0.8556908369064331, "step": 195 }, { "epoch": 0.6533278889342589, "grad_norm": 1.8591830730438232, "learning_rate": 4.436928561226087e-05, "logits/chosen": -0.3138996660709381, "logits/rejected": -1.2280880212783813, "logps/chosen": -0.901713490486145, "logps/rejected": -4.212186336517334, "loss": 0.9045, "odds_ratio_loss": 0.027884572744369507, "rewards/accuracies": 1.0, "rewards/chosen": -0.09017134457826614, "rewards/margins": 0.3310473561286926, "rewards/rejected": -0.4212186932563782, "sft_loss": 0.901713490486145, "step": 200 }, { "epoch": 0.6696610861576153, "grad_norm": 1.8346513509750366, "learning_rate": 4.409600511837756e-05, "logits/chosen": -0.23156340420246124, "logits/rejected": -1.138501524925232, "logps/chosen": -0.9389227032661438, "logps/rejected": -4.50938606262207, "loss": 0.9416, "odds_ratio_loss": 0.02684590592980385, "rewards/accuracies": 1.0, "rewards/chosen": -0.09389226883649826, "rewards/margins": 0.35704633593559265, "rewards/rejected": -0.4509386122226715, "sft_loss": 0.9389227032661438, "step": 205 }, { "epoch": 0.6859942833809718, "grad_norm": 1.5812890529632568, "learning_rate": 4.381713366536311e-05, "logits/chosen": -0.24586188793182373, "logits/rejected": -1.1232062578201294, "logps/chosen": -0.859118640422821, "logps/rejected": -4.883942127227783, "loss": 0.8609, "odds_ratio_loss": 0.018183354288339615, "rewards/accuracies": 1.0, "rewards/chosen": -0.08591187745332718, "rewards/margins": 0.4024823307991028, "rewards/rejected": -0.48839420080184937, "sft_loss": 0.859118640422821, "step": 210 }, { "epoch": 0.7023274806043283, "grad_norm": 1.8375471830368042, "learning_rate": 4.353275290165066e-05, "logits/chosen": -0.15520325303077698, "logits/rejected": -0.985150933265686, "logps/chosen": -0.8214719891548157, "logps/rejected": -4.9630866050720215, "loss": 0.823, "odds_ratio_loss": 0.014944888651371002, "rewards/accuracies": 1.0, "rewards/chosen": -0.08214719593524933, "rewards/margins": 0.4141615033149719, "rewards/rejected": -0.49630871415138245, "sft_loss": 0.8214719891548157, "step": 215 }, { "epoch": 0.7186606778276847, "grad_norm": 2.556330919265747, "learning_rate": 4.324294608869817e-05, "logits/chosen": -0.3221624791622162, "logits/rejected": -0.9542757272720337, "logps/chosen": -0.9679568409919739, "logps/rejected": -5.273629188537598, "loss": 0.9693, "odds_ratio_loss": 0.013113953173160553, "rewards/accuracies": 1.0, "rewards/chosen": -0.0967956930398941, "rewards/margins": 0.4305671751499176, "rewards/rejected": -0.5273628830909729, "sft_loss": 0.9679568409919739, "step": 220 }, { "epoch": 0.7349938750510412, "grad_norm": 2.346754312515259, "learning_rate": 4.294779807661105e-05, "logits/chosen": -0.2690557539463043, "logits/rejected": -1.0382682085037231, "logps/chosen": -0.8765946626663208, "logps/rejected": -5.27601432800293, "loss": 0.8783, "odds_ratio_loss": 0.016597110778093338, "rewards/accuracies": 1.0, "rewards/chosen": -0.08765946328639984, "rewards/margins": 0.43994197249412537, "rewards/rejected": -0.527601420879364, "sft_loss": 0.8765946626663208, "step": 225 }, { "epoch": 0.7513270722743978, "grad_norm": 1.543803095817566, "learning_rate": 4.270789202181204e-05, "logits/chosen": -0.3406456410884857, "logits/rejected": -1.0243726968765259, "logps/chosen": -0.938999354839325, "logps/rejected": -5.179265022277832, "loss": 0.9411, "odds_ratio_loss": 0.021242566406726837, "rewards/accuracies": 1.0, "rewards/chosen": -0.09389995038509369, "rewards/margins": 