{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.0,
  "eval_steps": 100,
  "global_step": 681,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0014684287812041115,
      "epsilon_dpo/beta": 0.09979122877120972,
      "epsilon_dpo/beta_margin_grad_mean": -0.4994426667690277,
      "epsilon_dpo/beta_margin_grad_std": 0.007109890226274729,
      "epsilon_dpo/beta_margin_mean": 0.0022308863699436188,
      "epsilon_dpo/beta_margin_std": 0.028449052944779396,
      "epsilon_dpo/loss_margin_mean": 0.024216145277023315,
      "grad_norm": 153.6117401123047,
      "kl/avg_steps": 0.21875,
      "kl/beta": 0.10000000149011612,
      "kl/n_epsilon_steps": 0.390625,
      "kl/p_epsilon_steps": 0.609375,
      "learning_rate": 0.0,
      "logits/chosen": -3.435748338699341,
      "logits/rejected": -3.461001396179199,
      "logps/chosen": -31.484785079956055,
      "logps/ref_chosen": -31.500404357910156,
      "logps/ref_rejected": -79.35133361816406,
      "logps/rejected": -79.35993194580078,
      "loss": 1.3843,
      "rewards/accuracies": 0.59375,
      "rewards/chosen": 0.0014821073273196816,
      "rewards/margins": 0.002230936661362648,
      "rewards/rejected": -0.0007488295086659491,
      "step": 1
    },
    {
      "epoch": 0.002936857562408223,
      "epsilon_dpo/beta": 0.09972933679819107,
      "epsilon_dpo/beta_margin_grad_mean": -0.5002588629722595,
      "epsilon_dpo/beta_margin_grad_std": 0.006068441551178694,
      "epsilon_dpo/beta_margin_mean": -0.0010360678425058722,
      "epsilon_dpo/beta_margin_std": 0.024278299883008003,
      "epsilon_dpo/loss_margin_mean": -0.008541211485862732,
      "grad_norm": 155.6954803466797,
      "kl/avg_steps": 0.0625,
      "kl/beta": 0.09978172928094864,
      "kl/n_epsilon_steps": 0.46875,
      "kl/p_epsilon_steps": 0.53125,
      "learning_rate": 7.246376811594203e-09,
      "logits/chosen": -3.517317533493042,
      "logits/rejected": -3.4876961708068848,
      "logps/chosen": -36.63703918457031,
      "logps/ref_chosen": -36.63695526123047,
      "logps/ref_rejected": -80.44895935058594,
      "logps/rejected": -80.44050598144531,
      "loss": 1.3875,
      "rewards/accuracies": 0.546875,
      "rewards/chosen": -8.885323768481612e-05,
      "rewards/margins": -0.0010360784363001585,
      "rewards/rejected": 0.0009472252568230033,
      "step": 2
    },
    {
      "epoch": 0.004405286343612335,
      "epsilon_dpo/beta": 0.09971363842487335,
      "epsilon_dpo/beta_margin_grad_mean": -0.4988003969192505,
      "epsilon_dpo/beta_margin_grad_std": 0.007179732900112867,
      "epsilon_dpo/beta_margin_mean": 0.004800259135663509,
      "epsilon_dpo/beta_margin_std": 0.028725432232022285,
      "epsilon_dpo/loss_margin_mean": 0.050338417291641235,
      "grad_norm": 160.3519744873047,
      "kl/avg_steps": 0.015625,
      "kl/beta": 0.09971940517425537,
      "kl/n_epsilon_steps": 0.484375,
      "kl/p_epsilon_steps": 0.5,
      "learning_rate": 1.4492753623188406e-08,
      "logits/chosen": -3.5040597915649414,
      "logits/rejected": -3.4887866973876953,
      "logps/chosen": -37.792213439941406,
      "logps/ref_chosen": -37.83708190917969,
      "logps/ref_rejected": -73.14408874511719,
      "logps/rejected": -73.14955139160156,
      "loss": 1.3817,
      "rewards/accuracies": 0.515625,
      "rewards/chosen": 0.004386060871183872,
      "rewards/margins": 0.004800218157470226,
      "rewards/rejected": -0.00041415661689825356,
      "step": 3
    },
    {
      "epoch": 0.005873715124816446,
      "epsilon_dpo/beta": 0.09960457682609558,
      "epsilon_dpo/beta_margin_grad_mean": -0.4994582235813141,
      "epsilon_dpo/beta_margin_grad_std": 0.007399262860417366,
      "epsilon_dpo/beta_margin_mean": 0.002166971331462264,
      "epsilon_dpo/beta_margin_std": 0.029604651033878326,
      "epsilon_dpo/loss_margin_mean": 0.02400125563144684,
      "grad_norm": 155.65963745117188,
      "kl/avg_steps": 0.109375,
      "kl/beta": 0.0997038260102272,
      "kl/n_epsilon_steps": 0.4375,
      "kl/p_epsilon_steps": 0.546875,
      "learning_rate": 2.1739130434782606e-08,
      "logits/chosen": -3.468817710876465,
      "logits/rejected": -3.508214235305786,
      "logps/chosen": -43.339508056640625,
      "logps/ref_chosen": -43.336036682128906,
      "logps/ref_rejected": -93.7607650756836,
      "logps/rejected": -93.78823852539062,
      "loss": 1.3843,
      "rewards/accuracies": 0.53125,
      "rewards/chosen": -0.000430062209488824,
      "rewards/margins": 0.002166960621252656,
      "rewards/rejected": -0.0025970228016376495,
      "step": 4
    },
    {
      "epoch": 0.007342143906020558,
      "epsilon_dpo/beta": 0.09963598102331161,
      "epsilon_dpo/beta_margin_grad_mean": -0.5009759068489075,
      "epsilon_dpo/beta_margin_grad_std": 0.008376221172511578,
      "epsilon_dpo/beta_margin_mean": -0.003906682599335909,
      "epsilon_dpo/beta_margin_std": 0.033521223813295364,
      "epsilon_dpo/loss_margin_mean": -0.036883413791656494,
      "grad_norm": 179.6461639404297,
      "kl/avg_steps": -0.03125,
      "kl/beta": 0.09959489107131958,
      "kl/n_epsilon_steps": 0.515625,
      "kl/p_epsilon_steps": 0.484375,
      "learning_rate": 2.898550724637681e-08,
      "logits/chosen": -3.436485767364502,
      "logits/rejected": -3.4411563873291016,
      "logps/chosen": -32.93244934082031,
      "logps/ref_chosen": -32.90675354003906,
      "logps/ref_rejected": -90.30764770507812,
      "logps/rejected": -90.29646301269531,
      "loss": 1.3905,
      "rewards/accuracies": 0.484375,
      "rewards/chosen": -0.0026276579592376947,
      "rewards/margins": -0.003906670026481152,
      "rewards/rejected": 0.0012790121836587787,
      "step": 5
    },
    {
      "epoch": 0.00881057268722467,
      "epsilon_dpo/beta": 0.09954258054494858,
      "epsilon_dpo/beta_margin_grad_mean": -0.4993135929107666,
      "epsilon_dpo/beta_margin_grad_std": 0.006349037401378155,
      "epsilon_dpo/beta_margin_mean": 0.002748197177425027,
      "epsilon_dpo/beta_margin_std": 0.025408554822206497,
      "epsilon_dpo/loss_margin_mean": 0.029358193278312683,
      "grad_norm": 168.01638793945312,
      "kl/avg_steps": 0.09375,
      "kl/beta": 0.09962602704763412,
      "kl/n_epsilon_steps": 0.453125,
      "kl/p_epsilon_steps": 0.546875,
      "learning_rate": 3.6231884057971014e-08,
      "logits/chosen": -3.476027011871338,
      "logits/rejected": -3.391662836074829,
      "logps/chosen": -40.59587478637695,
      "logps/ref_chosen": -40.57701110839844,
      "logps/ref_rejected": -98.47296142578125,
      "logps/rejected": -98.52117919921875,
      "loss": 1.3837,
      "rewards/accuracies": 0.53125,
      "rewards/chosen": -0.0019228225573897362,
      "rewards/margins": 0.0027480702847242355,
      "rewards/rejected": -0.004670892842113972,
      "step": 6
    },
    {
      "epoch": 0.010279001468428781,
      "epsilon_dpo/beta": 0.09951155632734299,
      "epsilon_dpo/beta_margin_grad_mean": -0.49992430210113525,
      "epsilon_dpo/beta_margin_grad_std": 0.007149491459131241,
      "epsilon_dpo/beta_margin_mean": 0.0003023834724444896,
      "epsilon_dpo/beta_margin_std": 0.028604645282030106,
      "epsilon_dpo/loss_margin_mean": 0.00525471568107605,
      "grad_norm": 165.30645751953125,
      "kl/avg_steps": 0.03125,
      "kl/beta": 0.09953271597623825,
      "kl/n_epsilon_steps": 0.484375,
      "kl/p_epsilon_steps": 0.515625,
      "learning_rate": 4.347826086956521e-08,
      "logits/chosen": -3.4681951999664307,
      "logits/rejected": -3.4873459339141846,
      "logps/chosen": -41.3924674987793,
      "logps/ref_chosen": -41.414642333984375,
      "logps/ref_rejected": -111.01716613769531,
      "logps/rejected": -111.000244140625,
      "loss": 1.3862,
      "rewards/accuracies": 0.515625,
      "rewards/chosen": 0.002097110729664564,
      "rewards/margins": 0.0003024058823939413,
      "rewards/rejected": 0.0017947049345821142,
      "step": 7
    },
    {
      "epoch": 0.011747430249632892,
      "epsilon_dpo/beta": 0.09983792901039124,
      "epsilon_dpo/beta_margin_grad_mean": -0.5019076466560364,
      "epsilon_dpo/beta_margin_grad_std": 0.006241849157959223,
      "epsilon_dpo/beta_margin_mean": -0.007631635759025812,
      "epsilon_dpo/beta_margin_std": 0.02497241646051407,
      "epsilon_dpo/loss_margin_mean": -0.07474762201309204,
      "grad_norm": 170.1224822998047,
      "kl/avg_steps": -0.328125,
      "kl/beta": 0.09950161725282669,
      "kl/n_epsilon_steps": 0.65625,
      "kl/p_epsilon_steps": 0.328125,
      "learning_rate": 5.0724637681159424e-08,
      "logits/chosen": -3.5701963901519775,
      "logits/rejected": -3.46230411529541,
      "logps/chosen": -39.27941131591797,
      "logps/ref_chosen": -39.25566482543945,
      "logps/ref_rejected": -85.94947814941406,
      "logps/rejected": -85.89848327636719,
      "loss": 1.3941,
      "rewards/accuracies": 0.328125,
      "rewards/chosen": -0.0024497562553733587,
      "rewards/margins": -0.007631613872945309,
      "rewards/rejected": 0.005181857850402594,
      "step": 8
    },
    {
      "epoch": 0.013215859030837005,
      "epsilon_dpo/beta": 0.09983916580677032,
      "epsilon_dpo/beta_margin_grad_mean": -0.5000754594802856,
      "epsilon_dpo/beta_margin_grad_std": 0.006013626232743263,
      "epsilon_dpo/beta_margin_mean": -0.0003015802940353751,
      "epsilon_dpo/beta_margin_std": 0.02405872382223606,
      "epsilon_dpo/loss_margin_mean": -0.0012376010417938232,
      "grad_norm": 179.10533142089844,
      "kl/avg_steps": 0.0,
      "kl/beta": 0.09982918202877045,
      "kl/n_epsilon_steps": 0.5,
      "kl/p_epsilon_steps": 0.5,
      "learning_rate": 5.797101449275362e-08,
      "logits/chosen": -3.526092529296875,
      "logits/rejected": -3.532437801361084,
      "logps/chosen": -38.925682067871094,
      "logps/ref_chosen": -38.9265251159668,
      "logps/ref_rejected": -106.43075561523438,
      "logps/rejected": -106.42868041992188,
      "loss": 1.3867,
      "rewards/accuracies": 0.453125,
      "rewards/chosen": -4.6280911192297935e-07,
      "rewards/margins": -0.0003015893744304776,
      "rewards/rejected": 0.00030112662352621555,
      "step": 9
    },
    {
      "epoch": 0.014684287812041116,
      "epsilon_dpo/beta": 0.09972980618476868,
      "epsilon_dpo/beta_margin_grad_mean": -0.49933433532714844,
      "epsilon_dpo/beta_margin_grad_std": 0.005578126758337021,
      "epsilon_dpo/beta_margin_mean": 0.0026634030509740114,
      "epsilon_dpo/beta_margin_std": 0.02231536991894245,
      "epsilon_dpo/loss_margin_mean": 0.028553977608680725,
      "grad_norm": 161.84414672851562,
      "kl/avg_steps": 0.109375,
      "kl/beta": 0.09982918202877045,
      "kl/n_epsilon_steps": 0.4375,
      "kl/p_epsilon_steps": 0.546875,
      "learning_rate": 6.521739130434782e-08,
      "logits/chosen": -3.487415313720703,
      "logits/rejected": -3.4311609268188477,
      "logps/chosen": -37.931678771972656,
      "logps/ref_chosen": -37.94172668457031,
      "logps/ref_rejected": -84.4234848022461,
      "logps/rejected": -84.44198608398438,
      "loss": 1.3838,
      "rewards/accuracies": 0.578125,
      "rewards/chosen": 0.0009347056620754302,
      "rewards/margins": 0.0026633867528289557,
      "rewards/rejected": -0.0017286810325458646,
      "step": 10
    },
    {
      "epoch": 0.016152716593245228,
      "epsilon_dpo/beta": 0.09954308718442917,
      "epsilon_dpo/beta_margin_grad_mean": -0.49776411056518555,
      "epsilon_dpo/beta_margin_grad_std": 0.006337402854114771,
      "epsilon_dpo/beta_margin_mean": 0.008945857174694538,
      "epsilon_dpo/beta_margin_std": 0.02535552717745304,
      "epsilon_dpo/loss_margin_mean": 0.09173570573329926,
      "grad_norm": 167.265380859375,
      "kl/avg_steps": 0.1875,
      "kl/beta": 0.09972011297941208,
      "kl/n_epsilon_steps": 0.40625,
      "kl/p_epsilon_steps": 0.59375,
      "learning_rate": 7.246376811594203e-08,
      "logits/chosen": -3.4978814125061035,
      "logits/rejected": -3.5033230781555176,
      "logps/chosen": -32.73089599609375,
      "logps/ref_chosen": -32.742462158203125,
      "logps/ref_rejected": -94.11013793945312,
      "logps/rejected": -94.19029998779297,
      "loss": 1.3775,
      "rewards/accuracies": 0.578125,
      "rewards/chosen": 0.0010777543066069484,
      "rewards/margins": 0.008945956826210022,
      "rewards/rejected": -0.007868202403187752,
      "step": 11
    },
    {
      "epoch": 0.01762114537444934,
      "epsilon_dpo/beta": 0.09929458051919937,
      "epsilon_dpo/beta_margin_grad_mean": -0.49879196286201477,
      "epsilon_dpo/beta_margin_grad_std": 0.008286659605801105,
      "epsilon_dpo/beta_margin_mean": 0.004831877537071705,
      "epsilon_dpo/beta_margin_std": 0.03315826877951622,
      "epsilon_dpo/loss_margin_mean": 0.05105578899383545,
      "grad_norm": 184.79959106445312,
      "kl/avg_steps": 0.25,
      "kl/beta": 0.09953349083662033,
      "kl/n_epsilon_steps": 0.375,
      "kl/p_epsilon_steps": 0.625,
      "learning_rate": 7.971014492753623e-08,
      "logits/chosen": -3.5019748210906982,
      "logits/rejected": -3.507927417755127,
      "logps/chosen": -43.87688446044922,
      "logps/ref_chosen": -43.85453796386719,
      "logps/ref_rejected": -111.72984313964844,
      "logps/rejected": -111.80323791503906,
      "loss": 1.3817,
      "rewards/accuracies": 0.625,
      "rewards/chosen": -0.0023162683937698603,
      "rewards/margins": 0.004831977654248476,
      "rewards/rejected": -0.007148245815187693,
      "step": 12
    },
    {
      "epoch": 0.01908957415565345,
      "epsilon_dpo/beta": 0.09935726225376129,
      "epsilon_dpo/beta_margin_grad_mean": -0.49919238686561584,
      "epsilon_dpo/beta_margin_grad_std": 0.006513380445539951,
      "epsilon_dpo/beta_margin_mean": 0.00323132099583745,
      "epsilon_dpo/beta_margin_std": 0.026057543233036995,
      "epsilon_dpo/loss_margin_mean": 0.034654468297958374,
      "grad_norm": 170.2859344482422,
      "kl/avg_steps": -0.0625,
      "kl/beta": 0.09928527474403381,
      "kl/n_epsilon_steps": 0.53125,
      "kl/p_epsilon_steps": 0.46875,
      "learning_rate": 8.695652173913042e-08,
      "logits/chosen": -3.540365219116211,
      "logits/rejected": -3.4680285453796387,
      "logps/chosen": -41.653934478759766,
      "logps/ref_chosen": -41.67176818847656,
      "logps/ref_rejected": -91.09086608886719,
      "logps/rejected": -91.1076889038086,
      "loss": 1.3832,
      "rewards/accuracies": 0.484375,
      "rewards/chosen": 0.0017041495302692056,
      "rewards/margins": 0.0032313289120793343,
      "rewards/rejected": -0.0015271796146407723,
      "step": 13
    },
    {
      "epoch": 0.020558002936857563,
      "epsilon_dpo/beta": 0.09892261028289795,
      "epsilon_dpo/beta_margin_grad_mean": -0.49505484104156494,
      "epsilon_dpo/beta_margin_grad_std": 0.007620512507855892,
      "epsilon_dpo/beta_margin_mean": 0.01978609710931778,
      "epsilon_dpo/beta_margin_std": 0.03049391135573387,
      "epsilon_dpo/loss_margin_mean": 0.20201179385185242,
      "grad_norm": 191.28602600097656,
      "kl/avg_steps": 0.4375,
      "kl/beta": 0.09934736788272858,
      "kl/n_epsilon_steps": 0.28125,
      "kl/p_epsilon_steps": 0.71875,
      "learning_rate": 9.420289855072464e-08,
      "logits/chosen": -3.5063905715942383,
      "logits/rejected": -3.5407614707946777,
      "logps/chosen": -32.60367202758789,
      "logps/ref_chosen": -32.668601989746094,
      "logps/ref_rejected": -111.8526611328125,
      "logps/rejected": -111.98973846435547,
      "loss": 1.3668,
      "rewards/accuracies": 0.734375,
      "rewards/chosen": 0.0063568041659891605,
      "rewards/margins": 0.019785966724157333,
      "rewards/rejected": -0.01342916302382946,
      "step": 14
    },
    {
      "epoch": 0.022026431718061675,
      "epsilon_dpo/beta": 0.09844519197940826,
      "epsilon_dpo/beta_margin_grad_mean": -0.4958465099334717,
      "epsilon_dpo/beta_margin_grad_std": 0.006347531918436289,
      "epsilon_dpo/beta_margin_mean": 0.016617843881249428,
      "epsilon_dpo/beta_margin_std": 0.025397835299372673,
      "epsilon_dpo/loss_margin_mean": 0.17031680047512054,
      "grad_norm": 144.098388671875,
      "kl/avg_steps": 0.484375,
      "kl/beta": 0.09891461580991745,
      "kl/n_epsilon_steps": 0.25,
      "kl/p_epsilon_steps": 0.734375,
      "learning_rate": 1.0144927536231885e-07,
      "logits/chosen": -3.4299309253692627,
      "logits/rejected": -3.4826550483703613,
      "logps/chosen": -41.79482650756836,
      "logps/ref_chosen": -41.78295135498047,
      "logps/ref_rejected": -86.42213439941406,
      "logps/rejected": -86.60433197021484,
      "loss": 1.3699,
      "rewards/accuracies": 0.734375,
      "rewards/chosen": -0.0012117127189412713,
      "rewards/margins": 0.01661786250770092,
      "rewards/rejected": -0.01782957650721073,
      "step": 15
    },
    {
      "epoch": 0.023494860499265784,
      "epsilon_dpo/beta": 0.09795541316270828,
      "epsilon_dpo/beta_margin_grad_mean": -0.4958525598049164,
      "epsilon_dpo/beta_margin_grad_std": 0.007386185694485903,
      "epsilon_dpo/beta_margin_mean": 0.016594676300883293,
      "epsilon_dpo/beta_margin_std": 0.029557235538959503,
      "epsilon_dpo/loss_margin_mean": 0.1711507886648178,
      "grad_norm": 150.39041137695312,
      "kl/avg_steps": 0.5,
      "kl/beta": 0.09843780845403671,
      "kl/n_epsilon_steps": 0.25,
      "kl/p_epsilon_steps": 0.75,
      "learning_rate": 1.0869565217391303e-07,
      "logits/chosen": -3.477813959121704,
      "logits/rejected": -3.4168307781219482,
      "logps/chosen": -40.885353088378906,
      "logps/ref_chosen": -40.9011116027832,
      "logps/ref_rejected": -88.89961242675781,
      "logps/rejected": -89.05500793457031,
      "loss": 1.37,
      "rewards/accuracies": 0.75,
      "rewards/chosen": 0.0014686340000480413,
      "rewards/margins": 0.016594652086496353,
      "rewards/rejected": -0.015126017853617668,
      "step": 16
    },
    {
      "epoch": 0.024963289280469897,
      "epsilon_dpo/beta": 0.09734562784433365,
      "epsilon_dpo/beta_margin_grad_mean": -0.49166470766067505,
      "epsilon_dpo/beta_margin_grad_std": 0.009398871101439,
      "epsilon_dpo/beta_margin_mean": 0.033359427005052567,
      "epsilon_dpo/beta_margin_std": 0.037631068378686905,
      "epsilon_dpo/loss_margin_mean": 0.34458500146865845,
      "grad_norm": 171.44161987304688,
      "kl/avg_steps": 0.625,
      "kl/beta": 0.09794806689023972,
      "kl/n_epsilon_steps": 0.1875,
      "kl/p_epsilon_steps": 0.8125,
      "learning_rate": 1.1594202898550725e-07,
      "logits/chosen": -3.526141405105591,
      "logits/rejected": -3.4046664237976074,
      "logps/chosen": -38.45612335205078,
      "logps/ref_chosen": -38.555274963378906,
      "logps/ref_rejected": -91.66143798828125,
      "logps/rejected": -91.90687561035156,
      "loss": 1.3536,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": 0.00957987830042839,
      "rewards/margins": 0.03335944563150406,
      "rewards/rejected": -0.023779571056365967,
      "step": 17
    },
    {
      "epoch": 0.02643171806167401,
      "epsilon_dpo/beta": 0.09655846655368805,
      "epsilon_dpo/beta_margin_grad_mean": -0.4906516373157501,
      "epsilon_dpo/beta_margin_grad_std": 0.009394010528922081,
      "epsilon_dpo/beta_margin_mean": 0.03741518035531044,
      "epsilon_dpo/beta_margin_std": 0.03761782497167587,
      "epsilon_dpo/loss_margin_mean": 0.3884708285331726,
      "grad_norm": 174.4280242919922,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0973396971821785,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.2318840579710146e-07,
      "logits/chosen": -3.457716464996338,
      "logits/rejected": -3.3878774642944336,
      "logps/chosen": -26.462953567504883,
      "logps/ref_chosen": -26.55130386352539,
      "logps/ref_rejected": -82.15496063232422,
      "logps/rejected": -82.455078125,
      "loss": 1.3496,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": 0.008510860614478588,
      "rewards/margins": 0.03741515427827835,
      "rewards/rejected": -0.028904292732477188,
      "step": 18
    },
    {
      "epoch": 0.027900146842878122,
      "epsilon_dpo/beta": 0.0959613099694252,
      "epsilon_dpo/beta_margin_grad_mean": -0.4871312379837036,
      "epsilon_dpo/beta_margin_grad_std": 0.014039521105587482,
      "epsilon_dpo/beta_margin_mean": 0.05153882876038551,
      "epsilon_dpo/beta_margin_std": 0.05628020688891411,
      "epsilon_dpo/loss_margin_mean": 0.5397706031799316,
      "grad_norm": 158.5471649169922,
      "kl/avg_steps": 0.625,
      "kl/beta": 0.0965551808476448,
      "kl/n_epsilon_steps": 0.1875,
      "kl/p_epsilon_steps": 0.8125,
      "learning_rate": 1.3043478260869563e-07,
      "logits/chosen": -3.5245118141174316,
      "logits/rejected": -3.4201385974884033,
      "logps/chosen": -51.44847869873047,
      "logps/ref_chosen": -51.548377990722656,
      "logps/ref_rejected": -95.98385620117188,
      "logps/rejected": -96.42372131347656,
      "loss": 1.3362,
      "rewards/accuracies": 0.828125,
      "rewards/chosen": 0.009450599551200867,
      "rewards/margins": 0.0515388660132885,
      "rewards/rejected": -0.04208826646208763,
      "step": 19
    },
    {
      "epoch": 0.02936857562408223,
      "epsilon_dpo/beta": 0.09518533945083618,
      "epsilon_dpo/beta_margin_grad_mean": -0.4861924350261688,
      "epsilon_dpo/beta_margin_grad_std": 0.01256165187805891,
      "epsilon_dpo/beta_margin_mean": 0.05528602749109268,
      "epsilon_dpo/beta_margin_std": 0.05034392327070236,
      "epsilon_dpo/loss_margin_mean": 0.5822827816009521,
      "grad_norm": 148.6184844970703,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.09595546126365662,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.3768115942028986e-07,
      "logits/chosen": -3.5266153812408447,
      "logits/rejected": -3.454347848892212,
      "logps/chosen": -32.44221878051758,
      "logps/ref_chosen": -32.57563781738281,
      "logps/ref_rejected": -83.72441101074219,
      "logps/rejected": -84.17327880859375,
      "loss": 1.3324,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": 0.012646486982703209,
      "rewards/margins": 0.055285997688770294,
      "rewards/rejected": -0.042639512568712234,
      "step": 20
    },
    {
      "epoch": 0.030837004405286344,
      "epsilon_dpo/beta": 0.09435869753360748,
      "epsilon_dpo/beta_margin_grad_mean": -0.485392689704895,
      "epsilon_dpo/beta_margin_grad_std": 0.01355255488306284,
      "epsilon_dpo/beta_margin_mean": 0.058501191437244415,
      "epsilon_dpo/beta_margin_std": 0.054356660693883896,
      "epsilon_dpo/loss_margin_mean": 0.6211739182472229,
      "grad_norm": 138.44097900390625,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.09518210589885712,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 1.4492753623188405e-07,
      "logits/chosen": -3.416736602783203,
      "logits/rejected": -3.4210104942321777,
      "logps/chosen": -36.161651611328125,
      "logps/ref_chosen": -36.24628448486328,
      "logps/ref_rejected": -82.68882751464844,
      "logps/rejected": -83.22535705566406,
      "loss": 1.3294,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": 0.007934953086078167,
      "rewards/margins": 0.05850117653608322,
      "rewards/rejected": -0.05056622251868248,
      "step": 21
    },
    {
      "epoch": 0.032305433186490456,
      "epsilon_dpo/beta": 0.09351073950529099,
      "epsilon_dpo/beta_margin_grad_mean": -0.48178407549858093,
      "epsilon_dpo/beta_margin_grad_std": 0.01465072762221098,
      "epsilon_dpo/beta_margin_mean": 0.07297518849372864,
      "epsilon_dpo/beta_margin_std": 0.05880572646856308,
      "epsilon_dpo/loss_margin_mean": 0.7812974452972412,
      "grad_norm": 163.26019287109375,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.0943564921617508,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 1.5217391304347825e-07,
      "logits/chosen": -3.4850616455078125,
      "logits/rejected": -3.4672698974609375,
      "logps/chosen": -44.6229248046875,
      "logps/ref_chosen": -44.70884704589844,
      "logps/ref_rejected": -103.42787170410156,
      "logps/rejected": -104.12324523925781,
      "loss": 1.3155,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": 0.007986144162714481,
      "rewards/margins": 0.07297517359256744,
      "rewards/rejected": -0.06498903036117554,
      "step": 22
    },
    {
      "epoch": 0.033773861967694566,
      "epsilon_dpo/beta": 0.09270013123750687,
      "epsilon_dpo/beta_margin_grad_mean": -0.47907260060310364,
      "epsilon_dpo/beta_margin_grad_std": 0.01780831441283226,
      "epsilon_dpo/beta_margin_mean": 0.08389898389577866,
      "epsilon_dpo/beta_margin_std": 0.07160933315753937,
      "epsilon_dpo/loss_margin_mean": 0.9064887762069702,
      "grad_norm": 153.97116088867188,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.09350906312465668,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 1.5942028985507245e-07,
      "logits/chosen": -3.4643330574035645,
      "logits/rejected": -3.442781686782837,
      "logps/chosen": -40.120094299316406,
      "logps/ref_chosen": -40.26862335205078,
      "logps/ref_rejected": -85.85059356689453,
      "logps/rejected": -86.60855102539062,
      "loss": 1.3054,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": 0.013693436048924923,
      "rewards/margins": 0.08389902114868164,
      "rewards/rejected": -0.07020558416843414,
      "step": 23
    },
    {
      "epoch": 0.03524229074889868,
      "epsilon_dpo/beta": 0.0919250100851059,
      "epsilon_dpo/beta_margin_grad_mean": -0.4745597541332245,
      "epsilon_dpo/beta_margin_grad_std": 0.019036216661334038,
      "epsilon_dpo/beta_margin_mean": 0.10201232135295868,
      "epsilon_dpo/beta_margin_std": 0.07649125903844833,
      "epsilon_dpo/loss_margin_mean": 1.1116865873336792,
      "grad_norm": 161.09317016601562,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.09269795566797256,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 1.6666666666666665e-07,
      "logits/chosen": -3.4035227298736572,
      "logits/rejected": -3.4438557624816895,
      "logps/chosen": -27.038997650146484,
      "logps/ref_chosen": -27.20970916748047,
      "logps/ref_rejected": -106.27947998046875,
      "logps/rejected": -107.220458984375,
      "loss": 1.2883,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": 0.015648098662495613,
      "rewards/margins": 0.10201247036457062,
      "rewards/rejected": -0.08636436611413956,
      "step": 24
    },
    {
      "epoch": 0.03671071953010279,
      "epsilon_dpo/beta": 0.09127078950405121,
      "epsilon_dpo/beta_margin_grad_mean": -0.47578126192092896,
      "epsilon_dpo/beta_margin_grad_std": 0.023359432816505432,
      "epsilon_dpo/beta_margin_mean": 0.09719490259885788,
      "epsilon_dpo/beta_margin_std": 0.09393724054098129,
      "epsilon_dpo/loss_margin_mean": 1.0689027309417725,
      "grad_norm": 128.98117065429688,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.09192235767841339,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 1.7391304347826085e-07,
      "logits/chosen": -3.487727403640747,
      "logits/rejected": -3.466651439666748,
      "logps/chosen": -36.42045593261719,
      "logps/ref_chosen": -36.47064208984375,
      "logps/ref_rejected": -93.89593505859375,
      "logps/rejected": -94.91465759277344,
      "loss": 1.2937,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": 0.004450969398021698,
      "rewards/margins": 0.09719488024711609,
      "rewards/rejected": -0.0927439033985138,
      "step": 25
    },
    {
      "epoch": 0.0381791483113069,
      "epsilon_dpo/beta": 0.0905909463763237,
      "epsilon_dpo/beta_margin_grad_mean": -0.4600984454154968,
      "epsilon_dpo/beta_margin_grad_std": 0.04119768738746643,
      "epsilon_dpo/beta_margin_mean": 0.16133488714694977,
      "epsilon_dpo/beta_margin_std": 0.1687132865190506,
      "epsilon_dpo/loss_margin_mean": 1.7871633768081665,
      "grad_norm": 150.258056640625,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.09126638621091843,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 1.8115942028985507e-07,
      "logits/chosen": -3.4571170806884766,
      "logits/rejected": -3.5464980602264404,
      "logps/chosen": -39.704803466796875,
      "logps/ref_chosen": -39.82624816894531,
      "logps/ref_rejected": -109.0130615234375,
      "logps/rejected": -110.67877197265625,
      "loss": 1.2385,
      "rewards/accuracies": 0.875,
      "rewards/chosen": 0.010786900296807289,
      "rewards/margins": 0.16133487224578857,
      "rewards/rejected": -0.15054798126220703,
      "step": 26
    },
    {
      "epoch": 0.039647577092511016,
      "epsilon_dpo/beta": 0.08971839398145676,
      "epsilon_dpo/beta_margin_grad_mean": -0.4532409906387329,
      "epsilon_dpo/beta_margin_grad_std": 0.02980455383658409,
      "epsilon_dpo/beta_margin_mean": 0.1883021891117096,
      "epsilon_dpo/beta_margin_std": 0.1207134947180748,
      "epsilon_dpo/loss_margin_mean": 2.099519729614258,
      "grad_norm": 152.13011169433594,
      "kl/avg_steps": 0.96875,
      "kl/beta": 0.09058698266744614,
      "kl/n_epsilon_steps": 0.015625,
      "kl/p_epsilon_steps": 0.984375,
      "learning_rate": 1.8840579710144927e-07,
      "logits/chosen": -3.417879104614258,
      "logits/rejected": -3.4293429851531982,
      "logps/chosen": -19.790260314941406,
      "logps/ref_chosen": -19.93426513671875,
      "logps/ref_rejected": -107.32525634765625,
      "logps/rejected": -109.28077697753906,
      "loss": 1.2104,
      "rewards/accuracies": 0.984375,
      "rewards/chosen": 0.012909766286611557,
      "rewards/margins": 0.18830212950706482,
      "rewards/rejected": -0.17539237439632416,
      "step": 27
    },
    {
      "epoch": 0.041116005873715125,
      "epsilon_dpo/beta": 0.08888562768697739,
      "epsilon_dpo/beta_margin_grad_mean": -0.4523911476135254,
      "epsilon_dpo/beta_margin_grad_std": 0.03366325423121452,
      "epsilon_dpo/beta_margin_mean": 0.1920655220746994,
      "epsilon_dpo/beta_margin_std": 0.13740301132202148,
      "epsilon_dpo/loss_margin_mean": 2.162940263748169,
      "grad_norm": 144.26730346679688,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.08971784263849258,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 1.9565217391304347e-07,
      "logits/chosen": -3.5261058807373047,
      "logits/rejected": -3.475046157836914,
      "logps/chosen": -43.32648849487305,
      "logps/ref_chosen": -43.6025390625,
      "logps/ref_rejected": -96.1494140625,
      "logps/rejected": -98.03630065917969,
      "loss": 1.2081,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": 0.02439035288989544,
      "rewards/margins": 0.19206556677818298,
      "rewards/rejected": -0.1676751971244812,
      "step": 28
    },
    {
      "epoch": 0.042584434654919234,
      "epsilon_dpo/beta": 0.08808783441781998,
      "epsilon_dpo/beta_margin_grad_mean": -0.4418832063674927,
      "epsilon_dpo/beta_margin_grad_std": 0.04080166295170784,
      "epsilon_dpo/beta_margin_mean": 0.23540179431438446,
      "epsilon_dpo/beta_margin_std": 0.1678141951560974,
      "epsilon_dpo/loss_margin_mean": 2.675217866897583,
      "grad_norm": 141.24661254882812,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.08888454735279083,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 2.028985507246377e-07,
      "logits/chosen": -3.4727797508239746,
      "logits/rejected": -3.4470648765563965,
      "logps/chosen": -32.09503936767578,
      "logps/ref_chosen": -32.44408416748047,
      "logps/ref_rejected": -101.98307037353516,
      "logps/rejected": -104.30924224853516,
      "loss": 1.1716,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": 0.03065740317106247,
      "rewards/margins": 0.2354017198085785,
      "rewards/rejected": -0.20474430918693542,
      "step": 29
    },
    {
      "epoch": 0.04405286343612335,
      "epsilon_dpo/beta": 0.08729670941829681,
      "epsilon_dpo/beta_margin_grad_mean": -0.44231948256492615,
      "epsilon_dpo/beta_margin_grad_std": 0.04499894008040428,
      "epsilon_dpo/beta_margin_mean": 0.23390042781829834,
      "epsilon_dpo/beta_margin_std": 0.1849268078804016,
      "epsilon_dpo/loss_margin_mean": 2.6831705570220947,
      "grad_norm": 139.83612060546875,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.08808626234531403,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 2.1014492753623187e-07,
      "logits/chosen": -3.5117549896240234,
      "logits/rejected": -3.389063835144043,
      "logps/chosen": -39.259483337402344,
      "logps/ref_chosen": -39.2830810546875,
      "logps/ref_rejected": -103.8922119140625,
      "logps/rejected": -106.5517807006836,
      "loss": 1.1744,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": 0.0018627983517944813,
      "rewards/margins": 0.23390044271945953,
      "rewards/rejected": -0.23203766345977783,
      "step": 30
    },
    {
      "epoch": 0.04552129221732746,
      "epsilon_dpo/beta": 0.08656726032495499,
      "epsilon_dpo/beta_margin_grad_mean": -0.44433295726776123,
      "epsilon_dpo/beta_margin_grad_std": 0.04770767688751221,
      "epsilon_dpo/beta_margin_mean": 0.22623342275619507,
      "epsilon_dpo/beta_margin_std": 0.1985906958580017,
      "epsilon_dpo/loss_margin_mean": 2.6196024417877197,
      "grad_norm": 121.26469421386719,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.08729515224695206,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 2.1739130434782607e-07,
      "logits/chosen": -3.4779539108276367,
      "logits/rejected": -3.431878089904785,
      "logps/chosen": -35.737491607666016,
      "logps/ref_chosen": -36.05577850341797,
      "logps/ref_rejected": -78.35195922851562,
      "logps/rejected": -80.65327453613281,
      "loss": 1.1824,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": 0.02715596929192543,
      "rewards/margins": 0.22623339295387268,
      "rewards/rejected": -0.19907742738723755,
      "step": 31
    },
    {
      "epoch": 0.04698972099853157,
      "epsilon_dpo/beta": 0.08576179295778275,
      "epsilon_dpo/beta_margin_grad_mean": -0.42864087224006653,
      "epsilon_dpo/beta_margin_grad_std": 0.05508127063512802,
      "epsilon_dpo/beta_margin_mean": 0.292092889547348,
      "epsilon_dpo/beta_margin_std": 0.2318342626094818,
      "epsilon_dpo/loss_margin_mean": 3.4083411693573,
      "grad_norm": 131.14877319335938,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.08656476438045502,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 2.2463768115942027e-07,
      "logits/chosen": -3.4760398864746094,
      "logits/rejected": -3.5285589694976807,
      "logps/chosen": -32.82677459716797,
      "logps/ref_chosen": -33.10527420043945,
      "logps/ref_rejected": -95.47318267822266,
      "logps/rejected": -98.60302734375,
      "loss": 1.1284,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": 0.023832818493247032,
      "rewards/margins": 0.2920929789543152,
      "rewards/rejected": -0.2682601809501648,
      "step": 32
    },
    {
      "epoch": 0.048458149779735685,
      "epsilon_dpo/beta": 0.08496525138616562,
      "epsilon_dpo/beta_margin_grad_mean": -0.43119847774505615,
      "epsilon_dpo/beta_margin_grad_std": 0.058498919010162354,
      "epsilon_dpo/beta_margin_mean": 0.2807844281196594,
      "epsilon_dpo/beta_margin_std": 0.24786221981048584,
      "epsilon_dpo/loss_margin_mean": 3.309786796569824,
      "grad_norm": 127.04460906982422,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.08576075732707977,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 2.318840579710145e-07,
      "logits/chosen": -3.5147736072540283,
      "logits/rejected": -3.4672505855560303,
      "logps/chosen": -40.38764953613281,
      "logps/ref_chosen": -40.39752960205078,
      "logps/ref_rejected": -94.48348999023438,
      "logps/rejected": -97.78340148925781,
      "loss": 1.1399,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": 0.00047481246292591095,
      "rewards/margins": 0.28078436851501465,
      "rewards/rejected": -0.2803095579147339,
      "step": 33
    },
    {
      "epoch": 0.049926578560939794,
      "epsilon_dpo/beta": 0.08430886268615723,
      "epsilon_dpo/beta_margin_grad_mean": -0.42384782433509827,
      "epsilon_dpo/beta_margin_grad_std": 0.06483103334903717,
      "epsilon_dpo/beta_margin_mean": 0.3134419322013855,
      "epsilon_dpo/beta_margin_std": 0.27257490158081055,
      "epsilon_dpo/loss_margin_mean": 3.726898193359375,
      "grad_norm": 127.52790069580078,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.08496421575546265,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 2.391304347826087e-07,
      "logits/chosen": -3.4469704627990723,
      "logits/rejected": -3.5294151306152344,
      "logps/chosen": -35.16386413574219,
      "logps/ref_chosen": -35.10262680053711,
      "logps/ref_rejected": -106.70514678955078,
      "logps/rejected": -110.49327850341797,
      "loss": 1.1152,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.0053707570768892765,
      "rewards/margins": 0.3134418725967407,
      "rewards/rejected": -0.31881266832351685,
      "step": 34
    },
    {
      "epoch": 0.0513950073421439,
      "epsilon_dpo/beta": 0.0835762619972229,
      "epsilon_dpo/beta_margin_grad_mean": -0.4079773724079132,
      "epsilon_dpo/beta_margin_grad_std": 0.07277688384056091,
      "epsilon_dpo/beta_margin_mean": 0.3832019567489624,
      "epsilon_dpo/beta_margin_std": 0.3155387043952942,
      "epsilon_dpo/loss_margin_mean": 4.590485095977783,
      "grad_norm": 120.23053741455078,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.08430557698011398,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 2.463768115942029e-07,
      "logits/chosen": -3.4881181716918945,
      "logits/rejected": -3.5153722763061523,
      "logps/chosen": -34.6436767578125,
      "logps/ref_chosen": -34.41180419921875,
      "logps/ref_rejected": -111.88399505615234,
      "logps/rejected": -116.70634460449219,
      "loss": 1.0629,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.019653314724564552,
      "rewards/margins": 0.3832019567489624,
      "rewards/rejected": -0.4028552770614624,
      "step": 35
    },
    {
      "epoch": 0.05286343612334802,
      "epsilon_dpo/beta": 0.08292967081069946,
      "epsilon_dpo/beta_margin_grad_mean": -0.3959466516971588,
      "epsilon_dpo/beta_margin_grad_std": 0.09478563815355301,
      "epsilon_dpo/beta_margin_mean": 0.4413573443889618,
      "epsilon_dpo/beta_margin_std": 0.41608670353889465,
      "epsilon_dpo/loss_margin_mean": 5.337898254394531,
      "grad_norm": 105.79562377929688,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.08357430249452591,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 2.536231884057971e-07,
      "logits/chosen": -3.4250617027282715,
      "logits/rejected": -3.506716251373291,
      "logps/chosen": -32.58708572387695,
      "logps/ref_chosen": -32.743473052978516,
      "logps/ref_rejected": -90.1633529663086,
      "logps/rejected": -95.34486389160156,
      "loss": 1.0334,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": 0.012363580986857414,
      "rewards/margins": 0.4413573741912842,
      "rewards/rejected": -0.428993821144104,
      "step": 36
    },
    {
      "epoch": 0.05433186490455213,
      "epsilon_dpo/beta": 0.08228681236505508,
      "epsilon_dpo/beta_margin_grad_mean": -0.3971908390522003,
      "epsilon_dpo/beta_margin_grad_std": 0.09092327207326889,
      "epsilon_dpo/beta_margin_mean": 0.4376983940601349,
      "epsilon_dpo/beta_margin_std": 0.4152112901210785,
      "epsilon_dpo/loss_margin_mean": 5.334095478057861,
      "grad_norm": 100.5900650024414,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.082926444709301,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 2.6086956521739126e-07,
      "logits/chosen": -3.4775028228759766,
      "logits/rejected": -3.44918155670166,
      "logps/chosen": -40.18016052246094,
      "logps/ref_chosen": -39.88025665283203,
      "logps/ref_rejected": -83.95890808105469,
      "logps/rejected": -89.59291076660156,
      "loss": 1.0351,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.025536730885505676,
      "rewards/margins": 0.4376984238624573,
      "rewards/rejected": -0.46323513984680176,
      "step": 37
    },
    {
      "epoch": 0.055800293685756244,
      "epsilon_dpo/beta": 0.08167463541030884,
      "epsilon_dpo/beta_margin_grad_mean": -0.36130231618881226,
      "epsilon_dpo/beta_margin_grad_std": 0.11581370234489441,
      "epsilon_dpo/beta_margin_mean": 0.6206669211387634,
      "epsilon_dpo/beta_margin_std": 0.5732053518295288,
      "epsilon_dpo/loss_margin_mean": 7.621293067932129,
      "grad_norm": 99.69669342041016,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.08228360116481781,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 2.681159420289855e-07,
      "logits/chosen": -3.436084270477295,
      "logits/rejected": -3.5204224586486816,
      "logps/chosen": -33.92012405395508,
      "logps/ref_chosen": -33.85154342651367,
      "logps/ref_rejected": -104.96053314208984,
      "logps/rejected": -112.65040588378906,
      "loss": 0.9294,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.0062013790011405945,
      "rewards/margins": 0.6206669807434082,
      "rewards/rejected": -0.6268683671951294,
      "step": 38
    },
    {
      "epoch": 0.05726872246696035,
      "epsilon_dpo/beta": 0.0809134915471077,
      "epsilon_dpo/beta_margin_grad_mean": -0.35245731472969055,
      "epsilon_dpo/beta_margin_grad_std": 0.09772588312625885,
      "epsilon_dpo/beta_margin_mean": 0.6451180577278137,
      "epsilon_dpo/beta_margin_std": 0.48367542028427124,
      "epsilon_dpo/loss_margin_mean": 7.98002290725708,
      "grad_norm": 102.14472198486328,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.08167106658220291,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 2.753623188405797e-07,
      "logits/chosen": -3.435272216796875,
      "logits/rejected": -3.441884994506836,
      "logps/chosen": -31.436540603637695,
      "logps/ref_chosen": -31.883747100830078,
      "logps/ref_rejected": -84.24908447265625,
      "logps/rejected": -91.78189086914062,
      "loss": 0.8922,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": 0.035918496549129486,
      "rewards/margins": 0.645117998123169,
      "rewards/rejected": -0.6091995239257812,
      "step": 39
    },
    {
      "epoch": 0.05873715124816446,
      "epsilon_dpo/beta": 0.08031369745731354,
      "epsilon_dpo/beta_margin_grad_mean": -0.35732436180114746,
      "epsilon_dpo/beta_margin_grad_std": 0.10807473212480545,
      "epsilon_dpo/beta_margin_mean": 0.6359905004501343,
      "epsilon_dpo/beta_margin_std": 0.5488929152488708,
      "epsilon_dpo/loss_margin_mean": 7.938919544219971,
      "grad_norm": 94.2869644165039,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0809125155210495,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 2.8260869565217386e-07,
      "logits/chosen": -3.48494291305542,
      "logits/rejected": -3.507589101791382,
      "logps/chosen": -36.815399169921875,
      "logps/ref_chosen": -36.59412384033203,
      "logps/ref_rejected": -80.92609405517578,
      "logps/rejected": -89.08628845214844,
      "loss": 0.9118,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.018150903284549713,
      "rewards/margins": 0.6359904408454895,
      "rewards/rejected": -0.654141366481781,
      "step": 40
    },
    {
      "epoch": 0.06020558002936858,
      "epsilon_dpo/beta": 0.07975336164236069,
      "epsilon_dpo/beta_margin_grad_mean": -0.3376636803150177,
      "epsilon_dpo/beta_margin_grad_std": 0.1441744863986969,
      "epsilon_dpo/beta_margin_mean": 0.762270450592041,
      "epsilon_dpo/beta_margin_std": 0.7963690161705017,
      "epsilon_dpo/loss_margin_mean": 9.593799591064453,
      "grad_norm": 101.93110656738281,
      "kl/avg_steps": 0.703125,
      "kl/beta": 0.08031018823385239,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 2.898550724637681e-07,
      "logits/chosen": -3.483686923980713,
      "logits/rejected": -3.4968438148498535,
      "logps/chosen": -41.283260345458984,
      "logps/ref_chosen": -39.986053466796875,
      "logps/ref_rejected": -105.53334045410156,
      "logps/rejected": -116.42434692382812,
      "loss": 0.889,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.10497994720935822,
      "rewards/margins": 0.762270450592041,
      "rewards/rejected": -0.8672504425048828,
      "step": 41
    },
    {
      "epoch": 0.06167400881057269,
      "epsilon_dpo/beta": 0.07915924489498138,
      "epsilon_dpo/beta_margin_grad_mean": -0.3248225450515747,
      "epsilon_dpo/beta_margin_grad_std": 0.14031846821308136,
      "epsilon_dpo/beta_margin_mean": 0.8673705458641052,
      "epsilon_dpo/beta_margin_std": 0.8684898018836975,
      "epsilon_dpo/loss_margin_mean": 10.988822937011719,
      "grad_norm": 98.95552062988281,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0797494500875473,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 2.971014492753623e-07,
      "logits/chosen": -3.436079502105713,
      "logits/rejected": -3.551501989364624,
      "logps/chosen": -47.57762145996094,
      "logps/ref_chosen": -45.769351959228516,
      "logps/ref_rejected": -118.03570556640625,
      "logps/rejected": -130.83279418945312,
      "loss": 0.8297,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.1441619098186493,
      "rewards/margins": 0.86737060546875,
      "rewards/rejected": -1.0115324258804321,
      "step": 42
    },
    {
      "epoch": 0.0631424375917768,
      "epsilon_dpo/beta": 0.07852049171924591,
      "epsilon_dpo/beta_margin_grad_mean": -0.29879364371299744,
      "epsilon_dpo/beta_margin_grad_std": 0.14372758567333221,
      "epsilon_dpo/beta_margin_mean": 0.9733436703681946,
      "epsilon_dpo/beta_margin_std": 0.7786263227462769,
      "epsilon_dpo/loss_margin_mean": 12.422151565551758,
      "grad_norm": 86.0045166015625,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.07915578037500381,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 3.043478260869565e-07,
      "logits/chosen": -3.4531750679016113,
      "logits/rejected": -3.502927303314209,
      "logps/chosen": -36.81052017211914,
      "logps/ref_chosen": -36.684478759765625,
      "logps/ref_rejected": -104.91730499267578,
      "logps/rejected": -117.46549224853516,
      "loss": 0.7552,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.010491969995200634,
      "rewards/margins": 0.9733436703681946,
      "rewards/rejected": -0.9838355779647827,
      "step": 43
    },
    {
      "epoch": 0.06461086637298091,
      "epsilon_dpo/beta": 0.07783857733011246,
      "epsilon_dpo/beta_margin_grad_mean": -0.2732856273651123,
      "epsilon_dpo/beta_margin_grad_std": 0.15621285140514374,
      "epsilon_dpo/beta_margin_mean": 1.1567952632904053,
      "epsilon_dpo/beta_margin_std": 0.9610118269920349,
      "epsilon_dpo/loss_margin_mean": 14.889684677124023,
      "grad_norm": 85.54306030273438,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.07851782441139221,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 3.115942028985507e-07,
      "logits/chosen": -3.3832082748413086,
      "logits/rejected": -3.4922008514404297,
      "logps/chosen": -28.09503936767578,
      "logps/ref_chosen": -27.785930633544922,
      "logps/ref_rejected": -117.58551788330078,
      "logps/rejected": -132.7843017578125,
      "loss": 0.6985,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.02512773871421814,
      "rewards/margins": 1.1567951440811157,
      "rewards/rejected": -1.1819229125976562,
      "step": 44
    },
    {
      "epoch": 0.06607929515418502,
      "epsilon_dpo/beta": 0.07733368128538132,
      "epsilon_dpo/beta_margin_grad_mean": -0.3332192301750183,
      "epsilon_dpo/beta_margin_grad_std": 0.15007154643535614,
      "epsilon_dpo/beta_margin_mean": 0.8284286260604858,
      "epsilon_dpo/beta_margin_std": 0.8817453980445862,
      "epsilon_dpo/loss_margin_mean": 10.75483226776123,
      "grad_norm": 86.84064483642578,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.07783675193786621,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 3.188405797101449e-07,
      "logits/chosen": -3.4090023040771484,
      "logits/rejected": -3.488529920578003,
      "logps/chosen": -38.12892150878906,
      "logps/ref_chosen": -36.33074951171875,
      "logps/ref_rejected": -83.34062194824219,
      "logps/rejected": -95.89362335205078,
      "loss": 0.8642,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.14034625887870789,
      "rewards/margins": 0.8284286260604858,
      "rewards/rejected": -0.9687749147415161,
      "step": 45
    },
    {
      "epoch": 0.06754772393538913,
      "epsilon_dpo/beta": 0.07675698399543762,
      "epsilon_dpo/beta_margin_grad_mean": -0.2891290783882141,
      "epsilon_dpo/beta_margin_grad_std": 0.1580805480480194,
      "epsilon_dpo/beta_margin_mean": 1.1003601551055908,
      "epsilon_dpo/beta_margin_std": 1.0318849086761475,
      "epsilon_dpo/loss_margin_mean": 14.38028621673584,
      "grad_norm": 96.92198181152344,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.07732927799224854,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 3.260869565217391e-07,
      "logits/chosen": -3.375389814376831,
      "logits/rejected": -3.439635992050171,
      "logps/chosen": -37.75640106201172,
      "logps/ref_chosen": -36.0171012878418,
      "logps/ref_rejected": -101.40194702148438,
      "logps/rejected": -117.52153015136719,
      "loss": 0.7386,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.13517028093338013,
      "rewards/margins": 1.1003601551055908,
      "rewards/rejected": -1.2355304956436157,
      "step": 46
    },
    {
      "epoch": 0.06901615271659324,
      "epsilon_dpo/beta": 0.07616160064935684,
      "epsilon_dpo/beta_margin_grad_mean": -0.31451866030693054,
      "epsilon_dpo/beta_margin_grad_std": 0.16808247566223145,
      "epsilon_dpo/beta_margin_mean": 0.9304282069206238,
      "epsilon_dpo/beta_margin_std": 1.0573234558105469,
      "epsilon_dpo/loss_margin_mean": 12.260807037353516,
      "grad_norm": 94.45354461669922,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.07675362378358841,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 3.333333333333333e-07,
      "logits/chosen": -3.4388890266418457,
      "logits/rejected": -3.5206193923950195,
      "logps/chosen": -44.316593170166016,
      "logps/ref_chosen": -41.82904815673828,
      "logps/ref_rejected": -101.29283142089844,
      "logps/rejected": -116.04118347167969,
      "loss": 0.8604,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.19145357608795166,
      "rewards/margins": 0.9304282665252686,
      "rewards/rejected": -1.1218818426132202,
      "step": 47
    },
    {
      "epoch": 0.07048458149779736,
      "epsilon_dpo/beta": 0.07571401447057724,
      "epsilon_dpo/beta_margin_grad_mean": -0.29027602076530457,
      "epsilon_dpo/beta_margin_grad_std": 0.17632031440734863,
      "epsilon_dpo/beta_margin_mean": 1.1656516790390015,
      "epsilon_dpo/beta_margin_std": 1.2263123989105225,
      "epsilon_dpo/loss_margin_mean": 15.469880104064941,
      "grad_norm": 90.87052154541016,
      "kl/avg_steps": 0.59375,
      "kl/beta": 0.07615863531827927,
      "kl/n_epsilon_steps": 0.203125,
      "kl/p_epsilon_steps": 0.796875,
      "learning_rate": 3.4057971014492755e-07,
      "logits/chosen": -3.463271141052246,
      "logits/rejected": -3.4467902183532715,
      "logps/chosen": -38.7716064453125,
      "logps/ref_chosen": -36.18339920043945,
      "logps/ref_rejected": -95.41502380371094,
      "logps/rejected": -113.47309875488281,
      "loss": 0.7607,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.19789756834506989,
      "rewards/margins": 1.1656516790390015,
      "rewards/rejected": -1.3635492324829102,
      "step": 48
    },
    {
      "epoch": 0.07195301027900147,
      "epsilon_dpo/beta": 0.075148805975914,
      "epsilon_dpo/beta_margin_grad_mean": -0.2587027847766876,
      "epsilon_dpo/beta_margin_grad_std": 0.17777220904827118,
      "epsilon_dpo/beta_margin_mean": 1.3605566024780273,
      "epsilon_dpo/beta_margin_std": 1.2057913541793823,
      "epsilon_dpo/loss_margin_mean": 18.15953826904297,
      "grad_norm": 81.80361938476562,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.07570911198854446,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 3.478260869565217e-07,
      "logits/chosen": -3.472126007080078,
      "logits/rejected": -3.5396313667297363,
      "logps/chosen": -37.51390075683594,
      "logps/ref_chosen": -35.96546936035156,
      "logps/ref_rejected": -91.31820678710938,
      "logps/rejected": -111.02618408203125,
      "loss": 0.6703,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.118332639336586,
      "rewards/margins": 1.3605566024780273,
      "rewards/rejected": -1.4788892269134521,
      "step": 49
    },
    {
      "epoch": 0.07342143906020558,
      "epsilon_dpo/beta": 0.07468333095312119,
      "epsilon_dpo/beta_margin_grad_mean": -0.27890822291374207,
      "epsilon_dpo/beta_margin_grad_std": 0.21263383328914642,
      "epsilon_dpo/beta_margin_mean": 1.3234366178512573,
      "epsilon_dpo/beta_margin_std": 1.4268280267715454,
      "epsilon_dpo/loss_margin_mean": 17.81157684326172,
      "grad_norm": 100.63481903076172,
      "kl/avg_steps": 0.625,
      "kl/beta": 0.07514552026987076,
      "kl/n_epsilon_steps": 0.1875,
      "kl/p_epsilon_steps": 0.8125,
      "learning_rate": 3.5507246376811595e-07,
      "logits/chosen": -3.4349403381347656,
      "logits/rejected": -3.533860445022583,
      "logps/chosen": -46.16408920288086,
      "logps/ref_chosen": -42.138206481933594,
      "logps/ref_rejected": -100.4173583984375,
      "logps/rejected": -122.25482177734375,
      "loss": 0.7726,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.30415183305740356,
      "rewards/margins": 1.3234367370605469,
      "rewards/rejected": -1.6275885105133057,
      "step": 50
    },
    {
      "epoch": 0.07488986784140969,
      "epsilon_dpo/beta": 0.07421945780515671,
      "epsilon_dpo/beta_margin_grad_mean": -0.28537219762802124,
      "epsilon_dpo/beta_margin_grad_std": 0.19426609575748444,
      "epsilon_dpo/beta_margin_mean": 1.3490850925445557,
      "epsilon_dpo/beta_margin_std": 1.5637013912200928,
      "epsilon_dpo/loss_margin_mean": 18.262516021728516,
      "grad_norm": 91.88065338134766,
      "kl/avg_steps": 0.625,
      "kl/beta": 0.07467877864837646,
      "kl/n_epsilon_steps": 0.1875,
      "kl/p_epsilon_steps": 0.8125,
      "learning_rate": 3.6231884057971015e-07,
      "logits/chosen": -3.377105236053467,
      "logits/rejected": -3.4192728996276855,
      "logps/chosen": -43.83557891845703,
      "logps/ref_chosen": -39.016597747802734,
      "logps/ref_rejected": -80.60652160644531,
      "logps/rejected": -103.68801879882812,
      "loss": 0.7605,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.3597980737686157,
      "rewards/margins": 1.3490850925445557,
      "rewards/rejected": -1.7088831663131714,
      "step": 51
    },
    {
      "epoch": 0.0763582966226138,
      "epsilon_dpo/beta": 0.07364249974489212,
      "epsilon_dpo/beta_margin_grad_mean": -0.23005272448062897,
      "epsilon_dpo/beta_margin_grad_std": 0.1819276660680771,
      "epsilon_dpo/beta_margin_mean": 1.7073160409927368,
      "epsilon_dpo/beta_margin_std": 1.5912011861801147,
      "epsilon_dpo/loss_margin_mean": 23.246597290039062,
      "grad_norm": 87.76582336425781,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.07421493530273438,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 3.695652173913043e-07,
      "logits/chosen": -3.4000444412231445,
      "logits/rejected": -3.473278284072876,
      "logps/chosen": -39.097511291503906,
      "logps/ref_chosen": -34.285945892333984,
      "logps/ref_rejected": -85.96109008789062,
      "logps/rejected": -114.01925659179688,
      "loss": 0.6001,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.3565700650215149,
      "rewards/margins": 1.7073161602020264,
      "rewards/rejected": -2.0638861656188965,
      "step": 52
    },
    {
      "epoch": 0.07782672540381791,
      "epsilon_dpo/beta": 0.07307162135839462,
      "epsilon_dpo/beta_margin_grad_mean": -0.21757791936397552,
      "epsilon_dpo/beta_margin_grad_std": 0.17687389254570007,
      "epsilon_dpo/beta_margin_mean": 1.8149864673614502,
      "epsilon_dpo/beta_margin_std": 1.6574300527572632,
      "epsilon_dpo/loss_margin_mean": 24.905494689941406,
      "grad_norm": 85.12645721435547,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.07363962382078171,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 3.7681159420289855e-07,
      "logits/chosen": -3.3786673545837402,
      "logits/rejected": -3.4282350540161133,
      "logps/chosen": -42.113983154296875,
      "logps/ref_chosen": -34.706817626953125,
      "logps/ref_rejected": -97.64952087402344,
      "logps/rejected": -129.96218872070312,
      "loss": 0.558,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.5423527359962463,
      "rewards/margins": 1.8149864673614502,
      "rewards/rejected": -2.3573391437530518,
      "step": 53
    },
    {
      "epoch": 0.07929515418502203,
      "epsilon_dpo/beta": 0.07252801209688187,
      "epsilon_dpo/beta_margin_grad_mean": -0.23283910751342773,
      "epsilon_dpo/beta_margin_grad_std": 0.18114687502384186,
      "epsilon_dpo/beta_margin_mean": 1.7452175617218018,
      "epsilon_dpo/beta_margin_std": 1.7455625534057617,
      "epsilon_dpo/loss_margin_mean": 24.13280487060547,
      "grad_norm": 83.51258850097656,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0730687752366066,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 3.8405797101449274e-07,
      "logits/chosen": -3.400343656539917,
      "logits/rejected": -3.5012292861938477,
      "logps/chosen": -46.09777069091797,
      "logps/ref_chosen": -39.777854919433594,
      "logps/ref_rejected": -98.70614624023438,
      "logps/rejected": -129.1588592529297,
      "loss": 0.6017,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.46132320165634155,
      "rewards/margins": 1.7452175617218018,
      "rewards/rejected": -2.206540822982788,
      "step": 54
    },
    {
      "epoch": 0.08076358296622614,
      "epsilon_dpo/beta": 0.07214676588773727,
      "epsilon_dpo/beta_margin_grad_mean": -0.2661394476890564,
      "epsilon_dpo/beta_margin_grad_std": 0.21455174684524536,
      "epsilon_dpo/beta_margin_mean": 1.643436074256897,
      "epsilon_dpo/beta_margin_std": 1.8868768215179443,
      "epsilon_dpo/loss_margin_mean": 22.906444549560547,
      "grad_norm": 104.0904541015625,
      "kl/avg_steps": 0.53125,
      "kl/beta": 0.07252483814954758,
      "kl/n_epsilon_steps": 0.234375,
      "kl/p_epsilon_steps": 0.765625,
      "learning_rate": 3.9130434782608694e-07,
      "logits/chosen": -3.390665054321289,
      "logits/rejected": -3.4186453819274902,
      "logps/chosen": -58.18317794799805,
      "logps/ref_chosen": -48.42914962768555,
      "logps/ref_rejected": -93.72831726074219,
      "logps/rejected": -126.38877868652344,
      "loss": 0.7261,
      "rewards/accuracies": 0.796875,
      "rewards/chosen": -0.7076321840286255,
      "rewards/margins": 1.643436074256897,
      "rewards/rejected": -2.3510682582855225,
      "step": 55
    },
    {
      "epoch": 0.08223201174743025,
      "epsilon_dpo/beta": 0.07168648391962051,
      "epsilon_dpo/beta_margin_grad_mean": -0.2267267107963562,
      "epsilon_dpo/beta_margin_grad_std": 0.24484246969223022,
      "epsilon_dpo/beta_margin_mean": 1.977746605873108,
      "epsilon_dpo/beta_margin_std": 2.0250532627105713,
      "epsilon_dpo/loss_margin_mean": 27.73088836669922,
      "grad_norm": 123.8470458984375,
      "kl/avg_steps": 0.640625,
      "kl/beta": 0.07214158773422241,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.8125,
      "learning_rate": 3.9855072463768114e-07,
      "logits/chosen": -3.2844839096069336,
      "logits/rejected": -3.3951363563537598,
      "logps/chosen": -38.82612991333008,
      "logps/ref_chosen": -31.692344665527344,
      "logps/ref_rejected": -100.61968994140625,
      "logps/rejected": -135.484375,
      "loss": 0.713,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.5154626369476318,
      "rewards/margins": 1.9777467250823975,
      "rewards/rejected": -2.4932093620300293,
      "step": 56
    },
    {
      "epoch": 0.08370044052863436,
      "epsilon_dpo/beta": 0.07115186750888824,
      "epsilon_dpo/beta_margin_grad_mean": -0.21097464859485626,
      "epsilon_dpo/beta_margin_grad_std": 0.22179222106933594,
      "epsilon_dpo/beta_margin_mean": 2.176724910736084,
      "epsilon_dpo/beta_margin_std": 2.1261887550354004,
      "epsilon_dpo/loss_margin_mean": 30.698436737060547,
      "grad_norm": 102.96115112304688,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.07168237119913101,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 4.057971014492754e-07,
      "logits/chosen": -3.386384963989258,
      "logits/rejected": -3.439547538757324,
      "logps/chosen": -45.702720642089844,
      "logps/ref_chosen": -38.302345275878906,
      "logps/ref_rejected": -101.74482727050781,
      "logps/rejected": -139.84365844726562,
      "loss": 0.6021,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.5308903455734253,
      "rewards/margins": 2.176724910736084,
      "rewards/rejected": -2.707615375518799,
      "step": 57
    },
    {
      "epoch": 0.08516886930983847,
      "epsilon_dpo/beta": 0.07062220573425293,
      "epsilon_dpo/beta_margin_grad_mean": -0.1960882991552353,
      "epsilon_dpo/beta_margin_grad_std": 0.17183293402194977,
      "epsilon_dpo/beta_margin_mean": 2.15674090385437,
      "epsilon_dpo/beta_margin_std": 2.0642378330230713,
      "epsilon_dpo/loss_margin_mean": 30.6201114654541,
      "grad_norm": 87.39006805419922,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.07114876061677933,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 4.1304347826086954e-07,
      "logits/chosen": -3.3047280311584473,
      "logits/rejected": -3.393859386444092,
      "logps/chosen": -46.135406494140625,
      "logps/ref_chosen": -38.44845962524414,
      "logps/ref_rejected": -89.55912780761719,
      "logps/rejected": -127.86618041992188,
      "loss": 0.4979,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.5473577976226807,
      "rewards/margins": 2.15674090385437,
      "rewards/rejected": -2.704098701477051,
      "step": 58
    },
    {
      "epoch": 0.08663729809104258,
      "epsilon_dpo/beta": 0.0700632631778717,
      "epsilon_dpo/beta_margin_grad_mean": -0.21717199683189392,
      "epsilon_dpo/beta_margin_grad_std": 0.2148067206144333,
      "epsilon_dpo/beta_margin_mean": 2.177267074584961,
      "epsilon_dpo/beta_margin_std": 2.287863254547119,
      "epsilon_dpo/loss_margin_mean": 31.16427993774414,
      "grad_norm": 87.90422058105469,
      "kl/avg_steps": 0.796875,
      "kl/beta": 0.0706191137433052,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 4.2028985507246374e-07,
      "logits/chosen": -3.3299951553344727,
      "logits/rejected": -3.408975124359131,
      "logps/chosen": -46.44553756713867,
      "logps/ref_chosen": -38.029998779296875,
      "logps/ref_rejected": -94.10072326660156,
      "logps/rejected": -133.6805419921875,
      "loss": 0.6172,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.5922181010246277,
      "rewards/margins": 2.177267074584961,
      "rewards/rejected": -2.7694854736328125,
      "step": 59
    },
    {
      "epoch": 0.0881057268722467,
      "epsilon_dpo/beta": 0.06962990015745163,
      "epsilon_dpo/beta_margin_grad_mean": -0.24627114832401276,
      "epsilon_dpo/beta_margin_grad_std": 0.2656000852584839,
      "epsilon_dpo/beta_margin_mean": 1.7643014192581177,
      "epsilon_dpo/beta_margin_std": 2.2814714908599854,
      "epsilon_dpo/loss_margin_mean": 25.50347328186035,
      "grad_norm": 167.74856567382812,
      "kl/avg_steps": 0.625,
      "kl/beta": 0.07006081938743591,
      "kl/n_epsilon_steps": 0.1875,
      "kl/p_epsilon_steps": 0.8125,
      "learning_rate": 4.2753623188405794e-07,
      "logits/chosen": -3.3077573776245117,
      "logits/rejected": -3.2960987091064453,
      "logps/chosen": -62.78964614868164,
      "logps/ref_chosen": -48.789947509765625,
      "logps/ref_rejected": -91.3543701171875,
      "logps/rejected": -130.8575439453125,
      "loss": 0.9335,
      "rewards/accuracies": 0.828125,
      "rewards/chosen": -0.9826135635375977,
      "rewards/margins": 1.7643013000488281,
      "rewards/rejected": -2.746914863586426,
      "step": 60
    },
    {
      "epoch": 0.08957415565345081,
      "epsilon_dpo/beta": 0.06911037117242813,
      "epsilon_dpo/beta_margin_grad_mean": -0.20471161603927612,
      "epsilon_dpo/beta_margin_grad_std": 0.22111034393310547,
      "epsilon_dpo/beta_margin_mean": 2.207209825515747,
      "epsilon_dpo/beta_margin_std": 2.3618977069854736,
      "epsilon_dpo/loss_margin_mean": 32.056705474853516,
      "grad_norm": 96.6867446899414,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.06962565332651138,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 4.3478260869565214e-07,
      "logits/chosen": -3.313138484954834,
      "logits/rejected": -3.4350461959838867,
      "logps/chosen": -47.05156707763672,
      "logps/ref_chosen": -35.972103118896484,
      "logps/ref_rejected": -95.45098876953125,
      "logps/rejected": -138.587158203125,
      "loss": 0.6385,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.7702499628067017,
      "rewards/margins": 2.207209825515747,
      "rewards/rejected": -2.9774599075317383,
      "step": 61
    },
    {
      "epoch": 0.09104258443465492,
      "epsilon_dpo/beta": 0.0686391070485115,
      "epsilon_dpo/beta_margin_grad_mean": -0.2324148714542389,
      "epsilon_dpo/beta_margin_grad_std": 0.20171168446540833,
      "epsilon_dpo/beta_margin_mean": 1.8510175943374634,
      "epsilon_dpo/beta_margin_std": 1.8581922054290771,
      "epsilon_dpo/loss_margin_mean": 27.06845474243164,
      "grad_norm": 91.73394775390625,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.06910735368728638,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 4.420289855072464e-07,
      "logits/chosen": -3.2718987464904785,
      "logits/rejected": -3.3174455165863037,
      "logps/chosen": -45.381500244140625,
      "logps/ref_chosen": -35.904327392578125,
      "logps/ref_rejected": -82.9093017578125,
      "logps/rejected": -119.45492553710938,
      "loss": 0.6249,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.6530295610427856,
      "rewards/margins": 1.8510175943374634,
      "rewards/rejected": -2.504047155380249,
      "step": 62
    },
    {
      "epoch": 0.09251101321585903,
      "epsilon_dpo/beta": 0.06817042827606201,
      "epsilon_dpo/beta_margin_grad_mean": -0.22793808579444885,
      "epsilon_dpo/beta_margin_grad_std": 0.2142634242773056,
      "epsilon_dpo/beta_margin_mean": 2.1623551845550537,
      "epsilon_dpo/beta_margin_std": 2.539600372314453,
      "epsilon_dpo/loss_margin_mean": 31.848644256591797,
      "grad_norm": 102.05279541015625,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0686354786157608,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 4.4927536231884053e-07,
      "logits/chosen": -3.3514161109924316,
      "logits/rejected": -3.4100210666656494,
      "logps/chosen": -60.41229248046875,
      "logps/ref_chosen": -46.25957107543945,
      "logps/ref_rejected": -104.15571594238281,
      "logps/rejected": -150.15708923339844,
      "loss": 0.6897,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.9690430164337158,
      "rewards/margins": 2.1623551845550537,
      "rewards/rejected": -3.1313982009887695,
      "step": 63
    },
    {
      "epoch": 0.09397944199706314,
      "epsilon_dpo/beta": 0.06770496070384979,
      "epsilon_dpo/beta_margin_grad_mean": -0.19847899675369263,
      "epsilon_dpo/beta_margin_grad_std": 0.19424547255039215,
      "epsilon_dpo/beta_margin_mean": 2.3565456867218018,
      "epsilon_dpo/beta_margin_std": 2.4017789363861084,
      "epsilon_dpo/loss_margin_mean": 34.92362594604492,
      "grad_norm": 95.53367614746094,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0681668370962143,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 4.5652173913043473e-07,
      "logits/chosen": -3.3310346603393555,
      "logits/rejected": -3.3657891750335693,
      "logps/chosen": -44.803466796875,
      "logps/ref_chosen": -34.512210845947266,
