{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.995276334435522,
  "eval_steps": 50,
  "global_step": 352,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "dpo_loss": 0.6931471824645996,
      "epoch": 0.005668398677373642,
      "grad_norm": 13.413637591400326,
      "learning_rate": 2.7777777777777776e-09,
      "logits": -1.3147305250167847,
      "logps": -88.0877456665039,
      "loss": 0.6931,
      "objective": 0.6931471824645996,
      "ranking_idealized": 0.625,
      "ranking_idealized_expo": 0.5208333134651184,
      "ranking_simple": 0.5208333134651184,
      "regularize": 0.6931471824645996,
      "step": 1
    },
    {
      "dpo_loss": 0.6931517124176025,
      "epoch": 0.02834199338686821,
      "grad_norm": 13.314796352051717,
      "learning_rate": 1.3888888888888889e-08,
      "logits": -1.3678232431411743,
      "logps": -84.45550537109375,
      "loss": 0.6931,
      "objective": 0.6931517124176025,
      "ranking_idealized": 0.6145833134651184,
      "ranking_idealized_expo": 0.546875,
      "ranking_simple": 0.546875,
      "regularize": 0.6931517124176025,
      "step": 5
    },
    {
      "dpo_loss": 0.6927531957626343,
      "epoch": 0.05668398677373642,
      "grad_norm": 13.032412874817698,
      "learning_rate": 2.7777777777777777e-08,
      "logits": -1.4463595151901245,
      "logps": -83.46998596191406,
      "loss": 0.6929,
      "objective": 0.6927531957626343,
      "ranking_idealized": 0.6000000238418579,
      "ranking_idealized_expo": 0.512499988079071,
      "ranking_simple": 0.5083333253860474,
      "regularize": 0.6927531957626343,
      "step": 10
    },
    {
      "dpo_loss": 0.6928622722625732,
      "epoch": 0.08502598016060463,
      "grad_norm": 12.566734219683084,
      "learning_rate": 4.166666666666667e-08,
      "logits": -1.4098025560379028,
      "logps": -83.67684936523438,
      "loss": 0.6927,
      "objective": 0.6928622722625732,
      "ranking_idealized": 0.5833333134651184,
      "ranking_idealized_expo": 0.5041666626930237,
      "ranking_simple": 0.5083333253860474,
      "regularize": 0.6928622722625732,
      "step": 15
    },
    {
      "dpo_loss": 0.6923068761825562,
      "epoch": 0.11336797354747284,
      "grad_norm": 13.25147394148046,
      "learning_rate": 5.5555555555555555e-08,
      "logits": -1.3977584838867188,
      "logps": -84.3187484741211,
      "loss": 0.6919,
      "objective": 0.6923068761825562,
      "ranking_idealized": 0.625,
      "ranking_idealized_expo": 0.5375000238418579,
      "ranking_simple": 0.5375000238418579,
      "regularize": 0.6923068761825562,
      "step": 20
    },
    {
      "dpo_loss": 0.6906546354293823,
      "epoch": 0.14170996693434104,
      "grad_norm": 12.656689096549028,
      "learning_rate": 6.944444444444444e-08,
      "logits": -1.4436664581298828,
      "logps": -84.10287475585938,
      "loss": 0.6912,
      "objective": 0.6906546354293823,
      "ranking_idealized": 0.5958333611488342,
      "ranking_idealized_expo": 0.4958333373069763,
      "ranking_simple": 0.4958333373069763,
      "regularize": 0.6906546354293823,
      "step": 25
    },
    {
      "dpo_loss": 0.6908212900161743,
      "epoch": 0.17005196032120926,
      "grad_norm": 13.15545428739909,
      "learning_rate": 8.333333333333334e-08,
      "logits": -1.4165135622024536,
      "logps": -84.5628433227539,
      "loss": 0.6902,
      "objective": 0.6908212900161743,
      "ranking_idealized": 0.6041666865348816,
      "ranking_idealized_expo": 0.5416666865348816,
      "ranking_simple": 0.5375000238418579,
      "regularize": 0.6908212900161743,
      "step": 30
    },
    {
      "dpo_loss": 0.6884526610374451,
      "epoch": 0.19839395370807747,
      "grad_norm": 12.961038019497309,
      "learning_rate": 9.722222222222221e-08,
      "logits": -1.3882626295089722,
      "logps": -83.0355453491211,
      "loss": 0.6879,
      "objective": 0.6884526610374451,
      "ranking_idealized": 0.6166666746139526,
      "ranking_idealized_expo": 0.5041666626930237,
      "ranking_simple": 0.5083333253860474,
      "regularize": 0.6884526610374451,
      "step": 35
    },
    {
      "dpo_loss": 0.6872146129608154,
      "epoch": 0.22673594709494568,
      "grad_norm": 15.265896606535208,
      "learning_rate": 9.996046986136509e-08,
      "logits": -1.3777297735214233,
      "logps": -83.15966796875,
      "loss": 0.6889,
      "objective": 0.6872146129608154,
      "ranking_idealized": 0.6000000238418579,
      "ranking_idealized_expo": 0.5208333134651184,
      "ranking_simple": 0.5166666507720947,
      "regularize": 0.6872146129608154,
      "step": 40
    },
    {
      "dpo_loss": 0.6848792433738708,
      "epoch": 0.25507794048181387,
      "grad_norm": 13.166431360837958,
      "learning_rate": 9.979998579289983e-08,
      "logits": -1.5064284801483154,
      "logps": -84.50239562988281,
      "loss": 0.6857,
      "objective": 0.6848792433738708,
      "ranking_idealized": 0.6208333373069763,
      "ranking_idealized_expo": 0.5375000238418579,