0.4240265488624573, "rewards/rejected": -0.5179265141487122, "sft_loss": 0.938999354839325, "step": 230 }, { "epoch": 0.7676602694977542, "grad_norm": 1.8233140707015991, "learning_rate": 4.2403348649073174e-05, "logits/chosen": -0.2546311020851135, "logits/rejected": -0.9541198015213013, "logps/chosen": -0.7934538722038269, "logps/rejected": -5.206686973571777, "loss": 0.7944, "odds_ratio_loss": 0.00939333625137806, "rewards/accuracies": 1.0, "rewards/chosen": -0.07934539020061493, "rewards/margins": 0.4413233697414398, "rewards/rejected": -0.5206687450408936, "sft_loss": 0.7934538722038269, "step": 235 }, { "epoch": 0.7839934667211107, "grad_norm": 1.5795395374298096, "learning_rate": 4.209370989589013e-05, "logits/chosen": -0.3499365448951721, "logits/rejected": -1.0148460865020752, "logps/chosen": -0.7999550104141235, "logps/rejected": -5.092876434326172, "loss": 0.8013, "odds_ratio_loss": 0.013519616797566414, "rewards/accuracies": 1.0, "rewards/chosen": -0.07999551296234131, "rewards/margins": 0.4292922019958496, "rewards/rejected": -0.5092877149581909, "sft_loss": 0.7999550104141235, "step": 240 }, { "epoch": 0.8003266639444672, "grad_norm": 1.8533421754837036, "learning_rate": 4.177906641879575e-05, "logits/chosen": -0.3404558598995209, "logits/rejected": -1.0349606275558472, "logps/chosen": -0.9274562001228333, "logps/rejected": -4.853649616241455, "loss": 0.9298, "odds_ratio_loss": 0.023337265476584435, "rewards/accuracies": 1.0, "rewards/chosen": -0.09274561703205109, "rewards/margins": 0.3926192820072174, "rewards/rejected": -0.4853649139404297, "sft_loss": 0.9274562001228333, "step": 245 }, { "epoch": 0.8166598611678236, "grad_norm": 1.792864441871643, "learning_rate": 4.145951033961395e-05, "logits/chosen": -0.39237070083618164, "logits/rejected": -1.20218825340271, "logps/chosen": -0.8956495523452759, "logps/rejected": -4.416346549987793, "loss": 0.8988, "odds_ratio_loss": 0.03134030103683472, "rewards/accuracies": 1.0, "rewards/chosen": -0.08956494927406311, "rewards/margins": 0.3520697355270386, "rewards/rejected": -0.4416346549987793, "sft_loss": 0.8956495523452759, "step": 250 }, { "epoch": 0.8329930583911801, "grad_norm": 1.388491153717041, "learning_rate": 4.1135135218488206e-05, "logits/chosen": -0.4966661036014557, "logits/rejected": -1.301102638244629, "logps/chosen": -0.822670578956604, "logps/rejected": -4.420852184295654, "loss": 0.8251, "odds_ratio_loss": 0.02416636422276497, "rewards/accuracies": 1.0, "rewards/chosen": -0.08226706087589264, "rewards/margins": 0.3598182201385498, "rewards/rejected": -0.4420852065086365, "sft_loss": 0.822670578956604, "step": 255 }, { "epoch": 0.8493262556145366, "grad_norm": 1.7927485704421997, "learning_rate": 4.080603602648887e-05, "logits/chosen": -0.38711199164390564, "logits/rejected": -1.1762360334396362, "logps/chosen": -0.8747900128364563, "logps/rejected": -4.844551086425781, "loss": 0.8772, "odds_ratio_loss": 0.024142403155565262, "rewards/accuracies": 1.0, "rewards/chosen": -0.08747899532318115, "rewards/margins": 0.3969760835170746, "rewards/rejected": -0.4844551086425781, "sft_loss": 0.8747900128364563, "step": 260 }, { "epoch": 0.865659452837893, "grad_norm": 1.763637661933899, "learning_rate": 4.047230911780737e-05, "logits/chosen": -0.4611227512359619, "logits/rejected": -1.092810034751892, "logps/chosen": -0.8380341529846191, "logps/rejected": -4.902915000915527, "loss": 0.8414, "odds_ratio_loss": 0.03330729901790619, "rewards/accuracies": 1.0, "rewards/chosen": -0.08380343019962311, "rewards/margins": 