zephyr-orpo-141b-A35b-v0.1-bpw3.5 / trainer_state.json
blockblockblock's picture
Upload folder using huggingface_hub
1509b40 verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 3.0,
"eval_steps": 500,
"global_step": 651,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.05,
"grad_norm": 117760.0,
"learning_rate": 5.000000000000001e-07,
"log_odds_chosen": 0.36438828706741333,
"log_odds_ratio": -0.6397662162780762,
"logits/chosen": 3.8861491680145264,
"logits/rejected": 5.231001853942871,
"logps/chosen": -0.9861465692520142,
"logps/rejected": -1.2529093027114868,
"loss": 1.953,
"nll_loss": 3.2415008544921875,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.04930732399225235,
"rewards/margins": 0.013338141143321991,
"rewards/rejected": -0.06264545768499374,
"step": 10
},
{
"epoch": 0.09,
"grad_norm": 29184.0,
"learning_rate": 1.0000000000000002e-06,
"log_odds_chosen": 0.17107267677783966,
"log_odds_ratio": -0.6301043033599854,
"logits/chosen": 4.779696464538574,
"logits/rejected": 5.251872539520264,
"logps/chosen": -1.1045284271240234,
"logps/rejected": -1.2445374727249146,
"loss": 1.7108,
"nll_loss": 1.8614288568496704,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.05522642284631729,
"rewards/margins": 0.007000453770160675,
"rewards/rejected": -0.06222687289118767,
"step": 20
},
{
"epoch": 0.14,
"grad_norm": 3932160.0,
"learning_rate": 1.5e-06,
"log_odds_chosen": 0.478428453207016,
"log_odds_ratio": -0.5682710409164429,
"logits/chosen": 4.58956241607666,
"logits/rejected": 5.215265274047852,
"logps/chosen": -0.9884525537490845,
"logps/rejected": -1.2604442834854126,
"loss": 2.1071,
"nll_loss": 1.525723934173584,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.04942262917757034,
"rewards/margins": 0.013599586673080921,
"rewards/rejected": -0.06302221864461899,
"step": 30
},
{
"epoch": 0.18,
"grad_norm": 63232.0,
"learning_rate": 2.0000000000000003e-06,
"log_odds_chosen": 0.2763148248195648,
"log_odds_ratio": -0.6428317427635193,
"logits/chosen": 5.248695373535156,
"logits/rejected": 5.335747718811035,
"logps/chosen": -0.9019734263420105,
"logps/rejected": -1.058569312095642,
"loss": 1.6367,
"nll_loss": 1.17539644241333,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.045098677277565,
"rewards/margins": 0.007829795591533184,
"rewards/rejected": -0.052928466349840164,
"step": 40
},
{
"epoch": 0.23,
"grad_norm": 1802240.0,
"learning_rate": 2.5e-06,
"log_odds_chosen": -0.07109338045120239,
"log_odds_ratio": -0.9068069458007812,
"logits/chosen": 4.34699821472168,
"logits/rejected": 5.148941993713379,
"logps/chosen": -1.0307292938232422,
"logps/rejected": -1.0001896619796753,
"loss": 2.0499,
"nll_loss": 2.4581282138824463,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.05153647065162659,
"rewards/margins": -0.0015269846189767122,
"rewards/rejected": -0.05000948905944824,
"step": 50
},
{
"epoch": 0.28,
"grad_norm": 473088.0,
"learning_rate": 3e-06,
"log_odds_chosen": 0.7022095918655396,
"log_odds_ratio": -0.47877854108810425,
"logits/chosen": 5.137725830078125,
"logits/rejected": 5.073107719421387,
"logps/chosen": -0.7480964660644531,
"logps/rejected": -1.172572374343872,
"loss": 1.9116,
"nll_loss": 1.2398216724395752,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.037404827773571014,
"rewards/margins": 0.021223794668912888,
"rewards/rejected": -0.058628618717193604,
"step": 60
},
{
"epoch": 0.32,
"grad_norm": 4161536.0,
"learning_rate": 3.5e-06,
"log_odds_chosen": -0.30822521448135376,
"log_odds_ratio": -1.0616459846496582,
"logits/chosen": 4.378929615020752,
"logits/rejected": 5.239219665527344,
"logps/chosen": -1.115562081336975,
"logps/rejected": -0.8684147596359253,
"loss": 2.0166,
"nll_loss": 2.142368793487549,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.05577809736132622,
"rewards/margins": -0.012357364408671856,
"rewards/rejected": -0.043420739471912384,
"step": 70
},
{
"epoch": 0.37,
"grad_norm": 211968.0,
"learning_rate": 4.000000000000001e-06,
"log_odds_chosen": 0.38707518577575684,
"log_odds_ratio": -0.5776039361953735,
"logits/chosen": 5.019408226013184,
"logits/rejected": 5.3371453285217285,
"logps/chosen": -0.9375723004341125,
"logps/rejected": -1.1710981130599976,
"loss": 1.8918,
"nll_loss": 1.6274166107177734,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.046878617256879807,
