{ "best_metric": null, "best_model_checkpoint": null, "epoch": 3.0, "eval_steps": 500, "global_step": 1476, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.02, "grad_norm": 420.0, "learning_rate": 5.000000000000001e-07, "log_odds_chosen": 0.4207151532173157, "log_odds_ratio": -0.5201840400695801, "logits/chosen": -2.347763776779175, "logits/rejected": -2.3541953563690186, "logps/chosen": -1.7608455419540405, "logps/rejected": -2.1218907833099365, "loss": 3.9956, "nll_loss": 4.183717250823975, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": -0.08804227411746979, "rewards/margins": 0.018052268773317337, "rewards/rejected": -0.10609455406665802, "step": 10 }, { "epoch": 0.04, "grad_norm": 44.25, "learning_rate": 1.0000000000000002e-06, "log_odds_chosen": 0.5392915606498718, "log_odds_ratio": -0.4769781231880188, "logits/chosen": -2.866349458694458, "logits/rejected": -2.866671323776245, "logps/chosen": -1.5143439769744873, "logps/rejected": -1.9585195779800415, "loss": 2.0453, "nll_loss": 1.9479516744613647, "rewards/accuracies": 0.925000011920929, "rewards/chosen": -0.07571719586849213, "rewards/margins": 0.02220877632498741, "rewards/rejected": -0.09792597591876984, "step": 20 }, { "epoch": 0.06, "grad_norm": 7.5, "learning_rate": 1.5e-06, "log_odds_chosen": 0.9706680178642273, "log_odds_ratio": -0.3677331805229187, "logits/chosen": -2.6566712856292725, "logits/rejected": -2.660308361053467, "logps/chosen": -0.7488349676132202, "logps/rejected": -1.312819480895996, "loss": 1.0008, "nll_loss": 0.8921732902526855, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": -0.03744174912571907, "rewards/margins": 0.028199229389429092, "rewards/rejected": -0.06564097106456757, "step": 30 }, { "epoch": 0.08, "grad_norm": 7.40625, "learning_rate": 2.0000000000000003e-06, "log_odds_chosen": 2.282021999359131, "log_odds_ratio": -0.20370522141456604, "logits/chosen": -2.1984050273895264, "logits/rejected": -2.202855110168457, "logps/chosen": -0.27147307991981506, "logps/rejected": -1.2435765266418457, "loss": 0.8107, "nll_loss": 0.683063268661499, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.013573653995990753, "rewards/margins": 0.04860517010092735, "rewards/rejected": -0.06217882037162781, "step": 40 }, { "epoch": 0.1, "grad_norm": 6.375, "learning_rate": 2.5e-06, "log_odds_chosen": 2.8677659034729004, "log_odds_ratio": -0.14154179394245148, "logits/chosen": -2.279298782348633, "logits/rejected": -2.2862045764923096, "logps/chosen": -0.216520756483078, "logps/rejected": -1.3591465950012207, "loss": 0.7978, "nll_loss": 0.828228771686554, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.01082603819668293, "rewards/margins": 0.057131290435791016, "rewards/rejected": -0.0679573267698288, "step": 50 }, { "epoch": 0.12, "grad_norm": 4.3125, "learning_rate": 3e-06, "log_odds_chosen": 3.2677154541015625, "log_odds_ratio": -0.12433552742004395, "logits/chosen": -2.2281601428985596, "logits/rejected": -2.2347497940063477, "logps/chosen": -0.20724907517433167, "logps/rejected": -1.5352025032043457, "loss": 0.7919, "nll_loss": 0.8021117448806763, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": -0.010362453758716583, "rewards/margins": 0.06639767438173294, "rewards/rejected": -0.07676012814044952, "step": 60 }, { "epoch": 0.14, "grad_norm": 4.21875, "learning_rate": 3.5e-06, "log_odds_chosen": 3.9659018516540527, "log_odds_ratio": -0.11224009841680527, "logits/chosen": -2.0779483318328857, "logits/rejected": -2.0807433128356934, "logps/chosen": -0.1909479945898056, "logps/rejected": -1.9270519018173218, "loss": 0.7197, "nll_loss": 0.6386193037033081, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.00954739935696125, "rewards/margins": 0.08680519461631775, "rewards/rejected": -0.09635259956121445, "step": 70 }, { "epoch": 0.16, "grad_norm": 4.15625, "learning_rate": 4.000000000000001e-06, "log_odds_chosen": 3.9601542949676514, "log_odds_ratio": -0.11061631143093109, "logits/chosen": -2.1697168350219727, "logits/rejected": -2.1794517040252686, "logps/chosen": -0.185487762093544, "logps/rejected": -1.8811283111572266, "loss": 0.7033, "nll_loss": 0.7545000314712524, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.009274388663470745, "rewards/margins": 0.08478203415870667, "rewards/rejected": -0.09405642002820969, "step": 80 }, { "epoch": 0.18, "grad_norm": 4.71875, "learning_rate": 4.5e-06, "log_odds_chosen": 4.599364280700684, "log_odds_ratio": -0.07719282805919647, "logits/chosen": -2.150642156600952, "logits/rejected": -2.15342378616333, "logps/chosen": -0.15149493515491486, "logps/rejected": -2.1990609169006348, "loss": 0.6116, "nll_loss": 0.6107597947120667, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.007574746850878, "rewards/margins": 0.10237829387187958, "rewards/rejected": -0.10995303094387054, "step": 90 }, { "epoch": 0.2, "grad_norm": 4.53125, "learning_rate": 5e-06, "log_odds_chosen": 4.817965507507324, "log_odds_ratio": -0.07800901681184769, "logits/chosen": -2.1689393520355225, "logits/rejected": -2.1756036281585693, "logps/chosen": -0.1730736345052719, "logps/rejected": -2.246196746826172, "loss": 0.6043, "nll_loss": 0.6302344799041748, "rewards/accuracies": 1.0, "rewards/chosen": -0.008653681725263596, "rewards/margins": 0.10365615785121918, "rewards/rejected": -0.11230983585119247, "step": 100 }, { "epoch": 0.22, "grad_norm": 5.625, "learning_rate": 4.767312946227961e-06, "log_odds_chosen": 4.97894287109375, "log_odds_ratio": -0.10877744108438492, "logits/chosen": -2.2094526290893555, "logits/rejected": -2.216020107269287, "logps/chosen": -0.18459849059581757, "logps/rejected": -2.573103904724121, "loss": 0.5724, "nll_loss": 0.6197156310081482, "rewards/accuracies": 0.9375, "rewards/chosen": -0.009229923598468304, "rewards/margins": 0.11942527443170547, "rewards/rejected": -0.12865519523620605, "step": 110 }, { "epoch": 0.24, "grad_norm": 5.8125, "learning_rate": 4.564354645876385e-06, "log_odds_chosen": 5.132683277130127, "log_odds_ratio": -0.13557349145412445, "logits/chosen": -2.178772449493408, "logits/rejected": -2.184910774230957, "logps/chosen": -0.154138445854187, "logps/rejected": -2.2600817680358887, "loss": 0.4748, "nll_loss": 0.459114134311676, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.0077069224789738655, "rewards/margins": 0.10529716312885284, "rewards/rejected": -0.11300408840179443, "step": 120 }, { "epoch": 0.26, "grad_norm": 5.0, "learning_rate": 4.385290096535147e-06, "log_odds_chosen": 5.768979072570801, "log_odds_ratio": -0.08052898198366165, "logits/chosen": -2.101267099380493, "logits/rejected": -2.104369878768921, "logps/chosen": -0.1255403310060501, "logps/rejected": -2.850447416305542, "loss": 0.4379, "nll_loss": 0.40499457716941833, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.006277016364037991, "rewards/margins": 0.13624534010887146, "rewards/rejected": -0.14252236485481262, "step": 130 }, { "epoch": 0.28, "grad_norm": 6.0, "learning_rate": 4.2257712736425835e-06, "log_odds_chosen": 5.547473907470703, "log_odds_ratio": -0.056151650846004486, "logits/chosen": -2.0569746494293213, "logits/rejected": -2.059227466583252, "logps/chosen": -0.12779685854911804, "logps/rejected": -2.394996166229248, "loss": 0.4122, "nll_loss": 0.33014988899230957, "rewards/accuracies": 1.0, "rewards/chosen": -0.006389842834323645, "rewards/margins": 0.1133599653840065, "rewards/rejected": -0.11974982172250748, "step": 140 }, { "epoch": 0.3, "grad_norm": 5.21875, "learning_rate": 4.082482904638631e-06, "log_odds_chosen": 5.67286491394043, "log_odds_ratio": -0.06435154378414154, "logits/chosen": -2.166313886642456, "logits/rejected": -2.1680521965026855, "logps/chosen": -0.13122043013572693, "logps/rejected": -2.701577663421631, "loss": 0.4055, "nll_loss": 0.4321494996547699, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.0065610213205218315, "rewards/margins": 0.12851786613464355, "rewards/rejected": -0.13507887721061707, "step": 150 }, { "epoch": 0.33, "grad_norm": 6.5625, "learning_rate": 3.952847075210474e-06, "log_odds_chosen": 5.581357002258301, "log_odds_ratio": -0.094394750893116, "logits/chosen": -2.121497392654419, "logits/rejected": -2.1272549629211426, "logps/chosen": -0.15410242974758148, "logps/rejected": -2.376707077026367, "loss": 0.3695, "nll_loss": 0.43908724188804626, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.007705122232437134, "rewards/margins": 0.1111302375793457, "rewards/rejected": -0.11883536726236343, "step": 160 }, { "epoch": 0.35, "grad_norm": 6.0, "learning_rate": 3.834824944236852e-06, "log_odds_chosen": 6.281979560852051, "log_odds_ratio": -0.059505023062229156, "logits/chosen": -2.1019725799560547, "logits/rejected": -2.1044390201568604, "logps/chosen": -0.15183117985725403, "logps/rejected": -3.044116497039795, "loss": 0.3264, "nll_loss": 0.3386436998844147, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.007591559551656246, "rewards/margins": 0.14461426436901093, "rewards/rejected": -0.15220583975315094, "step": 170 }, { "epoch": 0.37, "grad_norm": 6.3125, "learning_rate": 3.72677996249965e-06, "log_odds_chosen": 6.111880302429199, "log_odds_ratio": -0.07030317932367325, "logits/chosen": -2.0706937313079834, "logits/rejected": -2.067783832550049, "logps/chosen": -0.14989063143730164, "logps/rejected": -3.172281265258789, "loss": 0.3324, "nll_loss": 0.38789159059524536, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.007494532503187656, "rewards/margins": 0.15111951529979706, "rewards/rejected": -0.15861406922340393, "step": 180 }, { "epoch": 0.39, "grad_norm": 6.0625, "learning_rate": 3.6273812505500587e-06, "log_odds_chosen": 6.445823669433594, "log_odds_ratio": -0.05266156792640686, "logits/chosen": -2.0448713302612305, "logits/rejected": -2.0481276512145996, "logps/chosen": -0.13432614505290985, "logps/rejected": -2.9074063301086426, "loss": 