      "logps/ref_rejected": -101.21166229248047,
      "logps/rejected": -146.42654418945312,
      "loss": 0.5255,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.6991441249847412,
      "rewards/margins": 2.3565456867218018,
      "rewards/rejected": -3.055689811706543,
      "step": 64
    },
    {
      "epoch": 0.09544787077826726,
      "epsilon_dpo/beta": 0.06730613857507706,
      "epsilon_dpo/beta_margin_grad_mean": -0.21261604130268097,
      "epsilon_dpo/beta_margin_grad_std": 0.22306111454963684,
      "epsilon_dpo/beta_margin_mean": 2.48264741897583,
      "epsilon_dpo/beta_margin_std": 2.5546720027923584,
      "epsilon_dpo/loss_margin_mean": 37.058799743652344,
      "grad_norm": 93.06818389892578,
      "kl/avg_steps": 0.59375,
      "kl/beta": 0.06770138442516327,
      "kl/n_epsilon_steps": 0.203125,
      "kl/p_epsilon_steps": 0.796875,
      "learning_rate": 4.63768115942029e-07,
      "logits/chosen": -3.3164443969726562,
      "logits/rejected": -3.343629837036133,
      "logps/chosen": -53.97098159790039,
      "logps/ref_chosen": -43.620361328125,
      "logps/ref_rejected": -115.21531677246094,
      "logps/rejected": -162.62474060058594,
      "loss": 0.5986,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.6990154981613159,
      "rewards/margins": 2.48264741897583,
      "rewards/rejected": -3.1816627979278564,
      "step": 65
    },
    {
      "epoch": 0.09691629955947137,
      "epsilon_dpo/beta": 0.0669299066066742,
      "epsilon_dpo/beta_margin_grad_mean": -0.24468040466308594,
      "epsilon_dpo/beta_margin_grad_std": 0.2503945231437683,
      "epsilon_dpo/beta_margin_mean": 2.0747146606445312,
      "epsilon_dpo/beta_margin_std": 2.3843464851379395,
      "epsilon_dpo/loss_margin_mean": 31.18071746826172,
      "grad_norm": 116.8039321899414,
      "kl/avg_steps": 0.5625,
      "kl/beta": 0.06730178743600845,
      "kl/n_epsilon_steps": 0.21875,
      "kl/p_epsilon_steps": 0.78125,
      "learning_rate": 4.7101449275362313e-07,
      "logits/chosen": -3.2685537338256836,
      "logits/rejected": -3.3573737144470215,
      "logps/chosen": -49.09389877319336,
      "logps/ref_chosen": -37.514625549316406,
      "logps/ref_rejected": -80.34272766113281,
      "logps/rejected": -123.10272216796875,
      "loss": 0.7582,
      "rewards/accuracies": 0.796875,
      "rewards/chosen": -0.7809337973594666,
      "rewards/margins": 2.0747146606445312,
      "rewards/rejected": -2.8556485176086426,
      "step": 66
    },
    {
      "epoch": 0.09838472834067548,
      "epsilon_dpo/beta": 0.0665346160531044,
      "epsilon_dpo/beta_margin_grad_mean": -0.2480592578649521,
      "epsilon_dpo/beta_margin_grad_std": 0.23476538062095642,
      "epsilon_dpo/beta_margin_mean": 1.869264006614685,
      "epsilon_dpo/beta_margin_std": 2.120335340499878,
      "epsilon_dpo/loss_margin_mean": 28.2425479888916,
      "grad_norm": 115.71697998046875,
      "kl/avg_steps": 0.59375,
      "kl/beta": 0.06692533195018768,
      "kl/n_epsilon_steps": 0.203125,
      "kl/p_epsilon_steps": 0.796875,
      "learning_rate": 4.782608695652174e-07,
      "logits/chosen": -3.2035374641418457,
      "logits/rejected": -3.216823101043701,
      "logps/chosen": -51.816200256347656,
      "logps/ref_chosen": -38.82200622558594,
      "logps/ref_rejected": -78.41658782958984,
      "logps/rejected": -119.65333557128906,
      "loss": 0.7213,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.8721305727958679,
      "rewards/margins": 1.8692641258239746,
      "rewards/rejected": -2.741394519805908,
      "step": 67
    },
    {
      "epoch": 0.09985315712187959,
      "epsilon_dpo/beta": 0.06605872511863708,
      "epsilon_dpo/beta_margin_grad_mean": -0.19077911972999573,
      "epsilon_dpo/beta_margin_grad_std": 0.20632153749465942,
      "epsilon_dpo/beta_margin_mean": 2.237042188644409,
      "epsilon_dpo/beta_margin_std": 1.8940068483352661,
      "epsilon_dpo/loss_margin_mean": 33.978355407714844,
      "grad_norm": 89.41032409667969,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.06653030216693878,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 4.855072463768116e-07,
      "logits/chosen": -3.2750229835510254,
      "logits/rejected": -3.2742607593536377,
      "logps/chosen": -53.25446319580078,
      "logps/ref_chosen": -41.910316467285156,
      "logps/ref_rejected": -82.59764862060547,
      "logps/rejected": -127.92015075683594,
      "loss": 0.5268,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.7523297071456909,
      "rewards/margins": 2.237042188644409,
      "rewards/rejected": -2.9893717765808105,
      "step": 68
    },
    {
      "epoch": 0.1013215859030837,
      "epsilon_dpo/beta": 0.06550473719835281,
      "epsilon_dpo/beta_margin_grad_mean": -0.1768295019865036,
      "epsilon_dpo/beta_margin_grad_std": 0.19139200448989868,
      "epsilon_dpo/beta_margin_mean": 2.4585933685302734,
      "epsilon_dpo/beta_margin_std": 2.252986431121826,
      "epsilon_dpo/loss_margin_mean": 37.620914459228516,
      "grad_norm": 85.71179962158203,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.066055528819561,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.927536231884058e-07,
      "logits/chosen": -3.309999704360962,
      "logits/rejected": -3.2920103073120117,
      "logps/chosen": -56.83376693725586,
      "logps/ref_chosen": -42.98963165283203,
      "logps/ref_rejected": -111.28137969970703,
      "logps/rejected": -162.74642944335938,
      "loss": 0.5513,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.910045325756073,
      "rewards/margins": 2.4585933685302734,
      "rewards/rejected": -3.3686389923095703,
      "step": 69
    },
    {
      "epoch": 0.1027900146842878,
      "epsilon_dpo/beta": 0.06497713923454285,
      "epsilon_dpo/beta_margin_grad_mean": -0.1917823851108551,
      "epsilon_dpo/beta_margin_grad_std": 0.18596571683883667,
      "epsilon_dpo/beta_margin_mean": 2.3181331157684326,
      "epsilon_dpo/beta_margin_std": 2.1517505645751953,
      "epsilon_dpo/loss_margin_mean": 35.75630187988281,
      "grad_norm": 88.73102569580078,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.06550285220146179,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 5e-07,
      "logits/chosen": -3.2341670989990234,
      "logits/rejected": -3.3183038234710693,
      "logps/chosen": -55.655487060546875,
      "logps/ref_chosen": -43.68109130859375,
      "logps/ref_rejected": -97.17718505859375,
      "logps/rejected": -144.90789794921875,
      "loss": 0.5016,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.781528651714325,
      "rewards/margins": 2.3181333541870117,
      "rewards/rejected": -3.0996618270874023,
      "step": 70
    },
    {
      "epoch": 0.10425844346549193,
      "epsilon_dpo/beta": 0.0644940659403801,
      "epsilon_dpo/beta_margin_grad_mean": -0.18719229102134705,
      "epsilon_dpo/beta_margin_grad_std": 0.23163102567195892,
      "epsilon_dpo/beta_margin_mean": 2.6485307216644287,
      "epsilon_dpo/beta_margin_std": 2.504869222640991,
      "epsilon_dpo/loss_margin_mean": 41.206172943115234,
      "grad_norm": 100.56282806396484,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.06497492641210556,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 4.999967061337492e-07,
      "logits/chosen": -3.228590488433838,
      "logits/rejected": -3.3432466983795166,
      "logps/chosen": -54.0555419921875,
      "logps/ref_chosen": -40.898582458496094,
      "logps/ref_rejected": -104.50498962402344,
      "logps/rejected": -158.8681182861328,
      "loss": 0.5652,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.8535667657852173,
      "rewards/margins": 2.6485307216644287,
      "rewards/rejected": -3.5020976066589355,
      "step": 71
    },
    {
      "epoch": 0.10572687224669604,
      "epsilon_dpo/beta": 0.06397364288568497,
      "epsilon_dpo/beta_margin_grad_mean": -0.15650521218776703,
      "epsilon_dpo/beta_margin_grad_std": 0.17222045361995697,
      "epsilon_dpo/beta_margin_mean": 2.5931060314178467,
      "epsilon_dpo/beta_margin_std": 2.0323598384857178,
      "epsilon_dpo/loss_margin_mean": 40.60845184326172,
      "grad_norm": 71.43647766113281,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.06449124217033386,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.999868246217933e-07,
      "logits/chosen": -3.2843093872070312,
      "logits/rejected": -3.316100835800171,
      "logps/chosen": -54.79957580566406,
      "logps/ref_chosen": -42.15618896484375,
      "logps/ref_rejected": -102.02656555175781,
      "logps/rejected": -155.27841186523438,
      "loss": 0.4074,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -0.8116366863250732,
      "rewards/margins": 2.593106269836426,
      "rewards/rejected": -3.40474271774292,
      "step": 72
    },
    {
      "epoch": 0.10719530102790015,
      "epsilon_dpo/beta": 0.06355801969766617,
      "epsilon_dpo/beta_margin_grad_mean": -0.2032501995563507,
      "epsilon_dpo/beta_margin_grad_std": 0.21837185323238373,
      "epsilon_dpo/beta_margin_mean": 2.5123815536499023,
      "epsilon_dpo/beta_margin_std": 2.4150218963623047,
      "epsilon_dpo/loss_margin_mean": 39.681556701660156,
      "grad_norm": 75.35091400146484,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.06397147476673126,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 4.999703557245192e-07,
      "logits/chosen": -3.2197837829589844,
      "logits/rejected": -3.2296009063720703,
      "logps/chosen": -55.1505126953125,
      "logps/ref_chosen": -43.86912155151367,
      "logps/ref_rejected": -96.146728515625,
      "logps/rejected": -147.10968017578125,
      "loss": 0.5577,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.7197288870811462,
      "rewards/margins": 2.5123815536499023,
      "rewards/rejected": -3.2321105003356934,
      "step": 73
    },
    {
      "epoch": 0.10866372980910426,
      "epsilon_dpo/beta": 0.06304432451725006,
      "epsilon_dpo/beta_margin_grad_mean": -0.15927036106586456,
      "epsilon_dpo/beta_margin_grad_std": 0.18017472326755524,
      "epsilon_dpo/beta_margin_mean": 2.8419299125671387,
      "epsilon_dpo/beta_margin_std": 2.680656909942627,
      "epsilon_dpo/loss_margin_mean": 45.170841217041016,
      "grad_norm": 65.50677490234375,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.06355439871549606,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.999472998758977e-07,
      "logits/chosen": -3.147824764251709,
      "logits/rejected": -3.291006088256836,
      "logps/chosen": -38.77723693847656,
      "logps/ref_chosen": -29.008399963378906,
      "logps/ref_rejected": -102.72833251953125,
      "logps/rejected": -157.6680145263672,
      "loss": 0.4061,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.6169643998146057,
      "rewards/margins": 2.8419299125671387,
      "rewards/rejected": -3.4588942527770996,
      "step": 74
    },
    {
      "epoch": 0.11013215859030837,
      "epsilon_dpo/beta": 0.062496818602085114,
      "epsilon_dpo/beta_margin_grad_mean": -0.15049783885478973,
      "epsilon_dpo/beta_margin_grad_std": 0.1585647463798523,
      "epsilon_dpo/beta_margin_mean": 2.8132150173187256,
      "epsilon_dpo/beta_margin_std": 2.3551828861236572,
      "epsilon_dpo/loss_margin_mean": 45.07414245605469,
      "grad_norm": 62.55946731567383,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.06304218620061874,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.999176576834721e-07,
      "logits/chosen": -3.151859760284424,
      "logits/rejected": -3.281860828399658,
      "logps/chosen": -45.443458557128906,
      "logps/ref_chosen": -30.710708618164062,
      "logps/ref_rejected": -117.44107818603516,
      "logps/rejected": -177.2479705810547,
      "loss": 0.3702,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.9213298559188843,
      "rewards/margins": 2.8132152557373047,
      "rewards/rejected": -3.7345452308654785,
      "step": 75
    },
    {
      "epoch": 0.11160058737151249,
      "epsilon_dpo/beta": 0.062013305723667145,
      "epsilon_dpo/beta_margin_grad_mean": -0.19430118799209595,
      "epsilon_dpo/beta_margin_grad_std": 0.2106785923242569,
      "epsilon_dpo/beta_margin_mean": 2.20247745513916,
      "epsilon_dpo/beta_margin_std": 1.9251240491867065,
      "epsilon_dpo/loss_margin_mean": 35.627044677734375,
      "grad_norm": 96.22077941894531,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.06249535083770752,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 4.998814299283415e-07,
      "logits/chosen": -3.207365036010742,
      "logits/rejected": -3.260375499725342,
      "logps/chosen": -48.6707763671875,
      "logps/ref_chosen": -35.03684997558594,
      "logps/ref_rejected": -84.84458923339844,
      "logps/rejected": -134.10556030273438,
      "loss": 0.5828,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.8483954668045044,
      "rewards/margins": 2.20247745513916,
      "rewards/rejected": -3.050872802734375,
      "step": 76
    },
    {
      "epoch": 0.1130690161527166,
      "epsilon_dpo/beta": 0.06151320040225983,
      "epsilon_dpo/beta_margin_grad_mean": -0.16589321196079254,
      "epsilon_dpo/beta_margin_grad_std": 0.19776158034801483,
      "epsilon_dpo/beta_margin_mean": 2.548051595687866,
      "epsilon_dpo/beta_margin_std": 2.222637414932251,
      "epsilon_dpo/loss_margin_mean": 41.525611877441406,
      "grad_norm": 109.90064239501953,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.06201088801026344,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.998386175651409e-07,
      "logits/chosen": -3.159731864929199,
      "logits/rejected": -3.2670416831970215,
      "logps/chosen": -57.969879150390625,
      "logps/ref_chosen": -44.09752655029297,
      "logps/ref_rejected": -98.75190734863281,
      "logps/rejected": -154.14987182617188,
      "loss": 0.5577,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.8574961423873901,
      "rewards/margins": 2.548051357269287,
      "rewards/rejected": -3.405547618865967,
      "step": 77
    },
    {
      "epoch": 0.1145374449339207,
      "epsilon_dpo/beta": 0.06109432876110077,
      "epsilon_dpo/beta_margin_grad_mean": -0.19816353917121887,
      "epsilon_dpo/beta_margin_grad_std": 0.19675467908382416,
      "epsilon_dpo/beta_margin_mean": 2.1238291263580322,
      "epsilon_dpo/beta_margin_std": 1.9207454919815063,
      "epsilon_dpo/loss_margin_mean": 34.886810302734375,
      "grad_norm": 88.5641860961914,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.06151111051440239,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 4.997892217220159e-07,
      "logits/chosen": -3.2637405395507812,
      "logits/rejected": -3.250288963317871,
      "logps/chosen": -52.061004638671875,
      "logps/ref_chosen": -38.710906982421875,
      "logps/ref_rejected": -91.00759887695312,
      "logps/rejected": -139.2445068359375,
      "loss": 0.5417,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.8202856779098511,
      "rewards/margins": 2.123828887939453,
      "rewards/rejected": -2.9441146850585938,
      "step": 78
    },
    {
      "epoch": 0.11600587371512482,
      "epsilon_dpo/beta": 0.060581713914871216,
      "epsilon_dpo/beta_margin_grad_mean": -0.1641714721918106,
      "epsilon_dpo/beta_margin_grad_std": 0.17812295258045197,
      "epsilon_dpo/beta_margin_mean": 2.5997278690338135,
      "epsilon_dpo/beta_margin_std": 2.096586227416992,
      "epsilon_dpo/loss_margin_mean": 42.99335479736328,
      "grad_norm": 86.2493667602539,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.0610911101102829,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.997332437005931e-07,
      "logits/chosen": -3.1998233795166016,
      "logits/rejected": -3.2696633338928223,
      "logps/chosen": -42.79743957519531,
      "logps/ref_chosen": -32.905845642089844,
      "logps/ref_rejected": -95.70394897460938,
      "logps/rejected": -148.58889770507812,
      "loss": 0.4274,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.6022886633872986,
      "rewards/margins": 2.5997278690338135,
      "rewards/rejected": -3.202016592025757,
      "step": 79
    },
    {
      "epoch": 0.11747430249632893,
      "epsilon_dpo/beta": 0.06015056371688843,
      "epsilon_dpo/beta_margin_grad_mean": -0.20884603261947632,
      "epsilon_dpo/beta_margin_grad_std": 0.22885563969612122,
      "epsilon_dpo/beta_margin_mean": 2.3807008266448975,
      "epsilon_dpo/beta_margin_std": 2.437171220779419,
      "epsilon_dpo/loss_margin_mean": 39.722938537597656,
      "grad_norm": 94.2277603149414,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.06057996675372124,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 4.996706849759452e-07,
      "logits/chosen": -3.2733001708984375,
      "logits/rejected": -3.3127360343933105,
      "logps/chosen": -55.867835998535156,
      "logps/ref_chosen": -42.08654022216797,
      "logps/ref_rejected": -93.93815612792969,
      "logps/rejected": -147.4423828125,
      "loss": 0.6202,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.8336386680603027,
      "rewards/margins": 2.3807008266448975,
      "rewards/rejected": -3.2143394947052,
      "step": 80
    },
    {
      "epoch": 0.11894273127753303,
      "epsilon_dpo/beta": 0.05966491997241974,
      "epsilon_dpo/beta_margin_grad_mean": -0.17548146843910217,
      "epsilon_dpo/beta_margin_grad_std": 0.19263611733913422,
      "epsilon_dpo/beta_margin_mean": 2.601060628890991,
      "epsilon_dpo/beta_margin_std": 2.4161243438720703,
      "epsilon_dpo/loss_margin_mean": 43.69050598144531,
      "grad_norm": 70.84126281738281,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.06014765426516533,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.996015471965529e-07,
      "logits/chosen": -3.1918153762817383,
      "logits/rejected": -3.2612764835357666,
      "logps/chosen": -52.02825927734375,
      "logps/ref_chosen": -39.808433532714844,
      "logps/ref_rejected": -132.9473876953125,
      "logps/rejected": -188.8577117919922,
      "loss": 0.4684,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.7309558391571045,
      "rewards/margins": 2.601060628890991,
      "rewards/rejected": -3.3320164680480957,
      "step": 81
    },
    {
      "epoch": 0.12041116005873716,
      "epsilon_dpo/beta": 0.05922134220600128,
      "epsilon_dpo/beta_margin_grad_mean": -0.18756245076656342,
      "epsilon_dpo/beta_margin_grad_std": 0.23345990478992462,
      "epsilon_dpo/beta_margin_mean": 2.376243829727173,
      "epsilon_dpo/beta_margin_std": 2.1569864749908447,
      "epsilon_dpo/loss_margin_mean": 40.271209716796875,
      "grad_norm": 98.73262023925781,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.059662893414497375,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 4.995258321842611e-07,
      "logits/chosen": -3.1882076263427734,
      "logits/rejected": -3.236912488937378,
      "logps/chosen": -46.98686218261719,
      "logps/ref_chosen": -33.495845794677734,
      "logps/ref_rejected": -96.71635437011719,
      "logps/rejected": -150.47857666015625,
      "loss": 0.6066,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.8024958372116089,
      "rewards/margins": 2.376243829727173,
      "rewards/rejected": -3.178739547729492,
      "step": 82
    },
    {
      "epoch": 0.12187958883994127,
      "epsilon_dpo/beta": 0.05872496962547302,
      "epsilon_dpo/beta_margin_grad_mean": -0.16398201882839203,
      "epsilon_dpo/beta_margin_grad_std": 0.18266217410564423,
      "epsilon_dpo/beta_margin_mean": 2.464262008666992,
      "epsilon_dpo/beta_margin_std": 1.7969151735305786,
      "epsilon_dpo/loss_margin_mean": 42.03822326660156,
      "grad_norm": 67.05672454833984,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.05921875312924385,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.994435419342304e-07,
      "logits/chosen": -3.163329839706421,
      "logits/rejected": -3.297056198120117,
      "logps/chosen": -36.729583740234375,
      "logps/ref_chosen": -25.916236877441406,
      "logps/ref_rejected": -108.29981994628906,
      "logps/rejected": -161.15139770507812,
      "loss": 0.424,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.6363058090209961,
      "rewards/margins": 2.464262008666992,
      "rewards/rejected": -3.1005678176879883,
      "step": 83
    },
    {
      "epoch": 0.12334801762114538,
      "epsilon_dpo/beta": 0.058215267956256866,
      "epsilon_dpo/beta_margin_grad_mean": -0.1662423461675644,
      "epsilon_dpo/beta_margin_grad_std": 0.16180217266082764,
      "epsilon_dpo/beta_margin_mean": 2.257798910140991,
      "epsilon_dpo/beta_margin_std": 1.5832382440567017,
      "epsilon_dpo/loss_margin_mean": 38.83984375,
      "grad_norm": 68.96977996826172,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.05872327461838722,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.993546786148857e-07,
      "logits/chosen": -3.2171850204467773,
      "logits/rejected": -3.2014272212982178,
      "logps/chosen": -51.26725769042969,
      "logps/ref_chosen": -36.62953567504883,
      "logps/ref_rejected": -95.27814483642578,
      "logps/rejected": -148.75570678710938,
      "loss": 0.4147,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.8531267642974854,
      "rewards/margins": 2.2577991485595703,
      "rewards/rejected": -3.1109256744384766,
      "step": 84
    },
    {
      "epoch": 0.12481644640234948,
      "epsilon_dpo/beta": 0.05781027674674988,
      "epsilon_dpo/beta_margin_grad_mean": -0.20659653842449188,
      "epsilon_dpo/beta_margin_grad_std": 0.2083936482667923,
      "epsilon_dpo/beta_margin_mean": 2.0086328983306885,
      "epsilon_dpo/beta_margin_std": 1.7025460004806519,
      "epsilon_dpo/loss_margin_mean": 34.859657287597656,
      "grad_norm": 81.10045623779297,
      "kl/avg_steps": 0.703125,
      "kl/beta": 0.05821390450000763,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 4.992592445678582e-07,
      "logits/chosen": -3.2968966960906982,
      "logits/rejected": -3.2394418716430664,
      "logps/chosen": -59.8691291809082,
      "logps/ref_chosen": -43.555397033691406,
      "logps/ref_rejected": -83.9044418334961,
      "logps/rejected": -135.07781982421875,
      "loss": 0.564,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.9453153610229492,
      "rewards/margins": 2.0086328983306885,
      "rewards/rejected": -2.953948497772217,
      "step": 85
    },
    {
      "epoch": 0.1262848751835536,
      "epsilon_dpo/beta": 0.057433828711509705,
      "epsilon_dpo/beta_margin_grad_mean": -0.2296195775270462,
      "epsilon_dpo/beta_margin_grad_std": 0.22347640991210938,
      "epsilon_dpo/beta_margin_mean": 2.128232479095459,
      "epsilon_dpo/beta_margin_std": 2.417414903640747,
      "epsilon_dpo/loss_margin_mean": 37.22154998779297,
      "grad_norm": 107.6495132446289,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.05780744552612305,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 4.991572423079235e-07,
      "logits/chosen": -3.272704601287842,
      "logits/rejected": -3.2372689247131348,
      "logps/chosen": -54.787696838378906,
      "logps/ref_chosen": -38.846839904785156,
      "logps/ref_rejected": -92.59671020507812,
      "logps/rejected": -145.7591094970703,
      "loss": 0.6967,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.9202961921691895,
      "rewards/margins": 2.128232479095459,
      "rewards/rejected": -3.0485289096832275,
      "step": 86
    },
    {
      "epoch": 0.1277533039647577,
      "epsilon_dpo/beta": 0.05698757991194725,
      "epsilon_dpo/beta_margin_grad_mean": -0.18504740297794342,
      "epsilon_dpo/beta_margin_grad_std": 0.18343256413936615,
      "epsilon_dpo/beta_margin_mean": 2.3569390773773193,
      "epsilon_dpo/beta_margin_std": 1.9608224630355835,
      "epsilon_dpo/loss_margin_mean": 41.45563507080078,
      "grad_norm": 59.761165618896484,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.057430557906627655,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 4.990486745229364e-07,
      "logits/chosen": -3.230192184448242,
      "logits/rejected": -3.3376736640930176,
      "logps/chosen": -52.74066162109375,
      "logps/ref_chosen": -38.653236389160156,
      "logps/ref_rejected": -102.44976043701172,
      "logps/rejected": -157.99281311035156,
      "loss": 0.4717,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.8040882349014282,
      "rewards/margins": 2.3569388389587402,
      "rewards/rejected": -3.161027193069458,
      "step": 87
    },
    {
      "epoch": 0.12922173274596183,
      "epsilon_dpo/beta": 0.05656362697482109,
      "epsilon_dpo/beta_margin_grad_mean": -0.1606673151254654,
      "epsilon_dpo/beta_margin_grad_std": 0.18233220279216766,
      "epsilon_dpo/beta_margin_mean": 2.5598440170288086,
      "epsilon_dpo/beta_margin_std": 1.910704493522644,
      "epsilon_dpo/loss_margin_mean": 45.38097381591797,
      "grad_norm": 58.7109375,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.05698535963892937,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 4.989335440737586e-07,
      "logits/chosen": -3.27421236038208,
      "logits/rejected": -3.3343310356140137,
      "logps/chosen": -51.133506774902344,
      "logps/ref_chosen": -37.23695373535156,
      "logps/ref_rejected": -116.12947082519531,
      "logps/rejected": -175.40699768066406,
      "loss": 0.4141,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.7879468202590942,
      "rewards/margins": 2.5598437786102295,
      "rewards/rejected": -3.3477907180786133,
      "step": 88
    },
    {
      "epoch": 0.13069016152716592,
      "epsilon_dpo/beta": 0.05614255368709564,
      "epsilon_dpo/beta_margin_grad_mean": -0.18962043523788452,
      "epsilon_dpo/beta_margin_grad_std": 0.18511323630809784,
      "epsilon_dpo/beta_margin_mean": 2.2266876697540283,
      "epsilon_dpo/beta_margin_std": 1.9912763833999634,
      "epsilon_dpo/loss_margin_mean": 39.772979736328125,
      "grad_norm": 69.53501892089844,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.056561149656772614,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 4.988118539941847e-07,
      "logits/chosen": -3.2087016105651855,
      "logits/rejected": -3.23443603515625,
      "logps/chosen": -52.141876220703125,
      "logps/ref_chosen": -39.35747146606445,
      "logps/ref_rejected": -88.01043701171875,
      "logps/rejected": -140.5678253173828,
      "loss": 0.4951,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.7197608947753906,
      "rewards/margins": 2.2266876697540283,
      "rewards/rejected": -2.946448802947998,
      "step": 89
    },
    {
      "epoch": 0.13215859030837004,
      "epsilon_dpo/beta": 0.055777255445718765,
      "epsilon_dpo/beta_margin_grad_mean": -0.19303679466247559,
      "epsilon_dpo/beta_margin_grad_std": 0.21628044545650482,
      "epsilon_dpo/beta_margin_mean": 2.5097200870513916,
      "epsilon_dpo/beta_margin_std": 2.4309592247009277,
      "epsilon_dpo/loss_margin_mean": 45.174312591552734,
      "grad_norm": 91.28634643554688,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.056140098720788956,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 4.986836074908615e-07,
      "logits/chosen": -3.129134178161621,
      "logits/rejected": -3.3651089668273926,
      "logps/chosen": -46.801185607910156,
      "logps/ref_chosen": -30.30811882019043,
      "logps/ref_rejected": -119.35741424560547,
      "logps/rejected": -181.02479553222656,
      "loss": 0.5447,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.9231457710266113,
      "rewards/margins": 2.5097203254699707,
      "rewards/rejected": -3.432866096496582,
      "step": 90
    },
    {
      "epoch": 0.13362701908957417,
      "epsilon_dpo/beta": 0.05532645061612129,
      "epsilon_dpo/beta_margin_grad_mean": -0.16726556420326233,
      "epsilon_dpo/beta_margin_grad_std": 0.17135490477085114,
      "epsilon_dpo/beta_margin_mean": 2.357635736465454,
      "epsilon_dpo/beta_margin_std": 1.7421871423721313,
      "epsilon_dpo/loss_margin_mean": 42.69865036010742,
      "grad_norm": 68.51940155029297,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.055774081498384476,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.985488079432037e-07,
      "logits/chosen": -3.2310478687286377,
      "logits/rejected": -3.2350800037384033,
      "logps/chosen": -50.61489486694336,
      "logps/ref_chosen": -35.484596252441406,
      "logps/ref_rejected": -94.16378784179688,
      "logps/rejected": -151.99273681640625,
      "loss": 0.4213,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.8394383192062378,
      "rewards/margins": 2.357635736465454,
      "rewards/rejected": -3.1970739364624023,
      "step": 91
    },
    {
      "epoch": 0.13509544787077826,
      "epsilon_dpo/beta": 0.05489783734083176,
      "epsilon_dpo/beta_margin_grad_mean": -0.2024046629667282,
      "epsilon_dpo/beta_margin_grad_std": 0.220162495970726,
      "epsilon_dpo/beta_margin_mean": 2.2357397079467773,
      "epsilon_dpo/beta_margin_std": 2.0685596466064453,
      "epsilon_dpo/loss_margin_mean": 40.8467903137207,
      "grad_norm": 87.93189239501953,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.05532456934452057,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 4.984074589033043e-07,
      "logits/chosen": -3.121596574783325,
      "logits/rejected": -3.1575112342834473,
      "logps/chosen": -52.659706115722656,
      "logps/ref_chosen": -37.970062255859375,
      "logps/ref_rejected": -84.28839111328125,
      "logps/rejected": -139.8248291015625,
      "loss": 0.5756,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.808807373046875,
      "rewards/margins": 2.2357397079467773,
      "rewards/rejected": -3.0445470809936523,
      "step": 92
    },
    {
      "epoch": 0.13656387665198239,
      "epsilon_dpo/beta": 0.05447227507829666,
      "epsilon_dpo/beta_margin_grad_mean": -0.18667982518672943,
      "epsilon_dpo/beta_margin_grad_std": 0.18083734810352325,
      "epsilon_dpo/beta_margin_mean": 2.251617908477783,
      "epsilon_dpo/beta_margin_std": 1.9179073572158813,
      "epsilon_dpo/loss_margin_mean": 41.435726165771484,
      "grad_norm": 66.98503875732422,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.05489569902420044,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 4.982595640958425e-07,
      "logits/chosen": -3.2347559928894043,
      "logits/rejected": -3.2382359504699707,
      "logps/chosen": -51.651126861572266,
      "logps/ref_chosen": -35.3890266418457,
      "logps/ref_rejected": -81.84159851074219,
      "logps/rejected": -139.5394287109375,
      "loss": 0.4779,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.88683021068573,
      "rewards/margins": 2.251617908477783,
      "rewards/rejected": -3.1384482383728027,
      "step": 93
    },
    {
      "epoch": 0.13803230543318648,
      "epsilon_dpo/beta": 0.05399893596768379,
      "epsilon_dpo/beta_margin_grad_mean": -0.15303176641464233,
      "epsilon_dpo/beta_margin_grad_std": 0.14751987159252167,
      "epsilon_dpo/beta_margin_mean": 2.574958562850952,
      "epsilon_dpo/beta_margin_std": 1.9459197521209717,
      "epsilon_dpo/loss_margin_mean": 47.74397659301758,
      "grad_norm": 63.9381103515625,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.05447014793753624,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.98105127417984e-07,
      "logits/chosen": -3.194207191467285,
      "logits/rejected": -3.2492258548736572,
      "logps/chosen": -53.30908203125,
      "logps/ref_chosen": -38.974853515625,
      "logps/ref_rejected": -106.16789245605469,
      "logps/rejected": -168.24609375,
      "loss": 0.3679,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -0.7756279706954956,
      "rewards/margins": 2.574958324432373,
      "rewards/rejected": -3.350586414337158,
      "step": 94
    },
    {
      "epoch": 0.1395007342143906,
      "epsilon_dpo/beta": 0.05351366475224495,
      "epsilon_dpo/beta_margin_grad_mean": -0.176425963640213,
      "epsilon_dpo/beta_margin_grad_std": 0.15948951244354248,
      "epsilon_dpo/beta_margin_mean": 2.165757417678833,
      "epsilon_dpo/beta_margin_std": 1.6699368953704834,
      "epsilon_dpo/loss_margin_mean": 40.51607131958008,
      "grad_norm": 63.63764572143555,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.053997669368982315,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 4.979441529392784e-07,
      "logits/chosen": -3.180877208709717,
      "logits/rejected": -3.2767512798309326,
      "logps/chosen": -41.924564361572266,
      "logps/ref_chosen": -29.644317626953125,
      "logps/ref_rejected": -80.65695190429688,
      "logps/rejected": -133.45326232910156,
      "loss": 0.4364,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.6584010124206543,
      "rewards/margins": 2.165757417678833,
      "rewards/rejected": -2.824158191680908,
      "step": 95
    },
    {
      "epoch": 0.14096916299559473,
      "epsilon_dpo/beta": 0.05311667546629906,
      "epsilon_dpo/beta_margin_grad_mean": -0.18641680479049683,
      "epsilon_dpo/beta_margin_grad_std": 0.20043620467185974,
      "epsilon_dpo/beta_margin_mean": 2.554248094558716,
      "epsilon_dpo/beta_margin_std": 2.2736546993255615,
      "epsilon_dpo/loss_margin_mean": 48.22751998901367,
      "grad_norm": 85.53929901123047,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.05351271107792854,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 4.977766449015534e-07,
      "logits/chosen": -3.2188198566436768,
      "logits/rejected": -3.264829635620117,
      "logps/chosen": -65.12588500976562,
      "logps/ref_chosen": -48.42084884643555,
      "logps/ref_rejected": -102.56741333007812,
      "logps/rejected": -167.49996948242188,
      "loss": 0.4965,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.8906632661819458,
      "rewards/margins": 2.554248094558716,
      "rewards/rejected": -3.444911479949951,
      "step": 96
    },
    {
      "epoch": 0.14243759177679882,
      "epsilon_dpo/beta": 0.05267146974802017,
      "epsilon_dpo/beta_margin_grad_mean": -0.15445828437805176,
      "epsilon_dpo/beta_margin_grad_std": 0.16520872712135315,
      "epsilon_dpo/beta_margin_mean": 2.608966827392578,
      "epsilon_dpo/beta_margin_std": 1.9469996690750122,
      "epsilon_dpo/loss_margin_mean": 49.61277770996094,
      "grad_norm": 52.84062194824219,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.053114354610443115,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.976026077188012e-07,
      "logits/chosen": -3.178095817565918,
      "logits/rejected": -3.169985294342041,
      "logps/chosen": -47.99535369873047,
      "logps/ref_chosen": -35.1164436340332,
      "logps/ref_rejected": -83.36341857910156,
      "logps/rejected": -145.8551025390625,
      "loss": 0.3851,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -0.6797794103622437,
      "rewards/margins": 2.6089670658111572,
      "rewards/rejected": -3.2887463569641113,
      "step": 97
    },
    {
      "epoch": 0.14390602055800295,
      "epsilon_dpo/beta": 0.05228015407919884,
      "epsilon_dpo/beta_margin_grad_mean": -0.19064751267433167,
      "epsilon_dpo/beta_margin_grad_std": 0.22201648354530334,
      "epsilon_dpo/beta_margin_mean": 2.2540199756622314,
      "epsilon_dpo/beta_margin_std": 1.9207807779312134,
      "epsilon_dpo/loss_margin_mean": 43.26023864746094,
      "grad_norm": 101.56369018554688,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0526699498295784,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 4.974220459770639e-07,
      "logits/chosen": -3.108753204345703,
      "logits/rejected": -3.2454357147216797,
      "logps/chosen": -65.423828125,
      "logps/ref_chosen": -44.868499755859375,
      "logps/ref_rejected": -101.7425537109375,
      "logps/rejected": -165.55812072753906,
      "loss": 0.5783,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -1.079300880432129,
      "rewards/margins": 2.2540199756622314,
      "rewards/rejected": -3.3333208560943604,
      "step": 98
    },
    {
      "epoch": 0.14537444933920704,
      "epsilon_dpo/beta": 0.051825616508722305,
      "epsilon_dpo/beta_margin_grad_mean": -0.1456242799758911,
      "epsilon_dpo/beta_margin_grad_std": 0.16527007520198822,
      "epsilon_dpo/beta_margin_mean": 2.7669012546539307,
      "epsilon_dpo/beta_margin_std": 2.086336135864258,
      "epsilon_dpo/loss_margin_mean": 53.46054458618164,
      "grad_norm": 56.95939254760742,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.05227786675095558,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.972349644343108e-07,
      "logits/chosen": -3.0614562034606934,
      "logits/rejected": -3.1564629077911377,
      "logps/chosen": -40.343666076660156,
      "logps/ref_chosen": -25.89197540283203,
      "logps/ref_rejected": -91.06307983398438,
      "logps/rejected": -158.97531127929688,
      "loss": 0.3636,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -0.7500526905059814,
      "rewards/margins": 2.7669014930725098,
      "rewards/rejected": -3.516953945159912,
      "step": 99
    },
    {
      "epoch": 0.14684287812041116,
      "epsilon_dpo/beta": 0.05150564759969711,
      "epsilon_dpo/beta_margin_grad_mean": -0.23118621110916138,
      "epsilon_dpo/beta_margin_grad_std": 0.23917478322982788,
      "epsilon_dpo/beta_margin_mean": 1.9697211980819702,
      "epsilon_dpo/beta_margin_std": 1.9218071699142456,
      "epsilon_dpo/loss_margin_mean": 38.432437896728516,
      "grad_norm": 109.34486389160156,
      "kl/avg_steps": 0.625,
      "kl/beta": 0.05182440206408501,
      "kl/n_epsilon_steps": 0.1875,
      "kl/p_epsilon_steps": 0.8125,
      "learning_rate": 4.970413680203148e-07,
      "logits/chosen": -3.0767757892608643,
      "logits/rejected": -3.0463757514953613,
      "logps/chosen": -60.350921630859375,
      "logps/ref_chosen": -41.60627746582031,
      "logps/ref_rejected": -79.52035522460938,
      "logps/rejected": -136.6974334716797,
      "loss": 0.6663,
      "rewards/accuracies": 0.828125,
      "rewards/chosen": -0.9696159362792969,
      "rewards/margins": 1.9697211980819702,
      "rewards/rejected": -2.9393372535705566,
      "step": 100
    },
    {
      "epoch": 0.14684287812041116,
      "eval_epsilon_dpo/beta": 0.051277898252010345,
      "eval_epsilon_dpo/beta_margin_grad_mean": -0.30362221598625183,
      "eval_epsilon_dpo/beta_margin_grad_std": 0.2585260272026062,
      "eval_epsilon_dpo/beta_margin_mean": 1.5045814514160156,
      "eval_epsilon_dpo/beta_margin_std": 2.153695583343506,
      "eval_epsilon_dpo/loss_margin_mean": 29.590499877929688,
      "eval_kl/n_epsilon_steps": 0.2765410840511322,
      "eval_kl/p_epsilon_steps": 0.7226027250289917,
      "eval_logits/chosen": -3.276595115661621,
      "eval_logits/rejected": -3.2684147357940674,
      "eval_logps/chosen": -96.53782653808594,
      "eval_logps/ref_chosen": -68.29110717773438,
      "eval_logps/ref_rejected": -92.08038330078125,
      "eval_logps/rejected": -149.91758728027344,
      "eval_loss": 0.49363189935684204,
      "eval_rewards/accuracies": 0.7555650472640991,
      "eval_rewards/chosen": -1.4540938138961792,
      "eval_rewards/margins": 1.5045814514160156,
      "eval_rewards/rejected": -2.9586753845214844,
      "eval_runtime": 38.4605,
      "eval_samples_per_second": 60.816,
      "eval_steps_per_second": 1.924,
      "step": 100
    },
    {
      "epoch": 0.14831130690161526,
      "epsilon_dpo/beta": 0.051185738295316696,
      "epsilon_dpo/beta_margin_grad_mean": -0.23383933305740356,
      "epsilon_dpo/beta_margin_grad_std": 0.23400039970874786,
      "epsilon_dpo/beta_margin_mean": 1.9769922494888306,
      "epsilon_dpo/beta_margin_std": 2.0824813842773438,
      "epsilon_dpo/loss_margin_mean": 38.81412124633789,
      "grad_norm": 104.73916625976562,
      "kl/avg_steps": 0.625,
      "kl/beta": 0.05150251090526581,
      "kl/n_epsilon_steps": 0.1875,
      "kl/p_epsilon_steps": 0.8125,
      "learning_rate": 4.968412618365215e-07,
      "logits/chosen": -3.2080070972442627,
      "logits/rejected": -3.206873655319214,
      "logps/chosen": -67.021484375,
      "logps/ref_chosen": -46.34172439575195,
      "logps/ref_rejected": -89.14061737060547,
      "logps/rejected": -148.63449096679688,
      "loss": 0.6759,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": -1.0630580186843872,
      "rewards/margins": 1.976992130279541,
      "rewards/rejected": -3.0400502681732178,
      "step": 101
    },
    {
      "epoch": 0.14977973568281938,
      "epsilon_dpo/beta": 0.05080382898449898,
      "epsilon_dpo/beta_margin_grad_mean": -0.20627860724925995,
      "epsilon_dpo/beta_margin_grad_std": 0.19808727502822876,
      "epsilon_dpo/beta_margin_mean": 2.0157113075256348,
      "epsilon_dpo/beta_margin_std": 1.7352498769760132,
      "epsilon_dpo/loss_margin_mean": 39.79585647583008,
      "grad_norm": 70.99539184570312,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.051182620227336884,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 4.966346511559149e-07,
      "logits/chosen": -3.2234458923339844,
      "logits/rejected": -3.1278076171875,
      "logps/chosen": -59.05921936035156,
      "logps/ref_chosen": -40.793182373046875,
      "logps/ref_rejected": -74.26336669921875,
      "logps/rejected": -132.32525634765625,
      "loss": 0.5495,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.9300694465637207,
      "rewards/margins": 2.0157113075256348,
      "rewards/rejected": -2.9457807540893555,
      "step": 102
    },
    {
      "epoch": 0.1512481644640235,
      "epsilon_dpo/beta": 0.05037800967693329,
      "epsilon_dpo/beta_margin_grad_mean": -0.14925150573253632,
      "epsilon_dpo/beta_margin_grad_std": 0.1886918544769287,
      "epsilon_dpo/beta_margin_mean": 2.8233063220977783,
      "epsilon_dpo/beta_margin_std": 2.1322638988494873,
      "epsilon_dpo/loss_margin_mean": 56.148807525634766,
      "grad_norm": 70.43810272216797,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.050801608711481094,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.964215414228785e-07,
      "logits/chosen": -3.152428388595581,
      "logits/rejected": -3.057560920715332,
      "logps/chosen": -53.713714599609375,
      "logps/ref_chosen": -37.85403823852539,
      "logps/ref_rejected": -101.43891906738281,
      "logps/rejected": -173.4473876953125,
      "loss": 0.3972,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.8012863397598267,
      "rewards/margins": 2.823306083679199,
      "rewards/rejected": -3.6245927810668945,
      "step": 103
    },
    {
      "epoch": 0.1527165932452276,
      "epsilon_dpo/beta": 0.04997224360704422,
      "epsilon_dpo/beta_margin_grad_mean": -0.16119545698165894,
      "epsilon_dpo/beta_margin_grad_std": 0.1801510900259018,
      "epsilon_dpo/beta_margin_mean": 2.779792547225952,
      "epsilon_dpo/beta_margin_std": 2.224303960800171,
      "epsilon_dpo/loss_margin_mean": 55.73902893066406,
      "grad_norm": 61.21420669555664,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.05037655681371689,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.96201938253052e-07,
      "logits/chosen": -3.150313377380371,
      "logits/rejected": -3.136101245880127,
      "logps/chosen": -51.264808654785156,
      "logps/ref_chosen": -35.89470672607422,
      "logps/ref_rejected": -97.03848266601562,
      "logps/rejected": -168.14761352539062,
      "loss": 0.4096,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.769996166229248,
      "rewards/margins": 2.779792308807373,
      "rewards/rejected": -3.549788475036621,
      "step": 104
    },
    {
      "epoch": 0.15418502202643172,
      "epsilon_dpo/beta": 0.04952263832092285,
      "epsilon_dpo/beta_margin_grad_mean": -0.12934987246990204,
      "epsilon_dpo/beta_margin_grad_std": 0.15336252748966217,
      "epsilon_dpo/beta_margin_mean": 2.9811060428619385,
      "epsilon_dpo/beta_margin_std": 2.177426338195801,
      "epsilon_dpo/loss_margin_mean": 60.252281188964844,
      "grad_norm": 61.274295806884766,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.04997054487466812,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 4.959758474331832e-07,
      "logits/chosen": -3.1032373905181885,
      "logits/rejected": -3.1879420280456543,
      "logps/chosen": -56.80089569091797,
      "logps/ref_chosen": -39.414154052734375,
      "logps/ref_rejected": -105.14703369140625,
      "logps/rejected": -182.7860565185547,
      "loss": 0.3212,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.8625220060348511,
      "rewards/margins": 2.9811058044433594,
      "rewards/rejected": -3.8436279296875,
      "step": 105
    },
    {
      "epoch": 0.15565345080763582,
      "epsilon_dpo/beta": 0.04914744198322296,
      "epsilon_dpo/beta_margin_grad_mean": -0.18032847344875336,
      "epsilon_dpo/beta_margin_grad_std": 0.19534119963645935,
      "epsilon_dpo/beta_margin_mean": 2.3009119033813477,
      "epsilon_dpo/beta_margin_std": 1.8214956521987915,
      "epsilon_dpo/loss_margin_mean": 46.947818756103516,
      "grad_norm": 60.775062561035156,
      "kl/avg_steps": 0.765625,
      "kl/beta": 0.049521755427122116,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 4.957432749209755e-07,
      "logits/chosen": -3.1399245262145996,
      "logits/rejected": -3.1790456771850586,
      "logps/chosen": -58.34198760986328,
      "logps/ref_chosen": -39.36569595336914,
      "logps/ref_rejected": -90.00059509277344,
      "logps/rejected": -155.92471313476562,
      "loss": 0.4771,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.9345280528068542,
      "rewards/margins": 2.3009119033813477,
      "rewards/rejected": -3.2354397773742676,
      "step": 106
    },
    {
      "epoch": 0.15712187958883994,
      "epsilon_dpo/beta": 0.04868961498141289,
      "epsilon_dpo/beta_margin_grad_mean": -0.14175310730934143,
      "epsilon_dpo/beta_margin_grad_std": 0.17217276990413666,
      "epsilon_dpo/beta_margin_mean": 2.8992745876312256,
      "epsilon_dpo/beta_margin_std": 2.3069019317626953,
      "epsilon_dpo/loss_margin_mean": 59.605445861816406,
      "grad_norm": 64.9974365234375,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.04914548620581627,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 4.955042268449307e-07,
      "logits/chosen": -3.1247940063476562,
      "logits/rejected": -3.1252992153167725,
      "logps/chosen": -56.38127899169922,
      "logps/ref_chosen": -40.38301086425781,
      "logps/ref_rejected": -100.88917541503906,
      "logps/rejected": -176.49288940429688,
      "loss": 0.383,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -0.7801457643508911,
      "rewards/margins": 2.8992745876312256,
      "rewards/rejected": -3.6794204711914062,
      "step": 107
    },
    {
      "epoch": 0.15859030837004406,
      "epsilon_dpo/beta": 0.048267822712659836,
      "epsilon_dpo/beta_margin_grad_mean": -0.13621723651885986,
      "epsilon_dpo/beta_margin_grad_std": 0.16647490859031677,
      "epsilon_dpo/beta_margin_mean": 3.0122811794281006,
      "epsilon_dpo/beta_margin_std": 2.167310953140259,
      "epsilon_dpo/loss_margin_mean": 62.49602508544922,
      "grad_norm": 53.83271026611328,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.04868902638554573,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.952587095041881e-07,
      "logits/chosen": -3.1003808975219727,
      "logits/rejected": -3.143059492111206,
      "logps/chosen": -49.35261917114258,
      "logps/ref_chosen": -31.513410568237305,
      "logps/ref_rejected": -104.42971801757812,
      "logps/rejected": -184.76495361328125,
      "loss": 0.3473,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.8619762063026428,
      "rewards/margins": 3.0122811794281006,
      "rewards/rejected": -3.8742575645446777,
      "step": 108
    },
    {
      "epoch": 0.16005873715124816,
      "epsilon_dpo/beta": 0.04787931218743324,
      "epsilon_dpo/beta_margin_grad_mean": -0.17085351049900055,
      "epsilon_dpo/beta_margin_grad_std": 0.21672260761260986,
      "epsilon_dpo/beta_margin_mean": 2.650977611541748,
      "epsilon_dpo/beta_margin_std": 2.2566730976104736,
      "epsilon_dpo/loss_margin_mean": 55.51750564575195,
      "grad_norm": 100.95081329345703,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0482666902244091,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.95006729368358e-07,
      "logits/chosen": -3.113433361053467,
      "logits/rejected": -3.186359167098999,
      "logps/chosen": -64.14362335205078,
      "logps/ref_chosen": -45.64976501464844,
      "logps/ref_rejected": -104.12217712402344,
      "logps/rejected": -178.133544921875,
      "loss": 0.5342,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.8883336782455444,
      "rewards/margins": 2.650977611541748,
      "rewards/rejected": -3.539311408996582,
      "step": 109
    },
    {
      "epoch": 0.16152716593245228,
      "epsilon_dpo/beta": 0.04758320748806,
      "epsilon_dpo/beta_margin_grad_mean": -0.20529796183109283,
      "epsilon_dpo/beta_margin_grad_std": 0.21916480362415314,
      "epsilon_dpo/beta_margin_mean": 2.1833033561706543,
      "epsilon_dpo/beta_margin_std": 2.0075676441192627,
      "epsilon_dpo/loss_margin_mean": 46.09016799926758,
      "grad_norm": 87.10065460205078,
      "kl/avg_steps": 0.625,
      "kl/beta": 0.047877684235572815,
      "kl/n_epsilon_steps": 0.1875,
      "kl/p_epsilon_steps": 0.8125,
      "learning_rate": 4.947482930773511e-07,
      "logits/chosen": -3.0851223468780518,
      "logits/rejected": -3.0515449047088623,
      "logps/chosen": -74.82107543945312,
      "logps/ref_chosen": -50.388492584228516,
      "logps/ref_rejected": -85.45807647705078,
      "logps/rejected": -155.9808349609375,
      "loss": 0.5765,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -1.1659729480743408,
      "rewards/margins": 2.1833033561706543,
      "rewards/rejected": -3.349276065826416,
      "step": 110
    },
    {
      "epoch": 0.16299559471365638,
      "epsilon_dpo/beta": 0.04724304750561714,
      "epsilon_dpo/beta_margin_grad_mean": -0.17633913457393646,
      "epsilon_dpo/beta_margin_grad_std": 0.20233316719532013,
      "epsilon_dpo/beta_margin_mean": 2.6269123554229736,
      "epsilon_dpo/beta_margin_std": 2.2809414863586426,
      "epsilon_dpo/loss_margin_mean": 55.77882766723633,
      "grad_norm": 77.24600219726562,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.047580309212207794,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 4.944834074412042e-07,
      "logits/chosen": -3.054626941680908,
      "logits/rejected": -3.125295877456665,
      "logps/chosen": -75.80860900878906,
      "logps/ref_chosen": -50.71482849121094,
      "logps/ref_rejected": -103.22978210449219,
      "logps/rejected": -184.10238647460938,
      "loss": 0.4814,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.1880950927734375,
      "rewards/margins": 2.6269125938415527,
      "rewards/rejected": -3.8150076866149902,
      "step": 111
    },
    {
      "epoch": 0.1644640234948605,
      "epsilon_dpo/beta": 0.04686162248253822,
      "epsilon_dpo/beta_margin_grad_mean": -0.19428886473178864,
      "epsilon_dpo/beta_margin_grad_std": 0.19283361732959747,
      "epsilon_dpo/beta_margin_mean": 2.084317207336426,
      "epsilon_dpo/beta_margin_std": 1.6938108205795288,
      "epsilon_dpo/loss_margin_mean": 44.58252716064453,
      "grad_norm": 59.31022262573242,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.04724076762795448,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.942120794399002e-07,
      "logits/chosen": -3.0048656463623047,
      "logits/rejected": -3.0176172256469727,
      "logps/chosen": -50.82137680053711,
      "logps/ref_chosen": -31.809959411621094,
      "logps/ref_rejected": -68.63668823242188,
      "logps/rejected": -132.23062133789062,
      "loss": 0.5174,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.8925973176956177,
      "rewards/margins": 2.084317207336426,
      "rewards/rejected": -2.976914405822754,
      "step": 112
    },
    {
      "epoch": 0.16593245227606462,
      "epsilon_dpo/beta": 0.04648394137620926,
      "epsilon_dpo/beta_margin_grad_mean": -0.1680552214384079,
      "epsilon_dpo/beta_margin_grad_std": 0.16932794451713562,
      "epsilon_dpo/beta_margin_mean": 2.274296760559082,
      "epsilon_dpo/beta_margin_std": 1.6202584505081177,
      "epsilon_dpo/loss_margin_mean": 49.026641845703125,
      "grad_norm": 57.292938232421875,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.04686002805829048,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.939343162231841e-07,
      "logits/chosen": -3.054755687713623,
      "logits/rejected": -3.0067625045776367,
      "logps/chosen": -64.85562896728516,
      "logps/ref_chosen": -42.08271408081055,
      "logps/ref_rejected": -83.40155029296875,
      "logps/rejected": -155.20111083984375,
      "loss": 0.4213,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.0600247383117676,
      "rewards/margins": 2.274296760559082,
      "rewards/rejected": -3.3343217372894287,
      "step": 113
    },
    {
      "epoch": 0.16740088105726872,
      "epsilon_dpo/beta": 0.04610930010676384,
      "epsilon_dpo/beta_margin_grad_mean": -0.1812613159418106,
      "epsilon_dpo/beta_margin_grad_std": 0.20109733939170837,
      "epsilon_dpo/beta_margin_mean": 2.5884387493133545,
      "epsilon_dpo/beta_margin_std": 2.3022119998931885,
      "epsilon_dpo/loss_margin_mean": 56.26390075683594,
      "grad_norm": 70.24754333496094,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.04648235812783241,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.936501251103751e-07,
      "logits/chosen": -2.998676300048828,
      "logits/rejected": -3.029796600341797,
      "logps/chosen": -63.09971237182617,
      "logps/ref_chosen": -42.380760192871094,
      "logps/ref_rejected": -91.27511596679688,
      "logps/rejected": -168.25796508789062,
      "loss": 0.4844,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.9574916362762451,
      "rewards/margins": 2.5884387493133545,
      "rewards/rejected": -3.5459303855895996,
      "step": 114
    },
    {
      "epoch": 0.16886930983847284,
      "epsilon_dpo/beta": 0.04570886492729187,
      "epsilon_dpo/beta_margin_grad_mean": -0.16286557912826538,
      "epsilon_dpo/beta_margin_grad_std": 0.18301641941070557,
      "epsilon_dpo/beta_margin_mean": 2.5425362586975098,
      "epsilon_dpo/beta_margin_std": 1.9873340129852295,
      "epsilon_dpo/loss_margin_mean": 55.71139144897461,
      "grad_norm": 66.69357299804688,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.046107735484838486,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.933595135901732e-07,
      "logits/chosen": -3.068523406982422,
      "logits/rejected": -3.093740463256836,
      "logps/chosen": -61.14916229248047,
      "logps/ref_chosen": -40.20535659790039,
      "logps/ref_rejected": -106.33647155761719,
      "logps/rejected": -182.99166870117188,
      "loss": 0.4575,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.9589067697525024,
      "rewards/margins": 2.5425362586975098,
      "rewards/rejected": -3.5014431476593018,
      "step": 115
    },
    {
      "epoch": 0.17033773861967694,
      "epsilon_dpo/beta": 0.04532666504383087,
      "epsilon_dpo/beta_margin_grad_mean": -0.17079366743564606,
      "epsilon_dpo/beta_margin_grad_std": 0.1683167964220047,
      "epsilon_dpo/beta_margin_mean": 2.3608319759368896,
      "epsilon_dpo/beta_margin_std": 1.805658221244812,
      "epsilon_dpo/loss_margin_mean": 52.17644500732422,
      "grad_norm": 69.93063354492188,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.04570779204368591,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.930624893204624e-07,
      "logits/chosen": -2.8386454582214355,
      "logits/rejected": -2.940783977508545,
      "logps/chosen": -52.52592086791992,
      "logps/ref_chosen": -30.788612365722656,
      "logps/ref_rejected": -85.05826568603516,
      "logps/rejected": -158.97201538085938,
      "loss": 0.4288,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.9880549907684326,
      "rewards/margins": 2.3608322143554688,
      "rewards/rejected": -3.3488869667053223,
      "step": 116
    },
    {
      "epoch": 0.17180616740088106,
      "epsilon_dpo/beta": 0.04501824453473091,
      "epsilon_dpo/beta_margin_grad_mean": -0.18530869483947754,
      "epsilon_dpo/beta_margin_grad_std": 0.18177969753742218,
      "epsilon_dpo/beta_margin_mean": 2.1722805500030518,
      "epsilon_dpo/beta_margin_std": 1.6980078220367432,
      "epsilon_dpo/loss_margin_mean": 48.40717697143555,
      "grad_norm": 82.57999420166016,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.04532535746693611,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 4.927590601281083e-07,
      "logits/chosen": -2.9002761840820312,
      "logits/rejected": -2.8406801223754883,
      "logps/chosen": -77.98645782470703,
      "logps/ref_chosen": -49.41318893432617,
      "logps/ref_rejected": -72.67740631103516,
      "logps/rejected": -149.6578369140625,
      "loss": 0.4764,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.2893564701080322,
      "rewards/margins": 2.1722805500030518,
      "rewards/rejected": -3.461637020111084,
      "step": 117
    },
    {
      "epoch": 0.17327459618208516,
      "epsilon_dpo/beta": 0.044626448303461075,
      "epsilon_dpo/beta_margin_grad_mean": -0.15689337253570557,
      "epsilon_dpo/beta_margin_grad_std": 0.17096935212612152,
      "epsilon_dpo/beta_margin_mean": 2.458299398422241,
      "epsilon_dpo/beta_margin_std": 1.716274380683899,
      "epsilon_dpo/loss_margin_mean": 55.162723541259766,
      "grad_norm": 55.6863899230957,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.04501587525010109,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.924492340087524e-07,
      "logits/chosen": -2.8982632160186768,
      "logits/rejected": -2.9912490844726562,
      "logps/chosen": -58.43132400512695,
      "logps/ref_chosen": -35.58433532714844,
      "logps/ref_rejected": -80.31232452392578,
      "logps/rejected": -158.32205200195312,
      "loss": 0.3965,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.0198127031326294,
      "rewards/margins": 2.458299398422241,
      "rewards/rejected": -3.47811222076416,
      "step": 118
    },
    {
      "epoch": 0.17474302496328928,
      "epsilon_dpo/beta": 0.04427415132522583,
      "epsilon_dpo/beta_margin_grad_mean": -0.16944344341754913,
      "epsilon_dpo/beta_margin_grad_std": 0.20828606188297272,
      "epsilon_dpo/beta_margin_mean": 2.63626766204834,
      "epsilon_dpo/beta_margin_std": 2.4180660247802734,
      "epsilon_dpo/loss_margin_mean": 59.708961486816406,
      "grad_norm": 105.04483795166016,
      "kl/avg_steps": 0.796875,
      "kl/beta": 0.04462540149688721,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 4.92133019126601e-07,
      "logits/chosen": -2.9290332794189453,
      "logits/rejected": -2.974984645843506,
      "logps/chosen": -68.64191436767578,
      "logps/ref_chosen": -41.200706481933594,
      "logps/ref_rejected": -109.11971282958984,
      "logps/rejected": -196.2698974609375,
      "loss": 0.5632,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.2193212509155273,
      "rewards/margins": 2.63626766204834,
      "rewards/rejected": -3.855588912963867,
      "step": 119
    },
    {
      "epoch": 0.1762114537444934,
      "epsilon_dpo/beta": 0.043917279690504074,
      "epsilon_dpo/beta_margin_grad_mean": -0.1446925848722458,
      "epsilon_dpo/beta_margin_grad_std": 0.17125862836837769,
      "epsilon_dpo/beta_margin_mean": 2.938814163208008,
      "epsilon_dpo/beta_margin_std": 2.2117486000061035,
      "epsilon_dpo/loss_margin_mean": 67.05432891845703,
      "grad_norm": 58.40270233154297,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.04427260532975197,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.918104238142103e-07,
      "logits/chosen": -2.980833053588867,
      "logits/rejected": -2.9574685096740723,
      "logps/chosen": -79.7890625,
      "logps/ref_chosen": -53.9609489440918,
      "logps/ref_rejected": -114.34416961669922,
      "logps/rejected": -207.2266082763672,
      "loss": 0.3704,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.1369682550430298,
      "rewards/margins": 2.938814163208008,
      "rewards/rejected": -4.075782775878906,
      "step": 120
    },
    {
      "epoch": 0.1776798825256975,
      "epsilon_dpo/beta": 0.04350843280553818,
      "epsilon_dpo/beta_margin_grad_mean": -0.12654456496238708,
      "epsilon_dpo/beta_margin_grad_std": 0.1407029926776886,
      "epsilon_dpo/beta_margin_mean": 3.131282091140747,
      "epsilon_dpo/beta_margin_std": 2.3148951530456543,
      "epsilon_dpo/loss_margin_mean": 72.01702117919922,
      "grad_norm": 43.766937255859375,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.04391578957438469,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 4.91481456572267e-07,
      "logits/chosen": -2.7587661743164062,
      "logits/rejected": -2.9042224884033203,
      "logps/chosen": -57.087093353271484,
      "logps/ref_chosen": -32.518829345703125,
      "logps/ref_rejected": -119.22189331054688,
      "logps/rejected": -215.80718994140625,
      "loss": 0.3024,
      "rewards/accuracies": 0.984375,
      "rewards/chosen": -1.0692138671875,
      "rewards/margins": 3.131281852722168,
      "rewards/rejected": -4.200495719909668,
      "step": 121
    },
    {
      "epoch": 0.17914831130690162,
      "epsilon_dpo/beta": 0.04314511641860008,
      "epsilon_dpo/beta_margin_grad_mean": -0.13571257889270782,
      "epsilon_dpo/beta_margin_grad_std": 0.1690312772989273,
      "epsilon_dpo/beta_margin_mean": 3.182982921600342,
      "epsilon_dpo/beta_margin_std": 2.4074745178222656,
      "epsilon_dpo/loss_margin_mean": 73.9016342163086,
      "grad_norm": 72.01390075683594,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.0435079000890255,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.911461260693638e-07,
      "logits/chosen": -2.679837942123413,
      "logits/rejected": -2.8655993938446045,
      "logps/chosen": -59.01881790161133,
      "logps/ref_chosen": -31.410724639892578,
      "logps/ref_rejected": -114.03694152832031,
      "logps/rejected": -215.54666137695312,
      "loss": 0.3434,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.1941652297973633,
      "rewards/margins": 3.182982921600342,
      "rewards/rejected": -4.377148151397705,
      "step": 122
    },
    {
      "epoch": 0.18061674008810572,
      "epsilon_dpo/beta": 0.04283806309103966,
      "epsilon_dpo/beta_margin_grad_mean": -0.16673751175403595,
      "epsilon_dpo/beta_margin_grad_std": 0.19917066395282745,
      "epsilon_dpo/beta_margin_mean": 2.654533624649048,
      "epsilon_dpo/beta_margin_std": 2.180321216583252,
      "epsilon_dpo/loss_margin_mean": 62.16706085205078,
      "grad_norm": 75.45635223388672,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.043143875896930695,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 4.908044411417711e-07,
      "logits/chosen": -2.7718982696533203,
      "logits/rejected": -2.8315348625183105,
      "logps/chosen": -72.34667205810547,
      "logps/ref_chosen": -45.39820098876953,
      "logps/ref_rejected": -95.36239624023438,
      "logps/rejected": -184.47792053222656,
      "loss": 0.4491,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -1.1574070453643799,
      "rewards/margins": 2.654533624649048,
      "rewards/rejected": -3.8119406700134277,
      "step": 123
    },
    {
      "epoch": 0.18208516886930984,
      "epsilon_dpo/beta": 0.04243864864110947,
      "epsilon_dpo/beta_margin_grad_mean": -0.12405575811862946,
      "epsilon_dpo/beta_margin_grad_std": 0.14530256390571594,
      "epsilon_dpo/beta_margin_mean": 3.2046613693237305,
      "epsilon_dpo/beta_margin_std": 2.2337818145751953,
      "epsilon_dpo/loss_margin_mean": 75.55775451660156,
      "grad_norm": 42.3536376953125,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.04283599182963371,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 4.904564107932048e-07,
      "logits/chosen": -2.7853384017944336,
      "logits/rejected": -2.904179096221924,
      "logps/chosen": -68.72981262207031,
      "logps/ref_chosen": -42.682899475097656,
      "logps/ref_rejected": -124.48335266113281,
      "logps/rejected": -226.0880126953125,
      "loss": 0.2979,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -1.1055363416671753,
      "rewards/margins": 3.2046613693237305,
      "rewards/rejected": -4.310197830200195,
      "step": 124
    },
    {
      "epoch": 0.18355359765051396,
      "epsilon_dpo/beta": 0.04211079701781273,
      "epsilon_dpo/beta_margin_grad_mean": -0.16087745130062103,
      "epsilon_dpo/beta_margin_grad_std": 0.1763782650232315,
      "epsilon_dpo/beta_margin_mean": 2.658567428588867,
      "epsilon_dpo/beta_margin_std": 2.0561933517456055,
      "epsilon_dpo/loss_margin_mean": 63.280696868896484,
      "grad_norm": 81.41658020019531,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.04243813455104828,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 4.90102044194588e-07,
      "logits/chosen": -2.5554890632629395,
      "logits/rejected": -2.767099618911743,
      "logps/chosen": -63.22058868408203,
      "logps/ref_chosen": -30.618457794189453,
      "logps/ref_rejected": -88.96886444091797,
      "logps/rejected": -184.8516845703125,
      "loss": 0.4111,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.3759876489639282,
      "rewards/margins": 2.658567428588867,
      "rewards/rejected": -4.034554958343506,
      "step": 125
    },
    {
      "epoch": 0.18502202643171806,
      "epsilon_dpo/beta": 0.041758038103580475,
      "epsilon_dpo/beta_margin_grad_mean": -0.1536438912153244,
      "epsilon_dpo/beta_margin_grad_std": 0.21458838880062103,
      "epsilon_dpo/beta_margin_mean": 2.926898956298828,
      "epsilon_dpo/beta_margin_std": 2.228339433670044,
      "epsilon_dpo/loss_margin_mean": 70.24527740478516,
      "grad_norm": 81.50450897216797,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.042109157890081406,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.897413506838102e-07,
      "logits/chosen": -2.6770665645599365,
      "logits/rejected": -2.785977840423584,
      "logps/chosen": -72.39849090576172,
      "logps/ref_chosen": -38.92976379394531,
      "logps/ref_rejected": -104.5777587890625,
      "logps/rejected": -208.291748046875,
      "loss": 0.4561,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.4002439975738525,
      "rewards/margins": 2.926898956298828,
      "rewards/rejected": -4.327142715454102,
      "step": 126
    },
    {
      "epoch": 0.18649045521292218,
      "epsilon_dpo/beta": 0.04140865430235863,
      "epsilon_dpo/beta_margin_grad_mean": -0.1614929735660553,
      "epsilon_dpo/beta_margin_grad_std": 0.1582755744457245,
      "epsilon_dpo/beta_margin_mean": 2.3239240646362305,
      "epsilon_dpo/beta_margin_std": 1.5966986417770386,
      "epsilon_dpo/loss_margin_mean": 56.2106819152832,
      "grad_norm": 61.22636413574219,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.04175683483481407,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.89374339765481e-07,
      "logits/chosen": -2.705301523208618,
      "logits/rejected": -2.7008934020996094,
      "logps/chosen": -69.19281005859375,
      "logps/ref_chosen": -39.06876754760742,
      "logps/ref_rejected": -81.60107421875,
      "logps/rejected": -167.935791015625,
      "loss": 0.395,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.2484067678451538,
      "rewards/margins": 2.3239240646362305,
      "rewards/rejected": -3.5723307132720947,
      "step": 127
    },
    {
      "epoch": 0.18795888399412627,
      "epsilon_dpo/beta": 0.04112689569592476,
      "epsilon_dpo/beta_margin_grad_mean": -0.18886899948120117,
      "epsilon_dpo/beta_margin_grad_std": 0.20472285151481628,
      "epsilon_dpo/beta_margin_mean": 2.6414670944213867,
      "epsilon_dpo/beta_margin_std": 2.408226251602173,
      "epsilon_dpo/loss_margin_mean": 64.43711853027344,
      "grad_norm": 69.2333984375,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.04140745848417282,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 4.890010211106795e-07,
      "logits/chosen": -2.7107484340667725,
      "logits/rejected": -2.624185085296631,
      "logps/chosen": -70.23330688476562,
      "logps/ref_chosen": -37.46366882324219,
      "logps/ref_rejected": -81.10179138183594,
      "logps/rejected": -178.3085479736328,