"rewards/margins": 0.011676294729113579,
"rewards/rejected": -0.058554910123348236,
"step": 80
},
{
"epoch": 0.41,
"grad_norm": 1704.0,
"learning_rate": 4.5e-06,
"log_odds_chosen": 0.5051761865615845,
"log_odds_ratio": -0.5127500295639038,
"logits/chosen": 4.478859901428223,
"logits/rejected": 4.748915672302246,
"logps/chosen": -0.8121053576469421,
"logps/rejected": -1.1007237434387207,
"loss": 1.8015,
"nll_loss": 1.57842218875885,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.04060526192188263,
"rewards/margins": 0.014430919662117958,
"rewards/rejected": -0.055036187171936035,
"step": 90
},
{
"epoch": 0.46,
"grad_norm": 8.9375,
"learning_rate": 5e-06,
"log_odds_chosen": 1.037414789199829,
"log_odds_ratio": -0.3248421549797058,
"logits/chosen": 4.653676509857178,
"logits/rejected": 5.350204944610596,
"logps/chosen": -0.6695261001586914,
"logps/rejected": -1.300445795059204,
"loss": 0.9356,
"nll_loss": 0.7057152986526489,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.03347630053758621,
"rewards/margins": 0.031545985490083694,
"rewards/rejected": -0.0650222972035408,
"step": 100
},
{
"epoch": 0.51,
"grad_norm": 2.8125,
"learning_rate": 4.767312946227961e-06,
"log_odds_chosen": 0.6677854061126709,
"log_odds_ratio": -0.5610898733139038,
"logits/chosen": 4.671368598937988,
"logits/rejected": 5.119524002075195,
"logps/chosen": -0.8684868812561035,
"logps/rejected": -1.2020485401153564,
"loss": 0.8288,
"nll_loss": 0.9947482347488403,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.043424345552921295,
"rewards/margins": 0.01667807623744011,
"rewards/rejected": -0.06010241433978081,
"step": 110
},
{
"epoch": 0.55,
"grad_norm": 1.9921875,
"learning_rate": 4.564354645876385e-06,
"log_odds_chosen": 0.5597886443138123,
"log_odds_ratio": -0.5193617343902588,
"logits/chosen": 5.5248026847839355,
"logits/rejected": 6.067958354949951,
"logps/chosen": -0.9015194773674011,
"logps/rejected": -1.2406196594238281,
"loss": 0.7451,
"nll_loss": 0.8342186212539673,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.045075975358486176,
"rewards/margins": 0.01695500686764717,
"rewards/rejected": -0.06203098222613335,
"step": 120
},
{
"epoch": 0.6,
"grad_norm": 2.875,
"learning_rate": 4.385290096535147e-06,
"log_odds_chosen": 0.26957136392593384,
"log_odds_ratio": -0.7732787728309631,
"logits/chosen": 4.8973588943481445,
"logits/rejected": 5.552582263946533,
"logps/chosen": -0.877202033996582,
"logps/rejected": -0.9518612623214722,
"loss": 0.7319,
"nll_loss": 0.6940464377403259,
"rewards/accuracies": 0.5,
"rewards/chosen": -0.04386010393500328,
"rewards/margins": 0.0037329583428800106,
"rewards/rejected": -0.04759306460618973,
"step": 130
},
{
"epoch": 0.65,
"grad_norm": 2.5,
"learning_rate": 4.2257712736425835e-06,
"log_odds_chosen": 0.7680839896202087,
"log_odds_ratio": -0.5321913957595825,
"logits/chosen": 5.471996307373047,
"logits/rejected": 5.644137382507324,
"logps/chosen": -0.6714180111885071,
"logps/rejected": -0.9587985277175903,
"loss": 0.732,
"nll_loss": 0.6440631151199341,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.033570900559425354,
"rewards/margins": 0.014369020238518715,
"rewards/rejected": -0.047939930111169815,
"step": 140
},
{
"epoch": 0.69,
"grad_norm": 2.375,
"learning_rate": 4.082482904638631e-06,
"log_odds_chosen": 0.5068908929824829,
"log_odds_ratio": -0.604145884513855,
"logits/chosen": 5.474297523498535,
"logits/rejected": 5.376832485198975,
"logps/chosen": -0.8357957005500793,
"logps/rejected": -1.0136160850524902,
"loss": 0.706,
"nll_loss": 0.6737378835678101,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.04178978502750397,
"rewards/margins": 0.00889101903885603,
"rewards/rejected": -0.05068080872297287,
"step": 150
},
{
"epoch": 0.74,
"grad_norm": 2.015625,
"learning_rate": 3.952847075210474e-06,
"log_odds_chosen": 0.615983784198761,
"log_odds_ratio": -0.4876289963722229,
"logits/chosen": 5.473410129547119,
"logits/rejected": 6.06318998336792,
"logps/chosen": -0.9676389694213867,
"logps/rejected": -1.349385142326355,
"loss": 0.6996,
"nll_loss": 0.6852242350578308,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.048381954431533813,
"rewards/margins": 0.019087309017777443,
"rewards/rejected": -0.06746925413608551,
"step": 160
},
{
"epoch": 0.78,
"grad_norm": 2.171875,
"learning_rate": 3.834824944236852e-06,
"log_odds_chosen": 0.4551977515220642,