0.4064881205558777, "rewards/rejected": -0.4902915358543396, "sft_loss": 0.8380341529846191, "step": 265 }, { "epoch": 0.8819926500612495, "grad_norm": 1.7203322649002075, "learning_rate": 4.0134052201545447e-05, "logits/chosen": -0.24241454899311066, "logits/rejected": -0.9308720827102661, "logps/chosen": -0.8675628900527954, "logps/rejected": -5.390278339385986, "loss": 0.8685, "odds_ratio_loss": 0.009246011264622211, "rewards/accuracies": 1.0, "rewards/chosen": -0.08675628900527954, "rewards/margins": 0.4522715210914612, "rewards/rejected": -0.5390278100967407, "sft_loss": 0.8675628900527954, "step": 270 }, { "epoch": 0.898325847284606, "grad_norm": 1.9116899967193604, "learning_rate": 3.9791364313107804e-05, "logits/chosen": -0.401073157787323, "logits/rejected": -0.9959608912467957, "logps/chosen": -0.8706862330436707, "logps/rejected": -5.204241752624512, "loss": 0.872, "odds_ratio_loss": 0.0133589506149292, "rewards/accuracies": 1.0, "rewards/chosen": -0.08706863224506378, "rewards/margins": 0.4333556294441223, "rewards/rejected": -0.5204241871833801, "sft_loss": 0.8706862330436707, "step": 275 }, { "epoch": 0.9146590445079624, "grad_norm": 1.9296520948410034, "learning_rate": 3.9444345785206285e-05, "logits/chosen": -0.40887147188186646, "logits/rejected": -0.9652082324028015, "logps/chosen": -0.8309628367424011, "logps/rejected": -5.662296295166016, "loss": 0.8318, "odds_ratio_loss": 0.008708246983587742, "rewards/accuracies": 1.0, "rewards/chosen": -0.08309628069400787, "rewards/margins": 0.48313331604003906, "rewards/rejected": -0.5662296414375305, "sft_loss": 0.8309628367424011, "step": 280 }, { "epoch": 0.9309922417313189, "grad_norm": 3.055554151535034, "learning_rate": 3.909309821848432e-05, "logits/chosen": -0.3440246284008026, "logits/rejected": -0.8914414644241333, "logps/chosen": -0.8914461135864258, "logps/rejected": -5.75334358215332, "loss": 0.8921, "odds_ratio_loss": 0.0068351454101502895, "rewards/accuracies": 1.0, "rewards/chosen": -0.08914461731910706, "rewards/margins": 0.48618975281715393, "rewards/rejected": -0.575334370136261, "sft_loss": 0.8914461135864258, "step": 285 }, { "epoch": 0.9473254389546754, "grad_norm": 1.5384926795959473, "learning_rate": 3.873772445177015e-05, "logits/chosen": -0.37930798530578613, "logits/rejected": -0.9451101422309875, "logps/chosen": -0.8741447329521179, "logps/rejected": -5.84790563583374, "loss": 0.8751, "odds_ratio_loss": 0.009465084411203861, "rewards/accuracies": 1.0, "rewards/chosen": -0.08741448074579239, "rewards/margins": 0.49737611413002014, "rewards/rejected": -0.5847905874252319, "sft_loss": 0.8741447329521179, "step": 290 }, { "epoch": 0.9636586361780318, "grad_norm": 2.4715216159820557, "learning_rate": 3.837832853196751e-05, "logits/chosen": -0.3153218626976013, "logits/rejected": -0.9612942934036255, "logps/chosen": -0.8935874700546265, "logps/rejected": -5.140650749206543, "loss": 0.8956, "odds_ratio_loss": 0.01970650628209114, "rewards/accuracies": 1.0, "rewards/chosen": -0.08935876190662384, "rewards/margins": 0.4247063100337982, "rewards/rejected": -0.514065146446228, "sft_loss": 0.8935874700546265, "step": 295 }, { "epoch": 0.9799918334013883, "grad_norm": 1.9313322305679321, "learning_rate": 3.8015015683592575e-05, "logits/chosen": -0.30432718992233276, "logits/rejected": -1.099277377128601, "logps/chosen": -0.8711298108100891, "logps/rejected": -4.722751140594482, "loss": 0.873, "odds_ratio_loss": 0.018964512273669243, "rewards/accuracies": 