0.3063, "nll_loss": 0.28705888986587524, "rewards/accuracies": 1.0, "rewards/chosen": -0.006716308183968067, "rewards/margins": 0.13865402340888977, "rewards/rejected": -0.14537033438682556, "step": 190 }, { "epoch": 0.41, "grad_norm": 6.0, "learning_rate": 3.5355339059327378e-06, "log_odds_chosen": 5.775099754333496, "log_odds_ratio": -0.08722616732120514, "logits/chosen": -2.087759256362915, "logits/rejected": -2.0876097679138184, "logps/chosen": -0.11597369611263275, "logps/rejected": -2.6853129863739014, "loss": 0.2962, "nll_loss": 0.31049779057502747, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.005798685364425182, "rewards/margins": 0.12846694886684418, "rewards/rejected": -0.13426566123962402, "step": 200 }, { "epoch": 0.43, "grad_norm": 5.90625, "learning_rate": 3.450327796711771e-06, "log_odds_chosen": 7.5723066329956055, "log_odds_ratio": -0.040629759430885315, "logits/chosen": -1.973818063735962, "logits/rejected": -1.976941704750061, "logps/chosen": -0.09340523928403854, "logps/rejected": -3.5397675037384033, "loss": 0.2637, "nll_loss": 0.20889191329479218, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.00467026187106967, "rewards/margins": 0.17231810092926025, "rewards/rejected": -0.1769883632659912, "step": 210 }, { "epoch": 0.45, "grad_norm": 6.25, "learning_rate": 3.3709993123162106e-06, "log_odds_chosen": 7.618361473083496, "log_odds_ratio": -0.074308380484581, "logits/chosen": -2.0433554649353027, "logits/rejected": -2.0466837882995605, "logps/chosen": -0.09071488678455353, "logps/rejected": -3.5269036293029785, "loss": 0.2727, "nll_loss": 0.24802179634571075, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": -0.004535744432359934, "rewards/margins": 0.17180944979190826, "rewards/rejected": -0.17634519934654236, "step": 220 }, { "epoch": 0.47, "grad_norm": 5.875, "learning_rate": 3.296902366978936e-06, "log_odds_chosen": 6.698166847229004, "log_odds_ratio": -0.04654809832572937, "logits/chosen": -1.968505620956421, "logits/rejected": -1.9712941646575928, "logps/chosen": -0.1543513834476471, "logps/rejected": -3.186583995819092, "loss": 0.2015, "nll_loss": 0.1845930814743042, "rewards/accuracies": 1.0, "rewards/chosen": -0.00771756935864687, "rewards/margins": 0.15161164104938507, "rewards/rejected": -0.15932922065258026, "step": 230 }, { "epoch": 0.49, "grad_norm": 5.28125, "learning_rate": 3.2274861218395142e-06, "log_odds_chosen": 7.480630397796631, "log_odds_ratio": -0.06602514535188675, "logits/chosen": -1.995495080947876, "logits/rejected": -1.9951270818710327, "logps/chosen": -0.12399391829967499, "logps/rejected": -3.7262351512908936, "loss": 0.2181, "nll_loss": 0.1835182011127472, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.006199696101248264, "rewards/margins": 0.1801120638847351, "rewards/rejected": -0.1863117516040802, "step": 240 }, { "epoch": 0.51, "grad_norm": 6.65625, "learning_rate": 3.1622776601683796e-06, "log_odds_chosen": 7.167886257171631, "log_odds_ratio": -0.020891310647130013, "logits/chosen": -1.9453086853027344, "logits/rejected": -1.9443309307098389, "logps/chosen": -0.07993815094232559, "logps/rejected": -3.2065398693084717, "loss": 0.1966, "nll_loss": 0.21496787667274475, "rewards/accuracies": 1.0, "rewards/chosen": -0.003996907267719507, "rewards/margins": 0.15633007884025574, "rewards/rejected": -0.1603269875049591, "step": 250 }, { "epoch": 0.53, "grad_norm": 7.03125, "learning_rate": 3.1008683647302113e-06, "log_odds_chosen": 6.189971446990967, "log_odds_ratio": -0.06337615847587585, "logits/chosen": -1.983068823814392, "logits/rejected": -1.9870706796646118, "logps/chosen": -0.13599030673503876, "logps/rejected": -2.809053897857666, "loss": 0.1981, "nll_loss": 0.23529119789600372, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.0067995162680745125, "rewards/margins": 0.1336531937122345, "rewards/rejected": -0.14045271277427673, "step": 260 }, { "epoch": 0.55, "grad_norm": 5.125, "learning_rate": 3.0429030972509227e-06, "log_odds_chosen": 6.417479038238525, "log_odds_ratio": -0.05179642513394356, "logits/chosen": -1.9953663349151611, "logits/rejected": -1.9997583627700806, "logps/chosen": -0.12870252132415771, "logps/rejected": -2.9075326919555664, "loss": 0.1753, "nll_loss": 0.22951626777648926, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.006435126066207886, "rewards/margins": 0.13894154131412506, "rewards/rejected": -0.14537665247917175, "step": 270 }, { "epoch": 0.57, "grad_norm": 5.90625, "learning_rate": 2.988071523335984e-06, "log_odds_chosen": 6.808651924133301, "log_odds_ratio": -0.06405957788228989, "logits/chosen": -2.0137691497802734, "logits/rejected": -2.015270709991455, "logps/chosen": -0.11106196790933609, "logps/rejected": -3.0401816368103027, "loss": 0.1885, "nll_loss": 0.1923011988401413, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.005553097929805517, "rewards/margins": 0.14645597338676453, "rewards/rejected": -0.15200909972190857, "step": 280 }, { "epoch": 0.59, "grad_norm": 5.375, "learning_rate": 2.9361010975735177e-06, "log_odds_chosen": 8.074934959411621, "log_odds_ratio": -0.022188549861311913, "logits/chosen": -1.94329833984375, "logits/rejected": -1.943985939025879, "logps/chosen": -0.06887535750865936, "logps/rejected": -3.7694950103759766, "loss": 0.1729, "nll_loss": 0.15832534432411194, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.0034437670838087797, "rewards/margins": 0.185030996799469, "rewards/rejected": -0.18847474455833435, "step": 290 }, { "epoch": 0.61, "grad_norm": 4.875, "learning_rate": 2.8867513459481293e-06, "log_odds_chosen": 6.823982238769531, "log_odds_ratio": -0.04056983441114426, "logits/chosen": -1.9236243963241577, "logits/rejected": -1.9263416528701782, "logps/chosen": -0.15296845138072968, "logps/rejected": -3.20212984085083, "loss": 0.1766, "nll_loss": 0.1584191620349884, "rewards/accuracies": 1.0, "rewards/chosen": -0.0076484219171106815, "rewards/margins": 0.152458056807518, "rewards/rejected": -0.16010649502277374, "step": 300 }, { "epoch": 0.63, "grad_norm": 5.0, "learning_rate": 2.839809171235324e-06, "log_odds_chosen": 7.289086818695068, "log_odds_ratio": -0.07533205300569534, "logits/chosen": -2.0198066234588623, "logits/rejected": -2.022125720977783, "logps/chosen": -0.12243582308292389, "logps/rejected": -3.1500906944274902, "loss": 0.1636, "nll_loss": 0.1975417137145996, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": -0.0061217909678816795, "rewards/margins": 0.15138274431228638, "rewards/rejected": -0.15750452876091003, "step": 310 }, { "epoch": 0.65, "grad_norm": 9.0, "learning_rate": 2.7950849718747376e-06, "log_odds_chosen": 6.9169487953186035, "log_odds_ratio": -0.03769087791442871, "logits/chosen": -1.9343087673187256, "logits/rejected": -1.9343109130859375, "logps/chosen": -0.10367435216903687, "logps/rejected": -3.0478665828704834, "loss": 0.1696, "nll_loss": 0.15094847977161407, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.005183717701584101, "rewards/margins": 0.14720964431762695, "rewards/rejected": -0.15239334106445312, "step": 320 }, { "epoch": 0.67, "grad_norm": 5.90625, "learning_rate": 2.752409412815902e-06, "log_odds_chosen": 6.86618185043335, "log_odds_ratio": -0.038672395050525665, "logits/chosen": -2.047116279602051, "logits/rejected": -2.044132947921753, "logps/chosen": -0.13211274147033691, "logps/rejected": -3.158360004425049, "loss": 0.1625, "nll_loss": 0.17262741923332214, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.006605637725442648, "rewards/margins": 0.15131238102912903, "rewards/rejected": -0.15791800618171692, "step": 330 }, { "epoch": 0.69, "grad_norm": 5.40625, "learning_rate": 2.711630722733202e-06, "log_odds_chosen": 7.859139442443848, "log_odds_ratio": -0.05241536349058151, "logits/chosen": -1.951807975769043, "logits/rejected": -1.9466063976287842, "logps/chosen": -0.08200397342443466, "logps/rejected": -3.5420470237731934, "loss": 0.1465, "nll_loss": 0.1444871723651886, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.004100198391824961, "rewards/margins": 0.17300216853618622, "rewards/rejected": -0.17710235714912415, "step": 340 }, { "epoch": 0.71, "grad_norm": 5.71875, "learning_rate": 2.6726124191242444e-06, "log_odds_chosen": 7.041175842285156, "log_odds_ratio": -0.03843230381608009, "logits/chosen": -1.9341367483139038, "logits/rejected": -1.9377237558364868, "logps/chosen": -0.105894073843956, "logps/rejected": -2.9890971183776855, "loss": 0.145, "nll_loss": 0.15412884950637817, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.005294703878462315, "rewards/margins": 0.14416013658046722, "rewards/rejected": -0.14945486187934875, "step": 350 }, { "epoch": 0.73, "grad_norm": 6.90625, "learning_rate": 2.6352313834736496e-06, "log_odds_chosen": 6.8190507888793945, "log_odds_ratio": -0.04403644800186157, "logits/chosen": -1.984197974205017, "logits/rejected": -1.987479567527771, "logps/chosen": -0.11883167922496796, "logps/rejected": -3.1214818954467773, "loss": 0.1577, "nll_loss": 0.12197007238864899, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.0059415847063064575, "rewards/margins": 0.15013250708580017, "rewards/rejected": -0.15607409179210663, "step": 360 }, { "epoch": 0.75, "grad_norm": 5.28125, "learning_rate": 2.599376224550182e-06, "log_odds_chosen": 6.533886909484863, "log_odds_ratio": -0.03479010984301567, "logits/chosen": -1.898244857788086, "logits/rejected": -1.8991117477416992, "logps/chosen": -0.10942695289850235, "logps/rejected": -2.946617841720581, "loss": 0.1199, "nll_loss": 0.11430631577968597, "rewards/accuracies": 1.0, "rewards/chosen": -0.005471347831189632, "rewards/margins": 