      "ranking_simple": 0.5416666865348816,
      "regularize": 0.6848792433738708,
      "step": 45
    },
    {
      "dpo_loss": 0.6835355162620544,
      "epoch": 0.2834199338686821,
      "grad_norm": 12.111837571939198,
      "learning_rate": 9.95164733236251e-08,
      "logits": -1.4091321229934692,
      "logps": -82.50495147705078,
      "loss": 0.6855,
      "objective": 0.6835355162620544,
      "ranking_idealized": 0.5416666865348816,
      "ranking_idealized_expo": 0.42500001192092896,
      "ranking_simple": 0.4333333373069763,
      "regularize": 0.6835355162620544,
      "step": 50
    },
    {
      "epoch": 0.2834199338686821,
      "eval_dpo_loss": 0.6918472647666931,
      "eval_logits": -1.4343044757843018,
      "eval_logps": -90.7668685913086,
      "eval_loss": 0.6888882517814636,
      "eval_objective": 0.6918472647666931,
      "eval_ranking_idealized": 0.5888429880142212,
      "eval_ranking_idealized_expo": 0.5103305578231812,
      "eval_ranking_simple": 0.5103305578231812,
      "eval_regularize": 0.6918472647666931,
      "eval_runtime": 261.146,
      "eval_samples_per_second": 22.172,
      "eval_steps_per_second": 0.927,
      "step": 50
    },
    {
      "dpo_loss": 0.6787654161453247,
      "epoch": 0.3117619272555503,
      "grad_norm": 14.068180692870671,
      "learning_rate": 9.911063285706807e-08,
      "logits": -1.4499430656433105,
      "logps": -84.21847534179688,
      "loss": 0.6825,
      "objective": 0.6787654161453247,
      "ranking_idealized": 0.625,
      "ranking_idealized_expo": 0.5416666865348816,
      "ranking_simple": 0.550000011920929,
      "regularize": 0.6787654161453247,
      "step": 55
    },
    {
      "dpo_loss": 0.6828632950782776,
      "epoch": 0.3401039206424185,
      "grad_norm": 13.070276554702126,
      "learning_rate": 9.85834670020205e-08,
      "logits": -1.4364044666290283,
      "logps": -83.14185333251953,
      "loss": 0.6829,
      "objective": 0.6828632950782776,
      "ranking_idealized": 0.637499988079071,
      "ranking_idealized_expo": 0.512499988079071,
      "ranking_simple": 0.5249999761581421,
      "regularize": 0.6828632950782776,
      "step": 60
    },
    {
      "dpo_loss": 0.6835117340087891,
      "epoch": 0.3684459140292867,
      "grad_norm": 13.497944122383998,
      "learning_rate": 9.793627809564323e-08,
      "logits": -1.3944473266601562,
      "logps": -83.09821319580078,
      "loss": 0.6822,
      "objective": 0.6835117340087891,
      "ranking_idealized": 0.5708333253860474,
      "ranking_idealized_expo": 0.5083333253860474,
      "ranking_simple": 0.4958333373069763,
      "regularize": 0.6835117340087891,
      "step": 65
    },
    {
      "dpo_loss": 0.6772111058235168,
      "epoch": 0.39678790741615494,
      "grad_norm": 13.151701497634273,
      "learning_rate": 9.717066498610672e-08,
      "logits": -1.480456829071045,
      "logps": -84.27593231201172,
      "loss": 0.6818,
      "objective": 0.6772111058235168,
      "ranking_idealized": 0.6000000238418579,
      "ranking_idealized_expo": 0.5166666507720947,
      "ranking_simple": 0.5166666507720947,
      "regularize": 0.6772111058235168,
      "step": 70
    },
    {
      "dpo_loss": 0.679884672164917,
      "epoch": 0.42512990080302315,
      "grad_norm": 12.601045745018736,
      "learning_rate": 9.628851908271572e-08,
      "logits": -1.5056648254394531,
      "logps": -84.74910736083984,
      "loss": 0.6776,
      "objective": 0.679884672164917,
      "ranking_idealized": 0.6291666626930237,
      "ranking_idealized_expo": 0.5458333492279053,
      "ranking_simple": 0.5458333492279053,
      "regularize": 0.679884672164917,
      "step": 75
    },
    {
      "dpo_loss": 0.683576762676239,
      "epoch": 0.45347189418989137,
      "grad_norm": 13.218876579787791,
      "learning_rate": 9.529201968327615e-08,
      "logits": -1.550328016281128,
      "logps": -84.59281158447266,
      "loss": 0.6772,
      "objective": 0.683576762676239,
      "ranking_idealized": 0.5833333134651184,
      "ranking_idealized_expo": 0.5166666507720947,
      "ranking_simple": 0.5208333134651184,
      "regularize": 0.683576762676239,
      "step": 80
    },
    {
      "dpo_loss": 0.6882224678993225,
      "epoch": 0.4818138875767596,
      "grad_norm": 12.712021197638556,
      "learning_rate": 9.41836285902478e-08,
      "logits": -1.507049560546875,
      "logps": -82.99642944335938,
      "loss": 0.6781,
      "objective": 0.6882224678993225,
      "ranking_idealized": 0.6000000238418579,
      "ranking_idealized_expo": 0.5249999761581421,
      "ranking_simple": 0.5291666388511658,
      "regularize": 0.6882224678993225,
      "step": 85
    },
    {
      "dpo_loss": 0.6753127574920654,
      "epoch": 0.5101558809636277,
      "grad_norm": 13.425590079214597,
      "learning_rate": 9.296608402898305e-08,
      "logits": -1.3210861682891846,
      "logps": -84.38108825683594,
      "loss": 0.6726,
      "objective": 0.6753127574920654,
      "ranking_idealized": 0.550000011920929,