"log_odds_ratio": -0.5428072214126587,
"logits/chosen": 4.785284042358398,
"logits/rejected": 6.005092620849609,
"logps/chosen": -0.7350739240646362,
"logps/rejected": -1.0496256351470947,
"loss": 0.6959,
"nll_loss": 0.5339438319206238,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.03675369173288345,
"rewards/margins": 0.015727588906884193,
"rewards/rejected": -0.052481282502412796,
"step": 170
},
{
"epoch": 0.83,
"grad_norm": 2.015625,
"learning_rate": 3.72677996249965e-06,
"log_odds_chosen": 0.5587902665138245,
"log_odds_ratio": -0.6063727140426636,
"logits/chosen": 4.6595892906188965,
"logits/rejected": 5.4700422286987305,
"logps/chosen": -0.7482207417488098,
"logps/rejected": -0.9887701272964478,
"loss": 0.7233,
"nll_loss": 0.5874465703964233,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.03741103783249855,
"rewards/margins": 0.012027469463646412,
"rewards/rejected": -0.04943850636482239,
"step": 180
},
{
"epoch": 0.88,
"grad_norm": 2.0625,
"learning_rate": 3.6273812505500587e-06,
"log_odds_chosen": 0.9965683817863464,
"log_odds_ratio": -0.4162277281284332,
"logits/chosen": 5.304540157318115,
"logits/rejected": 5.486930847167969,
"logps/chosen": -0.7579169869422913,
"logps/rejected": -1.1843591928482056,
"loss": 0.7298,
"nll_loss": 0.6787526607513428,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.03789585083723068,
"rewards/margins": 0.021322116255760193,
"rewards/rejected": -0.059217967092990875,
"step": 190
},
{
"epoch": 0.92,
"grad_norm": 1.9609375,
"learning_rate": 3.5355339059327378e-06,
"log_odds_chosen": 0.2911016047000885,
"log_odds_ratio": -0.6208275556564331,
"logits/chosen": 5.865508556365967,
"logits/rejected": 5.9140448570251465,
"logps/chosen": -1.0318800210952759,
"logps/rejected": -1.2233208417892456,
"loss": 0.6888,
"nll_loss": 0.8277570009231567,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.051594000309705734,
"rewards/margins": 0.009572046808898449,
"rewards/rejected": -0.061166055500507355,
"step": 200
},
{
"epoch": 0.97,
"grad_norm": 2.546875,
"learning_rate": 3.450327796711771e-06,
"log_odds_chosen": 0.3929597735404968,
"log_odds_ratio": -0.6252869367599487,
"logits/chosen": 5.480368137359619,
"logits/rejected": 5.818605899810791,
"logps/chosen": -0.8382253646850586,
"logps/rejected": -1.1194109916687012,
"loss": 0.703,
"nll_loss": 0.7914389967918396,
"rewards/accuracies": 0.5,
"rewards/chosen": -0.04191126674413681,
"rewards/margins": 0.014059278182685375,
"rewards/rejected": -0.05597054958343506,
"step": 210
},
{
"epoch": 1.01,
"grad_norm": 2.234375,
"learning_rate": 3.3709993123162106e-06,
"log_odds_chosen": 1.1686198711395264,
"log_odds_ratio": -0.39844751358032227,
"logits/chosen": 4.818378448486328,
"logits/rejected": 5.660789966583252,
"logps/chosen": -0.5040851831436157,
"logps/rejected": -0.9685913324356079,
"loss": 0.6554,
"nll_loss": 0.49605101346969604,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.025204259902238846,
"rewards/margins": 0.02322530373930931,
"rewards/rejected": -0.04842956364154816,
"step": 220
},
{
"epoch": 1.06,
"grad_norm": 2.046875,
"learning_rate": 3.296902366978936e-06,
"log_odds_chosen": 0.7159255743026733,
"log_odds_ratio": -0.5276229977607727,
"logits/chosen": 4.3275017738342285,
"logits/rejected": 5.1829423904418945,
"logps/chosen": -0.7593253254890442,
"logps/rejected": -1.0148638486862183,
"loss": 0.6289,
"nll_loss": 0.609928548336029,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.03796626627445221,
"rewards/margins": 0.012776928022503853,
"rewards/rejected": -0.05074319988489151,
"step": 230
},
{
"epoch": 1.11,
"grad_norm": 2.375,
"learning_rate": 3.2274861218395142e-06,
"log_odds_chosen": 0.7326894998550415,
"log_odds_ratio": -0.5214331150054932,
"logits/chosen": 4.783654689788818,
"logits/rejected": 5.283537864685059,
"logps/chosen": -0.7465990781784058,
"logps/rejected": -0.9910147786140442,
"loss": 0.6382,
"nll_loss": 0.7347540855407715,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.03732995316386223,
"rewards/margins": 0.012220785021781921,
"rewards/rejected": -0.04955074191093445,
"step": 240
},
{
"epoch": 1.15,
"grad_norm": 2.0625,
"learning_rate": 3.1622776601683796e-06,
"log_odds_chosen": 0.040362291038036346,
"log_odds_ratio": -0.7654204964637756,
"logits/chosen": 4.929324150085449,
"logits/rejected": 4.940483570098877,