      "loss": 0.5065,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -1.3495298624038696,
      "rewards/margins": 2.6414670944213867,
      "rewards/rejected": -3.990996837615967,
      "step": 128
    },
    {
      "epoch": 0.1894273127753304,
      "epsilon_dpo/beta": 0.040820371359586716,
      "epsilon_dpo/beta_margin_grad_mean": -0.16870415210723877,
      "epsilon_dpo/beta_margin_grad_std": 0.19453568756580353,
      "epsilon_dpo/beta_margin_mean": 2.5504720211029053,
      "epsilon_dpo/beta_margin_std": 2.0157787799835205,
      "epsilon_dpo/loss_margin_mean": 62.65648651123047,
      "grad_norm": 62.1826057434082,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.041124723851680756,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 4.88621404556699e-07,
      "logits/chosen": -2.753101348876953,
      "logits/rejected": -2.694021701812744,
      "logps/chosen": -82.27946472167969,
      "logps/ref_chosen": -48.18524932861328,
      "logps/ref_rejected": -98.1295166015625,
      "logps/rejected": -194.88021850585938,
      "loss": 0.4566,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.3921856880187988,
      "rewards/margins": 2.5504720211029053,
      "rewards/rejected": -3.942657709121704,
      "step": 129
    },
    {
      "epoch": 0.19089574155653452,
      "epsilon_dpo/beta": 0.04049098491668701,
      "epsilon_dpo/beta_margin_grad_mean": -0.14500190317630768,
      "epsilon_dpo/beta_margin_grad_std": 0.1896047294139862,
      "epsilon_dpo/beta_margin_mean": 2.9150590896606445,
      "epsilon_dpo/beta_margin_std": 2.0989155769348145,
      "epsilon_dpo/loss_margin_mean": 72.14749145507812,
      "grad_norm": 67.861572265625,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.04081858694553375,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.882355001067891e-07,
      "logits/chosen": -2.629741668701172,
      "logits/rejected": -2.686960220336914,
      "logps/chosen": -60.728355407714844,
      "logps/ref_chosen": -28.03460121154785,
      "logps/ref_rejected": -87.78228759765625,
      "logps/rejected": -192.62353515625,
      "loss": 0.3878,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.3256336450576782,
      "rewards/margins": 2.9150590896606445,
      "rewards/rejected": -4.240693092346191,
      "step": 130
    },
    {
      "epoch": 0.19236417033773862,
      "epsilon_dpo/beta": 0.040177300572395325,
      "epsilon_dpo/beta_margin_grad_mean": -0.18838103115558624,
      "epsilon_dpo/beta_margin_grad_std": 0.19518207013607025,
      "epsilon_dpo/beta_margin_mean": 2.35014009475708,
      "epsilon_dpo/beta_margin_std": 1.9053841829299927,
      "epsilon_dpo/loss_margin_mean": 58.63149642944336,
      "grad_norm": 55.52901077270508,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.04048960655927658,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 4.878433179298909e-07,
      "logits/chosen": -2.6353015899658203,
      "logits/rejected": -2.799935817718506,
      "logps/chosen": -64.12167358398438,
      "logps/ref_chosen": -34.23711013793945,
      "logps/ref_rejected": -94.21966552734375,
      "logps/rejected": -182.7357177734375,
      "loss": 0.4887,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -1.201073169708252,
      "rewards/margins": 2.35014009475708,
      "rewards/rejected": -3.551213264465332,
      "step": 131
    },
    {
      "epoch": 0.19383259911894274,
      "epsilon_dpo/beta": 0.0398343950510025,
      "epsilon_dpo/beta_margin_grad_mean": -0.15264593064785004,
      "epsilon_dpo/beta_margin_grad_std": 0.15929141640663147,
      "epsilon_dpo/beta_margin_mean": 2.4843099117279053,
      "epsilon_dpo/beta_margin_std": 1.845669150352478,
      "epsilon_dpo/loss_margin_mean": 62.46304702758789,
      "grad_norm": 69.0653305053711,
      "kl/avg_steps": 0.859375,
      "kl/beta": 0.04017573595046997,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.874448683603694e-07,
      "logits/chosen": -2.7254421710968018,
      "logits/rejected": -2.769134044647217,
      "logps/chosen": -86.7735824584961,
      "logps/ref_chosen": -48.03201675415039,
      "logps/ref_rejected": -93.43745422363281,
      "logps/rejected": -194.64205932617188,
      "loss": 0.3811,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -1.5439248085021973,
      "rewards/margins": 2.4843099117279053,
      "rewards/rejected": -4.028234958648682,
      "step": 132
    },
    {
      "epoch": 0.19530102790014683,
      "epsilon_dpo/beta": 0.03946392610669136,
      "epsilon_dpo/beta_margin_grad_mean": -0.14380493760108948,
      "epsilon_dpo/beta_margin_grad_std": 0.15240231156349182,
      "epsilon_dpo/beta_margin_mean": 2.68829607963562,
      "epsilon_dpo/beta_margin_std": 1.8531904220581055,
      "epsilon_dpo/loss_margin_mean": 68.16297149658203,
      "grad_norm": 58.320655822753906,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.03983341529965401,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 4.870401618977415e-07,
      "logits/chosen": -2.6911277770996094,
      "logits/rejected": -2.8012609481811523,
      "logps/chosen": -73.2670669555664,
      "logps/ref_chosen": -38.89875793457031,
      "logps/ref_rejected": -103.64884948730469,
      "logps/rejected": -206.1801300048828,
      "loss": 0.3486,
      "rewards/accuracies": 0.984375,
      "rewards/chosen": -1.3570879697799683,
      "rewards/margins": 2.688295841217041,
      "rewards/rejected": -4.045383930206299,
      "step": 133
    },
    {
      "epoch": 0.19676945668135096,
      "epsilon_dpo/beta": 0.03912205249071121,
      "epsilon_dpo/beta_margin_grad_mean": -0.15723571181297302,
      "epsilon_dpo/beta_margin_grad_std": 0.16265857219696045,
      "epsilon_dpo/beta_margin_mean": 2.531024217605591,
      "epsilon_dpo/beta_margin_std": 1.835780143737793,
      "epsilon_dpo/loss_margin_mean": 64.7786865234375,
      "grad_norm": 53.25428009033203,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.03946344554424286,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.866292092063986e-07,
      "logits/chosen": -2.665362596511841,
      "logits/rejected": -2.729599952697754,
      "logps/chosen": -71.48147583007812,
      "logps/ref_chosen": -37.08332061767578,
      "logps/ref_rejected": -93.31185150146484,
      "logps/rejected": -192.48870849609375,
      "loss": 0.3876,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.346604824066162,
      "rewards/margins": 2.531024217605591,
      "rewards/rejected": -3.877628803253174,
      "step": 134
    },
    {
      "epoch": 0.19823788546255505,
      "epsilon_dpo/beta": 0.03875825181603432,
      "epsilon_dpo/beta_margin_grad_mean": -0.14636199176311493,
      "epsilon_dpo/beta_margin_grad_std": 0.1549866497516632,
      "epsilon_dpo/beta_margin_mean": 2.7066292762756348,
      "epsilon_dpo/beta_margin_std": 2.084244966506958,
      "epsilon_dpo/loss_margin_mean": 69.87920379638672,
      "grad_norm": 58.53507995605469,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.03912113606929779,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 4.862120211153265e-07,
      "logits/chosen": -2.4796104431152344,
      "logits/rejected": -2.7320666313171387,
      "logps/chosen": -68.14396667480469,
      "logps/ref_chosen": -28.130008697509766,
      "logps/ref_rejected": -121.04113006591797,
      "logps/rejected": -230.93429565429688,
      "loss": 0.3559,
      "rewards/accuracies": 0.984375,
      "rewards/chosen": -1.5515646934509277,
      "rewards/margins": 2.7066292762756348,
      "rewards/rejected": -4.2581939697265625,
      "step": 135
    },
    {
      "epoch": 0.19970631424375918,
      "epsilon_dpo/beta": 0.03843460604548454,
      "epsilon_dpo/beta_margin_grad_mean": -0.1646306961774826,
      "epsilon_dpo/beta_margin_grad_std": 0.19557887315750122,
      "epsilon_dpo/beta_margin_mean": 2.8479745388031006,
      "epsilon_dpo/beta_margin_std": 2.358449935913086,
      "epsilon_dpo/loss_margin_mean": 74.22594451904297,
      "grad_norm": 71.29747009277344,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.03875778242945671,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.857886086178193e-07,
      "logits/chosen": -2.642530918121338,
      "logits/rejected": -2.726252794265747,
      "logps/chosen": -80.63860321044922,
      "logps/ref_chosen": -41.21574401855469,
      "logps/ref_rejected": -103.0965805053711,
      "logps/rejected": -216.74537658691406,
      "loss": 0.4334,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.5163445472717285,
      "rewards/margins": 2.8479745388031006,
      "rewards/rejected": -4.36431884765625,
      "step": 136
    },
    {
      "epoch": 0.2011747430249633,
      "epsilon_dpo/beta": 0.03807699307799339,
      "epsilon_dpo/beta_margin_grad_mean": -0.125637486577034,
      "epsilon_dpo/beta_margin_grad_std": 0.15655429661273956,
      "epsilon_dpo/beta_margin_mean": 3.427241325378418,
      "epsilon_dpo/beta_margin_std": 2.608391523361206,
      "epsilon_dpo/loss_margin_mean": 90.0654296875,
      "grad_norm": 50.14603805541992,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.0384334996342659,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 4.853589828711902e-07,
      "logits/chosen": -2.384611129760742,
      "logits/rejected": -2.656006336212158,
      "logps/chosen": -75.71855163574219,
      "logps/ref_chosen": -32.656131744384766,
      "logps/ref_rejected": -119.52920532226562,
      "logps/rejected": -252.65704345703125,
      "loss": 0.3075,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -1.6396162509918213,
      "rewards/margins": 3.427241325378418,
      "rewards/rejected": -5.06685733795166,
      "step": 137
    },
    {
      "epoch": 0.2026431718061674,
      "epsilon_dpo/beta": 0.03774713724851608,
      "epsilon_dpo/beta_margin_grad_mean": -0.13305923342704773,
      "epsilon_dpo/beta_margin_grad_std": 0.17672809958457947,
      "epsilon_dpo/beta_margin_mean": 3.189133644104004,
      "epsilon_dpo/beta_margin_std": 2.1990420818328857,
      "epsilon_dpo/loss_margin_mean": 84.6056137084961,
      "grad_norm": 64.70891571044922,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.038076531141996384,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.849231551964771e-07,
      "logits/chosen": -2.3941986560821533,
      "logits/rejected": -2.439361095428467,
      "logps/chosen": -81.6050796508789,
      "logps/ref_chosen": -31.68457794189453,
      "logps/ref_rejected": -99.58272552490234,
      "logps/rejected": -234.10885620117188,
      "loss": 0.3421,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.8868746757507324,
      "rewards/margins": 3.189133644104004,
      "rewards/rejected": -5.076008319854736,
      "step": 138
    },
    {
      "epoch": 0.20411160058737152,
      "epsilon_dpo/beta": 0.03741971403360367,
      "epsilon_dpo/beta_margin_grad_mean": -0.10873990505933762,
      "epsilon_dpo/beta_margin_grad_std": 0.18296301364898682,
      "epsilon_dpo/beta_margin_mean": 3.4576268196105957,
      "epsilon_dpo/beta_margin_std": 2.2132863998413086,
      "epsilon_dpo/loss_margin_mean": 92.54183959960938,
      "grad_norm": 64.89179992675781,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.03774625062942505,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.844811370781446e-07,
      "logits/chosen": -2.3702569007873535,
      "logits/rejected": -2.3330612182617188,
      "logps/chosen": -74.70332336425781,
      "logps/ref_chosen": -30.504302978515625,
      "logps/ref_rejected": -85.29287719726562,
      "logps/rejected": -222.0337371826172,
      "loss": 0.3096,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.656851053237915,
      "rewards/margins": 3.4576268196105957,
      "rewards/rejected": -5.11447811126709,
      "step": 139
    },
    {
      "epoch": 0.2055800293685756,
      "epsilon_dpo/beta": 0.037118520587682724,
      "epsilon_dpo/beta_margin_grad_mean": -0.14209268987178802,
      "epsilon_dpo/beta_margin_grad_std": 0.18320953845977783,
      "epsilon_dpo/beta_margin_mean": 3.001073122024536,
      "epsilon_dpo/beta_margin_std": 2.1810362339019775,
      "epsilon_dpo/loss_margin_mean": 81.0239486694336,
      "grad_norm": 64.97393035888672,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.03741883859038353,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.840329401637809e-07,
      "logits/chosen": -2.463718891143799,
      "logits/rejected": -2.454908609390259,
      "logps/chosen": -87.36968994140625,
      "logps/ref_chosen": -36.32595443725586,
      "logps/ref_rejected": -89.32574462890625,
      "logps/rejected": -221.3934326171875,
      "loss": 0.3733,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.8963526487350464,
      "rewards/margins": 3.001072883605957,
      "rewards/rejected": -4.897425651550293,
      "step": 140
    },
    {
      "epoch": 0.20704845814977973,
      "epsilon_dpo/beta": 0.03681936860084534,
      "epsilon_dpo/beta_margin_grad_mean": -0.12768100202083588,
      "epsilon_dpo/beta_margin_grad_std": 0.1734439581632614,
      "epsilon_dpo/beta_margin_mean": 3.1639914512634277,
      "epsilon_dpo/beta_margin_std": 2.328209638595581,
      "epsilon_dpo/loss_margin_mean": 86.11234283447266,
      "grad_norm": 73.34292602539062,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.037117261439561844,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.83578576263792e-07,
      "logits/chosen": -2.636939287185669,
      "logits/rejected": -2.580254316329956,
      "logps/chosen": -108.33308410644531,
      "logps/ref_chosen": -54.31216049194336,
      "logps/ref_rejected": -105.19206237792969,
      "logps/rejected": -245.32533264160156,
      "loss": 0.3351,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.991283655166626,
      "rewards/margins": 3.1639914512634277,
      "rewards/rejected": -5.155275344848633,
      "step": 141
    },
    {
      "epoch": 0.20851688693098386,
      "epsilon_dpo/beta": 0.03649960458278656,
      "epsilon_dpo/beta_margin_grad_mean": -0.14174862205982208,
      "epsilon_dpo/beta_margin_grad_std": 0.17707964777946472,
      "epsilon_dpo/beta_margin_mean": 3.183145761489868,
      "epsilon_dpo/beta_margin_std": 2.435292959213257,
      "epsilon_dpo/loss_margin_mean": 87.32428741455078,
      "grad_norm": 72.56526947021484,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.036818113178014755,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.83118057351089e-07,
      "logits/chosen": -2.3550641536712646,
      "logits/rejected": -2.433764934539795,
      "logps/chosen": -91.45923614501953,
      "logps/ref_chosen": -37.70738220214844,
      "logps/ref_rejected": -100.41778564453125,
      "logps/rejected": -241.49392700195312,
      "loss": 0.3654,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.9631787538528442,
      "rewards/margins": 3.183145523071289,
      "rewards/rejected": -5.146324157714844,
      "step": 142
    },
    {
      "epoch": 0.20998531571218795,
      "epsilon_dpo/beta": 0.03624003753066063,
      "epsilon_dpo/beta_margin_grad_mean": -0.1898718923330307,
      "epsilon_dpo/beta_margin_grad_std": 0.2213219702243805,
      "epsilon_dpo/beta_margin_mean": 2.72891902923584,
      "epsilon_dpo/beta_margin_std": 2.580869674682617,
      "epsilon_dpo/loss_margin_mean": 75.5594482421875,
      "grad_norm": 89.1434555053711,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.036498747766017914,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 4.826513955607734e-07,
      "logits/chosen": -2.482374668121338,
      "logits/rejected": -2.3714542388916016,
      "logps/chosen": -89.65013122558594,
      "logps/ref_chosen": -40.36812210083008,
      "logps/ref_rejected": -88.28787994384766,
      "logps/rejected": -213.12933349609375,
      "loss": 0.5517,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.7896441221237183,
      "rewards/margins": 2.72891902923584,
      "rewards/rejected": -4.518563270568848,
      "step": 143
    },
    {
      "epoch": 0.21145374449339208,
      "epsilon_dpo/beta": 0.03593612089753151,
      "epsilon_dpo/beta_margin_grad_mean": -0.14966890215873718,
      "epsilon_dpo/beta_margin_grad_std": 0.1911356896162033,
      "epsilon_dpo/beta_margin_mean": 2.930758237838745,
      "epsilon_dpo/beta_margin_std": 2.2233402729034424,
      "epsilon_dpo/loss_margin_mean": 81.71424102783203,
      "grad_norm": 74.61936950683594,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.03623828664422035,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.821786031898176e-07,
      "logits/chosen": -2.5220513343811035,
      "logits/rejected": -2.4428181648254395,
      "logps/chosen": -90.16825103759766,
      "logps/ref_chosen": -41.09404373168945,
      "logps/ref_rejected": -87.43101501464844,
      "logps/rejected": -218.21946716308594,
      "loss": 0.4112,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.765446662902832,
      "rewards/margins": 2.930758476257324,
      "rewards/rejected": -4.696205139160156,
      "step": 144
    },
    {
      "epoch": 0.21292217327459617,
      "epsilon_dpo/beta": 0.035612985491752625,
      "epsilon_dpo/beta_margin_grad_mean": -0.11719842255115509,
      "epsilon_dpo/beta_margin_grad_std": 0.13962174952030182,
      "epsilon_dpo/beta_margin_mean": 3.221985101699829,
      "epsilon_dpo/beta_margin_std": 2.2744228839874268,
      "epsilon_dpo/loss_margin_mean": 90.55402374267578,
      "grad_norm": 62.392303466796875,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.03593508526682854,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 4.816996926967401e-07,
      "logits/chosen": -2.4150798320770264,
      "logits/rejected": -2.3624231815338135,
      "logps/chosen": -86.23258972167969,
      "logps/ref_chosen": -35.7225341796875,
      "logps/ref_rejected": -84.17701721191406,
      "logps/rejected": -225.2410888671875,
      "loss": 0.2806,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -1.800328016281128,
      "rewards/margins": 3.221985340118408,
      "rewards/rejected": -5.022313117980957,
      "step": 145
    },
    {
      "epoch": 0.2143906020558003,
      "epsilon_dpo/beta": 0.03531539812684059,
      "epsilon_dpo/beta_margin_grad_mean": -0.1365576684474945,
      "epsilon_dpo/beta_margin_grad_std": 0.18428391218185425,
      "epsilon_dpo/beta_margin_mean": 3.0302278995513916,
      "epsilon_dpo/beta_margin_std": 2.1277804374694824,
      "epsilon_dpo/loss_margin_mean": 85.95875549316406,
      "grad_norm": 61.81754684448242,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.03561234846711159,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.812146767012779e-07,
      "logits/chosen": -2.4440078735351562,
      "logits/rejected": -2.263402223587036,
      "logps/chosen": -100.90562438964844,
      "logps/ref_chosen": -41.776546478271484,
      "logps/ref_rejected": -87.99478149414062,
      "logps/rejected": -233.08261108398438,
      "loss": 0.3739,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -2.0904293060302734,
      "rewards/margins": 3.0302281379699707,
      "rewards/rejected": -5.120657444000244,
      "step": 146
    },
    {
      "epoch": 0.21585903083700442,
      "epsilon_dpo/beta": 0.03503095358610153,
      "epsilon_dpo/beta_margin_grad_mean": -0.14339102804660797,
      "epsilon_dpo/beta_margin_grad_std": 0.21028229594230652,
      "epsilon_dpo/beta_margin_mean": 3.1163737773895264,
      "epsilon_dpo/beta_margin_std": 2.363018274307251,
      "epsilon_dpo/loss_margin_mean": 89.17818450927734,
      "grad_norm": 94.37625122070312,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.03531438112258911,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.807235679840536e-07,
      "logits/chosen": -2.4340085983276367,
      "logits/rejected": -2.291194200515747,
      "logps/chosen": -89.63932800292969,
      "logps/ref_chosen": -35.02455139160156,
      "logps/ref_rejected": -76.69776153564453,
      "logps/rejected": -220.49072265625,
      "loss": 0.4282,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.9178650379180908,
      "rewards/margins": 3.1163740158081055,
      "rewards/rejected": -5.034238815307617,
      "step": 147
    },
    {
      "epoch": 0.2173274596182085,
      "epsilon_dpo/beta": 0.03473767638206482,
      "epsilon_dpo/beta_margin_grad_mean": -0.18217557668685913,
      "epsilon_dpo/beta_margin_grad_std": 0.1949581652879715,
      "epsilon_dpo/beta_margin_mean": 2.4017815589904785,
      "epsilon_dpo/beta_margin_std": 2.13431715965271,
      "epsilon_dpo/loss_margin_mean": 69.28678894042969,
      "grad_norm": 83.76101684570312,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.03502976521849632,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.802263794862384e-07,
      "logits/chosen": -2.529165744781494,
      "logits/rejected": -2.569955825805664,
      "logps/chosen": -100.62034606933594,
      "logps/ref_chosen": -43.158294677734375,
      "logps/ref_rejected": -109.0255126953125,
      "logps/rejected": -235.7743682861328,
      "loss": 0.4897,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.9978649616241455,
      "rewards/margins": 2.4017815589904785,
      "rewards/rejected": -4.399646759033203,
      "step": 148
    },
    {
      "epoch": 0.21879588839941264,
      "epsilon_dpo/beta": 0.0344361737370491,
      "epsilon_dpo/beta_margin_grad_mean": -0.15145255625247955,
      "epsilon_dpo/beta_margin_grad_std": 0.1748467981815338,
      "epsilon_dpo/beta_margin_mean": 2.92539644241333,
      "epsilon_dpo/beta_margin_std": 2.3435420989990234,
      "epsilon_dpo/loss_margin_mean": 85.06315612792969,
      "grad_norm": 69.50299835205078,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.03473667427897453,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.797231243092118e-07,
      "logits/chosen": -2.407914638519287,
      "logits/rejected": -2.4749755859375,
      "logps/chosen": -93.2674331665039,
      "logps/ref_chosen": -34.801856994628906,
      "logps/ref_rejected": -101.17015075683594,
      "logps/rejected": -244.69888305664062,
      "loss": 0.383,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -2.014852285385132,
      "rewards/margins": 2.92539644241333,
      "rewards/rejected": -4.940248489379883,
      "step": 149
    },
    {
      "epoch": 0.22026431718061673,
      "epsilon_dpo/beta": 0.034148234874010086,
      "epsilon_dpo/beta_margin_grad_mean": -0.13967867195606232,
      "epsilon_dpo/beta_margin_grad_std": 0.1724717617034912,
      "epsilon_dpo/beta_margin_mean": 2.8002302646636963,
      "epsilon_dpo/beta_margin_std": 1.951874852180481,
      "epsilon_dpo/loss_margin_mean": 82.1502456665039,
      "grad_norm": 67.23019409179688,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.03443536534905434,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.792138157142157e-07,
      "logits/chosen": -2.3103485107421875,
      "logits/rejected": -2.399022102355957,
      "logps/chosen": -79.86786651611328,
      "logps/ref_chosen": -27.168495178222656,
      "logps/ref_rejected": -88.6546859741211,
      "logps/rejected": -223.50430297851562,
      "loss": 0.3576,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.8020522594451904,
      "rewards/margins": 2.800230026245117,
      "rewards/rejected": -4.602282524108887,
      "step": 150
    },
    {
      "epoch": 0.22173274596182085,
      "epsilon_dpo/beta": 0.033841170370578766,
      "epsilon_dpo/beta_margin_grad_mean": -0.12026865780353546,
      "epsilon_dpo/beta_margin_grad_std": 0.14584773778915405,
      "epsilon_dpo/beta_margin_mean": 2.9734950065612793,
      "epsilon_dpo/beta_margin_std": 1.9034627676010132,
      "epsilon_dpo/loss_margin_mean": 87.95604705810547,
      "grad_norm": 56.315635681152344,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.034147247672080994,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 4.786984671220053e-07,
      "logits/chosen": -2.5218310356140137,
      "logits/rejected": -2.627288341522217,
      "logps/chosen": -98.35040283203125,
      "logps/ref_chosen": -44.517967224121094,
      "logps/ref_rejected": -107.56175994873047,
      "logps/rejected": -249.35023498535156,
      "loss": 0.2918,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.8228228092193604,
      "rewards/margins": 2.9734950065612793,
      "rewards/rejected": -4.796318054199219,
      "step": 151
    },
    {
      "epoch": 0.22320117474302498,
      "epsilon_dpo/beta": 0.033547814935445786,
      "epsilon_dpo/beta_margin_grad_mean": -0.12259558588266373,
      "epsilon_dpo/beta_margin_grad_std": 0.1732029914855957,
      "epsilon_dpo/beta_margin_mean": 3.1676061153411865,
      "epsilon_dpo/beta_margin_std": 2.1665868759155273,
      "epsilon_dpo/loss_margin_mean": 94.56542205810547,
      "grad_norm": 68.73706817626953,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.033840566873550415,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.78177092112495e-07,
      "logits/chosen": -2.4150760173797607,
      "logits/rejected": -2.4403088092803955,
      "logps/chosen": -102.739501953125,
      "logps/ref_chosen": -45.36680603027344,
      "logps/ref_rejected": -112.48715209960938,
      "logps/rejected": -264.4252624511719,
      "loss": 0.3243,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.9261062145233154,
      "rewards/margins": 3.1676058769226074,
      "rewards/rejected": -5.093712329864502,
      "step": 152
    },
    {
      "epoch": 0.22466960352422907,
      "epsilon_dpo/beta": 0.033309243619441986,
      "epsilon_dpo/beta_margin_grad_mean": -0.17181988060474396,
      "epsilon_dpo/beta_margin_grad_std": 0.2117866724729538,
      "epsilon_dpo/beta_margin_mean": 2.878479480743408,
      "epsilon_dpo/beta_margin_std": 2.4571328163146973,
      "epsilon_dpo/loss_margin_mean": 86.69181060791016,
      "grad_norm": 84.02366638183594,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.03354702889919281,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 4.776497044244016e-07,
      "logits/chosen": -2.4134445190429688,
      "logits/rejected": -2.408932685852051,
      "logps/chosen": -99.41288757324219,
      "logps/ref_chosen": -39.70823669433594,
      "logps/ref_rejected": -91.66893768310547,
      "logps/rejected": -238.06539916992188,
      "loss": 0.468,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -1.9907963275909424,
      "rewards/margins": 2.878479480743408,
      "rewards/rejected": -4.86927604675293,
      "step": 153
    },
    {
      "epoch": 0.2261380323054332,
      "epsilon_dpo/beta": 0.033061131834983826,
      "epsilon_dpo/beta_margin_grad_mean": -0.15857023000717163,
      "epsilon_dpo/beta_margin_grad_std": 0.21813619136810303,
      "epsilon_dpo/beta_margin_mean": 3.233847141265869,
      "epsilon_dpo/beta_margin_std": 2.5847911834716797,
      "epsilon_dpo/loss_margin_mean": 98.10140991210938,
      "grad_norm": 101.56253051757812,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.03330763056874275,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 4.771163179548808e-07,
      "logits/chosen": -2.3385679721832275,
      "logits/rejected": -2.41574764251709,
      "logps/chosen": -91.09129333496094,
      "logps/ref_chosen": -35.708839416503906,
      "logps/ref_rejected": -111.15846252441406,
      "logps/rejected": -264.642333984375,
      "loss": 0.4488,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -1.834795594215393,
      "rewards/margins": 3.233847141265869,
      "rewards/rejected": -5.068642616271973,
      "step": 154
    },
    {
      "epoch": 0.2276064610866373,
      "epsilon_dpo/beta": 0.03278402239084244,
      "epsilon_dpo/beta_margin_grad_mean": -0.1480940580368042,
      "epsilon_dpo/beta_margin_grad_std": 0.18447977304458618,
      "epsilon_dpo/beta_margin_mean": 2.7261838912963867,
      "epsilon_dpo/beta_margin_std": 1.8952271938323975,
      "epsilon_dpo/loss_margin_mean": 83.30589294433594,
      "grad_norm": 63.52682876586914,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.033059682697057724,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.7657694675916247e-07,
      "logits/chosen": -2.438190460205078,
      "logits/rejected": -2.357145071029663,
      "logps/chosen": -102.16056823730469,
      "logps/ref_chosen": -42.04505157470703,
      "logps/ref_rejected": -84.00552368164062,
      "logps/rejected": -227.42694091796875,
      "loss": 0.384,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.9715728759765625,
      "rewards/margins": 2.7261838912963867,
      "rewards/rejected": -4.697756767272949,
      "step": 155
    },
    {
      "epoch": 0.2290748898678414,
      "epsilon_dpo/beta": 0.03256094828248024,
      "epsilon_dpo/beta_margin_grad_mean": -0.16920383274555206,
      "epsilon_dpo/beta_margin_grad_std": 0.22653749585151672,
      "epsilon_dpo/beta_margin_mean": 2.8284294605255127,
      "epsilon_dpo/beta_margin_std": 2.360419750213623,
      "epsilon_dpo/loss_margin_mean": 87.19612121582031,
      "grad_norm": 146.7109375,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.03278307616710663,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 4.7603160505017893e-07,
      "logits/chosen": -2.4288077354431152,
      "logits/rejected": -2.2443032264709473,
      "logps/chosen": -115.40594482421875,
      "logps/ref_chosen": -49.038185119628906,
      "logps/ref_rejected": -84.26383209228516,
      "logps/rejected": -237.8277130126953,
      "loss": 0.5071,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -2.1660971641540527,
      "rewards/margins": 2.828429698944092,
      "rewards/rejected": -4.9945268630981445,
      "step": 156
    },
    {
      "epoch": 0.2305433186490455,
      "epsilon_dpo/beta": 0.03226739168167114,
      "epsilon_dpo/beta_margin_grad_mean": -0.11788501590490341,
      "epsilon_dpo/beta_margin_grad_std": 0.1734883189201355,
      "epsilon_dpo/beta_margin_mean": 3.567206382751465,
      "epsilon_dpo/beta_margin_std": 2.317911386489868,
      "epsilon_dpo/loss_margin_mean": 110.67003631591797,
      "grad_norm": 74.27545928955078,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.03255923092365265,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 4.7548030719819154e-07,
      "logits/chosen": -2.416853904724121,
      "logits/rejected": -2.5221433639526367,
      "logps/chosen": -85.81341552734375,
      "logps/ref_chosen": -30.534713745117188,
      "logps/ref_rejected": -114.2703628540039,
      "logps/rejected": -280.2191162109375,
      "loss": 0.3039,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.7866594791412354,
      "rewards/margins": 3.567206382751465,
      "rewards/rejected": -5.353865623474121,
      "step": 157
    },
    {
      "epoch": 0.23201174743024963,
      "epsilon_dpo/beta": 0.03198767825961113,
      "epsilon_dpo/beta_margin_grad_mean": -0.11316555738449097,
      "epsilon_dpo/beta_margin_grad_std": 0.1629510372877121,
      "epsilon_dpo/beta_margin_mean": 3.3184967041015625,
      "epsilon_dpo/beta_margin_std": 2.208585739135742,
      "epsilon_dpo/loss_margin_mean": 103.88768768310547,
      "grad_norm": 52.514793395996094,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.03226681426167488,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.7492306773041136e-07,
      "logits/chosen": -2.434067726135254,
      "logits/rejected": -2.5916080474853516,
      "logps/chosen": -77.87710571289062,
      "logps/ref_chosen": -30.166948318481445,
      "logps/ref_rejected": -119.61189270019531,
      "logps/rejected": -271.2097473144531,
      "loss": 0.2905,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.5289018154144287,
      "rewards/margins": 3.3184967041015625,
      "rewards/rejected": -4.84739875793457,
      "step": 158
    },
    {
      "epoch": 0.23348017621145375,
      "epsilon_dpo/beta": 0.03173020854592323,
      "epsilon_dpo/beta_margin_grad_mean": -0.1506548821926117,
      "epsilon_dpo/beta_margin_grad_std": 0.20841404795646667,
      "epsilon_dpo/beta_margin_mean": 3.110417604446411,
      "epsilon_dpo/beta_margin_std": 2.455885410308838,
      "epsilon_dpo/loss_margin_mean": 98.24658203125,
      "grad_norm": 72.99739074707031,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.03198692947626114,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.743599013306165e-07,
      "logits/chosen": -2.645659923553467,
      "logits/rejected": -2.5504307746887207,
      "logps/chosen": -96.61356353759766,
      "logps/ref_chosen": -43.95257568359375,
      "logps/ref_rejected": -96.23291015625,
      "logps/rejected": -247.14047241210938,
      "loss": 0.4372,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.6737987995147705,
      "rewards/margins": 3.110417366027832,
      "rewards/rejected": -4.784216403961182,
      "step": 159
    },
    {
      "epoch": 0.23494860499265785,
      "epsilon_dpo/beta": 0.03151414170861244,
      "epsilon_dpo/beta_margin_grad_mean": -0.17995795607566833,
      "epsilon_dpo/beta_margin_grad_std": 0.22594332695007324,
      "epsilon_dpo/beta_margin_mean": 2.8287112712860107,
      "epsilon_dpo/beta_margin_std": 2.4387524127960205,
      "epsilon_dpo/loss_margin_mean": 90.09088134765625,
      "grad_norm": 85.227294921875,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.03172912821173668,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 4.737908228387656e-07,
      "logits/chosen": -2.586042881011963,
      "logits/rejected": -2.5926198959350586,
      "logps/chosen": -97.44248962402344,
      "logps/ref_chosen": -43.95906066894531,
      "logps/ref_rejected": -104.28060913085938,
      "logps/rejected": -247.85491943359375,
      "loss": 0.5131,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -1.6889050006866455,
      "rewards/margins": 2.82871150970459,
      "rewards/rejected": -4.517616271972656,
      "step": 160
    },
    {
      "epoch": 0.23641703377386197,
      "epsilon_dpo/beta": 0.031249718740582466,
      "epsilon_dpo/beta_margin_grad_mean": -0.14812429249286652,
      "epsilon_dpo/beta_margin_grad_std": 0.21790093183517456,
      "epsilon_dpo/beta_margin_mean": 2.9674715995788574,
      "epsilon_dpo/beta_margin_std": 2.1990880966186523,
      "epsilon_dpo/loss_margin_mean": 95.17440795898438,
      "grad_norm": 69.41145324707031,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.03151248022913933,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.7321584725060594e-07,
      "logits/chosen": -2.514247417449951,
      "logits/rejected": -2.5099971294403076,
      "logps/chosen": -81.56330108642578,
      "logps/ref_chosen": -30.6236572265625,
      "logps/ref_rejected": -88.5151138305664,
      "logps/rejected": -234.62916564941406,
      "loss": 0.4746,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.5938321352005005,
      "rewards/margins": 2.9674715995788574,
      "rewards/rejected": -4.561304092407227,
      "step": 161
    },
    {
      "epoch": 0.23788546255506607,
      "epsilon_dpo/beta": 0.030998021364212036,
      "epsilon_dpo/beta_margin_grad_mean": -0.14223085343837738,
      "epsilon_dpo/beta_margin_grad_std": 0.18433333933353424,
      "epsilon_dpo/beta_margin_mean": 2.894646406173706,
      "epsilon_dpo/beta_margin_std": 2.0557403564453125,
      "epsilon_dpo/loss_margin_mean": 93.58312225341797,
      "grad_norm": 68.47539520263672,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.031248819082975388,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.7263498971727905e-07,
      "logits/chosen": -2.5756616592407227,
      "logits/rejected": -2.518211603164673,
      "logps/chosen": -85.86666870117188,
      "logps/ref_chosen": -37.11417007446289,
      "logps/ref_rejected": -94.5440673828125,
      "logps/rejected": -236.87969970703125,
      "loss": 0.3757,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.514453411102295,
      "rewards/margins": 2.894646167755127,
      "rewards/rejected": -4.409099578857422,
      "step": 162
    },
    {
      "epoch": 0.2393538913362702,
      "epsilon_dpo/beta": 0.030757879838347435,
      "epsilon_dpo/beta_margin_grad_mean": -0.15289483964443207,
      "epsilon_dpo/beta_margin_grad_std": 0.209597647190094,
      "epsilon_dpo/beta_margin_mean": 2.800816535949707,
      "epsilon_dpo/beta_margin_std": 1.9919666051864624,
      "epsilon_dpo/loss_margin_mean": 91.29692077636719,
      "grad_norm": 56.615814208984375,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.03099696896970272,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 4.720482655449212e-07,
      "logits/chosen": -2.5021204948425293,
      "logits/rejected": -2.4259862899780273,
      "logps/chosen": -75.5699234008789,
      "logps/ref_chosen": -30.90496063232422,
      "logps/ref_rejected": -81.61336517333984,
      "logps/rejected": -217.57525634765625,
      "loss": 0.4252,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -1.3745607137680054,
      "rewards/margins": 2.800816535949707,
      "rewards/rejected": -4.175376892089844,
      "step": 163
    },
    {
      "epoch": 0.24082232011747431,
      "epsilon_dpo/beta": 0.03046177327632904,
      "epsilon_dpo/beta_margin_grad_mean": -0.12999504804611206,
      "epsilon_dpo/beta_margin_grad_std": 0.14756108820438385,
      "epsilon_dpo/beta_margin_mean": 2.6089589595794678,
      "epsilon_dpo/beta_margin_std": 1.5135873556137085,
      "epsilon_dpo/loss_margin_mean": 85.68292999267578,
      "grad_norm": 59.15823745727539,
      "kl/avg_steps": 0.96875,
      "kl/beta": 0.03075668215751648,
      "kl/n_epsilon_steps": 0.015625,
      "kl/p_epsilon_steps": 0.984375,
      "learning_rate": 4.714556901942599e-07,
      "logits/chosen": -2.5339465141296387,
      "logits/rejected": -2.4976041316986084,
      "logps/chosen": -84.63601684570312,
      "logps/ref_chosen": -32.731651306152344,
      "logps/ref_rejected": -87.29302215576172,
      "logps/rejected": -224.88031005859375,
      "loss": 0.3178,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -1.5812957286834717,
      "rewards/margins": 2.6089587211608887,
      "rewards/rejected": -4.190255165100098,
      "step": 164
    },
    {
      "epoch": 0.2422907488986784,
      "epsilon_dpo/beta": 0.03020758554339409,
      "epsilon_dpo/beta_margin_grad_mean": -0.17471088469028473,
      "epsilon_dpo/beta_margin_grad_std": 0.19172601401805878,
      "epsilon_dpo/beta_margin_mean": 2.3461360931396484,
      "epsilon_dpo/beta_margin_std": 1.85222589969635,
      "epsilon_dpo/loss_margin_mean": 77.82381439208984,
      "grad_norm": 75.63594055175781,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.030461585149168968,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.708572792802069e-07,
      "logits/chosen": -2.676835536956787,
      "logits/rejected": -2.463170051574707,
      "logps/chosen": -100.64947509765625,
      "logps/ref_chosen": -42.845787048339844,
      "logps/ref_rejected": -77.92173767089844,
      "logps/rejected": -213.5492401123047,
      "loss": 0.4735,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.7484538555145264,
      "rewards/margins": 2.3461360931396484,
      "rewards/rejected": -4.094590187072754,
      "step": 165
    },
    {
      "epoch": 0.24375917767988253,
      "epsilon_dpo/beta": 0.03000204637646675,
      "epsilon_dpo/beta_margin_grad_mean": -0.1740747094154358,
      "epsilon_dpo/beta_margin_grad_std": 0.21660596132278442,
      "epsilon_dpo/beta_margin_mean": 2.6607251167297363,
      "epsilon_dpo/beta_margin_std": 2.3676037788391113,
      "epsilon_dpo/loss_margin_mean": 89.0173110961914,
      "grad_norm": 83.7566909790039,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.030206717550754547,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 4.702530485714461e-07,
      "logits/chosen": -2.4904098510742188,
      "logits/rejected": -2.599632740020752,
      "logps/chosen": -106.1424560546875,
      "logps/ref_chosen": -41.773406982421875,
      "logps/ref_rejected": -105.80018615722656,
      "logps/rejected": -259.1865234375,
      "loss": 0.4996,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -1.9366867542266846,
      "rewards/margins": 2.6607251167297363,
      "rewards/rejected": -4.597411632537842,
      "step": 166
    },
    {
      "epoch": 0.24522760646108663,
      "epsilon_dpo/beta": 0.02974093332886696,
      "epsilon_dpo/beta_margin_grad_mean": -0.12238018214702606,
      "epsilon_dpo/beta_margin_grad_std": 0.16405010223388672,
      "epsilon_dpo/beta_margin_mean": 2.9359490871429443,
      "epsilon_dpo/beta_margin_std": 1.7715035676956177,
      "epsilon_dpo/loss_margin_mean": 98.85689544677734,
      "grad_norm": 57.45083999633789,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.030000463128089905,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.6964301399001877e-07,
      "logits/chosen": -2.4179039001464844,
      "logits/rejected": -2.5366764068603516,
      "logps/chosen": -93.92123413085938,
      "logps/ref_chosen": -32.289695739746094,
      "logps/ref_rejected": -99.84996795654297,
      "logps/rejected": -260.33837890625,
      "loss": 0.3122,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.8347046375274658,
      "rewards/margins": 2.9359493255615234,
      "rewards/rejected": -4.77065372467041,
      "step": 167
    },
    {
      "epoch": 0.24669603524229075,
      "epsilon_dpo/beta": 0.02948295697569847,
      "epsilon_dpo/beta_margin_grad_mean": -0.15910814702510834,
      "epsilon_dpo/beta_margin_grad_std": 0.168760284781456,
      "epsilon_dpo/beta_margin_mean": 2.5459718704223633,
      "epsilon_dpo/beta_margin_std": 1.8775861263275146,
      "epsilon_dpo/loss_margin_mean": 86.47795104980469,
      "grad_norm": 70.27559661865234,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.029740236699581146,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.690271916109034e-07,
      "logits/chosen": -2.536531448364258,
      "logits/rejected": -2.3941898345947266,
      "logps/chosen": -95.75901794433594,
      "logps/ref_chosen": -35.88120651245117,
      "logps/ref_rejected": -80.32093811035156,
      "logps/rejected": -226.67669677734375,
      "loss": 0.4007,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.767836093902588,
      "rewards/margins": 2.5459718704223633,
      "rewards/rejected": -4.313807964324951,
      "step": 168
    },
    {
      "epoch": 0.24816446402349487,
      "epsilon_dpo/beta": 0.029254861176013947,
      "epsilon_dpo/beta_margin_grad_mean": -0.19129043817520142,
      "epsilon_dpo/beta_margin_grad_std": 0.22424982488155365,
      "epsilon_dpo/beta_margin_mean": 2.5321929454803467,
      "epsilon_dpo/beta_margin_std": 2.3219850063323975,
      "epsilon_dpo/loss_margin_mean": 86.81468963623047,
      "grad_norm": 84.80979919433594,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.029482265934348106,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 4.6840559766159235e-07,
      "logits/chosen": -2.558346748352051,
      "logits/rejected": -2.5368800163269043,
      "logps/chosen": -96.68545532226562,
      "logps/ref_chosen": -35.75175476074219,
      "logps/ref_rejected": -88.39851379394531,
      "logps/rejected": -236.14691162109375,
      "loss": 0.5701,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -1.786689281463623,
      "rewards/margins": 2.532193183898926,
      "rewards/rejected": -4.318882465362549,
      "step": 169
    },
    {
      "epoch": 0.24963289280469897,
      "epsilon_dpo/beta": 0.028982365503907204,
      "epsilon_dpo/beta_margin_grad_mean": -0.11600478738546371,
      "epsilon_dpo/beta_margin_grad_std": 0.14576373994350433,
      "epsilon_dpo/beta_margin_mean": 2.932541847229004,
      "epsilon_dpo/beta_margin_std": 1.7402071952819824,
      "epsilon_dpo/loss_margin_mean": 101.24751281738281,
      "grad_norm": 43.760440826416016,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.029253721237182617,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 4.6777824852166437e-07,
      "logits/chosen": -2.3630564212799072,
      "logits/rejected": -2.373150587081909,
      "logps/chosen": -83.35134887695312,
      "logps/ref_chosen": -31.416187286376953,
      "logps/ref_rejected": -82.51651000976562,
      "logps/rejected": -235.6991729736328,
      "loss": 0.2813,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -1.5055475234985352,
      "rewards/margins": 2.932541847229004,
      "rewards/rejected": -4.438089370727539,
      "step": 170
    },
    {
      "epoch": 0.2511013215859031,
      "epsilon_dpo/beta": 0.028740353882312775,
      "epsilon_dpo/beta_margin_grad_mean": -0.13344722986221313,
      "epsilon_dpo/beta_margin_grad_std": 0.16077838838100433,
      "epsilon_dpo/beta_margin_mean": 2.9016919136047363,
      "epsilon_dpo/beta_margin_std": 2.042875051498413,
      "epsilon_dpo/loss_margin_mean": 101.12266540527344,
      "grad_norm": 72.2908706665039,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.028982015326619148,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.6714516072235273e-07,
      "logits/chosen": -2.617936611175537,
      "logits/rejected": -2.546191692352295,
      "logps/chosen": -120.83259582519531,
      "logps/ref_chosen": -55.1216926574707,
      "logps/ref_rejected": -117.03608703613281,
      "logps/rejected": -283.8696594238281,
      "loss": 0.3355,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.8903237581253052,
      "rewards/margins": 2.9016919136047363,
      "rewards/rejected": -4.792015552520752,
      "step": 171
    },
    {
      "epoch": 0.2525697503671072,
      "epsilon_dpo/beta": 0.028499886393547058,
      "epsilon_dpo/beta_margin_grad_mean": -0.12330116331577301,
      "epsilon_dpo/beta_margin_grad_std": 0.18182381987571716,
      "epsilon_dpo/beta_margin_mean": 2.9543943405151367,
      "epsilon_dpo/beta_margin_std": 1.9407035112380981,
      "epsilon_dpo/loss_margin_mean": 103.86387634277344,
      "grad_norm": 62.24739456176758,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.028739525005221367,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.6650635094610966e-07,
      "logits/chosen": -2.4678688049316406,
      "logits/rejected": -2.340376853942871,
      "logps/chosen": -103.77349090576172,
      "logps/ref_chosen": -42.184104919433594,
      "logps/ref_rejected": -90.7464599609375,
      "logps/rejected": -256.19970703125,
      "loss": 0.3457,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.7575478553771973,
      "rewards/margins": 2.9543943405151367,
      "rewards/rejected": -4.711941719055176,
      "step": 172
    },
    {
      "epoch": 0.2540381791483113,
      "epsilon_dpo/beta": 0.028279244899749756,
      "epsilon_dpo/beta_margin_grad_mean": -0.1761482208967209,
      "epsilon_dpo/beta_margin_grad_std": 0.1999354213476181,
      "epsilon_dpo/beta_margin_mean": 2.418248414993286,
      "epsilon_dpo/beta_margin_std": 1.9150099754333496,
      "epsilon_dpo/loss_margin_mean": 85.72386932373047,
      "grad_norm": 93.45008850097656,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.028499064967036247,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 4.6586183602616687e-07,
      "logits/chosen": -2.362546443939209,
      "logits/rejected": -2.276439666748047,
      "logps/chosen": -126.22164154052734,
      "logps/ref_chosen": -46.40528869628906,
      "logps/ref_rejected": -84.17361450195312,
      "logps/rejected": -249.71383666992188,
      "loss": 0.4831,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -2.261880874633789,
      "rewards/margins": 2.418248414993286,
      "rewards/rejected": -4.680129051208496,
      "step": 173
    },
    {
      "epoch": 0.2555066079295154,
      "epsilon_dpo/beta": 0.028042349964380264,
      "epsilon_dpo/beta_margin_grad_mean": -0.15671639144420624,
      "epsilon_dpo/beta_margin_grad_std": 0.1755547821521759,
      "epsilon_dpo/beta_margin_mean": 2.786201238632202,
      "epsilon_dpo/beta_margin_std": 2.2568979263305664,
      "epsilon_dpo/loss_margin_mean": 99.53160858154297,
      "grad_norm": 68.29830932617188,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.02827814221382141,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.652116329460919e-07,
      "logits/chosen": -2.276648759841919,
      "logits/rejected": -2.400620460510254,
      "logps/chosen": -106.728515625,
      "logps/ref_chosen": -33.12921142578125,
      "logps/ref_rejected": -107.51828002929688,
      "logps/rejected": -280.6492004394531,
      "loss": 0.4027,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -2.0658488273620605,
      "rewards/margins": 2.786201238632202,
      "rewards/rejected": -4.852049827575684,
      "step": 174
    },
    {
      "epoch": 0.25697503671071953,
      "epsilon_dpo/beta": 0.027807721868157387,
      "epsilon_dpo/beta_margin_grad_mean": -0.12103898078203201,
      "epsilon_dpo/beta_margin_grad_std": 0.18520337343215942,
      "epsilon_dpo/beta_margin_mean": 3.309805393218994,
      "epsilon_dpo/beta_margin_std": 2.263869524002075,
      "epsilon_dpo/loss_margin_mean": 119.24435424804688,
      "grad_norm": 80.84042358398438,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.028041541576385498,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.645557588393406e-07,
      "logits/chosen": -2.220841884613037,
      "logits/rejected": -2.1725313663482666,
      "logps/chosen": -105.72348022460938,
      "logps/ref_chosen": -31.185977935791016,
      "logps/ref_rejected": -96.78857421875,
      "logps/rejected": -290.5704345703125,
      "loss": 0.3301,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -2.0765433311462402,
      "rewards/margins": 3.309805393218994,
      "rewards/rejected": -5.386348724365234,
      "step": 175
    },
    {
      "epoch": 0.25844346549192365,
      "epsilon_dpo/beta": 0.02758374810218811,
      "epsilon_dpo/beta_margin_grad_mean": -0.1377795934677124,
      "epsilon_dpo/beta_margin_grad_std": 0.17307686805725098,
      "epsilon_dpo/beta_margin_mean": 2.979292154312134,
      "epsilon_dpo/beta_margin_std": 2.1362128257751465,
      "epsilon_dpo/loss_margin_mean": 108.23030853271484,
      "grad_norm": 60.13013458251953,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.027806920930743217,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.638942309888058e-07,
      "logits/chosen": -2.2297158241271973,
      "logits/rejected": -2.2983927726745605,
      "logps/chosen": -116.0774917602539,
      "logps/ref_chosen": -34.4107666015625,
      "logps/ref_rejected": -103.36322021484375,
      "logps/rejected": -293.26025390625,
      "loss": 0.3514,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -2.254232168197632,
      "rewards/margins": 2.979292392730713,
      "rewards/rejected": -5.233524322509766,
      "step": 176
    },
    {
      "epoch": 0.2599118942731278,
      "epsilon_dpo/beta": 0.027404537424445152,
      "epsilon_dpo/beta_margin_grad_mean": -0.17347557842731476,
      "epsilon_dpo/beta_margin_grad_std": 0.22418558597564697,
      "epsilon_dpo/beta_margin_mean": 2.8900716304779053,
      "epsilon_dpo/beta_margin_std": 2.369600296020508,
      "epsilon_dpo/loss_margin_mean": 105.86678314208984,
      "grad_norm": 88.59068298339844,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.027582809329032898,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 4.6322706682636137e-07,
      "logits/chosen": -2.3562376499176025,
      "logits/rejected": -2.3722541332244873,
      "logps/chosen": -122.57586669921875,
      "logps/ref_chosen": -42.30748748779297,
      "logps/ref_rejected": -101.40332794189453,
      "logps/rejected": -287.5384826660156,
      "loss": 0.4875,
      "rewards/accuracies": 0.828125,
      "rewards/chosen": -2.2037370204925537,
      "rewards/margins": 2.8900716304779053,
      "rewards/rejected": -5.093808650970459,
      "step": 177
    },
    {
      "epoch": 0.26138032305433184,
      "epsilon_dpo/beta": 0.027148790657520294,
      "epsilon_dpo/beta_margin_grad_mean": -0.11924877017736435,
      "epsilon_dpo/beta_margin_grad_std": 0.15244512259960175,
      "epsilon_dpo/beta_margin_mean": 3.4991955757141113,
      "epsilon_dpo/beta_margin_std": 2.5016074180603027,
      "epsilon_dpo/loss_margin_mean": 128.963623046875,
      "grad_norm": 60.96648025512695,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.027402978390455246,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 4.6255428393240354e-07,
      "logits/chosen": -2.2090775966644287,
      "logits/rejected": -2.2421045303344727,
      "logps/chosen": -107.5852279663086,
      "logps/ref_chosen": -32.16440963745117,
      "logps/ref_rejected": -111.8971939086914,
      "logps/rejected": -316.2816467285156,
      "loss": 0.2914,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -2.04891300201416,
      "rewards/margins": 3.4991955757141113,
      "rewards/rejected": -5.5481085777282715,
      "step": 178
    },
    {
      "epoch": 0.26284875183553597,
      "epsilon_dpo/beta": 0.026943257078528404,
      "epsilon_dpo/beta_margin_grad_mean": -0.16198381781578064,
      "epsilon_dpo/beta_margin_grad_std": 0.22395355999469757,
      "epsilon_dpo/beta_margin_mean": 2.953178882598877,
      "epsilon_dpo/beta_margin_std": 2.3476455211639404,
      "epsilon_dpo/loss_margin_mean": 109.93001556396484,
      "grad_norm": 76.76069641113281,
      "kl/avg_steps": 0.765625,
      "kl/beta": 0.02714846096932888,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 4.6187590003538724e-07,
      "logits/chosen": -2.2808642387390137,
      "logits/rejected": -2.315495252609253,
      "logps/chosen": -120.60657501220703,
      "logps/ref_chosen": -41.662925720214844,
      "logps/ref_rejected": -96.57734680175781,
      "logps/rejected": -285.45098876953125,
      "loss": 0.4834,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -2.1310267448425293,
      "rewards/margins": 2.953178882598877,
      "rewards/rejected": -5.084205627441406,
      "step": 179
    },
    {
      "epoch": 0.2643171806167401,
      "epsilon_dpo/beta": 0.02671753242611885,
      "epsilon_dpo/beta_margin_grad_mean": -0.13764944672584534,
      "epsilon_dpo/beta_margin_grad_std": 0.20955896377563477,
      "epsilon_dpo/beta_margin_mean": 3.045346975326538,
      "epsilon_dpo/beta_margin_std": 2.1593165397644043,
      "epsilon_dpo/loss_margin_mean": 114.23480987548828,
      "grad_norm": 90.68710327148438,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.026942186057567596,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.611919330113591e-07,
      "logits/chosen": -2.333063840866089,
      "logits/rejected": -2.4003429412841797,
      "logps/chosen": -128.1096649169922,
      "logps/ref_chosen": -45.704586029052734,
      "logps/ref_rejected": -106.64442443847656,
      "logps/rejected": -303.2843017578125,
      "loss": 0.4102,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -2.2053043842315674,
      "rewards/margins": 3.045346975326538,
      "rewards/rejected": -5.2506513595581055,
      "step": 180
    },
    {
      "epoch": 0.2657856093979442,
      "epsilon_dpo/beta": 0.026519039645791054,
      "epsilon_dpo/beta_margin_grad_mean": -0.17692674696445465,
      "epsilon_dpo/beta_margin_grad_std": 0.21322111785411835,
      "epsilon_dpo/beta_margin_mean": 2.4334042072296143,
      "epsilon_dpo/beta_margin_std": 2.0837597846984863,
      "epsilon_dpo/loss_margin_mean": 92.05476379394531,
      "grad_norm": 90.5920181274414,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.02671676315367222,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 4.605024008834863e-07,
      "logits/chosen": -2.3339180946350098,
      "logits/rejected": -2.1919121742248535,
      "logps/chosen": -121.45870971679688,
      "logps/ref_chosen": -42.68443298339844,
      "logps/ref_rejected": -66.97550201416016,
      "logps/rejected": -237.80453491210938,
      "loss": 0.5481,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -2.093709945678711,
      "rewards/margins": 2.4334044456481934,
      "rewards/rejected": -4.527114391326904,
      "step": 181
    },
    {
      "epoch": 0.26725403817914833,
      "epsilon_dpo/beta": 0.026296764612197876,
      "epsilon_dpo/beta_margin_grad_mean": -0.14079764485359192,
      "epsilon_dpo/beta_margin_grad_std": 0.20722897350788116,
      "epsilon_dpo/beta_margin_mean": 3.104597330093384,
      "epsilon_dpo/beta_margin_std": 2.1538186073303223,
      "epsilon_dpo/loss_margin_mean": 118.28819274902344,
      "grad_norm": 73.83556365966797,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.026517879217863083,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.598073218215817e-07,
      "logits/chosen": -2.208120822906494,
      "logits/rejected": -2.3168540000915527,
      "logps/chosen": -95.87370300292969,
      "logps/ref_chosen": -22.746728897094727,
      "logps/ref_rejected": -96.14221954345703,
      "logps/rejected": -287.557373046875,
      "loss": 0.402,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.9256913661956787,
      "rewards/margins": 3.104597568511963,
      "rewards/rejected": -5.0302886962890625,
      "step": 182
    },
    {
      "epoch": 0.2687224669603524,
      "epsilon_dpo/beta": 0.02606852352619171,
      "epsilon_dpo/beta_margin_grad_mean": -0.12661683559417725,
      "epsilon_dpo/beta_margin_grad_std": 0.14383679628372192,
      "epsilon_dpo/beta_margin_mean": 2.813413619995117,
      "epsilon_dpo/beta_margin_std": 1.8080170154571533,
      "epsilon_dpo/loss_margin_mean": 108.05697631835938,
      "grad_norm": 49.043949127197266,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.02629600651562214,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.5910671414162484e-07,
      "logits/chosen": -2.3619985580444336,
      "logits/rejected": -2.227496862411499,
      "logps/chosen": -111.11225891113281,
      "logps/ref_chosen": -37.3929557800293,
      "logps/ref_rejected": -82.01314544677734,
      "logps/rejected": -263.7894287109375,
      "loss": 0.305,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -1.921753168106079,
      "rewards/margins": 2.813413619995117,
      "rewards/rejected": -4.735166549682617,
      "step": 183
    },
    {
      "epoch": 0.2701908957415565,
      "epsilon_dpo/beta": 0.02590757980942726,
      "epsilon_dpo/beta_margin_grad_mean": -0.22445106506347656,
      "epsilon_dpo/beta_margin_grad_std": 0.2272690087556839,
      "epsilon_dpo/beta_margin_mean": 2.006890058517456,
      "epsilon_dpo/beta_margin_std": 1.832121729850769,
      "epsilon_dpo/loss_margin_mean": 77.82298278808594,
      "grad_norm": 91.00837707519531,
      "kl/avg_steps": 0.625,
      "kl/beta": 0.026067912578582764,
      "kl/n_epsilon_steps": 0.1875,
      "kl/p_epsilon_steps": 0.8125,
      "learning_rate": 4.5840059630527985e-07,
      "logits/chosen": -2.4576053619384766,
      "logits/rejected": -2.333487033843994,
      "logps/chosen": -125.68778991699219,
      "logps/ref_chosen": -43.53949737548828,
      "logps/ref_rejected": -81.33880615234375,
      "logps/rejected": -241.31008911132812,
      "loss": 0.6265,
      "rewards/accuracies": 0.828125,
      "rewards/chosen": -2.1323318481445312,
      "rewards/margins": 2.006890058517456,
      "rewards/rejected": -4.139222145080566,
      "step": 184
    },
    {
      "epoch": 0.27165932452276065,
      "epsilon_dpo/beta": 0.0256899856030941,
      "epsilon_dpo/beta_margin_grad_mean": -0.16735970973968506,
      "epsilon_dpo/beta_margin_grad_std": 0.17599676549434662,
      "epsilon_dpo/beta_margin_mean": 2.3658905029296875,
      "epsilon_dpo/beta_margin_std": 1.7580143213272095,
      "epsilon_dpo/loss_margin_mean": 92.25860595703125,
      "grad_norm": 67.63098907470703,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.025905998423695564,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.5768898691940836e-07,
      "logits/chosen": -2.360046863555908,
      "logits/rejected": -2.2244629859924316,
      "logps/chosen": -111.3865737915039,
      "logps/ref_chosen": -37.372806549072266,
      "logps/ref_rejected": -78.53436279296875,
      "logps/rejected": -244.80673217773438,
      "loss": 0.4255,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.9024994373321533,
      "rewards/margins": 2.3658905029296875,
      "rewards/rejected": -4.268389701843262,
      "step": 185
    },
    {
      "epoch": 0.27312775330396477,
      "epsilon_dpo/beta": 0.02545095421373844,
      "epsilon_dpo/beta_margin_grad_mean": -0.09410841017961502,
      "epsilon_dpo/beta_margin_grad_std": 0.11750077456235886,
      "epsilon_dpo/beta_margin_mean": 3.1497642993927,
      "epsilon_dpo/beta_margin_std": 1.6914327144622803,
      "epsilon_dpo/loss_margin_mean": 123.8324966430664,
      "grad_norm": 39.774940490722656,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.025689246132969856,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 4.5697190473557947e-07,
      "logits/chosen": -2.5002646446228027,
      "logits/rejected": -2.3675756454467773,
      "logps/chosen": -100.51115417480469,
      "logps/ref_chosen": -41.56720733642578,
      "logps/ref_rejected": -93.58636474609375,
      "logps/rejected": -276.3628234863281,
      "loss": 0.2184,
      "rewards/accuracies": 1.0,
      "rewards/chosen": -1.5010137557983398,
      "rewards/margins": 3.1497642993927,
      "rewards/rejected": -4.650777816772461,
      "step": 186
    },
    {
      "epoch": 0.2745961820851689,
      "epsilon_dpo/beta": 0.025262294337153435,
      "epsilon_dpo/beta_margin_grad_mean": -0.18915115296840668,
      "epsilon_dpo/beta_margin_grad_std": 0.22492022812366486,
      "epsilon_dpo/beta_margin_mean": 2.369025468826294,
      "epsilon_dpo/beta_margin_std": 1.9370827674865723,
      "epsilon_dpo/loss_margin_mean": 94.07389831542969,
      "grad_norm": 85.82039642333984,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.02545064687728882,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 4.5624936864957555e-07,
      "logits/chosen": -2.322283983230591,
      "logits/rejected": -2.289004325866699,
      "logps/chosen": -108.8873291015625,
      "logps/ref_chosen": -35.3756217956543,
      "logps/ref_rejected": -86.07025146484375,
      "logps/rejected": -253.65586853027344,
      "loss": 0.5347,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -1.861724615097046,
      "rewards/margins": 2.369025230407715,
      "rewards/rejected": -4.23075008392334,
      "step": 187
    },
    {
      "epoch": 0.27606461086637296,
      "epsilon_dpo/beta": 0.025042656809091568,
      "epsilon_dpo/beta_margin_grad_mean": -0.15640155971050262,
      "epsilon_dpo/beta_margin_grad_std": 0.18032178282737732,
      "epsilon_dpo/beta_margin_mean": 2.593634843826294,
      "epsilon_dpo/beta_margin_std": 1.8406410217285156,
      "epsilon_dpo/loss_margin_mean": 103.7129898071289,
      "grad_norm": 62.04313659667969,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.02526118792593479,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.5552139770089454e-07,
      "logits/chosen": -2.326641321182251,
      "logits/rejected": -2.3683671951293945,
      "logps/chosen": -104.10310363769531,
      "logps/ref_chosen": -34.50104522705078,
      "logps/ref_rejected": -96.81013488769531,
      "logps/rejected": -270.12518310546875,
      "loss": 0.4002,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.7442595958709717,
      "rewards/margins": 2.593634843826294,
      "rewards/rejected": -4.337894439697266,
      "step": 188
    },
    {
      "epoch": 0.2775330396475771,
      "epsilon_dpo/beta": 0.024817608296871185,
      "epsilon_dpo/beta_margin_grad_mean": -0.13081274926662445,
      "epsilon_dpo/beta_margin_grad_std": 0.16288872063159943,
      "epsilon_dpo/beta_margin_mean": 2.6774120330810547,
      "epsilon_dpo/beta_margin_std": 1.7680622339248657,
      "epsilon_dpo/loss_margin_mean": 108.02799224853516,
      "grad_norm": 57.21524429321289,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.02504207007586956,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 4.5478801107224794e-07,
      "logits/chosen": -2.281872272491455,
      "logits/rejected": -2.1657588481903076,
      "logps/chosen": -94.84893798828125,
      "logps/ref_chosen": -30.625194549560547,
      "logps/ref_rejected": -78.02359008789062,
      "logps/rejected": -250.27532958984375,
      "loss": 0.3608,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -1.5961487293243408,
      "rewards/margins": 2.6774120330810547,
      "rewards/rejected": -4.273560523986816,
      "step": 189
    },
    {
      "epoch": 0.2790014684287812,
      "epsilon_dpo/beta": 0.024594716727733612,
      "epsilon_dpo/beta_margin_grad_mean": -0.14998801052570343,
      "epsilon_dpo/beta_margin_grad_std": 0.1647987961769104,
      "epsilon_dpo/beta_margin_mean": 2.706925392150879,
      "epsilon_dpo/beta_margin_std": 1.9622080326080322,
      "epsilon_dpo/loss_margin_mean": 110.1700210571289,
      "grad_norm": 55.00584030151367,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.02481716312468052,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 4.5404922808905543e-07,
      "logits/chosen": -2.403958797454834,
      "logits/rejected": -2.4575366973876953,
      "logps/chosen": -105.57707214355469,
      "logps/ref_chosen": -41.5223388671875,
      "logps/ref_rejected": -109.71112823486328,
      "logps/rejected": -283.9358825683594,
      "loss": 0.3711,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.5758315324783325,
      "rewards/margins": 2.706925392150879,
      "rewards/rejected": -4.282756805419922,
      "step": 190
    },
    {
      "epoch": 0.28046989720998533,
      "epsilon_dpo/beta": 0.024389203637838364,
      "epsilon_dpo/beta_margin_grad_mean": -0.1167060062289238,
      "epsilon_dpo/beta_margin_grad_std": 0.15317706763744354,
      "epsilon_dpo/beta_margin_mean": 3.051992416381836,
      "epsilon_dpo/beta_margin_std": 1.9232109785079956,
      "epsilon_dpo/loss_margin_mean": 125.34806060791016,
      "grad_norm": 55.07308578491211,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.024594279006123543,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.5330506821893565e-07,
      "logits/chosen": -2.455369234085083,
      "logits/rejected": -2.3924498558044434,
      "logps/chosen": -107.12852478027344,
      "logps/ref_chosen": -43.46269989013672,
      "logps/ref_rejected": -115.54437255859375,
      "logps/rejected": -304.5582580566406,
      "loss": 0.2901,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.554311990737915,
      "rewards/margins": 3.051992416381836,
      "rewards/rejected": -4.606304168701172,
      "step": 191
    },
    {
      "epoch": 0.28193832599118945,
      "epsilon_dpo/beta": 0.024181291460990906,
      "epsilon_dpo/beta_margin_grad_mean": -0.16211238503456116,
      "epsilon_dpo/beta_margin_grad_std": 0.18894152343273163,
      "epsilon_dpo/beta_margin_mean": 2.575836181640625,
      "epsilon_dpo/beta_margin_std": 1.9563552141189575,
      "epsilon_dpo/loss_margin_mean": 106.70730590820312,
      "grad_norm": 76.85231018066406,
      "kl/avg_steps": 0.859375,
      "kl/beta": 0.024388499557971954,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.5255555107119336e-07,
      "logits/chosen": -2.2935123443603516,
      "logits/rejected": -2.236546754837036,
      "logps/chosen": -124.95817565917969,
      "logps/ref_chosen": -45.30467224121094,
      "logps/ref_rejected": -109.86914825439453,
      "logps/rejected": -296.22998046875,
      "loss": 0.4295,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.9286646842956543,
      "rewards/margins": 2.575836181640625,
      "rewards/rejected": -4.504500865936279,
      "step": 192
    },
    {
      "epoch": 0.2834067547723935,
      "epsilon_dpo/beta": 0.02397906966507435,
      "epsilon_dpo/beta_margin_grad_mean": -0.1852301061153412,
      "epsilon_dpo/beta_margin_grad_std": 0.20167653262615204,
      "epsilon_dpo/beta_margin_mean": 2.189572334289551,
      "epsilon_dpo/beta_margin_std": 1.6879116296768188,
      "epsilon_dpo/loss_margin_mean": 91.50328063964844,
      "grad_norm": 68.04452514648438,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.0241806972771883,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.5180069639630236e-07,
      "logits/chosen": -2.299919605255127,
      "logits/rejected": -2.2096662521362305,
      "logps/chosen": -114.80885314941406,
      "logps/ref_chosen": -38.554039001464844,
      "logps/ref_rejected": -81.17373657226562,
      "logps/rejected": -248.93182373046875,
      "loss": 0.5015,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -1.82987380027771,
      "rewards/margins": 2.189572334289551,
      "rewards/rejected": -4.01944637298584,
      "step": 193
    },
    {
      "epoch": 0.28487518355359764,
      "epsilon_dpo/beta": 0.023755960166454315,
      "epsilon_dpo/beta_margin_grad_mean": -0.15306001901626587,
      "epsilon_dpo/beta_margin_grad_std": 0.14770975708961487,