      "ranking_idealized_expo": 0.4791666567325592,
      "ranking_simple": 0.48750001192092896,
      "regularize": 0.6753127574920654,
      "step": 90
    },
    {
      "dpo_loss": 0.6701774001121521,
      "epoch": 0.538497874350496,
      "grad_norm": 13.30415786759025,
      "learning_rate": 9.164239388307666e-08,
      "logits": -1.4094645977020264,
      "logps": -84.12966918945312,
      "loss": 0.6736,
      "objective": 0.6701774001121521,
      "ranking_idealized": 0.625,
      "ranking_idealized_expo": 0.5416666865348816,
      "ranking_simple": 0.550000011920929,
      "regularize": 0.6701774001121521,
      "step": 95
    },
    {
      "dpo_loss": 0.6689857840538025,
      "epoch": 0.5668398677373642,
      "grad_norm": 14.262472601225529,
      "learning_rate": 9.021582826353824e-08,
      "logits": -1.4942399263381958,
      "logps": -85.0649185180664,
      "loss": 0.6746,
      "objective": 0.6689857840538025,
      "ranking_idealized": 0.612500011920929,
      "ranking_idealized_expo": 0.5416666865348816,
      "ranking_simple": 0.5375000238418579,
      "regularize": 0.6689857840538025,
      "step": 100
    },
    {
      "epoch": 0.5668398677373642,
      "eval_dpo_loss": 0.6899410486221313,
      "eval_logits": -1.476444125175476,
      "eval_logps": -90.9747543334961,
      "eval_loss": 0.6857675909996033,
      "eval_objective": 0.6899410486221313,
      "eval_ranking_idealized": 0.5888429880142212,
      "eval_ranking_idealized_expo": 0.5103305578231812,
      "eval_ranking_simple": 0.5092975497245789,
      "eval_regularize": 0.6899410486221313,
      "eval_runtime": 259.9569,
      "eval_samples_per_second": 22.273,
      "eval_steps_per_second": 0.931,
      "step": 100
    },
    {
      "dpo_loss": 0.6589028239250183,
      "epoch": 0.5951818611242324,
      "grad_norm": 13.550145448172236,
      "learning_rate": 8.868991143014468e-08,
      "logits": -1.3795539140701294,
      "logps": -84.87275695800781,
      "loss": 0.6674,
      "objective": 0.6589028239250183,
      "ranking_idealized": 0.6208333373069763,
      "ranking_idealized_expo": 0.5333333611488342,
      "ranking_simple": 0.5458333492279053,
      "regularize": 0.6589028239250183,
      "step": 105
    },
    {
      "dpo_loss": 0.6661778688430786,
      "epoch": 0.6235238545111006,
      "grad_norm": 15.598239406382287,
      "learning_rate": 8.706841308493092e-08,
      "logits": -1.561141848564148,
      "logps": -83.28865814208984,
      "loss": 0.668,
      "objective": 0.6661778688430786,
      "ranking_idealized": 0.5708333253860474,
      "ranking_idealized_expo": 0.4541666805744171,
      "ranking_simple": 0.46666666865348816,
      "regularize": 0.6661778688430786,
      "step": 110
    },
    {
      "dpo_loss": 0.6700491309165955,
      "epoch": 0.6518658478979689,
      "grad_norm": 14.005634900395668,
      "learning_rate": 8.535533905932736e-08,
      "logits": -1.5146809816360474,
      "logps": -83.37279510498047,
      "loss": 0.6669,
      "objective": 0.6700491309165955,
      "ranking_idealized": 0.574999988079071,
      "ranking_idealized_expo": 0.4833333194255829,
      "ranking_simple": 0.4958333373069763,
      "regularize": 0.6700491309165955,
      "step": 115
    },
    {
      "dpo_loss": 0.6605091691017151,
      "epoch": 0.680207841284837,
      "grad_norm": 13.532342794289065,
      "learning_rate": 8.355492141795184e-08,
      "logits": -1.4583431482315063,
      "logps": -84.31383514404297,
      "loss": 0.6659,
      "objective": 0.6605091691017151,
      "ranking_idealized": 0.5874999761581421,
      "ranking_idealized_expo": 0.5333333611488342,
      "ranking_simple": 0.5375000238418579,
      "regularize": 0.6605091691017151,
      "step": 120
    },
    {
      "dpo_loss": 0.6656448841094971,
      "epoch": 0.7085498346717053,
      "grad_norm": 13.79567400382208,
      "learning_rate": 8.167160800350306e-08,
      "logits": -1.4243894815444946,
      "logps": -83.22982025146484,
      "loss": 0.6688,
      "objective": 0.6656448841094971,
      "ranking_idealized": 0.6291666626930237,
      "ranking_idealized_expo": 0.550000011920929,
      "ranking_simple": 0.5458333492279053,
      "regularize": 0.6656448841094971,
      "step": 125
    },
    {
      "dpo_loss": 0.6677243113517761,
      "epoch": 0.7368918280585735,
      "grad_norm": 13.221986997309093,
      "learning_rate": 7.971005144858553e-08,
      "logits": -1.564219355583191,
      "logps": -82.65666961669922,
      "loss": 0.6701,
      "objective": 0.6677243113517761,
      "ranking_idealized": 0.5416666865348816,
      "ranking_idealized_expo": 0.44999998807907104,
      "ranking_simple": 0.47083333134651184,
      "regularize": 0.6677243113517761,
      "step": 130
    },
    {
      "dpo_loss": 0.6758949160575867,
      "epoch": 0.7652338214454416,
      "grad_norm": 13.535024146063408,
      "learning_rate": 7.767509768161078e-08,
      "logits": -1.5405229330062866,
      "logps": -84.79542541503906,
      "loss": 0.6661,