0.14185954630374908, "rewards/rejected": -0.14733090996742249, "step": 370 }, { "epoch": 0.77, "grad_norm": 5.6875, "learning_rate": 2.564945880212886e-06, "log_odds_chosen": 6.550747871398926, "log_odds_ratio": -0.0334211066365242, "logits/chosen": -1.9049135446548462, "logits/rejected": -1.9040100574493408, "logps/chosen": -0.12769648432731628, "logps/rejected": -3.1679463386535645, "loss": 0.1321, "nll_loss": 0.12015843391418457, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.006384824402630329, "rewards/margins": 0.15201251208782196, "rewards/rejected": -0.15839733183383942, "step": 380 }, { "epoch": 0.79, "grad_norm": 4.46875, "learning_rate": 2.5318484177091667e-06, "log_odds_chosen": 6.963167667388916, "log_odds_ratio": -0.07767531275749207, "logits/chosen": -1.9564456939697266, "logits/rejected": -1.957360029220581, "logps/chosen": -0.09807981550693512, "logps/rejected": -2.8958752155303955, "loss": 0.1334, "nll_loss": 0.13979199528694153, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.004903990775346756, "rewards/margins": 0.13988977670669556, "rewards/rejected": -0.1447937786579132, "step": 390 }, { "epoch": 0.81, "grad_norm": 6.46875, "learning_rate": 2.5e-06, "log_odds_chosen": 7.737925052642822, "log_odds_ratio": -0.011308287270367146, "logits/chosen": -1.8833544254302979, "logits/rejected": -1.8816306591033936, "logps/chosen": -0.07808590680360794, "logps/rejected": -3.517468214035034, "loss": 0.116, "nll_loss": 0.09308944642543793, "rewards/accuracies": 1.0, "rewards/chosen": -0.0039042953867465258, "rewards/margins": 0.17196913063526154, "rewards/rejected": -0.17587339878082275, "step": 400 }, { "epoch": 0.83, "grad_norm": 5.09375, "learning_rate": 2.4693239916239746e-06, "log_odds_chosen": 8.026070594787598, "log_odds_ratio": -0.009079935029149055, "logits/chosen": -1.8916471004486084, "logits/rejected": -1.8905131816864014, "logps/chosen": -0.08094353973865509, "logps/rejected": -3.705895185470581, "loss": 0.1292, "nll_loss": 0.10612641274929047, "rewards/accuracies": 1.0, "rewards/chosen": -0.00404717680066824, "rewards/margins": 0.18124757707118988, "rewards/rejected": -0.1852947473526001, "step": 410 }, { "epoch": 0.85, "grad_norm": 4.15625, "learning_rate": 2.4397501823713327e-06, "log_odds_chosen": 6.706048488616943, "log_odds_ratio": -0.04705270379781723, "logits/chosen": -1.9574733972549438, "logits/rejected": -1.9553906917572021, "logps/chosen": -0.13810355961322784, "logps/rejected": -3.244007110595703, "loss": 0.1116, "nll_loss": 0.13714413344860077, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.00690517807379365, "rewards/margins": 0.15529517829418182, "rewards/rejected": -0.16220036149024963, "step": 420 }, { "epoch": 0.87, "grad_norm": 4.9375, "learning_rate": 2.411214110852061e-06, "log_odds_chosen": 7.185335636138916, "log_odds_ratio": -0.049024853855371475, "logits/chosen": -1.9474496841430664, "logits/rejected": -1.94386887550354, "logps/chosen": -0.12190796434879303, "logps/rejected": -3.319507122039795, "loss": 0.1158, "nll_loss": 0.13795073330402374, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.0060953982174396515, "rewards/margins": 0.15987998247146606, "rewards/rejected": -0.16597537696361542, "step": 430 }, { "epoch": 0.89, "grad_norm": 4.8125, "learning_rate": 2.3836564731139807e-06, "log_odds_chosen": 7.590296268463135, "log_odds_ratio": -0.021134180948138237, "logits/chosen": -1.8886020183563232, "logits/rejected": -1.886425256729126, "logps/chosen": -0.09185922890901566, "logps/rejected": -3.507038116455078, "loss": 0.1033, "nll_loss": 0.06822694092988968, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.00459296116605401, "rewards/margins": 0.1707589477300644, "rewards/rejected": -0.17535191774368286, "step": 440 }, { "epoch": 0.91, "grad_norm": 6.28125, "learning_rate": 2.357022603955159e-06, "log_odds_chosen": 6.999331474304199, "log_odds_ratio": -0.03660160303115845, "logits/chosen": -1.861541748046875, "logits/rejected": -1.861534833908081, "logps/chosen": -0.12671752274036407, "logps/rejected": -3.262852430343628, "loss": 0.1048, "nll_loss": 0.09448351711034775, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.0063358754850924015, "rewards/margins": 0.15680675208568573, "rewards/rejected": -0.1631426364183426, "step": 450 }, { "epoch": 0.93, "grad_norm": 4.78125, "learning_rate": 2.3312620206007847e-06, "log_odds_chosen": 6.333981037139893, "log_odds_ratio": -0.049268852919340134, "logits/chosen": -1.8751580715179443, "logits/rejected": -1.8740007877349854, "logps/chosen": -0.13210389018058777, "logps/rejected": -2.7039647102355957, "loss": 0.1094, "nll_loss": 0.09672755748033524, "rewards/accuracies": 1.0, "rewards/chosen": -0.006605193950235844, "rewards/margins": 0.12859304249286652, "rewards/rejected": -0.13519823551177979, "step": 460 }, { "epoch": 0.96, "grad_norm": 6.65625, "learning_rate": 2.3063280200722128e-06, "log_odds_chosen": 7.438237190246582, "log_odds_ratio": -0.02831345424056053, "logits/chosen": -1.9090907573699951, "logits/rejected": -1.9076954126358032, "logps/chosen": -0.07157810777425766, "logps/rejected": -3.022064447402954, "loss": 0.0996, "nll_loss": 0.09075023233890533, "rewards/accuracies": 1.0, "rewards/chosen": -0.003578905714675784, "rewards/margins": 0.14752431213855743, "rewards/rejected": -0.151103213429451, "step": 470 }, { "epoch": 0.98, "grad_norm": 5.46875, "learning_rate": 2.2821773229381924e-06, "log_odds_chosen": 8.533849716186523, "log_odds_ratio": -0.010285028256475925, "logits/chosen": -1.958752989768982, "logits/rejected": -1.961133599281311, "logps/chosen": -0.07796324789524078, "logps/rejected": -3.159803628921509, "loss": 0.1074, "nll_loss": 0.10170219838619232, "rewards/accuracies": 1.0, "rewards/chosen": -0.0038981616962701082, "rewards/margins": 0.15409204363822937, "rewards/rejected": -0.15799018740653992, "step": 480 }, { "epoch": 1.0, "grad_norm": 4.40625, "learning_rate": 2.2587697572631284e-06, "log_odds_chosen": 7.246984958648682, "log_odds_ratio": -0.03945430740714073, "logits/chosen": -1.9564424753189087, "logits/rejected": -1.9600261449813843, "logps/chosen": -0.1045081838965416, "logps/rejected": -3.074450731277466, "loss": 0.1118, "nll_loss": 0.11936734616756439, "rewards/accuracies": 1.0, "rewards/chosen": -0.00522540882229805, "rewards/margins": 0.14849711954593658, "rewards/rejected": -0.15372253954410553, "step": 490 }, { "epoch": 1.02, "grad_norm": 4.78125, "learning_rate": 2.23606797749979e-06, "log_odds_chosen": 7.313413143157959, "log_odds_ratio": -0.06336325407028198, "logits/chosen": -1.925005316734314, "logits/rejected": -1.9246110916137695, "logps/chosen": -0.06910906732082367, "logps/rejected": -3.088639974594116, "loss": 0.0748, "nll_loss": 0.06068875640630722, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.0034554533194750547, "rewards/margins": 0.15097656846046448, "rewards/rejected": -0.1544320285320282, "step": 500 }, { "epoch": 1.04, "grad_norm": 3.984375, "learning_rate": 2.2140372138502386e-06, "log_odds_chosen": 8.353876113891602, "log_odds_ratio": -0.011013227514922619, "logits/chosen": -1.9190175533294678, "logits/rejected": -1.9186252355575562, "logps/chosen": -0.060610998421907425, "logps/rejected": -3.5158791542053223, "loss": 0.073, "nll_loss": 0.0933486819267273, "rewards/accuracies": 1.0, "rewards/chosen": -0.0030305504333227873, "rewards/margins": 0.1727634072303772, "rewards/rejected": -0.17579396069049835, "step": 510 }, { "epoch": 1.06, "grad_norm": 4.6875, "learning_rate": 2.1926450482675734e-06, "log_odds_chosen": 8.398624420166016, "log_odds_ratio": -0.01034296303987503, "logits/chosen": -1.9005162715911865, "logits/rejected": -1.8965606689453125, "logps/chosen": -0.08098746836185455, "logps/rejected": -3.6455941200256348, "loss": 0.0753, "nll_loss": 0.0635688304901123, "rewards/accuracies": 1.0, "rewards/chosen": -0.00404937332496047, "rewards/margins": 0.17823031544685364, "rewards/rejected": -0.18227970600128174, "step": 520 }, { "epoch": 1.08, "grad_norm": 5.4375, "learning_rate": 2.1718612138153473e-06, "log_odds_chosen": 8.60987663269043, "log_odds_ratio": -0.00940676499158144, "logits/chosen": -1.8967382907867432, "logits/rejected": -1.8937251567840576, "logps/chosen": -0.07309839874505997, "logps/rejected": -3.786154270172119, "loss": 0.0827, "nll_loss": 0.07813505828380585, "rewards/accuracies": 1.0, "rewards/chosen": -0.003654920030385256, "rewards/margins": 0.18565279245376587, "rewards/rejected": -0.18930771946907043, "step": 530 }, { "epoch": 1.1, "grad_norm": 4.34375, "learning_rate": 2.151657414559676e-06, "log_odds_chosen": 9.486946105957031, "log_odds_ratio": -0.006830220576375723, "logits/chosen": -1.8446986675262451, "logits/rejected": -1.839769721031189, "logps/chosen": -0.045620597898960114, "logps/rejected": -4.278280258178711, "loss": 0.0759, "nll_loss": 0.061400435864925385, "rewards/accuracies": 1.0, "rewards/chosen": -0.002281030174344778, "rewards/margins": 0.21163301169872284, "rewards/rejected": -0.21391403675079346, "step": 540 }, { "epoch": 1.12, "grad_norm": 3.328125, "learning_rate": 2.132007163556104e-06, "log_odds_chosen": 8.17650032043457, "log_odds_ratio": -0.015669699758291245, "logits/chosen": -1.83901047706604, "logits/rejected": -1.835837960243225, "logps/chosen": -0.08997470885515213, "logps/rejected": -3.697235584259033, "loss": 0.0766, "nll_loss": 0.05432655289769173, "rewards/accuracies": 1.0, "rewards/chosen": -0.004498735070228577, "rewards/margins": 0.1803630292415619, "rewards/rejected": -0.18486176431179047, "step": 550 }, { "epoch": 1.14, "grad_norm": 4.75, "learning_rate": 2.1128856368212917e-06, "log_odds_chosen": 7.758215427398682, "log_odds_ratio": -0.027782147750258446, "logits/chosen": -1.887261152267456, "logits/rejected": -1.8827438354492188, "logps/chosen": -0.14876721799373627, "logps/rejected": -3.6743240356445312, "loss": 