      "epsilon_dpo/beta_margin_mean": 2.215096950531006,
      "epsilon_dpo/beta_margin_std": 1.2999720573425293,
      "epsilon_dpo/loss_margin_mean": 93.31468200683594,
      "grad_norm": 58.625755310058594,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.023978380486369133,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 4.510405240853854e-07,
      "logits/chosen": -2.02514386177063,
      "logits/rejected": -1.9236395359039307,
      "logps/chosen": -111.03370666503906,
      "logps/ref_chosen": -26.635162353515625,
      "logps/ref_rejected": -64.14307403564453,
      "logps/rejected": -241.85629272460938,
      "loss": 0.3712,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -2.005694627761841,
      "rewards/margins": 2.215096950531006,
      "rewards/rejected": -4.220791816711426,
      "step": 194
    },
    {
      "epoch": 0.28634361233480177,
      "epsilon_dpo/beta": 0.023535314947366714,
      "epsilon_dpo/beta_margin_grad_mean": -0.128713458776474,
      "epsilon_dpo/beta_margin_grad_std": 0.14728419482707977,
      "epsilon_dpo/beta_margin_mean": 2.749429702758789,
      "epsilon_dpo/beta_margin_std": 1.724953055381775,
      "epsilon_dpo/loss_margin_mean": 116.91402435302734,
      "grad_norm": 68.01368713378906,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.023755669593811035,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 4.5027505416968985e-07,
      "logits/chosen": -2.1741151809692383,
      "logits/rejected": -2.245378255844116,
      "logps/chosen": -107.69198608398438,
      "logps/ref_chosen": -33.191192626953125,
      "logps/ref_rejected": -103.19746398925781,
      "logps/rejected": -294.6122741699219,
      "loss": 0.316,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -1.754541277885437,
      "rewards/margins": 2.749429702758789,
      "rewards/rejected": -4.503971099853516,
      "step": 195
    },
    {
      "epoch": 0.2878120411160059,
      "epsilon_dpo/beta": 0.023338787257671356,
      "epsilon_dpo/beta_margin_grad_mean": -0.13086631894111633,
      "epsilon_dpo/beta_margin_grad_std": 0.16106443107128143,
      "epsilon_dpo/beta_margin_mean": 2.784287929534912,
      "epsilon_dpo/beta_margin_std": 1.7888661623001099,
      "epsilon_dpo/loss_margin_mean": 119.4956283569336,
      "grad_norm": 46.899410247802734,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.023535029962658882,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.495043068200599e-07,
      "logits/chosen": -2.1250100135803223,
      "logits/rejected": -1.9867154359817505,
      "logps/chosen": -101.93032836914062,
      "logps/ref_chosen": -30.988731384277344,
      "logps/ref_rejected": -82.06416320800781,
      "logps/rejected": -272.50140380859375,
      "loss": 0.3278,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.6580393314361572,
      "rewards/margins": 2.784287929534912,
      "rewards/rejected": -4.442327499389648,
      "step": 196
    },
    {
      "epoch": 0.28928046989721,
      "epsilon_dpo/beta": 0.02313622087240219,
      "epsilon_dpo/beta_margin_grad_mean": -0.15594223141670227,
      "epsilon_dpo/beta_margin_grad_std": 0.17379499971866608,
      "epsilon_dpo/beta_margin_mean": 2.3834075927734375,
      "epsilon_dpo/beta_margin_std": 1.601683259010315,
      "epsilon_dpo/loss_margin_mean": 103.17386627197266,
      "grad_norm": 68.2065200805664,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.023338114842772484,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.4872830234640493e-07,
      "logits/chosen": -2.200084686279297,
      "logits/rejected": -2.1047005653381348,
      "logps/chosen": -115.05329895019531,
      "logps/ref_chosen": -38.329471588134766,
      "logps/ref_rejected": -81.55877685546875,
      "logps/rejected": -261.45648193359375,
      "loss": 0.4039,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.7775802612304688,
      "rewards/margins": 2.3834075927734375,
      "rewards/rejected": -4.160987854003906,
      "step": 197
    },
    {
      "epoch": 0.2907488986784141,
      "epsilon_dpo/beta": 0.0229355338960886,
      "epsilon_dpo/beta_margin_grad_mean": -0.13343603909015656,
      "epsilon_dpo/beta_margin_grad_std": 0.17914170026779175,
      "epsilon_dpo/beta_margin_mean": 2.9484434127807617,
      "epsilon_dpo/beta_margin_std": 2.0632452964782715,
      "epsilon_dpo/loss_margin_mean": 128.7584228515625,
      "grad_norm": 67.33406829833984,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.023135676980018616,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.479470611971645e-07,
      "logits/chosen": -2.1627941131591797,
      "logits/rejected": -2.1062378883361816,
      "logps/chosen": -111.89299011230469,
      "logps/ref_chosen": -33.095436096191406,
      "logps/ref_rejected": -102.81793212890625,
      "logps/rejected": -310.3739013671875,
      "loss": 0.3647,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.8085954189300537,
      "rewards/margins": 2.9484434127807617,
      "rewards/rejected": -4.7570390701293945,
      "step": 198
    },
    {
      "epoch": 0.2922173274596182,
      "epsilon_dpo/beta": 0.0227294210344553,
      "epsilon_dpo/beta_margin_grad_mean": -0.13506709039211273,
      "epsilon_dpo/beta_margin_grad_std": 0.15224339067935944,
      "epsilon_dpo/beta_margin_mean": 2.6718122959136963,
      "epsilon_dpo/beta_margin_std": 1.751692295074463,
      "epsilon_dpo/loss_margin_mean": 117.67414855957031,
      "grad_norm": 61.87571334838867,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.022934995591640472,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 4.471606039587695e-07,
      "logits/chosen": -2.119852066040039,
      "logits/rejected": -1.9955350160598755,
      "logps/chosen": -117.20360565185547,
      "logps/ref_chosen": -34.307518005371094,
      "logps/ref_rejected": -90.8328857421875,
      "logps/rejected": -291.4031066894531,
      "loss": 0.3347,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.884255290031433,
      "rewards/margins": 2.6718125343322754,
      "rewards/rejected": -4.55606746673584,
      "step": 199
    },
    {
      "epoch": 0.2936857562408223,
      "epsilon_dpo/beta": 0.022539492696523666,
      "epsilon_dpo/beta_margin_grad_mean": -0.1900298297405243,
      "epsilon_dpo/beta_margin_grad_std": 0.20542176067829132,
      "epsilon_dpo/beta_margin_mean": 2.5110652446746826,
      "epsilon_dpo/beta_margin_std": 2.381103754043579,
      "epsilon_dpo/loss_margin_mean": 111.65966033935547,
      "grad_norm": 93.91866302490234,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.022729014977812767,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.4636895135509966e-07,
      "logits/chosen": -2.0332164764404297,
      "logits/rejected": -1.9690152406692505,
      "logps/chosen": -130.70755004882812,
      "logps/ref_chosen": -43.81787872314453,
      "logps/ref_rejected": -86.25389099121094,
      "logps/rejected": -284.80322265625,
      "loss": 0.559,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.9615219831466675,
      "rewards/margins": 2.5110652446746826,
      "rewards/rejected": -4.4725871086120605,
      "step": 200
    },
    {
      "epoch": 0.2936857562408223,
      "eval_epsilon_dpo/beta": 0.022413820028305054,
      "eval_epsilon_dpo/beta_margin_grad_mean": -0.26273271441459656,
      "eval_epsilon_dpo/beta_margin_grad_std": 0.2419932633638382,
      "eval_epsilon_dpo/beta_margin_mean": 1.7521748542785645,
      "eval_epsilon_dpo/beta_margin_std": 1.9999192953109741,
      "eval_epsilon_dpo/loss_margin_mean": 78.65644836425781,
      "eval_kl/n_epsilon_steps": 0.21746575832366943,
      "eval_kl/p_epsilon_steps": 0.7821061611175537,
      "eval_logits/chosen": -2.2068309783935547,
      "eval_logits/rejected": -2.105496644973755,
      "eval_logps/chosen": -175.93331909179688,
      "eval_logps/ref_chosen": -68.29110717773438,
      "eval_logps/ref_rejected": -92.08038330078125,
      "eval_logps/rejected": -278.3790588378906,
      "eval_loss": 0.4013729393482208,
      "eval_rewards/accuracies": 0.798373281955719,
      "eval_rewards/chosen": -2.4182353019714355,
      "eval_rewards/margins": 1.7521746158599854,
      "eval_rewards/rejected": -4.170409679412842,
      "eval_runtime": 38.3575,
      "eval_samples_per_second": 60.979,
      "eval_steps_per_second": 1.929,
      "step": 200
    },
    {
      "epoch": 0.29515418502202645,
      "epsilon_dpo/beta": 0.02235795184969902,
      "epsilon_dpo/beta_margin_grad_mean": -0.1633475422859192,
      "epsilon_dpo/beta_margin_grad_std": 0.2009790688753128,
      "epsilon_dpo/beta_margin_mean": 2.613624095916748,
      "epsilon_dpo/beta_margin_std": 1.999254584312439,
      "epsilon_dpo/loss_margin_mean": 117.17089080810547,
      "grad_norm": 77.21038055419922,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.022538842633366585,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.455721242469372e-07,
      "logits/chosen": -2.2331690788269043,
      "logits/rejected": -2.180114269256592,
      "logps/chosen": -133.12782287597656,
      "logps/ref_chosen": -55.676116943359375,
      "logps/ref_rejected": -121.86392974853516,
      "logps/rejected": -316.4865417480469,
      "loss": 0.4443,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.7362768650054932,
      "rewards/margins": 2.613624095916748,
      "rewards/rejected": -4.34990119934082,
      "step": 201
    },
    {
      "epoch": 0.2966226138032305,
      "epsilon_dpo/beta": 0.022170767188072205,
      "epsilon_dpo/beta_margin_grad_mean": -0.16092374920845032,
      "epsilon_dpo/beta_margin_grad_std": 0.16366447508335114,
      "epsilon_dpo/beta_margin_mean": 2.40342116355896,
      "epsilon_dpo/beta_margin_std": 1.810045599937439,
      "epsilon_dpo/loss_margin_mean": 108.59467315673828,
      "grad_norm": 63.114845275878906,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.022357190027832985,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.4477014363141755e-07,
      "logits/chosen": -1.9032566547393799,
      "logits/rejected": -2.0256400108337402,
      "logps/chosen": -113.52767181396484,
      "logps/ref_chosen": -30.73172378540039,
      "logps/ref_rejected": -93.48927307128906,
      "logps/rejected": -284.8798828125,
      "loss": 0.4009,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.8372435569763184,
      "rewards/margins": 2.403420925140381,
      "rewards/rejected": -4.240664482116699,
      "step": 202
    },
    {
      "epoch": 0.29809104258443464,
      "epsilon_dpo/beta": 0.021971412003040314,
      "epsilon_dpo/beta_margin_grad_mean": -0.12936589121818542,
      "epsilon_dpo/beta_margin_grad_std": 0.16617560386657715,
      "epsilon_dpo/beta_margin_mean": 2.778576612472534,
      "epsilon_dpo/beta_margin_std": 1.7826684713363647,
      "epsilon_dpo/loss_margin_mean": 126.62322998046875,
      "grad_norm": 45.17288589477539,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.022170130163431168,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 4.439630306414758e-07,
      "logits/chosen": -2.110769748687744,
      "logits/rejected": -2.13863468170166,
      "logps/chosen": -107.68612670898438,
      "logps/ref_chosen": -38.8436393737793,
      "logps/ref_rejected": -92.13667297363281,
      "logps/rejected": -287.6023864746094,
      "loss": 0.3417,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.5144764184951782,
      "rewards/margins": 2.778576612472534,
      "rewards/rejected": -4.293052673339844,
      "step": 203
    },
    {
      "epoch": 0.29955947136563876,
      "epsilon_dpo/beta": 0.02179468236863613,
      "epsilon_dpo/beta_margin_grad_mean": -0.14814503490924835,
      "epsilon_dpo/beta_margin_grad_std": 0.18509677052497864,
      "epsilon_dpo/beta_margin_mean": 2.5697386264801025,
      "epsilon_dpo/beta_margin_std": 1.7625229358673096,
      "epsilon_dpo/loss_margin_mean": 118.17318725585938,
      "grad_norm": 67.1108169555664,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.021971017122268677,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.431508065452897e-07,
      "logits/chosen": -2.281491756439209,
      "logits/rejected": -2.114370822906494,
      "logps/chosen": -135.9166259765625,
      "logps/ref_chosen": -55.713932037353516,
      "logps/ref_rejected": -93.70796203613281,
      "logps/rejected": -292.0838623046875,
      "loss": 0.3937,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.7504637241363525,
      "rewards/margins": 2.5697386264801025,
      "rewards/rejected": -4.320202350616455,
      "step": 204
    },
    {
      "epoch": 0.3010279001468429,
      "epsilon_dpo/beta": 0.021612215787172318,
      "epsilon_dpo/beta_margin_grad_mean": -0.12213469296693802,
      "epsilon_dpo/beta_margin_grad_std": 0.18215671181678772,
      "epsilon_dpo/beta_margin_mean": 2.9263906478881836,
      "epsilon_dpo/beta_margin_std": 1.8439240455627441,
      "epsilon_dpo/loss_margin_mean": 135.66934204101562,
      "grad_norm": 67.83074951171875,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.021793941035866737,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.4233349274571974e-07,
      "logits/chosen": -2.015986680984497,
      "logits/rejected": -1.8826128244400024,
      "logps/chosen": -121.29883575439453,
      "logps/ref_chosen": -34.816200256347656,
      "logps/ref_rejected": -92.58261108398438,
      "logps/rejected": -314.7345886230469,
      "loss": 0.3485,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.872864007949829,
      "rewards/margins": 2.9263906478881836,
      "rewards/rejected": -4.799254417419434,
      "step": 205
    },
    {
      "epoch": 0.302496328928047,
      "epsilon_dpo/beta": 0.021431388333439827,
      "epsilon_dpo/beta_margin_grad_mean": -0.14491204917430878,
      "epsilon_dpo/beta_margin_grad_std": 0.19017614424228668,
      "epsilon_dpo/beta_margin_mean": 2.6911041736602783,
      "epsilon_dpo/beta_margin_std": 1.9302713871002197,
      "epsilon_dpo/loss_margin_mean": 125.82368469238281,
      "grad_norm": 70.67012023925781,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.021611593663692474,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.415111107797445e-07,
      "logits/chosen": -1.9591203927993774,
      "logits/rejected": -1.9519094228744507,
      "logps/chosen": -117.2315673828125,
      "logps/ref_chosen": -30.099918365478516,
      "logps/ref_rejected": -103.39237976074219,
      "logps/rejected": -316.34771728515625,
      "loss": 0.4266,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.8712209463119507,
      "rewards/margins": 2.6911041736602783,
      "rewards/rejected": -4.5623250007629395,
      "step": 206
    },
    {
      "epoch": 0.3039647577092511,
      "epsilon_dpo/beta": 0.02126547135412693,
      "epsilon_dpo/beta_margin_grad_mean": -0.1460837423801422,
      "epsilon_dpo/beta_margin_grad_std": 0.1840200126171112,
      "epsilon_dpo/beta_margin_mean": 2.8168625831604004,
      "epsilon_dpo/beta_margin_std": 1.992667317390442,
      "epsilon_dpo/loss_margin_mean": 132.78456115722656,
      "grad_norm": 59.43008041381836,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.02143077179789543,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 4.4068368231789365e-07,
      "logits/chosen": -2.092212200164795,
      "logits/rejected": -1.968192458152771,
      "logps/chosen": -101.26658630371094,
      "logps/ref_chosen": -31.34187889099121,
      "logps/ref_rejected": -89.86247253417969,
      "logps/rejected": -292.57171630859375,
      "loss": 0.3788,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.4902305603027344,
      "rewards/margins": 2.8168625831604004,
      "rewards/rejected": -4.307093143463135,
      "step": 207
    },
    {
      "epoch": 0.3054331864904552,
      "epsilon_dpo/beta": 0.02108733169734478,
      "epsilon_dpo/beta_margin_grad_mean": -0.1172553300857544,
      "epsilon_dpo/beta_margin_grad_std": 0.15206165611743927,
      "epsilon_dpo/beta_margin_mean": 2.8677847385406494,
      "epsilon_dpo/beta_margin_std": 1.6818532943725586,
      "epsilon_dpo/loss_margin_mean": 136.2245635986328,
      "grad_norm": 52.98038101196289,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.021264642477035522,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.398512291636768e-07,
      "logits/chosen": -2.082803726196289,
      "logits/rejected": -2.1149425506591797,
      "logps/chosen": -119.51541137695312,
      "logps/ref_chosen": -35.819129943847656,
      "logps/ref_rejected": -100.89794921875,
      "logps/rejected": -320.81878662109375,
      "loss": 0.2919,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.7668159008026123,
      "rewards/margins": 2.8677845001220703,
      "rewards/rejected": -4.634600639343262,
      "step": 208
    },
    {
      "epoch": 0.3069016152716593,
      "epsilon_dpo/beta": 0.020917484536767006,
      "epsilon_dpo/beta_margin_grad_mean": -0.1551363617181778,
      "epsilon_dpo/beta_margin_grad_std": 0.20075276494026184,
      "epsilon_dpo/beta_margin_mean": 2.5712711811065674,
      "epsilon_dpo/beta_margin_std": 1.814591646194458,
      "epsilon_dpo/loss_margin_mean": 123.22737884521484,
      "grad_norm": 68.60906219482422,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.02108672261238098,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.3901377325300857e-07,
      "logits/chosen": -2.075434446334839,
      "logits/rejected": -1.8750395774841309,
      "logps/chosen": -114.03466796875,
      "logps/ref_chosen": -38.91720199584961,
      "logps/ref_rejected": -86.70390319824219,
      "logps/rejected": -285.0487365722656,
      "loss": 0.4441,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.573714017868042,
      "rewards/margins": 2.5712709426879883,
      "rewards/rejected": -4.144985198974609,
      "step": 209
    },
    {
      "epoch": 0.30837004405286345,
      "epsilon_dpo/beta": 0.020716214552521706,
      "epsilon_dpo/beta_margin_grad_mean": -0.13410209119319916,
      "epsilon_dpo/beta_margin_grad_std": 0.13863912224769592,
      "epsilon_dpo/beta_margin_mean": 2.6166491508483887,
      "epsilon_dpo/beta_margin_std": 1.6349241733551025,
      "epsilon_dpo/loss_margin_mean": 126.35758972167969,
      "grad_norm": 44.407615661621094,
      "kl/avg_steps": 0.96875,
      "kl/beta": 0.02091677300632,
      "kl/n_epsilon_steps": 0.015625,
      "kl/p_epsilon_steps": 0.984375,
      "learning_rate": 4.381713366536311e-07,
      "logits/chosen": -2.012350082397461,
      "logits/rejected": -1.9750981330871582,
      "logps/chosen": -108.2236099243164,
      "logps/ref_chosen": -29.373889923095703,
      "logps/ref_rejected": -85.03504943847656,
      "logps/rejected": -290.24237060546875,
      "loss": 0.32,
      "rewards/accuracies": 0.984375,
      "rewards/chosen": -1.6338069438934326,
      "rewards/margins": 2.6166491508483887,
      "rewards/rejected": -4.250455856323242,
      "step": 210
    },
    {
      "epoch": 0.30983847283406757,
      "epsilon_dpo/beta": 0.02053687535226345,
      "epsilon_dpo/beta_margin_grad_mean": -0.1510934829711914,
      "epsilon_dpo/beta_margin_grad_std": 0.18279674649238586,
      "epsilon_dpo/beta_margin_mean": 2.54111647605896,
      "epsilon_dpo/beta_margin_std": 1.720736026763916,
      "epsilon_dpo/loss_margin_mean": 123.92025756835938,
      "grad_norm": 62.367835998535156,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.02071608603000641,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.373239415645323e-07,
      "logits/chosen": -2.1857690811157227,
      "logits/rejected": -2.038249969482422,
      "logps/chosen": -131.88340759277344,
      "logps/ref_chosen": -47.88237380981445,
      "logps/ref_rejected": -93.18321228027344,
      "logps/rejected": -301.1044921875,
      "loss": 0.4024,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.7274291515350342,
      "rewards/margins": 2.541116237640381,
      "rewards/rejected": -4.268545150756836,
      "step": 211
    },
    {
      "epoch": 0.31130690161527164,
      "epsilon_dpo/beta": 0.02035231702029705,
      "epsilon_dpo/beta_margin_grad_mean": -0.11501824855804443,
      "epsilon_dpo/beta_margin_grad_std": 0.16267718374729156,
      "epsilon_dpo/beta_margin_mean": 3.0987837314605713,
      "epsilon_dpo/beta_margin_std": 2.040536403656006,
      "epsilon_dpo/loss_margin_mean": 152.44503784179688,
      "grad_norm": 56.73617935180664,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.0205363929271698,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 4.3647161031536086e-07,
      "logits/chosen": -2.1070785522460938,
      "logits/rejected": -2.0808253288269043,
      "logps/chosen": -105.17568969726562,
      "logps/ref_chosen": -35.5427360534668,
      "logps/ref_rejected": -110.93476867675781,
      "logps/rejected": -333.01275634765625,
      "loss": 0.3267,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -1.4194426536560059,
      "rewards/margins": 3.0987837314605713,
      "rewards/rejected": -4.518226623535156,
      "step": 212
    },
    {
      "epoch": 0.31277533039647576,
      "epsilon_dpo/beta": 0.02018861286342144,
      "epsilon_dpo/beta_margin_grad_mean": -0.1576324999332428,
      "epsilon_dpo/beta_margin_grad_std": 0.1839437037706375,
      "epsilon_dpo/beta_margin_mean": 2.5347602367401123,
      "epsilon_dpo/beta_margin_std": 1.81096613407135,
      "epsilon_dpo/loss_margin_mean": 125.82254028320312,
      "grad_norm": 60.45912551879883,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.02035195380449295,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.3561436536583774e-07,
      "logits/chosen": -2.2190451622009277,
      "logits/rejected": -2.022303581237793,
      "logps/chosen": -121.68977355957031,
      "logps/ref_chosen": -46.382476806640625,
      "logps/ref_rejected": -98.22808074951172,
      "logps/rejected": -299.35791015625,
      "loss": 0.4116,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.522735595703125,
      "rewards/margins": 2.5347602367401123,
      "rewards/rejected": -4.057496070861816,
      "step": 213
    },
    {
      "epoch": 0.3142437591776799,
      "epsilon_dpo/beta": 0.02001328393816948,
      "epsilon_dpo/beta_margin_grad_mean": -0.18691565096378326,
      "epsilon_dpo/beta_margin_grad_std": 0.2173331379890442,
      "epsilon_dpo/beta_margin_mean": 2.255009889602661,
      "epsilon_dpo/beta_margin_std": 1.9002856016159058,
      "epsilon_dpo/loss_margin_mean": 112.92225646972656,
      "grad_norm": 80.94749450683594,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.020187927410006523,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.3475222930516473e-07,
      "logits/chosen": -1.925289273262024,
      "logits/rejected": -1.9057788848876953,
      "logps/chosen": -117.653564453125,
      "logps/ref_chosen": -33.69921112060547,
      "logps/ref_rejected": -83.6459732055664,
      "logps/rejected": -280.5225830078125,
      "loss": 0.5721,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.683471441268921,
      "rewards/margins": 2.255009889602661,
      "rewards/rejected": -3.938481330871582,
      "step": 214
    },
    {
      "epoch": 0.315712187958884,
      "epsilon_dpo/beta": 0.019833432510495186,
      "epsilon_dpo/beta_margin_grad_mean": -0.12425589561462402,
      "epsilon_dpo/beta_margin_grad_std": 0.13742640614509583,
      "epsilon_dpo/beta_margin_mean": 2.6745448112487793,
      "epsilon_dpo/beta_margin_std": 1.592254638671875,
      "epsilon_dpo/loss_margin_mean": 134.98655700683594,
      "grad_norm": 46.67967987060547,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.020012814551591873,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 4.3388522485142885e-07,
      "logits/chosen": -2.086688756942749,
      "logits/rejected": -2.0053024291992188,
      "logps/chosen": -107.9367446899414,
      "logps/ref_chosen": -30.35393714904785,
      "logps/ref_rejected": -99.64697265625,
      "logps/rejected": -312.21636962890625,
      "loss": 0.2971,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.5396337509155273,
      "rewards/margins": 2.6745448112487793,
      "rewards/rejected": -4.214178562164307,
      "step": 215
    },
    {
      "epoch": 0.31718061674008813,
      "epsilon_dpo/beta": 0.019673900678753853,
      "epsilon_dpo/beta_margin_grad_mean": -0.1697354018688202,
      "epsilon_dpo/beta_margin_grad_std": 0.19105024635791779,
      "epsilon_dpo/beta_margin_mean": 2.5016613006591797,
      "epsilon_dpo/beta_margin_std": 1.8782446384429932,
      "epsilon_dpo/loss_margin_mean": 127.44037628173828,
      "grad_norm": 60.99223709106445,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.019833076745271683,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.330133748510036e-07,
      "logits/chosen": -2.0426881313323975,
      "logits/rejected": -1.9267804622650146,
      "logps/chosen": -102.5655746459961,
      "logps/ref_chosen": -28.687610626220703,
      "logps/ref_rejected": -83.20097351074219,
      "logps/rejected": -284.519287109375,
      "loss": 0.4448,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.4557723999023438,
      "rewards/margins": 2.5016613006591797,
      "rewards/rejected": -3.9574337005615234,
      "step": 216
    },
    {
      "epoch": 0.3186490455212922,
      "epsilon_dpo/beta": 0.019503042101860046,
      "epsilon_dpo/beta_margin_grad_mean": -0.12006211280822754,
      "epsilon_dpo/beta_margin_grad_std": 0.1564583033323288,
      "epsilon_dpo/beta_margin_mean": 2.867358446121216,
      "epsilon_dpo/beta_margin_std": 1.7631230354309082,
      "epsilon_dpo/loss_margin_mean": 147.23101806640625,
      "grad_norm": 50.90324401855469,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.01967323198914528,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.3213670227794757e-07,
      "logits/chosen": -2.1286168098449707,
      "logits/rejected": -2.087315559387207,
      "logps/chosen": -98.74182891845703,
      "logps/ref_chosen": -31.528701782226562,
      "logps/ref_rejected": -107.33251190185547,
      "logps/rejected": -321.77667236328125,
      "loss": 0.3023,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.3119699954986572,
      "rewards/margins": 2.867358684539795,
      "rewards/rejected": -4.179328441619873,
      "step": 217
    },
    {
      "epoch": 0.3201174743024963,
      "epsilon_dpo/beta": 0.01932777464389801,
      "epsilon_dpo/beta_margin_grad_mean": -0.15312562882900238,
      "epsilon_dpo/beta_margin_grad_std": 0.161267951130867,
      "epsilon_dpo/beta_margin_mean": 2.22654128074646,
      "epsilon_dpo/beta_margin_std": 1.3763279914855957,
      "epsilon_dpo/loss_margin_mean": 115.3489990234375,
      "grad_norm": 58.98360824584961,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.019502583891153336,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 4.3125523023339815e-07,
      "logits/chosen": -2.162309169769287,
      "logits/rejected": -2.1342501640319824,
      "logps/chosen": -108.99090576171875,
      "logps/ref_chosen": -36.7948112487793,
      "logps/ref_rejected": -93.1485366821289,
      "logps/rejected": -280.693603515625,
      "loss": 0.3962,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.39695405960083,
      "rewards/margins": 2.226541519165039,
      "rewards/rejected": -3.623495578765869,
      "step": 218
    },
    {
      "epoch": 0.32158590308370044,
      "epsilon_dpo/beta": 0.019178351387381554,
      "epsilon_dpo/beta_margin_grad_mean": -0.178512305021286,
      "epsilon_dpo/beta_margin_grad_std": 0.184907004237175,
      "epsilon_dpo/beta_margin_mean": 2.2544329166412354,
      "epsilon_dpo/beta_margin_std": 1.7746726274490356,
      "epsilon_dpo/loss_margin_mean": 117.84127807617188,
      "grad_norm": 61.87991714477539,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.019327430054545403,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 4.303689819449636e-07,
      "logits/chosen": -2.2709426879882812,
      "logits/rejected": -2.1236162185668945,
      "logps/chosen": -113.684814453125,
      "logps/ref_chosen": -42.875755310058594,
      "logps/ref_rejected": -92.20575714111328,
      "logps/rejected": -280.8560791015625,
      "loss": 0.4669,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.3612611293792725,
      "rewards/margins": 2.2544326782226562,
      "rewards/rejected": -3.615694046020508,
      "step": 219
    },
    {
      "epoch": 0.32305433186490456,
      "epsilon_dpo/beta": 0.019017694517970085,
      "epsilon_dpo/beta_margin_grad_mean": -0.17227615416049957,
      "epsilon_dpo/beta_margin_grad_std": 0.1603304147720337,
      "epsilon_dpo/beta_margin_mean": 2.1362380981445312,
      "epsilon_dpo/beta_margin_std": 1.426170825958252,
      "epsilon_dpo/loss_margin_mean": 112.51897430419922,
      "grad_norm": 57.75115966796875,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.019177604466676712,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.2947798076611047e-07,
      "logits/chosen": -2.249175548553467,
      "logits/rejected": -2.228762626647949,
      "logps/chosen": -121.99055480957031,
      "logps/ref_chosen": -43.218231201171875,
      "logps/ref_rejected": -94.84095764160156,
      "logps/rejected": -286.1322326660156,
      "loss": 0.4237,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.4991801977157593,
      "rewards/margins": 2.1362380981445312,
      "rewards/rejected": -3.63541841506958,
      "step": 220
    },
    {
      "epoch": 0.3245227606461087,
      "epsilon_dpo/beta": 0.018852630630135536,
      "epsilon_dpo/beta_margin_grad_mean": -0.11038573086261749,
      "epsilon_dpo/beta_margin_grad_std": 0.14331035315990448,
      "epsilon_dpo/beta_margin_mean": 2.8602123260498047,
      "epsilon_dpo/beta_margin_std": 1.571157693862915,
      "epsilon_dpo/loss_margin_mean": 151.9083251953125,
      "grad_norm": 44.90319061279297,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.01901714690029621,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.285822501755485e-07,
      "logits/chosen": -2.0982179641723633,
      "logits/rejected": -2.1222004890441895,
      "logps/chosen": -105.45207214355469,
      "logps/ref_chosen": -36.884986877441406,
      "logps/ref_rejected": -112.87872314453125,
      "logps/rejected": -333.3541259765625,
      "loss": 0.2732,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.2945091724395752,
      "rewards/margins": 2.8602123260498047,
      "rewards/rejected": -4.154721260070801,
      "step": 221
    },
    {
      "epoch": 0.32599118942731276,
      "epsilon_dpo/beta": 0.018689103424549103,
      "epsilon_dpo/beta_margin_grad_mean": -0.15696805715560913,
      "epsilon_dpo/beta_margin_grad_std": 0.16546384990215302,
      "epsilon_dpo/beta_margin_mean": 2.4271247386932373,
      "epsilon_dpo/beta_margin_std": 1.7556877136230469,
      "epsilon_dpo/loss_margin_mean": 130.051025390625,
      "grad_norm": 68.18460845947266,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.018852191045880318,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.276818137766118e-07,
      "logits/chosen": -2.1893157958984375,
      "logits/rejected": -2.2103257179260254,
      "logps/chosen": -112.70558166503906,
      "logps/ref_chosen": -37.27526092529297,
      "logps/ref_rejected": -106.37206268310547,
      "logps/rejected": -311.8533935546875,
      "loss": 0.3942,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.4119031429290771,
      "rewards/margins": 2.4271249771118164,
      "rewards/rejected": -3.8390281200408936,
      "step": 222
    },
    {
      "epoch": 0.3274596182085169,
      "epsilon_dpo/beta": 0.018556196242570877,
      "epsilon_dpo/beta_margin_grad_mean": -0.1877458542585373,
      "epsilon_dpo/beta_margin_grad_std": 0.21679145097732544,
      "epsilon_dpo/beta_margin_mean": 2.193220615386963,
      "epsilon_dpo/beta_margin_std": 1.8057856559753418,
      "epsilon_dpo/loss_margin_mean": 118.63549041748047,
      "grad_norm": 77.07755279541016,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.018688665702939034,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 4.2677669529663686e-07,
      "logits/chosen": -2.1241304874420166,
      "logits/rejected": -2.0650038719177246,
      "logps/chosen": -117.65826416015625,
      "logps/ref_chosen": -32.709083557128906,
      "logps/ref_rejected": -86.52430725097656,
      "logps/rejected": -290.10894775390625,
      "loss": 0.542,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -1.581796407699585,
      "rewards/margins": 2.193220615386963,
      "rewards/rejected": -3.775017261505127,
      "step": 223
    },
    {
      "epoch": 0.328928046989721,
      "epsilon_dpo/beta": 0.018406376242637634,
      "epsilon_dpo/beta_margin_grad_mean": -0.15500150620937347,
      "epsilon_dpo/beta_margin_grad_std": 0.1830359399318695,
      "epsilon_dpo/beta_margin_mean": 2.481572389602661,
      "epsilon_dpo/beta_margin_std": 1.7067841291427612,
      "epsilon_dpo/loss_margin_mean": 135.1320037841797,
      "grad_norm": 64.31571197509766,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.018555298447608948,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.2586691858633747e-07,
      "logits/chosen": -2.1345582008361816,
      "logits/rejected": -1.9911662340164185,
      "logps/chosen": -113.04464721679688,
      "logps/ref_chosen": -35.54627990722656,
      "logps/ref_rejected": -81.60932922363281,
      "logps/rejected": -294.23968505859375,
      "loss": 0.4078,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.429018497467041,
      "rewards/margins": 2.481572389602661,
      "rewards/rejected": -3.910590648651123,
      "step": 224
    },
    {
      "epoch": 0.3303964757709251,
      "epsilon_dpo/beta": 0.018246525898575783,
      "epsilon_dpo/beta_margin_grad_mean": -0.1459973156452179,
      "epsilon_dpo/beta_margin_grad_std": 0.19097650051116943,
      "epsilon_dpo/beta_margin_mean": 2.6388025283813477,
      "epsilon_dpo/beta_margin_std": 1.800585150718689,
      "epsilon_dpo/loss_margin_mean": 144.87014770507812,
      "grad_norm": 65.03907775878906,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.018405752256512642,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.249525076191759e-07,
      "logits/chosen": -2.1871337890625,
      "logits/rejected": -2.1202986240386963,
      "logps/chosen": -111.3397216796875,
      "logps/ref_chosen": -34.65919876098633,
      "logps/ref_rejected": -106.95365905761719,
      "logps/rejected": -328.50433349609375,
      "loss": 0.4097,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.4013426303863525,
      "rewards/margins": 2.6388025283813477,
      "rewards/rejected": -4.040144920349121,
      "step": 225
    },
    {
      "epoch": 0.33186490455212925,
      "epsilon_dpo/beta": 0.01808825507760048,
      "epsilon_dpo/beta_margin_grad_mean": -0.17010191082954407,
      "epsilon_dpo/beta_margin_grad_std": 0.16130515933036804,
      "epsilon_dpo/beta_margin_mean": 2.194584846496582,
      "epsilon_dpo/beta_margin_std": 1.5229578018188477,
      "epsilon_dpo/loss_margin_mean": 121.48944854736328,
      "grad_norm": 53.89804458618164,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.01824609935283661,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.2403348649073167e-07,
      "logits/chosen": -2.2922072410583496,
      "logits/rejected": -2.074962615966797,
      "logps/chosen": -120.92872619628906,
      "logps/ref_chosen": -44.4660758972168,
      "logps/ref_rejected": -84.042724609375,
      "logps/rejected": -281.99481201171875,
      "loss": 0.4196,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.3841276168823242,
      "rewards/margins": 2.194584846496582,
      "rewards/rejected": -3.5787124633789062,
      "step": 226
    },
    {
      "epoch": 0.3333333333333333,
      "epsilon_dpo/beta": 0.017948314547538757,
      "epsilon_dpo/beta_margin_grad_mean": -0.15359929203987122,
      "epsilon_dpo/beta_margin_grad_std": 0.18651175498962402,
      "epsilon_dpo/beta_margin_mean": 2.6601450443267822,
      "epsilon_dpo/beta_margin_std": 1.9379621744155884,
      "epsilon_dpo/loss_margin_mean": 148.5768585205078,
      "grad_norm": 62.88763427734375,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.018087830394506454,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 4.2310987941806615e-07,
      "logits/chosen": -2.217491865158081,
      "logits/rejected": -2.0915043354034424,
      "logps/chosen": -111.29264831542969,
      "logps/ref_chosen": -33.7484245300293,
      "logps/ref_rejected": -106.498291015625,
      "logps/rejected": -332.619384765625,
      "loss": 0.4017,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.3941301107406616,
      "rewards/margins": 2.6601450443267822,
      "rewards/rejected": -4.054275035858154,
      "step": 227
    },
    {
      "epoch": 0.33480176211453744,
      "epsilon_dpo/beta": 0.0177923534065485,
      "epsilon_dpo/beta_margin_grad_mean": -0.15493857860565186,
      "epsilon_dpo/beta_margin_grad_std": 0.1513359248638153,
      "epsilon_dpo/beta_margin_mean": 2.2347021102905273,
      "epsilon_dpo/beta_margin_std": 1.3810467720031738,
      "epsilon_dpo/loss_margin_mean": 125.76899719238281,
      "grad_norm": 55.945281982421875,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.017947614192962646,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.2218171073908463e-07,
      "logits/chosen": -2.300729990005493,
      "logits/rejected": -2.2182745933532715,
      "logps/chosen": -128.66030883789062,
      "logps/ref_chosen": -45.935726165771484,
      "logps/ref_rejected": -96.16656494140625,
      "logps/rejected": -304.66015625,
      "loss": 0.3771,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -1.473421335220337,
      "rewards/margins": 2.2347021102905273,
      "rewards/rejected": -3.7081234455108643,
      "step": 228
    },
    {
      "epoch": 0.33627019089574156,
      "epsilon_dpo/beta": 0.01766026020050049,
      "epsilon_dpo/beta_margin_grad_mean": -0.155641108751297,
      "epsilon_dpo/beta_margin_grad_std": 0.19922208786010742,
      "epsilon_dpo/beta_margin_mean": 2.632094383239746,
      "epsilon_dpo/beta_margin_std": 1.839905023574829,
      "epsilon_dpo/loss_margin_mean": 149.47091674804688,
      "grad_norm": 64.01463317871094,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.017791934311389923,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 4.212490049118951e-07,
      "logits/chosen": -2.2102646827697754,
      "logits/rejected": -2.084036350250244,
      "logps/chosen": -100.0278091430664,
      "logps/ref_chosen": -35.16382598876953,
      "logps/ref_rejected": -89.91634368896484,
      "logps/rejected": -304.251220703125,
      "loss": 0.4188,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.1469919681549072,
      "rewards/margins": 2.632094383239746,
      "rewards/rejected": -3.779086112976074,
      "step": 229
    },
    {
      "epoch": 0.3377386196769457,
      "epsilon_dpo/beta": 0.017512237653136253,
      "epsilon_dpo/beta_margin_grad_mean": -0.15799804031848907,
      "epsilon_dpo/beta_margin_grad_std": 0.1725914478302002,
      "epsilon_dpo/beta_margin_mean": 2.30881667137146,
      "epsilon_dpo/beta_margin_std": 1.4573560953140259,
      "epsilon_dpo/loss_margin_mean": 132.07550048828125,
      "grad_norm": 57.6724853515625,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.017659489065408707,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.203117865141635e-07,
      "logits/chosen": -1.9427410364151,
      "logits/rejected": -1.9851582050323486,
      "logps/chosen": -114.36758422851562,
      "logps/ref_chosen": -28.29522705078125,
      "logps/ref_rejected": -89.92157745361328,
      "logps/rejected": -308.0694274902344,
      "loss": 0.4007,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.5095224380493164,
      "rewards/margins": 2.30881667137146,
      "rewards/rejected": -3.8183393478393555,
      "step": 230
    },
    {
      "epoch": 0.3392070484581498,
      "epsilon_dpo/beta": 0.01736571453511715,
      "epsilon_dpo/beta_margin_grad_mean": -0.15104635059833527,
      "epsilon_dpo/beta_margin_grad_std": 0.16789193451404572,
      "epsilon_dpo/beta_margin_mean": 2.344306468963623,
      "epsilon_dpo/beta_margin_std": 1.5224120616912842,
      "epsilon_dpo/loss_margin_mean": 135.25741577148438,
      "grad_norm": 58.66816711425781,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.01751173473894596,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.1937008024246625e-07,
      "logits/chosen": -2.2687087059020996,
      "logits/rejected": -2.0110936164855957,
      "logps/chosen": -122.01680755615234,
      "logps/ref_chosen": -39.274810791015625,
      "logps/ref_rejected": -80.1042709350586,
      "logps/rejected": -298.10369873046875,
      "loss": 0.3873,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.4385372400283813,
      "rewards/margins": 2.344306468963623,
      "rewards/rejected": -3.782843828201294,
      "step": 231
    },
    {
      "epoch": 0.3406754772393539,
      "epsilon_dpo/beta": 0.0172258447855711,
      "epsilon_dpo/beta_margin_grad_mean": -0.16703417897224426,
      "epsilon_dpo/beta_margin_grad_std": 0.17445211112499237,
      "epsilon_dpo/beta_margin_mean": 2.3062567710876465,
      "epsilon_dpo/beta_margin_std": 1.650961995124817,
      "epsilon_dpo/loss_margin_mean": 134.1649169921875,
      "grad_norm": 55.898075103759766,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.017365215346217155,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.1842391091163933e-07,
      "logits/chosen": -2.2190351486206055,
      "logits/rejected": -2.2044219970703125,
      "logps/chosen": -121.8990707397461,
      "logps/ref_chosen": -42.393104553222656,
      "logps/ref_rejected": -88.90144348144531,
      "logps/rejected": -302.57232666015625,
      "loss": 0.4288,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.3723829984664917,
      "rewards/margins": 2.3062567710876465,
      "rewards/rejected": -3.6786396503448486,
      "step": 232
    },
    {
      "epoch": 0.342143906020558,
      "epsilon_dpo/beta": 0.017076246440410614,
      "epsilon_dpo/beta_margin_grad_mean": -0.11840350925922394,
      "epsilon_dpo/beta_margin_grad_std": 0.15504033863544464,
      "epsilon_dpo/beta_margin_mean": 2.873262882232666,
      "epsilon_dpo/beta_margin_std": 1.6749653816223145,
      "epsilon_dpo/loss_margin_mean": 168.5035858154297,
      "grad_norm": 49.05929183959961,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.01722525991499424,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.174733034541245e-07,
      "logits/chosen": -2.11576509475708,
      "logits/rejected": -2.1665868759155273,
      "logps/chosen": -108.28323364257812,
      "logps/ref_chosen": -30.483036041259766,
      "logps/ref_rejected": -115.03839111328125,
      "logps/rejected": -361.3421630859375,
      "loss": 0.2959,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.3310226202011108,
      "rewards/margins": 2.873262882232666,
      "rewards/rejected": -4.204285621643066,
      "step": 233
    },
    {
      "epoch": 0.3436123348017621,
      "epsilon_dpo/beta": 0.01694413460791111,
      "epsilon_dpo/beta_margin_grad_mean": -0.15992534160614014,
      "epsilon_dpo/beta_margin_grad_std": 0.19790522754192352,
      "epsilon_dpo/beta_margin_mean": 2.6408679485321045,
      "epsilon_dpo/beta_margin_std": 2.0294158458709717,
      "epsilon_dpo/loss_margin_mean": 156.26454162597656,
      "grad_norm": 67.56781005859375,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.01707584597170353,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 4.165182829193126e-07,
      "logits/chosen": -2.0071873664855957,
      "logits/rejected": -2.1877129077911377,
      "logps/chosen": -111.21868133544922,
      "logps/ref_chosen": -30.016942977905273,
      "logps/ref_rejected": -108.75608825683594,
      "logps/rejected": -346.22235107421875,
      "loss": 0.4328,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.3788411617279053,
      "rewards/margins": 2.6408681869506836,
      "rewards/rejected": -4.01970911026001,
      "step": 234
    },
    {
      "epoch": 0.34508076358296624,
      "epsilon_dpo/beta": 0.01681278459727764,
      "epsilon_dpo/beta_margin_grad_mean": -0.175105020403862,
      "epsilon_dpo/beta_margin_grad_std": 0.1960555464029312,
      "epsilon_dpo/beta_margin_mean": 2.271127700805664,
      "epsilon_dpo/beta_margin_std": 1.7841382026672363,
      "epsilon_dpo/loss_margin_mean": 135.44342041015625,
      "grad_norm": 77.5376205444336,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.016943475231528282,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 4.1555887447288255e-07,
      "logits/chosen": -2.111543655395508,
      "logits/rejected": -2.081702470779419,
      "logps/chosen": -141.7093963623047,
      "logps/ref_chosen": -46.081146240234375,
      "logps/ref_rejected": -96.02244567871094,
      "logps/rejected": -327.0941162109375,
      "loss": 0.4894,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.6119556427001953,
      "rewards/margins": 2.271127700805664,
      "rewards/rejected": -3.8830833435058594,
      "step": 235
    },
    {
      "epoch": 0.3465491923641703,
      "epsilon_dpo/beta": 0.01666669175028801,
      "epsilon_dpo/beta_margin_grad_mean": -0.1637529879808426,
      "epsilon_dpo/beta_margin_grad_std": 0.16410239040851593,
      "epsilon_dpo/beta_margin_mean": 2.3281750679016113,
      "epsilon_dpo/beta_margin_std": 1.6591843366622925,
      "epsilon_dpo/loss_margin_mean": 139.8822784423828,
      "grad_norm": 56.97810363769531,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.01681213080883026,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.1459510339613946e-07,
      "logits/chosen": -2.05222487449646,
      "logits/rejected": -2.147634267807007,
      "logps/chosen": -112.40109252929688,
      "logps/ref_chosen": -31.17489242553711,
      "logps/ref_rejected": -108.55508422851562,
      "logps/rejected": -329.66357421875,
      "loss": 0.4048,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.3549550771713257,
      "rewards/margins": 2.3281750679016113,
      "rewards/rejected": -3.6831302642822266,
      "step": 236
    },
    {
      "epoch": 0.34801762114537443,
      "epsilon_dpo/beta": 0.01652212254703045,
      "epsilon_dpo/beta_margin_grad_mean": -0.13068081438541412,
      "epsilon_dpo/beta_margin_grad_std": 0.15401048958301544,
      "epsilon_dpo/beta_margin_mean": 2.651111125946045,
      "epsilon_dpo/beta_margin_std": 1.6006019115447998,
      "epsilon_dpo/loss_margin_mean": 160.67715454101562,
      "grad_norm": 47.88125991821289,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.01666630059480667,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.136269950853473e-07,
      "logits/chosen": -2.0394325256347656,
      "logits/rejected": -2.007416248321533,
      "logps/chosen": -99.72041320800781,
      "logps/ref_chosen": -27.259479522705078,
      "logps/ref_rejected": -100.87033081054688,
      "logps/rejected": -334.0084228515625,
      "loss": 0.3225,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.198293924331665,
      "rewards/margins": 2.651111125946045,
      "rewards/rejected": -3.849404811859131,
      "step": 237
    },
    {
      "epoch": 0.34948604992657856,
      "epsilon_dpo/beta": 0.016383972018957138,
      "epsilon_dpo/beta_margin_grad_mean": -0.1836518496274948,
      "epsilon_dpo/beta_margin_grad_std": 0.17506077885627747,
      "epsilon_dpo/beta_margin_mean": 2.173346996307373,
      "epsilon_dpo/beta_margin_std": 1.7011436223983765,
      "epsilon_dpo/loss_margin_mean": 132.8949737548828,
      "grad_norm": 65.86769104003906,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.01652173511683941,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.126545750510605e-07,
      "logits/chosen": -2.0702290534973145,
      "logits/rejected": -2.123429298400879,
      "logps/chosen": -137.6107177734375,
      "logps/ref_chosen": -40.190574645996094,
      "logps/ref_rejected": -93.49894714355469,
      "logps/rejected": -323.8140563964844,
      "loss": 0.4646,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.598804235458374,
      "rewards/margins": 2.173346996307373,
      "rewards/rejected": -3.772151231765747,
      "step": 238
    },
    {
      "epoch": 0.3509544787077827,
      "epsilon_dpo/beta": 0.016236647963523865,
      "epsilon_dpo/beta_margin_grad_mean": -0.16019320487976074,
      "epsilon_dpo/beta_margin_grad_std": 0.1673043966293335,
      "epsilon_dpo/beta_margin_mean": 2.3489763736724854,
      "epsilon_dpo/beta_margin_std": 1.5995773077011108,
      "epsilon_dpo/loss_margin_mean": 144.83961486816406,
      "grad_norm": 56.121368408203125,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.016383498907089233,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 4.116778689174514e-07,
      "logits/chosen": -2.0602469444274902,
      "logits/rejected": -1.9942574501037598,
      "logps/chosen": -124.91606140136719,
      "logps/ref_chosen": -38.33943176269531,
      "logps/ref_rejected": -99.64225769042969,
      "logps/rejected": -331.0585021972656,
      "loss": 0.4034,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -1.4062585830688477,
      "rewards/margins": 2.3489763736724854,
      "rewards/rejected": -3.755234956741333,
      "step": 239
    },
    {
      "epoch": 0.3524229074889868,
      "epsilon_dpo/beta": 0.016100972890853882,
      "epsilon_dpo/beta_margin_grad_mean": -0.16005262732505798,
      "epsilon_dpo/beta_margin_grad_std": 0.16643419861793518,
      "epsilon_dpo/beta_margin_mean": 2.2492594718933105,
      "epsilon_dpo/beta_margin_std": 1.4863446950912476,
      "epsilon_dpo/loss_margin_mean": 139.95294189453125,
      "grad_norm": 66.01368713378906,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.016236357390880585,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 4.106969024216348e-07,
      "logits/chosen": -2.0398552417755127,
      "logits/rejected": -1.8625662326812744,
      "logps/chosen": -111.68765258789062,
      "logps/ref_chosen": -36.1579704284668,
      "logps/ref_rejected": -80.07916259765625,
      "logps/rejected": -295.5617980957031,
      "loss": 0.4095,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.2190780639648438,
      "rewards/margins": 2.2492594718933105,
      "rewards/rejected": -3.4683375358581543,
      "step": 240
    },
    {
      "epoch": 0.35389133627019087,
      "epsilon_dpo/beta": 0.015971289947628975,
      "epsilon_dpo/beta_margin_grad_mean": -0.16114287078380585,
      "epsilon_dpo/beta_margin_grad_std": 0.18923403322696686,
      "epsilon_dpo/beta_margin_mean": 2.457042932510376,
      "epsilon_dpo/beta_margin_std": 1.8046914339065552,
      "epsilon_dpo/loss_margin_mean": 154.20245361328125,
      "grad_norm": 74.83716583251953,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.016100509092211723,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.097117014129903e-07,
      "logits/chosen": -2.196474313735962,
      "logits/rejected": -1.913820505142212,
      "logps/chosen": -126.07571411132812,
      "logps/ref_chosen": -44.0040397644043,
      "logps/ref_rejected": -93.001220703125,
      "logps/rejected": -329.2753601074219,
      "loss": 0.426,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -1.3152108192443848,
      "rewards/margins": 2.457043170928955,
      "rewards/rejected": -3.7722537517547607,
      "step": 241
    },
    {
      "epoch": 0.355359765051395,
      "epsilon_dpo/beta": 0.015827594324946404,
      "epsilon_dpo/beta_margin_grad_mean": -0.14874504506587982,
      "epsilon_dpo/beta_margin_grad_std": 0.17723040282726288,
      "epsilon_dpo/beta_margin_mean": 2.45681095123291,
      "epsilon_dpo/beta_margin_std": 1.6646445989608765,
      "epsilon_dpo/loss_margin_mean": 155.43951416015625,
      "grad_norm": 78.47781372070312,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.0159707460552454,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 4.087222918524807e-07,
      "logits/chosen": -2.161656379699707,
      "logits/rejected": -1.9027358293533325,
      "logps/chosen": -109.82485961914062,
      "logps/ref_chosen": -32.014137268066406,
      "logps/ref_rejected": -91.38673400878906,
      "logps/rejected": -324.636962890625,
      "loss": 0.4175,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.2343041896820068,
      "rewards/margins": 2.45681095123291,
      "rewards/rejected": -3.691114902496338,
      "step": 242
    },
    {
      "epoch": 0.3568281938325991,
      "epsilon_dpo/beta": 0.01569039188325405,
      "epsilon_dpo/beta_margin_grad_mean": -0.14957940578460693,
      "epsilon_dpo/beta_margin_grad_std": 0.1580885350704193,
      "epsilon_dpo/beta_margin_mean": 2.3924245834350586,
      "epsilon_dpo/beta_margin_std": 1.4725757837295532,
      "epsilon_dpo/loss_margin_mean": 152.6823272705078,
      "grad_norm": 58.3444709777832,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.01582731120288372,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 4.07728699811968e-07,
      "logits/chosen": -2.0235860347747803,
      "logits/rejected": -1.7727184295654297,
      "logps/chosen": -119.57041931152344,
      "logps/ref_chosen": -37.93376922607422,
      "logps/ref_rejected": -81.79078674316406,
      "logps/rejected": -316.1097717285156,
      "loss": 0.3675,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.282745122909546,
      "rewards/margins": 2.3924245834350586,
      "rewards/rejected": -3.6751694679260254,
      "step": 243
    },
    {
      "epoch": 0.35829662261380324,
      "epsilon_dpo/beta": 0.015569004230201244,
      "epsilon_dpo/beta_margin_grad_mean": -0.18348611891269684,
      "epsilon_dpo/beta_margin_grad_std": 0.19074872136116028,
      "epsilon_dpo/beta_margin_mean": 2.0866005420684814,
      "epsilon_dpo/beta_margin_std": 1.515892744064331,
      "epsilon_dpo/loss_margin_mean": 134.38136291503906,
      "grad_norm": 60.811737060546875,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.015690024942159653,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 4.067309514735267e-07,
      "logits/chosen": -2.1224303245544434,
      "logits/rejected": -2.0141685009002686,
      "logps/chosen": -125.60481262207031,
      "logps/ref_chosen": -39.915008544921875,
      "logps/ref_rejected": -101.33251953125,
      "logps/rejected": -321.4036865234375,
      "loss": 0.4834,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -1.3352429866790771,
      "rewards/margins": 2.0866003036499023,
      "rewards/rejected": -3.4218435287475586,
      "step": 244
    },
    {
      "epoch": 0.35976505139500736,
      "epsilon_dpo/beta": 0.015458044596016407,
      "epsilon_dpo/beta_margin_grad_mean": -0.19498883187770844,
      "epsilon_dpo/beta_margin_grad_std": 0.21323725581169128,
      "epsilon_dpo/beta_margin_mean": 2.181067943572998,
      "epsilon_dpo/beta_margin_std": 1.7734534740447998,
      "epsilon_dpo/loss_margin_mean": 141.5851593017578,
      "grad_norm": 70.79053497314453,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.01556839607656002,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 4.057290731287531e-07,
      "logits/chosen": -2.187222957611084,
      "logits/rejected": -1.977565050125122,
      "logps/chosen": -115.84690856933594,
      "logps/ref_chosen": -40.404693603515625,
      "logps/ref_rejected": -93.24897766113281,
      "logps/rejected": -310.2763671875,
      "loss": 0.538,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -1.1700770854949951,
      "rewards/margins": 2.181067943572998,
      "rewards/rejected": -3.351145029067993,
      "step": 245
    },
    {
      "epoch": 0.36123348017621143,
      "epsilon_dpo/beta": 0.015333239920437336,
      "epsilon_dpo/beta_margin_grad_mean": -0.18123117089271545,
      "epsilon_dpo/beta_margin_grad_std": 0.17709243297576904,
      "epsilon_dpo/beta_margin_mean": 2.124358892440796,
      "epsilon_dpo/beta_margin_std": 1.5861018896102905,
      "epsilon_dpo/loss_margin_mean": 138.85049438476562,
      "grad_norm": 59.1838264465332,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.01545729674398899,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.047230911780736e-07,
      "logits/chosen": -2.1199212074279785,
      "logits/rejected": -1.9131314754486084,
      "logps/chosen": -134.75564575195312,
      "logps/ref_chosen": -46.212432861328125,
      "logps/ref_rejected": -90.18721008300781,
      "logps/rejected": -317.5809326171875,
      "loss": 0.4609,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.3602585792541504,
      "rewards/margins": 2.124358892440796,
      "rewards/rejected": -3.4846174716949463,
      "step": 246
    },
    {
      "epoch": 0.36270190895741555,
      "epsilon_dpo/beta": 0.015209660865366459,
      "epsilon_dpo/beta_margin_grad_mean": -0.16919678449630737,
      "epsilon_dpo/beta_margin_grad_std": 0.20299823582172394,
      "epsilon_dpo/beta_margin_mean": 2.335465669631958,
      "epsilon_dpo/beta_margin_std": 1.7916253805160522,
      "epsilon_dpo/loss_margin_mean": 153.95811462402344,
      "grad_norm": 73.05105590820312,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.015332718379795551,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.0371303213004814e-07,
      "logits/chosen": -2.0222108364105225,
      "logits/rejected": -2.00075626373291,
      "logps/chosen": -138.3466796875,
      "logps/ref_chosen": -41.23990249633789,
      "logps/ref_rejected": -111.74742126464844,
      "logps/rejected": -362.81231689453125,
      "loss": 0.4948,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.479392409324646,
      "rewards/margins": 2.335465431213379,
      "rewards/rejected": -3.8148579597473145,
      "step": 247
    },
    {
      "epoch": 0.3641703377386197,
      "epsilon_dpo/beta": 0.015063311904668808,
      "epsilon_dpo/beta_margin_grad_mean": -0.12139809876680374,
      "epsilon_dpo/beta_margin_grad_std": 0.12686574459075928,
      "epsilon_dpo/beta_margin_mean": 2.461768865585327,
      "epsilon_dpo/beta_margin_std": 1.234181523323059,
      "epsilon_dpo/loss_margin_mean": 163.502685546875,
      "grad_norm": 40.315460205078125,
      "kl/avg_steps": 0.96875,
      "kl/beta": 0.015209143981337547,
      "kl/n_epsilon_steps": 0.015625,
      "kl/p_epsilon_steps": 0.984375,
      "learning_rate": 4.0269892260067197e-07,
      "logits/chosen": -1.9299895763397217,
      "logits/rejected": -2.0007243156433105,
      "logps/chosen": -102.28887939453125,
      "logps/ref_chosen": -25.618247985839844,
      "logps/ref_rejected": -97.59014892578125,
      "logps/rejected": -337.76348876953125,
      "loss": 0.2904,
      "rewards/accuracies": 0.984375,
      "rewards/chosen": -1.1554369926452637,
      "rewards/margins": 2.461768865585327,
      "rewards/rejected": -3.61720609664917,
      "step": 248
    },
    {
      "epoch": 0.3656387665198238,
      "epsilon_dpo/beta": 0.014947031624615192,
      "epsilon_dpo/beta_margin_grad_mean": -0.18095283210277557,
      "epsilon_dpo/beta_margin_grad_std": 0.164636492729187,
      "epsilon_dpo/beta_margin_mean": 2.128697395324707,
      "epsilon_dpo/beta_margin_std": 1.604970097541809,
      "epsilon_dpo/loss_margin_mean": 142.73989868164062,
      "grad_norm": 58.40861129760742,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.015063218772411346,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 4.0168078931267426e-07,
      "logits/chosen": -2.040287494659424,
      "logits/rejected": -1.8837002515792847,
      "logps/chosen": -134.6277618408203,
      "logps/ref_chosen": -42.576805114746094,
      "logps/ref_rejected": -87.38154602050781,
      "logps/rejected": -322.1723937988281,
      "loss": 0.4484,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.377791404724121,
      "rewards/margins": 2.128697395324707,
      "rewards/rejected": -3.506488800048828,
      "step": 249
    },
    {
      "epoch": 0.3671071953010279,
      "epsilon_dpo/beta": 0.014812478795647621,
      "epsilon_dpo/beta_margin_grad_mean": -0.16431747376918793,
      "epsilon_dpo/beta_margin_grad_std": 0.16239361464977264,
      "epsilon_dpo/beta_margin_mean": 2.1128971576690674,
      "epsilon_dpo/beta_margin_std": 1.335999608039856,
      "epsilon_dpo/loss_margin_mean": 142.83154296875,
      "grad_norm": 54.212379455566406,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.014946449548006058,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 4.006586590948141e-07,
      "logits/chosen": -2.1342391967773438,
      "logits/rejected": -1.7132606506347656,
      "logps/chosen": -126.69745635986328,
      "logps/ref_chosen": -43.33977508544922,
      "logps/ref_rejected": -79.84855651855469,
      "logps/rejected": -306.03778076171875,
      "loss": 0.4175,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.2351545095443726,
      "rewards/margins": 2.1128971576690674,
      "rewards/rejected": -3.3480517864227295,
      "step": 250
    },
    {
      "epoch": 0.368575624082232,
      "epsilon_dpo/beta": 0.014693334698677063,
      "epsilon_dpo/beta_margin_grad_mean": -0.1908206194639206,
      "epsilon_dpo/beta_margin_grad_std": 0.17475423216819763,
      "epsilon_dpo/beta_margin_mean": 1.979662299156189,
      "epsilon_dpo/beta_margin_std": 1.4761016368865967,
      "epsilon_dpo/loss_margin_mean": 135.0380859375,
      "grad_norm": 58.540199279785156,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.01481221430003643,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 3.9963255888117325e-07,
      "logits/chosen": -2.064459800720215,
      "logits/rejected": -1.8460227251052856,
      "logps/chosen": -128.45263671875,
      "logps/ref_chosen": -37.8934211730957,
      "logps/ref_rejected": -82.71955871582031,
      "logps/rejected": -308.31683349609375,
      "loss": 0.4894,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.332367181777954,
      "rewards/margins": 1.979662299156189,
      "rewards/rejected": -3.3120296001434326,
      "step": 251
    },
    {
      "epoch": 0.3700440528634361,
      "epsilon_dpo/beta": 0.014561137184500694,
      "epsilon_dpo/beta_margin_grad_mean": -0.16394339501857758,
      "epsilon_dpo/beta_margin_grad_std": 0.15933236479759216,
      "epsilon_dpo/beta_margin_mean": 2.167374849319458,
      "epsilon_dpo/beta_margin_std": 1.4150327444076538,
      "epsilon_dpo/loss_margin_mean": 149.01345825195312,
      "grad_norm": 61.02168273925781,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.014692834578454494,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 3.9860251571044666e-07,
      "logits/chosen": -2.0972189903259277,
      "logits/rejected": -1.8574435710906982,
      "logps/chosen": -143.55337524414062,
      "logps/ref_chosen": -49.172019958496094,
      "logps/ref_rejected": -91.81843566894531,
      "logps/rejected": -335.2132568359375,
      "loss": 0.4052,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.3763771057128906,
      "rewards/margins": 2.167374610900879,
      "rewards/rejected": -3.5437517166137695,
      "step": 252
    },
    {
      "epoch": 0.37151248164464024,
      "epsilon_dpo/beta": 0.014448565430939198,
      "epsilon_dpo/beta_margin_grad_mean": -0.1948602944612503,
      "epsilon_dpo/beta_margin_grad_std": 0.1849391907453537,
      "epsilon_dpo/beta_margin_mean": 1.9554654359817505,
      "epsilon_dpo/beta_margin_std": 1.4330785274505615,
      "epsilon_dpo/loss_margin_mean": 135.68392944335938,
      "grad_norm": 65.06855773925781,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.014560877345502377,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 3.9756855672522986e-07,
      "logits/chosen": -2.0035042762756348,
      "logits/rejected": -2.029510021209717,
      "logps/chosen": -156.42469787597656,
      "logps/ref_chosen": -55.18561553955078,
      "logps/ref_rejected": -104.15153503417969,
      "logps/rejected": -341.0745544433594,
      "loss": 0.5018,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -1.4654229879379272,
      "rewards/margins": 1.95546555519104,
      "rewards/rejected": -3.4208884239196777,
      "step": 253
    },
    {
      "epoch": 0.37298091042584436,
      "epsilon_dpo/beta": 0.014350106939673424,
      "epsilon_dpo/beta_margin_grad_mean": -0.19600307941436768,
      "epsilon_dpo/beta_margin_grad_std": 0.19512991607189178,
      "epsilon_dpo/beta_margin_mean": 2.031984806060791,
      "epsilon_dpo/beta_margin_std": 1.5720576047897339,
      "epsilon_dpo/loss_margin_mean": 142.08840942382812,
      "grad_norm": 57.888275146484375,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.014448001980781555,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 3.965307091713037e-07,
      "logits/chosen": -2.0362446308135986,
      "logits/rejected": -1.8677754402160645,
      "logps/chosen": -133.6912078857422,
      "logps/ref_chosen": -44.80467224121094,
      "logps/ref_rejected": -93.50021362304688,
      "logps/rejected": -324.47515869140625,
      "loss": 0.5136,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -1.277719259262085,
      "rewards/margins": 2.031984806060791,
      "rewards/rejected": -3.309704065322876,
      "step": 254
    },
    {
      "epoch": 0.3744493392070485,
      "epsilon_dpo/beta": 0.01423418615013361,
      "epsilon_dpo/beta_margin_grad_mean": -0.16781477630138397,
      "epsilon_dpo/beta_margin_grad_std": 0.1848216950893402,
      "epsilon_dpo/beta_margin_mean": 2.3087897300720215,
      "epsilon_dpo/beta_margin_std": 1.6275063753128052,
      "epsilon_dpo/loss_margin_mean": 162.57176208496094,
      "grad_norm": 61.83622360229492,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.014349350705742836,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 3.954890003969163e-07,
      "logits/chosen": -1.7996987104415894,
      "logits/rejected": -1.752530574798584,
      "logps/chosen": -130.06533813476562,
      "logps/ref_chosen": -37.239234924316406,
      "logps/ref_rejected": -96.95054626464844,
      "logps/rejected": -352.3484191894531,
      "loss": 0.4448,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.3225433826446533,
      "rewards/margins": 2.3087897300720215,
      "rewards/rejected": -3.631333112716675,
      "step": 255
    },
    {
      "epoch": 0.37591776798825255,
      "epsilon_dpo/beta": 0.014110567979514599,
      "epsilon_dpo/beta_margin_grad_mean": -0.14517581462860107,
      "epsilon_dpo/beta_margin_grad_std": 0.1478438377380371,
      "epsilon_dpo/beta_margin_mean": 2.3481807708740234,
      "epsilon_dpo/beta_margin_std": 1.3911449909210205,
      "epsilon_dpo/loss_margin_mean": 166.6383819580078,
      "grad_norm": 47.72064208984375,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.01423370186239481,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 3.944434578520628e-07,
      "logits/chosen": -1.8363198041915894,
      "logits/rejected": -1.8857228755950928,
      "logps/chosen": -118.48297119140625,
      "logps/ref_chosen": -35.025508880615234,
      "logps/ref_rejected": -99.25279998779297,
      "logps/rejected": -349.3486328125,
      "loss": 0.3517,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.1789929866790771,
      "rewards/margins": 2.3481807708740234,
      "rewards/rejected": -3.5271737575531006,
      "step": 256
    },
    {
      "epoch": 0.37738619676945667,
      "epsilon_dpo/beta": 0.013992580585181713,
      "epsilon_dpo/beta_margin_grad_mean": -0.1678382307291031,
      "epsilon_dpo/beta_margin_grad_std": 0.16572539508342743,
      "epsilon_dpo/beta_margin_mean": 2.1580240726470947,
      "epsilon_dpo/beta_margin_std": 1.4499742984771729,
      "epsilon_dpo/loss_margin_mean": 154.5001983642578,
      "grad_norm": 58.855709075927734,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.014110236428678036,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.933941090877615e-07,
      "logits/chosen": -1.8602499961853027,
      "logits/rejected": -1.7255544662475586,
      "logps/chosen": -125.66242980957031,
      "logps/ref_chosen": -34.74375534057617,
      "logps/ref_rejected": -84.338134765625,
      "logps/rejected": -329.75701904296875,
      "loss": 0.4202,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.2731714248657227,
      "rewards/margins": 2.1580240726470947,
      "rewards/rejected": -3.4311952590942383,
      "step": 257
    },
    {
      "epoch": 0.3788546255506608,
      "epsilon_dpo/beta": 0.013875506818294525,
      "epsilon_dpo/beta_margin_grad_mean": -0.16494029760360718,
      "epsilon_dpo/beta_margin_grad_std": 0.19155991077423096,
      "epsilon_dpo/beta_margin_mean": 2.2436838150024414,
      "epsilon_dpo/beta_margin_std": 1.5077780485153198,
      "epsilon_dpo/loss_margin_mean": 162.05101013183594,