      "objective": 0.6758949160575867,
      "ranking_idealized": 0.5833333134651184,
      "ranking_idealized_expo": 0.5291666388511658,
      "ranking_simple": 0.5249999761581421,
      "regularize": 0.6758949160575867,
      "step": 135
    },
    {
      "dpo_loss": 0.6768299341201782,
      "epoch": 0.7935758148323099,
      "grad_norm": 13.276044449636085,
      "learning_rate": 7.557177395517111e-08,
      "logits": -1.4452087879180908,
      "logps": -83.74736022949219,
      "loss": 0.6665,
      "objective": 0.6768299341201782,
      "ranking_idealized": 0.5791666507720947,
      "ranking_idealized_expo": 0.5041666626930237,
      "ranking_simple": 0.5249999761581421,
      "regularize": 0.6768299341201782,
      "step": 140
    },
    {
      "dpo_loss": 0.6667078733444214,
      "epoch": 0.821917808219178,
      "grad_norm": 13.106504702831565,
      "learning_rate": 7.340527642646068e-08,
      "logits": -1.522230863571167,
      "logps": -83.9556655883789,
      "loss": 0.6628,
      "objective": 0.6667078733444214,
      "ranking_idealized": 0.5791666507720947,
      "ranking_idealized_expo": 0.4833333194255829,
      "ranking_simple": 0.4833333194255829,
      "regularize": 0.6667078733444214,
      "step": 145
    },
    {
      "dpo_loss": 0.6657266616821289,
      "epoch": 0.8502598016060463,
      "grad_norm": 13.484987339837966,
      "learning_rate": 7.118095732042642e-08,
      "logits": -1.494079351425171,
      "logps": -82.12902069091797,
      "loss": 0.6601,
      "objective": 0.6657266616821289,
      "ranking_idealized": 0.6166666746139526,
      "ranking_idealized_expo": 0.5416666865348816,
      "ranking_simple": 0.5541666746139526,
      "regularize": 0.6657266616821289,
      "step": 150
    },
    {
      "epoch": 0.8502598016060463,
      "eval_dpo_loss": 0.688634991645813,
      "eval_logits": -1.5179173946380615,
      "eval_logps": -90.80626678466797,
      "eval_loss": 0.682818591594696,
      "eval_objective": 0.688634991645813,
      "eval_ranking_idealized": 0.5888429880142212,
      "eval_ranking_idealized_expo": 0.5103305578231812,
      "eval_ranking_simple": 0.5134297609329224,
      "eval_regularize": 0.688634991645813,
      "eval_runtime": 259.9549,
      "eval_samples_per_second": 22.273,
      "eval_steps_per_second": 0.931,
      "step": 150
    },
    {
      "dpo_loss": 0.659411609172821,
      "epoch": 0.8786017949929145,
      "grad_norm": 15.013754194888069,
      "learning_rate": 6.89043117073609e-08,
      "logits": -1.5184494256973267,
      "logps": -82.16848754882812,
      "loss": 0.6586,
      "objective": 0.659411609172821,
      "ranking_idealized": 0.6166666746139526,
      "ranking_idealized_expo": 0.512499988079071,
      "ranking_simple": 0.5416666865348816,
      "regularize": 0.659411609172821,
      "step": 155
    },
    {
      "dpo_loss": 0.660663902759552,
      "epoch": 0.9069437883797827,
      "grad_norm": 13.528075587958016,
      "learning_rate": 6.65809639276034e-08,
      "logits": -1.4307061433792114,
      "logps": -83.12458038330078,
      "loss": 0.666,
      "objective": 0.660663902759552,
      "ranking_idealized": 0.5708333253860474,
      "ranking_idealized_expo": 0.4958333373069763,
      "ranking_simple": 0.512499988079071,
      "regularize": 0.660663902759552,
      "step": 160
    },
    {
      "dpo_loss": 0.6495606303215027,
      "epoch": 0.9352857817666509,
      "grad_norm": 14.079206144678064,
      "learning_rate": 6.4216653696885e-08,
      "logits": -1.4516769647598267,
      "logps": -84.19721984863281,
      "loss": 0.6591,
      "objective": 0.6495606303215027,
      "ranking_idealized": 0.6041666865348816,
      "ranking_idealized_expo": 0.5208333134651184,
      "ranking_simple": 0.5458333492279053,
      "regularize": 0.6495606303215027,
      "step": 165
    },
    {
      "dpo_loss": 0.6589613556861877,
      "epoch": 0.9636277751535192,
      "grad_norm": 15.945719305933032,
      "learning_rate": 6.181722192664524e-08,
      "logits": -1.4703598022460938,
      "logps": -84.60289764404297,
      "loss": 0.6579,
      "objective": 0.6589613556861877,
      "ranking_idealized": 0.5791666507720947,
      "ranking_idealized_expo": 0.512499988079071,
      "ranking_simple": 0.5291666388511658,
      "regularize": 0.6589613556861877,
      "step": 170
    },
    {
      "dpo_loss": 0.6631531119346619,
      "epoch": 0.9919697685403873,
      "grad_norm": 13.762050879048529,
      "learning_rate": 5.9388596294349124e-08,
      "logits": -1.4086403846740723,
      "logps": -85.38468933105469,
      "loss": 0.6613,
      "objective": 0.6631531119346619,
      "ranking_idealized": 0.6541666388511658,
      "ranking_idealized_expo": 0.574999988079071,
      "ranking_simple": 0.5791666507720947,
      "regularize": 0.6631531119346619,
      "step": 175
    },
    {
      "dpo_loss": 0.6530786156654358,
      "epoch": 1.0203117619272555,
      "grad_norm": 15.08770015055669,
      "learning_rate": 5.693677659945342e-08,
      "logits": -1.6005927324295044,