0.0828, "nll_loss": 0.11095689237117767, "rewards/accuracies": 1.0, "rewards/chosen": -0.007438362576067448, "rewards/margins": 0.17627784609794617, "rewards/rejected": -0.18371620774269104, "step": 560 }, { "epoch": 1.16, "grad_norm": 4.5625, "learning_rate": 2.0942695414584777e-06, "log_odds_chosen": 8.34414291381836, "log_odds_ratio": -0.021350722759962082, "logits/chosen": -1.8923943042755127, "logits/rejected": -1.8904056549072266, "logps/chosen": -0.09869461506605148, "logps/rejected": -3.8127264976501465, "loss": 0.0633, "nll_loss": 0.06047019362449646, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.004934730939567089, "rewards/margins": 0.18570157885551453, "rewards/rejected": -0.19063633680343628, "step": 570 }, { "epoch": 1.18, "grad_norm": 4.09375, "learning_rate": 2.0761369963434992e-06, "log_odds_chosen": 8.360025405883789, "log_odds_ratio": -0.022461308166384697, "logits/chosen": -1.9567257165908813, "logits/rejected": -1.9553306102752686, "logps/chosen": -0.057451434433460236, "logps/rejected": -3.338317394256592, "loss": 0.0647, "nll_loss": 0.09119249880313873, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.0028725718148052692, "rewards/margins": 0.1640433371067047, "rewards/rejected": -0.1669158935546875, "step": 580 }, { "epoch": 1.2, "grad_norm": 4.53125, "learning_rate": 2.058467423981546e-06, "log_odds_chosen": 8.788118362426758, "log_odds_ratio": -0.014825952239334583, "logits/chosen": -2.0043745040893555, "logits/rejected": -2.00369930267334, "logps/chosen": -0.06359820067882538, "logps/rejected": -3.9866626262664795, "loss": 0.0737, "nll_loss": 0.10766579210758209, "rewards/accuracies": 1.0, "rewards/chosen": -0.00317991035990417, "rewards/margins": 0.19615323841571808, "rewards/rejected": -0.19933313131332397, "step": 590 }, { "epoch": 1.22, "grad_norm": 4.53125, "learning_rate": 2.0412414523193154e-06, "log_odds_chosen": 8.787096977233887, "log_odds_ratio": -0.02094932086765766, "logits/chosen": -2.0005838871002197, "logits/rejected": -1.9980299472808838, "logps/chosen": -0.06464333832263947, "logps/rejected": -3.8292534351348877, "loss": 0.0665, "nll_loss": 0.07510735094547272, "rewards/accuracies": 1.0, "rewards/chosen": -0.0032321668695658445, "rewards/margins": 0.1882304847240448, "rewards/rejected": -0.1914626806974411, "step": 600 }, { "epoch": 1.24, "grad_norm": 5.0, "learning_rate": 2.0244408254472904e-06, "log_odds_chosen": 8.794069290161133, "log_odds_ratio": -0.0068406336940824986, "logits/chosen": -1.9200255870819092, "logits/rejected": -1.9168468713760376, "logps/chosen": -0.07637156546115875, "logps/rejected": -4.043463230133057, "loss": 0.0774, "nll_loss": 0.06419046223163605, "rewards/accuracies": 1.0, "rewards/chosen": -0.00381857855245471, "rewards/margins": 0.1983545869588852, "rewards/rejected": -0.2021731585264206, "step": 610 }, { "epoch": 1.26, "grad_norm": 3.359375, "learning_rate": 2.0080483222562476e-06, "log_odds_chosen": 9.725520133972168, "log_odds_ratio": -0.010235416702926159, "logits/chosen": -1.8615529537200928, "logits/rejected": -1.8584880828857422, "logps/chosen": -0.07194074988365173, "logps/rejected": -4.24954891204834, "loss": 0.0592, "nll_loss": 0.045302603393793106, "rewards/accuracies": 1.0, "rewards/chosen": -0.0035970378667116165, "rewards/margins": 0.20888039469718933, "rewards/rejected": -0.212477445602417, "step": 620 }, { "epoch": 1.28, "grad_norm": 3.703125, "learning_rate": 1.9920476822239895e-06, "log_odds_chosen": 8.134096145629883, "log_odds_ratio": -0.02648916281759739, "logits/chosen": -1.987565279006958, "logits/rejected": -1.985645055770874, "logps/chosen": -0.07914531975984573, "logps/rejected": -3.5829226970672607, "loss": 0.076, "nll_loss": 0.06963127106428146, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.003957265987992287, "rewards/margins": 0.17518886923789978, "rewards/rejected": -0.17914614081382751, "step": 630 }, { "epoch": 1.3, "grad_norm": 4.65625, "learning_rate": 1.976423537605237e-06, "log_odds_chosen": 8.905769348144531, "log_odds_ratio": -0.0063996403478085995, "logits/chosen": -1.8973668813705444, "logits/rejected": -1.895581841468811, "logps/chosen": -0.04883747547864914, "logps/rejected": -3.7183055877685547, "loss": 0.0691, "nll_loss": 0.060373760759830475, "rewards/accuracies": 1.0, "rewards/chosen": -0.0024418740067631006, "rewards/margins": 0.1834733784198761, "rewards/rejected": -0.1859152615070343, "step": 640 }, { "epoch": 1.32, "grad_norm": 3.671875, "learning_rate": 1.961161351381841e-06, "log_odds_chosen": 8.213096618652344, "log_odds_ratio": -0.012629789300262928, "logits/chosen": -1.8932087421417236, "logits/rejected": -1.8902513980865479, "logps/chosen": -0.07331077754497528, "logps/rejected": -3.4578137397766113, "loss": 0.0703, "nll_loss": 0.060921620577573776, "rewards/accuracies": 1.0, "rewards/chosen": -0.0036655396688729525, "rewards/margins": 0.16922515630722046, "rewards/rejected": -0.17289067804813385, "step": 650 }, { "epoch": 1.34, "grad_norm": 4.1875, "learning_rate": 1.9462473604038077e-06, "log_odds_chosen": 8.2732572555542, "log_odds_ratio": -0.015530670061707497, "logits/chosen": -1.9228805303573608, "logits/rejected": -1.9188134670257568, "logps/chosen": -0.08276908099651337, "logps/rejected": -3.572401523590088, "loss": 0.0672, "nll_loss": 0.03623514249920845, "rewards/accuracies": 1.0, "rewards/chosen": -0.004138454794883728, "rewards/margins": 0.1744816154241562, "rewards/rejected": -0.17862007021903992, "step": 660 }, { "epoch": 1.36, "grad_norm": 2.953125, "learning_rate": 1.9316685232156397e-06, "log_odds_chosen": 8.805624961853027, "log_odds_ratio": -0.0074573298916220665, "logits/chosen": -1.8942184448242188, "logits/rejected": -1.8891878128051758, "logps/chosen": -0.07351969927549362, "logps/rejected": -3.884014844894409, "loss": 0.0656, "nll_loss": 0.05570251867175102, "rewards/accuracies": 1.0, "rewards/chosen": -0.0036759853828698397, "rewards/margins": 0.19052477180957794, "rewards/rejected": -0.1942007690668106, "step": 670 }, { "epoch": 1.38, "grad_norm": 3.328125, "learning_rate": 1.917412472118426e-06, "log_odds_chosen": 8.748679161071777, "log_odds_ratio": -0.024869108572602272, "logits/chosen": -1.9717075824737549, "logits/rejected": -1.9675525426864624, "logps/chosen": -0.06451079994440079, "logps/rejected": -3.9762165546417236, "loss": 0.073, "nll_loss": 0.11092261970043182, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.003225540043786168, "rewards/margins": 0.19558526575565338, "rewards/rejected": -0.19881080090999603, "step": 680 }, { "epoch": 1.4, "grad_norm": 4.90625, "learning_rate": 1.9034674690672024e-06, "log_odds_chosen": 8.910208702087402, "log_odds_ratio": -0.005920908413827419, "logits/chosen": -1.8858320713043213, "logits/rejected": -1.8857864141464233, "logps/chosen": -0.06332524120807648, "logps/rejected": -3.758251667022705, "loss": 0.0613, "nll_loss": 0.07212791591882706, "rewards/accuracies": 1.0, "rewards/chosen": -0.003166262526065111, "rewards/margins": 0.18474632501602173, "rewards/rejected": -0.18791258335113525, "step": 690 }, { "epoch": 1.42, "grad_norm": 3.25, "learning_rate": 1.8898223650461362e-06, "log_odds_chosen": 8.65866756439209, "log_odds_ratio": -0.012265409342944622, "logits/chosen": -1.9048351049423218, "logits/rejected": -1.9031221866607666, "logps/chosen": -0.0720919519662857, "logps/rejected": -3.809000015258789, "loss": 0.0596, "nll_loss": 0.054995644837617874, "rewards/accuracies": 1.0, "rewards/chosen": -0.003604597644880414, "rewards/margins": 0.18684543669223785, "rewards/rejected": -0.1904500126838684, "step": 700 }, { "epoch": 1.44, "grad_norm": 3.34375, "learning_rate": 1.876466562602004e-06, "log_odds_chosen": 8.907628059387207, "log_odds_ratio": -0.01111848745495081, "logits/chosen": -1.945853590965271, "logits/rejected": -1.942375898361206, "logps/chosen": -0.07798963785171509, "logps/rejected": -4.1663408279418945, "loss": 0.0662, "nll_loss": 0.08023327589035034, "rewards/accuracies": 1.0, "rewards/chosen": -0.003899481613188982, "rewards/margins": 0.20441755652427673, "rewards/rejected": -0.20831704139709473, "step": 710 }, { "epoch": 1.46, "grad_norm": 5.28125, "learning_rate": 1.863389981249825e-06, "log_odds_chosen": 9.416102409362793, "log_odds_ratio": -0.007732940372079611, "logits/chosen": -1.8631290197372437, "logits/rejected": -1.8593025207519531, "logps/chosen": -0.06107889488339424, "logps/rejected": -4.503323554992676, "loss": 0.0585, "nll_loss": 0.05343944951891899, "rewards/accuracies": 1.0, "rewards/chosen": -0.0030539450235664845, "rewards/margins": 0.22211222350597382, "rewards/rejected": -0.2251661717891693, "step": 720 }, { "epoch": 1.48, "grad_norm": 2.78125, "learning_rate": 1.8505830254940132e-06, "log_odds_chosen": 9.134099960327148, "log_odds_ratio": -0.011219611391425133, "logits/chosen": -1.9615509510040283, "logits/rejected": -1.9607875347137451, "logps/chosen": -0.05891376733779907, "logps/rejected": -4.027497291564941, "loss": 0.0582, "nll_loss": 0.050385020673274994, "rewards/accuracies": 1.0, "rewards/chosen": -0.0029456880874931812, "rewards/margins": 0.1984291821718216, "rewards/rejected": -0.2013748586177826, "step": 730 }, { "epoch": 1.5, "grad_norm": 4.0625, "learning_rate": 1.8380365552345197e-06, "log_odds_chosen": 8.314208984375, "log_odds_ratio": -0.017240628600120544, "logits/chosen": -1.9306405782699585, "logits/rejected": -1.9276716709136963, "logps/chosen": -0.12811905145645142, "logps/rejected": -3.857971668243408, "loss": 0.0666, "nll_loss": 0.07230392098426819, "rewards/accuracies": 1.0, "rewards/chosen": -0.006405952386558056, "rewards/margins": 0.18649262189865112, "rewards/rejected": -0.19289858639240265, "step": 