"logps/chosen": -0.939703106880188,
"logps/rejected": -0.9395262598991394,
"loss": 0.6626,
"nll_loss": 0.7169132232666016,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.04698516055941582,
"rewards/margins": -8.843839168548584e-06,
"rewards/rejected": -0.04697632044553757,
"step": 250
},
{
"epoch": 1.2,
"grad_norm": 2.59375,
"learning_rate": 3.1008683647302113e-06,
"log_odds_chosen": 0.8304751515388489,
"log_odds_ratio": -0.4627406597137451,
"logits/chosen": 4.34907341003418,
"logits/rejected": 4.541801929473877,
"logps/chosen": -0.7797168493270874,
"logps/rejected": -1.0878037214279175,
"loss": 0.6408,
"nll_loss": 0.6424815058708191,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.03898584097623825,
"rewards/margins": 0.015404346399009228,
"rewards/rejected": -0.05439019203186035,
"step": 260
},
{
"epoch": 1.24,
"grad_norm": 2.328125,
"learning_rate": 3.0429030972509227e-06,
"log_odds_chosen": 0.2547241747379303,
"log_odds_ratio": -0.7041358351707458,
"logits/chosen": 4.1212077140808105,
"logits/rejected": 5.139257431030273,
"logps/chosen": -0.5988011360168457,
"logps/rejected": -0.7647382020950317,
"loss": 0.6441,
"nll_loss": 0.4384763836860657,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.029940057545900345,
"rewards/margins": 0.008296851068735123,
"rewards/rejected": -0.03823690861463547,
"step": 270
},
{
"epoch": 1.29,
"grad_norm": 1.9609375,
"learning_rate": 2.988071523335984e-06,
"log_odds_chosen": 0.7432643175125122,
"log_odds_ratio": -0.4928904175758362,
"logits/chosen": 4.240169525146484,
"logits/rejected": 4.746310234069824,
"logps/chosen": -0.7583116292953491,
"logps/rejected": -1.0217373371124268,
"loss": 0.6349,
"nll_loss": 0.5912537574768066,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.03791557624936104,
"rewards/margins": 0.013171288184821606,
"rewards/rejected": -0.05108686536550522,
"step": 280
},
{
"epoch": 1.34,
"grad_norm": 2.5,
"learning_rate": 2.9361010975735177e-06,
"log_odds_chosen": 0.6404408812522888,
"log_odds_ratio": -0.5461726784706116,
"logits/chosen": 4.347890377044678,
"logits/rejected": 5.2955708503723145,
"logps/chosen": -0.8145158886909485,
"logps/rejected": -1.124975323677063,
"loss": 0.6204,
"nll_loss": 0.5651360154151917,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.040725789964199066,
"rewards/margins": 0.015522971749305725,
"rewards/rejected": -0.05624876171350479,
"step": 290
},
{
"epoch": 1.38,
"grad_norm": 1.96875,
"learning_rate": 2.8867513459481293e-06,
"log_odds_chosen": 0.4704459607601166,
"log_odds_ratio": -0.6623938083648682,
"logits/chosen": 4.255876064300537,
"logits/rejected": 5.063040733337402,
"logps/chosen": -0.7718355059623718,
"logps/rejected": -1.144460916519165,
"loss": 0.6404,
"nll_loss": 0.6724303364753723,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.03859177231788635,
"rewards/margins": 0.01863126829266548,
"rewards/rejected": -0.05722304433584213,
"step": 300
},
{
"epoch": 1.43,
"grad_norm": 2.453125,
"learning_rate": 2.839809171235324e-06,
"log_odds_chosen": 1.5952459573745728,
"log_odds_ratio": -0.2707791328430176,
"logits/chosen": 2.7694969177246094,
"logits/rejected": 5.479510307312012,
"logps/chosen": -0.4962679445743561,
"logps/rejected": -1.2316776514053345,
"loss": 0.6428,
"nll_loss": 0.3623020648956299,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.024813394993543625,
"rewards/margins": 0.03677048534154892,
"rewards/rejected": -0.06158388406038284,
"step": 310
},
{
"epoch": 1.47,
"grad_norm": 2.078125,
"learning_rate": 2.7950849718747376e-06,
"log_odds_chosen": 0.4402007460594177,
"log_odds_ratio": -0.5388344526290894,
"logits/chosen": 4.8701372146606445,
"logits/rejected": 4.049181938171387,
"logps/chosen": -0.8427563905715942,
"logps/rejected": -1.1280080080032349,
"loss": 0.6661,
"nll_loss": 0.6774541735649109,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.04213782027363777,
"rewards/margins": 0.014262576587498188,
"rewards/rejected": -0.056400395929813385,
"step": 320
},
{
"epoch": 1.52,
"grad_norm": 1.9765625,
"learning_rate": 2.752409412815902e-06,
"log_odds_chosen": 1.4536019563674927,
"log_odds_ratio": -0.3178521990776062,
"logits/chosen": 4.046222686767578,
"logits/rejected": 4.855486869812012,
"logps/chosen": -0.4614998400211334,
"logps/rejected": -1.0025476217269897,
"loss": 0.6396,