1.0, "rewards/chosen": -0.08711298555135727, "rewards/margins": 0.3851621747016907, "rewards/rejected": -0.47227516770362854, "sft_loss": 0.8711298108100891, "step": 300 }, { "epoch": 0.9963250306247448, "grad_norm": 2.2331244945526123, "learning_rate": 3.764789227796616e-05, "logits/chosen": -0.3758210837841034, "logits/rejected": -1.1826483011245728, "logps/chosen": -0.8142495155334473, "logps/rejected": -4.7738471031188965, "loss": 0.816, "odds_ratio_loss": 0.017485955730080605, "rewards/accuracies": 1.0, "rewards/chosen": -0.08142495155334473, "rewards/margins": 0.3959597945213318, "rewards/rejected": -0.4773847460746765, "sft_loss": 0.8142495155334473, "step": 305 }, { "epoch": 1.0126582278481013, "grad_norm": 1.7077336311340332, "learning_rate": 3.727706580207021e-05, "logits/chosen": -0.38093143701553345, "logits/rejected": -1.096470594406128, "logps/chosen": -0.8485842943191528, "logps/rejected": -4.818546295166016, "loss": 0.8503, "odds_ratio_loss": 0.01746557094156742, "rewards/accuracies": 1.0, "rewards/chosen": -0.08485842496156693, "rewards/margins": 0.3969961702823639, "rewards/rejected": -0.4818546175956726, "sft_loss": 0.8485842943191528, "step": 310 }, { "epoch": 1.0289914250714578, "grad_norm": 1.8160737752914429, "learning_rate": 3.69026448270775e-05, "logits/chosen": -0.37943610548973083, "logits/rejected": -1.0238980054855347, "logps/chosen": -0.8509234189987183, "logps/rejected": -5.104341506958008, "loss": 0.8521, "odds_ratio_loss": 0.011288631707429886, "rewards/accuracies": 1.0, "rewards/chosen": -0.08509235084056854, "rewards/margins": 0.4253418445587158, "rewards/rejected": -0.5104341506958008, "sft_loss": 0.8509234189987183, "step": 315 }, { "epoch": 1.0453246222948143, "grad_norm": 2.9576852321624756, "learning_rate": 3.652473897656409e-05, "logits/chosen": -0.30405062437057495, "logits/rejected": -0.9463305473327637, "logps/chosen": -0.9074727296829224, "logps/rejected": -5.486522674560547, "loss": 0.9089, "odds_ratio_loss": 0.013948513194918633, "rewards/accuracies": 1.0, "rewards/chosen": -0.09074728190898895, "rewards/margins": 0.45790499448776245, "rewards/rejected": -0.5486522912979126, "sft_loss": 0.9074727296829224, "step": 320 }, { "epoch": 1.0616578195181707, "grad_norm": 1.9609084129333496, "learning_rate": 3.6143458894413465e-05, "logits/chosen": -0.39042356610298157, "logits/rejected": -1.0480425357818604, "logps/chosen": -0.7837303876876831, "logps/rejected": -5.29976749420166, "loss": 0.7846, "odds_ratio_loss": 0.008555177599191666, "rewards/accuracies": 1.0, "rewards/chosen": -0.0783730298280716, "rewards/margins": 0.45160380005836487, "rewards/rejected": -0.5299768447875977, "sft_loss": 0.7837303876876831, "step": 325 }, { "epoch": 1.0779910167415272, "grad_norm": 2.609851360321045, "learning_rate": 3.575891621242209e-05, "logits/chosen": -0.35802769660949707, "logits/rejected": -0.9912473559379578, "logps/chosen": -0.8010882139205933, "logps/rejected": -5.255009651184082, "loss": 0.8022, "odds_ratio_loss": 0.010739787481725216, "rewards/accuracies": 1.0, "rewards/chosen": -0.08010882139205933, "rewards/margins": 0.4453921914100647, "rewards/rejected": -0.525501012802124, "sft_loss": 0.8010882139205933, "step": 330 }, { "epoch": 1.0943242139648837, "grad_norm": 1.9164317846298218, "learning_rate": 3.537122351761568e-05, "logits/chosen": -0.29361772537231445, "logits/rejected": -0.9894407987594604, "logps/chosen": -0.8663752675056458, "logps/rejected": -5.1811041831970215, "loss": 0.8682, "odds_ratio_loss": 