740 }, { "epoch": 1.52, "grad_norm": 6.8125, "learning_rate": 1.8257418583505536e-06, "log_odds_chosen": 8.118074417114258, "log_odds_ratio": -0.04242696613073349, "logits/chosen": -1.951964020729065, "logits/rejected": -1.9483181238174438, "logps/chosen": -0.07924682646989822, "logps/rejected": -3.6926894187927246, "loss": 0.0608, "nll_loss": 0.05164622515439987, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.003962341230362654, "rewards/margins": 0.18067213892936707, "rewards/rejected": -0.1846344769001007, "step": 750 }, { "epoch": 1.54, "grad_norm": 2.890625, "learning_rate": 1.8136906252750293e-06, "log_odds_chosen": 8.884002685546875, "log_odds_ratio": -0.007163527421653271, "logits/chosen": -1.8522392511367798, "logits/rejected": -1.8525257110595703, "logps/chosen": -0.06406822055578232, "logps/rejected": -3.8873114585876465, "loss": 0.0508, "nll_loss": 0.052127063274383545, "rewards/accuracies": 1.0, "rewards/chosen": -0.003203411353752017, "rewards/margins": 0.19116216897964478, "rewards/rejected": -0.19436559081077576, "step": 760 }, { "epoch": 1.57, "grad_norm": 4.8125, "learning_rate": 1.801874925391118e-06, "log_odds_chosen": 8.48697566986084, "log_odds_ratio": -0.026209736242890358, "logits/chosen": -1.9563274383544922, "logits/rejected": -1.9566619396209717, "logps/chosen": -0.075563944876194, "logps/rejected": -3.6374893188476562, "loss": 0.0665, "nll_loss": 0.05961776524782181, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.0037781973369419575, "rewards/margins": 0.17809626460075378, "rewards/rejected": -0.18187445402145386, "step": 770 }, { "epoch": 1.59, "grad_norm": 3.703125, "learning_rate": 1.7902871850985824e-06, "log_odds_chosen": 9.119174003601074, "log_odds_ratio": -0.024956868961453438, "logits/chosen": -1.918013572692871, "logits/rejected": -1.9134963750839233, "logps/chosen": -0.05534970015287399, "logps/rejected": -4.010322570800781, "loss": 0.0595, "nll_loss": 0.04103861376643181, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.0027674853336066008, "rewards/margins": 0.19774861633777618, "rewards/rejected": -0.20051613450050354, "step": 780 }, { "epoch": 1.61, "grad_norm": 3.6875, "learning_rate": 1.7789201674120502e-06, "log_odds_chosen": 8.733691215515137, "log_odds_ratio": -0.008985884487628937, "logits/chosen": -1.8730590343475342, "logits/rejected": -1.8695628643035889, "logps/chosen": -0.05646519735455513, "logps/rejected": -3.503816604614258, "loss": 0.0506, "nll_loss": 0.047934651374816895, "rewards/accuracies": 1.0, "rewards/chosen": -0.002823260147124529, "rewards/margins": 0.17236760258674622, "rewards/rejected": -0.17519085109233856, "step": 790 }, { "epoch": 1.63, "grad_norm": 2.140625, "learning_rate": 1.7677669529663689e-06, "log_odds_chosen": 8.943314552307129, "log_odds_ratio": -0.00942598469555378, "logits/chosen": -1.8948551416397095, "logits/rejected": -1.892686128616333, "logps/chosen": -0.05206329748034477, "logps/rejected": -3.80363130569458, "loss": 0.0526, "nll_loss": 0.06757352501153946, "rewards/accuracies": 1.0, "rewards/chosen": -0.0026031648740172386, "rewards/margins": 0.18757839500904083, "rewards/rejected": -0.19018153846263885, "step": 800 }, { "epoch": 1.65, "grad_norm": 3.515625, "learning_rate": 1.7568209223157664e-06, "log_odds_chosen": 9.602978706359863, "log_odds_ratio": -0.0054091801866889, "logits/chosen": -1.8749698400497437, "logits/rejected": -1.8726646900177002, "logps/chosen": -0.049051497131586075, "logps/rejected": -4.167695045471191, "loss": 0.0527, "nll_loss": 0.08227180689573288, "rewards/accuracies": 1.0, "rewards/chosen": -0.0024525749031454325, "rewards/margins": 0.20593218505382538, "rewards/rejected": -0.20838478207588196, "step": 810 }, { "epoch": 1.67, "grad_norm": 4.1875, "learning_rate": 1.7460757394239458e-06, "log_odds_chosen": 8.731072425842285, "log_odds_ratio": -0.006553919520229101, "logits/chosen": -1.9533252716064453, "logits/rejected": -1.9529660940170288, "logps/chosen": -0.0727374479174614, "logps/rejected": -3.5265769958496094, "loss": 0.0659, "nll_loss": 0.0733790472149849, "rewards/accuracies": 1.0, "rewards/chosen": -0.003636872861534357, "rewards/margins": 0.17269198596477509, "rewards/rejected": -0.1763288527727127, "step": 820 }, { "epoch": 1.69, "grad_norm": 2.984375, "learning_rate": 1.7355253362515584e-06, "log_odds_chosen": 9.135687828063965, "log_odds_ratio": -0.03666002303361893, "logits/chosen": -1.9465566873550415, "logits/rejected": -1.9425933361053467, "logps/chosen": -0.07385378330945969, "logps/rejected": -3.989952564239502, "loss": 0.0478, "nll_loss": 0.06535151600837708, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.003692689584568143, "rewards/margins": 0.1958049088716507, "rewards/rejected": -0.19949761033058167, "step": 830 }, { "epoch": 1.71, "grad_norm": 2.9375, "learning_rate": 1.7251638983558855e-06, "log_odds_chosen": 8.61412239074707, "log_odds_ratio": -0.006854639388620853, "logits/chosen": -1.8970463275909424, "logits/rejected": -1.8957879543304443, "logps/chosen": -0.09370087087154388, "logps/rejected": -3.841550827026367, "loss": 0.0587, "nll_loss": 0.046143922954797745, "rewards/accuracies": 1.0, "rewards/chosen": -0.004685043357312679, "rewards/margins": 0.18739250302314758, "rewards/rejected": -0.19207754731178284, "step": 840 }, { "epoch": 1.73, "grad_norm": 3.40625, "learning_rate": 1.7149858514250883e-06, "log_odds_chosen": 9.044164657592773, "log_odds_ratio": -0.023376554250717163, "logits/chosen": -1.9251768589019775, "logits/rejected": -1.9241018295288086, "logps/chosen": -0.05019516497850418, "logps/rejected": -3.740339994430542, "loss": 0.0537, "nll_loss": 0.04762560874223709, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.0025097583420574665, "rewards/margins": 0.18450722098350525, "rewards/rejected": -0.18701700866222382, "step": 850 }, { "epoch": 1.75, "grad_norm": 2.703125, "learning_rate": 1.704985848676184e-06, "log_odds_chosen": 8.560470581054688, "log_odds_ratio": -0.015030615031719208, "logits/chosen": -1.893930435180664, "logits/rejected": -1.8902561664581299, "logps/chosen": -0.07954917848110199, "logps/rejected": -3.786511182785034, "loss": 0.0544, "nll_loss": 0.054652292281389236, "rewards/accuracies": 1.0, "rewards/chosen": -0.0039774589240550995, "rewards/margins": 0.18534812331199646, "rewards/rejected": -0.18932557106018066, "step": 860 }, { "epoch": 1.77, "grad_norm": 2.9375, "learning_rate": 1.6951587590520263e-06, "log_odds_chosen": 9.6688232421875, "log_odds_ratio": -0.010795338079333305, "logits/chosen": -1.9225839376449585, "logits/rejected": -1.9220905303955078, "logps/chosen": -0.048293761909008026, "logps/rejected": -4.323627948760986, "loss": 0.0488, "nll_loss": 0.043466441333293915, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.002414688002318144, "rewards/margins": 0.21376672387123108, "rewards/rejected": -0.2161814272403717, "step": 870 }, { "epoch": 1.79, "grad_norm": 3.234375, "learning_rate": 1.6854996561581053e-06, "log_odds_chosen": 9.363157272338867, "log_odds_ratio": -0.026759719476103783, "logits/chosen": -1.9485028982162476, "logits/rejected": -1.9452836513519287, "logps/chosen": -0.042739786207675934, "logps/rejected": -3.5607593059539795, "loss": 0.0555, "nll_loss": 0.04253282770514488, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.002136989263817668, "rewards/margins": 0.17590096592903137, "rewards/rejected": -0.17803797125816345, "step": 880 }, { "epoch": 1.81, "grad_norm": 4.1875, "learning_rate": 1.6760038078849776e-06, "log_odds_chosen": 8.319189071655273, "log_odds_ratio": -0.023240404203534126, "logits/chosen": -1.851123571395874, "logits/rejected": -1.849542260169983, "logps/chosen": -0.06788856536149979, "logps/rejected": -3.406628131866455, "loss": 0.0603, "nll_loss": 0.04571394622325897, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.003394428174942732, "rewards/margins": 0.1669369637966156, "rewards/rejected": -0.17033140361309052, "step": 890 }, { "epoch": 1.83, "grad_norm": 4.21875, "learning_rate": 1.6666666666666667e-06, "log_odds_chosen": 8.61945915222168, "log_odds_ratio": -0.017158176749944687, "logits/chosen": -1.9368407726287842, "logits/rejected": -1.9338048696517944, "logps/chosen": -0.05057498812675476, "logps/rejected": -3.4434471130371094, "loss": 0.0499, "nll_loss": 0.05269382521510124, "rewards/accuracies": 1.0, "rewards/chosen": -0.0025287498719990253, "rewards/margins": 0.1696436107158661, "rewards/rejected": -0.17217235267162323, "step": 900 }, { "epoch": 1.85, "grad_norm": 3.265625, "learning_rate": 1.6574838603294898e-06, "log_odds_chosen": 9.762075424194336, "log_odds_ratio": -0.0037377229891717434, "logits/chosen": -2.0028135776519775, "logits/rejected": -2.0014703273773193, "logps/chosen": -0.034680210053920746, "logps/rejected": -4.053231239318848, "loss": 0.0576, "nll_loss": 0.057746343314647675, "rewards/accuracies": 1.0, "rewards/chosen": -0.0017340105259791017, "rewards/margins": 0.20092757046222687, "rewards/rejected": -0.20266155898571014, "step": 910 }, { "epoch": 1.87, "grad_norm": 3.0, "learning_rate": 1.648451183489468e-06, "log_odds_chosen": 9.181684494018555, "log_odds_ratio": -0.007746308110654354, "logits/chosen": -1.935484528541565, "logits/rejected": -1.932363748550415, "logps/chosen": -0.08534862101078033, "logps/rejected": -4.044074535369873, "loss": 0.0526, "nll_loss": 0.06775641441345215, "rewards/accuracies": 1.0, "rewards/chosen": -0.004267430864274502, "rewards/margins": 0.1979363113641739, "rewards/rejected": -0.20220375061035156, "step": 920 }, { "epoch": 1.89, "grad_norm": 1.609375, "learning_rate": 1.6395645894598825e-06, "log_odds_chosen": 