"nll_loss": 0.46759381890296936,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.02307499200105667,
"rewards/margins": 0.027052391320466995,
"rewards/rejected": -0.05012737959623337,
"step": 330
},
{
"epoch": 1.57,
"grad_norm": 2.65625,
"learning_rate": 2.711630722733202e-06,
"log_odds_chosen": 0.4552677273750305,
"log_odds_ratio": -0.5441101789474487,
"logits/chosen": 4.233187198638916,
"logits/rejected": 4.776756286621094,
"logps/chosen": -0.9984881281852722,
"logps/rejected": -1.3039405345916748,
"loss": 0.6326,
"nll_loss": 0.7266319990158081,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.04992440715432167,
"rewards/margins": 0.01527262944728136,
"rewards/rejected": -0.06519703567028046,
"step": 340
},
{
"epoch": 1.61,
"grad_norm": 1.9609375,
"learning_rate": 2.6726124191242444e-06,
"log_odds_chosen": 0.3951299488544464,
"log_odds_ratio": -0.6442996263504028,
"logits/chosen": 4.592418193817139,
"logits/rejected": 4.885247707366943,
"logps/chosen": -0.9690208435058594,
"logps/rejected": -1.1191128492355347,
"loss": 0.6271,
"nll_loss": 0.7028160095214844,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.04845104366540909,
"rewards/margins": 0.007504602428525686,
"rewards/rejected": -0.055955640971660614,
"step": 350
},
{
"epoch": 1.66,
"grad_norm": 2.109375,
"learning_rate": 2.6352313834736496e-06,
"log_odds_chosen": 0.6397253274917603,
"log_odds_ratio": -0.4948647916316986,
"logits/chosen": 3.1035220623016357,
"logits/rejected": 4.4074320793151855,
"logps/chosen": -0.7063679695129395,
"logps/rejected": -1.086042881011963,
"loss": 0.6133,
"nll_loss": 0.5209956765174866,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.03531839698553085,
"rewards/margins": 0.0189837496727705,
"rewards/rejected": -0.054302144795656204,
"step": 360
},
{
"epoch": 1.71,
"grad_norm": 1.9296875,
"learning_rate": 2.599376224550182e-06,
"log_odds_chosen": 0.5072129368782043,
"log_odds_ratio": -0.5375211834907532,
"logits/chosen": 4.4618144035339355,
"logits/rejected": 4.897726535797119,
"logps/chosen": -0.8658114671707153,
"logps/rejected": -1.161678433418274,
"loss": 0.625,
"nll_loss": 0.7147814035415649,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.043290577828884125,
"rewards/margins": 0.014793348498642445,
"rewards/rejected": -0.058083921670913696,
"step": 370
},
{
"epoch": 1.75,
"grad_norm": 2.28125,
"learning_rate": 2.564945880212886e-06,
"log_odds_chosen": 0.5736058950424194,
"log_odds_ratio": -0.4948197305202484,
"logits/chosen": 4.31764554977417,
"logits/rejected": 4.153486251831055,
"logps/chosen": -0.8540223836898804,
"logps/rejected": -1.1471771001815796,
"loss": 0.6393,
"nll_loss": 0.6763076186180115,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.0427011176943779,
"rewards/margins": 0.014657735824584961,
"rewards/rejected": -0.05735884979367256,
"step": 380
},
{
"epoch": 1.8,
"grad_norm": 3.640625,
"learning_rate": 2.5318484177091667e-06,
"log_odds_chosen": 0.8381564021110535,
"log_odds_ratio": -0.5308811068534851,
"logits/chosen": 4.037534236907959,
"logits/rejected": 5.888669013977051,
"logps/chosen": -0.700161337852478,
"logps/rejected": -1.2042081356048584,
"loss": 0.6318,
"nll_loss": 0.5512461066246033,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.03500806540250778,
"rewards/margins": 0.025202345103025436,
"rewards/rejected": -0.060210417956113815,
"step": 390
},
{
"epoch": 1.84,
"grad_norm": 2.171875,
"learning_rate": 2.5e-06,
"log_odds_chosen": 0.7038768529891968,
"log_odds_ratio": -0.43052348494529724,
"logits/chosen": 3.822885036468506,
"logits/rejected": 4.210227012634277,
"logps/chosen": -0.6150542497634888,
"logps/rejected": -0.9889954328536987,
"loss": 0.6218,
"nll_loss": 0.5013046264648438,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.030752714723348618,
"rewards/margins": 0.01869705691933632,
"rewards/rejected": -0.04944976791739464,
"step": 400
},
{
"epoch": 1.89,
"grad_norm": 2.03125,
"learning_rate": 2.4693239916239746e-06,
"log_odds_chosen": 0.49417972564697266,
"log_odds_ratio": -0.5454962253570557,
"logits/chosen": 3.7158710956573486,
"logits/rejected": 4.625822067260742,
"logps/chosen": -0.7136448621749878,
"logps/rejected": -0.9806584119796753,
"loss": 0.6163,
"nll_loss": 0.5766875147819519,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.03568224236369133,
"rewards/margins": 0.013350683264434338,
"rewards/rejected": -0.04903292655944824,