      "logps": -84.76131439208984,
      "loss": 0.6531,
      "objective": 0.6530786156654358,
      "ranking_idealized": 0.6166666746139526,
      "ranking_idealized_expo": 0.5416666865348816,
      "ranking_simple": 0.550000011920929,
      "regularize": 0.6530786156654358,
      "step": 180
    },
    {
      "dpo_loss": 0.6419151425361633,
      "epoch": 1.0486537553141237,
      "grad_norm": 14.657084961337553,
      "learning_rate": 5.446781994119886e-08,
      "logits": -1.4578330516815186,
      "logps": -84.88520812988281,
      "loss": 0.6472,
      "objective": 0.6419151425361633,
      "ranking_idealized": 0.5791666507720947,
      "ranking_idealized_expo": 0.4833333194255829,
      "ranking_simple": 0.5166666507720947,
      "regularize": 0.6419151425361633,
      "step": 185
    },
    {
      "dpo_loss": 0.6436451077461243,
      "epoch": 1.076995748700992,
      "grad_norm": 13.750596760410684,
      "learning_rate": 5.198782575484629e-08,
      "logits": -1.5236998796463013,
      "logps": -86.2567138671875,
      "loss": 0.6487,
      "objective": 0.6436451077461243,
      "ranking_idealized": 0.625,
      "ranking_idealized_expo": 0.5541666746139526,
      "ranking_simple": 0.5791666507720947,
      "regularize": 0.6436451077461243,
      "step": 190
    },
    {
      "dpo_loss": 0.6471238732337952,
      "epoch": 1.10533774208786,
      "grad_norm": 16.400656168720136,
      "learning_rate": 4.9502920743323515e-08,
      "logits": -1.504638433456421,
      "logps": -85.38711547851562,
      "loss": 0.6526,
      "objective": 0.6471238732337952,
      "ranking_idealized": 0.6333333253860474,
      "ranking_idealized_expo": 0.5666666626930237,
      "ranking_simple": 0.5791666507720947,
      "regularize": 0.6471238732337952,
      "step": 195
    },
    {
      "dpo_loss": 0.6551472544670105,
      "epoch": 1.1336797354747283,
      "grad_norm": 13.299323953163684,
      "learning_rate": 4.701924374150901e-08,
      "logits": -1.5839470624923706,
      "logps": -84.28254699707031,
      "loss": 0.6473,
      "objective": 0.6551472544670105,
      "ranking_idealized": 0.6000000238418579,
      "ranking_idealized_expo": 0.5083333253860474,
      "ranking_simple": 0.512499988079071,
      "regularize": 0.6551472544670105,
      "step": 200
    },
    {
      "epoch": 1.1336797354747283,
      "eval_dpo_loss": 0.689020574092865,
      "eval_logits": -1.5427004098892212,
      "eval_logps": -91.97786712646484,
      "eval_loss": 0.6826460361480713,
      "eval_objective": 0.689020574092865,
      "eval_ranking_idealized": 0.5888429880142212,
      "eval_ranking_idealized_expo": 0.5103305578231812,
      "eval_ranking_simple": 0.5175619721412659,
      "eval_regularize": 0.689020574092865,
      "eval_runtime": 259.6574,
      "eval_samples_per_second": 22.299,
      "eval_steps_per_second": 0.932,
      "step": 200
    },
    {
      "dpo_loss": 0.6496204733848572,
      "epoch": 1.1620217288615966,
      "grad_norm": 13.335913932026553,
      "learning_rate": 4.454293055054397e-08,
      "logits": -1.5556657314300537,
      "logps": -84.84135437011719,
      "loss": 0.6507,
      "objective": 0.6496204733848572,
      "ranking_idealized": 0.5708333253860474,
      "ranking_idealized_expo": 0.48750001192092896,
      "ranking_simple": 0.5166666507720947,
      "regularize": 0.6496204733848572,
      "step": 205
    },
    {
      "dpo_loss": 0.6496918201446533,
      "epoch": 1.1903637222484649,
      "grad_norm": 14.028092373104228,
      "learning_rate": 4.208009877963925e-08,
      "logits": -1.5215808153152466,
      "logps": -85.02703094482422,
      "loss": 0.6464,
      "objective": 0.6496918201446533,
      "ranking_idealized": 0.5874999761581421,
      "ranking_idealized_expo": 0.5,
      "ranking_simple": 0.5083333253860474,
      "regularize": 0.6496918201446533,
      "step": 210
    },
    {
      "dpo_loss": 0.6471313834190369,
      "epoch": 1.2187057156353331,
      "grad_norm": 13.415699632283031,
      "learning_rate": 3.963683273282455e-08,
      "logits": -1.4972035884857178,
      "logps": -84.6708984375,
      "loss": 0.6444,
      "objective": 0.6471313834190369,
      "ranking_idealized": 0.5625,
      "ranking_idealized_expo": 0.4833333194255829,
      "ranking_simple": 0.49166667461395264,
      "regularize": 0.6471313834190369,
      "step": 215
    },
    {
      "dpo_loss": 0.6404972076416016,
      "epoch": 1.2470477090222012,
      "grad_norm": 15.126572956782434,
      "learning_rate": 3.721916837797627e-08,
      "logits": -1.4574114084243774,
      "logps": -83.83084106445312,
      "loss": 0.6425,
      "objective": 0.6404972076416016,
      "ranking_idealized": 0.6541666388511658,
      "ranking_idealized_expo": 0.6041666865348816,
      "ranking_simple": 0.6041666865348816,
      "regularize": 0.6404972076416016,
      "step": 220
    },
    {
      "dpo_loss": 0.64054936170578,
      "epoch": 1.2753897024090695,
      "grad_norm": 13.996717622001338,
      "learning_rate": 3.483307843525758e-08,