      "grad_norm": 56.01180648803711,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.01399217825382948,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.923409817553284e-07,
      "logits/chosen": -1.829561710357666,
      "logits/rejected": -1.8402704000473022,
      "logps/chosen": -128.53773498535156,
      "logps/ref_chosen": -38.32011032104492,
      "logps/ref_rejected": -104.34061431884766,
      "logps/rejected": -356.6092529296875,
      "loss": 0.4481,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.2534571886062622,
      "rewards/margins": 2.2436838150024414,
      "rewards/rejected": -3.497140884399414,
      "step": 258
    },
    {
      "epoch": 0.3803230543318649,
      "epsilon_dpo/beta": 0.013755074702203274,
      "epsilon_dpo/beta_margin_grad_mean": -0.17741423845291138,
      "epsilon_dpo/beta_margin_grad_std": 0.16296601295471191,
      "epsilon_dpo/beta_margin_mean": 2.0163936614990234,
      "epsilon_dpo/beta_margin_std": 1.3358848094940186,
      "epsilon_dpo/loss_margin_mean": 146.82223510742188,
      "grad_norm": 53.66272735595703,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.01387510634958744,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 3.9128410360564793e-07,
      "logits/chosen": -1.766379952430725,
      "logits/rejected": -1.7046854496002197,
      "logps/chosen": -133.28622436523438,
      "logps/ref_chosen": -37.96626663208008,
      "logps/ref_rejected": -94.62816619873047,
      "logps/rejected": -336.7703552246094,
      "loss": 0.4435,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.3125896453857422,
      "rewards/margins": 2.0163938999176025,
      "rewards/rejected": -3.3289833068847656,
      "step": 259
    },
    {
      "epoch": 0.38179148311306904,
      "epsilon_dpo/beta": 0.01364436000585556,
      "epsilon_dpo/beta_margin_grad_mean": -0.18185892701148987,
      "epsilon_dpo/beta_margin_grad_std": 0.1839314103126526,
      "epsilon_dpo/beta_margin_mean": 2.0660133361816406,
      "epsilon_dpo/beta_margin_std": 1.5317002534866333,
      "epsilon_dpo/loss_margin_mean": 151.79209899902344,
      "grad_norm": 71.4144287109375,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.013754752464592457,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 3.9022350248844246e-07,
      "logits/chosen": -1.7605104446411133,
      "logits/rejected": -1.844254732131958,
      "logps/chosen": -128.06361389160156,
      "logps/ref_chosen": -29.670434951782227,
      "logps/ref_rejected": -101.38003540039062,
      "logps/rejected": -351.5653076171875,
      "loss": 0.4876,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.3459320068359375,
      "rewards/margins": 2.0660133361816406,
      "rewards/rejected": -3.411945343017578,
      "step": 260
    },
    {
      "epoch": 0.3832599118942731,
      "epsilon_dpo/beta": 0.013534392230212688,
      "epsilon_dpo/beta_margin_grad_mean": -0.14695821702480316,
      "epsilon_dpo/beta_margin_grad_std": 0.183539479970932,
      "epsilon_dpo/beta_margin_mean": 2.5458641052246094,
      "epsilon_dpo/beta_margin_std": 1.7510801553726196,
      "epsilon_dpo/loss_margin_mean": 188.52330017089844,
      "grad_norm": 66.16899871826172,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.013643896207213402,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 3.891592063515376e-07,
      "logits/chosen": -1.7042864561080933,
      "logits/rejected": -1.5886824131011963,
      "logps/chosen": -113.61106872558594,
      "logps/ref_chosen": -31.892623901367188,
      "logps/ref_rejected": -94.94419860839844,
      "logps/rejected": -365.1859436035156,
      "loss": 0.396,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.1082427501678467,
      "rewards/margins": 2.5458641052246094,
      "rewards/rejected": -3.654106855392456,
      "step": 261
    },
    {
      "epoch": 0.38472834067547723,
      "epsilon_dpo/beta": 0.013416852802038193,
      "epsilon_dpo/beta_margin_grad_mean": -0.16779224574565887,
      "epsilon_dpo/beta_margin_grad_std": 0.15692253410816193,
      "epsilon_dpo/beta_margin_mean": 2.050642251968384,
      "epsilon_dpo/beta_margin_std": 1.2614741325378418,
      "epsilon_dpo/loss_margin_mean": 153.06997680664062,
      "grad_norm": 56.77939224243164,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.013533933088183403,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 3.880912432401264e-07,
      "logits/chosen": -1.6830520629882812,
      "logits/rejected": -1.5291244983673096,
      "logps/chosen": -134.6734619140625,
      "logps/ref_chosen": -34.26129913330078,
      "logps/ref_rejected": -90.29788208007812,
      "logps/rejected": -343.780029296875,
      "loss": 0.4156,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.348780632019043,
      "rewards/margins": 2.050642251968384,
      "rewards/rejected": -3.3994228839874268,
      "step": 262
    },
    {
      "epoch": 0.38619676945668135,
      "epsilon_dpo/beta": 0.013304665684700012,
      "epsilon_dpo/beta_margin_grad_mean": -0.14855755865573883,
      "epsilon_dpo/beta_margin_grad_std": 0.16537067294120789,
      "epsilon_dpo/beta_margin_mean": 2.3461291790008545,
      "epsilon_dpo/beta_margin_std": 1.4775606393814087,
      "epsilon_dpo/loss_margin_mean": 176.666748046875,
      "grad_norm": 59.517337799072266,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.013416538015007973,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.870196412960302e-07,
      "logits/chosen": -1.8858226537704468,
      "logits/rejected": -1.5893621444702148,
      "logps/chosen": -131.9771728515625,
      "logps/ref_chosen": -39.4767951965332,
      "logps/ref_rejected": -102.44886779785156,
      "logps/rejected": -371.615966796875,
      "loss": 0.3749,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.233377456665039,
      "rewards/margins": 2.3461294174194336,
      "rewards/rejected": -3.5795066356658936,
      "step": 263
    },
    {
      "epoch": 0.3876651982378855,
      "epsilon_dpo/beta": 0.01318918913602829,
      "epsilon_dpo/beta_margin_grad_mean": -0.15769457817077637,
      "epsilon_dpo/beta_margin_grad_std": 0.15815286338329315,
      "epsilon_dpo/beta_margin_mean": 2.2963368892669678,
      "epsilon_dpo/beta_margin_std": 1.5442652702331543,
      "epsilon_dpo/loss_margin_mean": 174.36825561523438,
      "grad_norm": 60.96957015991211,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.013304282911121845,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 3.8594442875695665e-07,
      "logits/chosen": -1.6896053552627563,
      "logits/rejected": -1.7085572481155396,
      "logps/chosen": -123.23445129394531,
      "logps/ref_chosen": -38.707645416259766,
      "logps/ref_rejected": -100.15180969238281,
      "logps/rejected": -359.046875,
      "loss": 0.3936,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.1168205738067627,
      "rewards/margins": 2.296337127685547,
      "rewards/rejected": -3.4131574630737305,
      "step": 264
    },
    {
      "epoch": 0.3891336270190896,
      "epsilon_dpo/beta": 0.013070664368569851,
      "epsilon_dpo/beta_margin_grad_mean": -0.17950545251369476,
      "epsilon_dpo/beta_margin_grad_std": 0.168075293302536,
      "epsilon_dpo/beta_margin_mean": 2.0962088108062744,
      "epsilon_dpo/beta_margin_std": 1.5082602500915527,
      "epsilon_dpo/loss_margin_mean": 160.54234313964844,
      "grad_norm": 61.71824645996094,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.013188880868256092,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 3.848656339557562e-07,
      "logits/chosen": -1.6642320156097412,
      "logits/rejected": -1.5806505680084229,
      "logps/chosen": -145.28750610351562,
      "logps/ref_chosen": -37.51420593261719,
      "logps/ref_rejected": -94.66896057128906,
      "logps/rejected": -362.984619140625,
      "loss": 0.4478,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.4101543426513672,
      "rewards/margins": 2.0962088108062744,
      "rewards/rejected": -3.5063629150390625,
      "step": 265
    },
    {
      "epoch": 0.39060205580029367,
      "epsilon_dpo/beta": 0.012973698787391186,
      "epsilon_dpo/beta_margin_grad_mean": -0.20105549693107605,
      "epsilon_dpo/beta_margin_grad_std": 0.19386924803256989,
      "epsilon_dpo/beta_margin_mean": 1.843229055404663,
      "epsilon_dpo/beta_margin_std": 1.3907108306884766,
      "epsilon_dpo/loss_margin_mean": 142.52218627929688,
      "grad_norm": 68.82777404785156,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.013070429675281048,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 3.8378328531967507e-07,
      "logits/chosen": -1.6910204887390137,
      "logits/rejected": -1.2894883155822754,
      "logps/chosen": -167.84719848632812,
      "logps/ref_chosen": -51.23357391357422,
      "logps/ref_rejected": -73.29232025146484,
      "logps/rejected": -332.4281311035156,
      "loss": 0.5442,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.5156571865081787,
      "rewards/margins": 1.843229055404663,
      "rewards/rejected": -3.358886241912842,
      "step": 266
    },
    {
      "epoch": 0.3920704845814978,
      "epsilon_dpo/beta": 0.012881174683570862,
      "epsilon_dpo/beta_margin_grad_mean": -0.19470059871673584,
      "epsilon_dpo/beta_margin_grad_std": 0.1797318160533905,
      "epsilon_dpo/beta_margin_mean": 1.9295697212219238,
      "epsilon_dpo/beta_margin_std": 1.4559646844863892,
      "epsilon_dpo/loss_margin_mean": 150.2775115966797,
      "grad_norm": 67.35930633544922,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.012973131611943245,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 3.8269741136960646e-07,
      "logits/chosen": -1.8555903434753418,
      "logits/rejected": -1.507951021194458,
      "logps/chosen": -154.36102294921875,
      "logps/ref_chosen": -50.439453125,
      "logps/ref_rejected": -95.28913879394531,
      "logps/rejected": -349.48822021484375,
      "loss": 0.4993,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -1.3415794372558594,
      "rewards/margins": 1.9295697212219238,
      "rewards/rejected": -3.271149158477783,
      "step": 267
    },
    {
      "epoch": 0.3935389133627019,
      "epsilon_dpo/beta": 0.012789253145456314,
      "epsilon_dpo/beta_margin_grad_mean": -0.21375124156475067,
      "epsilon_dpo/beta_margin_grad_std": 0.20960654318332672,
      "epsilon_dpo/beta_margin_mean": 1.9788737297058105,
      "epsilon_dpo/beta_margin_std": 1.7514647245407104,
      "epsilon_dpo/loss_margin_mean": 155.29205322265625,
      "grad_norm": 80.12260437011719,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.012880552560091019,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 3.8160804071933894e-07,
      "logits/chosen": -1.6909327507019043,
      "logits/rejected": -1.6416935920715332,
      "logps/chosen": -156.72665405273438,
      "logps/ref_chosen": -42.84587097167969,
      "logps/ref_rejected": -108.26278686523438,
      "logps/rejected": -377.4356384277344,
      "loss": 0.5814,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -1.461806058883667,
      "rewards/margins": 1.9788737297058105,
      "rewards/rejected": -3.4406800270080566,
      "step": 268
    },
    {
      "epoch": 0.39500734214390604,
      "epsilon_dpo/beta": 0.012697985395789146,
      "epsilon_dpo/beta_margin_grad_mean": -0.15136900544166565,
      "epsilon_dpo/beta_margin_grad_std": 0.18286660313606262,
      "epsilon_dpo/beta_margin_mean": 2.484262228012085,
      "epsilon_dpo/beta_margin_std": 1.5870592594146729,
      "epsilon_dpo/loss_margin_mean": 196.23475646972656,
      "grad_norm": 59.1228141784668,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.012788633815944195,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 3.8051520207480204e-07,
      "logits/chosen": -1.7581638097763062,
      "logits/rejected": -1.7033562660217285,
      "logps/chosen": -142.43875122070312,
      "logps/ref_chosen": -43.99567413330078,
      "logps/ref_rejected": -115.62770080566406,
      "logps/rejected": -410.3055419921875,
      "loss": 0.3916,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.253930926322937,
      "rewards/margins": 2.484261989593506,
      "rewards/rejected": -3.7381930351257324,
      "step": 269
    },
    {
      "epoch": 0.3964757709251101,
      "epsilon_dpo/beta": 0.012599433772265911,
      "epsilon_dpo/beta_margin_grad_mean": -0.18488836288452148,
      "epsilon_dpo/beta_margin_grad_std": 0.2021288424730301,
      "epsilon_dpo/beta_margin_mean": 2.0731112957000732,
      "epsilon_dpo/beta_margin_std": 1.5216357707977295,
      "epsilon_dpo/loss_margin_mean": 165.01644897460938,
      "grad_norm": 79.74817657470703,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.012697371654212475,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 3.794189242333106e-07,
      "logits/chosen": -1.8194841146469116,
      "logits/rejected": -1.6801855564117432,
      "logps/chosen": -164.27764892578125,
      "logps/ref_chosen": -44.127079010009766,
      "logps/ref_rejected": -116.14840698242188,
      "logps/rejected": -401.3154296875,
      "loss": 0.512,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -1.5177865028381348,
      "rewards/margins": 2.0731112957000732,
      "rewards/rejected": -3.590897798538208,
      "step": 270
    },
    {
      "epoch": 0.39794419970631423,
      "epsilon_dpo/beta": 0.012493887916207314,
      "epsilon_dpo/beta_margin_grad_mean": -0.19335998594760895,
      "epsilon_dpo/beta_margin_grad_std": 0.19798798859119415,
      "epsilon_dpo/beta_margin_mean": 2.0564560890197754,
      "epsilon_dpo/beta_margin_std": 1.6271862983703613,
      "epsilon_dpo/loss_margin_mean": 164.9576873779297,
      "grad_norm": 58.63235855102539,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.012598942033946514,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.7831923608280514e-07,
      "logits/chosen": -1.7537434101104736,
      "logits/rejected": -1.6001136302947998,
      "logps/chosen": -133.10488891601562,
      "logps/ref_chosen": -35.64381408691406,
      "logps/ref_rejected": -97.9295883178711,
      "logps/rejected": -360.34832763671875,
      "loss": 0.5335,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.219379186630249,
      "rewards/margins": 2.0564560890197754,
      "rewards/rejected": -3.2758355140686035,
      "step": 271
    },
    {
      "epoch": 0.39941262848751835,
      "epsilon_dpo/beta": 0.012385448440909386,
      "epsilon_dpo/beta_margin_grad_mean": -0.12951304018497467,
      "epsilon_dpo/beta_margin_grad_std": 0.1683792918920517,
      "epsilon_dpo/beta_margin_mean": 2.5986411571502686,
      "epsilon_dpo/beta_margin_std": 1.48728609085083,
      "epsilon_dpo/loss_margin_mean": 210.13800048828125,
      "grad_norm": 43.1225471496582,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.012493528425693512,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 3.772161666010912e-07,
      "logits/chosen": -1.6689045429229736,
      "logits/rejected": -1.6010162830352783,
      "logps/chosen": -112.50562286376953,
      "logps/ref_chosen": -26.52655792236328,
      "logps/ref_rejected": -112.5738525390625,
      "logps/rejected": -408.69091796875,
      "loss": 0.336,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.065723180770874,
      "rewards/margins": 2.5986409187316895,
      "rewards/rejected": -3.6643643379211426,
      "step": 272
    },
    {
      "epoch": 0.4008810572687225,
      "epsilon_dpo/beta": 0.012266403064131737,
      "epsilon_dpo/beta_margin_grad_mean": -0.14163939654827118,
      "epsilon_dpo/beta_margin_grad_std": 0.14621266722679138,
      "epsilon_dpo/beta_margin_mean": 2.353142261505127,
      "epsilon_dpo/beta_margin_std": 1.3905528783798218,
      "epsilon_dpo/loss_margin_mean": 191.923583984375,
      "grad_norm": 50.07746505737305,
      "kl/avg_steps": 0.96875,
      "kl/beta": 0.012385157868266106,
      "kl/n_epsilon_steps": 0.015625,
      "kl/p_epsilon_steps": 0.984375,
      "learning_rate": 3.761097448550755e-07,
      "logits/chosen": -1.71974515914917,
      "logits/rejected": -1.6216449737548828,
      "logps/chosen": -137.0894317626953,
      "logps/ref_chosen": -34.63496017456055,
      "logps/ref_rejected": -97.36636352539062,
      "logps/rejected": -391.7444152832031,
      "loss": 0.3461,
      "rewards/accuracies": 0.984375,
      "rewards/chosen": -1.2572829723358154,
      "rewards/margins": 2.353142261505127,
      "rewards/rejected": -3.6104252338409424,
      "step": 273
    },
    {
      "epoch": 0.4023494860499266,
      "epsilon_dpo/beta": 0.012164046056568623,
      "epsilon_dpo/beta_margin_grad_mean": -0.1828116923570633,
      "epsilon_dpo/beta_margin_grad_std": 0.1835782825946808,
      "epsilon_dpo/beta_margin_mean": 2.068690061569214,
      "epsilon_dpo/beta_margin_std": 1.446864366531372,
      "epsilon_dpo/loss_margin_mean": 170.40960693359375,
      "grad_norm": 58.21686553955078,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.012266327627003193,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.75e-07,
      "logits/chosen": -1.5487819910049438,
      "logits/rejected": -1.4224430322647095,
      "logps/chosen": -150.64378356933594,
      "logps/ref_chosen": -32.576805114746094,
      "logps/ref_rejected": -83.13209533691406,
      "logps/rejected": -371.6086730957031,
      "loss": 0.4745,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.4374797344207764,
      "rewards/margins": 2.0686898231506348,
      "rewards/rejected": -3.5061697959899902,
      "step": 274
    },
    {
      "epoch": 0.40381791483113066,
      "epsilon_dpo/beta": 0.012066072784364223,
      "epsilon_dpo/beta_margin_grad_mean": -0.1988518238067627,
      "epsilon_dpo/beta_margin_grad_std": 0.19736242294311523,
      "epsilon_dpo/beta_margin_mean": 1.9684127569198608,
      "epsilon_dpo/beta_margin_std": 1.5511200428009033,
      "epsilon_dpo/loss_margin_mean": 163.5451202392578,
      "grad_norm": 58.6602783203125,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.012163696810603142,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 3.738869612786737e-07,
      "logits/chosen": -1.6947541236877441,
      "logits/rejected": -1.5999658107757568,
      "logps/chosen": -152.82989501953125,
      "logps/ref_chosen": -34.1890983581543,
      "logps/ref_rejected": -98.71599578857422,
      "logps/rejected": -380.90191650390625,
      "loss": 0.5343,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.4348502159118652,
      "rewards/margins": 1.9684127569198608,
      "rewards/rejected": -3.4032628536224365,
      "step": 275
    },
    {
      "epoch": 0.4052863436123348,
      "epsilon_dpo/beta": 0.011965055949985981,
      "epsilon_dpo/beta_margin_grad_mean": -0.2083517462015152,
      "epsilon_dpo/beta_margin_grad_std": 0.1770058423280716,
      "epsilon_dpo/beta_margin_mean": 1.7445530891418457,
      "epsilon_dpo/beta_margin_std": 1.3083961009979248,
      "epsilon_dpo/loss_margin_mean": 146.08535766601562,
      "grad_norm": 62.463478088378906,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.012065663002431393,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.7277065802070204e-07,
      "logits/chosen": -1.75152587890625,
      "logits/rejected": -1.3374929428100586,
      "logps/chosen": -188.8147735595703,
      "logps/ref_chosen": -47.06272888183594,
      "logps/ref_rejected": -76.37776947021484,
      "logps/rejected": -364.2151794433594,
      "loss": 0.537,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.6980412006378174,
      "rewards/margins": 1.7445529699325562,
      "rewards/rejected": -3.442594289779663,
      "step": 276
    },
    {
      "epoch": 0.4067547723935389,
      "epsilon_dpo/beta": 0.01186494529247284,
      "epsilon_dpo/beta_margin_grad_mean": -0.17211416363716125,
      "epsilon_dpo/beta_margin_grad_std": 0.16370359063148499,
      "epsilon_dpo/beta_margin_mean": 2.1374924182891846,
      "epsilon_dpo/beta_margin_std": 1.4325969219207764,
      "epsilon_dpo/loss_margin_mean": 180.449951171875,
      "grad_norm": 52.0661735534668,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.011964711360633373,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.71651119641714e-07,
      "logits/chosen": -1.622573971748352,
      "logits/rejected": -1.471212387084961,
      "logps/chosen": -159.37274169921875,
      "logps/ref_chosen": -35.24298095703125,
      "logps/ref_rejected": -99.66352844238281,
      "logps/rejected": -404.24322509765625,
      "loss": 0.4258,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.4755170345306396,
      "rewards/margins": 2.1374926567077637,
      "rewards/rejected": -3.613009452819824,
      "step": 277
    },
    {
      "epoch": 0.40822320117474303,
      "epsilon_dpo/beta": 0.011769380420446396,
      "epsilon_dpo/beta_margin_grad_mean": -0.18432171642780304,
      "epsilon_dpo/beta_margin_grad_std": 0.17036166787147522,
      "epsilon_dpo/beta_margin_mean": 2.0264534950256348,
      "epsilon_dpo/beta_margin_std": 1.5373589992523193,
      "epsilon_dpo/loss_margin_mean": 172.5458526611328,
      "grad_norm": 54.76714324951172,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.011864603497087955,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 3.705283756425872e-07,
      "logits/chosen": -1.7193242311477661,
      "logits/rejected": -1.5687487125396729,
      "logps/chosen": -169.8358154296875,
      "logps/ref_chosen": -38.0869140625,
      "logps/ref_rejected": -96.20486450195312,
      "logps/rejected": -400.4996337890625,
      "loss": 0.4658,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.5532996654510498,
      "rewards/margins": 2.0264534950256348,
      "rewards/rejected": -3.5797531604766846,
      "step": 278
    },
    {
      "epoch": 0.40969162995594716,
      "epsilon_dpo/beta": 0.011667169630527496,
      "epsilon_dpo/beta_margin_grad_mean": -0.17751751840114594,
      "epsilon_dpo/beta_margin_grad_std": 0.16590261459350586,
      "epsilon_dpo/beta_margin_mean": 2.106186628341675,
      "epsilon_dpo/beta_margin_std": 1.458591341972351,
      "epsilon_dpo/loss_margin_mean": 180.77297973632812,
      "grad_norm": 56.205360412597656,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.011768980883061886,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 3.6940245560867e-07,
      "logits/chosen": -1.4648022651672363,
      "logits/rejected": -1.4414713382720947,
      "logps/chosen": -151.58514404296875,
      "logps/ref_chosen": -29.908262252807617,
      "logps/ref_rejected": -93.1087646484375,
      "logps/rejected": -395.55865478515625,
      "loss": 0.4424,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -1.4212896823883057,
      "rewards/margins": 2.106186628341675,
      "rewards/rejected": -3.5274763107299805,
      "step": 279
    },
    {
      "epoch": 0.4111600587371512,
      "epsilon_dpo/beta": 0.01156961265951395,
      "epsilon_dpo/beta_margin_grad_mean": -0.16350157558918,
      "epsilon_dpo/beta_margin_grad_std": 0.16755622625350952,
      "epsilon_dpo/beta_margin_mean": 2.1660408973693848,
      "epsilon_dpo/beta_margin_std": 1.4027117490768433,
      "epsilon_dpo/loss_margin_mean": 187.58596801757812,
      "grad_norm": 52.6536865234375,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.011666894890367985,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.6827338920900253e-07,
      "logits/chosen": -1.6275845766067505,
      "logits/rejected": -1.6037421226501465,
      "logps/chosen": -165.87985229492188,
      "logps/ref_chosen": -40.1248664855957,
      "logps/ref_rejected": -105.53138732910156,
      "logps/rejected": -418.8723449707031,
      "loss": 0.417,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.4572311639785767,
      "rewards/margins": 2.1660408973693848,
      "rewards/rejected": -3.623271942138672,
      "step": 280
    },
    {
      "epoch": 0.41262848751835535,
      "epsilon_dpo/beta": 0.011469194665551186,
      "epsilon_dpo/beta_margin_grad_mean": -0.18068785965442657,
      "epsilon_dpo/beta_margin_grad_std": 0.1649240106344223,
      "epsilon_dpo/beta_margin_mean": 2.053178548812866,
      "epsilon_dpo/beta_margin_std": 1.504626750946045,
      "epsilon_dpo/loss_margin_mean": 179.28672790527344,
      "grad_norm": 54.2619743347168,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.011569279246032238,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 3.6714120619553435e-07,
      "logits/chosen": -1.6719520092010498,
      "logits/rejected": -1.367035150527954,
      "logps/chosen": -169.83157348632812,
      "logps/ref_chosen": -39.501502990722656,
      "logps/ref_rejected": -87.23008728027344,
      "logps/rejected": -396.84686279296875,
      "loss": 0.45,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.4972405433654785,
      "rewards/margins": 2.053178548812866,
      "rewards/rejected": -3.5504188537597656,
      "step": 281
    },
    {
      "epoch": 0.41409691629955947,
      "epsilon_dpo/beta": 0.011376879177987576,
      "epsilon_dpo/beta_margin_grad_mean": -0.18481247127056122,
      "epsilon_dpo/beta_margin_grad_std": 0.17733608186244965,
      "epsilon_dpo/beta_margin_mean": 1.9530349969863892,
      "epsilon_dpo/beta_margin_std": 1.3995331525802612,
      "epsilon_dpo/loss_margin_mean": 172.08128356933594,
      "grad_norm": 54.99889373779297,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.011468926444649696,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 3.660059364023408e-07,
      "logits/chosen": -1.6971466541290283,
      "logits/rejected": -1.5857012271881104,
      "logps/chosen": -174.48223876953125,
      "logps/ref_chosen": -46.00492858886719,
      "logps/ref_rejected": -101.88005828857422,
      "logps/rejected": -402.43865966796875,
      "loss": 0.4842,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.4642865657806396,
      "rewards/margins": 1.9530349969863892,
      "rewards/rejected": -3.4173216819763184,
      "step": 282
    },
    {
      "epoch": 0.4155653450807636,
      "epsilon_dpo/beta": 0.01127807516604662,
      "epsilon_dpo/beta_margin_grad_mean": -0.1464729756116867,
      "epsilon_dpo/beta_margin_grad_std": 0.15092232823371887,
      "epsilon_dpo/beta_margin_mean": 2.340468645095825,
      "epsilon_dpo/beta_margin_std": 1.3866698741912842,
      "epsilon_dpo/loss_margin_mean": 207.81991577148438,
      "grad_norm": 47.481971740722656,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.011376491747796535,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 3.6486760974483685e-07,
      "logits/chosen": -1.6741735935211182,
      "logits/rejected": -1.502656102180481,
      "logps/chosen": -153.2581329345703,
      "logps/ref_chosen": -35.3682861328125,
      "logps/ref_rejected": -103.27058410644531,
      "logps/rejected": -428.9803466796875,
      "loss": 0.3582,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.330566644668579,
      "rewards/margins": 2.340468645095825,
      "rewards/rejected": -3.6710352897644043,
      "step": 283
    },
    {
      "epoch": 0.4170337738619677,
      "epsilon_dpo/beta": 0.011197871528565884,
      "epsilon_dpo/beta_margin_grad_mean": -0.18323473632335663,
      "epsilon_dpo/beta_margin_grad_std": 0.1965819150209427,
      "epsilon_dpo/beta_margin_mean": 2.156057596206665,
      "epsilon_dpo/beta_margin_std": 1.6328465938568115,
      "epsilon_dpo/loss_margin_mean": 193.1970977783203,
      "grad_norm": 64.37258911132812,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.011277811601758003,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 3.6372625621898863e-07,
      "logits/chosen": -1.67313551902771,
      "logits/rejected": -1.5210282802581787,
      "logps/chosen": -168.32333374023438,
      "logps/ref_chosen": -41.10857391357422,
      "logps/ref_rejected": -99.55363464355469,
      "logps/rejected": -419.96551513671875,
      "loss": 0.4919,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.4272245168685913,
      "rewards/margins": 2.156057357788086,
      "rewards/rejected": -3.583281993865967,
      "step": 284
    },
    {
      "epoch": 0.4185022026431718,
      "epsilon_dpo/beta": 0.011096964590251446,
      "epsilon_dpo/beta_margin_grad_mean": -0.14801117777824402,
      "epsilon_dpo/beta_margin_grad_std": 0.17136132717132568,
      "epsilon_dpo/beta_margin_mean": 2.2969985008239746,
      "epsilon_dpo/beta_margin_std": 1.3974249362945557,
      "epsilon_dpo/loss_margin_mean": 207.27685546875,
      "grad_norm": 60.48746871948242,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.011197330430150032,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 3.625819059005228e-07,
      "logits/chosen": -1.64085054397583,
      "logits/rejected": -1.4849854707717896,
      "logps/chosen": -163.57217407226562,
      "logps/ref_chosen": -35.757301330566406,
      "logps/ref_rejected": -108.66427612304688,
      "logps/rejected": -443.7559814453125,
      "loss": 0.4048,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.4201911687850952,
      "rewards/margins": 2.2969985008239746,
      "rewards/rejected": -3.7171897888183594,
      "step": 285
    },
    {
      "epoch": 0.4199706314243759,
      "epsilon_dpo/beta": 0.011004237458109856,
      "epsilon_dpo/beta_margin_grad_mean": -0.16391685605049133,
      "epsilon_dpo/beta_margin_grad_std": 0.18253783881664276,
      "epsilon_dpo/beta_margin_mean": 2.2222766876220703,
      "epsilon_dpo/beta_margin_std": 1.464618444442749,
      "epsilon_dpo/loss_margin_mean": 202.3627471923828,
      "grad_norm": 60.148529052734375,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.011096766218543053,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.614345889441346e-07,
      "logits/chosen": -1.7381117343902588,
      "logits/rejected": -1.4690245389938354,
      "logps/chosen": -175.14715576171875,
      "logps/ref_chosen": -45.06391525268555,
      "logps/ref_rejected": -95.78263854980469,
      "logps/rejected": -428.2286376953125,
      "loss": 0.4342,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.4335534572601318,
      "rewards/margins": 2.2222766876220703,
      "rewards/rejected": -3.6558303833007812,
      "step": 286
    },
    {
      "epoch": 0.42143906020558003,
      "epsilon_dpo/beta": 0.010929361917078495,
      "epsilon_dpo/beta_margin_grad_mean": -0.23106051981449127,
      "epsilon_dpo/beta_margin_grad_std": 0.19667471945285797,
      "epsilon_dpo/beta_margin_mean": 1.6739410161972046,
      "epsilon_dpo/beta_margin_std": 1.4721882343292236,
      "epsilon_dpo/loss_margin_mean": 153.76026916503906,
      "grad_norm": 71.68363189697266,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.011003920808434486,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 3.6028433558269275e-07,
      "logits/chosen": -1.7347445487976074,
      "logits/rejected": -1.4353196620941162,
      "logps/chosen": -186.29312133789062,
      "logps/ref_chosen": -44.68206787109375,
      "logps/ref_rejected": -82.90010070800781,
      "logps/rejected": -378.27142333984375,
      "loss": 0.611,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -1.551443099975586,
      "rewards/margins": 1.6739410161972046,
      "rewards/rejected": -3.22538423538208,
      "step": 287
    },
    {
      "epoch": 0.42290748898678415,
      "epsilon_dpo/beta": 0.010827410966157913,
      "epsilon_dpo/beta_margin_grad_mean": -0.15691792964935303,
      "epsilon_dpo/beta_margin_grad_std": 0.16194118559360504,
      "epsilon_dpo/beta_margin_mean": 2.1758952140808105,
      "epsilon_dpo/beta_margin_std": 1.2892285585403442,
      "epsilon_dpo/loss_margin_mean": 201.15391540527344,
      "grad_norm": 56.249759674072266,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.010928785428404808,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 3.5913117612644327e-07,
      "logits/chosen": -1.6811950206756592,
      "logits/rejected": -1.4292771816253662,
      "logps/chosen": -160.89662170410156,
      "logps/ref_chosen": -35.92053985595703,
      "logps/ref_rejected": -92.28993225097656,
      "logps/rejected": -418.419921875,
      "loss": 0.3983,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -1.3546841144561768,
      "rewards/margins": 2.1758952140808105,
      "rewards/rejected": -3.5305793285369873,
      "step": 288
    },
    {
      "epoch": 0.4243759177679883,
      "epsilon_dpo/beta": 0.010733614675700665,
      "epsilon_dpo/beta_margin_grad_mean": -0.19883452355861664,
      "epsilon_dpo/beta_margin_grad_std": 0.18830937147140503,
      "epsilon_dpo/beta_margin_mean": 1.978835105895996,
      "epsilon_dpo/beta_margin_std": 1.5322097539901733,
      "epsilon_dpo/loss_margin_mean": 184.6563262939453,
      "grad_norm": 53.693756103515625,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.010827279649674892,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 3.5797514096221024e-07,
      "logits/chosen": -1.7554616928100586,
      "logits/rejected": -1.5511056184768677,
      "logps/chosen": -151.41265869140625,
      "logps/ref_chosen": -37.65406036376953,
      "logps/ref_rejected": -92.58161163330078,
      "logps/rejected": -390.99652099609375,
      "loss": 0.5158,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.222680687904358,
      "rewards/margins": 1.978835105895996,
      "rewards/rejected": -3.2015156745910645,
      "step": 289
    },
    {
      "epoch": 0.42584434654919234,
      "epsilon_dpo/beta": 0.01063715573400259,
      "epsilon_dpo/beta_margin_grad_mean": -0.17113091051578522,
      "epsilon_dpo/beta_margin_grad_std": 0.15478722751140594,
      "epsilon_dpo/beta_margin_mean": 2.094456911087036,
      "epsilon_dpo/beta_margin_std": 1.3732136487960815,
      "epsilon_dpo/loss_margin_mean": 197.09515380859375,
      "grad_norm": 51.92376708984375,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.010733362287282944,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 3.568162605525952e-07,
      "logits/chosen": -1.6529500484466553,
      "logits/rejected": -1.7434873580932617,
      "logps/chosen": -165.3158416748047,
      "logps/ref_chosen": -43.256103515625,
      "logps/ref_rejected": -123.40228271484375,
      "logps/rejected": -442.55718994140625,
      "loss": 0.4168,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.2994060516357422,
      "rewards/margins": 2.094456911087036,
      "rewards/rejected": -3.3938629627227783,
      "step": 290
    },
    {
      "epoch": 0.42731277533039647,
      "epsilon_dpo/beta": 0.010556564666330814,
      "epsilon_dpo/beta_margin_grad_mean": -0.1979316920042038,
      "epsilon_dpo/beta_margin_grad_std": 0.18217770755290985,
      "epsilon_dpo/beta_margin_mean": 1.9196021556854248,
      "epsilon_dpo/beta_margin_std": 1.422903060913086,
      "epsilon_dpo/loss_margin_mean": 182.32635498046875,
      "grad_norm": 62.94866943359375,
      "kl/avg_steps": 0.765625,
      "kl/beta": 0.010636964812874794,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 3.5565456543517485e-07,
      "logits/chosen": -1.8479794263839722,
      "logits/rejected": -1.6821517944335938,
      "logps/chosen": -165.5596466064453,
      "logps/ref_chosen": -43.823760986328125,
      "logps/ref_rejected": -94.49006652832031,
      "logps/rejected": -398.55230712890625,
      "loss": 0.5064,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.2878754138946533,
      "rewards/margins": 1.9196021556854248,
      "rewards/rejected": -3.207477569580078,
      "step": 291
    },
    {
      "epoch": 0.4287812041116006,
      "epsilon_dpo/beta": 0.010471423156559467,
      "epsilon_dpo/beta_margin_grad_mean": -0.1748674064874649,
      "epsilon_dpo/beta_margin_grad_std": 0.17317946255207062,
      "epsilon_dpo/beta_margin_mean": 2.048370122909546,
      "epsilon_dpo/beta_margin_std": 1.3248964548110962,
      "epsilon_dpo/loss_margin_mean": 196.04180908203125,
      "grad_norm": 62.74113464355469,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.010556144639849663,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 3.5449008622169583e-07,
      "logits/chosen": -1.7095630168914795,
      "logits/rejected": -1.6350113153457642,
      "logps/chosen": -141.83103942871094,
      "logps/ref_chosen": -35.38202667236328,
      "logps/ref_rejected": -94.85586547851562,
      "logps/rejected": -397.3466796875,
      "loss": 0.4426,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.1173462867736816,
      "rewards/margins": 2.048370361328125,
      "rewards/rejected": -3.1657166481018066,
      "step": 292
    },
    {
      "epoch": 0.4302496328928047,
      "epsilon_dpo/beta": 0.010403390973806381,
      "epsilon_dpo/beta_margin_grad_mean": -0.2334311306476593,
      "epsilon_dpo/beta_margin_grad_std": 0.19696100056171417,
      "epsilon_dpo/beta_margin_mean": 1.6216365098953247,
      "epsilon_dpo/beta_margin_std": 1.3715225458145142,
      "epsilon_dpo/loss_margin_mean": 156.5295867919922,
      "grad_norm": 59.1533088684082,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.010471067391335964,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 3.5332285359726846e-07,
      "logits/chosen": -1.8089414834976196,
      "logits/rejected": -1.613842248916626,
      "logps/chosen": -150.8321075439453,
      "logps/ref_chosen": -36.39015197753906,
      "logps/ref_rejected": -83.55977630615234,
      "logps/rejected": -354.53131103515625,
      "loss": 0.6168,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -1.193084478378296,
      "rewards/margins": 1.6216366291046143,
      "rewards/rejected": -2.81472110748291,
      "step": 293
    },
    {
      "epoch": 0.43171806167400884,
      "epsilon_dpo/beta": 0.01031605713069439,
      "epsilon_dpo/beta_margin_grad_mean": -0.20596943795681,
      "epsilon_dpo/beta_margin_grad_std": 0.16830576956272125,
      "epsilon_dpo/beta_margin_mean": 1.6836501359939575,
      "epsilon_dpo/beta_margin_std": 1.2057768106460571,
      "epsilon_dpo/loss_margin_mean": 163.5525360107422,
      "grad_norm": 53.15205001831055,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.010402798652648926,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.5215289831955786e-07,
      "logits/chosen": -1.6690821647644043,
      "logits/rejected": -1.641129732131958,
      "logps/chosen": -139.99928283691406,
      "logps/ref_chosen": -32.32667541503906,
      "logps/ref_rejected": -87.2591552734375,
      "logps/rejected": -358.48431396484375,
      "loss": 0.5324,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.1122854948043823,
      "rewards/margins": 1.6836501359939575,
      "rewards/rejected": -2.79593563079834,
      "step": 294
    },
    {
      "epoch": 0.4331864904552129,
      "epsilon_dpo/beta": 0.010229743085801601,
      "epsilon_dpo/beta_margin_grad_mean": -0.2001977413892746,
      "epsilon_dpo/beta_margin_grad_std": 0.19301089644432068,
      "epsilon_dpo/beta_margin_mean": 1.9150567054748535,
      "epsilon_dpo/beta_margin_std": 1.5144377946853638,
      "epsilon_dpo/loss_margin_mean": 187.62313842773438,
      "grad_norm": 73.17021179199219,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.010315759107470512,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.509802512179737e-07,
      "logits/chosen": -1.7157707214355469,
      "logits/rejected": -1.6699479818344116,
      "logps/chosen": -145.589599609375,
      "logps/ref_chosen": -32.976951599121094,
      "logps/ref_rejected": -96.25013732910156,
      "logps/rejected": -396.48590087890625,
      "loss": 0.5393,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.1549984216690063,
      "rewards/margins": 1.9150567054748535,
      "rewards/rejected": -3.0700550079345703,
      "step": 295
    },
    {
      "epoch": 0.434654919236417,
      "epsilon_dpo/beta": 0.010147349908947945,
      "epsilon_dpo/beta_margin_grad_mean": -0.21665216982364655,
      "epsilon_dpo/beta_margin_grad_std": 0.17736758291721344,
      "epsilon_dpo/beta_margin_mean": 1.7295547723770142,
      "epsilon_dpo/beta_margin_std": 1.38618004322052,
      "epsilon_dpo/loss_margin_mean": 170.8570556640625,
      "grad_norm": 54.07697296142578,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.010229448787868023,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 3.498049431928577e-07,
      "logits/chosen": -1.944000005722046,
      "logits/rejected": -1.7023565769195557,
      "logps/chosen": -159.61997985839844,
      "logps/ref_chosen": -41.81062316894531,
      "logps/ref_rejected": -99.36541748046875,
      "logps/rejected": -388.03179931640625,
      "loss": 0.5562,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.1971094608306885,
      "rewards/margins": 1.7295548915863037,
      "rewards/rejected": -2.926664352416992,
      "step": 296
    },
    {
      "epoch": 0.43612334801762115,
      "epsilon_dpo/beta": 0.010062395595014095,
      "epsilon_dpo/beta_margin_grad_mean": -0.19078318774700165,
      "epsilon_dpo/beta_margin_grad_std": 0.15408045053482056,
      "epsilon_dpo/beta_margin_mean": 1.7881520986557007,
      "epsilon_dpo/beta_margin_std": 1.1349061727523804,
      "epsilon_dpo/loss_margin_mean": 178.01626586914062,
      "grad_norm": 52.08464431762695,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.010147004388272762,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.486270052146694e-07,
      "logits/chosen": -1.8551688194274902,
      "logits/rejected": -1.8467998504638672,
      "logps/chosen": -146.10855102539062,
      "logps/ref_chosen": -35.64509582519531,
      "logps/ref_rejected": -101.33485412597656,
      "logps/rejected": -389.8145751953125,
      "loss": 0.469,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.112633466720581,
      "rewards/margins": 1.7881522178649902,
      "rewards/rejected": -2.900785446166992,
      "step": 297
    },
    {
      "epoch": 0.43759177679882527,
      "epsilon_dpo/beta": 0.009971914812922478,
      "epsilon_dpo/beta_margin_grad_mean": -0.17282237112522125,
      "epsilon_dpo/beta_margin_grad_std": 0.14743012189865112,
      "epsilon_dpo/beta_margin_mean": 2.0463857650756836,
      "epsilon_dpo/beta_margin_std": 1.3721272945404053,
      "epsilon_dpo/loss_margin_mean": 205.4117889404297,
      "grad_norm": 47.372406005859375,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.010062105022370815,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 3.474464683231698e-07,
      "logits/chosen": -1.854588508605957,
      "logits/rejected": -1.974242091178894,
      "logps/chosen": -136.94308471679688,
      "logps/ref_chosen": -39.13259506225586,
      "logps/ref_rejected": -125.148193359375,
      "logps/rejected": -428.3704833984375,
      "loss": 0.4176,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -0.9769392013549805,
      "rewards/margins": 2.0463857650756836,
      "rewards/rejected": -3.023324966430664,
      "step": 298
    },
    {
      "epoch": 0.4390602055800294,
      "epsilon_dpo/beta": 0.009882355108857155,
      "epsilon_dpo/beta_margin_grad_mean": -0.17700159549713135,
      "epsilon_dpo/beta_margin_grad_std": 0.15176741778850555,
      "epsilon_dpo/beta_margin_mean": 1.9076223373413086,
      "epsilon_dpo/beta_margin_std": 1.1600240468978882,
      "epsilon_dpo/loss_margin_mean": 193.25885009765625,
      "grad_norm": 50.078041076660156,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.009971735998988152,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 3.462633636266041e-07,
      "logits/chosen": -1.7307016849517822,
      "logits/rejected": -1.819953203201294,
      "logps/chosen": -122.997314453125,
      "logps/ref_chosen": -28.626670837402344,
      "logps/ref_rejected": -87.74382781982422,
      "logps/rejected": -375.3733215332031,
      "loss": 0.4348,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.933772087097168,
      "rewards/margins": 1.9076223373413086,
      "rewards/rejected": -2.8413944244384766,
      "step": 299
    },
    {
      "epoch": 0.44052863436123346,
      "epsilon_dpo/beta": 0.009802866727113724,
      "epsilon_dpo/beta_margin_grad_mean": -0.1854911744594574,
      "epsilon_dpo/beta_margin_grad_std": 0.19396579265594482,
      "epsilon_dpo/beta_margin_mean": 2.040740728378296,
      "epsilon_dpo/beta_margin_std": 1.4987692832946777,
      "epsilon_dpo/loss_margin_mean": 208.71090698242188,
      "grad_norm": 58.53467559814453,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.009882179088890553,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 3.4507772230088147e-07,
      "logits/chosen": -1.7490193843841553,
      "logits/rejected": -1.7381911277770996,
      "logps/chosen": -144.44647216796875,
      "logps/ref_chosen": -33.894203186035156,
      "logps/ref_rejected": -103.88007354736328,
      "logps/rejected": -423.14324951171875,
      "loss": 0.4996,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.0862467288970947,
      "rewards/margins": 2.040740728378296,
      "rewards/rejected": -3.1269874572753906,
      "step": 300
    },
    {
      "epoch": 0.44052863436123346,
      "eval_epsilon_dpo/beta": 0.009748955257236958,
      "eval_epsilon_dpo/beta_margin_grad_mean": -0.29610493779182434,
      "eval_epsilon_dpo/beta_margin_grad_std": 0.20769952237606049,
      "eval_epsilon_dpo/beta_margin_mean": 1.1843825578689575,
      "eval_epsilon_dpo/beta_margin_std": 1.3051005601882935,
      "eval_epsilon_dpo/loss_margin_mean": 122.25267791748047,
      "eval_kl/n_epsilon_steps": 0.22174657881259918,
      "eval_kl/p_epsilon_steps": 0.778253436088562,
      "eval_logits/chosen": -1.9427273273468018,
      "eval_logits/rejected": -1.7790117263793945,
      "eval_logps/chosen": -225.67529296875,
      "eval_logps/ref_chosen": -68.29110717773438,
      "eval_logps/ref_rejected": -92.08038330078125,
      "eval_logps/rejected": -371.7172546386719,
      "eval_loss": 0.41313475370407104,
      "eval_rewards/accuracies": 0.7928082346916199,
      "eval_rewards/chosen": -1.5382959842681885,
      "eval_rewards/margins": 1.1843825578689575,
      "eval_rewards/rejected": -2.7226786613464355,
      "eval_runtime": 38.3815,
      "eval_samples_per_second": 60.941,
      "eval_steps_per_second": 1.928,
      "step": 300
    },
    {
      "epoch": 0.4419970631424376,
      "epsilon_dpo/beta": 0.009714669547975063,
      "epsilon_dpo/beta_margin_grad_mean": -0.16645051538944244,
      "epsilon_dpo/beta_margin_grad_std": 0.14359311759471893,
      "epsilon_dpo/beta_margin_mean": 1.9759942293167114,
      "epsilon_dpo/beta_margin_std": 1.1226637363433838,
      "epsilon_dpo/loss_margin_mean": 203.62547302246094,
      "grad_norm": 49.01324462890625,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.009802533313632011,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 3.4388957558875316e-07,
      "logits/chosen": -1.7920889854431152,
      "logits/rejected": -1.7111059427261353,
      "logps/chosen": -140.640380859375,
      "logps/ref_chosen": -35.052242279052734,
      "logps/ref_rejected": -98.54647827148438,
      "logps/rejected": -407.7601013183594,
      "loss": 0.402,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -1.0266605615615845,
      "rewards/margins": 1.9759942293167114,
      "rewards/rejected": -3.002654790878296,
      "step": 301
    },
    {
      "epoch": 0.4434654919236417,
      "epsilon_dpo/beta": 0.009633492678403854,
      "epsilon_dpo/beta_margin_grad_mean": -0.1874367892742157,
      "epsilon_dpo/beta_margin_grad_std": 0.1728639304637909,
      "epsilon_dpo/beta_margin_mean": 1.9052062034606934,
      "epsilon_dpo/beta_margin_std": 1.2975436449050903,
      "epsilon_dpo/loss_margin_mean": 198.14389038085938,
      "grad_norm": 63.02902603149414,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.00971449539065361,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.426989547989902e-07,
      "logits/chosen": -1.8670696020126343,
      "logits/rejected": -1.852245807647705,
      "logps/chosen": -124.57978820800781,
      "logps/ref_chosen": -31.218624114990234,
      "logps/ref_rejected": -103.73237609863281,
      "logps/rejected": -395.2374267578125,
      "loss": 0.4794,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.9009679555892944,
      "rewards/margins": 1.9052062034606934,
      "rewards/rejected": -2.8061742782592773,
      "step": 302
    },
    {
      "epoch": 0.44493392070484583,
      "epsilon_dpo/beta": 0.009558911435306072,
      "epsilon_dpo/beta_margin_grad_mean": -0.1947356015443802,
      "epsilon_dpo/beta_margin_grad_std": 0.19330711662769318,
      "epsilon_dpo/beta_margin_mean": 1.8838841915130615,
      "epsilon_dpo/beta_margin_std": 1.4101883172988892,
      "epsilon_dpo/loss_margin_mean": 197.6653594970703,
      "grad_norm": 64.47979736328125,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.009633215144276619,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 3.4150589130555773e-07,
      "logits/chosen": -1.9428577423095703,
      "logits/rejected": -1.761818289756775,
      "logps/chosen": -145.26632690429688,
      "logps/ref_chosen": -42.143306732177734,
      "logps/ref_rejected": -93.17671203613281,
      "logps/rejected": -393.965087890625,
      "loss": 0.5374,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.9886916875839233,
      "rewards/margins": 1.8838841915130615,
      "rewards/rejected": -2.8725757598876953,
      "step": 303
    },
    {
      "epoch": 0.44640234948604995,
      "epsilon_dpo/beta": 0.009478837251663208,
      "epsilon_dpo/beta_margin_grad_mean": -0.17279693484306335,
      "epsilon_dpo/beta_margin_grad_std": 0.15856152772903442,
      "epsilon_dpo/beta_margin_mean": 1.9841113090515137,
      "epsilon_dpo/beta_margin_std": 1.1967487335205078,
      "epsilon_dpo/loss_margin_mean": 209.67984008789062,
      "grad_norm": 50.6738395690918,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.0095585398375988,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.403104165467883e-07,
      "logits/chosen": -1.9802241325378418,
      "logits/rejected": -1.827461838722229,
      "logps/chosen": -144.84146118164062,
      "logps/ref_chosen": -42.97766876220703,
      "logps/ref_rejected": -97.31230926513672,
      "logps/rejected": -408.85595703125,
      "loss": 0.4258,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.9677666425704956,
      "rewards/margins": 1.9841113090515137,
      "rewards/rejected": -2.951878070831299,
      "step": 304
    },
    {
      "epoch": 0.447870778267254,
      "epsilon_dpo/beta": 0.009399528615176678,
      "epsilon_dpo/beta_margin_grad_mean": -0.1955670863389969,
      "epsilon_dpo/beta_margin_grad_std": 0.16656768321990967,
      "epsilon_dpo/beta_margin_mean": 1.814961552619934,
      "epsilon_dpo/beta_margin_std": 1.2615888118743896,
      "epsilon_dpo/loss_margin_mean": 193.46107482910156,
      "grad_norm": 49.391204833984375,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.009478564374148846,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.391125620245535e-07,
      "logits/chosen": -1.9485129117965698,
      "logits/rejected": -1.8468644618988037,
      "logps/chosen": -159.82958984375,
      "logps/ref_chosen": -46.15346145629883,
      "logps/ref_rejected": -98.42967224121094,
      "logps/rejected": -405.56689453125,
      "loss": 0.4923,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.0703450441360474,
      "rewards/margins": 1.8149614334106445,
      "rewards/rejected": -2.8853065967559814,
      "step": 305
    },
    {
      "epoch": 0.44933920704845814,
      "epsilon_dpo/beta": 0.009312070906162262,
      "epsilon_dpo/beta_margin_grad_mean": -0.17502400279045105,
      "epsilon_dpo/beta_margin_grad_std": 0.13139215111732483,
      "epsilon_dpo/beta_margin_mean": 1.8663511276245117,
      "epsilon_dpo/beta_margin_std": 1.0456689596176147,
      "epsilon_dpo/loss_margin_mean": 200.55418395996094,
      "grad_norm": 44.052284240722656,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.009399257600307465,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 3.3791235930343417e-07,
      "logits/chosen": -1.9102896451950073,
      "logits/rejected": -1.7865833044052124,
      "logps/chosen": -163.22915649414062,
      "logps/ref_chosen": -46.639854431152344,
      "logps/ref_rejected": -91.45211791992188,
      "logps/rejected": -408.5956115722656,
      "loss": 0.4142,
      "rewards/accuracies": 0.984375,
      "rewards/chosen": -1.0861380100250244,
      "rewards/margins": 1.8663511276245117,
      "rewards/rejected": -2.952488899230957,
      "step": 306
    },
    {
      "epoch": 0.45080763582966227,
      "epsilon_dpo/beta": 0.009248863905668259,
      "epsilon_dpo/beta_margin_grad_mean": -0.2182687371969223,
      "epsilon_dpo/beta_margin_grad_std": 0.1931067556142807,
      "epsilon_dpo/beta_margin_mean": 1.6918431520462036,
      "epsilon_dpo/beta_margin_std": 1.3128682374954224,
      "epsilon_dpo/loss_margin_mean": 183.6102294921875,
      "grad_norm": 60.684417724609375,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.009311958216130733,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 3.367098400098881e-07,
      "logits/chosen": -1.928983449935913,
      "logits/rejected": -1.8208749294281006,
      "logps/chosen": -150.8052978515625,
      "logps/ref_chosen": -44.40485382080078,
      "logps/ref_rejected": -92.84294128417969,
      "logps/rejected": -382.8536376953125,
      "loss": 0.5766,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.9867569208145142,
      "rewards/margins": 1.6918432712554932,
      "rewards/rejected": -2.6786000728607178,
      "step": 307
    },
    {
      "epoch": 0.4522760646108664,
      "epsilon_dpo/beta": 0.009177041240036488,
      "epsilon_dpo/beta_margin_grad_mean": -0.21642521023750305,
      "epsilon_dpo/beta_margin_grad_std": 0.18705737590789795,
      "epsilon_dpo/beta_margin_mean": 1.6384843587875366,
      "epsilon_dpo/beta_margin_std": 1.205591082572937,
      "epsilon_dpo/loss_margin_mean": 179.0634307861328,
      "grad_norm": 64.62168884277344,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.00924837589263916,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 3.355050358314172e-07,
      "logits/chosen": -1.9160457849502563,
      "logits/rejected": -1.7217934131622314,
      "logps/chosen": -164.06167602539062,
      "logps/ref_chosen": -37.731178283691406,
      "logps/ref_rejected": -85.77114868164062,
      "logps/rejected": -391.16510009765625,
      "loss": 0.5679,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.161744236946106,
      "rewards/margins": 1.638484239578247,
      "rewards/rejected": -2.8002285957336426,
      "step": 308
    },
    {
      "epoch": 0.45374449339207046,
      "epsilon_dpo/beta": 0.009103032760322094,
      "epsilon_dpo/beta_margin_grad_mean": -0.20181921124458313,
      "epsilon_dpo/beta_margin_grad_std": 0.18235483765602112,
      "epsilon_dpo/beta_margin_mean": 1.7625192403793335,
      "epsilon_dpo/beta_margin_std": 1.2436327934265137,
      "epsilon_dpo/loss_margin_mean": 194.1045379638672,
      "grad_norm": 56.110897064208984,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.00917668268084526,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 3.3429797851573183e-07,
      "logits/chosen": -1.9111260175704956,
      "logits/rejected": -1.7678544521331787,
      "logps/chosen": -139.52688598632812,
      "logps/ref_chosen": -35.310665130615234,
      "logps/ref_rejected": -86.65730285644531,
      "logps/rejected": -384.9780578613281,
      "loss": 0.5253,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.9513264894485474,
      "rewards/margins": 1.7625192403793335,
      "rewards/rejected": -2.713845729827881,
      "step": 309
    },
    {
      "epoch": 0.4552129221732746,
      "epsilon_dpo/beta": 0.009035356342792511,
      "epsilon_dpo/beta_margin_grad_mean": -0.2072700560092926,
      "epsilon_dpo/beta_margin_grad_std": 0.18662025034427643,
      "epsilon_dpo/beta_margin_mean": 1.7331546545028687,
      "epsilon_dpo/beta_margin_std": 1.237532138824463,
      "epsilon_dpo/loss_margin_mean": 192.4178924560547,
      "grad_norm": 57.408546447753906,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.009102723561227322,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 3.3308869986991487e-07,
      "logits/chosen": -1.9054917097091675,
      "logits/rejected": -1.831470012664795,
      "logps/chosen": -152.9647674560547,
      "logps/ref_chosen": -42.4453125,
      "logps/ref_rejected": -85.33100128173828,
      "logps/rejected": -388.26837158203125,
      "loss": 0.5391,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -1.0019582509994507,
      "rewards/margins": 1.7331546545028687,
      "rewards/rejected": -2.7351129055023193,
      "step": 310
    },
    {
      "epoch": 0.4566813509544787,
      "epsilon_dpo/beta": 0.008951152674853802,
      "epsilon_dpo/beta_margin_grad_mean": -0.1625707447528839,
      "epsilon_dpo/beta_margin_grad_std": 0.14155903458595276,
      "epsilon_dpo/beta_margin_mean": 2.1258339881896973,
      "epsilon_dpo/beta_margin_std": 1.3122553825378418,
      "epsilon_dpo/loss_margin_mean": 237.6621856689453,
      "grad_norm": 37.80696487426758,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.0090349605306983,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 3.3187723175958346e-07,
      "logits/chosen": -1.8577749729156494,
      "logits/rejected": -1.6029454469680786,
      "logps/chosen": -135.50538635253906,
      "logps/ref_chosen": -30.15520477294922,
      "logps/ref_rejected": -82.74742126464844,
      "logps/rejected": -425.759765625,
      "loss": 0.3892,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -0.9435994029045105,
      "rewards/margins": 2.1258339881896973,
      "rewards/rejected": -3.0694334506988525,
      "step": 311
    },
    {
      "epoch": 0.4581497797356828,
      "epsilon_dpo/beta": 0.008882002905011177,
      "epsilon_dpo/beta_margin_grad_mean": -0.209417924284935,
      "epsilon_dpo/beta_margin_grad_std": 0.18264873325824738,
      "epsilon_dpo/beta_margin_mean": 1.7365641593933105,
      "epsilon_dpo/beta_margin_std": 1.3034473657608032,
      "epsilon_dpo/loss_margin_mean": 196.04774475097656,
      "grad_norm": 56.1854362487793,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.008951044641435146,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 3.306636061080487e-07,
      "logits/chosen": -1.8750476837158203,
      "logits/rejected": -1.6363931894302368,
      "logps/chosen": -154.71817016601562,
      "logps/ref_chosen": -32.57682418823242,
      "logps/ref_rejected": -81.77426147460938,
      "logps/rejected": -399.96337890625,
      "loss": 0.5417,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -1.088057041168213,
      "rewards/margins": 1.7365641593933105,
      "rewards/rejected": -2.8246212005615234,
      "step": 312
    },
    {
      "epoch": 0.45961820851688695,
      "epsilon_dpo/beta": 0.008807598613202572,
      "epsilon_dpo/beta_margin_grad_mean": -0.19396935403347015,
      "epsilon_dpo/beta_margin_grad_std": 0.16426628828048706,
      "epsilon_dpo/beta_margin_mean": 1.8069427013397217,
      "epsilon_dpo/beta_margin_std": 1.2012414932250977,
      "epsilon_dpo/loss_margin_mean": 205.55426025390625,
      "grad_norm": 47.559181213378906,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.008881657384335995,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.2944785489547537e-07,
      "logits/chosen": -1.9602954387664795,
      "logits/rejected": -1.78814697265625,
      "logps/chosen": -147.1026153564453,
      "logps/ref_chosen": -28.45586395263672,
      "logps/ref_rejected": -93.60973358154297,
      "logps/rejected": -417.8107604980469,
      "loss": 0.4862,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.0461211204528809,
      "rewards/margins": 1.8069427013397217,
      "rewards/rejected": -2.8530638217926025,
      "step": 313
    },
    {
      "epoch": 0.461086637298091,
      "epsilon_dpo/beta": 0.008725648745894432,
      "epsilon_dpo/beta_margin_grad_mean": -0.19880293309688568,
      "epsilon_dpo/beta_margin_grad_std": 0.14619764685630798,
      "epsilon_dpo/beta_margin_mean": 1.7408623695373535,
      "epsilon_dpo/beta_margin_std": 1.1878901720046997,
      "epsilon_dpo/loss_margin_mean": 199.67056274414062,
      "grad_norm": 43.1519660949707,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.008807345293462276,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 3.2823001015803857e-07,
      "logits/chosen": -1.9734838008880615,
      "logits/rejected": -1.930768370628357,
      "logps/chosen": -144.1108856201172,
      "logps/ref_chosen": -34.89348220825195,
      "logps/ref_rejected": -102.7928695678711,
      "logps/rejected": -411.68084716796875,
      "loss": 0.4832,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -0.953238308429718,
      "rewards/margins": 1.7408623695373535,
      "rewards/rejected": -2.694100856781006,
      "step": 314
    },
    {
      "epoch": 0.46255506607929514,
      "epsilon_dpo/beta": 0.008652785792946815,
      "epsilon_dpo/beta_margin_grad_mean": -0.23646311461925507,
      "epsilon_dpo/beta_margin_grad_std": 0.16848552227020264,
      "epsilon_dpo/beta_margin_mean": 1.4849480390548706,
      "epsilon_dpo/beta_margin_std": 1.1557073593139648,
      "epsilon_dpo/loss_margin_mean": 171.94798278808594,
      "grad_norm": 62.91958999633789,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.008725542575120926,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.270101039870797e-07,
      "logits/chosen": -1.8116509914398193,
      "logits/rejected": -1.7590763568878174,
      "logps/chosen": -159.944091796875,
      "logps/ref_chosen": -33.06333923339844,
      "logps/ref_rejected": -85.8106918334961,
      "logps/rejected": -384.639404296875,
      "loss": 0.5973,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.0994541645050049,
      "rewards/margins": 1.4849481582641602,
      "rewards/rejected": -2.584402084350586,
      "step": 315
    },
    {
      "epoch": 0.46402349486049926,
      "epsilon_dpo/beta": 0.008572276681661606,
      "epsilon_dpo/beta_margin_grad_mean": -0.19783826172351837,
      "epsilon_dpo/beta_margin_grad_std": 0.14127233624458313,
      "epsilon_dpo/beta_margin_mean": 1.7659293413162231,
      "epsilon_dpo/beta_margin_std": 1.222463607788086,
      "epsilon_dpo/loss_margin_mean": 206.15966796875,
      "grad_norm": 53.759849548339844,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.008652537129819393,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 3.2578816852826086e-07,
      "logits/chosen": -1.918544054031372,
      "logits/rejected": -1.8977160453796387,
      "logps/chosen": -171.6690216064453,
      "logps/ref_chosen": -39.93564224243164,
      "logps/ref_rejected": -107.95684814453125,
      "logps/rejected": -445.84991455078125,
      "loss": 0.4775,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -1.1298701763153076,
      "rewards/margins": 1.7659293413162231,
      "rewards/rejected": -2.8957996368408203,
      "step": 316
    },
    {
      "epoch": 0.4654919236417034,
      "epsilon_dpo/beta": 0.00850069522857666,
      "epsilon_dpo/beta_margin_grad_mean": -0.1773056834936142,
      "epsilon_dpo/beta_margin_grad_std": 0.1531369686126709,
      "epsilon_dpo/beta_margin_mean": 1.9689700603485107,
      "epsilon_dpo/beta_margin_std": 1.239305853843689,
      "epsilon_dpo/loss_margin_mean": 232.01678466796875,
      "grad_norm": 45.539825439453125,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.008572173304855824,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.2456423598071783e-07,
      "logits/chosen": -1.982346773147583,
      "logits/rejected": -1.7732250690460205,
      "logps/chosen": -157.90228271484375,
      "logps/ref_chosen": -33.96290969848633,
      "logps/ref_rejected": -107.23854064941406,
      "logps/rejected": -463.1947021484375,
      "loss": 0.4319,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.0545722246170044,
      "rewards/margins": 1.9689700603485107,
      "rewards/rejected": -3.0235424041748047,
      "step": 317
    },
    {
      "epoch": 0.4669603524229075,
      "epsilon_dpo/beta": 0.00842957105487585,
      "epsilon_dpo/beta_margin_grad_mean": -0.21183888614177704,
      "epsilon_dpo/beta_margin_grad_std": 0.17660567164421082,
      "epsilon_dpo/beta_margin_mean": 1.6326032876968384,
      "epsilon_dpo/beta_margin_std": 1.1618354320526123,
      "epsilon_dpo/loss_margin_mean": 194.12109375,
      "grad_norm": 55.2903938293457,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.008500450290739536,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.233383385962115e-07,
      "logits/chosen": -2.0958797931671143,
      "logits/rejected": -1.8145737648010254,
      "logps/chosen": -180.26669311523438,
      "logps/ref_chosen": -40.75099563598633,
      "logps/ref_rejected": -89.30494689941406,
      "logps/rejected": -422.9417419433594,
      "loss": 0.5493,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.177966833114624,
      "rewards/margins": 1.6326031684875488,
      "rewards/rejected": -2.810570240020752,
      "step": 318
    },
    {
      "epoch": 0.4684287812041116,
      "epsilon_dpo/beta": 0.008361675776541233,
      "epsilon_dpo/beta_margin_grad_mean": -0.18624837696552277,
      "epsilon_dpo/beta_margin_grad_std": 0.18912473320960999,
      "epsilon_dpo/beta_margin_mean": 1.9518612623214722,
      "epsilon_dpo/beta_margin_std": 1.3541351556777954,
      "epsilon_dpo/loss_margin_mean": 234.00582885742188,
      "grad_norm": 45.54665756225586,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.008429327979683876,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 3.2211050867827805e-07,
      "logits/chosen": -2.003638982772827,
      "logits/rejected": -2.138455867767334,
      "logps/chosen": -135.60565185546875,
      "logps/ref_chosen": -31.367862701416016,
      "logps/ref_rejected": -121.58282470703125,
      "logps/rejected": -459.826416015625,
      "loss": 0.4951,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.8743832111358643,
      "rewards/margins": 1.9518612623214722,
      "rewards/rejected": -2.826244354248047,
      "step": 319
    },
    {
      "epoch": 0.4698972099853157,
      "epsilon_dpo/beta": 0.008289058692753315,
      "epsilon_dpo/beta_margin_grad_mean": -0.18520162999629974,
      "epsilon_dpo/beta_margin_grad_std": 0.15284550189971924,
      "epsilon_dpo/beta_margin_mean": 1.8797887563705444,
      "epsilon_dpo/beta_margin_std": 1.241491675376892,
      "epsilon_dpo/loss_margin_mean": 227.08258056640625,
      "grad_norm": 42.13555908203125,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.008361391723155975,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 3.208807785813777e-07,
      "logits/chosen": -2.008018970489502,
      "logits/rejected": -1.9689502716064453,
      "logps/chosen": -159.43008422851562,
      "logps/ref_chosen": -38.88483428955078,
      "logps/ref_rejected": -104.38986206054688,
      "logps/rejected": -452.0176696777344,
      "loss": 0.4546,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.9999314546585083,
      "rewards/margins": 1.8797886371612549,
      "rewards/rejected": -2.8797202110290527,
      "step": 320
    },
    {
      "epoch": 0.4713656387665198,
      "epsilon_dpo/beta": 0.008217157796025276,
      "epsilon_dpo/beta_margin_grad_mean": -0.19574427604675293,
      "epsilon_dpo/beta_margin_grad_std": 0.17788065969944,
      "epsilon_dpo/beta_margin_mean": 1.8304779529571533,
      "epsilon_dpo/beta_margin_std": 1.294824481010437,
      "epsilon_dpo/loss_margin_mean": 223.14154052734375,
      "grad_norm": 57.55246353149414,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.00828886404633522,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 3.1964918071004217e-07,
      "logits/chosen": -1.9321234226226807,
      "logits/rejected": -1.7463536262512207,
      "logps/chosen": -163.596923828125,
      "logps/ref_chosen": -33.60523986816406,
      "logps/ref_rejected": -97.33091735839844,
      "logps/rejected": -450.46417236328125,
      "loss": 0.5089,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.0699717998504639,
      "rewards/margins": 1.8304779529571533,
      "rewards/rejected": -2.900449752807617,
      "step": 321
    },
    {
      "epoch": 0.47283406754772395,
      "epsilon_dpo/beta": 0.008156154304742813,
      "epsilon_dpo/beta_margin_grad_mean": -0.1957569420337677,
      "epsilon_dpo/beta_margin_grad_std": 0.16668935120105743,
      "epsilon_dpo/beta_margin_mean": 1.7986748218536377,
      "epsilon_dpo/beta_margin_std": 1.202728509902954,
      "epsilon_dpo/loss_margin_mean": 221.13067626953125,
      "grad_norm": 51.22137451171875,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.008216965943574905,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 3.184157475180207e-07,