"step": 410
},
{
"epoch": 1.94,
"grad_norm": 1.96875,
"learning_rate": 2.4397501823713327e-06,
"log_odds_chosen": 1.2905668020248413,
"log_odds_ratio": -0.3054632544517517,
"logits/chosen": 4.375031471252441,
"logits/rejected": 5.165828704833984,
"logps/chosen": -0.6634560823440552,
"logps/rejected": -1.2297804355621338,
"loss": 0.6299,
"nll_loss": 0.5654190182685852,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.03317280486226082,
"rewards/margins": 0.02831621840596199,
"rewards/rejected": -0.06148902326822281,
"step": 420
},
{
"epoch": 1.98,
"grad_norm": 2.234375,
"learning_rate": 2.411214110852061e-06,
"log_odds_chosen": 0.4614163041114807,
"log_odds_ratio": -0.5477044582366943,
"logits/chosen": 3.945091724395752,
"logits/rejected": 4.783943176269531,
"logps/chosen": -0.670985758304596,
"logps/rejected": -0.8528381586074829,
"loss": 0.6328,
"nll_loss": 0.5353778004646301,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.03354928642511368,
"rewards/margins": 0.009092616848647594,
"rewards/rejected": -0.04264190047979355,
"step": 430
},
{
"epoch": 2.03,
"grad_norm": 2.140625,
"learning_rate": 2.3836564731139807e-06,
"log_odds_chosen": 0.519318699836731,
"log_odds_ratio": -0.5034213066101074,
"logits/chosen": 3.990828037261963,
"logits/rejected": 4.283727645874023,
"logps/chosen": -0.7843809723854065,
"logps/rejected": -1.1084554195404053,
"loss": 0.598,
"nll_loss": 0.6064985394477844,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.03921904414892197,
"rewards/margins": 0.01620371639728546,
"rewards/rejected": -0.055422764271497726,
"step": 440
},
{
"epoch": 2.07,
"grad_norm": 2.015625,
"learning_rate": 2.357022603955159e-06,
"log_odds_chosen": 1.2161670923233032,
"log_odds_ratio": -0.5558447241783142,
"logits/chosen": 2.7631869316101074,
"logits/rejected": 4.014997959136963,
"logps/chosen": -0.4891352653503418,
"logps/rejected": -1.057556390762329,
"loss": 0.6063,
"nll_loss": 0.5005042552947998,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.02445676364004612,
"rewards/margins": 0.028421055525541306,
"rewards/rejected": -0.052877821028232574,
"step": 450
},
{
"epoch": 2.12,
"grad_norm": 2.0625,
"learning_rate": 2.3312620206007847e-06,
"log_odds_chosen": 0.8278636932373047,
"log_odds_ratio": -0.43884754180908203,
"logits/chosen": 4.009448051452637,
"logits/rejected": 4.671367645263672,
"logps/chosen": -0.7134698629379272,
"logps/rejected": -1.146784782409668,
"loss": 0.5862,
"nll_loss": 0.5619599223136902,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.03567349165678024,
"rewards/margins": 0.021665748208761215,
"rewards/rejected": -0.05733924359083176,
"step": 460
},
{
"epoch": 2.17,
"grad_norm": 2.609375,
"learning_rate": 2.3063280200722128e-06,
"log_odds_chosen": 1.677671194076538,
"log_odds_ratio": -0.2895694375038147,
"logits/chosen": 2.985790491104126,
"logits/rejected": 4.190914630889893,
"logps/chosen": -0.5018793344497681,
"logps/rejected": -1.0572091341018677,
"loss": 0.5765,
"nll_loss": 0.5001329183578491,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.025093963369727135,
"rewards/margins": 0.02776649035513401,
"rewards/rejected": -0.05286044999957085,
"step": 470
},
{
"epoch": 2.21,
"grad_norm": 2.0,
"learning_rate": 2.2821773229381924e-06,
"log_odds_chosen": 1.0791471004486084,
"log_odds_ratio": -0.37350553274154663,
"logits/chosen": 3.676426649093628,
"logits/rejected": 3.8374907970428467,
"logps/chosen": -0.7438164353370667,
"logps/rejected": -1.29355788230896,
"loss": 0.5652,
"nll_loss": 0.6555451154708862,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.037190817296504974,
"rewards/margins": 0.027487074956297874,
"rewards/rejected": -0.0646779015660286,
"step": 480
},
{
"epoch": 2.26,
"grad_norm": 2.21875,
"learning_rate": 2.2587697572631284e-06,
"log_odds_chosen": 0.275502473115921,
"log_odds_ratio": -0.7135687470436096,
"logits/chosen": 4.321534156799316,
"logits/rejected": 4.41732120513916,
"logps/chosen": -0.9727070927619934,
"logps/rejected": -1.0810346603393555,
"loss": 0.5952,
"nll_loss": 0.7110171914100647,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.04863535612821579,
"rewards/margins": 0.005416377447545528,
"rewards/rejected": -0.05405173450708389,
"step": 490
},
{
"epoch": 2.3,
"grad_norm": 2.15625,
"learning_rate": 2.23606797749979e-06,
"log_odds_chosen": 0.34863442182540894,