0.01779218763113022, "rewards/accuracies": 1.0, "rewards/chosen": -0.08663753420114517, "rewards/margins": 0.43147286772727966, "rewards/rejected": -0.5181103944778442, "sft_loss": 0.8663752675056458, "step": 335 }, { "epoch": 1.1106574111882401, "grad_norm": 1.8551751375198364, "learning_rate": 3.4980494319285775e-05, "logits/chosen": -0.2934134900569916, "logits/rejected": -0.9298968315124512, "logps/chosen": -0.8729950189590454, "logps/rejected": -5.402509689331055, "loss": 0.8739, "odds_ratio_loss": 0.009193124249577522, "rewards/accuracies": 1.0, "rewards/chosen": -0.08729950338602066, "rewards/margins": 0.45295149087905884, "rewards/rejected": -0.5402510166168213, "sft_loss": 0.8729950189590454, "step": 340 }, { "epoch": 1.1269906084115966, "grad_norm": 1.8746849298477173, "learning_rate": 3.45868430157563e-05, "logits/chosen": -0.27233654260635376, "logits/rejected": -0.9365426301956177, "logps/chosen": -0.8479824066162109, "logps/rejected": -5.225064754486084, "loss": 0.8493, "odds_ratio_loss": 0.013331398367881775, "rewards/accuracies": 1.0, "rewards/chosen": -0.08479823917150497, "rewards/margins": 0.43770831823349, "rewards/rejected": -0.522506594657898, "sft_loss": 0.8479824066162109, "step": 345 }, { "epoch": 1.143323805634953, "grad_norm": 1.7111650705337524, "learning_rate": 3.4190384860889824e-05, "logits/chosen": -0.366852343082428, "logits/rejected": -1.0381531715393066, "logps/chosen": -0.8182533383369446, "logps/rejected": -5.278301239013672, "loss": 0.8204, "odds_ratio_loss": 0.02150268480181694, "rewards/accuracies": 1.0, "rewards/chosen": -0.08182532340288162, "rewards/margins": 0.4460047781467438, "rewards/rejected": -0.5278300642967224, "sft_loss": 0.8182533383369446, "step": 350 }, { "epoch": 1.1596570028583095, "grad_norm": 1.9346224069595337, "learning_rate": 3.379123593034342e-05, "logits/chosen": -0.2924169898033142, "logits/rejected": -1.028192400932312, "logps/chosen": -0.9392467737197876, "logps/rejected": -5.247817516326904, "loss": 0.9409, "odds_ratio_loss": 0.0163104310631752, "rewards/accuracies": 1.0, "rewards/chosen": -0.09392467886209488, "rewards/margins": 0.4308570921421051, "rewards/rejected": -0.5247817635536194, "sft_loss": 0.9392467737197876, "step": 355 }, { "epoch": 1.175990200081666, "grad_norm": 1.9083410501480103, "learning_rate": 3.3389513087583754e-05, "logits/chosen": -0.1519567221403122, "logits/rejected": -1.0742218494415283, "logps/chosen": -0.8435864448547363, "logps/rejected": -4.644381523132324, "loss": 0.8465, "odds_ratio_loss": 0.028982138261198997, "rewards/accuracies": 1.0, "rewards/chosen": -0.08435864001512527, "rewards/margins": 0.3800795078277588, "rewards/rejected": -0.4644381105899811, "sft_loss": 0.8435864448547363, "step": 360 }, { "epoch": 1.1923233973050225, "grad_norm": 1.8590240478515625, "learning_rate": 3.2985333949671685e-05, "logits/chosen": -0.43951839208602905, "logits/rejected": -1.1476421356201172, "logps/chosen": -0.8573881387710571, "logps/rejected": -4.641860485076904, "loss": 0.8605, "odds_ratio_loss": 0.030726104974746704, "rewards/accuracies": 1.0, "rewards/chosen": -0.08573882281780243, "rewards/margins": 0.3784472346305847, "rewards/rejected": -0.4641861021518707, "sft_loss": 0.8573881387710571, "step": 365 }, { "epoch": 1.208656594528379, "grad_norm": 2.064635753631592, "learning_rate": 3.257881685282609e-05, "logits/chosen": -0.2586503028869629, "logits/rejected": -1.1886835098266602, "logps/chosen": -0.7855366468429565, "logps/rejected": -4.374009132385254, "loss": 