      "logits": -1.545446515083313,
      "logps": -85.11936950683594,
      "loss": 0.6452,
      "objective": 0.64054936170578,
      "ranking_idealized": 0.612500011920929,
      "ranking_idealized_expo": 0.5291666388511658,
      "ranking_simple": 0.550000011920929,
      "regularize": 0.64054936170578,
      "step": 225
    },
    {
      "dpo_loss": 0.6458050608634949,
      "epoch": 1.3037316957959377,
      "grad_norm": 13.111039704803117,
      "learning_rate": 3.248445762180878e-08,
      "logits": -1.5097877979278564,
      "logps": -84.0020523071289,
      "loss": 0.6439,
      "objective": 0.6458050608634949,
      "ranking_idealized": 0.5833333134651184,
      "ranking_idealized_expo": 0.5083333253860474,
      "ranking_simple": 0.5333333611488342,
      "regularize": 0.6458050608634949,
      "step": 230
    },
    {
      "dpo_loss": 0.6388878226280212,
      "epoch": 1.3320736891828058,
      "grad_norm": 14.535772291959058,
      "learning_rate": 3.01791080891403e-08,
      "logits": -1.5895107984542847,
      "logps": -82.74007415771484,
      "loss": 0.6453,
      "objective": 0.6388878226280212,
      "ranking_idealized": 0.6041666865348816,
      "ranking_idealized_expo": 0.4833333194255829,
      "ranking_simple": 0.5375000238418579,
      "regularize": 0.6388878226280212,
      "step": 235
    },
    {
      "dpo_loss": 0.6343578100204468,
      "epoch": 1.360415682569674,
      "grad_norm": 14.17039051493401,
      "learning_rate": 2.7922725089204424e-08,
      "logits": -1.5228545665740967,
      "logps": -84.07835388183594,
      "loss": 0.6392,
      "objective": 0.6343578100204468,
      "ranking_idealized": 0.5541666746139526,
      "ranking_idealized_expo": 0.4791666567325592,
      "ranking_simple": 0.5083333253860474,
      "regularize": 0.6343578100204468,
      "step": 240
    },
    {
      "dpo_loss": 0.6442958116531372,
      "epoch": 1.3887576759565423,
      "grad_norm": 13.5370550702067,
      "learning_rate": 2.572088290455715e-08,
      "logits": -1.551520586013794,
      "logps": -84.03118133544922,
      "loss": 0.6442,
      "objective": 0.6442958116531372,
      "ranking_idealized": 0.5916666388511658,
      "ranking_idealized_expo": 0.5208333134651184,
      "ranking_simple": 0.5375000238418579,
      "regularize": 0.6442958116531372,
      "step": 245
    },
    {
      "dpo_loss": 0.6429386138916016,
      "epoch": 1.4170996693434104,
      "grad_norm": 13.133220637867716,
      "learning_rate": 2.3579021077369044e-08,
      "logits": -1.4924685955047607,
      "logps": -83.64738464355469,
      "loss": 0.6449,
      "objective": 0.6429386138916016,
      "ranking_idealized": 0.574999988079071,
      "ranking_idealized_expo": 0.47083333134651184,
      "ranking_simple": 0.49166667461395264,
      "regularize": 0.6429386138916016,
      "step": 250
    },
    {
      "epoch": 1.4170996693434104,
      "eval_dpo_loss": 0.6887210607528687,
      "eval_logits": -1.5537456274032593,
      "eval_logps": -91.60443878173828,
      "eval_loss": 0.6813494563102722,
      "eval_objective": 0.6887210607528687,
      "eval_ranking_idealized": 0.5888429880142212,
      "eval_ranking_idealized_expo": 0.5103305578231812,
      "eval_ranking_simple": 0.5175619721412659,
      "eval_regularize": 0.6887210607528687,
      "eval_runtime": 259.4717,
      "eval_samples_per_second": 22.315,
      "eval_steps_per_second": 0.933,
      "step": 250
    },
    {
      "dpo_loss": 0.6430085301399231,
      "epoch": 1.4454416627302786,
      "grad_norm": 14.835038849969866,
      "learning_rate": 2.1502430971305285e-08,
      "logits": -1.5067726373672485,
      "logps": -84.9552993774414,
      "loss": 0.6394,
      "objective": 0.6430085301399231,
      "ranking_idealized": 0.6166666746139526,
      "ranking_idealized_expo": 0.5458333492279053,
      "ranking_simple": 0.5583333373069763,
      "regularize": 0.6430085301399231,
      "step": 255
    },
    {
      "dpo_loss": 0.6379219889640808,
      "epoch": 1.473783656117147,
      "grad_norm": 15.014357418345154,
      "learning_rate": 1.9496242699473782e-08,
      "logits": -1.570318579673767,
      "logps": -86.18223571777344,
      "loss": 0.6438,
      "objective": 0.6379219889640808,
      "ranking_idealized": 0.5666666626930237,
      "ranking_idealized_expo": 0.48750001192092896,
      "ranking_simple": 0.5166666507720947,
      "regularize": 0.6379219889640808,
      "step": 260
    },
    {
      "dpo_loss": 0.6470407843589783,
      "epoch": 1.5021256495040152,
      "grad_norm": 13.71802645949903,
      "learning_rate": 1.7565412450734317e-08,
      "logits": -1.525256633758545,
      "logps": -83.78214263916016,
      "loss": 0.643,
      "objective": 0.6470407843589783,
      "ranking_idealized": 0.574999988079071,
      "ranking_idealized_expo": 0.5041666626930237,
      "ranking_simple": 0.5333333611488342,
      "regularize": 0.6470407843589783,
      "step": 265
    },
    {
      "dpo_loss": 0.6488485932350159,
      "epoch": 1.5304676428908834,