"log_odds_ratio": -0.6463712453842163,
"logits/chosen": 4.6876606941223145,
"logits/rejected": 5.054124355316162,
"logps/chosen": -0.9338000416755676,
"logps/rejected": -1.1037800312042236,
"loss": 0.5953,
"nll_loss": 0.8528131246566772,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.04669000208377838,
"rewards/margins": 0.008499005809426308,
"rewards/rejected": -0.05518900603055954,
"step": 500
},
{
"epoch": 2.35,
"grad_norm": 2.171875,
"learning_rate": 2.2140372138502386e-06,
"log_odds_chosen": 0.9548345804214478,
"log_odds_ratio": -0.39882007241249084,
"logits/chosen": 3.5289406776428223,
"logits/rejected": 3.8287463188171387,
"logps/chosen": -0.6570809483528137,
"logps/rejected": -1.1388274431228638,
"loss": 0.609,
"nll_loss": 0.5968061685562134,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.032854050397872925,
"rewards/margins": 0.024087321013212204,
"rewards/rejected": -0.05694136768579483,
"step": 510
},
{
"epoch": 2.4,
"grad_norm": 1.9609375,
"learning_rate": 2.1926450482675734e-06,
"log_odds_chosen": 0.4539831280708313,
"log_odds_ratio": -0.5872747302055359,
"logits/chosen": 3.2061939239501953,
"logits/rejected": 4.589787006378174,
"logps/chosen": -0.7979894280433655,
"logps/rejected": -1.0285401344299316,
"loss": 0.5827,
"nll_loss": 0.6084668636322021,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.039899468421936035,
"rewards/margins": 0.011527536436915398,
"rewards/rejected": -0.051426999270915985,
"step": 520
},
{
"epoch": 2.44,
"grad_norm": 2.484375,
"learning_rate": 2.1718612138153473e-06,
"log_odds_chosen": 0.8493059277534485,
"log_odds_ratio": -0.6372500658035278,
"logits/chosen": 3.078615665435791,
"logits/rejected": 4.099945068359375,
"logps/chosen": -0.6704202890396118,
"logps/rejected": -0.7899671792984009,
"loss": 0.5788,
"nll_loss": 0.5733928084373474,
"rewards/accuracies": 0.5,
"rewards/chosen": -0.03352101519703865,
"rewards/margins": 0.005977341439574957,
"rewards/rejected": -0.039498358964920044,
"step": 530
},
{
"epoch": 2.49,
"grad_norm": 1.859375,
"learning_rate": 2.151657414559676e-06,
"log_odds_chosen": 0.6374627351760864,
"log_odds_ratio": -0.5592355728149414,
"logits/chosen": 3.680483341217041,
"logits/rejected": 3.9816291332244873,
"logps/chosen": -0.8559755086898804,
"logps/rejected": -1.1612054109573364,
"loss": 0.6003,
"nll_loss": 0.6403124928474426,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.04279877990484238,
"rewards/margins": 0.015261486172676086,
"rewards/rejected": -0.05806026607751846,
"step": 540
},
{
"epoch": 2.53,
"grad_norm": 1.8984375,
"learning_rate": 2.132007163556104e-06,
"log_odds_chosen": 1.399209976196289,
"log_odds_ratio": -0.5735031366348267,
"logits/chosen": 3.132289171218872,
"logits/rejected": 3.5427193641662598,
"logps/chosen": -0.5963010191917419,
"logps/rejected": -0.9639393091201782,
"loss": 0.5984,
"nll_loss": 0.5058175325393677,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.029815051704645157,
"rewards/margins": 0.018381912261247635,
"rewards/rejected": -0.04819696769118309,
"step": 550
},
{
"epoch": 2.58,
"grad_norm": 1.859375,
"learning_rate": 2.1128856368212917e-06,
"log_odds_chosen": 0.688880443572998,
"log_odds_ratio": -0.4902462959289551,
"logits/chosen": 2.6950721740722656,
"logits/rejected": 3.1528286933898926,
"logps/chosen": -0.6383022665977478,
"logps/rejected": -0.9691828489303589,
"loss": 0.5718,
"nll_loss": 0.4289799630641937,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.03191510960459709,
"rewards/margins": 0.016544032841920853,
"rewards/rejected": -0.048459142446517944,
"step": 560
},
{
"epoch": 2.63,
"grad_norm": 2.421875,
"learning_rate": 2.0942695414584777e-06,
"log_odds_chosen": 1.3283271789550781,
"log_odds_ratio": -0.3012233078479767,
"logits/chosen": 3.4564871788024902,
"logits/rejected": 4.7043867111206055,
"logps/chosen": -0.6779360771179199,
"logps/rejected": -1.523970365524292,
"loss": 0.6138,
"nll_loss": 0.5768535137176514,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.033896803855895996,
"rewards/margins": 0.042301714420318604,
"rewards/rejected": -0.0761985182762146,
"step": 570
},
{
"epoch": 2.67,
"grad_norm": 1.953125,
"learning_rate": 2.0761369963434992e-06,
"log_odds_chosen": 1.4566174745559692,
"log_odds_ratio": -0.32581037282943726,
"logits/chosen": 2.691676616668701,
"logits/rejected": 4.661564826965332,