0.7886, "odds_ratio_loss": 0.03049846924841404, "rewards/accuracies": 1.0, "rewards/chosen": -0.07855366170406342, "rewards/margins": 0.3588472604751587, "rewards/rejected": -0.4374009668827057, "sft_loss": 0.7855366468429565, "step": 370 }, { "epoch": 1.2249897917517354, "grad_norm": 1.986666202545166, "learning_rate": 3.217008081777726e-05, "logits/chosen": -0.3716577887535095, "logits/rejected": -1.2067687511444092, "logps/chosen": -0.8298207521438599, "logps/rejected": -4.59980583190918, "loss": 0.8315, "odds_ratio_loss": 0.016955239698290825, "rewards/accuracies": 1.0, "rewards/chosen": -0.08298207074403763, "rewards/margins": 0.3769984841346741, "rewards/rejected": -0.4599805772304535, "sft_loss": 0.8298207521438599, "step": 375 }, { "epoch": 1.2413229889750919, "grad_norm": 1.8263311386108398, "learning_rate": 3.175924551491977e-05, "logits/chosen": -0.3963642716407776, "logits/rejected": -1.1271188259124756, "logps/chosen": -0.8212121725082397, "logps/rejected": -4.6982741355896, "loss": 0.8234, "odds_ratio_loss": 0.02219826728105545, "rewards/accuracies": 1.0, "rewards/chosen": -0.08212121576070786, "rewards/margins": 0.3877061605453491, "rewards/rejected": -0.4698273539543152, "sft_loss": 0.8212121725082397, "step": 380 }, { "epoch": 1.2576561861984483, "grad_norm": 1.946442723274231, "learning_rate": 3.13464312292752e-05, "logits/chosen": -0.343479186296463, "logits/rejected": -0.9710591435432434, "logps/chosen": -0.8622458577156067, "logps/rejected": -5.370631217956543, "loss": 0.8646, "odds_ratio_loss": 0.023645440116524696, "rewards/accuracies": 1.0, "rewards/chosen": -0.08622457832098007, "rewards/margins": 0.450838565826416, "rewards/rejected": -0.5370631217956543, "sft_loss": 0.8622458577156067, "step": 385 }, { "epoch": 1.2739893834218048, "grad_norm": 1.7212151288986206, "learning_rate": 3.093175882527491e-05, "logits/chosen": -0.3060234785079956, "logits/rejected": -1.022698163986206, "logps/chosen": -0.8830709457397461, "logps/rejected": -5.35630464553833, "loss": 0.8848, "odds_ratio_loss": 0.01716751977801323, "rewards/accuracies": 1.0, "rewards/chosen": -0.08830709010362625, "rewards/margins": 0.44732338190078735, "rewards/rejected": -0.535630464553833, "sft_loss": 0.8830709457397461, "step": 390 }, { "epoch": 1.2903225806451613, "grad_norm": 2.029407024383545, "learning_rate": 3.0515349711373146e-05, "logits/chosen": -0.13591887056827545, "logits/rejected": -0.9809386134147644, "logps/chosen": -0.880872368812561, "logps/rejected": -5.28397798538208, "loss": 0.8826, "odds_ratio_loss": 0.01720813475549221, "rewards/accuracies": 1.0, "rewards/chosen": -0.08808723837137222, "rewards/margins": 0.4403105676174164, "rewards/rejected": -0.5283978581428528, "sft_loss": 0.880872368812561, "step": 395 }, { "epoch": 1.3066557778685177, "grad_norm": 1.8320516347885132, "learning_rate": 3.0097325804500864e-05, "logits/chosen": -0.3814406991004944, "logits/rejected": -1.0283949375152588, "logps/chosen": -0.7927986979484558, "logps/rejected": -5.100050926208496, "loss": 0.7945, "odds_ratio_loss": 0.016741547733545303, "rewards/accuracies": 1.0, "rewards/chosen": -0.07927986979484558, "rewards/margins": 0.4307252764701843, "rewards/rejected": -0.5100051164627075, "sft_loss": 0.7927986979484558, "step": 400 } ], "logging_steps": 5, "max_steps": 918, "num_input_tokens_seen": 0, "num_train_epochs": 3, "save_steps": 100, "total_flos": 1.663517229609124e+17, "train_batch_size": 1, "trial_name": null, "trial_params": null }