      "grad_norm": 13.736321065743684,
      "learning_rate": 1.5714710245679348e-08,
      "logits": -1.5463277101516724,
      "logps": -85.4636459350586,
      "loss": 0.6396,
      "objective": 0.6488485932350159,
      "ranking_idealized": 0.5791666507720947,
      "ranking_idealized_expo": 0.5083333253860474,
      "ranking_simple": 0.5333333611488342,
      "regularize": 0.6488485932350159,
      "step": 270
    },
    {
      "dpo_loss": 0.6298065781593323,
      "epoch": 1.5588096362777515,
      "grad_norm": 13.998439522929331,
      "learning_rate": 1.394870815253416e-08,
      "logits": -1.5613772869110107,
      "logps": -86.0442886352539,
      "loss": 0.644,
      "objective": 0.6298065781593323,
      "ranking_idealized": 0.6208333373069763,
      "ranking_idealized_expo": 0.5249999761581421,
      "ranking_simple": 0.5458333492279053,
      "regularize": 0.6298065781593323,
      "step": 275
    },
    {
      "dpo_loss": 0.6345570087432861,
      "epoch": 1.5871516296646198,
      "grad_norm": 14.368158677946898,
      "learning_rate": 1.2271768992088489e-08,
      "logits": -1.4692186117172241,
      "logps": -84.12680053710938,
      "loss": 0.643,
      "objective": 0.6345570087432861,
      "ranking_idealized": 0.574999988079071,
      "ranking_idealized_expo": 0.5041666626930237,
      "ranking_simple": 0.5208333134651184,
      "regularize": 0.6345570087432861,
      "step": 280
    },
    {
      "dpo_loss": 0.6354427337646484,
      "epoch": 1.615493623051488,
      "grad_norm": 15.792908553123706,
      "learning_rate": 1.0688035559563669e-08,
      "logits": -1.5792218446731567,
      "logps": -85.75080108642578,
      "loss": 0.6396,
      "objective": 0.6354427337646484,
      "ranking_idealized": 0.5874999761581421,
      "ranking_idealized_expo": 0.5166666507720947,
      "ranking_simple": 0.5291666388511658,
      "regularize": 0.6354427337646484,
      "step": 285
    },
    {
      "dpo_loss": 0.6333154439926147,
      "epoch": 1.643835616438356,
      "grad_norm": 13.281161084922974,
      "learning_rate": 9.201420390041964e-09,
      "logits": -1.62575364112854,
      "logps": -84.09397888183594,
      "loss": 0.6388,
      "objective": 0.6333154439926147,
      "ranking_idealized": 0.612500011920929,
      "ranking_idealized_expo": 0.5625,
      "ranking_simple": 0.5708333253860474,
      "regularize": 0.6333154439926147,
      "step": 290
    },
    {
      "dpo_loss": 0.6362510919570923,
      "epoch": 1.6721776098252243,
      "grad_norm": 13.068465728291164,
      "learning_rate": 7.815596092742277e-09,
      "logits": -1.5190918445587158,
      "logps": -83.1363296508789,
      "loss": 0.6395,
      "objective": 0.6362510919570923,
      "ranking_idealized": 0.5791666507720947,
      "ranking_idealized_expo": 0.4791666567325592,
      "ranking_simple": 0.512499988079071,
      "regularize": 0.6362510919570923,
      "step": 295
    },
    {
      "dpo_loss": 0.6396799087524414,
      "epoch": 1.7005196032120926,
      "grad_norm": 13.517747947290621,
      "learning_rate": 6.533986278020875e-09,
      "logits": -1.5812262296676636,
      "logps": -84.42427062988281,
      "loss": 0.6384,
      "objective": 0.6396799087524414,
      "ranking_idealized": 0.6583333611488342,
      "ranking_idealized_expo": 0.574999988079071,
      "ranking_simple": 0.5958333611488342,
      "regularize": 0.6396799087524414,
      "step": 300
    },
    {
      "epoch": 1.7005196032120926,
      "eval_dpo_loss": 0.6890466213226318,
      "eval_logits": -1.5627375841140747,
      "eval_logps": -92.01404571533203,
      "eval_loss": 0.6818000674247742,
      "eval_objective": 0.6890466213226318,
      "eval_ranking_idealized": 0.5888429880142212,
      "eval_ranking_idealized_expo": 0.5103305578231812,
      "eval_ranking_simple": 0.5185950398445129,
      "eval_regularize": 0.6890466213226318,
      "eval_runtime": 261.8927,
      "eval_samples_per_second": 22.108,
      "eval_steps_per_second": 0.924,
      "step": 300
    },
    {
      "dpo_loss": 0.6523339748382568,
      "epoch": 1.7288615965989607,
      "grad_norm": 14.833162083591201,
      "learning_rate": 5.359757099511236e-09,
      "logits": -1.5387723445892334,
      "logps": -85.74983978271484,
      "loss": 0.6443,
      "objective": 0.6523339748382568,
      "ranking_idealized": 0.5625,
      "ranking_idealized_expo": 0.5083333253860474,
      "ranking_simple": 0.5208333134651184,
      "regularize": 0.6523339748382568,
      "step": 305
    },
    {
      "dpo_loss": 0.6306132078170776,
      "epoch": 1.7572035899858292,
      "grad_norm": 14.09816487492066,
      "learning_rate": 4.295809432298269e-09,
      "logits": -1.616551160812378,
      "logps": -83.39339447021484,
      "loss": 0.6374,
      "objective": 0.6306132078170776,
      "ranking_idealized": 0.6083333492279053,
      "ranking_idealized_expo": 0.5249999761581421,
      "ranking_simple": 0.5458333492279053,
      "regularize": 0.6306132078170776,
      "step": 310
    },
    {
      "dpo_loss": 0.6414229273796082,