      "logits/chosen": -2.0422754287719727,
      "logits/rejected": -1.863848090171814,
      "logps/chosen": -169.1143798828125,
      "logps/ref_chosen": -39.760765075683594,
      "logps/ref_rejected": -101.8095703125,
      "logps/rejected": -452.2938537597656,
      "loss": 0.4902,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -1.057375431060791,
      "rewards/margins": 1.7986748218536377,
      "rewards/rejected": -2.8560502529144287,
      "step": 322
    },
    {
      "epoch": 0.47430249632892807,
      "epsilon_dpo/beta": 0.008091602474451065,
      "epsilon_dpo/beta_margin_grad_mean": -0.2292575240135193,
      "epsilon_dpo/beta_margin_grad_std": 0.17154912650585175,
      "epsilon_dpo/beta_margin_mean": 1.5250296592712402,
      "epsilon_dpo/beta_margin_std": 1.1593869924545288,
      "epsilon_dpo/loss_margin_mean": 188.9366912841797,
      "grad_norm": 59.20304489135742,
      "kl/avg_steps": 0.796875,
      "kl/beta": 0.008155797608196735,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 3.171805115074251e-07,
      "logits/chosen": -1.9689267873764038,
      "logits/rejected": -1.8758494853973389,
      "logps/chosen": -173.9646759033203,
      "logps/ref_chosen": -39.59267807006836,
      "logps/ref_rejected": -78.83342742919922,
      "logps/rejected": -402.1421203613281,
      "loss": 0.5838,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -1.0898997783660889,
      "rewards/margins": 1.5250296592712402,
      "rewards/rejected": -2.614929437637329,
      "step": 323
    },
    {
      "epoch": 0.47577092511013214,
      "epsilon_dpo/beta": 0.008031437173485756,
      "epsilon_dpo/beta_margin_grad_mean": -0.19054661691188812,
      "epsilon_dpo/beta_margin_grad_std": 0.18834634125232697,
      "epsilon_dpo/beta_margin_mean": 1.9402786493301392,
      "epsilon_dpo/beta_margin_std": 1.3676698207855225,
      "epsilon_dpo/loss_margin_mean": 242.30368041992188,
      "grad_norm": 56.6846923828125,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.008091319352388382,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 3.1594350522787295e-07,
      "logits/chosen": -2.086808443069458,
      "logits/rejected": -1.915328025817871,
      "logps/chosen": -158.4698486328125,
      "logps/ref_chosen": -39.379974365234375,
      "logps/ref_rejected": -94.40634155273438,
      "logps/rejected": -455.7998962402344,
      "loss": 0.4977,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.9597259163856506,
      "rewards/margins": 1.9402787685394287,
      "rewards/rejected": -2.9000043869018555,
      "step": 324
    },
    {
      "epoch": 0.47723935389133626,
      "epsilon_dpo/beta": 0.007969140075147152,
      "epsilon_dpo/beta_margin_grad_mean": -0.22008143365383148,
      "epsilon_dpo/beta_margin_grad_std": 0.1740773767232895,
      "epsilon_dpo/beta_margin_mean": 1.603606939315796,
      "epsilon_dpo/beta_margin_std": 1.2323416471481323,
      "epsilon_dpo/loss_margin_mean": 201.80323791503906,
      "grad_norm": 57.17675018310547,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.008031086064875126,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 3.147047612756302e-07,
      "logits/chosen": -2.2039828300476074,
      "logits/rejected": -1.9359973669052124,
      "logps/chosen": -174.04409790039062,
      "logps/ref_chosen": -44.78877258300781,
      "logps/ref_rejected": -76.1741943359375,
      "logps/rejected": -407.23272705078125,
      "loss": 0.5678,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.032346248626709,
      "rewards/margins": 1.603606939315796,
      "rewards/rejected": -2.635953426361084,
      "step": 325
    },
    {
      "epoch": 0.4787077826725404,
      "epsilon_dpo/beta": 0.007899893447756767,
      "epsilon_dpo/beta_margin_grad_mean": -0.19375857710838318,
      "epsilon_dpo/beta_margin_grad_std": 0.14700661599636078,
      "epsilon_dpo/beta_margin_mean": 1.8040101528167725,
      "epsilon_dpo/beta_margin_std": 1.1729263067245483,
      "epsilon_dpo/loss_margin_mean": 228.64596557617188,
      "grad_norm": 43.5681037902832,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.007968829944729805,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 3.134643122927519e-07,
      "logits/chosen": -2.2271931171417236,
      "logits/rejected": -1.8989125490188599,
      "logps/chosen": -165.11163330078125,
      "logps/ref_chosen": -39.91899871826172,
      "logps/ref_rejected": -87.51998901367188,
      "logps/rejected": -441.35858154296875,
      "loss": 0.4677,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -0.9895492792129517,
      "rewards/margins": 1.8040101528167725,
      "rewards/rejected": -2.7935595512390137,
      "step": 326
    },
    {
      "epoch": 0.4801762114537445,
      "epsilon_dpo/beta": 0.007831367664039135,
      "epsilon_dpo/beta_margin_grad_mean": -0.16008856892585754,
      "epsilon_dpo/beta_margin_grad_std": 0.15593233704566956,
      "epsilon_dpo/beta_margin_mean": 2.0613248348236084,
      "epsilon_dpo/beta_margin_std": 1.1618621349334717,
      "epsilon_dpo/loss_margin_mean": 263.6210632324219,
      "grad_norm": 45.13627624511719,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.007899707183241844,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 3.1222219096622264e-07,
      "logits/chosen": -2.139697551727295,
      "logits/rejected": -1.981135368347168,
      "logps/chosen": -143.63931274414062,
      "logps/ref_chosen": -38.00127410888672,
      "logps/ref_rejected": -119.96142578125,
      "logps/rejected": -489.22052001953125,
      "loss": 0.3983,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.8290703296661377,
      "rewards/margins": 2.0613250732421875,
      "rewards/rejected": -2.890395164489746,
      "step": 327
    },
    {
      "epoch": 0.48164464023494863,
      "epsilon_dpo/beta": 0.0077707814052701,
      "epsilon_dpo/beta_margin_grad_mean": -0.18991978466510773,
      "epsilon_dpo/beta_margin_grad_std": 0.1622304469347,
      "epsilon_dpo/beta_margin_mean": 1.8376761674880981,
      "epsilon_dpo/beta_margin_std": 1.1939152479171753,
      "epsilon_dpo/loss_margin_mean": 237.0498046875,
      "grad_norm": 43.36030197143555,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.00783118512481451,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 3.1097843002709427e-07,
      "logits/chosen": -2.1575894355773926,
      "logits/rejected": -1.941756010055542,
      "logps/chosen": -156.41744995117188,
      "logps/ref_chosen": -34.87682342529297,
      "logps/ref_rejected": -96.5279769897461,
      "logps/rejected": -455.118408203125,
      "loss": 0.4714,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.9466592669487,
      "rewards/margins": 1.8376761674880981,
      "rewards/rejected": -2.7843356132507324,
      "step": 328
    },
    {
      "epoch": 0.4831130690161527,
      "epsilon_dpo/beta": 0.007708113174885511,
      "epsilon_dpo/beta_margin_grad_mean": -0.18650057911872864,
      "epsilon_dpo/beta_margin_grad_std": 0.1677255630493164,
      "epsilon_dpo/beta_margin_mean": 1.8852558135986328,
      "epsilon_dpo/beta_margin_std": 1.2298924922943115,
      "epsilon_dpo/loss_margin_mean": 245.12493896484375,
      "grad_norm": 50.19667053222656,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.007770477794110775,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 3.0973306224962437e-07,
      "logits/chosen": -2.229661464691162,
      "logits/rejected": -2.011414051055908,
      "logps/chosen": -148.80914306640625,
      "logps/ref_chosen": -35.48017120361328,
      "logps/ref_rejected": -105.25934600830078,
      "logps/rejected": -463.7132568359375,
      "loss": 0.4686,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.8763324022293091,
      "rewards/margins": 1.8852558135986328,
      "rewards/rejected": -2.7615880966186523,
      "step": 329
    },
    {
      "epoch": 0.4845814977973568,
      "epsilon_dpo/beta": 0.007645989768207073,
      "epsilon_dpo/beta_margin_grad_mean": -0.1875944435596466,
      "epsilon_dpo/beta_margin_grad_std": 0.16437454521656036,
      "epsilon_dpo/beta_margin_mean": 1.8213557004928589,
      "epsilon_dpo/beta_margin_std": 1.1398029327392578,
      "epsilon_dpo/loss_margin_mean": 238.7486572265625,
      "grad_norm": 45.98007583618164,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0077078514732420444,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 3.084861204504122e-07,
      "logits/chosen": -2.0862879753112793,
      "logits/rejected": -2.0670981407165527,
      "logps/chosen": -126.71240234375,
      "logps/ref_chosen": -29.089645385742188,
      "logps/ref_rejected": -93.00184631347656,
      "logps/rejected": -429.37322998046875,
      "loss": 0.4709,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.7483941912651062,
      "rewards/margins": 1.8213555812835693,
      "rewards/rejected": -2.5697498321533203,
      "step": 330
    },
    {
      "epoch": 0.48604992657856094,
      "epsilon_dpo/beta": 0.0075819771736860275,
      "epsilon_dpo/beta_margin_grad_mean": -0.1762796938419342,
      "epsilon_dpo/beta_margin_grad_std": 0.15704338252544403,
      "epsilon_dpo/beta_margin_mean": 1.8491023778915405,
      "epsilon_dpo/beta_margin_std": 1.0536081790924072,
      "epsilon_dpo/loss_margin_mean": 244.35574340820312,
      "grad_norm": 42.173667907714844,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.0076457299292087555,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.072376374875335e-07,
      "logits/chosen": -2.128115177154541,
      "logits/rejected": -1.9702262878417969,
      "logps/chosen": -145.30062866210938,
      "logps/ref_chosen": -30.57064437866211,
      "logps/ref_rejected": -95.55513000488281,
      "logps/rejected": -454.640869140625,
      "loss": 0.4442,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.8718706369400024,
      "rewards/margins": 1.84910249710083,
      "rewards/rejected": -2.720973014831543,
      "step": 331
    },
    {
      "epoch": 0.48751835535976507,
      "epsilon_dpo/beta": 0.007520908955484629,
      "epsilon_dpo/beta_margin_grad_mean": -0.2063533514738083,
      "epsilon_dpo/beta_margin_grad_std": 0.1539819985628128,
      "epsilon_dpo/beta_margin_mean": 1.6616438627243042,
      "epsilon_dpo/beta_margin_std": 1.0939134359359741,
      "epsilon_dpo/loss_margin_mean": 221.3557891845703,
      "grad_norm": 46.4879035949707,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.00758175877854228,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 3.059876462596758e-07,
      "logits/chosen": -2.075554370880127,
      "logits/rejected": -1.9067268371582031,
      "logps/chosen": -150.6046142578125,
      "logps/ref_chosen": -31.9512882232666,
      "logps/ref_rejected": -81.75474548339844,
      "logps/rejected": -421.76385498046875,
      "loss": 0.5063,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.8930763006210327,
      "rewards/margins": 1.6616438627243042,
      "rewards/rejected": -2.554720163345337,
      "step": 332
    },
    {
      "epoch": 0.4889867841409692,
      "epsilon_dpo/beta": 0.007457944098860025,
      "epsilon_dpo/beta_margin_grad_mean": -0.1893877238035202,
      "epsilon_dpo/beta_margin_grad_std": 0.14721722900867462,
      "epsilon_dpo/beta_margin_mean": 1.7599835395812988,
      "epsilon_dpo/beta_margin_std": 1.0482655763626099,
      "epsilon_dpo/loss_margin_mean": 236.4034423828125,
      "grad_norm": 41.35095977783203,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.007520653773099184,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.0473617970527015e-07,
      "logits/chosen": -2.1496317386627197,
      "logits/rejected": -2.04693603515625,
      "logps/chosen": -148.9061737060547,
      "logps/ref_chosen": -35.37367630004883,
      "logps/ref_rejected": -102.79765319824219,
      "logps/rejected": -452.73358154296875,
      "loss": 0.4612,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.8483164310455322,
      "rewards/margins": 1.7599835395812988,
      "rewards/rejected": -2.608299732208252,
      "step": 333
    },
    {
      "epoch": 0.49045521292217326,
      "epsilon_dpo/beta": 0.007395543623715639,
      "epsilon_dpo/beta_margin_grad_mean": -0.20312325656414032,
      "epsilon_dpo/beta_margin_grad_std": 0.1642070859670639,
      "epsilon_dpo/beta_margin_mean": 1.6376922130584717,
      "epsilon_dpo/beta_margin_std": 1.04725182056427,
      "epsilon_dpo/loss_margin_mean": 221.89352416992188,
      "grad_norm": 43.43370819091797,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.007457728963345289,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 3.034832708016243e-07,
      "logits/chosen": -2.1904244422912598,
      "logits/rejected": -2.034341812133789,
      "logps/chosen": -160.87864685058594,
      "logps/ref_chosen": -44.046653747558594,
      "logps/ref_rejected": -101.81437683105469,
      "logps/rejected": -440.5398864746094,
      "loss": 0.5177,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.8658192157745361,
      "rewards/margins": 1.6376922130584717,
      "rewards/rejected": -2.503511428833008,
      "step": 334
    },
    {
      "epoch": 0.4919236417033774,
      "epsilon_dpo/beta": 0.007335977163165808,
      "epsilon_dpo/beta_margin_grad_mean": -0.24230733513832092,
      "epsilon_dpo/beta_margin_grad_std": 0.15432856976985931,
      "epsilon_dpo/beta_margin_mean": 1.3553001880645752,
      "epsilon_dpo/beta_margin_std": 0.970168948173523,
      "epsilon_dpo/loss_margin_mean": 185.18431091308594,
      "grad_norm": 51.230499267578125,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0073953308165073395,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 3.022289525640531e-07,
      "logits/chosen": -2.224094867706299,
      "logits/rejected": -2.0437393188476562,
      "logps/chosen": -182.1242218017578,
      "logps/ref_chosen": -45.84385681152344,
      "logps/ref_rejected": -93.38748931884766,
      "logps/rejected": -414.8521728515625,
      "loss": 0.6057,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.0017235279083252,
      "rewards/margins": 1.3553001880645752,
      "rewards/rejected": -2.3570237159729004,
      "step": 335
    },
    {
      "epoch": 0.4933920704845815,
      "epsilon_dpo/beta": 0.00726768234744668,
      "epsilon_dpo/beta_margin_grad_mean": -0.18038199841976166,
      "epsilon_dpo/beta_margin_grad_std": 0.13710618019104004,
      "epsilon_dpo/beta_margin_mean": 1.829659104347229,
      "epsilon_dpo/beta_margin_std": 1.08203125,
      "epsilon_dpo/loss_margin_mean": 251.95303344726562,
      "grad_norm": 50.833736419677734,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.007335728034377098,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 3.009732580450086e-07,
      "logits/chosen": -2.130596876144409,
      "logits/rejected": -2.1031112670898438,
      "logps/chosen": -138.39865112304688,
      "logps/ref_chosen": -27.110448837280273,
      "logps/ref_rejected": -113.38800048828125,
      "logps/rejected": -476.6292419433594,
      "loss": 0.4338,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -0.8097207546234131,
      "rewards/margins": 1.829659104347229,
      "rewards/rejected": -2.6393799781799316,
      "step": 336
    },
    {
      "epoch": 0.4948604992657856,
      "epsilon_dpo/beta": 0.007211537566035986,
      "epsilon_dpo/beta_margin_grad_mean": -0.2015434354543686,
      "epsilon_dpo/beta_margin_grad_std": 0.17217382788658142,
      "epsilon_dpo/beta_margin_mean": 1.7117812633514404,
      "epsilon_dpo/beta_margin_std": 1.1512197256088257,
      "epsilon_dpo/loss_margin_mean": 238.00546264648438,
      "grad_norm": 54.00065612792969,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.007267594337463379,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 2.9971622033320914e-07,
      "logits/chosen": -2.1570067405700684,
      "logits/rejected": -2.0568230152130127,
      "logps/chosen": -177.97787475585938,
      "logps/ref_chosen": -46.36784362792969,
      "logps/ref_rejected": -109.42700958251953,
      "logps/rejected": -479.04248046875,
      "loss": 0.5162,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.9518444538116455,
      "rewards/margins": 1.7117812633514404,
      "rewards/rejected": -2.663625717163086,
      "step": 337
    },
    {
      "epoch": 0.49632892804698975,
      "epsilon_dpo/beta": 0.007151126395910978,
      "epsilon_dpo/beta_margin_grad_mean": -0.21953918039798737,
      "epsilon_dpo/beta_margin_grad_std": 0.16535867750644684,
      "epsilon_dpo/beta_margin_mean": 1.531717300415039,
      "epsilon_dpo/beta_margin_std": 1.0485196113586426,
      "epsilon_dpo/loss_margin_mean": 214.6240692138672,
      "grad_norm": 54.01927185058594,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.007211256306618452,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 2.984578725527675e-07,
      "logits/chosen": -2.093170166015625,
      "logits/rejected": -2.0884156227111816,
      "logps/chosen": -169.5230712890625,
      "logps/ref_chosen": -40.62709045410156,
      "logps/ref_rejected": -94.02056121826172,
      "logps/rejected": -437.54058837890625,
      "loss": 0.5573,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.9232571125030518,
      "rewards/margins": 1.531717300415039,
      "rewards/rejected": -2.454974412918091,
      "step": 338
    },
    {
      "epoch": 0.4977973568281938,
      "epsilon_dpo/beta": 0.007089058868587017,
      "epsilon_dpo/beta_margin_grad_mean": -0.19239673018455505,
      "epsilon_dpo/beta_margin_grad_std": 0.1555846780538559,
      "epsilon_dpo/beta_margin_mean": 1.7873952388763428,
      "epsilon_dpo/beta_margin_std": 1.163671851158142,
      "epsilon_dpo/loss_margin_mean": 252.5406494140625,
      "grad_norm": 48.617767333984375,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.007150920573621988,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 2.9719824786231796e-07,
      "logits/chosen": -2.1362557411193848,
      "logits/rejected": -2.0298609733581543,
      "logps/chosen": -171.6763458251953,
      "logps/ref_chosen": -42.66804885864258,
      "logps/ref_rejected": -113.68071746826172,
      "logps/rejected": -495.22967529296875,
      "loss": 0.4775,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.9160549640655518,
      "rewards/margins": 1.7873952388763428,
      "rewards/rejected": -2.7034502029418945,
      "step": 339
    },
    {
      "epoch": 0.49926578560939794,
      "epsilon_dpo/beta": 0.007029782980680466,
      "epsilon_dpo/beta_margin_grad_mean": -0.22466416656970978,
      "epsilon_dpo/beta_margin_grad_std": 0.18301187455654144,
      "epsilon_dpo/beta_margin_mean": 1.512877106666565,
      "epsilon_dpo/beta_margin_std": 1.1476131677627563,
      "epsilon_dpo/loss_margin_mean": 215.7501983642578,
      "grad_norm": 53.56328582763672,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.007088892627507448,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 2.959373794541426e-07,
      "logits/chosen": -2.1749508380889893,
      "logits/rejected": -1.981499195098877,
      "logps/chosen": -152.33245849609375,
      "logps/ref_chosen": -33.344947814941406,
      "logps/ref_rejected": -78.67686462402344,
      "logps/rejected": -413.41455078125,
      "loss": 0.6076,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.8387173414230347,
      "rewards/margins": 1.5128772258758545,
      "rewards/rejected": -2.3515944480895996,
      "step": 340
    },
    {
      "epoch": 0.5007342143906021,
      "epsilon_dpo/beta": 0.006977556273341179,
      "epsilon_dpo/beta_margin_grad_mean": -0.1975635439157486,
      "epsilon_dpo/beta_margin_grad_std": 0.181951105594635,
      "epsilon_dpo/beta_margin_mean": 1.7857365608215332,
      "epsilon_dpo/beta_margin_std": 1.1933374404907227,
      "epsilon_dpo/loss_margin_mean": 256.70587158203125,
      "grad_norm": 51.781864166259766,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.007029580418020487,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 2.946753005532965e-07,
      "logits/chosen": -2.1569323539733887,
      "logits/rejected": -1.9963412284851074,
      "logps/chosen": -133.99024963378906,
      "logps/ref_chosen": -26.7175235748291,
      "logps/ref_rejected": -108.1113510131836,
      "logps/rejected": -472.0899353027344,
      "loss": 0.5095,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.7515710592269897,
      "rewards/margins": 1.7857365608215332,
      "rewards/rejected": -2.5373077392578125,
      "step": 341
    },
    {
      "epoch": 0.5022026431718062,
      "epsilon_dpo/beta": 0.006912530865520239,
      "epsilon_dpo/beta_margin_grad_mean": -0.18946228921413422,
      "epsilon_dpo/beta_margin_grad_std": 0.1443195641040802,
      "epsilon_dpo/beta_margin_mean": 1.8066987991333008,
      "epsilon_dpo/beta_margin_std": 1.138454794883728,
      "epsilon_dpo/loss_margin_mean": 261.54833984375,
      "grad_norm": 45.30876159667969,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.006977251265197992,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 2.934120444167326e-07,
      "logits/chosen": -2.289546489715576,
      "logits/rejected": -1.9754924774169922,
      "logps/chosen": -175.69277954101562,
      "logps/ref_chosen": -55.37556838989258,
      "logps/ref_rejected": -100.79505920410156,
      "logps/rejected": -482.6606140136719,
      "loss": 0.4569,
      "rewards/accuracies": 0.984375,
      "rewards/chosen": -0.8333628177642822,
      "rewards/margins": 1.8066987991333008,
      "rewards/rejected": -2.640061378479004,
      "step": 342
    },
    {
      "epoch": 0.5036710719530103,
      "epsilon_dpo/beta": 0.006846167147159576,
      "epsilon_dpo/beta_margin_grad_mean": -0.21721310913562775,
      "epsilon_dpo/beta_margin_grad_std": 0.11748269945383072,
      "epsilon_dpo/beta_margin_mean": 1.4517802000045776,
      "epsilon_dpo/beta_margin_std": 0.8032383322715759,
      "epsilon_dpo/loss_margin_mean": 212.13150024414062,
      "grad_norm": 49.670379638671875,
      "kl/avg_steps": 0.96875,
      "kl/beta": 0.0069124470464885235,
      "kl/n_epsilon_steps": 0.015625,
      "kl/p_epsilon_steps": 0.984375,
      "learning_rate": 2.9214764433242476e-07,
      "logits/chosen": -2.164459228515625,
      "logits/rejected": -2.1661453247070312,
      "logps/chosen": -178.66543579101562,
      "logps/ref_chosen": -42.08940124511719,
      "logps/ref_rejected": -117.90960693359375,
      "logps/rejected": -466.6171569824219,
      "loss": 0.514,
      "rewards/accuracies": 0.984375,
      "rewards/chosen": -0.9355216026306152,
      "rewards/margins": 1.451780080795288,
      "rewards/rejected": -2.3873019218444824,
      "step": 343
    },
    {
      "epoch": 0.5051395007342144,
      "epsilon_dpo/beta": 0.006797598209232092,
      "epsilon_dpo/beta_margin_grad_mean": -0.24506476521492004,
      "epsilon_dpo/beta_margin_grad_std": 0.1786363273859024,
      "epsilon_dpo/beta_margin_mean": 1.4522466659545898,
      "epsilon_dpo/beta_margin_std": 1.2337263822555542,
      "epsilon_dpo/loss_margin_mean": 214.36805725097656,
      "grad_norm": 57.320960998535156,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.006846125237643719,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 2.9088213361849126e-07,
      "logits/chosen": -2.179155111312866,
      "logits/rejected": -2.0898489952087402,
      "logps/chosen": -162.51026916503906,
      "logps/ref_chosen": -34.435752868652344,
      "logps/ref_rejected": -96.5870361328125,
      "logps/rejected": -439.02960205078125,
      "loss": 0.6315,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.8735264539718628,
      "rewards/margins": 1.4522466659545898,
      "rewards/rejected": -2.325773239135742,
      "step": 344
    },
    {
      "epoch": 0.5066079295154186,
      "epsilon_dpo/beta": 0.0067384676076471806,
      "epsilon_dpo/beta_margin_grad_mean": -0.18444088101387024,
      "epsilon_dpo/beta_margin_grad_std": 0.13682545721530914,
      "epsilon_dpo/beta_margin_mean": 1.7831730842590332,
      "epsilon_dpo/beta_margin_std": 1.0325002670288086,
      "epsilon_dpo/loss_margin_mean": 264.975341796875,
      "grad_norm": 50.00996398925781,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.00679726991802454,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 2.896155456223163e-07,
      "logits/chosen": -2.24761962890625,
      "logits/rejected": -2.086604595184326,
      "logps/chosen": -173.568603515625,
      "logps/ref_chosen": -40.498329162597656,
      "logps/ref_rejected": -107.90937805175781,
      "logps/rejected": -505.9549560546875,
      "loss": 0.4414,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -0.8974334001541138,
      "rewards/margins": 1.7831730842590332,
      "rewards/rejected": -2.6806066036224365,
      "step": 345
    },
    {
      "epoch": 0.5080763582966226,
      "epsilon_dpo/beta": 0.00667791161686182,
      "epsilon_dpo/beta_margin_grad_mean": -0.18874618411064148,
      "epsilon_dpo/beta_margin_grad_std": 0.13201214373111725,
      "epsilon_dpo/beta_margin_mean": 1.7088000774383545,
      "epsilon_dpo/beta_margin_std": 0.9547905325889587,
      "epsilon_dpo/loss_margin_mean": 256.1742858886719,
      "grad_norm": 42.087677001953125,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.006738309748470783,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 2.883479137196714e-07,
      "logits/chosen": -2.0703492164611816,
      "logits/rejected": -1.9229077100753784,
      "logps/chosen": -160.73776245117188,
      "logps/ref_chosen": -33.41410827636719,
      "logps/ref_rejected": -83.742919921875,
      "logps/rejected": -467.2408447265625,
      "loss": 0.4511,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.8508467078208923,
      "rewards/margins": 1.7088000774383545,
      "rewards/rejected": -2.5596466064453125,
      "step": 346
    },
    {
      "epoch": 0.5095447870778267,
      "epsilon_dpo/beta": 0.006634632125496864,
      "epsilon_dpo/beta_margin_grad_mean": -0.26513612270355225,
      "epsilon_dpo/beta_margin_grad_std": 0.2060505747795105,
      "epsilon_dpo/beta_margin_mean": 1.3007481098175049,
      "epsilon_dpo/beta_margin_std": 1.2315279245376587,
      "epsilon_dpo/loss_margin_mean": 197.03175354003906,
      "grad_norm": 72.8255615234375,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.006677791941910982,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 2.8707927131383614e-07,
      "logits/chosen": -2.1761817932128906,
      "logits/rejected": -2.0537772178649902,
      "logps/chosen": -213.68258666992188,
      "logps/ref_chosen": -46.14110565185547,
      "logps/ref_rejected": -97.28417205810547,
      "logps/rejected": -461.857421875,
      "loss": 0.7379,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -1.1159558296203613,
      "rewards/margins": 1.3007481098175049,
      "rewards/rejected": -2.416703701019287,
      "step": 347
    },
    {
      "epoch": 0.5110132158590308,
      "epsilon_dpo/beta": 0.0065768626518547535,
      "epsilon_dpo/beta_margin_grad_mean": -0.20659483969211578,
      "epsilon_dpo/beta_margin_grad_std": 0.15565478801727295,
      "epsilon_dpo/beta_margin_mean": 1.618215560913086,
      "epsilon_dpo/beta_margin_std": 1.0140705108642578,
      "epsilon_dpo/loss_margin_mean": 246.40516662597656,
      "grad_norm": 40.67171859741211,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.006634254939854145,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 2.858096518347179e-07,
      "logits/chosen": -2.202709674835205,
      "logits/rejected": -2.092670440673828,
      "logps/chosen": -162.21255493164062,
      "logps/ref_chosen": -36.474021911621094,
      "logps/ref_rejected": -93.86297607421875,
      "logps/rejected": -466.0066833496094,
      "loss": 0.5108,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.8284679651260376,
      "rewards/margins": 1.618215560913086,
      "rewards/rejected": -2.446683645248413,
      "step": 348
    },
    {
      "epoch": 0.5124816446402349,
      "epsilon_dpo/beta": 0.006519814487546682,
      "epsilon_dpo/beta_margin_grad_mean": -0.2202288955450058,
      "epsilon_dpo/beta_margin_grad_std": 0.171220600605011,
      "epsilon_dpo/beta_margin_mean": 1.6194798946380615,
      "epsilon_dpo/beta_margin_std": 1.1991218328475952,
      "epsilon_dpo/loss_margin_mean": 248.80596923828125,
      "grad_norm": 54.361026763916016,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.006576708517968655,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 2.845390887379706e-07,
      "logits/chosen": -2.1096177101135254,
      "logits/rejected": -2.068809986114502,
      "logps/chosen": -157.87994384765625,
      "logps/ref_chosen": -32.72327423095703,
      "logps/ref_rejected": -104.45381927490234,
      "logps/rejected": -478.4164733886719,
      "loss": 0.5577,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.8179513216018677,
      "rewards/margins": 1.6194798946380615,
      "rewards/rejected": -2.4374313354492188,
      "step": 349
    },
    {
      "epoch": 0.5139500734214391,
      "epsilon_dpo/beta": 0.006471411325037479,
      "epsilon_dpo/beta_margin_grad_mean": -0.2324196696281433,
      "epsilon_dpo/beta_margin_grad_std": 0.18096159398555756,
      "epsilon_dpo/beta_margin_mean": 1.507636308670044,
      "epsilon_dpo/beta_margin_std": 1.1591410636901855,
      "epsilon_dpo/loss_margin_mean": 233.73513793945312,
      "grad_norm": 66.0088119506836,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.006519661750644445,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 2.8326761550411346e-07,
      "logits/chosen": -2.133009910583496,
      "logits/rejected": -2.081470489501953,
      "logps/chosen": -188.4402618408203,
      "logps/ref_chosen": -41.924171447753906,
      "logps/ref_rejected": -96.81851196289062,
      "logps/rejected": -477.0697326660156,
      "loss": 0.6014,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.9508112668991089,
      "rewards/margins": 1.507636308670044,
      "rewards/rejected": -2.4584474563598633,
      "step": 350
    },
    {
      "epoch": 0.5154185022026432,
      "epsilon_dpo/beta": 0.006419192533940077,
      "epsilon_dpo/beta_margin_grad_mean": -0.22091074287891388,
      "epsilon_dpo/beta_margin_grad_std": 0.16051997244358063,
      "epsilon_dpo/beta_margin_mean": 1.5670552253723145,
      "epsilon_dpo/beta_margin_std": 1.1509732007980347,
      "epsilon_dpo/loss_margin_mean": 244.6831817626953,
      "grad_norm": 63.894622802734375,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.006471128202974796,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 2.819952656376487e-07,
      "logits/chosen": -2.141003131866455,
      "logits/rejected": -2.1702256202697754,
      "logps/chosen": -178.15701293945312,
      "logps/ref_chosen": -38.231632232666016,
      "logps/ref_rejected": -109.67044830322266,
      "logps/rejected": -494.2790222167969,
      "loss": 0.5544,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.9005885124206543,
      "rewards/margins": 1.5670552253723145,
      "rewards/rejected": -2.4676437377929688,
      "step": 351
    },
    {
      "epoch": 0.5168869309838473,
      "epsilon_dpo/beta": 0.006367456633597612,
      "epsilon_dpo/beta_margin_grad_mean": -0.22667014598846436,
      "epsilon_dpo/beta_margin_grad_std": 0.1572793573141098,
      "epsilon_dpo/beta_margin_mean": 1.4911013841629028,
      "epsilon_dpo/beta_margin_std": 1.0576071739196777,
      "epsilon_dpo/loss_margin_mean": 234.71234130859375,
      "grad_norm": 50.314727783203125,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0064189741387963295,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 2.8072207266617854e-07,
      "logits/chosen": -2.2193918228149414,
      "logits/rejected": -2.0443153381347656,
      "logps/chosen": -167.52468872070312,
      "logps/ref_chosen": -39.300880432128906,
      "logps/ref_rejected": -82.15660095214844,
      "logps/rejected": -445.0927734375,
      "loss": 0.5649,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.8179040551185608,
      "rewards/margins": 1.4911013841629028,
      "rewards/rejected": -2.3090052604675293,
      "step": 352
    },
    {
      "epoch": 0.5183553597650514,
      "epsilon_dpo/beta": 0.006326087284833193,
      "epsilon_dpo/beta_margin_grad_mean": -0.24461467564105988,
      "epsilon_dpo/beta_margin_grad_std": 0.182557612657547,
      "epsilon_dpo/beta_margin_mean": 1.4262304306030273,
      "epsilon_dpo/beta_margin_std": 1.1893874406814575,
      "epsilon_dpo/loss_margin_mean": 226.36761474609375,
      "grad_norm": 57.24125289916992,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.006367240101099014,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 2.794480701395219e-07,
      "logits/chosen": -2.104984760284424,
      "logits/rejected": -2.00747013092041,
      "logps/chosen": -186.15036010742188,
      "logps/ref_chosen": -37.74017333984375,
      "logps/ref_rejected": -86.70198059082031,
      "logps/rejected": -461.47979736328125,
      "loss": 0.6444,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.9420816898345947,
      "rewards/margins": 1.4262304306030273,
      "rewards/rejected": -2.368312120437622,
      "step": 353
    },
    {
      "epoch": 0.5198237885462555,
      "epsilon_dpo/beta": 0.006276935804635286,
      "epsilon_dpo/beta_margin_grad_mean": -0.21808402240276337,
      "epsilon_dpo/beta_margin_grad_std": 0.14849531650543213,
      "epsilon_dpo/beta_margin_mean": 1.4829356670379639,
      "epsilon_dpo/beta_margin_std": 0.9004656672477722,
      "epsilon_dpo/loss_margin_mean": 236.8414306640625,
      "grad_norm": 44.26091384887695,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.006325727794319391,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 2.781732916288303e-07,
      "logits/chosen": -2.21701717376709,
      "logits/rejected": -2.05487322807312,
      "logps/chosen": -189.85113525390625,
      "logps/ref_chosen": -46.64361572265625,
      "logps/ref_rejected": -94.82032775878906,
      "logps/rejected": -474.8692626953125,
      "loss": 0.5369,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.8998274803161621,
      "rewards/margins": 1.4829356670379639,
      "rewards/rejected": -2.382763147354126,
      "step": 354
    },
    {
      "epoch": 0.5212922173274597,
      "epsilon_dpo/beta": 0.006220430601388216,
      "epsilon_dpo/beta_margin_grad_mean": -0.19742940366268158,
      "epsilon_dpo/beta_margin_grad_std": 0.1338631510734558,
      "epsilon_dpo/beta_margin_mean": 1.626767873764038,
      "epsilon_dpo/beta_margin_std": 0.9073279500007629,
      "epsilon_dpo/loss_margin_mean": 261.8030700683594,
      "grad_norm": 43.440181732177734,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.0062766908667981625,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 2.7689777072570284e-07,
      "logits/chosen": -2.233151435852051,
      "logits/rejected": -2.0062508583068848,
      "logps/chosen": -158.43057250976562,
      "logps/ref_chosen": -38.99018859863281,
      "logps/ref_rejected": -88.16902160644531,
      "logps/rejected": -469.4125061035156,
      "loss": 0.473,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.7434677481651306,
      "rewards/margins": 1.626767873764038,
      "rewards/rejected": -2.3702354431152344,
      "step": 355
    },
    {
      "epoch": 0.5227606461086637,
      "epsilon_dpo/beta": 0.006174284033477306,
      "epsilon_dpo/beta_margin_grad_mean": -0.2459879219532013,
      "epsilon_dpo/beta_margin_grad_std": 0.1747446060180664,
      "epsilon_dpo/beta_margin_mean": 1.3851886987686157,
      "epsilon_dpo/beta_margin_std": 1.0974996089935303,
      "epsilon_dpo/loss_margin_mean": 225.0797882080078,
      "grad_norm": 52.8966178894043,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0062203193083405495,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 2.7562154104130176e-07,
      "logits/chosen": -2.149020195007324,
      "logits/rejected": -1.9297568798065186,
      "logps/chosen": -177.79510498046875,
      "logps/ref_chosen": -36.000328063964844,
      "logps/ref_rejected": -81.04385375976562,
      "logps/rejected": -447.9184265136719,
      "loss": 0.6355,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.8773345351219177,
      "rewards/margins": 1.3851888179779053,
      "rewards/rejected": -2.2625231742858887,
      "step": 356
    },
    {
      "epoch": 0.5242290748898678,
      "epsilon_dpo/beta": 0.006120603997260332,
      "epsilon_dpo/beta_margin_grad_mean": -0.22637690603733063,
      "epsilon_dpo/beta_margin_grad_std": 0.17489291727542877,
      "epsilon_dpo/beta_margin_mean": 1.5144197940826416,
      "epsilon_dpo/beta_margin_std": 1.1098681688308716,
      "epsilon_dpo/loss_margin_mean": 247.9174041748047,
      "grad_norm": 53.406307220458984,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.006174014415591955,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 2.7434463620546594e-07,
      "logits/chosen": -2.1703100204467773,
      "logits/rejected": -2.005626916885376,
      "logps/chosen": -169.40367126464844,
      "logps/ref_chosen": -37.324180603027344,
      "logps/ref_rejected": -90.94546508789062,
      "logps/rejected": -470.9423522949219,
      "loss": 0.5852,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.8109502792358398,
      "rewards/margins": 1.5144197940826416,
      "rewards/rejected": -2.3253703117370605,
      "step": 357
    },
    {
      "epoch": 0.5256975036710719,
      "epsilon_dpo/beta": 0.0060713388957083225,
      "epsilon_dpo/beta_margin_grad_mean": -0.2422979176044464,
      "epsilon_dpo/beta_margin_grad_std": 0.17289938032627106,
      "epsilon_dpo/beta_margin_mean": 1.4026376008987427,
      "epsilon_dpo/beta_margin_std": 1.0886473655700684,
      "epsilon_dpo/loss_margin_mean": 231.60694885253906,
      "grad_norm": 54.73785400390625,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.006120460107922554,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 2.730670898658255e-07,
      "logits/chosen": -2.2487030029296875,
      "logits/rejected": -2.109694480895996,
      "logps/chosen": -169.3449249267578,
      "logps/ref_chosen": -45.07238006591797,
      "logps/ref_rejected": -94.94227600097656,
      "logps/rejected": -450.82177734375,
      "loss": 0.6267,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.7564330101013184,
      "rewards/margins": 1.4026376008987427,
      "rewards/rejected": -2.1590704917907715,
      "step": 358
    },
    {
      "epoch": 0.527165932452276,
      "epsilon_dpo/beta": 0.0060280985198915005,
      "epsilon_dpo/beta_margin_grad_mean": -0.23849013447761536,
      "epsilon_dpo/beta_margin_grad_std": 0.1872168630361557,
      "epsilon_dpo/beta_margin_mean": 1.4863260984420776,
      "epsilon_dpo/beta_margin_std": 1.1895928382873535,
      "epsilon_dpo/loss_margin_mean": 247.45826721191406,
      "grad_norm": 62.71131134033203,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.006071132142096758,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 2.717889356869146e-07,
      "logits/chosen": -2.1315059661865234,
      "logits/rejected": -1.9415364265441895,
      "logps/chosen": -162.13319396972656,
      "logps/ref_chosen": -34.454952239990234,
      "logps/ref_rejected": -87.54586791992188,
      "logps/rejected": -462.682373046875,
      "loss": 0.6242,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.7723500728607178,
      "rewards/margins": 1.486325979232788,
      "rewards/rejected": -2.258676052093506,
      "step": 359
    },
    {
      "epoch": 0.5286343612334802,
      "epsilon_dpo/beta": 0.005986965261399746,
      "epsilon_dpo/beta_margin_grad_mean": -0.2441015988588333,
      "epsilon_dpo/beta_margin_grad_std": 0.1673990935087204,
      "epsilon_dpo/beta_margin_mean": 1.3607901334762573,
      "epsilon_dpo/beta_margin_std": 0.9962857365608215,
      "epsilon_dpo/loss_margin_mean": 228.12159729003906,
      "grad_norm": 53.97544860839844,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.006027807481586933,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 2.7051020734928443e-07,
      "logits/chosen": -2.0680994987487793,
      "logits/rejected": -1.957366704940796,
      "logps/chosen": -168.40216064453125,
      "logps/ref_chosen": -37.17042922973633,
      "logps/ref_rejected": -76.79757690429688,
      "logps/rejected": -436.15087890625,
      "loss": 0.6196,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.7889617681503296,
      "rewards/margins": 1.3607900142669678,
      "rewards/rejected": -2.149751901626587,
      "step": 360
    },
    {
      "epoch": 0.5301027900146843,
      "epsilon_dpo/beta": 0.005942343734204769,
      "epsilon_dpo/beta_margin_grad_mean": -0.25044846534729004,
      "epsilon_dpo/beta_margin_grad_std": 0.18136946856975555,
      "epsilon_dpo/beta_margin_mean": 1.3711673021316528,
      "epsilon_dpo/beta_margin_std": 1.1313101053237915,
      "epsilon_dpo/loss_margin_mean": 231.51885986328125,
      "grad_norm": 63.17055892944336,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.005986649077385664,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 2.6923093854861593e-07,
      "logits/chosen": -2.1590943336486816,
      "logits/rejected": -2.121795177459717,
      "logps/chosen": -164.0542755126953,
      "logps/ref_chosen": -33.520301818847656,
      "logps/ref_rejected": -95.78604125976562,
      "logps/rejected": -457.8388671875,
      "loss": 0.6539,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.7777143120765686,
      "rewards/margins": 1.3711671829223633,
      "rewards/rejected": -2.148881435394287,
      "step": 361
    },
    {
      "epoch": 0.5315712187958884,
      "epsilon_dpo/beta": 0.005890679080039263,
      "epsilon_dpo/beta_margin_grad_mean": -0.23398736119270325,
      "epsilon_dpo/beta_margin_grad_std": 0.1614319384098053,
      "epsilon_dpo/beta_margin_mean": 1.4454318284988403,
      "epsilon_dpo/beta_margin_std": 1.0421142578125,
      "epsilon_dpo/loss_margin_mean": 245.76043701171875,
      "grad_norm": 57.27875900268555,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.005942083429545164,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 2.679511629948319e-07,
      "logits/chosen": -2.2858972549438477,
      "logits/rejected": -2.1964077949523926,
      "logps/chosen": -185.8004913330078,
      "logps/ref_chosen": -44.4540901184082,
      "logps/ref_rejected": -110.75244140625,
      "logps/rejected": -497.8592834472656,
      "loss": 0.5872,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.8333665728569031,
      "rewards/margins": 1.4454319477081299,
      "rewards/rejected": -2.2787985801696777,
      "step": 362
    },
    {
      "epoch": 0.5330396475770925,
      "epsilon_dpo/beta": 0.005839582998305559,
      "epsilon_dpo/beta_margin_grad_mean": -0.18884162604808807,
      "epsilon_dpo/beta_margin_grad_std": 0.14109890162944794,
      "epsilon_dpo/beta_margin_mean": 1.7546730041503906,
      "epsilon_dpo/beta_margin_std": 1.0252107381820679,
      "epsilon_dpo/loss_margin_mean": 300.8768615722656,
      "grad_norm": 44.063777923583984,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.00589054124429822,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 2.6667091441120816e-07,
      "logits/chosen": -2.20405650138855,
      "logits/rejected": -2.084122657775879,
      "logps/chosen": -119.686767578125,
      "logps/ref_chosen": -23.62018585205078,
      "logps/ref_rejected": -79.18755340576172,
      "logps/rejected": -476.1310119628906,
      "loss": 0.4544,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -0.5620150566101074,
      "rewards/margins": 1.7546730041503906,
      "rewards/rejected": -2.316688060760498,
      "step": 363
    },
    {
      "epoch": 0.5345080763582967,
      "epsilon_dpo/beta": 0.005788929760456085,
      "epsilon_dpo/beta_margin_grad_mean": -0.23159046471118927,
      "epsilon_dpo/beta_margin_grad_std": 0.15345177054405212,
      "epsilon_dpo/beta_margin_mean": 1.4228205680847168,
      "epsilon_dpo/beta_margin_std": 0.979676365852356,
      "epsilon_dpo/loss_margin_mean": 246.19557189941406,
      "grad_norm": 54.10791015625,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.005839446093887091,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 2.6539022653348575e-07,
      "logits/chosen": -2.1630282402038574,
      "logits/rejected": -2.1442086696624756,
      "logps/chosen": -151.59072875976562,
      "logps/ref_chosen": -31.6948184967041,
      "logps/ref_rejected": -98.25422668457031,
      "logps/rejected": -464.345703125,
      "loss": 0.5787,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.6957234144210815,
      "rewards/margins": 1.4228205680847168,
      "rewards/rejected": -2.118544101715088,
      "step": 364
    },
    {
      "epoch": 0.5359765051395007,
      "epsilon_dpo/beta": 0.005736907012760639,
      "epsilon_dpo/beta_margin_grad_mean": -0.22249744832515717,
      "epsilon_dpo/beta_margin_grad_std": 0.13856177031993866,
      "epsilon_dpo/beta_margin_mean": 1.4774224758148193,
      "epsilon_dpo/beta_margin_std": 0.9550539255142212,
      "epsilon_dpo/loss_margin_mean": 257.8067932128906,
      "grad_norm": 45.98625564575195,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.005788794253021479,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 2.641091331089811e-07,
      "logits/chosen": -2.2106542587280273,
      "logits/rejected": -2.1826648712158203,
      "logps/chosen": -125.05105590820312,
      "logps/ref_chosen": -28.370468139648438,
      "logps/ref_rejected": -97.9366226196289,
      "logps/rejected": -452.42401123046875,
      "loss": 0.5397,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.5555899739265442,
      "rewards/margins": 1.4774224758148193,
      "rewards/rejected": -2.0330123901367188,
      "step": 365
    },
    {
      "epoch": 0.5374449339207048,
      "epsilon_dpo/beta": 0.005688969045877457,
      "epsilon_dpo/beta_margin_grad_mean": -0.241429403424263,
      "epsilon_dpo/beta_margin_grad_std": 0.15970946848392487,
      "epsilon_dpo/beta_margin_mean": 1.3868852853775024,
      "epsilon_dpo/beta_margin_std": 1.0170059204101562,
      "epsilon_dpo/loss_margin_mean": 244.25833129882812,
      "grad_norm": 45.932926177978516,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.0057368045672774315,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 2.6282766789569736e-07,
      "logits/chosen": -2.1703920364379883,
      "logits/rejected": -2.1496224403381348,
      "logps/chosen": -141.24298095703125,
      "logps/ref_chosen": -29.695941925048828,
      "logps/ref_rejected": -91.68377685546875,
      "logps/rejected": -447.4891357421875,
      "loss": 0.605,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.6359978914260864,
      "rewards/margins": 1.3868852853775024,
      "rewards/rejected": -2.022883176803589,
      "step": 366
    },
    {
      "epoch": 0.5389133627019089,
      "epsilon_dpo/beta": 0.005634258035570383,
      "epsilon_dpo/beta_margin_grad_mean": -0.23779228329658508,
      "epsilon_dpo/beta_margin_grad_std": 0.14979609847068787,
      "epsilon_dpo/beta_margin_mean": 1.3488836288452148,
      "epsilon_dpo/beta_margin_std": 0.8744153380393982,
      "epsilon_dpo/loss_margin_mean": 239.53407287597656,
      "grad_norm": 42.18586730957031,
      "kl/avg_steps": 0.96875,
      "kl/beta": 0.005688805133104324,
      "kl/n_epsilon_steps": 0.015625,
      "kl/p_epsilon_steps": 0.984375,
      "learning_rate": 2.615458646614349e-07,
      "logits/chosen": -2.2993650436401367,
      "logits/rejected": -2.1403145790100098,
      "logps/chosen": -153.87405395507812,
      "logps/ref_chosen": -43.07976531982422,
      "logps/ref_rejected": -82.26808166503906,
      "logps/rejected": -432.596435546875,
      "loss": 0.5899,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -0.624735951423645,
      "rewards/margins": 1.3488836288452148,
      "rewards/rejected": -1.9736195802688599,
      "step": 367
    },
    {
      "epoch": 0.540381791483113,
      "epsilon_dpo/beta": 0.005581961013376713,
      "epsilon_dpo/beta_margin_grad_mean": -0.2129683494567871,
      "epsilon_dpo/beta_margin_grad_std": 0.1246427670121193,
      "epsilon_dpo/beta_margin_mean": 1.4776982069015503,
      "epsilon_dpo/beta_margin_std": 0.806984007358551,
      "epsilon_dpo/loss_margin_mean": 264.9310607910156,
      "grad_norm": 42.2542610168457,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.005634223576635122,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 2.6026375718290083e-07,
      "logits/chosen": -2.189484119415283,
      "logits/rejected": -2.1851706504821777,
      "logps/chosen": -142.2340545654297,
      "logps/ref_chosen": -28.834022521972656,
      "logps/ref_rejected": -104.38568115234375,
      "logps/rejected": -482.7167663574219,
      "loss": 0.5089,
      "rewards/accuracies": 0.984375,
      "rewards/chosen": -0.633552074432373,
      "rewards/margins": 1.4776982069015503,
      "rewards/rejected": -2.111250400543213,
      "step": 368
    },
    {
      "epoch": 0.5418502202643172,
      "epsilon_dpo/beta": 0.0055318609811365604,
      "epsilon_dpo/beta_margin_grad_mean": -0.223898246884346,
      "epsilon_dpo/beta_margin_grad_std": 0.13859553635120392,
      "epsilon_dpo/beta_margin_mean": 1.431330919265747,
      "epsilon_dpo/beta_margin_std": 0.8688164353370667,
      "epsilon_dpo/loss_margin_mean": 259.0464782714844,
      "grad_norm": 39.924076080322266,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.005581893492490053,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 2.589813792448196e-07,
      "logits/chosen": -2.307586193084717,
      "logits/rejected": -2.1981377601623535,
      "logps/chosen": -161.64993286132812,
      "logps/ref_chosen": -46.21360778808594,
      "logps/ref_rejected": -98.12519836425781,
      "logps/rejected": -472.6080017089844,
      "loss": 0.5458,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -0.640053391456604,
      "rewards/margins": 1.431330919265747,
      "rewards/rejected": -2.0713844299316406,
      "step": 369
    },
    {
      "epoch": 0.5433186490455213,
      "epsilon_dpo/beta": 0.005487365182489157,
      "epsilon_dpo/beta_margin_grad_mean": -0.26117002964019775,
      "epsilon_dpo/beta_margin_grad_std": 0.16114374995231628,
      "epsilon_dpo/beta_margin_mean": 1.2338489294052124,
      "epsilon_dpo/beta_margin_std": 0.9565902352333069,
      "epsilon_dpo/loss_margin_mean": 225.38717651367188,
      "grad_norm": 44.1972770690918,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.005531761795282364,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 2.5769876463904263e-07,
      "logits/chosen": -2.3571860790252686,
      "logits/rejected": -2.2904090881347656,
      "logps/chosen": -190.09152221679688,
      "logps/ref_chosen": -51.13551330566406,
      "logps/ref_rejected": -101.64096069335938,
      "logps/rejected": -465.984130859375,
      "loss": 0.6656,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.7640899419784546,
      "rewards/margins": 1.2338488101959229,
      "rewards/rejected": -1.997938871383667,
      "step": 370
    },
    {
      "epoch": 0.5447870778267254,
      "epsilon_dpo/beta": 0.005443139001727104,
      "epsilon_dpo/beta_margin_grad_mean": -0.2511758804321289,
      "epsilon_dpo/beta_margin_grad_std": 0.17598623037338257,
      "epsilon_dpo/beta_margin_mean": 1.3466187715530396,
      "epsilon_dpo/beta_margin_std": 1.0779163837432861,
      "epsilon_dpo/loss_margin_mean": 248.08079528808594,
      "grad_norm": 59.9031982421875,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.005487178452312946,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 2.5641594716365744e-07,
      "logits/chosen": -2.2778027057647705,
      "logits/rejected": -2.235936164855957,
      "logps/chosen": -181.4659881591797,
      "logps/ref_chosen": -44.41009521484375,
      "logps/ref_rejected": -106.13936614990234,
      "logps/rejected": -491.27606201171875,
      "loss": 0.65,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.7477129697799683,
      "rewards/margins": 1.34661865234375,
      "rewards/rejected": -2.094331741333008,
      "step": 371
    },
    {
      "epoch": 0.5462555066079295,
      "epsilon_dpo/beta": 0.005397569388151169,
      "epsilon_dpo/beta_margin_grad_mean": -0.2212544083595276,
      "epsilon_dpo/beta_margin_grad_std": 0.16570396721363068,
      "epsilon_dpo/beta_margin_mean": 1.5037719011306763,
      "epsilon_dpo/beta_margin_std": 1.0139695405960083,
      "epsilon_dpo/loss_margin_mean": 279.17572021484375,
      "grad_norm": 42.31486511230469,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.00544295459985733,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 2.551329606220976e-07,
      "logits/chosen": -2.188450336456299,
      "logits/rejected": -1.9817776679992676,
      "logps/chosen": -168.07699584960938,
      "logps/ref_chosen": -37.96990203857422,
      "logps/ref_rejected": -85.50196838378906,
      "logps/rejected": -494.7847595214844,
      "loss": 0.5632,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.7031924724578857,
      "rewards/margins": 1.5037720203399658,
      "rewards/rejected": -2.2069644927978516,
      "step": 372
    },
    {
      "epoch": 0.5477239353891337,
      "epsilon_dpo/beta": 0.005349034443497658,
      "epsilon_dpo/beta_margin_grad_mean": -0.2134443074464798,
      "epsilon_dpo/beta_margin_grad_std": 0.14159946143627167,
      "epsilon_dpo/beta_margin_mean": 1.5223017930984497,
      "epsilon_dpo/beta_margin_std": 0.9210037589073181,
      "epsilon_dpo/loss_margin_mean": 284.92840576171875,
      "grad_norm": 46.6475944519043,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.005397413857281208,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 2.538498388222517e-07,
      "logits/chosen": -2.178375005722046,
      "logits/rejected": -2.014909267425537,
      "logps/chosen": -159.17428588867188,
      "logps/ref_chosen": -31.804325103759766,
      "logps/ref_rejected": -93.370361328125,
      "logps/rejected": -505.668701171875,
      "loss": 0.5204,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.6824929118156433,
      "rewards/margins": 1.5223019123077393,
      "rewards/rejected": -2.2047948837280273,
      "step": 373
    },
    {
      "epoch": 0.5491923641703378,
      "epsilon_dpo/beta": 0.005304337944835424,
      "epsilon_dpo/beta_margin_grad_mean": -0.24420872330665588,
      "epsilon_dpo/beta_margin_grad_std": 0.14603427052497864,
      "epsilon_dpo/beta_margin_mean": 1.3295738697052002,
      "epsilon_dpo/beta_margin_std": 0.9483336806297302,
      "epsilon_dpo/loss_margin_mean": 251.12574768066406,
      "grad_norm": 65.01420593261719,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.00534893898293376,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 2.525666155755725e-07,
      "logits/chosen": -2.1958394050598145,
      "logits/rejected": -2.096358060836792,
      "logps/chosen": -186.239501953125,
      "logps/ref_chosen": -47.937625885009766,
      "logps/ref_rejected": -99.9871826171875,
      "logps/rejected": -489.414794921875,
      "loss": 0.6052,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.7352691888809204,
      "rewards/margins": 1.3295738697052002,
      "rewards/rejected": -2.06484317779541,
      "step": 374
    },
    {
      "epoch": 0.5506607929515418,
      "epsilon_dpo/beta": 0.005254983436316252,
      "epsilon_dpo/beta_margin_grad_mean": -0.20512135326862335,
      "epsilon_dpo/beta_margin_grad_std": 0.1477726697921753,
      "epsilon_dpo/beta_margin_mean": 1.6202079057693481,
      "epsilon_dpo/beta_margin_std": 1.0169481039047241,
      "epsilon_dpo/loss_margin_mean": 308.59869384765625,
      "grad_norm": 44.36325454711914,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.0053041847422719,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 2.512833246961859e-07,
      "logits/chosen": -2.20353364944458,
      "logits/rejected": -2.1390933990478516,
      "logps/chosen": -151.54449462890625,
      "logps/ref_chosen": -37.31241989135742,
      "logps/ref_rejected": -96.08183288574219,
      "logps/rejected": -518.91259765625,
      "loss": 0.5067,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -0.6010797023773193,
      "rewards/margins": 1.6202080249786377,
      "rewards/rejected": -2.221287727355957,
      "step": 375
    },
    {
      "epoch": 0.5521292217327459,
      "epsilon_dpo/beta": 0.005209460388869047,
      "epsilon_dpo/beta_margin_grad_mean": -0.20545507967472076,
      "epsilon_dpo/beta_margin_grad_std": 0.1631988137960434,
      "epsilon_dpo/beta_margin_mean": 1.6594352722167969,
      "epsilon_dpo/beta_margin_std": 1.0614945888519287,
      "epsilon_dpo/loss_margin_mean": 319.01361083984375,
      "grad_norm": 37.207130432128906,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.005254920106381178,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 2.5e-07,
      "logits/chosen": -2.2137274742126465,
      "logits/rejected": -2.0475354194641113,
      "logps/chosen": -162.23599243164062,
      "logps/ref_chosen": -46.223838806152344,
      "logps/ref_rejected": -112.95925903320312,
      "logps/rejected": -547.9850463867188,
      "loss": 0.5114,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.605557918548584,
      "rewards/margins": 1.6594352722167969,
      "rewards/rejected": -2.264993190765381,
      "step": 376
    },
    {
      "epoch": 0.55359765051395,
      "epsilon_dpo/beta": 0.005162645131349564,
      "epsilon_dpo/beta_margin_grad_mean": -0.24356918036937714,
      "epsilon_dpo/beta_margin_grad_std": 0.13881158828735352,
      "epsilon_dpo/beta_margin_mean": 1.3458126783370972,
      "epsilon_dpo/beta_margin_std": 0.9675434827804565,
      "epsilon_dpo/loss_margin_mean": 260.9751281738281,
      "grad_norm": 44.168060302734375,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.0052093383856117725,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 2.487166753038141e-07,
      "logits/chosen": -2.0519776344299316,
      "logits/rejected": -2.01825213432312,
      "logps/chosen": -185.23043823242188,
      "logps/ref_chosen": -34.36379623413086,
      "logps/ref_rejected": -105.25419616699219,
      "logps/rejected": -517.095947265625,
      "loss": 0.5956,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.7789356708526611,
      "rewards/margins": 1.3458126783370972,
      "rewards/rejected": -2.1247482299804688,
      "step": 377
    },
    {
      "epoch": 0.5550660792951542,
      "epsilon_dpo/beta": 0.005117891822010279,
      "epsilon_dpo/beta_margin_grad_mean": -0.22632895410060883,
      "epsilon_dpo/beta_margin_grad_std": 0.1629311591386795,
      "epsilon_dpo/beta_margin_mean": 1.4298152923583984,
      "epsilon_dpo/beta_margin_std": 0.9799423217773438,
      "epsilon_dpo/loss_margin_mean": 279.919189453125,
      "grad_norm": 52.31544876098633,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.005162552464753389,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 2.4743338442442754e-07,
      "logits/chosen": -2.059394359588623,
      "logits/rejected": -1.9235854148864746,
      "logps/chosen": -163.18124389648438,
      "logps/ref_chosen": -31.191333770751953,
      "logps/ref_rejected": -94.8112564086914,
      "logps/rejected": -506.7203674316406,
      "loss": 0.5897,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.6769013404846191,
      "rewards/margins": 1.4298152923583984,
      "rewards/rejected": -2.1067166328430176,
      "step": 378
    },
    {
      "epoch": 0.5565345080763583,
      "epsilon_dpo/beta": 0.005076698027551174,
      "epsilon_dpo/beta_margin_grad_mean": -0.23987461626529694,
      "epsilon_dpo/beta_margin_grad_std": 0.16330820322036743,
      "epsilon_dpo/beta_margin_mean": 1.4291728734970093,
      "epsilon_dpo/beta_margin_std": 1.0851447582244873,
      "epsilon_dpo/loss_margin_mean": 282.147705078125,
      "grad_norm": 51.661922454833984,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.005117772147059441,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 2.461501611777483e-07,
      "logits/chosen": -2.0814695358276367,
      "logits/rejected": -2.1467087268829346,
      "logps/chosen": -197.74441528320312,
      "logps/ref_chosen": -35.68910217285156,
      "logps/ref_rejected": -120.69367980957031,
      "logps/rejected": -564.896728515625,
      "loss": 0.6029,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.8247177004814148,
      "rewards/margins": 1.4291728734970093,
      "rewards/rejected": -2.2538905143737793,
      "step": 379
    },
    {
      "epoch": 0.5580029368575624,
      "epsilon_dpo/beta": 0.005032609216868877,
      "epsilon_dpo/beta_margin_grad_mean": -0.20714667439460754,
      "epsilon_dpo/beta_margin_grad_std": 0.1527094542980194,
      "epsilon_dpo/beta_margin_mean": 1.6478782892227173,
      "epsilon_dpo/beta_margin_std": 1.1102581024169922,
      "epsilon_dpo/loss_margin_mean": 327.93218994140625,
      "grad_norm": 39.429542541503906,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.005076525267213583,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 2.4486703937790243e-07,
      "logits/chosen": -2.105628728866577,
      "logits/rejected": -2.135561466217041,
      "logps/chosen": -184.11827087402344,
      "logps/ref_chosen": -30.20886993408203,
      "logps/ref_rejected": -111.17587280273438,
      "logps/rejected": -593.0174560546875,
      "loss": 0.5118,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.7753435373306274,
      "rewards/margins": 1.6478781700134277,
      "rewards/rejected": -2.4232218265533447,
      "step": 380
    },
    {
      "epoch": 0.5594713656387665,
      "epsilon_dpo/beta": 0.004996819421648979,
      "epsilon_dpo/beta_margin_grad_mean": -0.2540000379085541,
      "epsilon_dpo/beta_margin_grad_std": 0.1617269366979599,
      "epsilon_dpo/beta_margin_mean": 1.3064846992492676,
      "epsilon_dpo/beta_margin_std": 1.0072511434555054,
      "epsilon_dpo/loss_margin_mean": 262.2966613769531,
      "grad_norm": 51.32357406616211,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.005032490938901901,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 2.435840528363426e-07,
      "logits/chosen": -2.112166404724121,
      "logits/rejected": -1.9345085620880127,
      "logps/chosen": -195.97369384765625,
      "logps/ref_chosen": -35.8305778503418,
      "logps/ref_rejected": -84.19041442871094,
      "logps/rejected": -506.63018798828125,
      "loss": 0.64,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.8029674291610718,
      "rewards/margins": 1.3064846992492676,
      "rewards/rejected": -2.109452247619629,
      "step": 381
    },
    {
      "epoch": 0.5609397944199707,
      "epsilon_dpo/beta": 0.004958038218319416,
      "epsilon_dpo/beta_margin_grad_mean": -0.23661740124225616,
      "epsilon_dpo/beta_margin_grad_std": 0.16259969770908356,
      "epsilon_dpo/beta_margin_mean": 1.4187428951263428,
      "epsilon_dpo/beta_margin_std": 1.0073864459991455,
      "epsilon_dpo/loss_margin_mean": 286.858642578125,
      "grad_norm": 47.54762268066406,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.004996578209102154,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 2.4230123536095745e-07,
      "logits/chosen": -2.1727161407470703,
      "logits/rejected": -2.1786251068115234,
      "logps/chosen": -211.81280517578125,
      "logps/ref_chosen": -42.912322998046875,
      "logps/ref_rejected": -116.84707641601562,
      "logps/rejected": -572.606201171875,
      "loss": 0.5931,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.839522123336792,
      "rewards/margins": 1.4187428951263428,
      "rewards/rejected": -2.2582650184631348,
      "step": 382
    },
    {
      "epoch": 0.5624082232011748,
      "epsilon_dpo/beta": 0.00491960346698761,
      "epsilon_dpo/beta_margin_grad_mean": -0.2571463882923126,
      "epsilon_dpo/beta_margin_grad_std": 0.1763610541820526,
      "epsilon_dpo/beta_margin_mean": 1.305212140083313,
      "epsilon_dpo/beta_margin_std": 1.0758684873580933,
      "epsilon_dpo/loss_margin_mean": 266.0962219238281,
      "grad_norm": 56.97015380859375,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.004957844968885183,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 2.4101862075518037e-07,
      "logits/chosen": -2.0988528728485107,
      "logits/rejected": -2.077052354812622,
      "logps/chosen": -209.27877807617188,
      "logps/ref_chosen": -35.82965087890625,
      "logps/ref_rejected": -100.10868072509766,
      "logps/rejected": -539.654052734375,
      "loss": 0.6678,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.8556368947029114,
      "rewards/margins": 1.3052120208740234,
      "rewards/rejected": -2.160849094390869,
      "step": 383
    },
    {
      "epoch": 0.5638766519823789,
      "epsilon_dpo/beta": 0.004878391977399588,
      "epsilon_dpo/beta_margin_grad_mean": -0.261491596698761,
      "epsilon_dpo/beta_margin_grad_std": 0.14361578226089478,
      "epsilon_dpo/beta_margin_mean": 1.1763639450073242,
      "epsilon_dpo/beta_margin_std": 0.7963204979896545,
      "epsilon_dpo/loss_margin_mean": 241.60084533691406,
      "grad_norm": 49.46163558959961,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.004919412080198526,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 2.397362428170992e-07,
      "logits/chosen": -2.1388659477233887,
      "logits/rejected": -2.1503121852874756,
      "logps/chosen": -204.53836059570312,
      "logps/ref_chosen": -36.32765579223633,
      "logps/ref_rejected": -88.09075927734375,
      "logps/rejected": -497.9023132324219,
      "loss": 0.6509,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.8223646879196167,
      "rewards/margins": 1.1763639450073242,
      "rewards/rejected": -1.998728632926941,
      "step": 384
    },
    {
      "epoch": 0.5653450807635829,
      "epsilon_dpo/beta": 0.004834526218473911,
      "epsilon_dpo/beta_margin_grad_mean": -0.24184301495552063,
      "epsilon_dpo/beta_margin_grad_std": 0.1403016448020935,
      "epsilon_dpo/beta_margin_mean": 1.3218892812728882,
      "epsilon_dpo/beta_margin_std": 0.876196563243866,
      "epsilon_dpo/loss_margin_mean": 273.7486267089844,
      "grad_norm": 42.43720245361328,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.004878251813352108,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 2.3845413533856514e-07,
      "logits/chosen": -2.1851940155029297,
      "logits/rejected": -1.9630916118621826,
      "logps/chosen": -205.151123046875,
      "logps/ref_chosen": -43.28904724121094,
      "logps/ref_rejected": -85.31612396240234,
      "logps/rejected": -520.9268188476562,
      "loss": 0.595,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.7839208841323853,
      "rewards/margins": 1.3218892812728882,
      "rewards/rejected": -2.1058101654052734,
      "step": 385
    },
    {
      "epoch": 0.566813509544787,
      "epsilon_dpo/beta": 0.004791106563061476,
      "epsilon_dpo/beta_margin_grad_mean": -0.22500666975975037,
      "epsilon_dpo/beta_margin_grad_std": 0.1310817301273346,
      "epsilon_dpo/beta_margin_mean": 1.448143482208252,
      "epsilon_dpo/beta_margin_std": 0.916283369064331,
      "epsilon_dpo/loss_margin_mean": 302.5584411621094,
      "grad_norm": 44.11907196044922,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.004834439605474472,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 2.3717233210430254e-07,
      "logits/chosen": -2.1383421421051025,
      "logits/rejected": -2.0959932804107666,
      "logps/chosen": -186.66197204589844,
      "logps/ref_chosen": -34.78173828125,
      "logps/ref_rejected": -98.69821166992188,
      "logps/rejected": -553.1368408203125,
      "loss": 0.5412,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -0.7281551957130432,
      "rewards/margins": 1.448143482208252,
      "rewards/rejected": -2.1762986183166504,
      "step": 386
    },
    {
      "epoch": 0.5682819383259912,
      "epsilon_dpo/beta": 0.004748077597469091,
      "epsilon_dpo/beta_margin_grad_mean": -0.2660858631134033,
      "epsilon_dpo/beta_margin_grad_std": 0.1432296484708786,
      "epsilon_dpo/beta_margin_mean": 1.1700471639633179,
      "epsilon_dpo/beta_margin_std": 0.8362559080123901,
      "epsilon_dpo/loss_margin_mean": 246.71267700195312,
      "grad_norm": 52.421470642089844,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.0047910213470458984,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 2.3589086689101889e-07,
      "logits/chosen": -2.315922737121582,
      "logits/rejected": -2.0701866149902344,
      "logps/chosen": -236.2892608642578,
      "logps/ref_chosen": -47.62559509277344,
      "logps/ref_rejected": -97.41331481933594,