9.165306091308594, "log_odds_ratio": -0.01449726801365614, "logits/chosen": -1.9488122463226318, "logits/rejected": -1.9472631216049194, "logps/chosen": -0.07069545984268188, "logps/rejected": -3.8849570751190186, "loss": 0.057, "nll_loss": 0.052025146782398224, "rewards/accuracies": 1.0, "rewards/chosen": -0.0035347728990018368, "rewards/margins": 0.1907130777835846, "rewards/rejected": -0.19424787163734436, "step": 930 }, { "epoch": 1.91, "grad_norm": 3.90625, "learning_rate": 1.6308201826336057e-06, "log_odds_chosen": 9.020658493041992, "log_odds_ratio": -0.00849425233900547, "logits/chosen": -1.916486144065857, "logits/rejected": -1.9109046459197998, "logps/chosen": -0.06480909883975983, "logps/rejected": -4.059305667877197, "loss": 0.0504, "nll_loss": 0.058377087116241455, "rewards/accuracies": 1.0, "rewards/chosen": -0.0032404554076492786, "rewards/margins": 0.19972485303878784, "rewards/rejected": -0.20296530425548553, "step": 940 }, { "epoch": 1.93, "grad_norm": 3.75, "learning_rate": 1.6222142113076255e-06, "log_odds_chosen": 9.191986083984375, "log_odds_ratio": -0.007987757213413715, "logits/chosen": -1.8939701318740845, "logits/rejected": -1.892139196395874, "logps/chosen": -0.042362868785858154, "logps/rejected": -3.922192335128784, "loss": 0.0543, "nll_loss": 0.0392763689160347, "rewards/accuracies": 1.0, "rewards/chosen": -0.0021181434858590364, "rewards/margins": 0.19399148225784302, "rewards/rejected": -0.1961096078157425, "step": 950 }, { "epoch": 1.95, "grad_norm": 3.65625, "learning_rate": 1.6137430609197571e-06, "log_odds_chosen": 8.950517654418945, "log_odds_ratio": -0.01621365174651146, "logits/chosen": -1.9614036083221436, "logits/rejected": -1.9606664180755615, "logps/chosen": -0.07848796993494034, "logps/rejected": -4.038422584533691, "loss": 0.0554, "nll_loss": 0.0618586540222168, "rewards/accuracies": 1.0, "rewards/chosen": -0.003924398683011532, "rewards/margins": 0.19799675047397614, "rewards/rejected": -0.20192115008831024, "step": 960 }, { "epoch": 1.97, "grad_norm": 2.875, "learning_rate": 1.605403247669839e-06, "log_odds_chosen": 8.41962718963623, "log_odds_ratio": -0.020061850547790527, "logits/chosen": -1.908539056777954, "logits/rejected": -1.9044193029403687, "logps/chosen": -0.07218164205551147, "logps/rejected": -3.6821017265319824, "loss": 0.0528, "nll_loss": 0.03915850818157196, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.003609082894399762, "rewards/margins": 0.1804960072040558, "rewards/rejected": -0.18410511314868927, "step": 970 }, { "epoch": 1.99, "grad_norm": 4.4375, "learning_rate": 1.59719141249985e-06, "log_odds_chosen": 8.960528373718262, "log_odds_ratio": -0.007182210683822632, "logits/chosen": -2.059950351715088, "logits/rejected": -2.058908224105835, "logps/chosen": -0.07827770709991455, "logps/rejected": -3.9805305004119873, "loss": 0.0589, "nll_loss": 0.08598147332668304, "rewards/accuracies": 1.0, "rewards/chosen": -0.003913884982466698, "rewards/margins": 0.19511264562606812, "rewards/rejected": -0.19902652502059937, "step": 980 }, { "epoch": 2.01, "grad_norm": 2.421875, "learning_rate": 1.5891043154093205e-06, "log_odds_chosen": 8.75261402130127, "log_odds_ratio": -0.008246051147580147, "logits/chosen": -1.9300010204315186, "logits/rejected": -1.9284627437591553, "logps/chosen": -0.059024132788181305, "logps/rejected": -3.7112643718719482, "loss": 0.0501, "nll_loss": 0.0379130020737648, "rewards/accuracies": 1.0, "rewards/chosen": -0.0029512066394090652, "rewards/margins": 0.18261203169822693, "rewards/rejected": -0.18556323647499084, "step": 990 }, { "epoch": 2.03, "grad_norm": 2.4375, "learning_rate": 1.5811388300841898e-06, "log_odds_chosen": 9.132776260375977, "log_odds_ratio": -0.010855279862880707, "logits/chosen": -1.9048831462860107, "logits/rejected": -1.9021356105804443, "logps/chosen": -0.05851290374994278, "logps/rejected": -3.9342761039733887, "loss": 0.0399, "nll_loss": 0.045368582010269165, "rewards/accuracies": 1.0, "rewards/chosen": -0.002925645327195525, "rewards/margins": 0.19378817081451416, "rewards/rejected": -0.19671380519866943, "step": 1000 }, { "epoch": 2.05, "grad_norm": 3.34375, "learning_rate": 1.5732919388188816e-06, "log_odds_chosen": 9.254007339477539, "log_odds_ratio": -0.004844398703426123, "logits/chosen": -1.906938910484314, "logits/rejected": -1.9027900695800781, "logps/chosen": -0.0639704093337059, "logps/rejected": -4.133778095245361, "loss": 0.0377, "nll_loss": 0.03271421417593956, "rewards/accuracies": 1.0, "rewards/chosen": -0.003198520513251424, "rewards/margins": 0.20349040627479553, "rewards/rejected": -0.20668891072273254, "step": 1010 }, { "epoch": 2.07, "grad_norm": 2.34375, "learning_rate": 1.565560727712874e-06, "log_odds_chosen": 9.79780387878418, "log_odds_ratio": -0.0035797320306301117, "logits/chosen": -1.934853196144104, "logits/rejected": -1.9338003396987915, "logps/chosen": -0.05114441365003586, "logps/rejected": -4.095462799072266, "loss": 0.0456, "nll_loss": 0.055598996579647064, "rewards/accuracies": 1.0, "rewards/chosen": -0.002557220868766308, "rewards/margins": 0.20221590995788574, "rewards/rejected": -0.20477311313152313, "step": 1020 }, { "epoch": 2.09, "grad_norm": 2.140625, "learning_rate": 1.5579423821243897e-06, "log_odds_chosen": 10.34232234954834, "log_odds_ratio": -0.002075557829812169, "logits/chosen": -1.908393144607544, "logits/rejected": -1.9075828790664673, "logps/chosen": -0.03475125879049301, "logps/rejected": -4.130476474761963, "loss": 0.0436, "nll_loss": 0.04652316868305206, "rewards/accuracies": 1.0, "rewards/chosen": -0.0017375629395246506, "rewards/margins": 0.2047862708568573, "rewards/rejected": -0.2065238207578659, "step": 1030 }, { "epoch": 2.11, "grad_norm": 3.796875, "learning_rate": 1.5504341823651056e-06, "log_odds_chosen": 9.534278869628906, "log_odds_ratio": -0.014330191537737846, "logits/chosen": -1.8937572240829468, "logits/rejected": -1.8925514221191406, "logps/chosen": -0.05281580239534378, "logps/rejected": -3.8376965522766113, "loss": 0.0506, "nll_loss": 0.04847393184900284, "rewards/accuracies": 1.0, "rewards/chosen": -0.0026407900732010603, "rewards/margins": 0.18924403190612793, "rewards/rejected": -0.1918848156929016, "step": 1040 }, { "epoch": 2.13, "grad_norm": 2.765625, "learning_rate": 1.5430334996209192e-06, "log_odds_chosen": 9.982309341430664, "log_odds_ratio": -0.0034835089463740587, "logits/chosen": -1.9073753356933594, "logits/rejected": -1.9033172130584717, "logps/chosen": -0.03734086453914642, "logps/rejected": -4.198154449462891, "loss": 0.0399, "nll_loss": 0.03343256562948227, "rewards/accuracies": 1.0, "rewards/chosen": -0.0018670432036742568, "rewards/margins": 0.20804063975811005, "rewards/rejected": -0.20990769565105438, "step": 1050 }, { "epoch": 2.15, "grad_norm": 2.015625, "learning_rate": 1.5357377920848783e-06, "log_odds_chosen": 9.932341575622559, "log_odds_ratio": -0.00664928276091814, "logits/chosen": -1.968000054359436, "logits/rejected": -1.9650490283966064, "logps/chosen": -0.053252361714839935, "logps/rejected": -4.3154191970825195, "loss": 0.0398, "nll_loss": 0.03708261623978615, "rewards/accuracies": 1.0, "rewards/chosen": -0.0026626181788742542, "rewards/margins": 0.2131083756685257, "rewards/rejected": -0.21577095985412598, "step": 1060 }, { "epoch": 2.17, "grad_norm": 3.203125, "learning_rate": 1.5285446012893579e-06, "log_odds_chosen": 10.412023544311523, "log_odds_ratio": -0.004382356535643339, "logits/chosen": -1.8630377054214478, "logits/rejected": -1.858724594116211, "logps/chosen": -0.03156786039471626, "logps/rejected": -4.467475414276123, "loss": 0.0435, "nll_loss": 0.043990008533000946, "rewards/accuracies": 1.0, "rewards/chosen": -0.0015783928101882339, "rewards/margins": 0.22179536521434784, "rewards/rejected": -0.22337380051612854, "step": 1070 }, { "epoch": 2.2, "grad_norm": 2.859375, "learning_rate": 1.5214515486254614e-06, "log_odds_chosen": 9.860450744628906, "log_odds_ratio": -0.00421870406717062, "logits/chosen": -1.9334264993667603, "logits/rejected": -1.9309759140014648, "logps/chosen": -0.04163174703717232, "logps/rejected": -4.278031349182129, "loss": 0.0399, "nll_loss": 0.03083736263215542, "rewards/accuracies": 1.0, "rewards/chosen": -0.0020815874449908733, "rewards/margins": 0.21182000637054443, "rewards/rejected": -0.2139015644788742, "step": 1080 }, { "epoch": 2.22, "grad_norm": 3.375, "learning_rate": 1.5144563320384566e-06, "log_odds_chosen": 9.61739444732666, "log_odds_ratio": -0.007051125168800354, "logits/chosen": -1.9654079675674438, "logits/rejected": -1.9627532958984375, "logps/chosen": -0.04155623912811279, "logps/rejected": -3.9517130851745605, "loss": 0.0419, "nll_loss": 0.0399235263466835, "rewards/accuracies": 1.0, "rewards/chosen": -0.002077811863273382, "rewards/margins": 0.19550786912441254, "rewards/rejected": -0.19758570194244385, "step": 1090 }, { "epoch": 2.24, "grad_norm": 2.5625, "learning_rate": 1.5075567228888182e-06, "log_odds_chosen": 9.472233772277832, "log_odds_ratio": -0.004241319373250008, "logits/chosen": -1.8757314682006836, "logits/rejected": -1.875422477722168, "logps/chosen": -0.04146653413772583, "logps/rejected": -3.9626433849334717, "loss": 0.0388, "nll_loss": 0.044396962970495224, "rewards/accuracies": 1.0, "rewards/chosen": -0.002073326613754034, "rewards/margins": 0.19605886936187744, "rewards/rejected": -0.19813218712806702, "step": 1100 }, { "epoch": 2.26, "grad_norm": 2.84375, "learning_rate": 1.5007505629691608e-06, "log_odds_chosen": 10.060752868652344, "log_odds_ratio": -0.005245304200798273, "logits/chosen": -1.884495735168457, "logits/rejected": -1.8812463283538818, "logps/chosen": -0.042206455022096634, "logps/rejected": -4.164919853210449, "loss": 0.0437, "nll_loss": 0.03563828393816948, "rewards/accuracies": 1.0, "rewards/chosen": -0.002110322704538703, "rewards/margins": 