      "epoch": 1.7855455833726972,
      "grad_norm": 12.85646357819718,
      "learning_rate": 3.344771706449956e-09,
      "logits": -1.5274823904037476,
      "logps": -84.38652801513672,
      "loss": 0.6393,
      "objective": 0.6414229273796082,
      "ranking_idealized": 0.6041666865348816,
      "ranking_idealized_expo": 0.5083333253860474,
      "ranking_simple": 0.5291666388511658,
      "regularize": 0.6414229273796082,
      "step": 315
    },
    {
      "dpo_loss": 0.6470603942871094,
      "epoch": 1.8138875767595655,
      "grad_norm": 14.912269898190576,
      "learning_rate": 2.508993413610866e-09,
      "logits": -1.5946826934814453,
      "logps": -86.11390686035156,
      "loss": 0.6416,
      "objective": 0.6470603942871094,
      "ranking_idealized": 0.6000000238418579,
      "ranking_idealized_expo": 0.5041666626930237,
      "ranking_simple": 0.5166666507720947,
      "regularize": 0.6470603942871094,
      "step": 320
    },
    {
      "dpo_loss": 0.6489089727401733,
      "epoch": 1.8422295701464337,
      "grad_norm": 16.310745862402126,
      "learning_rate": 1.790539302699351e-09,
      "logits": -1.6386195421218872,
      "logps": -86.83934020996094,
      "loss": 0.6366,
      "objective": 0.6489089727401733,
      "ranking_idealized": 0.6333333253860474,
      "ranking_idealized_expo": 0.5166666507720947,
      "ranking_simple": 0.5666666626930237,
      "regularize": 0.6489089727401733,
      "step": 325
    },
    {
      "dpo_loss": 0.6407893896102905,
      "epoch": 1.8705715635333018,
      "grad_norm": 14.36720629246665,
      "learning_rate": 1.1911842790474635e-09,
      "logits": -1.3639094829559326,
      "logps": -84.11365509033203,
      "loss": 0.6382,
      "objective": 0.6407893896102905,
      "ranking_idealized": 0.5708333253860474,
      "ranking_idealized_expo": 0.4749999940395355,
      "ranking_simple": 0.5,
      "regularize": 0.6407893896102905,
      "step": 330
    },
    {
      "dpo_loss": 0.642834484577179,
      "epoch": 1.89891355692017,
      "grad_norm": 12.837846805234047,
      "learning_rate": 7.124090195851806e-10,
      "logits": -1.5466150045394897,
      "logps": -84.1384048461914,
      "loss": 0.6395,
      "objective": 0.642834484577179,
      "ranking_idealized": 0.5916666388511658,
      "ranking_idealized_expo": 0.4958333373069763,
      "ranking_simple": 0.5458333492279053,
      "regularize": 0.642834484577179,
      "step": 335
    },
    {
      "dpo_loss": 0.6367443799972534,
      "epoch": 1.9272555503070383,
      "grad_norm": 13.618225760097143,
      "learning_rate": 3.553963149013295e-10,
      "logits": -1.5659750699996948,
      "logps": -86.13041687011719,
      "loss": 0.644,
      "objective": 0.6367443799972534,
      "ranking_idealized": 0.6416666507720947,
      "ranking_idealized_expo": 0.5416666865348816,
      "ranking_simple": 0.5791666507720947,
      "regularize": 0.6367443799972534,
      "step": 340
    },
    {
      "dpo_loss": 0.6491357088088989,
      "epoch": 1.9555975436939064,
      "grad_norm": 13.98814029327361,
      "learning_rate": 1.2102814721791643e-10,
      "logits": -1.5671348571777344,
      "logps": -84.2940902709961,
      "loss": 0.6417,
      "objective": 0.6491357088088989,
      "ranking_idealized": 0.574999988079071,
      "ranking_idealized_expo": 0.5333333611488342,
      "ranking_simple": 0.5375000238418579,
      "regularize": 0.6491357088088989,
      "step": 345
    },
    {
      "dpo_loss": 0.6474497318267822,
      "epoch": 1.9839395370807746,
      "grad_norm": 13.915220160397528,
      "learning_rate": 9.883511496722174e-12,
      "logits": -1.4725474119186401,
      "logps": -85.95130157470703,
      "loss": 0.6431,
      "objective": 0.6474497318267822,
      "ranking_idealized": 0.574999988079071,
      "ranking_idealized_expo": 0.48750001192092896,
      "ranking_simple": 0.4958333373069763,
      "regularize": 0.6474497318267822,
      "step": 350
    },
    {
      "epoch": 1.9839395370807746,
      "eval_dpo_loss": 0.6890938878059387,
      "eval_logits": -1.5636365413665771,
      "eval_logps": -92.05502319335938,
      "eval_loss": 0.6817956566810608,
      "eval_objective": 0.6890938878059387,
      "eval_ranking_idealized": 0.5888429880142212,
      "eval_ranking_idealized_expo": 0.5103305578231812,
      "eval_ranking_simple": 0.51962810754776,
      "eval_regularize": 0.6890938878059387,
      "eval_runtime": 259.5397,
      "eval_samples_per_second": 22.309,
      "eval_steps_per_second": 0.932,
      "step": 350
    },
    {
      "epoch": 1.995276334435522,
      "step": 352,
      "total_flos": 0.0,
      "train_loss": 0.6593505936590108,
      "train_runtime": 15975.3291,
      "train_samples_per_second": 6.36,
      "train_steps_per_second": 0.022
    }
  ],
  "logging_steps": 5,
  "max_steps": 352,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 2,
  "save_steps": 500,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": false,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}