      "logps/rejected": -532.7896728515625,
      "loss": 0.6613,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.8966184854507446,
      "rewards/margins": 1.1700471639633179,
      "rewards/rejected": -2.0666656494140625,
      "step": 387
    },
    {
      "epoch": 0.5697503671071953,
      "epsilon_dpo/beta": 0.004706918261945248,
      "epsilon_dpo/beta_margin_grad_mean": -0.23977012932300568,
      "epsilon_dpo/beta_margin_grad_std": 0.14799103140830994,
      "epsilon_dpo/beta_margin_mean": 1.386051893234253,
      "epsilon_dpo/beta_margin_std": 0.9813371300697327,
      "epsilon_dpo/loss_margin_mean": 294.90777587890625,
      "grad_norm": 50.50879669189453,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.004747992381453514,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 2.3460977346651428e-07,
      "logits/chosen": -2.154555320739746,
      "logits/rejected": -2.115201473236084,
      "logps/chosen": -198.59793090820312,
      "logps/ref_chosen": -35.743316650390625,
      "logps/ref_rejected": -110.60121154785156,
      "logps/rejected": -568.3635864257812,
      "loss": 0.5907,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.767310619354248,
      "rewards/margins": 1.386051893234253,
      "rewards/rejected": -2.153362512588501,
      "step": 388
    },
    {
      "epoch": 0.5712187958883994,
      "epsilon_dpo/beta": 0.004673445131629705,
      "epsilon_dpo/beta_margin_grad_mean": -0.26743534207344055,
      "epsilon_dpo/beta_margin_grad_std": 0.18217839300632477,
      "epsilon_dpo/beta_margin_mean": 1.2179065942764282,
      "epsilon_dpo/beta_margin_std": 1.0329232215881348,
      "epsilon_dpo/loss_margin_mean": 261.6222839355469,
      "grad_norm": 57.76890563964844,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.004706807900220156,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 2.3332908558879177e-07,
      "logits/chosen": -2.1483242511749268,
      "logits/rejected": -2.025583267211914,
      "logps/chosen": -195.3277587890625,
      "logps/ref_chosen": -40.81889724731445,
      "logps/ref_rejected": -82.08807373046875,
      "logps/rejected": -498.21917724609375,
      "loss": 0.7075,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.7245763540267944,
      "rewards/margins": 1.2179067134857178,
      "rewards/rejected": -1.9424829483032227,
      "step": 389
    },
    {
      "epoch": 0.5726872246696035,
      "epsilon_dpo/beta": 0.004632791969925165,
      "epsilon_dpo/beta_margin_grad_mean": -0.24001896381378174,
      "epsilon_dpo/beta_margin_grad_std": 0.13718777894973755,
      "epsilon_dpo/beta_margin_mean": 1.3450864553451538,
      "epsilon_dpo/beta_margin_std": 0.9246547222137451,
      "epsilon_dpo/loss_margin_mean": 290.7643737792969,
      "grad_norm": 47.92219161987305,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.004673219285905361,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 2.320488370051681e-07,
      "logits/chosen": -2.114452838897705,
      "logits/rejected": -1.998061180114746,
      "logps/chosen": -161.92718505859375,
      "logps/ref_chosen": -29.450889587402344,
      "logps/ref_rejected": -92.78984069824219,
      "logps/rejected": -516.030517578125,
      "loss": 0.5862,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.6144026517868042,
      "rewards/margins": 1.3450864553451538,
      "rewards/rejected": -1.959489107131958,
      "step": 390
    },
    {
      "epoch": 0.5741556534508077,
      "epsilon_dpo/beta": 0.004595502279698849,
      "epsilon_dpo/beta_margin_grad_mean": -0.273894727230072,
      "epsilon_dpo/beta_margin_grad_std": 0.1488787978887558,
      "epsilon_dpo/beta_margin_mean": 1.1127126216888428,
      "epsilon_dpo/beta_margin_std": 0.8292704820632935,
      "epsilon_dpo/loss_margin_mean": 242.7264404296875,
      "grad_norm": 53.62302017211914,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.004632683005183935,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 2.3076906145138405e-07,
      "logits/chosen": -2.1756703853607178,
      "logits/rejected": -2.104757785797119,
      "logps/chosen": -198.61831665039062,
      "logps/ref_chosen": -41.40957260131836,
      "logps/ref_rejected": -93.46768951416016,
      "logps/rejected": -493.40289306640625,
      "loss": 0.6924,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.7247334718704224,
      "rewards/margins": 1.1127126216888428,
      "rewards/rejected": -1.8374460935592651,
      "step": 391
    },
    {
      "epoch": 0.5756240822320118,
      "epsilon_dpo/beta": 0.004555592313408852,
      "epsilon_dpo/beta_margin_grad_mean": -0.2217557728290558,
      "epsilon_dpo/beta_margin_grad_std": 0.1372254192829132,
      "epsilon_dpo/beta_margin_mean": 1.4560474157333374,
      "epsilon_dpo/beta_margin_std": 0.9119820594787598,
      "epsilon_dpo/loss_margin_mean": 320.1082763671875,
      "grad_norm": 46.09185791015625,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.004595345817506313,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 2.294897926507156e-07,
      "logits/chosen": -2.129703998565674,
      "logits/rejected": -2.0612103939056396,
      "logps/chosen": -166.53561401367188,
      "logps/ref_chosen": -29.435636520385742,
      "logps/ref_rejected": -109.47046661376953,
      "logps/rejected": -566.6787109375,
      "loss": 0.5405,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.6257168054580688,
      "rewards/margins": 1.456047534942627,
      "rewards/rejected": -2.0817642211914062,
      "step": 392
    },
    {
      "epoch": 0.5770925110132159,
      "epsilon_dpo/beta": 0.004520347807556391,
      "epsilon_dpo/beta_margin_grad_mean": -0.2750376760959625,
      "epsilon_dpo/beta_margin_grad_std": 0.16051527857780457,
      "epsilon_dpo/beta_margin_mean": 1.1350780725479126,
      "epsilon_dpo/beta_margin_std": 0.9084685444831848,
      "epsilon_dpo/loss_margin_mean": 251.88442993164062,
      "grad_norm": 51.396175384521484,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.004555485676974058,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 2.2821106431308543e-07,
      "logits/chosen": -2.204120635986328,
      "logits/rejected": -2.0319063663482666,
      "logps/chosen": -160.30789184570312,
      "logps/ref_chosen": -27.59358024597168,
      "logps/ref_rejected": -77.63695526123047,
      "logps/rejected": -462.23565673828125,
      "loss": 0.7035,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.6019172668457031,
      "rewards/margins": 1.1350780725479126,
      "rewards/rejected": -1.7369953393936157,
      "step": 393
    },
    {
      "epoch": 0.57856093979442,
      "epsilon_dpo/beta": 0.004478242713958025,
      "epsilon_dpo/beta_margin_grad_mean": -0.22447995841503143,
      "epsilon_dpo/beta_margin_grad_std": 0.1293361932039261,
      "epsilon_dpo/beta_margin_mean": 1.4324853420257568,
      "epsilon_dpo/beta_margin_std": 0.8796045184135437,
      "epsilon_dpo/loss_margin_mean": 320.1289367675781,
      "grad_norm": 36.82295227050781,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.0045201717875897884,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 2.2693291013417452e-07,
      "logits/chosen": -2.1519179344177246,
      "logits/rejected": -2.0851173400878906,
      "logps/chosen": -171.11074829101562,
      "logps/ref_chosen": -33.609615325927734,
      "logps/ref_rejected": -97.24693298339844,
      "logps/rejected": -554.876953125,
      "loss": 0.5407,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -0.61604905128479,
      "rewards/margins": 1.4324853420257568,
      "rewards/rejected": -2.048534393310547,
      "step": 394
    },
    {
      "epoch": 0.580029368575624,
      "epsilon_dpo/beta": 0.004446446429938078,
      "epsilon_dpo/beta_margin_grad_mean": -0.2658129632472992,
      "epsilon_dpo/beta_margin_grad_std": 0.17667248845100403,
      "epsilon_dpo/beta_margin_mean": 1.248099446296692,
      "epsilon_dpo/beta_margin_std": 1.0518742799758911,
      "epsilon_dpo/loss_margin_mean": 281.7148132324219,
      "grad_norm": 55.038551330566406,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.004478188697248697,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 2.2565536379453404e-07,
      "logits/chosen": -2.1458773612976074,
      "logits/rejected": -2.072359085083008,
      "logps/chosen": -198.61692810058594,
      "logps/ref_chosen": -42.13022994995117,
      "logps/ref_rejected": -89.85330963134766,
      "logps/rejected": -528.0548095703125,
      "loss": 0.6914,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.6987878084182739,
      "rewards/margins": 1.248099446296692,
      "rewards/rejected": -1.9468872547149658,
      "step": 395
    },
    {
      "epoch": 0.5814977973568282,
      "epsilon_dpo/beta": 0.004410546738654375,
      "epsilon_dpo/beta_margin_grad_mean": -0.25534769892692566,
      "epsilon_dpo/beta_margin_grad_std": 0.17317545413970947,
      "epsilon_dpo/beta_margin_mean": 1.317757248878479,
      "epsilon_dpo/beta_margin_std": 1.0815587043762207,
      "epsilon_dpo/loss_margin_mean": 299.5541687011719,
      "grad_norm": 54.677188873291016,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.004446231294423342,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 2.2437845895869825e-07,
      "logits/chosen": -2.223626136779785,
      "logits/rejected": -2.068613290786743,
      "logps/chosen": -198.652099609375,
      "logps/ref_chosen": -39.016197204589844,
      "logps/ref_rejected": -94.28376007080078,
      "logps/rejected": -553.473876953125,
      "loss": 0.6594,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.706026017665863,
      "rewards/margins": 1.3177571296691895,
      "rewards/rejected": -2.0237832069396973,
      "step": 396
    },
    {
      "epoch": 0.5829662261380323,
      "epsilon_dpo/beta": 0.004374999552965164,
      "epsilon_dpo/beta_margin_grad_mean": -0.2254161238670349,
      "epsilon_dpo/beta_margin_grad_std": 0.1620989739894867,
      "epsilon_dpo/beta_margin_mean": 1.489986777305603,
      "epsilon_dpo/beta_margin_std": 1.04046630859375,
      "epsilon_dpo/loss_margin_mean": 341.36114501953125,
      "grad_norm": 48.635677337646484,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.004410396795719862,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 2.2310222927429716e-07,
      "logits/chosen": -2.2994096279144287,
      "logits/rejected": -2.131638526916504,
      "logps/chosen": -192.5784149169922,
      "logps/ref_chosen": -42.487823486328125,
      "logps/ref_rejected": -111.77519226074219,
      "logps/rejected": -603.2269287109375,
      "loss": 0.5673,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.6591649055480957,
      "rewards/margins": 1.4899868965148926,
      "rewards/rejected": -2.1491518020629883,
      "step": 397
    },
    {
      "epoch": 0.5844346549192364,
      "epsilon_dpo/beta": 0.004342474043369293,
      "epsilon_dpo/beta_margin_grad_mean": -0.2421146184206009,
      "epsilon_dpo/beta_margin_grad_std": 0.1806570291519165,
      "epsilon_dpo/beta_margin_mean": 1.4041334390640259,
      "epsilon_dpo/beta_margin_std": 1.1149554252624512,
      "epsilon_dpo/loss_margin_mean": 324.445556640625,
      "grad_norm": 49.05772399902344,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.004374851007014513,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 2.2182670837116972e-07,
      "logits/chosen": -2.292893409729004,
      "logits/rejected": -2.2275965213775635,
      "logps/chosen": -206.82467651367188,
      "logps/ref_chosen": -47.66093444824219,
      "logps/ref_rejected": -116.59112548828125,
      "logps/rejected": -600.200439453125,
      "loss": 0.6398,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.6942225694656372,
      "rewards/margins": 1.4041335582733154,
      "rewards/rejected": -2.098356246948242,
      "step": 398
    },
    {
      "epoch": 0.5859030837004405,
      "epsilon_dpo/beta": 0.004307433497160673,
      "epsilon_dpo/beta_margin_grad_mean": -0.24833783507347107,
      "epsilon_dpo/beta_margin_grad_std": 0.16362829506397247,
      "epsilon_dpo/beta_margin_mean": 1.3256672620773315,
      "epsilon_dpo/beta_margin_std": 0.9880812168121338,
      "epsilon_dpo/loss_margin_mean": 308.531494140625,
      "grad_norm": 54.13144302368164,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.004342284053564072,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 2.2055192986047804e-07,
      "logits/chosen": -2.2859010696411133,
      "logits/rejected": -2.0517654418945312,
      "logps/chosen": -172.4034423828125,
      "logps/ref_chosen": -35.37146759033203,
      "logps/ref_rejected": -83.08477020263672,
      "logps/rejected": -528.648193359375,
      "loss": 0.6303,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.5922940969467163,
      "rewards/margins": 1.325667142868042,
      "rewards/rejected": -1.9179613590240479,
      "step": 399
    },
    {
      "epoch": 0.5873715124816447,
      "epsilon_dpo/beta": 0.0042700255289673805,
      "epsilon_dpo/beta_margin_grad_mean": -0.23046433925628662,
      "epsilon_dpo/beta_margin_grad_std": 0.13983549177646637,
      "epsilon_dpo/beta_margin_mean": 1.362850308418274,
      "epsilon_dpo/beta_margin_std": 0.8101769089698792,
      "epsilon_dpo/loss_margin_mean": 319.66986083984375,
      "grad_norm": 42.21733093261719,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.0043072872795164585,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 2.192779273338215e-07,
      "logits/chosen": -2.346558094024658,
      "logits/rejected": -2.2209692001342773,
      "logps/chosen": -214.174560546875,
      "logps/ref_chosen": -52.37220001220703,
      "logps/ref_rejected": -110.96676635742188,
      "logps/rejected": -592.43896484375,
      "loss": 0.5665,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.6924352645874023,
      "rewards/margins": 1.3628504276275635,
      "rewards/rejected": -2.0552854537963867,
      "step": 400
    },
    {
      "epoch": 0.5873715124816447,
      "eval_epsilon_dpo/beta": 0.004249695222824812,
      "eval_epsilon_dpo/beta_margin_grad_mean": -0.35010644793510437,
      "eval_epsilon_dpo/beta_margin_grad_std": 0.18771161139011383,
      "eval_epsilon_dpo/beta_margin_mean": 0.7586736679077148,
      "eval_epsilon_dpo/beta_margin_std": 0.9838672876358032,
      "eval_epsilon_dpo/loss_margin_mean": 180.02191162109375,
      "eval_kl/n_epsilon_steps": 0.25813356041908264,
      "eval_kl/p_epsilon_steps": 0.741866409778595,
      "eval_logits/chosen": -2.246497631072998,
      "eval_logits/rejected": -2.1281936168670654,
      "eval_logps/chosen": -320.8791198730469,
      "eval_logps/ref_chosen": -68.29110717773438,
      "eval_logps/ref_rejected": -92.08038330078125,
      "eval_logps/rejected": -524.6903076171875,
      "eval_loss": 0.4863179922103882,
      "eval_rewards/accuracies": 0.7564212083816528,
      "eval_rewards/chosen": -1.077276587486267,
      "eval_rewards/margins": 0.7586737275123596,
      "eval_rewards/rejected": -1.835950255393982,
      "eval_runtime": 38.3555,
      "eval_samples_per_second": 60.982,
      "eval_steps_per_second": 1.929,
      "step": 400
    },
    {
      "epoch": 0.5888399412628488,
      "epsilon_dpo/beta": 0.0042383247055113316,
      "epsilon_dpo/beta_margin_grad_mean": -0.2859787940979004,
      "epsilon_dpo/beta_margin_grad_std": 0.17702849209308624,
      "epsilon_dpo/beta_margin_mean": 1.153497338294983,
      "epsilon_dpo/beta_margin_std": 1.093888521194458,
      "epsilon_dpo/loss_margin_mean": 273.0599365234375,
      "grad_norm": 55.85029983520508,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.004269925411790609,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 2.1800473436235136e-07,
      "logits/chosen": -2.2151618003845215,
      "logits/rejected": -2.1000635623931885,
      "logps/chosen": -206.0033721923828,
      "logps/ref_chosen": -39.69450378417969,
      "logps/ref_rejected": -90.86283874511719,
      "logps/rejected": -530.2316284179688,
      "loss": 0.7457,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.7069000005722046,
      "rewards/margins": 1.153497338294983,
      "rewards/rejected": -1.8603973388671875,
      "step": 401
    },
    {
      "epoch": 0.5903083700440529,
      "epsilon_dpo/beta": 0.004204125143587589,
      "epsilon_dpo/beta_margin_grad_mean": -0.251901239156723,
      "epsilon_dpo/beta_margin_grad_std": 0.14963571727275848,
      "epsilon_dpo/beta_margin_mean": 1.2545827627182007,
      "epsilon_dpo/beta_margin_std": 0.8603823184967041,
      "epsilon_dpo/loss_margin_mean": 299.1294860839844,
      "grad_norm": 44.19596862792969,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.004238139372318983,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 2.1673238449588665e-07,
      "logits/chosen": -2.216102123260498,
      "logits/rejected": -2.0876975059509277,
      "logps/chosen": -187.8063201904297,
      "logps/ref_chosen": -38.76295852661133,
      "logps/ref_rejected": -86.50106811523438,
      "logps/rejected": -534.6738891601562,
      "loss": 0.6286,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.6281874179840088,
      "rewards/margins": 1.2545827627182007,
      "rewards/rejected": -1.8827701807022095,
      "step": 402
    },
    {
      "epoch": 0.591776798825257,
      "epsilon_dpo/beta": 0.004170241765677929,
      "epsilon_dpo/beta_margin_grad_mean": -0.2570376694202423,
      "epsilon_dpo/beta_margin_grad_std": 0.15749378502368927,
      "epsilon_dpo/beta_margin_mean": 1.245935082435608,
      "epsilon_dpo/beta_margin_std": 0.9069526791572571,
      "epsilon_dpo/loss_margin_mean": 299.4497985839844,
      "grad_norm": 45.329769134521484,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.004203982185572386,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 2.154609112620295e-07,
      "logits/chosen": -2.254790782928467,
      "logits/rejected": -2.070467472076416,
      "logps/chosen": -162.46539306640625,
      "logps/ref_chosen": -29.60453224182129,
      "logps/ref_rejected": -82.97395324707031,
      "logps/rejected": -515.2846069335938,
      "loss": 0.6468,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.5560154914855957,
      "rewards/margins": 1.2459352016448975,
      "rewards/rejected": -1.8019505739212036,
      "step": 403
    },
    {
      "epoch": 0.593245227606461,
      "epsilon_dpo/beta": 0.004143148194998503,
      "epsilon_dpo/beta_margin_grad_mean": -0.2758185565471649,
      "epsilon_dpo/beta_margin_grad_std": 0.16760212182998657,
      "epsilon_dpo/beta_margin_mean": 1.1664576530456543,
      "epsilon_dpo/beta_margin_std": 0.9855588674545288,
      "epsilon_dpo/loss_margin_mean": 282.7168884277344,
      "grad_norm": 52.8735466003418,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.004170100204646587,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 2.1419034816528218e-07,
      "logits/chosen": -2.253624439239502,
      "logits/rejected": -2.092195987701416,
      "logps/chosen": -182.46783447265625,
      "logps/ref_chosen": -32.369415283203125,
      "logps/ref_rejected": -84.27439880371094,
      "logps/rejected": -517.0897216796875,
      "loss": 0.7084,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.6246769428253174,
      "rewards/margins": 1.1664576530456543,
      "rewards/rejected": -1.7911345958709717,
      "step": 404
    },
    {
      "epoch": 0.5947136563876652,
      "epsilon_dpo/beta": 0.004105777945369482,
      "epsilon_dpo/beta_margin_grad_mean": -0.27076855301856995,
      "epsilon_dpo/beta_margin_grad_std": 0.15409667789936066,
      "epsilon_dpo/beta_margin_mean": 1.1753402948379517,
      "epsilon_dpo/beta_margin_std": 0.9294463396072388,
      "epsilon_dpo/loss_margin_mean": 286.6537170410156,
      "grad_norm": 45.45560836791992,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.004142912104725838,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 2.129207286861638e-07,
      "logits/chosen": -2.2662670612335205,
      "logits/rejected": -2.1083085536956787,
      "logps/chosen": -207.39706420898438,
      "logps/ref_chosen": -45.16187286376953,
      "logps/ref_rejected": -93.87014770507812,
      "logps/rejected": -542.759033203125,
      "loss": 0.6822,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.6669937372207642,
      "rewards/margins": 1.1753404140472412,
      "rewards/rejected": -1.8423340320587158,
      "step": 405
    },
    {
      "epoch": 0.5961820851688693,
      "epsilon_dpo/beta": 0.004071469884365797,
      "epsilon_dpo/beta_margin_grad_mean": -0.24695971608161926,
      "epsilon_dpo/beta_margin_grad_std": 0.14529289305210114,
      "epsilon_dpo/beta_margin_mean": 1.2771347761154175,
      "epsilon_dpo/beta_margin_std": 0.8445379137992859,
      "epsilon_dpo/loss_margin_mean": 314.2767639160156,
      "grad_norm": 44.63689422607422,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.004105704370886087,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 2.1165208628032861e-07,
      "logits/chosen": -2.2001147270202637,
      "logits/rejected": -2.1016151905059814,
      "logps/chosen": -175.9896240234375,
      "logps/ref_chosen": -32.108238220214844,
      "logps/ref_rejected": -99.3056640625,
      "logps/rejected": -557.4638671875,
      "loss": 0.612,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.5877770185470581,
      "rewards/margins": 1.2771347761154175,
      "rewards/rejected": -1.8649117946624756,
      "step": 406
    },
    {
      "epoch": 0.5976505139500734,
      "epsilon_dpo/beta": 0.0040450384840369225,
      "epsilon_dpo/beta_margin_grad_mean": -0.2901814579963684,
      "epsilon_dpo/beta_margin_grad_std": 0.17480535805225372,
      "epsilon_dpo/beta_margin_mean": 1.060813307762146,
      "epsilon_dpo/beta_margin_std": 0.9306771159172058,
      "epsilon_dpo/loss_margin_mean": 263.4065246582031,
      "grad_norm": 52.973270416259766,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.004071352072060108,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 2.1038445437768375e-07,
      "logits/chosen": -2.2982513904571533,
      "logits/rejected": -2.1020822525024414,
      "logps/chosen": -181.23577880859375,
      "logps/ref_chosen": -34.63081359863281,
      "logps/ref_rejected": -83.40669250488281,
      "logps/rejected": -493.41815185546875,
      "loss": 0.7582,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.5955917835235596,
      "rewards/margins": 1.060813307762146,
      "rewards/rejected": -1.6564050912857056,
      "step": 407
    },
    {
      "epoch": 0.5991189427312775,
      "epsilon_dpo/beta": 0.004012344870716333,
      "epsilon_dpo/beta_margin_grad_mean": -0.2729249894618988,
      "epsilon_dpo/beta_margin_grad_std": 0.14225371181964874,
      "epsilon_dpo/beta_margin_mean": 1.1006709337234497,
      "epsilon_dpo/beta_margin_std": 0.758716344833374,
      "epsilon_dpo/loss_margin_mean": 274.95880126953125,
      "grad_norm": 44.9481201171875,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.004044807981699705,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 2.0911786638150872e-07,
      "logits/chosen": -2.361016273498535,
      "logits/rejected": -2.139800548553467,
      "logps/chosen": -198.23095703125,
      "logps/ref_chosen": -46.1392822265625,
      "logps/ref_rejected": -96.3233642578125,
      "logps/rejected": -523.3738403320312,
      "loss": 0.682,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.6118342280387878,
      "rewards/margins": 1.1006710529327393,
      "rewards/rejected": -1.7125052213668823,
      "step": 408
    },
    {
      "epoch": 0.6005873715124816,
      "epsilon_dpo/beta": 0.003982515539973974,
      "epsilon_dpo/beta_margin_grad_mean": -0.294739693403244,
      "epsilon_dpo/beta_margin_grad_std": 0.16220302879810333,
      "epsilon_dpo/beta_margin_mean": 1.007659673690796,
      "epsilon_dpo/beta_margin_std": 0.8669053316116333,
      "epsilon_dpo/loss_margin_mean": 253.92869567871094,
      "grad_norm": 45.330448150634766,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0040122088976204395,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 2.0785235566757517e-07,
      "logits/chosen": -2.2953333854675293,
      "logits/rejected": -2.1765618324279785,
      "logps/chosen": -214.66848754882812,
      "logps/ref_chosen": -48.41924285888672,
      "logps/ref_rejected": -88.46084594726562,
      "logps/rejected": -508.6387939453125,
      "loss": 0.7654,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.6647806167602539,
      "rewards/margins": 1.007659673690796,
      "rewards/rejected": -1.6724402904510498,
      "step": 409
    },
    {
      "epoch": 0.6020558002936858,
      "epsilon_dpo/beta": 0.003950380254536867,
      "epsilon_dpo/beta_margin_grad_mean": -0.27121517062187195,
      "epsilon_dpo/beta_margin_grad_std": 0.15460430085659027,
      "epsilon_dpo/beta_margin_mean": 1.1261425018310547,
      "epsilon_dpo/beta_margin_std": 0.8273832201957703,
      "epsilon_dpo/loss_margin_mean": 285.80035400390625,
      "grad_norm": 39.14712905883789,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.003982341382652521,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 2.065879555832674e-07,
      "logits/chosen": -2.258359909057617,
      "logits/rejected": -2.2356643676757812,
      "logps/chosen": -171.04891967773438,
      "logps/ref_chosen": -32.20702362060547,
      "logps/ref_rejected": -90.97166442871094,
      "logps/rejected": -515.6138916015625,
      "loss": 0.6897,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.5501143336296082,
      "rewards/margins": 1.1261425018310547,
      "rewards/rejected": -1.676256775856018,
      "step": 410
    },
    {
      "epoch": 0.6035242290748899,
      "epsilon_dpo/beta": 0.0039160726591944695,
      "epsilon_dpo/beta_margin_grad_mean": -0.2674121856689453,
      "epsilon_dpo/beta_margin_grad_std": 0.12682108581066132,
      "epsilon_dpo/beta_margin_mean": 1.1220488548278809,
      "epsilon_dpo/beta_margin_std": 0.7217124104499817,
      "epsilon_dpo/loss_margin_mean": 286.949951171875,
      "grad_norm": 38.4173698425293,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.003950245678424835,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 2.0532469944670343e-07,
      "logits/chosen": -2.2671849727630615,
      "logits/rejected": -2.18176007270813,
      "logps/chosen": -163.12564086914062,
      "logps/ref_chosen": -27.866039276123047,
      "logps/ref_rejected": -87.75438690185547,
      "logps/rejected": -509.96392822265625,
      "loss": 0.6557,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.5305934548377991,
      "rewards/margins": 1.1220488548278809,
      "rewards/rejected": -1.6526423692703247,
      "step": 411
    },
    {
      "epoch": 0.604992657856094,
      "epsilon_dpo/beta": 0.0038821042980998755,
      "epsilon_dpo/beta_margin_grad_mean": -0.254239022731781,
      "epsilon_dpo/beta_margin_grad_std": 0.13515068590641022,
      "epsilon_dpo/beta_margin_mean": 1.2237414121627808,
      "epsilon_dpo/beta_margin_std": 0.7976933121681213,
      "epsilon_dpo/loss_margin_mean": 315.66949462890625,
      "grad_norm": 39.3294563293457,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.00391598092392087,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 2.0406262054585738e-07,
      "logits/chosen": -2.2864794731140137,
      "logits/rejected": -2.262669563293457,
      "logps/chosen": -154.23080444335938,
      "logps/ref_chosen": -31.307266235351562,
      "logps/ref_rejected": -107.21038818359375,
      "logps/rejected": -545.803466796875,
      "loss": 0.6232,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.4781460464000702,
      "rewards/margins": 1.2237414121627808,
      "rewards/rejected": -1.7018874883651733,
      "step": 412
    },
    {
      "epoch": 0.6064610866372981,
      "epsilon_dpo/beta": 0.003849643748253584,
      "epsilon_dpo/beta_margin_grad_mean": -0.2595198452472687,
      "epsilon_dpo/beta_margin_grad_std": 0.14788804948329926,
      "epsilon_dpo/beta_margin_mean": 1.1739351749420166,
      "epsilon_dpo/beta_margin_std": 0.7997394800186157,
      "epsilon_dpo/loss_margin_mean": 305.6657409667969,
      "grad_norm": 38.31694412231445,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.003882013261318207,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 2.0280175213768205e-07,
      "logits/chosen": -2.353483200073242,
      "logits/rejected": -2.2761778831481934,
      "logps/chosen": -189.5083465576172,
      "logps/ref_chosen": -43.01777648925781,
      "logps/ref_rejected": -106.10704040527344,
      "logps/rejected": -558.2633666992188,
      "loss": 0.6576,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.5653949975967407,
      "rewards/margins": 1.1739351749420166,
      "rewards/rejected": -1.7393302917480469,
      "step": 413
    },
    {
      "epoch": 0.6079295154185022,
      "epsilon_dpo/beta": 0.0038186372257769108,
      "epsilon_dpo/beta_margin_grad_mean": -0.2614062428474426,
      "epsilon_dpo/beta_margin_grad_std": 0.14156121015548706,
      "epsilon_dpo/beta_margin_mean": 1.181972861289978,
      "epsilon_dpo/beta_margin_std": 0.8064230680465698,
      "epsilon_dpo/loss_margin_mean": 310.2454833984375,
      "grad_norm": 46.121341705322266,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.003849532688036561,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 2.0154212744723247e-07,
      "logits/chosen": -2.2362513542175293,
      "logits/rejected": -2.2514965534210205,
      "logps/chosen": -161.10008239746094,
      "logps/ref_chosen": -33.92742919921875,
      "logps/ref_rejected": -93.77487182617188,
      "logps/rejected": -531.1929931640625,
      "loss": 0.6491,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.4871448874473572,
      "rewards/margins": 1.181972861289978,
      "rewards/rejected": -1.6691176891326904,
      "step": 414
    },
    {
      "epoch": 0.6093979441997063,
      "epsilon_dpo/beta": 0.0037890539970248938,
      "epsilon_dpo/beta_margin_grad_mean": -0.281541645526886,
      "epsilon_dpo/beta_margin_grad_std": 0.14272256195545197,
      "epsilon_dpo/beta_margin_mean": 1.0638197660446167,
      "epsilon_dpo/beta_margin_std": 0.7809419631958008,
      "epsilon_dpo/loss_margin_mean": 281.4879150390625,
      "grad_norm": 42.91494369506836,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.003818507306277752,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 2.002837796667909e-07,
      "logits/chosen": -2.451190948486328,
      "logits/rejected": -2.4177732467651367,
      "logps/chosen": -195.28121948242188,
      "logps/ref_chosen": -45.32706832885742,
      "logps/ref_rejected": -108.54624938964844,
      "logps/rejected": -539.98828125,
      "loss": 0.7054,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.5699288845062256,
      "rewards/margins": 1.0638197660446167,
      "rewards/rejected": -1.6337485313415527,
      "step": 415
    },
    {
      "epoch": 0.6108663729809104,
      "epsilon_dpo/beta": 0.00375731335952878,
      "epsilon_dpo/beta_margin_grad_mean": -0.2511499524116516,
      "epsilon_dpo/beta_margin_grad_std": 0.1389520764350891,
      "epsilon_dpo/beta_margin_mean": 1.231960654258728,
      "epsilon_dpo/beta_margin_std": 0.7866246700286865,
      "epsilon_dpo/loss_margin_mean": 328.50146484375,
      "grad_norm": 41.69633483886719,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.003788906615227461,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 1.990267419549914e-07,
      "logits/chosen": -2.335287094116211,
      "logits/rejected": -2.2950291633605957,
      "logps/chosen": -167.67477416992188,
      "logps/ref_chosen": -38.674949645996094,
      "logps/ref_rejected": -98.17755889892578,
      "logps/rejected": -555.6788330078125,
      "loss": 0.6203,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.4858553111553192,
      "rewards/margins": 1.2319605350494385,
      "rewards/rejected": -1.71781587600708,
      "step": 416
    },
    {
      "epoch": 0.6123348017621145,
      "epsilon_dpo/beta": 0.0037270504981279373,
      "epsilon_dpo/beta_margin_grad_mean": -0.26951104402542114,
      "epsilon_dpo/beta_margin_grad_std": 0.15652604401111603,
      "epsilon_dpo/beta_margin_mean": 1.1333545446395874,
      "epsilon_dpo/beta_margin_std": 0.8189650774002075,
      "epsilon_dpo/loss_margin_mean": 304.8978271484375,
      "grad_norm": 43.83847427368164,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.003757205093279481,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.9777104743594686e-07,
      "logits/chosen": -2.3649911880493164,
      "logits/rejected": -2.098527431488037,
      "logps/chosen": -147.36915588378906,
      "logps/ref_chosen": -26.17800521850586,
      "logps/ref_rejected": -74.53215026855469,
      "logps/rejected": -500.62115478515625,
      "loss": 0.6858,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.4534587562084198,
      "rewards/margins": 1.1333545446395874,
      "rewards/rejected": -1.5868133306503296,
      "step": 417
    },
    {
      "epoch": 0.6138032305433186,
      "epsilon_dpo/beta": 0.0036970123182982206,
      "epsilon_dpo/beta_margin_grad_mean": -0.25392869114875793,
      "epsilon_dpo/beta_margin_grad_std": 0.1371525526046753,
      "epsilon_dpo/beta_margin_mean": 1.2147231101989746,
      "epsilon_dpo/beta_margin_std": 0.7788877487182617,
      "epsilon_dpo/loss_margin_mean": 329.29913330078125,
      "grad_norm": 42.56073760986328,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0037269238382577896,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.965167291983757e-07,
      "logits/chosen": -2.3986239433288574,
      "logits/rejected": -2.3206000328063965,
      "logps/chosen": -180.94284057617188,
      "logps/ref_chosen": -51.68669128417969,
      "logps/ref_rejected": -112.65290832519531,
      "logps/rejected": -571.2081298828125,
      "loss": 0.6258,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.47949230670928955,
      "rewards/margins": 1.2147231101989746,
      "rewards/rejected": -1.6942154169082642,
      "step": 418
    },
    {
      "epoch": 0.6152716593245228,
      "epsilon_dpo/beta": 0.003663749899715185,
      "epsilon_dpo/beta_margin_grad_mean": -0.25385352969169617,
      "epsilon_dpo/beta_margin_grad_std": 0.13596518337726593,
      "epsilon_dpo/beta_margin_mean": 1.1844698190689087,
      "epsilon_dpo/beta_margin_std": 0.7216576337814331,
      "epsilon_dpo/loss_margin_mean": 323.8036804199219,
      "grad_norm": 44.2584228515625,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.0036968865897506475,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 1.9526382029472988e-07,
      "logits/chosen": -2.3918709754943848,
      "logits/rejected": -2.2970142364501953,
      "logps/chosen": -165.10519409179688,
      "logps/ref_chosen": -34.45082473754883,
      "logps/ref_rejected": -98.03851318359375,
      "logps/rejected": -552.49658203125,
      "loss": 0.6311,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.4797900915145874,
      "rewards/margins": 1.1844696998596191,
      "rewards/rejected": -1.664259910583496,
      "step": 419
    },
    {
      "epoch": 0.6167400881057269,
      "epsilon_dpo/beta": 0.0036365704145282507,
      "epsilon_dpo/beta_margin_grad_mean": -0.27526789903640747,
      "epsilon_dpo/beta_margin_grad_std": 0.14264513552188873,
      "epsilon_dpo/beta_margin_mean": 1.0957496166229248,
      "epsilon_dpo/beta_margin_std": 0.781441330909729,
      "epsilon_dpo/loss_margin_mean": 302.1807556152344,
      "grad_norm": 52.36576843261719,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0036636844743043184,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 1.9401235374032425e-07,
      "logits/chosen": -2.424182891845703,
      "logits/rejected": -2.128852605819702,
      "logps/chosen": -174.64231872558594,
      "logps/ref_chosen": -37.82621765136719,
      "logps/ref_rejected": -76.69117736816406,
      "logps/rejected": -515.6880493164062,
      "loss": 0.6883,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.499639630317688,
      "rewards/margins": 1.0957496166229248,
      "rewards/rejected": -1.5953891277313232,
      "step": 420
    },
    {
      "epoch": 0.618208516886931,
      "epsilon_dpo/beta": 0.003604953410103917,
      "epsilon_dpo/beta_margin_grad_mean": -0.27535223960876465,
      "epsilon_dpo/beta_margin_grad_std": 0.13682620227336884,
      "epsilon_dpo/beta_margin_mean": 1.0881061553955078,
      "epsilon_dpo/beta_margin_std": 0.751617968082428,
      "epsilon_dpo/loss_margin_mean": 302.30877685546875,
      "grad_norm": 42.273597717285156,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.003636411391198635,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 1.9276236251246653e-07,
      "logits/chosen": -2.356675863265991,
      "logits/rejected": -2.3096842765808105,
      "logps/chosen": -170.40980529785156,
      "logps/ref_chosen": -33.550575256347656,
      "logps/ref_rejected": -93.99878692626953,
      "logps/rejected": -533.166748046875,
      "loss": 0.6841,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.49425339698791504,
      "rewards/margins": 1.0881061553955078,
      "rewards/rejected": -1.5823595523834229,
      "step": 421
    },
    {
      "epoch": 0.6196769456681351,
      "epsilon_dpo/beta": 0.0035804433282464743,
      "epsilon_dpo/beta_margin_grad_mean": -0.29801541566848755,
      "epsilon_dpo/beta_margin_grad_std": 0.17275013029575348,
      "epsilon_dpo/beta_margin_mean": 0.9903641939163208,
      "epsilon_dpo/beta_margin_std": 0.89506596326828,
      "epsilon_dpo/loss_margin_mean": 277.8594970703125,
      "grad_norm": 46.02511215209961,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0036048688925802708,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 1.9151387954958792e-07,
      "logits/chosen": -2.373016357421875,
      "logits/rejected": -2.3339807987213135,
      "logps/chosen": -231.19760131835938,
      "logps/ref_chosen": -53.665977478027344,
      "logps/ref_rejected": -91.93289947509766,
      "logps/rejected": -547.3240356445312,
      "loss": 0.7896,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.6386697292327881,
      "rewards/margins": 0.9903641939163208,
      "rewards/rejected": -1.6290339231491089,
      "step": 422
    },
    {
      "epoch": 0.6211453744493393,
      "epsilon_dpo/beta": 0.003549282206222415,
      "epsilon_dpo/beta_margin_grad_mean": -0.2701137661933899,
      "epsilon_dpo/beta_margin_grad_std": 0.13685345649719238,
      "epsilon_dpo/beta_margin_mean": 1.1097465753555298,
      "epsilon_dpo/beta_margin_std": 0.7358165383338928,
      "epsilon_dpo/loss_margin_mean": 313.14617919921875,
      "grad_norm": 43.717838287353516,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.003580254502594471,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 1.902669377503756e-07,
      "logits/chosen": -2.299365758895874,
      "logits/rejected": -2.2865123748779297,
      "logps/chosen": -190.4375762939453,
      "logps/ref_chosen": -35.00245666503906,
      "logps/ref_rejected": -91.82589721679688,
      "logps/rejected": -560.4072265625,
      "loss": 0.6702,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.5520890951156616,
      "rewards/margins": 1.1097465753555298,
      "rewards/rejected": -1.6618356704711914,
      "step": 423
    },
    {
      "epoch": 0.6226138032305433,
      "epsilon_dpo/beta": 0.0035229322966188192,
      "epsilon_dpo/beta_margin_grad_mean": -0.30060890316963196,
      "epsilon_dpo/beta_margin_grad_std": 0.15238063037395477,
      "epsilon_dpo/beta_margin_mean": 0.9499651789665222,
      "epsilon_dpo/beta_margin_std": 0.7828956842422485,
      "epsilon_dpo/loss_margin_mean": 270.63690185546875,
      "grad_norm": 42.60936737060547,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0035491990856826305,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 1.890215699729057e-07,
      "logits/chosen": -2.3963804244995117,
      "logits/rejected": -2.1616666316986084,
      "logps/chosen": -187.34298706054688,
      "logps/ref_chosen": -33.510772705078125,
      "logps/ref_rejected": -70.15070343017578,
      "logps/rejected": -494.61981201171875,
      "loss": 0.7749,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.543306291103363,
      "rewards/margins": 0.9499651789665222,
      "rewards/rejected": -1.4932714700698853,
      "step": 424
    },
    {
      "epoch": 0.6240822320117474,
      "epsilon_dpo/beta": 0.0034945050720125437,
      "epsilon_dpo/beta_margin_grad_mean": -0.27929848432540894,
      "epsilon_dpo/beta_margin_grad_std": 0.15515829622745514,
      "epsilon_dpo/beta_margin_mean": 1.0812467336654663,
      "epsilon_dpo/beta_margin_std": 0.8364912867546082,
      "epsilon_dpo/loss_margin_mean": 310.2441101074219,
      "grad_norm": 47.06675338745117,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0035227781627327204,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.8777780903377732e-07,
      "logits/chosen": -2.3238930702209473,
      "logits/rejected": -2.3823156356811523,
      "logps/chosen": -167.17559814453125,
      "logps/ref_chosen": -31.619510650634766,
      "logps/ref_rejected": -103.71711730957031,
      "logps/rejected": -549.517333984375,
      "loss": 0.7154,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.47508615255355835,
      "rewards/margins": 1.0812466144561768,
      "rewards/rejected": -1.5563328266143799,
      "step": 425
    },
    {
      "epoch": 0.6255506607929515,
      "epsilon_dpo/beta": 0.003468525130301714,
      "epsilon_dpo/beta_margin_grad_mean": -0.28968408703804016,
      "epsilon_dpo/beta_margin_grad_std": 0.1473844200372696,
      "epsilon_dpo/beta_margin_mean": 1.007444977760315,
      "epsilon_dpo/beta_margin_std": 0.7751356959342957,
      "epsilon_dpo/loss_margin_mean": 291.418701171875,
      "grad_norm": 66.66638946533203,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.00349438632838428,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 1.8653568770724803e-07,
      "logits/chosen": -2.42264986038208,
      "logits/rejected": -2.2819976806640625,
      "logps/chosen": -187.71844482421875,
      "logps/ref_chosen": -49.42609405517578,
      "logps/ref_rejected": -86.20869445800781,
      "logps/rejected": -515.9197387695312,
      "loss": 0.7381,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.48144960403442383,
      "rewards/margins": 1.007444977760315,
      "rewards/rejected": -1.4888947010040283,
      "step": 426
    },
    {
      "epoch": 0.6270190895741556,
      "epsilon_dpo/beta": 0.0034443254116922617,
      "epsilon_dpo/beta_margin_grad_mean": -0.3163089156150818,
      "epsilon_dpo/beta_margin_grad_std": 0.1538737267255783,
      "epsilon_dpo/beta_margin_mean": 0.8614251017570496,
      "epsilon_dpo/beta_margin_std": 0.760391354560852,
      "epsilon_dpo/loss_margin_mean": 251.1796112060547,
      "grad_norm": 60.45020294189453,
      "kl/avg_steps": 0.703125,
      "kl/beta": 0.0034683735575526953,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 1.8529523872436977e-07,
      "logits/chosen": -2.4879813194274902,
      "logits/rejected": -2.38423490524292,
      "logps/chosen": -194.640869140625,
      "logps/ref_chosen": -40.53302001953125,
      "logps/ref_rejected": -84.44095611572266,
      "logps/rejected": -489.7284240722656,
      "loss": 0.8242,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.5330853462219238,
      "rewards/margins": 0.8614251613616943,
      "rewards/rejected": -1.3945105075836182,
      "step": 427
    },
    {
      "epoch": 0.6284875183553598,
      "epsilon_dpo/beta": 0.0034143617376685143,
      "epsilon_dpo/beta_margin_grad_mean": -0.2803729176521301,
      "epsilon_dpo/beta_margin_grad_std": 0.14126239717006683,
      "epsilon_dpo/beta_margin_mean": 1.0690611600875854,
      "epsilon_dpo/beta_margin_std": 0.7979580163955688,
      "epsilon_dpo/loss_margin_mean": 313.6771545410156,
      "grad_norm": 48.36793899536133,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.0034441568423062563,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 1.8405649477212697e-07,
      "logits/chosen": -2.412050724029541,
      "logits/rejected": -2.416749954223633,
      "logps/chosen": -196.82138061523438,
      "logps/ref_chosen": -36.43036651611328,
      "logps/ref_rejected": -111.91488647460938,
      "logps/rejected": -585.9830322265625,
      "loss": 0.7038,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.5489051938056946,
      "rewards/margins": 1.069061040878296,
      "rewards/rejected": -1.6179664134979248,
      "step": 428
    },
    {
      "epoch": 0.6299559471365639,
      "epsilon_dpo/beta": 0.0033868795726448298,
      "epsilon_dpo/beta_margin_grad_mean": -0.2734263241291046,
      "epsilon_dpo/beta_margin_grad_std": 0.13189740478992462,
      "epsilon_dpo/beta_margin_mean": 1.0902152061462402,
      "epsilon_dpo/beta_margin_std": 0.7191058397293091,
      "epsilon_dpo/loss_margin_mean": 322.5122375488281,
      "grad_norm": 45.01988220214844,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.003414281876757741,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.828194884925749e-07,
      "logits/chosen": -2.5120902061462402,
      "logits/rejected": -2.3635692596435547,
      "logps/chosen": -211.84674072265625,
      "logps/ref_chosen": -49.30812072753906,
      "logps/ref_rejected": -98.94145965576172,
      "logps/rejected": -583.9923095703125,
      "loss": 0.6748,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.5519835352897644,
      "rewards/margins": 1.0902152061462402,
      "rewards/rejected": -1.6421988010406494,
      "step": 429
    },
    {
      "epoch": 0.631424375917768,
      "epsilon_dpo/beta": 0.0033638167660683393,
      "epsilon_dpo/beta_margin_grad_mean": -0.2967393100261688,
      "epsilon_dpo/beta_margin_grad_std": 0.14850302040576935,
      "epsilon_dpo/beta_margin_mean": 0.9720072150230408,
      "epsilon_dpo/beta_margin_std": 0.7672375440597534,
      "epsilon_dpo/loss_margin_mean": 290.09161376953125,
      "grad_norm": 43.82745361328125,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.003386764321476221,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 1.8158425248197928e-07,
      "logits/chosen": -2.4356656074523926,
      "logits/rejected": -2.5032784938812256,
      "logps/chosen": -194.5563507080078,
      "logps/ref_chosen": -45.3841438293457,
      "logps/ref_rejected": -110.27545928955078,
      "logps/rejected": -549.539306640625,
      "loss": 0.7561,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.5032367706298828,
      "rewards/margins": 0.9720072746276855,
      "rewards/rejected": -1.4752440452575684,
      "step": 430
    },
    {
      "epoch": 0.6328928046989721,
      "epsilon_dpo/beta": 0.00333664333447814,
      "epsilon_dpo/beta_margin_grad_mean": -0.2679252624511719,
      "epsilon_dpo/beta_margin_grad_std": 0.13725493848323822,
      "epsilon_dpo/beta_margin_mean": 1.1313966512680054,
      "epsilon_dpo/beta_margin_std": 0.7779242396354675,
      "epsilon_dpo/loss_margin_mean": 339.86541748046875,
      "grad_norm": 37.97023391723633,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.003363639349117875,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.8035081928995788e-07,
      "logits/chosen": -2.4460361003875732,
      "logits/rejected": -2.4488284587860107,
      "logps/chosen": -159.66354370117188,
      "logps/ref_chosen": -34.30770492553711,
      "logps/ref_rejected": -98.43756866455078,
      "logps/rejected": -563.6588134765625,
      "loss": 0.6656,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.4194945991039276,
      "rewards/margins": 1.1313966512680054,
      "rewards/rejected": -1.5508911609649658,
      "step": 431
    },
    {
      "epoch": 0.6343612334801763,
      "epsilon_dpo/beta": 0.0033087090123444796,
      "epsilon_dpo/beta_margin_grad_mean": -0.28203877806663513,
      "epsilon_dpo/beta_margin_grad_std": 0.12971222400665283,
      "epsilon_dpo/beta_margin_mean": 1.032950520515442,
      "epsilon_dpo/beta_margin_std": 0.699227511882782,
      "epsilon_dpo/loss_margin_mean": 312.8155822753906,
      "grad_norm": 44.22251892089844,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.003336530178785324,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 1.791192214186223e-07,
      "logits/chosen": -2.466712236404419,
      "logits/rejected": -2.4444732666015625,
      "logps/chosen": -184.24755859375,
      "logps/ref_chosen": -38.85405349731445,
      "logps/ref_rejected": -105.27049255371094,
      "logps/rejected": -563.4796142578125,
      "loss": 0.7004,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.4821677803993225,
      "rewards/margins": 1.0329504013061523,
      "rewards/rejected": -1.5151182413101196,
      "step": 432
    },
    {
      "epoch": 0.6358296622613803,
      "epsilon_dpo/beta": 0.0032830932177603245,
      "epsilon_dpo/beta_margin_grad_mean": -0.2790597379207611,
      "epsilon_dpo/beta_margin_grad_std": 0.14551861584186554,
      "epsilon_dpo/beta_margin_mean": 1.0783592462539673,
      "epsilon_dpo/beta_margin_std": 0.7789259552955627,
      "epsilon_dpo/loss_margin_mean": 329.2322692871094,
      "grad_norm": 45.43602752685547,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0033086135517805815,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 1.7788949132172193e-07,
      "logits/chosen": -2.3216044902801514,
      "logits/rejected": -2.3884963989257812,
      "logps/chosen": -213.7806396484375,
      "logps/ref_chosen": -35.81452178955078,
      "logps/ref_rejected": -103.11997985839844,
      "logps/rejected": -610.318359375,
      "loss": 0.6991,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.5854209661483765,
      "rewards/margins": 1.0783591270446777,
      "rewards/rejected": -1.6637802124023438,
      "step": 433
    },
    {
      "epoch": 0.6372980910425844,
      "epsilon_dpo/beta": 0.00325866905041039,
      "epsilon_dpo/beta_margin_grad_mean": -0.31264179944992065,
      "epsilon_dpo/beta_margin_grad_std": 0.1338089406490326,
      "epsilon_dpo/beta_margin_mean": 0.8834754228591919,
      "epsilon_dpo/beta_margin_std": 0.7322390079498291,
      "epsilon_dpo/loss_margin_mean": 271.9414367675781,
      "grad_norm": 43.43961715698242,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0032829653937369585,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 1.7666166140378853e-07,
      "logits/chosen": -2.4602980613708496,
      "logits/rejected": -2.3824265003204346,
      "logps/chosen": -245.33560180664062,
      "logps/ref_chosen": -49.93016052246094,
      "logps/ref_rejected": -83.06277465820312,
      "logps/rejected": -550.40966796875,
      "loss": 0.7919,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.6385765075683594,
      "rewards/margins": 0.8834754228591919,
      "rewards/rejected": -1.5220519304275513,
      "step": 434
    },
    {
      "epoch": 0.6387665198237885,
      "epsilon_dpo/beta": 0.003229319117963314,
      "epsilon_dpo/beta_margin_grad_mean": -0.2789570689201355,
      "epsilon_dpo/beta_margin_grad_std": 0.13298888504505157,
      "epsilon_dpo/beta_margin_mean": 1.0672143697738647,
      "epsilon_dpo/beta_margin_std": 0.7533097863197327,
      "epsilon_dpo/loss_margin_mean": 330.9088439941406,
      "grad_norm": 42.341400146484375,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.0032585265580564737,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 1.7543576401928218e-07,
      "logits/chosen": -2.4340808391571045,
      "logits/rejected": -2.4357552528381348,
      "logps/chosen": -165.0016326904297,
      "logps/ref_chosen": -33.67361831665039,
      "logps/ref_rejected": -93.73330688476562,
      "logps/rejected": -555.97021484375,
      "loss": 0.6928,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.4249953627586365,
      "rewards/margins": 1.0672142505645752,
      "rewards/rejected": -1.4922096729278564,
      "step": 435
    },
    {
      "epoch": 0.6402349486049926,
      "epsilon_dpo/beta": 0.0032013256568461657,
      "epsilon_dpo/beta_margin_grad_mean": -0.2841736078262329,
      "epsilon_dpo/beta_margin_grad_std": 0.11989546567201614,
      "epsilon_dpo/beta_margin_mean": 1.0001612901687622,
      "epsilon_dpo/beta_margin_std": 0.6187871694564819,
      "epsilon_dpo/loss_margin_mean": 312.92425537109375,
      "grad_norm": 51.16184616088867,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.003229261375963688,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 1.742118314717391e-07,
      "logits/chosen": -2.5231666564941406,
      "logits/rejected": -2.4640743732452393,
      "logps/chosen": -188.12747192382812,
      "logps/ref_chosen": -42.048667907714844,
      "logps/ref_rejected": -88.41279602050781,
      "logps/rejected": -547.4158325195312,
      "loss": 0.7008,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.4683375954627991,
      "rewards/margins": 1.0001612901687622,
      "rewards/rejected": -1.468498945236206,
      "step": 436
    },
    {
      "epoch": 0.6417033773861968,
      "epsilon_dpo/beta": 0.0031775590032339096,
      "epsilon_dpo/beta_margin_grad_mean": -0.30295178294181824,
      "epsilon_dpo/beta_margin_grad_std": 0.14245736598968506,
      "epsilon_dpo/beta_margin_mean": 0.9152756929397583,
      "epsilon_dpo/beta_margin_std": 0.6979488134384155,
      "epsilon_dpo/loss_margin_mean": 289.0637512207031,
      "grad_norm": 45.89207458496094,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0032012504525482655,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 1.7298989601292036e-07,
      "logits/chosen": -2.4697906970977783,
      "logits/rejected": -2.390430450439453,
      "logps/chosen": -226.48953247070312,
      "logps/ref_chosen": -44.77692413330078,
      "logps/ref_rejected": -86.48928833007812,
      "logps/rejected": -557.265625,
      "loss": 0.7736,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.5791641473770142,
      "rewards/margins": 0.9152756929397583,
      "rewards/rejected": -1.4944398403167725,
      "step": 437
    },
    {
      "epoch": 0.6431718061674009,
      "epsilon_dpo/beta": 0.0031519182957708836,
      "epsilon_dpo/beta_margin_grad_mean": -0.2808303236961365,
      "epsilon_dpo/beta_margin_grad_std": 0.12779134511947632,
      "epsilon_dpo/beta_margin_mean": 1.035179853439331,
      "epsilon_dpo/beta_margin_std": 0.6745886206626892,
      "epsilon_dpo/loss_margin_mean": 329.1087341308594,
      "grad_norm": 40.396175384521484,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0031774197705090046,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.7176998984196144e-07,
      "logits/chosen": -2.473397731781006,
      "logits/rejected": -2.3974690437316895,
      "logps/chosen": -164.937744140625,
      "logps/ref_chosen": -33.662109375,
      "logps/ref_rejected": -89.52166748046875,
      "logps/rejected": -549.906005859375,
      "loss": 0.6943,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.41470038890838623,
      "rewards/margins": 1.035179853439331,
      "rewards/rejected": -1.4498802423477173,
      "step": 438
    },
    {
      "epoch": 0.644640234948605,
      "epsilon_dpo/beta": 0.003126515308395028,
      "epsilon_dpo/beta_margin_grad_mean": -0.3024454414844513,
      "epsilon_dpo/beta_margin_grad_std": 0.1240093857049942,
      "epsilon_dpo/beta_margin_mean": 0.91387939453125,
      "epsilon_dpo/beta_margin_std": 0.6512243747711182,
      "epsilon_dpo/loss_margin_mean": 292.9576721191406,
      "grad_norm": 45.90419006347656,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0031518111936748028,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.7055214510452458e-07,
      "logits/chosen": -2.472698926925659,
      "logits/rejected": -2.5124192237854004,
      "logps/chosen": -196.35198974609375,
      "logps/ref_chosen": -34.986392974853516,
      "logps/ref_rejected": -87.497314453125,
      "logps/rejected": -541.820556640625,
      "loss": 0.7563,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.5056626200675964,
      "rewards/margins": 0.91387939453125,
      "rewards/rejected": -1.4195419549942017,
      "step": 439
    },
    {
      "epoch": 0.6461086637298091,
      "epsilon_dpo/beta": 0.003103271359577775,
      "epsilon_dpo/beta_margin_grad_mean": -0.3023149073123932,
      "epsilon_dpo/beta_margin_grad_std": 0.134578138589859,
      "epsilon_dpo/beta_margin_mean": 0.9336312413215637,
      "epsilon_dpo/beta_margin_std": 0.7152769565582275,
      "epsilon_dpo/loss_margin_mean": 301.76104736328125,
      "grad_norm": 51.92120361328125,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.003126409137621522,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 1.6933639389195134e-07,
      "logits/chosen": -2.5072128772735596,
      "logits/rejected": -2.5406949520111084,
      "logps/chosen": -238.4813232421875,
      "logps/ref_chosen": -53.56586837768555,
      "logps/ref_rejected": -104.3643569946289,
      "logps/rejected": -591.0408935546875,
      "loss": 0.761,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.5756343603134155,
      "rewards/margins": 0.9336312413215637,
      "rewards/rejected": -1.509265661239624,
      "step": 440
    },
    {
      "epoch": 0.6475770925110133,
      "epsilon_dpo/beta": 0.0030753209721297026,
      "epsilon_dpo/beta_margin_grad_mean": -0.2959996163845062,
      "epsilon_dpo/beta_margin_grad_std": 0.12961319088935852,
      "epsilon_dpo/beta_margin_mean": 0.9479500651359558,
      "epsilon_dpo/beta_margin_std": 0.6735605597496033,
      "epsilon_dpo/loss_margin_mean": 308.67352294921875,
      "grad_norm": 40.502708435058594,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.003103135619312525,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 1.681227682404166e-07,
      "logits/chosen": -2.5055270195007324,
      "logits/rejected": -2.469489812850952,
      "logps/chosen": -200.94158935546875,
      "logps/ref_chosen": -39.209449768066406,
      "logps/ref_rejected": -102.78851318359375,
      "logps/rejected": -573.1942138671875,
      "loss": 0.7438,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.49797898530960083,
      "rewards/margins": 0.947950005531311,
      "rewards/rejected": -1.4459290504455566,
      "step": 441
    },
    {
      "epoch": 0.6490455212922174,
      "epsilon_dpo/beta": 0.003050584578886628,
      "epsilon_dpo/beta_margin_grad_mean": -0.2755725681781769,
      "epsilon_dpo/beta_margin_grad_std": 0.11769750714302063,
      "epsilon_dpo/beta_margin_mean": 1.046021819114685,
      "epsilon_dpo/beta_margin_std": 0.6180484294891357,
      "epsilon_dpo/loss_margin_mean": 343.6279602050781,
      "grad_norm": 37.710105895996094,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0030752660240978003,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.669113001300851e-07,
      "logits/chosen": -2.491802930831909,
      "logits/rejected": -2.457737922668457,
      "logps/chosen": -159.370849609375,
      "logps/ref_chosen": -29.0069580078125,
      "logps/ref_rejected": -83.71453857421875,
      "logps/rejected": -557.7063598632812,
      "loss": 0.675,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.39891505241394043,
      "rewards/margins": 1.046021819114685,
      "rewards/rejected": -1.444936990737915,
      "step": 442
    },
    {
      "epoch": 0.6505139500734214,
      "epsilon_dpo/beta": 0.0030279052443802357,
      "epsilon_dpo/beta_margin_grad_mean": -0.3231920897960663,
      "epsilon_dpo/beta_margin_grad_std": 0.15162836015224457,
      "epsilon_dpo/beta_margin_mean": 0.8422841429710388,
      "epsilon_dpo/beta_margin_std": 0.7962964177131653,
      "epsilon_dpo/loss_margin_mean": 279.1518249511719,
      "grad_norm": 46.755645751953125,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.003050480969250202,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 1.6570202148426815e-07,
      "logits/chosen": -2.5168378353118896,
      "logits/rejected": -2.499556541442871,
      "logps/chosen": -233.57965087890625,
      "logps/ref_chosen": -52.68767166137695,
      "logps/ref_rejected": -93.39274597167969,
      "logps/rejected": -553.4365234375,
      "loss": 0.841,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.549048900604248,
      "rewards/margins": 0.8422842025756836,
      "rewards/rejected": -1.3913331031799316,
      "step": 443
    },
    {
      "epoch": 0.6519823788546255,
      "epsilon_dpo/beta": 0.0029996871016919613,
      "epsilon_dpo/beta_margin_grad_mean": -0.26602280139923096,
      "epsilon_dpo/beta_margin_grad_std": 0.12350434064865112,
      "epsilon_dpo/beta_margin_mean": 1.1127806901931763,
      "epsilon_dpo/beta_margin_std": 0.6654544472694397,
      "epsilon_dpo/loss_margin_mean": 371.2488708496094,
      "grad_norm": 38.41836929321289,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.003027772530913353,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 1.6449496416858282e-07,
      "logits/chosen": -2.462678909301758,
      "logits/rejected": -2.5427470207214355,
      "logps/chosen": -178.10812377929688,
      "logps/ref_chosen": -35.20741271972656,
      "logps/ref_rejected": -104.47367858886719,
      "logps/rejected": -618.6232299804688,
      "loss": 0.6502,
      "rewards/accuracies": 0.984375,
      "rewards/chosen": -0.4290057420730591,
      "rewards/margins": 1.1127806901931763,
      "rewards/rejected": -1.5417864322662354,
      "step": 444
    },
    {
      "epoch": 0.6534508076358296,
      "epsilon_dpo/beta": 0.0029783889185637236,
      "epsilon_dpo/beta_margin_grad_mean": -0.31596365571022034,
      "epsilon_dpo/beta_margin_grad_std": 0.1648440957069397,
      "epsilon_dpo/beta_margin_mean": 0.8814136385917664,
      "epsilon_dpo/beta_margin_std": 0.8328824639320374,
      "epsilon_dpo/loss_margin_mean": 297.23583984375,
      "grad_norm": 48.016639709472656,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.002999651012942195,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 1.6329015999011182e-07,
      "logits/chosen": -2.5741279125213623,
      "logits/rejected": -2.5458946228027344,
      "logps/chosen": -216.25303649902344,
      "logps/ref_chosen": -46.78347396850586,
      "logps/ref_rejected": -99.2047119140625,
      "logps/rejected": -565.91015625,
      "loss": 0.8334,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.5069864392280579,
      "rewards/margins": 0.8814135789871216,
      "rewards/rejected": -1.3884000778198242,
      "step": 445
    },
    {
      "epoch": 0.6549192364170338,
      "epsilon_dpo/beta": 0.0029506187420338392,
      "epsilon_dpo/beta_margin_grad_mean": -0.2692401707172394,
      "epsilon_dpo/beta_margin_grad_std": 0.11489293724298477,
      "epsilon_dpo/beta_margin_mean": 1.0957218408584595,
      "epsilon_dpo/beta_margin_std": 0.6562060713768005,
      "epsilon_dpo/loss_margin_mean": 371.6014099121094,
      "grad_norm": 50.814979553222656,
      "kl/avg_steps": 0.9375,
      "kl/beta": 0.002978244796395302,
      "kl/n_epsilon_steps": 0.03125,
      "kl/p_epsilon_steps": 0.96875,
      "learning_rate": 1.6208764069656578e-07,
      "logits/chosen": -2.528761863708496,
      "logits/rejected": -2.6162219047546387,
      "logps/chosen": -160.44252014160156,
      "logps/ref_chosen": -34.56015396118164,
      "logps/ref_rejected": -109.97004699707031,
      "logps/rejected": -607.453857421875,
      "loss": 0.6534,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -0.37178486585617065,
      "rewards/margins": 1.0957218408584595,
      "rewards/rejected": -1.4675066471099854,
      "step": 446
    },
    {
      "epoch": 0.6563876651982379,
      "epsilon_dpo/beta": 0.0029278243891894817,
      "epsilon_dpo/beta_margin_grad_mean": -0.2999436855316162,
      "epsilon_dpo/beta_margin_grad_std": 0.13803143799304962,
      "epsilon_dpo/beta_margin_mean": 0.9522778391838074,
      "epsilon_dpo/beta_margin_std": 0.7474956512451172,
      "epsilon_dpo/loss_margin_mean": 326.1642150878906,
      "grad_norm": 40.274986267089844,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.00295058311894536,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 1.608874379754465e-07,
      "logits/chosen": -2.502809762954712,
      "logits/rejected": -2.5910682678222656,
      "logps/chosen": -199.97299194335938,
      "logps/ref_chosen": -39.49730682373047,
      "logps/ref_rejected": -105.97085571289062,
      "logps/rejected": -592.6107788085938,
      "loss": 0.7578,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.471713662147522,
      "rewards/margins": 0.9522777795791626,
      "rewards/rejected": -1.4239914417266846,
      "step": 447
    },
    {
      "epoch": 0.657856093979442,
      "epsilon_dpo/beta": 0.0029014681931585073,
      "epsilon_dpo/beta_margin_grad_mean": -0.2846612334251404,
      "epsilon_dpo/beta_margin_grad_std": 0.12346479296684265,
      "epsilon_dpo/beta_margin_mean": 1.0155407190322876,
      "epsilon_dpo/beta_margin_std": 0.674639880657196,
      "epsilon_dpo/loss_margin_mean": 350.43487548828125,
      "grad_norm": 41.37987518310547,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.0029277103021740913,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 1.5968958345321177e-07,
      "logits/chosen": -2.5615386962890625,
      "logits/rejected": -2.621204376220703,
      "logps/chosen": -227.657958984375,
      "logps/ref_chosen": -42.827239990234375,
      "logps/ref_rejected": -109.36424255371094,
      "logps/rejected": -644.6298217773438,
      "loss": 0.7027,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.5373802781105042,
      "rewards/margins": 1.0155407190322876,
      "rewards/rejected": -1.5529210567474365,
      "step": 448
    },
    {
      "epoch": 0.6593245227606461,
      "epsilon_dpo/beta": 0.002878130180761218,
      "epsilon_dpo/beta_margin_grad_mean": -0.28992605209350586,
      "epsilon_dpo/beta_margin_grad_std": 0.13492745161056519,
      "epsilon_dpo/beta_margin_mean": 0.9840430617332458,
      "epsilon_dpo/beta_margin_std": 0.6940675973892212,
      "epsilon_dpo/loss_margin_mean": 342.7997131347656,
      "grad_norm": 43.268863677978516,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0029014162719249725,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.584941086944423e-07,
      "logits/chosen": -2.5850870609283447,
      "logits/rejected": -2.595974922180176,
      "logps/chosen": -194.70394897460938,
      "logps/ref_chosen": -36.90496826171875,
      "logps/ref_rejected": -95.95344543457031,
      "logps/rejected": -596.5521240234375,
      "loss": 0.7298,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.4556068181991577,
      "rewards/margins": 0.9840430021286011,
      "rewards/rejected": -1.4396498203277588,
      "step": 449
    },
    {
      "epoch": 0.6607929515418502,
      "epsilon_dpo/beta": 0.002853134647011757,
      "epsilon_dpo/beta_margin_grad_mean": -0.2585708200931549,
      "epsilon_dpo/beta_margin_grad_std": 0.11334740370512009,
      "epsilon_dpo/beta_margin_mean": 1.1357218027114868,
      "epsilon_dpo/beta_margin_std": 0.607581377029419,
      "epsilon_dpo/loss_margin_mean": 398.61895751953125,
      "grad_norm": 37.674678802490234,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.00287803215906024,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 1.573010452010098e-07,
      "logits/chosen": -2.546546459197998,
      "logits/rejected": -2.698519229888916,
      "logps/chosen": -175.88995361328125,
      "logps/ref_chosen": -34.65415573120117,
      "logps/ref_rejected": -108.24179077148438,
      "logps/rejected": -648.0965576171875,
      "loss": 0.6252,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.40372031927108765,
      "rewards/margins": 1.1357218027114868,
      "rewards/rejected": -1.5394420623779297,
      "step": 450
    },
    {
      "epoch": 0.6622613803230544,
      "epsilon_dpo/beta": 0.002834628103300929,
      "epsilon_dpo/beta_margin_grad_mean": -0.32642942667007446,
      "epsilon_dpo/beta_margin_grad_std": 0.15271244943141937,
      "epsilon_dpo/beta_margin_mean": 0.8098753690719604,
      "epsilon_dpo/beta_margin_std": 0.763114869594574,
      "epsilon_dpo/loss_margin_mean": 287.087158203125,
      "grad_norm": 46.581146240234375,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.0028530678246170282,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 1.5611042441124687e-07,
      "logits/chosen": -2.5236382484436035,
      "logits/rejected": -2.479279041290283,
      "logps/chosen": -234.43658447265625,
      "logps/ref_chosen": -42.703250885009766,
      "logps/ref_rejected": -79.43376159667969,
      "logps/rejected": -558.2542724609375,
      "loss": 0.8563,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.5459603071212769,