0.20613567531108856, "rewards/rejected": -0.20824599266052246, "step": 1110 }, { "epoch": 2.28, "grad_norm": 1.875, "learning_rate": 1.494035761667992e-06, "log_odds_chosen": 10.209965705871582, "log_odds_ratio": -0.00312315602786839, "logits/chosen": -1.921362280845642, "logits/rejected": -1.9180587530136108, "logps/chosen": -0.04360593855381012, "logps/rejected": -4.336488723754883, "loss": 0.0425, "nll_loss": 0.050070591270923615, "rewards/accuracies": 1.0, "rewards/chosen": -0.002180296927690506, "rewards/margins": 0.2146441489458084, "rewards/rejected": -0.21682441234588623, "step": 1120 }, { "epoch": 2.3, "grad_norm": 3.34375, "learning_rate": 1.487410293271824e-06, "log_odds_chosen": 9.571352005004883, "log_odds_ratio": -0.0124302227050066, "logits/chosen": -1.9245920181274414, "logits/rejected": -1.9231822490692139, "logps/chosen": -0.051782481372356415, "logps/rejected": -4.130411624908447, "loss": 0.042, "nll_loss": 0.04519005864858627, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.002589124022051692, "rewards/margins": 0.20393142104148865, "rewards/rejected": -0.20652055740356445, "step": 1130 }, { "epoch": 2.32, "grad_norm": 2.0625, "learning_rate": 1.480872194397731e-06, "log_odds_chosen": 9.836039543151855, "log_odds_ratio": -0.004647588822990656, "logits/chosen": -1.8972570896148682, "logits/rejected": -1.8939294815063477, "logps/chosen": -0.0661562830209732, "logps/rejected": -4.237457752227783, "loss": 0.0435, "nll_loss": 0.04812124744057655, "rewards/accuracies": 1.0, "rewards/chosen": -0.0033078144770115614, "rewards/margins": 0.20856507122516632, "rewards/rejected": -0.211872860789299, "step": 1140 }, { "epoch": 2.34, "grad_norm": 3.546875, "learning_rate": 1.4744195615489715e-06, "log_odds_chosen": 10.264336585998535, "log_odds_ratio": -0.0017655363772064447, "logits/chosen": -1.9511454105377197, "logits/rejected": -1.9473825693130493, "logps/chosen": -0.04254579171538353, "logps/rejected": -4.465078353881836, "loss": 0.0408, "nll_loss": 0.03954671695828438, "rewards/accuracies": 1.0, "rewards/chosen": -0.0021272895392030478, "rewards/margins": 0.22112664580345154, "rewards/rejected": -0.22325392067432404, "step": 1150 }, { "epoch": 2.36, "grad_norm": 2.203125, "learning_rate": 1.4680505487867589e-06, "log_odds_chosen": 9.718263626098633, "log_odds_ratio": -0.002675818046554923, "logits/chosen": -1.9049984216690063, "logits/rejected": -1.9013592004776, "logps/chosen": -0.04851164296269417, "logps/rejected": -4.104137420654297, "loss": 0.0396, "nll_loss": 0.03321639448404312, "rewards/accuracies": 1.0, "rewards/chosen": -0.0024255819153040648, "rewards/margins": 0.2027812898159027, "rewards/rejected": -0.20520684123039246, "step": 1160 }, { "epoch": 2.38, "grad_norm": 1.8828125, "learning_rate": 1.4617633655117156e-06, "log_odds_chosen": 9.280858993530273, "log_odds_ratio": -0.008265355601906776, "logits/chosen": -1.964536428451538, "logits/rejected": -1.9617042541503906, "logps/chosen": -0.06000961735844612, "logps/rejected": -4.104333400726318, "loss": 0.0435, "nll_loss": 0.04580981656908989, "rewards/accuracies": 1.0, "rewards/chosen": -0.003000481054186821, "rewards/margins": 0.20221619307994843, "rewards/rejected": -0.2052166759967804, "step": 1170 }, { "epoch": 2.4, "grad_norm": 2.578125, "learning_rate": 1.4555562743489552e-06, "log_odds_chosen": 10.110807418823242, "log_odds_ratio": -0.003054619999602437, "logits/chosen": -1.925559401512146, "logits/rejected": -1.924526572227478, "logps/chosen": -0.03492622449994087, "logps/rejected": -4.048843860626221, "loss": 0.0415, "nll_loss": 0.040362752974033356, "rewards/accuracies": 1.0, "rewards/chosen": -0.001746311434544623, "rewards/margins": 0.20069590210914612, "rewards/rejected": -0.2024422138929367, "step": 1180 }, { "epoch": 2.42, "grad_norm": 2.53125, "learning_rate": 1.4494275891311214e-06, "log_odds_chosen": 9.7469482421875, "log_odds_ratio": -0.012390781193971634, "logits/chosen": -1.8723634481430054, "logits/rejected": -1.8694031238555908, "logps/chosen": -0.045554567128419876, "logps/rejected": -4.15027379989624, "loss": 0.0354, "nll_loss": 0.03426050767302513, "rewards/accuracies": 1.0, "rewards/chosen": -0.002277728170156479, "rewards/margins": 0.20523595809936523, "rewards/rejected": -0.2075137197971344, "step": 1190 }, { "epoch": 2.44, "grad_norm": 2.140625, "learning_rate": 1.4433756729740647e-06, "log_odds_chosen": 10.098226547241211, "log_odds_ratio": -0.0019627574365586042, "logits/chosen": -1.9441982507705688, "logits/rejected": -1.9426124095916748, "logps/chosen": -0.02355217933654785, "logps/rejected": -4.209158897399902, "loss": 0.0457, "nll_loss": 0.053135812282562256, "rewards/accuracies": 1.0, "rewards/chosen": -0.0011776090832427144, "rewards/margins": 0.20928029716014862, "rewards/rejected": -0.2104579210281372, "step": 1200 }, { "epoch": 2.46, "grad_norm": 3.609375, "learning_rate": 1.4373989364401727e-06, "log_odds_chosen": 9.691657066345215, "log_odds_ratio": -0.004629252012819052, "logits/chosen": -1.9337596893310547, "logits/rejected": -1.931654930114746, "logps/chosen": -0.06260120123624802, "logps/rejected": -4.166058540344238, "loss": 0.0411, "nll_loss": 0.03751087933778763, "rewards/accuracies": 1.0, "rewards/chosen": -0.003130060387775302, "rewards/margins": 0.20517286658287048, "rewards/rejected": -0.20830294489860535, "step": 1210 }, { "epoch": 2.48, "grad_norm": 1.3984375, "learning_rate": 1.4314958357846706e-06, "log_odds_chosen": 9.472856521606445, "log_odds_ratio": -0.00650912756100297, "logits/chosen": -1.95502507686615, "logits/rejected": -1.951743721961975, "logps/chosen": -0.07376778870820999, "logps/rejected": -4.022158145904541, "loss": 0.0357, "nll_loss": 0.03547949343919754, "rewards/accuracies": 1.0, "rewards/chosen": -0.003688389901071787, "rewards/margins": 0.19741952419281006, "rewards/rejected": -0.2011079341173172, "step": 1220 }, { "epoch": 2.5, "grad_norm": 2.59375, "learning_rate": 1.4256648712805027e-06, "log_odds_chosen": 10.319469451904297, "log_odds_ratio": -0.001875861780717969, "logits/chosen": -1.9282852411270142, "logits/rejected": -1.9243885278701782, "logps/chosen": -0.0374315045773983, "logps/rejected": -4.336482524871826, "loss": 0.0412, "nll_loss": 0.0424610860645771, "rewards/accuracies": 1.0, "rewards/chosen": -0.0018715756013989449, "rewards/margins": 0.21495255827903748, "rewards/rejected": -0.21682414412498474, "step": 1230 }, { "epoch": 2.52, "grad_norm": 3.78125, "learning_rate": 1.419904585617662e-06, "log_odds_chosen": 10.347482681274414, "log_odds_ratio": -0.0033975064288824797, "logits/chosen": -1.9401060342788696, "logits/rejected": -1.9385757446289062, "logps/chosen": -0.04415961354970932, "logps/rejected": -4.426904201507568, "loss": 0.0403, "nll_loss": 0.04613417014479637, "rewards/accuracies": 1.0, "rewards/chosen": -0.0022079809568822384, "rewards/margins": 0.21913722157478333, "rewards/rejected": -0.2213451862335205, "step": 1240 }, { "epoch": 2.54, "grad_norm": 2.53125, "learning_rate": 1.4142135623730952e-06, "log_odds_chosen": 10.331408500671387, "log_odds_ratio": -0.0018885558238252997, "logits/chosen": -1.940860390663147, "logits/rejected": -1.9384444952011108, "logps/chosen": -0.03379340097308159, "logps/rejected": -4.056138038635254, "loss": 0.0393, "nll_loss": 0.030756259337067604, "rewards/accuracies": 1.0, "rewards/chosen": -0.0016896702582016587, "rewards/margins": 0.20111723244190216, "rewards/rejected": -0.20280690491199493, "step": 1250 }, { "epoch": 2.56, "grad_norm": 2.109375, "learning_rate": 1.4085904245475275e-06, "log_odds_chosen": 10.41731071472168, "log_odds_ratio": -0.012277336791157722, "logits/chosen": -1.9542973041534424, "logits/rejected": -1.953238844871521, "logps/chosen": -0.030469322577118874, "logps/rejected": -4.482726097106934, "loss": 0.0407, "nll_loss": 0.04968537390232086, "rewards/accuracies": 1.0, "rewards/chosen": -0.001523466082289815, "rewards/margins": 0.22261285781860352, "rewards/rejected": -0.2241363227367401, "step": 1260 }, { "epoch": 2.58, "grad_norm": 2.75, "learning_rate": 1.4030338331657844e-06, "log_odds_chosen": 10.663671493530273, "log_odds_ratio": -0.005900475196540356, "logits/chosen": -1.8991870880126953, "logits/rejected": -1.8935034275054932, "logps/chosen": -0.036128245294094086, "logps/rejected": -4.496649742126465, "loss": 0.036, "nll_loss": 0.032302360981702805, "rewards/accuracies": 1.0, "rewards/chosen": -0.00180641224142164, "rewards/margins": 0.2230260670185089, "rewards/rejected": -0.22483249008655548, "step": 1270 }, { "epoch": 2.6, "grad_norm": 1.984375, "learning_rate": 1.3975424859373688e-06, "log_odds_chosen": 9.953948020935059, "log_odds_ratio": -0.002804636722430587, "logits/chosen": -1.9552701711654663, "logits/rejected": -1.9517933130264282, "logps/chosen": -0.04314829781651497, "logps/rejected": -4.2899580001831055, "loss": 0.0389, "nll_loss": 0.03743810951709747, "rewards/accuracies": 1.0, "rewards/chosen": -0.002157414797693491, "rewards/margins": 0.21234047412872314, "rewards/rejected": -0.21449792385101318, "step": 1280 }, { "epoch": 2.62, "grad_norm": 2.046875, "learning_rate": 1.3921151159742616e-06, "log_odds_chosen": 9.60466194152832, "log_odds_ratio": -0.021028626710176468, "logits/chosen": -1.8756396770477295, "logits/rejected": -1.8728761672973633, "logps/chosen": -0.043241944164037704, "logps/rejected": -4.110201835632324, "loss": 0.0374, "nll_loss": 0.03517068549990654, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.0021620974875986576, "rewards/margins": 0.20334799587726593, "rewards/rejected": -0.20551009476184845, "step": 1290 }, { "epoch": 2.64, "grad_norm": 2.171875, "learning_rate": 1.386750490563073e-06, "log_odds_chosen": 