"logps/chosen": -0.4493564963340759,
"logps/rejected": -1.0139671564102173,
"loss": 0.5782,
"nll_loss": 0.37120580673217773,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.022467825561761856,
"rewards/margins": 0.028230536729097366,
"rewards/rejected": -0.05069836229085922,
"step": 580
},
{
"epoch": 2.72,
"grad_norm": 2.0625,
"learning_rate": 2.058467423981546e-06,
"log_odds_chosen": 1.0190517902374268,
"log_odds_ratio": -0.5730624198913574,
"logits/chosen": 3.407086133956909,
"logits/rejected": 4.482596397399902,
"logps/chosen": -0.7345553040504456,
"logps/rejected": -0.9309635162353516,
"loss": 0.5723,
"nll_loss": 0.5519307851791382,
"rewards/accuracies": 0.5,
"rewards/chosen": -0.03672776371240616,
"rewards/margins": 0.009820410050451756,
"rewards/rejected": -0.04654817283153534,
"step": 590
},
{
"epoch": 2.76,
"grad_norm": 2.375,
"learning_rate": 2.0412414523193154e-06,
"log_odds_chosen": 1.107779860496521,
"log_odds_ratio": -0.40593117475509644,
"logits/chosen": 3.215078830718994,
"logits/rejected": 4.503358840942383,
"logps/chosen": -0.663019597530365,
"logps/rejected": -1.2786920070648193,
"loss": 0.5815,
"nll_loss": 0.5633824467658997,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.03315097838640213,
"rewards/margins": 0.030783619731664658,
"rewards/rejected": -0.06393460184335709,
"step": 600
},
{
"epoch": 2.81,
"grad_norm": 2.09375,
"learning_rate": 2.0244408254472904e-06,
"log_odds_chosen": 0.7602224349975586,
"log_odds_ratio": -0.5018362998962402,
"logits/chosen": 3.604353666305542,
"logits/rejected": 4.481316089630127,
"logps/chosen": -0.7105517387390137,
"logps/rejected": -1.0740478038787842,
"loss": 0.5873,
"nll_loss": 0.5312780737876892,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.035527586936950684,
"rewards/margins": 0.018174810335040092,
"rewards/rejected": -0.05370239168405533,
"step": 610
},
{
"epoch": 2.86,
"grad_norm": 1.90625,
"learning_rate": 2.0080483222562476e-06,
"log_odds_chosen": 1.3286904096603394,
"log_odds_ratio": -0.36574870347976685,
"logits/chosen": 3.620469331741333,
"logits/rejected": 4.373411655426025,
"logps/chosen": -0.4990506172180176,
"logps/rejected": -0.953050971031189,
"loss": 0.5716,
"nll_loss": 0.5527733564376831,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.024952532723546028,
"rewards/margins": 0.022700021043419838,
"rewards/rejected": -0.047652553766965866,
"step": 620
},
{
"epoch": 2.9,
"grad_norm": 2.359375,
"learning_rate": 1.9920476822239895e-06,
"log_odds_chosen": 0.4847317636013031,
"log_odds_ratio": -0.5640643835067749,
"logits/chosen": 3.125113010406494,
"logits/rejected": 3.340205669403076,
"logps/chosen": -0.8360971212387085,
"logps/rejected": -1.0480194091796875,
"loss": 0.5738,
"nll_loss": 0.6136351823806763,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.041804857552051544,
"rewards/margins": 0.010596117004752159,
"rewards/rejected": -0.05240097641944885,
"step": 630
},
{
"epoch": 2.95,
"grad_norm": 2.09375,
"learning_rate": 1.976423537605237e-06,
"log_odds_chosen": 0.8931509256362915,
"log_odds_ratio": -0.40087467432022095,
"logits/chosen": 3.574153423309326,
"logits/rejected": 4.537802219390869,
"logps/chosen": -0.6440940499305725,
"logps/rejected": -1.088226556777954,
"loss": 0.5846,
"nll_loss": 0.5598152875900269,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.032204702496528625,
"rewards/margins": 0.0222066268324852,
"rewards/rejected": -0.054411329329013824,
"step": 640
},
{
"epoch": 3.0,
"grad_norm": 2.5625,
"learning_rate": 1.961161351381841e-06,
"log_odds_chosen": 1.2053475379943848,
"log_odds_ratio": -0.430248886346817,
"logits/chosen": 2.245370388031006,
"logits/rejected": 3.5309462547302246,
"logps/chosen": -0.5642444491386414,
"logps/rejected": -0.9910544157028198,
"loss": 0.5605,
"nll_loss": 0.45886915922164917,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.028212225064635277,
"rewards/margins": 0.021340493112802505,
"rewards/rejected": -0.04955272004008293,
"step": 650
},
{
"epoch": 3.0,
"step": 651,
"total_flos": 0.0,
"train_loss": 0.812556631554107,
"train_runtime": 4771.9621,
"train_samples_per_second": 4.358,
"train_steps_per_second": 0.136
}
],
"logging_steps": 10,
"max_steps": 651,
"num_input_tokens_seen": 0,
"num_train_epochs": 3,
"save_steps": 500,
"total_flos": 0.0,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}