10.577958106994629, "log_odds_ratio": -0.0030032056383788586, "logits/chosen": -1.9832937717437744, "logits/rejected": -1.9783260822296143, "logps/chosen": -0.06359104067087173, "logps/rejected": -4.623613357543945, "loss": 0.0396, "nll_loss": 0.04861373454332352, "rewards/accuracies": 1.0, "rewards/chosen": -0.003179552499204874, "rewards/margins": 0.22800114750862122, "rewards/rejected": -0.23118066787719727, "step": 1300 }, { "epoch": 2.66, "grad_norm": 4.1875, "learning_rate": 1.3814474099888442e-06, "log_odds_chosen": 10.487229347229004, "log_odds_ratio": -0.0037563578225672245, "logits/chosen": -1.9412161111831665, "logits/rejected": -1.9390010833740234, "logps/chosen": -0.04359542205929756, "logps/rejected": -4.6636810302734375, "loss": 0.0433, "nll_loss": 0.03972161188721657, "rewards/accuracies": 1.0, "rewards/chosen": -0.0021797711960971355, "rewards/margins": 0.2310042679309845, "rewards/rejected": -0.2331840544939041, "step": 1310 }, { "epoch": 2.68, "grad_norm": 1.796875, "learning_rate": 1.376204706407951e-06, "log_odds_chosen": 10.476730346679688, "log_odds_ratio": -0.002192417625337839, "logits/chosen": -1.9325330257415771, "logits/rejected": -1.930166244506836, "logps/chosen": -0.02953045442700386, "logps/rejected": -4.279176712036133, "loss": 0.0353, "nll_loss": 0.03177348151803017, "rewards/accuracies": 1.0, "rewards/chosen": -0.001476522651501, "rewards/margins": 0.2124823033809662, "rewards/rejected": -0.21395882964134216, "step": 1320 }, { "epoch": 2.7, "grad_norm": 2.484375, "learning_rate": 1.3710212427677044e-06, "log_odds_chosen": 10.246335983276367, "log_odds_ratio": -0.0470537394285202, "logits/chosen": -1.9275156259536743, "logits/rejected": -1.9267303943634033, "logps/chosen": -0.07270321995019913, "logps/rejected": -4.131697177886963, "loss": 0.036, "nll_loss": 0.031709469854831696, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.0036351613234728575, "rewards/margins": 0.20294968783855438, "rewards/rejected": -0.2065848559141159, "step": 1330 }, { "epoch": 2.72, "grad_norm": 2.578125, "learning_rate": 1.3658959117703826e-06, "log_odds_chosen": 10.568267822265625, "log_odds_ratio": -0.003112158738076687, "logits/chosen": -1.948743462562561, "logits/rejected": -1.9471248388290405, "logps/chosen": -0.028313973918557167, "logps/rejected": -4.5648627281188965, "loss": 0.0378, "nll_loss": 0.041449058800935745, "rewards/accuracies": 1.0, "rewards/chosen": -0.001415698672644794, "rewards/margins": 0.22682742774486542, "rewards/rejected": -0.2282431572675705, "step": 1340 }, { "epoch": 2.74, "grad_norm": 3.546875, "learning_rate": 1.3608276348795436e-06, "log_odds_chosen": 10.491666793823242, "log_odds_ratio": -0.0048934281803667545, "logits/chosen": -1.9433883428573608, "logits/rejected": -1.9420783519744873, "logps/chosen": -0.03949974104762077, "logps/rejected": -4.376409530639648, "loss": 0.0414, "nll_loss": 0.03402886912226677, "rewards/accuracies": 1.0, "rewards/chosen": -0.001974987331777811, "rewards/margins": 0.21684548258781433, "rewards/rejected": -0.2188204824924469, "step": 1350 }, { "epoch": 2.76, "grad_norm": 2.265625, "learning_rate": 1.355815361366601e-06, "log_odds_chosen": 10.092164993286133, "log_odds_ratio": -0.008559630252420902, "logits/chosen": -1.945858359336853, "logits/rejected": -1.9418405294418335, "logps/chosen": -0.06214701011776924, "logps/rejected": -4.4873552322387695, "loss": 0.0359, "nll_loss": 0.03512818366289139, "rewards/accuracies": 1.0, "rewards/chosen": -0.0031073507852852345, "rewards/margins": 0.2212604284286499, "rewards/rejected": -0.22436778247356415, "step": 1360 }, { "epoch": 2.78, "grad_norm": 1.9296875, "learning_rate": 1.350858067395748e-06, "log_odds_chosen": 10.725711822509766, "log_odds_ratio": -0.020754020661115646, "logits/chosen": -1.9590380191802979, "logits/rejected": -1.953980803489685, "logps/chosen": -0.08093880116939545, "logps/rejected": -4.96276330947876, "loss": 0.0419, "nll_loss": 0.0406344048678875, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.004046940244734287, "rewards/margins": 0.2440912276506424, "rewards/rejected": -0.2481381893157959, "step": 1370 }, { "epoch": 2.8, "grad_norm": 3.359375, "learning_rate": 1.345954755145414e-06, "log_odds_chosen": 10.852327346801758, "log_odds_ratio": -0.0012642950750887394, "logits/chosen": -1.8603143692016602, "logits/rejected": -1.85476815700531, "logps/chosen": -0.03734128549695015, "logps/rejected": -4.723127841949463, "loss": 0.04, "nll_loss": 0.025891494005918503, "rewards/accuracies": 1.0, "rewards/chosen": -0.0018670640420168638, "rewards/margins": 0.23428933322429657, "rewards/rejected": -0.2361563891172409, "step": 1380 }, { "epoch": 2.83, "grad_norm": 2.390625, "learning_rate": 1.3411044519645502e-06, "log_odds_chosen": 10.230730056762695, "log_odds_ratio": -0.002351417439058423, "logits/chosen": -1.8683515787124634, "logits/rejected": -1.86406672000885, "logps/chosen": -0.044098712503910065, "logps/rejected": -4.59433126449585, "loss": 0.0331, "nll_loss": 0.03638423979282379, "rewards/accuracies": 1.0, "rewards/chosen": -0.0022049355320632458, "rewards/margins": 0.22751164436340332, "rewards/rejected": -0.22971658408641815, "step": 1390 }, { "epoch": 2.85, "grad_norm": 2.296875, "learning_rate": 1.3363062095621222e-06, "log_odds_chosen": 9.968645095825195, "log_odds_ratio": -0.004948216024786234, "logits/chosen": -1.9218571186065674, "logits/rejected": -1.9199250936508179, "logps/chosen": -0.06343747675418854, "logps/rejected": -4.219951629638672, "loss": 0.0353, "nll_loss": 0.03614867106080055, "rewards/accuracies": 1.0, "rewards/chosen": -0.0031718737445771694, "rewards/margins": 0.20782573521137238, "rewards/rejected": -0.21099761128425598, "step": 1400 }, { "epoch": 2.87, "grad_norm": 2.609375, "learning_rate": 1.3315591032282687e-06, "log_odds_chosen": 10.9727783203125, "log_odds_ratio": -0.004531107842922211, "logits/chosen": -1.9281492233276367, "logits/rejected": -1.9244182109832764, "logps/chosen": -0.03406985104084015, "logps/rejected": -4.806418418884277, "loss": 0.0409, "nll_loss": 0.04478253796696663, "rewards/accuracies": 1.0, "rewards/chosen": -0.0017034925986081362, "rewards/margins": 0.2386174201965332, "rewards/rejected": -0.24032095074653625, "step": 1410 }, { "epoch": 2.89, "grad_norm": 2.21875, "learning_rate": 1.3268622310856882e-06, "log_odds_chosen": 11.176974296569824, "log_odds_ratio": -0.003099204506725073, "logits/chosen": -1.9768095016479492, "logits/rejected": -1.9735969305038452, "logps/chosen": -0.03214428573846817, "logps/rejected": -4.844766616821289, "loss": 0.0368, "nll_loss": 0.03034246899187565, "rewards/accuracies": 1.0, "rewards/chosen": -0.0016072141006588936, "rewards/margins": 0.240631103515625, "rewards/rejected": -0.24223831295967102, "step": 1420 }, { "epoch": 2.91, "grad_norm": 3.953125, "learning_rate": 1.3222147133698626e-06, "log_odds_chosen": 9.971961975097656, "log_odds_ratio": -0.009039332158863544, "logits/chosen": -1.8924024105072021, "logits/rejected": -1.8878147602081299, "logps/chosen": -0.05458871275186539, "logps/rejected": -4.532398700714111, "loss": 0.041, "nll_loss": 0.033775750547647476, "rewards/accuracies": 1.0, "rewards/chosen": -0.002729435684159398, "rewards/margins": 0.2238904982805252, "rewards/rejected": -0.2266199290752411, "step": 1430 }, { "epoch": 2.93, "grad_norm": 2.09375, "learning_rate": 1.3176156917368248e-06, "log_odds_chosen": 10.354188919067383, "log_odds_ratio": -0.0086409542709589, "logits/chosen": -1.871347427368164, "logits/rejected": -1.8664233684539795, "logps/chosen": -0.0295790396630764, "logps/rejected": -4.4826579093933105, "loss": 0.0313, "nll_loss": 0.030693087726831436, "rewards/accuracies": 1.0, "rewards/chosen": -0.0014789520064368844, "rewards/margins": 0.22265395522117615, "rewards/rejected": -0.22413286566734314, "step": 1440 }, { "epoch": 2.95, "grad_norm": 2.578125, "learning_rate": 1.3130643285972255e-06, "log_odds_chosen": 10.23839282989502, "log_odds_ratio": -0.008670750074088573, "logits/chosen": -1.9214890003204346, "logits/rejected": -1.9176349639892578, "logps/chosen": -0.04931662604212761, "logps/rejected": -4.222093105316162, "loss": 0.0354, "nll_loss": 0.03125245124101639, "rewards/accuracies": 1.0, "rewards/chosen": -0.002465831348672509, "rewards/margins": 0.20863883197307587, "rewards/rejected": -0.21110467612743378, "step": 1450 }, { "epoch": 2.97, "grad_norm": 1.890625, "learning_rate": 1.3085598064755342e-06, "log_odds_chosen": 10.399072647094727, "log_odds_ratio": -0.003747476963326335, "logits/chosen": -1.8761295080184937, "logits/rejected": -1.8740062713623047, "logps/chosen": -0.048522982746362686, "logps/rejected": -4.560256004333496, "loss": 0.0359, "nll_loss": 0.042360566556453705, "rewards/accuracies": 1.0, "rewards/chosen": -0.0024261490907520056, "rewards/margins": 0.2255866527557373, "rewards/rejected": -0.2280128002166748, "step": 1460 }, { "epoch": 2.99, "grad_norm": 2.140625, "learning_rate": 1.3041013273932528e-06, "log_odds_chosen": 11.28456974029541, "log_odds_ratio": -0.0021586515940725803, "logits/chosen": -1.9435927867889404, "logits/rejected": -1.9409191608428955, "logps/chosen": -0.03155381605029106, "logps/rejected": -5.088595390319824, "loss": 0.0357, "nll_loss": 0.028978243470191956, "rewards/accuracies": 1.0, "rewards/chosen": -0.0015776908257976174, "rewards/margins": 0.25285208225250244, "rewards/rejected": -0.25442975759506226, "step": 1470 }, { "epoch": 3.0, "step": 1476, "total_flos": 0.0, "train_loss": 0.1731818498170311, "train_runtime": 10691.148, "train_samples_per_second": 8.831, "train_steps_per_second": 0.138 } ], "logging_steps": 10, "max_steps": 1476, "num_input_tokens_seen": 0, "num_train_epochs": 3, "save_steps": 500, "total_flos": 0.0, "train_batch_size": 8, "trial_name": null, "trial_params": null }