      "rewards/margins": 0.8098753690719604,
      "rewards/rejected": -1.3558356761932373,
      "step": 451
    },
    {
      "epoch": 0.6637298091042585,
      "epsilon_dpo/beta": 0.0028108321130275726,
      "epsilon_dpo/beta_margin_grad_mean": -0.2964177429676056,
      "epsilon_dpo/beta_margin_grad_std": 0.12497884780168533,
      "epsilon_dpo/beta_margin_mean": 0.9474288821220398,
      "epsilon_dpo/beta_margin_std": 0.651831328868866,
      "epsilon_dpo/loss_margin_mean": 337.69549560546875,
      "grad_norm": 40.06062316894531,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.002834466751664877,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 1.549222776991186e-07,
      "logits/chosen": -2.5114731788635254,
      "logits/rejected": -2.680518627166748,
      "logps/chosen": -166.42276000976562,
      "logps/ref_chosen": -35.80718231201172,
      "logps/ref_rejected": -102.27734375,
      "logps/rejected": -570.5884399414062,
      "loss": 0.7379,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.3679235577583313,
      "rewards/margins": 0.9474288821220398,
      "rewards/rejected": -1.315352439880371,
      "step": 452
    },
    {
      "epoch": 0.6651982378854625,
      "epsilon_dpo/beta": 0.002787314122542739,
      "epsilon_dpo/beta_margin_grad_mean": -0.2953357696533203,
      "epsilon_dpo/beta_margin_grad_std": 0.12554973363876343,
      "epsilon_dpo/beta_margin_mean": 0.9552209973335266,
      "epsilon_dpo/beta_margin_std": 0.6552383899688721,
      "epsilon_dpo/loss_margin_mean": 343.3067321777344,
      "grad_norm": 39.58503723144531,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.002810751087963581,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 1.5373663637339584e-07,
      "logits/chosen": -2.605818748474121,
      "logits/rejected": -2.60537052154541,
      "logps/chosen": -195.30682373046875,
      "logps/ref_chosen": -41.37712860107422,
      "logps/ref_rejected": -87.86880493164062,
      "logps/rejected": -585.105224609375,
      "loss": 0.7343,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.4305250644683838,
      "rewards/margins": 0.9552209973335266,
      "rewards/rejected": -1.3857460021972656,
      "step": 453
    },
    {
      "epoch": 0.6666666666666666,
      "epsilon_dpo/beta": 0.0027683484368026257,
      "epsilon_dpo/beta_margin_grad_mean": -0.29989561438560486,
      "epsilon_dpo/beta_margin_grad_std": 0.14955060184001923,
      "epsilon_dpo/beta_margin_mean": 0.9520212411880493,
      "epsilon_dpo/beta_margin_std": 0.7603475451469421,
      "epsilon_dpo/loss_margin_mean": 345.2506103515625,
      "grad_norm": 55.10441207885742,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0027872337959706783,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 1.5255353167683017e-07,
      "logits/chosen": -2.5729262828826904,
      "logits/rejected": -2.6592321395874023,
      "logps/chosen": -236.1055145263672,
      "logps/ref_chosen": -44.58696746826172,
      "logps/ref_rejected": -90.57184600830078,
      "logps/rejected": -627.3410034179688,
      "loss": 0.7676,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.5324690341949463,
      "rewards/margins": 0.9520212411880493,
      "rewards/rejected": -1.4844902753829956,
      "step": 454
    },
    {
      "epoch": 0.6681350954478708,
      "epsilon_dpo/beta": 0.0027442548889666796,
      "epsilon_dpo/beta_margin_grad_mean": -0.26028749346733093,
      "epsilon_dpo/beta_margin_grad_std": 0.12892557680606842,
      "epsilon_dpo/beta_margin_mean": 1.1716121435165405,
      "epsilon_dpo/beta_margin_std": 0.7572227120399475,
      "epsilon_dpo/loss_margin_mean": 427.5795593261719,
      "grad_norm": 33.066959381103516,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.002768202219158411,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 1.5137299478533064e-07,
      "logits/chosen": -2.561234474182129,
      "logits/rejected": -2.784088611602783,
      "logps/chosen": -160.47991943359375,
      "logps/ref_chosen": -22.870223999023438,
      "logps/ref_rejected": -121.32386779785156,
      "logps/rejected": -686.5131225585938,
      "loss": 0.6379,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.37842369079589844,
      "rewards/margins": 1.171612024307251,
      "rewards/rejected": -1.5500357151031494,
      "step": 455
    },
    {
      "epoch": 0.6696035242290749,
      "epsilon_dpo/beta": 0.002723881509155035,
      "epsilon_dpo/beta_margin_grad_mean": -0.2826097905635834,
      "epsilon_dpo/beta_margin_grad_std": 0.13924254477024078,
      "epsilon_dpo/beta_margin_mean": 1.053339958190918,
      "epsilon_dpo/beta_margin_std": 0.7655569314956665,
      "epsilon_dpo/loss_margin_mean": 387.81005859375,
      "grad_norm": 35.979026794433594,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.002744190627709031,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 1.5019505680714232e-07,
      "logits/chosen": -2.559840202331543,
      "logits/rejected": -2.7625198364257812,
      "logps/chosen": -205.63754272460938,
      "logps/ref_chosen": -40.844276428222656,
      "logps/ref_rejected": -111.70032501220703,
      "logps/rejected": -664.3036499023438,
      "loss": 0.7064,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.4502452611923218,
      "rewards/margins": 1.053339958190918,
      "rewards/rejected": -1.5035851001739502,
      "step": 456
    },
    {
      "epoch": 0.671071953010279,
      "epsilon_dpo/beta": 0.0027019022963941097,
      "epsilon_dpo/beta_margin_grad_mean": -0.3032781481742859,
      "epsilon_dpo/beta_margin_grad_std": 0.13411802053451538,
      "epsilon_dpo/beta_margin_mean": 0.900545597076416,
      "epsilon_dpo/beta_margin_std": 0.6524316668510437,
      "epsilon_dpo/loss_margin_mean": 334.1934509277344,
      "grad_norm": 48.26573944091797,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.002723762532696128,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.4901974878202627e-07,
      "logits/chosen": -2.6022448539733887,
      "logits/rejected": -2.619676351547241,
      "logps/chosen": -215.42941284179688,
      "logps/ref_chosen": -38.554141998291016,
      "logps/ref_rejected": -90.09440612792969,
      "logps/rejected": -601.1631469726562,
      "loss": 0.7706,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.47887659072875977,
      "rewards/margins": 0.900545597076416,
      "rewards/rejected": -1.3794221878051758,
      "step": 457
    },
    {
      "epoch": 0.6725403817914831,
      "epsilon_dpo/beta": 0.002680970588698983,
      "epsilon_dpo/beta_margin_grad_mean": -0.29450592398643494,
      "epsilon_dpo/beta_margin_grad_std": 0.12352831661701202,
      "epsilon_dpo/beta_margin_mean": 0.951833963394165,
      "epsilon_dpo/beta_margin_std": 0.6352096199989319,
      "epsilon_dpo/loss_margin_mean": 355.9402770996094,
      "grad_norm": 49.126129150390625,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0027018103282898664,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 1.4784710168044212e-07,
      "logits/chosen": -2.625985622406006,
      "logits/rejected": -2.772733211517334,
      "logps/chosen": -206.733642578125,
      "logps/ref_chosen": -37.41191482543945,
      "logps/ref_rejected": -104.581298828125,
      "logps/rejected": -629.8433227539062,
      "loss": 0.732,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.4552924931049347,
      "rewards/margins": 0.951833963394165,
      "rewards/rejected": -1.4071264266967773,
      "step": 458
    },
    {
      "epoch": 0.6740088105726872,
      "epsilon_dpo/beta": 0.002659349935129285,
      "epsilon_dpo/beta_margin_grad_mean": -0.31732696294784546,
      "epsilon_dpo/beta_margin_grad_std": 0.12530642747879028,
      "epsilon_dpo/beta_margin_mean": 0.8335932493209839,
      "epsilon_dpo/beta_margin_std": 0.6189974546432495,
      "epsilon_dpo/loss_margin_mean": 314.1438903808594,
      "grad_norm": 36.275413513183594,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0026808660477399826,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.466771464027316e-07,
      "logits/chosen": -2.501951217651367,
      "logits/rejected": -2.667633533477783,
      "logps/chosen": -192.38290405273438,
      "logps/ref_chosen": -32.51487350463867,
      "logps/ref_rejected": -90.99087524414062,
      "logps/rejected": -565.0028076171875,
      "loss": 0.8004,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.4260733127593994,
      "rewards/margins": 0.8335932493209839,
      "rewards/rejected": -1.2596666812896729,
      "step": 459
    },
    {
      "epoch": 0.6754772393538914,
      "epsilon_dpo/beta": 0.0026354235596954823,
      "epsilon_dpo/beta_margin_grad_mean": -0.2933025658130646,
      "epsilon_dpo/beta_margin_grad_std": 0.11362046003341675,
      "epsilon_dpo/beta_margin_mean": 0.9537962079048157,
      "epsilon_dpo/beta_margin_std": 0.6000516414642334,
      "epsilon_dpo/loss_margin_mean": 362.2628479003906,
      "grad_norm": 46.03992462158203,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.0026592593640089035,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 1.4550991377830423e-07,
      "logits/chosen": -2.543520212173462,
      "logits/rejected": -2.7920000553131104,
      "logps/chosen": -206.57542419433594,
      "logps/ref_chosen": -31.02279281616211,
      "logps/ref_rejected": -110.9461669921875,
      "logps/rejected": -648.7615966796875,
      "loss": 0.7211,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -0.4630896747112274,
      "rewards/margins": 0.9537962079048157,
      "rewards/rejected": -1.4168858528137207,
      "step": 460
    },
    {
      "epoch": 0.6769456681350955,
      "epsilon_dpo/beta": 0.002612577984109521,
      "epsilon_dpo/beta_margin_grad_mean": -0.31691277027130127,
      "epsilon_dpo/beta_margin_grad_std": 0.11491651087999344,
      "epsilon_dpo/beta_margin_mean": 0.8249220252037048,
      "epsilon_dpo/beta_margin_std": 0.5639069676399231,
      "epsilon_dpo/loss_margin_mean": 316.2416687011719,
      "grad_norm": 37.75797653198242,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.0026353762950748205,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 1.4434543456482518e-07,
      "logits/chosen": -2.542848587036133,
      "logits/rejected": -2.7294955253601074,
      "logps/chosen": -206.56666564941406,
      "logps/ref_chosen": -35.32524108886719,
      "logps/ref_rejected": -93.41868591308594,
      "logps/rejected": -580.9017333984375,
      "loss": 0.7926,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.44818025827407837,
      "rewards/margins": 0.8249219655990601,
      "rewards/rejected": -1.2731022834777832,
      "step": 461
    },
    {
      "epoch": 0.6784140969162996,
      "epsilon_dpo/beta": 0.0025939985644072294,
      "epsilon_dpo/beta_margin_grad_mean": -0.3237418830394745,
      "epsilon_dpo/beta_margin_grad_std": 0.13112208247184753,
      "epsilon_dpo/beta_margin_mean": 0.8129886984825134,
      "epsilon_dpo/beta_margin_std": 0.6604252457618713,
      "epsilon_dpo/loss_margin_mean": 314.4653015136719,
      "grad_norm": 40.553749084472656,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.00261251674965024,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 1.4318373944740484e-07,
      "logits/chosen": -2.639768123626709,
      "logits/rejected": -2.776127338409424,
      "logps/chosen": -216.20462036132812,
      "logps/ref_chosen": -44.890872955322266,
      "logps/ref_rejected": -85.42142486572266,
      "logps/rejected": -571.200439453125,
      "loss": 0.8226,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.446227490901947,
      "rewards/margins": 0.8129886984825134,
      "rewards/rejected": -1.2592161893844604,
      "step": 462
    },
    {
      "epoch": 0.6798825256975036,
      "epsilon_dpo/beta": 0.0025706232991069555,
      "epsilon_dpo/beta_margin_grad_mean": -0.28741469979286194,
      "epsilon_dpo/beta_margin_grad_std": 0.12530295550823212,
      "epsilon_dpo/beta_margin_mean": 1.0049502849578857,
      "epsilon_dpo/beta_margin_std": 0.6909827589988708,
      "epsilon_dpo/loss_margin_mean": 391.4453125,
      "grad_norm": 36.294189453125,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.0025938733015209436,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 1.4202485903778976e-07,
      "logits/chosen": -2.6299610137939453,
      "logits/rejected": -2.773591995239258,
      "logps/chosen": -152.20263671875,
      "logps/ref_chosen": -27.038963317871094,
      "logps/ref_rejected": -98.03726196289062,
      "logps/rejected": -614.646240234375,
      "loss": 0.712,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.32232174277305603,
      "rewards/margins": 1.0049502849578857,
      "rewards/rejected": -1.3272720575332642,
      "step": 463
    },
    {
      "epoch": 0.6813509544787077,
      "epsilon_dpo/beta": 0.00254994654096663,
      "epsilon_dpo/beta_margin_grad_mean": -0.2655731737613678,
      "epsilon_dpo/beta_margin_grad_std": 0.12443613260984421,
      "epsilon_dpo/beta_margin_mean": 1.1106587648391724,
      "epsilon_dpo/beta_margin_std": 0.6612549424171448,
      "epsilon_dpo/loss_margin_mean": 436.4868469238281,
      "grad_norm": 48.614654541015625,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0025705774314701557,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.4086882387355658e-07,
      "logits/chosen": -2.5940399169921875,
      "logits/rejected": -2.805851697921753,
      "logps/chosen": -189.71588134765625,
      "logps/ref_chosen": -33.55242919921875,
      "logps/ref_rejected": -109.08905029296875,
      "logps/rejected": -701.7393798828125,
      "loss": 0.6506,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.3993784487247467,
      "rewards/margins": 1.1106586456298828,
      "rewards/rejected": -1.5100371837615967,
      "step": 464
    },
    {
      "epoch": 0.6828193832599119,
      "epsilon_dpo/beta": 0.0025285982992500067,
      "epsilon_dpo/beta_margin_grad_mean": -0.28670158982276917,
      "epsilon_dpo/beta_margin_grad_std": 0.11507753282785416,
      "epsilon_dpo/beta_margin_mean": 0.9909960627555847,
      "epsilon_dpo/beta_margin_std": 0.621567964553833,
      "epsilon_dpo/loss_margin_mean": 392.5635986328125,
      "grad_norm": 35.46124267578125,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.0025498599279671907,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 1.3971566441730714e-07,
      "logits/chosen": -2.6725172996520996,
      "logits/rejected": -2.8897171020507812,
      "logps/chosen": -178.53143310546875,
      "logps/ref_chosen": -35.28005599975586,
      "logps/ref_rejected": -116.3499755859375,
      "logps/rejected": -652.1649780273438,
      "loss": 0.7035,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.36298391222953796,
      "rewards/margins": 0.9909961223602295,
      "rewards/rejected": -1.3539800643920898,
      "step": 465
    },
    {
      "epoch": 0.684287812041116,
      "epsilon_dpo/beta": 0.0025074416771531105,
      "epsilon_dpo/beta_margin_grad_mean": -0.3023064434528351,
      "epsilon_dpo/beta_margin_grad_std": 0.11443696171045303,
      "epsilon_dpo/beta_margin_mean": 0.9071800112724304,
      "epsilon_dpo/beta_margin_std": 0.6066918969154358,
      "epsilon_dpo/loss_margin_mean": 362.46087646484375,
      "grad_norm": 32.29966354370117,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.002528525423258543,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 1.3856541105586545e-07,
      "logits/chosen": -2.5877268314361572,
      "logits/rejected": -2.7581279277801514,
      "logps/chosen": -175.187255859375,
      "logps/ref_chosen": -28.18646240234375,
      "logps/ref_rejected": -97.64432525634766,
      "logps/rejected": -607.10595703125,
      "loss": 0.749,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.3691789507865906,
      "rewards/margins": 0.9071800112724304,
      "rewards/rejected": -1.2763590812683105,
      "step": 466
    },
    {
      "epoch": 0.6857562408223201,
      "epsilon_dpo/beta": 0.0024880296550691128,
      "epsilon_dpo/beta_margin_grad_mean": -0.2942372262477875,
      "epsilon_dpo/beta_margin_grad_std": 0.15898768603801727,
      "epsilon_dpo/beta_margin_mean": 0.9923102259635925,
      "epsilon_dpo/beta_margin_std": 0.829995334148407,
      "epsilon_dpo/loss_margin_mean": 400.21441650390625,
      "grad_norm": 47.59917449951172,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.002507369499653578,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 1.3741809409947729e-07,
      "logits/chosen": -2.691765308380127,
      "logits/rejected": -2.9094629287719727,
      "logps/chosen": -223.679443359375,
      "logps/ref_chosen": -46.7025146484375,
      "logps/ref_rejected": -110.00337982177734,
      "logps/rejected": -687.1947021484375,
      "loss": 0.7652,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.44280189275741577,
      "rewards/margins": 0.9923102259635925,
      "rewards/rejected": -1.4351122379302979,
      "step": 467
    },
    {
      "epoch": 0.6872246696035242,
      "epsilon_dpo/beta": 0.0024679647758603096,
      "epsilon_dpo/beta_margin_grad_mean": -0.29025694727897644,
      "epsilon_dpo/beta_margin_grad_std": 0.14004936814308167,
      "epsilon_dpo/beta_margin_mean": 1.0126533508300781,
      "epsilon_dpo/beta_margin_std": 0.7630032896995544,
      "epsilon_dpo/loss_margin_mean": 411.1828918457031,
      "grad_norm": 40.286834716796875,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.002487932564690709,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.362737437810114e-07,
      "logits/chosen": -2.67857027053833,
      "logits/rejected": -2.87001895904541,
      "logps/chosen": -192.12745666503906,
      "logps/ref_chosen": -42.05735778808594,
      "logps/ref_rejected": -109.48826599121094,
      "logps/rejected": -670.7412109375,
      "loss": 0.728,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.37121495604515076,
      "rewards/margins": 1.0126533508300781,
      "rewards/rejected": -1.3838683366775513,
      "step": 468
    },
    {
      "epoch": 0.6886930983847284,
      "epsilon_dpo/beta": 0.002445760415866971,
      "epsilon_dpo/beta_margin_grad_mean": -0.2844110131263733,
      "epsilon_dpo/beta_margin_grad_std": 0.1195027157664299,
      "epsilon_dpo/beta_margin_mean": 1.0001988410949707,
      "epsilon_dpo/beta_margin_std": 0.6260125637054443,
      "epsilon_dpo/loss_margin_mean": 409.49169921875,
      "grad_norm": 48.018531799316406,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.002467880956828594,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 1.351323902551631e-07,
      "logits/chosen": -2.6168570518493652,
      "logits/rejected": -2.8470990657806396,
      "logps/chosen": -206.51370239257812,
      "logps/ref_chosen": -36.20582580566406,
      "logps/ref_rejected": -111.1355972290039,
      "logps/rejected": -690.9351806640625,
      "loss": 0.7019,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -0.4174906611442566,
      "rewards/margins": 1.0001988410949707,
      "rewards/rejected": -1.417689561843872,
      "step": 469
    },
    {
      "epoch": 0.6901615271659325,
      "epsilon_dpo/beta": 0.002425323473289609,
      "epsilon_dpo/beta_margin_grad_mean": -0.2954522371292114,
      "epsilon_dpo/beta_margin_grad_std": 0.12080518156290054,
      "epsilon_dpo/beta_margin_mean": 0.9496338963508606,
      "epsilon_dpo/beta_margin_std": 0.6434259414672852,
      "epsilon_dpo/loss_margin_mean": 392.281005859375,
      "grad_norm": 35.75682830810547,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.002445716643705964,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 1.339940635976592e-07,
      "logits/chosen": -2.6013760566711426,
      "logits/rejected": -2.87990665435791,
      "logps/chosen": -157.40719604492188,
      "logps/ref_chosen": -31.09160804748535,
      "logps/ref_rejected": -87.30916595458984,
      "logps/rejected": -605.90576171875,
      "loss": 0.7329,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.30687910318374634,
      "rewards/margins": 0.9496338367462158,
      "rewards/rejected": -1.2565128803253174,
      "step": 470
    },
    {
      "epoch": 0.6916299559471366,
      "epsilon_dpo/beta": 0.002405789215117693,
      "epsilon_dpo/beta_margin_grad_mean": -0.30035099387168884,
      "epsilon_dpo/beta_margin_grad_std": 0.11763791739940643,
      "epsilon_dpo/beta_margin_mean": 0.9167090654373169,
      "epsilon_dpo/beta_margin_std": 0.606320858001709,
      "epsilon_dpo/loss_margin_mean": 381.869140625,
      "grad_norm": 35.150997161865234,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.002425253624096513,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.3285879380446563e-07,
      "logits/chosen": -2.7026596069335938,
      "logits/rejected": -2.8277225494384766,
      "logps/chosen": -211.00067138671875,
      "logps/ref_chosen": -44.132484436035156,
      "logps/ref_rejected": -89.70744323730469,
      "logps/rejected": -638.4447631835938,
      "loss": 0.745,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.4026249945163727,
      "rewards/margins": 0.9167090654373169,
      "rewards/rejected": -1.3193340301513672,
      "step": 471
    },
    {
      "epoch": 0.6930983847283406,
      "epsilon_dpo/beta": 0.002386399544775486,
      "epsilon_dpo/beta_margin_grad_mean": -0.3045946955680847,
      "epsilon_dpo/beta_margin_grad_std": 0.14598874747753143,
      "epsilon_dpo/beta_margin_mean": 0.9562135934829712,
      "epsilon_dpo/beta_margin_std": 0.8416454792022705,
      "epsilon_dpo/loss_margin_mean": 401.76422119140625,
      "grad_norm": 36.8669548034668,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.002405707258731127,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.317266107909975e-07,
      "logits/chosen": -2.770143985748291,
      "logits/rejected": -2.981727361679077,
      "logps/chosen": -265.7175598144531,
      "logps/ref_chosen": -59.834197998046875,
      "logps/ref_rejected": -123.57960510253906,
      "logps/rejected": -731.2271728515625,
      "loss": 0.778,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.4928061366081238,
      "rewards/margins": 0.9562135934829712,
      "rewards/rejected": -1.4490196704864502,
      "step": 472
    },
    {
      "epoch": 0.6945668135095447,
      "epsilon_dpo/beta": 0.0023671663366258144,
      "epsilon_dpo/beta_margin_grad_mean": -0.3334435224533081,
      "epsilon_dpo/beta_margin_grad_std": 0.130352184176445,
      "epsilon_dpo/beta_margin_mean": 0.7624538540840149,
      "epsilon_dpo/beta_margin_std": 0.6814138889312744,
      "epsilon_dpo/loss_margin_mean": 323.02276611328125,
      "grad_norm": 50.768165588378906,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0023863185197114944,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.3059754439133002e-07,
      "logits/chosen": -2.749878406524658,
      "logits/rejected": -2.8519415855407715,
      "logps/chosen": -238.464599609375,
      "logps/ref_chosen": -44.87860870361328,
      "logps/ref_rejected": -85.82889556884766,
      "logps/rejected": -602.4376220703125,
      "loss": 0.8578,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.4595049023628235,
      "rewards/margins": 0.7624538540840149,
      "rewards/rejected": -1.2219587564468384,
      "step": 473
    },
    {
      "epoch": 0.6960352422907489,
      "epsilon_dpo/beta": 0.0023473482578992844,
      "epsilon_dpo/beta_margin_grad_mean": -0.3070417642593384,
      "epsilon_dpo/beta_margin_grad_std": 0.12120737135410309,
      "epsilon_dpo/beta_margin_mean": 0.889529287815094,
      "epsilon_dpo/beta_margin_std": 0.6321231126785278,
      "epsilon_dpo/loss_margin_mean": 379.68218994140625,
      "grad_norm": 44.010650634765625,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.00236708577722311,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 1.2947162435741277e-07,
      "logits/chosen": -2.6113648414611816,
      "logits/rejected": -2.8044967651367188,
      "logps/chosen": -195.01571655273438,
      "logps/ref_chosen": -30.269367218017578,
      "logps/ref_rejected": -97.37470245361328,
      "logps/rejected": -641.80322265625,
      "loss": 0.7665,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.38724249601364136,
      "rewards/margins": 0.889529287815094,
      "rewards/rejected": -1.2767717838287354,
      "step": 474
    },
    {
      "epoch": 0.697503671071953,
      "epsilon_dpo/beta": 0.002328441943973303,
      "epsilon_dpo/beta_margin_grad_mean": -0.32659581303596497,
      "epsilon_dpo/beta_margin_grad_std": 0.1249208152294159,
      "epsilon_dpo/beta_margin_mean": 0.8016044497489929,
      "epsilon_dpo/beta_margin_std": 0.6676663160324097,
      "epsilon_dpo/loss_margin_mean": 345.1247863769531,
      "grad_norm": 35.23362350463867,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0023472807370126247,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.2834888035828596e-07,
      "logits/chosen": -2.689065933227539,
      "logits/rejected": -2.8997483253479004,
      "logps/chosen": -206.45596313476562,
      "logps/ref_chosen": -34.96168518066406,
      "logps/ref_rejected": -94.91036987304688,
      "logps/rejected": -611.5294189453125,
      "loss": 0.8274,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.4007193148136139,
      "rewards/margins": 0.8016044497489929,
      "rewards/rejected": -1.2023236751556396,
      "step": 475
    },
    {
      "epoch": 0.6989720998531571,
      "epsilon_dpo/beta": 0.002310403622686863,
      "epsilon_dpo/beta_margin_grad_mean": -0.32414335012435913,
      "epsilon_dpo/beta_margin_grad_std": 0.12600372731685638,
      "epsilon_dpo/beta_margin_mean": 0.8043047189712524,
      "epsilon_dpo/beta_margin_std": 0.6316364407539368,
      "epsilon_dpo/loss_margin_mean": 349.0835876464844,
      "grad_norm": 40.693206787109375,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0023283627815544605,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 1.2722934197929802e-07,
      "logits/chosen": -2.668698310852051,
      "logits/rejected": -2.884255886077881,
      "logps/chosen": -192.21990966796875,
      "logps/ref_chosen": -28.34685516357422,
      "logps/ref_rejected": -78.29444885253906,
      "logps/rejected": -591.2510986328125,
      "loss": 0.8208,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.37969422340393066,
      "rewards/margins": 0.8043047189712524,
      "rewards/rejected": -1.183998942375183,
      "step": 476
    },
    {
      "epoch": 0.7004405286343612,
      "epsilon_dpo/beta": 0.0022924933582544327,
      "epsilon_dpo/beta_margin_grad_mean": -0.3193974196910858,
      "epsilon_dpo/beta_margin_grad_std": 0.11756820976734161,
      "epsilon_dpo/beta_margin_mean": 0.8161155581474304,
      "epsilon_dpo/beta_margin_std": 0.5828248858451843,
      "epsilon_dpo/loss_margin_mean": 356.9313049316406,
      "grad_norm": 37.30575942993164,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.002310313517227769,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 1.2611303872132631e-07,
      "logits/chosen": -2.7471470832824707,
      "logits/rejected": -2.842437267303467,
      "logps/chosen": -217.6685028076172,
      "logps/ref_chosen": -42.40911865234375,
      "logps/ref_rejected": -82.68942260742188,
      "logps/rejected": -614.880126953125,
      "loss": 0.802,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.40344566106796265,
      "rewards/margins": 0.8161154985427856,
      "rewards/rejected": -1.219561219215393,
      "step": 477
    },
    {
      "epoch": 0.7019089574155654,
      "epsilon_dpo/beta": 0.0022747220937162638,
      "epsilon_dpo/beta_margin_grad_mean": -0.3177875578403473,
      "epsilon_dpo/beta_margin_grad_std": 0.11528382450342178,
      "epsilon_dpo/beta_margin_mean": 0.8234969973564148,
      "epsilon_dpo/beta_margin_std": 0.577464759349823,
      "epsilon_dpo/loss_margin_mean": 362.92669677734375,
      "grad_norm": 39.92375564575195,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0022924039512872696,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 1.2500000000000005e-07,
      "logits/chosen": -2.6334609985351562,
      "logits/rejected": -2.921363592147827,
      "logps/chosen": -207.5630340576172,
      "logps/ref_chosen": -28.737815856933594,
      "logps/ref_rejected": -90.47331237792969,
      "logps/rejected": -632.2252197265625,
      "loss": 0.7954,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.407898485660553,
      "rewards/margins": 0.8234970569610596,
      "rewards/rejected": -1.2313954830169678,
      "step": 478
    },
    {
      "epoch": 0.7033773861967695,
      "epsilon_dpo/beta": 0.002257088664919138,
      "epsilon_dpo/beta_margin_grad_mean": -0.3112712800502777,
      "epsilon_dpo/beta_margin_grad_std": 0.1403777301311493,
      "epsilon_dpo/beta_margin_mean": 0.9059340357780457,
      "epsilon_dpo/beta_margin_std": 0.7668508291244507,
      "epsilon_dpo/loss_margin_mean": 402.3663330078125,
      "grad_norm": 42.03719711303711,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0022746333852410316,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 1.2389025514492456e-07,
      "logits/chosen": -2.710920810699463,
      "logits/rejected": -3.0226893424987793,
      "logps/chosen": -194.02374267578125,
      "logps/ref_chosen": -29.93898582458496,
      "logps/ref_rejected": -104.23573303222656,
      "logps/rejected": -670.686767578125,
      "loss": 0.7896,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.37124890089035034,
      "rewards/margins": 0.9059340357780457,
      "rewards/rejected": -1.2771828174591064,
      "step": 479
    },
    {
      "epoch": 0.7048458149779736,
      "epsilon_dpo/beta": 0.0022395916748791933,
      "epsilon_dpo/beta_margin_grad_mean": -0.3270338773727417,
      "epsilon_dpo/beta_margin_grad_std": 0.12246443331241608,
      "epsilon_dpo/beta_margin_mean": 0.7808691263198853,
      "epsilon_dpo/beta_margin_std": 0.6041759848594666,
      "epsilon_dpo/loss_margin_mean": 349.7110595703125,
      "grad_norm": 41.93212127685547,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.002257000654935837,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 1.227838333989088e-07,
      "logits/chosen": -2.7833218574523926,
      "logits/rejected": -2.883632183074951,
      "logps/chosen": -257.56707763671875,
      "logps/ref_chosen": -43.97426223754883,
      "logps/ref_rejected": -87.41323852539062,
      "logps/rejected": -650.7171020507812,
      "loss": 0.8298,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.4799705743789673,
      "rewards/margins": 0.7808691263198853,
      "rewards/rejected": -1.2608397006988525,
      "step": 480
    },
    {
      "epoch": 0.7063142437591777,
      "epsilon_dpo/beta": 0.0022222306579351425,
      "epsilon_dpo/beta_margin_grad_mean": -0.31972822546958923,
      "epsilon_dpo/beta_margin_grad_std": 0.13162846863269806,
      "epsilon_dpo/beta_margin_mean": 0.8373962044715881,
      "epsilon_dpo/beta_margin_std": 0.6917349100112915,
      "epsilon_dpo/loss_margin_mean": 377.9779968261719,
      "grad_norm": 38.608699798583984,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0022395045962184668,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 1.2168076391719489e-07,
      "logits/chosen": -2.710186004638672,
      "logits/rejected": -3.0134506225585938,
      "logps/chosen": -229.94204711914062,
      "logps/ref_chosen": -36.98882293701172,
      "logps/ref_rejected": -98.65377807617188,
      "logps/rejected": -669.5849609375,
      "loss": 0.8128,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.4300834834575653,
      "rewards/margins": 0.8373961448669434,
      "rewards/rejected": -1.267479658126831,
      "step": 481
    },
    {
      "epoch": 0.7077826725403817,
      "epsilon_dpo/beta": 0.0022050042171031237,
      "epsilon_dpo/beta_margin_grad_mean": -0.34148189425468445,
      "epsilon_dpo/beta_margin_grad_std": 0.12353808432817459,
      "epsilon_dpo/beta_margin_mean": 0.7151001691818237,
      "epsilon_dpo/beta_margin_std": 0.6155868172645569,
      "epsilon_dpo/loss_margin_mean": 325.2435607910156,
      "grad_norm": 43.73499298095703,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0022221440449357033,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 1.2058107576668938e-07,
      "logits/chosen": -2.777127504348755,
      "logits/rejected": -3.0118846893310547,
      "logps/chosen": -252.25537109375,
      "logps/ref_chosen": -47.419219970703125,
      "logps/ref_rejected": -92.47096252441406,
      "logps/rejected": -622.5506591796875,
      "loss": 0.875,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.4530524015426636,
      "rewards/margins": 0.7151001691818237,
      "rewards/rejected": -1.1681525707244873,
      "step": 482
    },
    {
      "epoch": 0.7092511013215859,
      "epsilon_dpo/beta": 0.0021872217766940594,
      "epsilon_dpo/beta_margin_grad_mean": -0.3099426329135895,
      "epsilon_dpo/beta_margin_grad_std": 0.1277041733264923,
      "epsilon_dpo/beta_margin_mean": 0.8772286176681519,
      "epsilon_dpo/beta_margin_std": 0.6511039733886719,
      "epsilon_dpo/loss_margin_mean": 402.0494689941406,
      "grad_norm": 37.85783386230469,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0022049180697649717,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.194847979251979e-07,
      "logits/chosen": -2.7550508975982666,
      "logits/rejected": -3.005962371826172,
      "logps/chosen": -219.78411865234375,
      "logps/ref_chosen": -39.672393798828125,
      "logps/ref_rejected": -100.94681549072266,
      "logps/rejected": -683.1080322265625,
      "loss": 0.7801,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.39537718892097473,
      "rewards/margins": 0.8772286176681519,
      "rewards/rejected": -1.2726057767868042,
      "step": 483
    },
    {
      "epoch": 0.71071953010279,
      "epsilon_dpo/beta": 0.002174378838390112,
      "epsilon_dpo/beta_margin_grad_mean": -0.35192134976387024,
      "epsilon_dpo/beta_margin_grad_std": 0.1350928395986557,
      "epsilon_dpo/beta_margin_mean": 0.6707573533058167,
      "epsilon_dpo/beta_margin_std": 0.6563224792480469,
      "epsilon_dpo/loss_margin_mean": 310.1882019042969,
      "grad_norm": 40.76497268676758,
      "kl/avg_steps": 0.59375,
      "kl/beta": 0.002187147503718734,
      "kl/n_epsilon_steps": 0.203125,
      "kl/p_epsilon_steps": 0.796875,
      "learning_rate": 1.1839195928066101e-07,
      "logits/chosen": -2.699538230895996,
      "logits/rejected": -2.9171078205108643,
      "logps/chosen": -237.5784149169922,
      "logps/ref_chosen": -43.43277359008789,
      "logps/ref_rejected": -89.96736145019531,
      "logps/rejected": -594.3012084960938,
      "loss": 0.9168,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": -0.4243118464946747,
      "rewards/margins": 0.6707573533058167,
      "rewards/rejected": -1.095069169998169,
      "step": 484
    },
    {
      "epoch": 0.7121879588839941,
      "epsilon_dpo/beta": 0.0021574674174189568,
      "epsilon_dpo/beta_margin_grad_mean": -0.3015494644641876,
      "epsilon_dpo/beta_margin_grad_std": 0.12973572313785553,
      "epsilon_dpo/beta_margin_mean": 0.9309406876564026,
      "epsilon_dpo/beta_margin_std": 0.6903952360153198,
      "epsilon_dpo/loss_margin_mean": 432.610107421875,
      "grad_norm": 32.71714782714844,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0021742379758507013,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 1.1730258863039347e-07,
      "logits/chosen": -2.6770548820495605,
      "logits/rejected": -2.999387264251709,
      "logps/chosen": -169.16915893554688,
      "logps/ref_chosen": -37.34454345703125,
      "logps/ref_rejected": -111.42447662353516,
      "logps/rejected": -675.8591918945312,
      "loss": 0.7557,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.28553080558776855,
      "rewards/margins": 0.9309406876564026,
      "rewards/rejected": -1.2164714336395264,
      "step": 485
    },
    {
      "epoch": 0.7136563876651982,
      "epsilon_dpo/beta": 0.002140068681910634,
      "epsilon_dpo/beta_margin_grad_mean": -0.3228636085987091,
      "epsilon_dpo/beta_margin_grad_std": 0.1375182420015335,
      "epsilon_dpo/beta_margin_mean": 0.8183300495147705,
      "epsilon_dpo/beta_margin_std": 0.6934942007064819,
      "epsilon_dpo/loss_margin_mean": 383.51116943359375,
      "grad_norm": 35.33386993408203,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0021573833655565977,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.1621671468032493e-07,
      "logits/chosen": -2.755924701690674,
      "logits/rejected": -3.058840274810791,
      "logps/chosen": -193.83453369140625,
      "logps/ref_chosen": -35.52677536010742,
      "logps/ref_rejected": -99.17495727539062,
      "logps/rejected": -640.993896484375,
      "loss": 0.8285,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.34016796946525574,
      "rewards/margins": 0.8183300495147705,
      "rewards/rejected": -1.1584980487823486,
      "step": 486
    },
    {
      "epoch": 0.7151248164464024,
      "epsilon_dpo/beta": 0.0021214832086116076,
      "epsilon_dpo/beta_margin_grad_mean": -0.30856984853744507,
      "epsilon_dpo/beta_margin_grad_std": 0.10273387283086777,
      "epsilon_dpo/beta_margin_mean": 0.8478483557701111,
      "epsilon_dpo/beta_margin_std": 0.48607146739959717,
      "epsilon_dpo/loss_margin_mean": 400.2981872558594,
      "grad_norm": 43.43318557739258,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.002139996038749814,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 1.1513436604424378e-07,
      "logits/chosen": -2.668581247329712,
      "logits/rejected": -3.001206636428833,
      "logps/chosen": -184.3863525390625,
      "logps/ref_chosen": -31.08715057373047,
      "logps/ref_rejected": -98.84352111816406,
      "logps/rejected": -652.44091796875,
      "loss": 0.763,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.3262515366077423,
      "rewards/margins": 0.8478483557701111,
      "rewards/rejected": -1.1740999221801758,
      "step": 487
    },
    {
      "epoch": 0.7165932452276065,
      "epsilon_dpo/beta": 0.0021050700452178717,
      "epsilon_dpo/beta_margin_grad_mean": -0.33707907795906067,
      "epsilon_dpo/beta_margin_grad_std": 0.12150773406028748,
      "epsilon_dpo/beta_margin_mean": 0.7329168319702148,
      "epsilon_dpo/beta_margin_std": 0.5872458815574646,
      "epsilon_dpo/loss_margin_mean": 349.0928039550781,
      "grad_norm": 35.68146514892578,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0021214333828538656,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 1.1405557124304335e-07,
      "logits/chosen": -2.7139856815338135,
      "logits/rejected": -2.9660186767578125,
      "logps/chosen": -184.39353942871094,
      "logps/ref_chosen": -35.27953338623047,
      "logps/ref_rejected": -89.09225463867188,
      "logps/rejected": -587.299072265625,
      "loss": 0.8576,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.3142842650413513,
      "rewards/margins": 0.7329168319702148,
      "rewards/rejected": -1.047201156616211,
      "step": 488
    },
    {
      "epoch": 0.7180616740088106,
      "epsilon_dpo/beta": 0.0020880938973277807,
      "epsilon_dpo/beta_margin_grad_mean": -0.34188124537467957,
      "epsilon_dpo/beta_margin_grad_std": 0.10775003582239151,
      "epsilon_dpo/beta_margin_mean": 0.6980612874031067,
      "epsilon_dpo/beta_margin_std": 0.5159537196159363,
      "epsilon_dpo/loss_margin_mean": 335.0322570800781,
      "grad_norm": 40.08867263793945,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0021049880888313055,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.1298035870396985e-07,
      "logits/chosen": -2.7437334060668945,
      "logits/rejected": -2.992414712905884,
      "logps/chosen": -199.35702514648438,
      "logps/ref_chosen": -37.423851013183594,
      "logps/ref_rejected": -87.10142517089844,
      "logps/rejected": -584.06689453125,
      "loss": 0.8644,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.3389902710914612,
      "rewards/margins": 0.6980613470077515,
      "rewards/rejected": -1.0370516777038574,
      "step": 489
    },
    {
      "epoch": 0.7195301027900147,
      "epsilon_dpo/beta": 0.002076485427096486,
      "epsilon_dpo/beta_margin_grad_mean": -0.3476724922657013,
      "epsilon_dpo/beta_margin_grad_std": 0.1495998054742813,
      "epsilon_dpo/beta_margin_mean": 0.7162193655967712,
      "epsilon_dpo/beta_margin_std": 0.7605259418487549,
      "epsilon_dpo/loss_margin_mean": 346.98077392578125,
      "grad_norm": 42.104305267333984,
      "kl/avg_steps": 0.5625,
      "kl/beta": 0.0020880228839814663,
      "kl/n_epsilon_steps": 0.21875,
      "kl/p_epsilon_steps": 0.78125,
      "learning_rate": 1.1190875675987355e-07,
      "logits/chosen": -2.7477211952209473,
      "logits/rejected": -3.100935935974121,
      "logps/chosen": -247.21926879882812,
      "logps/ref_chosen": -41.46424102783203,
      "logps/ref_rejected": -115.67326354980469,
      "logps/rejected": -668.4090576171875,
      "loss": 0.9131,
      "rewards/accuracies": 0.796875,
      "rewards/chosen": -0.42905905842781067,
      "rewards/margins": 0.7162193059921265,
      "rewards/rejected": -1.1452784538269043,
      "step": 490
    },
    {
      "epoch": 0.7209985315712188,
      "epsilon_dpo/beta": 0.002060000551864505,
      "epsilon_dpo/beta_margin_grad_mean": -0.3543255031108856,
      "epsilon_dpo/beta_margin_grad_std": 0.11612068116664886,
      "epsilon_dpo/beta_margin_mean": 0.6437206268310547,
      "epsilon_dpo/beta_margin_std": 0.547960102558136,
      "epsilon_dpo/loss_margin_mean": 313.3808288574219,
      "grad_norm": 40.64850997924805,
      "kl/avg_steps": 0.796875,
      "kl/beta": 0.002076343633234501,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 1.1084079364846241e-07,
      "logits/chosen": -2.7214245796203613,
      "logits/rejected": -3.0119452476501465,
      "logps/chosen": -226.95785522460938,
      "logps/ref_chosen": -41.33907699584961,
      "logps/ref_rejected": -79.69932556152344,
      "logps/rejected": -578.698974609375,
      "loss": 0.9096,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.38356825709342957,
      "rewards/margins": 0.6437206268310547,
      "rewards/rejected": -1.0272889137268066,
      "step": 491
    },
    {
      "epoch": 0.7224669603524229,
      "epsilon_dpo/beta": 0.0020440397784113884,
      "epsilon_dpo/beta_margin_grad_mean": -0.360959529876709,
      "epsilon_dpo/beta_margin_grad_std": 0.11504260450601578,
      "epsilon_dpo/beta_margin_mean": 0.6149997711181641,
      "epsilon_dpo/beta_margin_std": 0.5465402603149414,
      "epsilon_dpo/loss_margin_mean": 301.7585754394531,
      "grad_norm": 34.33412551879883,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0020599286071956158,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 1.097764975115576e-07,
      "logits/chosen": -2.7548234462738037,
      "logits/rejected": -2.9680824279785156,
      "logps/chosen": -195.3032989501953,
      "logps/ref_chosen": -31.90703582763672,
      "logps/ref_rejected": -79.67924499511719,
      "logps/rejected": -544.8341064453125,
      "loss": 0.929,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.3345920443534851,
      "rewards/margins": 0.6149997711181641,
      "rewards/rejected": -0.9495918154716492,
      "step": 492
    },
    {
      "epoch": 0.723935389133627,
      "epsilon_dpo/beta": 0.0020301109179854393,
      "epsilon_dpo/beta_margin_grad_mean": -0.3570352792739868,
      "epsilon_dpo/beta_margin_grad_std": 0.11298612505197525,
      "epsilon_dpo/beta_margin_mean": 0.6241260170936584,
      "epsilon_dpo/beta_margin_std": 0.5125545859336853,
      "epsilon_dpo/loss_margin_mean": 308.62005615234375,
      "grad_norm": 38.02676773071289,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0020439601503312588,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 1.0871589639435203e-07,
      "logits/chosen": -2.806032657623291,
      "logits/rejected": -3.0264089107513428,
      "logps/chosen": -245.80068969726562,
      "logps/ref_chosen": -52.45185089111328,
      "logps/ref_rejected": -91.2623291015625,
      "logps/rejected": -593.231201171875,
      "loss": 0.9167,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.3943876028060913,
      "rewards/margins": 0.6241260766983032,
      "rewards/rejected": -1.0185136795043945,
      "step": 493
    },
    {
      "epoch": 0.7254038179148311,
      "epsilon_dpo/beta": 0.0020137112587690353,
      "epsilon_dpo/beta_margin_grad_mean": -0.3272671103477478,
      "epsilon_dpo/beta_margin_grad_std": 0.12577760219573975,
      "epsilon_dpo/beta_margin_mean": 0.7729417085647583,
      "epsilon_dpo/beta_margin_std": 0.6120246052742004,
      "epsilon_dpo/loss_margin_mean": 384.9142150878906,
      "grad_norm": 42.489864349365234,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0020300038158893585,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.0765901824467166e-07,
      "logits/chosen": -2.750584125518799,
      "logits/rejected": -3.075985908508301,
      "logps/chosen": -170.007568359375,
      "logps/ref_chosen": -27.903043746948242,
      "logps/ref_rejected": -92.12089538574219,
      "logps/rejected": -619.1396484375,
      "loss": 0.8385,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.28744399547576904,
      "rewards/margins": 0.7729417085647583,
      "rewards/rejected": -1.0603857040405273,
      "step": 494
    },
    {
      "epoch": 0.7268722466960352,
      "epsilon_dpo/beta": 0.0019955940078943968,
      "epsilon_dpo/beta_margin_grad_mean": -0.3212648630142212,
      "epsilon_dpo/beta_margin_grad_std": 0.1038970947265625,
      "epsilon_dpo/beta_margin_mean": 0.7920843362808228,
      "epsilon_dpo/beta_margin_std": 0.502342939376831,
      "epsilon_dpo/loss_margin_mean": 397.37994384765625,
      "grad_norm": 40.98225021362305,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.0020136430393904448,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 1.0660589091223854e-07,
      "logits/chosen": -2.7651305198669434,
      "logits/rejected": -3.138665199279785,
      "logps/chosen": -212.49075317382812,
      "logps/ref_chosen": -37.603515625,
      "logps/ref_rejected": -97.60113525390625,
      "logps/rejected": -669.8682861328125,
      "loss": 0.8,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.3492244780063629,
      "rewards/margins": 0.7920843362808228,
      "rewards/rejected": -1.1413087844848633,
      "step": 495
    },
    {
      "epoch": 0.7283406754772394,
      "epsilon_dpo/beta": 0.001978294923901558,
      "epsilon_dpo/beta_margin_grad_mean": -0.3371340036392212,
      "epsilon_dpo/beta_margin_grad_std": 0.09828732907772064,
      "epsilon_dpo/beta_margin_mean": 0.7093020081520081,
      "epsilon_dpo/beta_margin_std": 0.45665252208709717,
      "epsilon_dpo/loss_margin_mean": 359.1010437011719,
      "grad_norm": 36.258426666259766,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.001995558151975274,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 1.0555654214793722e-07,
      "logits/chosen": -2.80366587638855,
      "logits/rejected": -3.1190853118896484,
      "logps/chosen": -235.90985107421875,
      "logps/ref_chosen": -45.088035583496094,
      "logps/ref_rejected": -92.02516174316406,
      "logps/rejected": -641.947998046875,
      "loss": 0.8456,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.3781554102897644,
      "rewards/margins": 0.7093019485473633,
      "rewards/rejected": -1.0874574184417725,
      "step": 496
    },
    {
      "epoch": 0.7298091042584435,
      "epsilon_dpo/beta": 0.001961135072633624,
      "epsilon_dpo/beta_margin_grad_mean": -0.3387848436832428,
      "epsilon_dpo/beta_margin_grad_std": 0.10064252465963364,
      "epsilon_dpo/beta_margin_mean": 0.7051173448562622,
      "epsilon_dpo/beta_margin_std": 0.4776591360569,
      "epsilon_dpo/loss_margin_mean": 360.11083984375,
      "grad_norm": 31.69774627685547,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.001978248590603471,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 1.0451099960308374e-07,
      "logits/chosen": -2.7518420219421387,
      "logits/rejected": -3.1248888969421387,
      "logps/chosen": -182.076416015625,
      "logps/ref_chosen": -30.02985382080078,
      "logps/ref_rejected": -81.73121643066406,
      "logps/rejected": -593.888671875,
      "loss": 0.8518,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.2986891269683838,
      "rewards/margins": 0.7051173448562622,
      "rewards/rejected": -1.003806471824646,
      "step": 497
    },
    {
      "epoch": 0.7312775330396476,
      "epsilon_dpo/beta": 0.0019471884006634355,
      "epsilon_dpo/beta_margin_grad_mean": -0.35530975461006165,
      "epsilon_dpo/beta_margin_grad_std": 0.13022439181804657,
      "epsilon_dpo/beta_margin_mean": 0.6506726145744324,
      "epsilon_dpo/beta_margin_std": 0.6256114840507507,
      "epsilon_dpo/loss_margin_mean": 335.5245056152344,
      "grad_norm": 43.087093353271484,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0019610889721661806,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 1.0346929082869641e-07,
      "logits/chosen": -2.85072660446167,
      "logits/rejected": -3.1245601177215576,
      "logps/chosen": -252.92184448242188,
      "logps/ref_chosen": -47.59989929199219,
      "logps/ref_rejected": -89.41059875488281,
      "logps/rejected": -630.257080078125,
      "loss": 0.9234,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.40171539783477783,
      "rewards/margins": 0.6506726741790771,
      "rewards/rejected": -1.0523879528045654,
      "step": 498
    },
    {
      "epoch": 0.7327459618208517,
      "epsilon_dpo/beta": 0.0019314672099426389,
      "epsilon_dpo/beta_margin_grad_mean": -0.3319936990737915,
      "epsilon_dpo/beta_margin_grad_std": 0.13098543882369995,
      "epsilon_dpo/beta_margin_mean": 0.765389621257782,
      "epsilon_dpo/beta_margin_std": 0.637798011302948,
      "epsilon_dpo/loss_margin_mean": 397.3015441894531,
      "grad_norm": 35.18621826171875,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0019470942206680775,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.0243144327477013e-07,
      "logits/chosen": -2.809879779815674,
      "logits/rejected": -3.214648485183716,
      "logps/chosen": -222.212158203125,
      "logps/ref_chosen": -34.13922882080078,
      "logps/ref_rejected": -108.009521484375,
      "logps/rejected": -693.384033203125,
      "loss": 0.849,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.3641180992126465,
      "rewards/margins": 0.765389621257782,
      "rewards/rejected": -1.1295077800750732,
      "step": 499
    },
    {
      "epoch": 0.7342143906020558,
      "epsilon_dpo/beta": 0.0019146933918818831,
      "epsilon_dpo/beta_margin_grad_mean": -0.3297097384929657,
      "epsilon_dpo/beta_margin_grad_std": 0.12436074763536453,
      "epsilon_dpo/beta_margin_mean": 0.7685431241989136,
      "epsilon_dpo/beta_margin_std": 0.6095584630966187,
      "epsilon_dpo/loss_margin_mean": 402.1811828613281,
      "grad_norm": 37.49474334716797,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.0019314016681164503,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 1.0139748428955333e-07,
      "logits/chosen": -2.810401201248169,
      "logits/rejected": -3.235410451889038,
      "logps/chosen": -230.16627502441406,
      "logps/ref_chosen": -36.92897033691406,
      "logps/ref_rejected": -100.48208618164062,
      "logps/rejected": -695.9005737304688,
      "loss": 0.8394,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.37080904841423035,
      "rewards/margins": 0.7685431241989136,
      "rewards/rejected": -1.1393522024154663,
      "step": 500
    },
    {
      "epoch": 0.7342143906020558,
      "eval_epsilon_dpo/beta": 0.0019060522317886353,
      "eval_epsilon_dpo/beta_margin_grad_mean": -0.4040687382221222,
      "eval_epsilon_dpo/beta_margin_grad_std": 0.1391373723745346,
      "eval_epsilon_dpo/beta_margin_mean": 0.4280838668346405,
      "eval_epsilon_dpo/beta_margin_std": 0.6343588829040527,
      "eval_epsilon_dpo/loss_margin_mean": 226.7985382080078,
      "eval_kl/n_epsilon_steps": 0.27011987566947937,
      "eval_kl/p_epsilon_steps": 0.7290239930152893,
      "eval_logits/chosen": -2.9733166694641113,
      "eval_logits/rejected": -3.179455518722534,
      "eval_logps/chosen": -384.0945129394531,
      "eval_logps/ref_chosen": -68.29110717773438,
      "eval_logps/ref_rejected": -92.08038330078125,
      "eval_logps/rejected": -634.682373046875,
      "eval_loss": 0.5502132177352905,
      "eval_rewards/accuracies": 0.7452911138534546,
      "eval_rewards/chosen": -0.6045443415641785,
      "eval_rewards/margins": 0.4280838668346405,
      "eval_rewards/rejected": -1.0326281785964966,
      "eval_runtime": 38.4009,
      "eval_samples_per_second": 60.91,
      "eval_steps_per_second": 1.927,
      "step": 500
    },
    {
      "epoch": 0.73568281938326,
      "epsilon_dpo/beta": 0.0018998802406713367,
      "epsilon_dpo/beta_margin_grad_mean": -0.3355919420719147,
      "epsilon_dpo/beta_margin_grad_std": 0.12509720027446747,
      "epsilon_dpo/beta_margin_mean": 0.7397939562797546,
      "epsilon_dpo/beta_margin_std": 0.6008874773979187,
      "epsilon_dpo/loss_margin_mean": 390.53387451171875,
      "grad_norm": 35.63896179199219,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.001914648455567658,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 1.0036744111882672e-07,
      "logits/chosen": -2.834397077560425,
      "logits/rejected": -3.2369680404663086,
      "logps/chosen": -224.27859497070312,
      "logps/ref_chosen": -39.85574722290039,
      "logps/ref_rejected": -85.64679718017578,
      "logps/rejected": -660.603515625,
      "loss": 0.8571,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.35178065299987793,
      "rewards/margins": 0.7397940158843994,
      "rewards/rejected": -1.0915746688842773,
      "step": 501
    },
    {
      "epoch": 0.737151248164464,
      "epsilon_dpo/beta": 0.0018857462564483285,
      "epsilon_dpo/beta_margin_grad_mean": -0.37188294529914856,
      "epsilon_dpo/beta_margin_grad_std": 0.12015336006879807,
      "epsilon_dpo/beta_margin_mean": 0.5563080906867981,
      "epsilon_dpo/beta_margin_std": 0.5363189578056335,
      "epsilon_dpo/loss_margin_mean": 296.284912109375,
      "grad_norm": 49.43293762207031,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0018998062005266547,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 9.934134090518592e-08,
      "logits/chosen": -2.929020404815674,
      "logits/rejected": -3.1855547428131104,
      "logps/chosen": -272.3970031738281,
      "logps/ref_chosen": -53.44129943847656,
      "logps/ref_rejected": -86.86759185791016,
      "logps/rejected": -602.1082153320312,
      "loss": 0.9721,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.41469526290893555,
      "rewards/margins": 0.5563080906867981,
      "rewards/rejected": -0.9710033535957336,
      "step": 502
    },
    {
      "epoch": 0.7386196769456681,
      "epsilon_dpo/beta": 0.0018699404317885637,
      "epsilon_dpo/beta_margin_grad_mean": -0.34032920002937317,
      "epsilon_dpo/beta_margin_grad_std": 0.11132641136646271,
      "epsilon_dpo/beta_margin_mean": 0.7089167833328247,
      "epsilon_dpo/beta_margin_std": 0.5404109358787537,
      "epsilon_dpo/loss_margin_mean": 379.900390625,
      "grad_norm": 37.46735382080078,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.0018856637179851532,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 9.831921068732571e-08,
      "logits/chosen": -2.8521652221679688,
      "logits/rejected": -3.250486135482788,
      "logps/chosen": -186.5693817138672,
      "logps/ref_chosen": -34.42343521118164,
      "logps/ref_rejected": -88.69686889648438,
      "logps/rejected": -620.7432250976562,
      "loss": 0.862,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.28525102138519287,
      "rewards/margins": 0.7089167833328247,
      "rewards/rejected": -0.9941678047180176,
      "step": 503
    },
    {
      "epoch": 0.7400881057268722,
      "epsilon_dpo/beta": 0.0018554636044427752,
      "epsilon_dpo/beta_margin_grad_mean": -0.31814929842948914,
      "epsilon_dpo/beta_margin_grad_std": 0.12207160890102386,
      "epsilon_dpo/beta_margin_mean": 0.8290372490882874,
      "epsilon_dpo/beta_margin_std": 0.6291282176971436,
      "epsilon_dpo/loss_margin_mean": 448.0475769042969,
      "grad_norm": 43.66539001464844,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0018698865314945579,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 9.730107739932805e-08,
      "logits/chosen": -2.828281879425049,
      "logits/rejected": -3.302220106124878,
      "logps/chosen": -209.00950622558594,
      "logps/ref_chosen": -46.35227966308594,
      "logps/ref_rejected": -111.0133056640625,
      "logps/rejected": -721.7181396484375,
      "loss": 0.8029,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.3029179573059082,
      "rewards/margins": 0.8290372490882874,
      "rewards/rejected": -1.1319552659988403,
      "step": 504
    },
    {
      "epoch": 0.7415565345080763,
      "epsilon_dpo/beta": 0.0018439794657751918,
      "epsilon_dpo/beta_margin_grad_mean": -0.372370183467865,
      "epsilon_dpo/beta_margin_grad_std": 0.15154238045215607,
      "epsilon_dpo/beta_margin_mean": 0.5709735155105591,
      "epsilon_dpo/beta_margin_std": 0.7030720710754395,
      "epsilon_dpo/loss_margin_mean": 311.79742431640625,
      "grad_norm": 45.315521240234375,
      "kl/avg_steps": 0.625,
      "kl/beta": 0.0018553913105279207,
      "kl/n_epsilon_steps": 0.1875,
      "kl/p_epsilon_steps": 0.8125,
      "learning_rate": 9.628696786995188e-08,
      "logits/chosen": -2.8761417865753174,
      "logits/rejected": -3.2082276344299316,
      "logps/chosen": -259.9178771972656,
      "logps/ref_chosen": -47.89013671875,
      "logps/ref_rejected": -96.4730453491211,
      "logps/rejected": -620.2982177734375,
      "loss": 1.0064,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": -0.39312461018562317,
      "rewards/margins": 0.5709735155105591,
      "rewards/rejected": -0.9640980958938599,
      "step": 505
    },
    {
      "epoch": 0.7430249632892805,
      "epsilon_dpo/beta": 0.0018290685256943107,
      "epsilon_dpo/beta_margin_grad_mean": -0.3460908532142639,
      "epsilon_dpo/beta_margin_grad_std": 0.11466103047132492,
      "epsilon_dpo/beta_margin_mean": 0.6695333123207092,
      "epsilon_dpo/beta_margin_std": 0.5264294743537903,
      "epsilon_dpo/loss_margin_mean": 367.1669616699219,
      "grad_norm": 39.87546920776367,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0018438671249896288,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 9.527690882192635e-08,
      "logits/chosen": -2.8096556663513184,
      "logits/rejected": -3.270047187805176,
      "logps/chosen": -212.9695281982422,
      "logps/ref_chosen": -31.555944442749023,
      "logps/ref_rejected": -84.6425552368164,
      "logps/rejected": -633.22314453125,
      "loss": 0.8885,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.3329414129257202,
      "rewards/margins": 0.669533371925354,
      "rewards/rejected": -1.0024747848510742,
      "step": 506
    },
    {
      "epoch": 0.7444933920704846,
      "epsilon_dpo/beta": 0.001816042116843164,
      "epsilon_dpo/beta_margin_grad_mean": -0.366396963596344,
      "epsilon_dpo/beta_margin_grad_std": 0.11917287856340408,
      "epsilon_dpo/beta_margin_mean": 0.5886867642402649,
      "epsilon_dpo/beta_margin_std": 0.5481820702552795,
      "epsilon_dpo/loss_margin_mean": 325.3662414550781,
      "grad_norm": 34.283470153808594,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.001829006476327777,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 9.427092687124691e-08,
      "logits/chosen": -2.9394993782043457,
      "logits/rejected": -3.2683815956115723,
      "logps/chosen": -282.3764953613281,
      "logps/ref_chosen": -50.91284942626953,
      "logps/ref_rejected": -101.67366790771484,
      "logps/rejected": -658.5035400390625,
      "loss": 0.9498,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.42130210995674133,
      "rewards/margins": 0.5886867642402649,
      "rewards/rejected": -1.009988784790039,
      "step": 507
    },
    {
      "epoch": 0.7459618208516887,
      "epsilon_dpo/beta": 0.0018047851044684649,
      "epsilon_dpo/beta_margin_grad_mean": -0.3578824996948242,
      "epsilon_dpo/beta_margin_grad_std": 0.12713228166103363,
      "epsilon_dpo/beta_margin_mean": 0.6354185342788696,
      "epsilon_dpo/beta_margin_std": 0.6018913984298706,
      "epsilon_dpo/loss_margin_mean": 353.8747863769531,
      "grad_norm": 40.939327239990234,
      "kl/avg_steps": 0.625,
      "kl/beta": 0.0018159543396905065,
      "kl/n_epsilon_steps": 0.1875,
      "kl/p_epsilon_steps": 0.8125,
      "learning_rate": 9.326904852647344e-08,
      "logits/chosen": -2.883289337158203,
      "logits/rejected": -3.2906653881073,
      "logps/chosen": -260.6716613769531,
      "logps/ref_chosen": -50.73012924194336,
      "logps/ref_rejected": -101.8434829711914,
      "logps/rejected": -665.6597900390625,
      "loss": 0.9286,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.38065105676651,
      "rewards/margins": 0.6354185342788696,
      "rewards/rejected": -1.0160696506500244,
      "step": 508
    },
    {
      "epoch": 0.7474302496328928,
      "epsilon_dpo/beta": 0.0017901910468935966,
      "epsilon_dpo/beta_margin_grad_mean": -0.35225531458854675,
      "epsilon_dpo/beta_margin_grad_std": 0.10537883639335632,
      "epsilon_dpo/beta_margin_mean": 0.6424784660339355,
      "epsilon_dpo/beta_margin_std": 0.4860381782054901,
      "epsilon_dpo/loss_margin_mean": 359.7619323730469,
      "grad_norm": 38.303802490234375,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0018046750919893384,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 9.227130018803195e-08,
      "logits/chosen": -2.86599063873291,
      "logits/rejected": -3.2632508277893066,
      "logps/chosen": -214.93182373046875,
      "logps/ref_chosen": -47.68657302856445,
      "logps/ref_rejected": -88.62358093261719,
      "logps/rejected": -615.6307373046875,
      "loss": 0.8974,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.30018889904022217,
      "rewards/margins": 0.6424784660339355,
      "rewards/rejected": -0.9426673650741577,
      "step": 509
    },
    {
      "epoch": 0.748898678414097,
      "epsilon_dpo/beta": 0.001774644129909575,
      "epsilon_dpo/beta_margin_grad_mean": -0.33876028656959534,
      "epsilon_dpo/beta_margin_grad_std": 0.0986710786819458,
      "epsilon_dpo/beta_margin_mean": 0.7086153626441956,
      "epsilon_dpo/beta_margin_std": 0.4846038222312927,
      "epsilon_dpo/loss_margin_mean": 399.8760070800781,
      "grad_norm": 32.25279235839844,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.001790130278095603,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 9.127770814751932e-08,
      "logits/chosen": -2.808089256286621,
      "logits/rejected": -3.3828306198120117,
      "logps/chosen": -220.7513427734375,
      "logps/ref_chosen": -32.308326721191406,
      "logps/ref_rejected": -102.69437408447266,
      "logps/rejected": -691.013427734375,
      "loss": 0.8498,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.33473625779151917,
      "rewards/margins": 0.7086154222488403,
      "rewards/rejected": -1.043351650238037,
      "step": 510
    },
    {
      "epoch": 0.750367107195301,
      "epsilon_dpo/beta": 0.0017620236612856388,
      "epsilon_dpo/beta_margin_grad_mean": -0.34761884808540344,
      "epsilon_dpo/beta_margin_grad_std": 0.10665444284677505,
      "epsilon_dpo/beta_margin_mean": 0.6692096590995789,
      "epsilon_dpo/beta_margin_std": 0.5043637752532959,
      "epsilon_dpo/loss_margin_mean": 380.9803466796875,
      "grad_norm": 34.81239318847656,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.001774602453224361,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 9.028829858700973e-08,
      "logits/chosen": -2.88916015625,
      "logits/rejected": -3.333850860595703,
      "logps/chosen": -226.4573974609375,
      "logps/ref_chosen": -41.02062225341797,
      "logps/ref_rejected": -98.91937255859375,
      "logps/rejected": -665.3365478515625,
      "loss": 0.8819,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.3278487026691437,
      "rewards/margins": 0.6692096590995789,
      "rewards/rejected": -0.9970583915710449,
      "step": 511
    },
    {
      "epoch": 0.7518355359765051,
      "epsilon_dpo/beta": 0.001748348237015307,
      "epsilon_dpo/beta_margin_grad_mean": -0.3473552465438843,
      "epsilon_dpo/beta_margin_grad_std": 0.12345966696739197,
      "epsilon_dpo/beta_margin_mean": 0.6806043386459351,
      "epsilon_dpo/beta_margin_std": 0.5956137180328369,
      "epsilon_dpo/loss_margin_mean": 390.4645690917969,
      "grad_norm": 34.644046783447266,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0017619385616853833,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 8.930309757836516e-08,
      "logits/chosen": -2.82757306098938,
      "logits/rejected": -3.3794026374816895,
      "logps/chosen": -213.54293823242188,
      "logps/ref_chosen": -37.26538848876953,
      "logps/ref_rejected": -88.17929077148438,
      "logps/rejected": -654.92138671875,
      "loss": 0.8953,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.30911028385162354,
      "rewards/margins": 0.6806043386459351,
      "rewards/rejected": -0.9897146224975586,
      "step": 512
    },
    {
      "epoch": 0.7533039647577092,
      "epsilon_dpo/beta": 0.0017353416187688708,
      "epsilon_dpo/beta_margin_grad_mean": -0.3601101040840149,
      "epsilon_dpo/beta_margin_grad_std": 0.1102689802646637,
      "epsilon_dpo/beta_margin_mean": 0.6162811517715454,
      "epsilon_dpo/beta_margin_std": 0.5283299684524536,
      "epsilon_dpo/loss_margin_mean": 356.2281494140625,
      "grad_norm": 37.334041595458984,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0017482801340520382,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 8.832213108254863e-08,
      "logits/chosen": -2.8295106887817383,
      "logits/rejected": -3.2998032569885254,
      "logps/chosen": -189.08787536621094,
      "logps/ref_chosen": -35.62997055053711,
      "logps/ref_rejected": -81.91015625,
      "logps/rejected": -591.59619140625,
      "loss": 0.9234,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.2673969268798828,
      "rewards/margins": 0.6162811517715454,
      "rewards/rejected": -0.8836780786514282,
      "step": 513
    },
    {
      "epoch": 0.7547723935389133,
      "epsilon_dpo/beta": 0.0017213387181982398,
      "epsilon_dpo/beta_margin_grad_mean": -0.35787054896354675,
      "epsilon_dpo/beta_margin_grad_std": 0.11788881570100784,
      "epsilon_dpo/beta_margin_mean": 0.6275299191474915,
      "epsilon_dpo/beta_margin_std": 0.5512734055519104,
      "epsilon_dpo/loss_margin_mean": 365.5804748535156,
      "grad_norm": 29.60846519470215,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0017352655995637178,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 8.734542494893954e-08,
      "logits/chosen": -2.9027023315429688,
      "logits/rejected": -3.4183146953582764,
      "logps/chosen": -241.52822875976562,
      "logps/ref_chosen": -43.213531494140625,
      "logps/ref_rejected": -84.48947143554688,
      "logps/rejected": -648.3846435546875,
      "loss": 0.9222,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.3422423005104065,
      "rewards/margins": 0.6275299787521362,
      "rewards/rejected": -0.969772219657898,
      "step": 514
    },
    {
      "epoch": 0.7562408223201175,
      "epsilon_dpo/beta": 0.0017085415311157703,
      "epsilon_dpo/beta_margin_grad_mean": -0.3623103201389313,
      "epsilon_dpo/beta_margin_grad_std": 0.13232813775539398,
      "epsilon_dpo/beta_margin_mean": 0.6002452969551086,
      "epsilon_dpo/beta_margin_std": 0.5947220921516418,
      "epsilon_dpo/loss_margin_mean": 352.9983825683594,
      "grad_norm": 36.903953552246094,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0017212802777066827,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 8.637300491465272e-08,
      "logits/chosen": -2.862236738204956,
      "logits/rejected": -3.4654173851013184,
      "logps/chosen": -235.07046508789062,
      "logps/ref_chosen": -33.94335174560547,
      "logps/ref_rejected": -92.82087707519531,
      "logps/rejected": -646.9464111328125,
      "loss": 0.9554,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.34564244747161865,
      "rewards/margins": 0.6002452969551086,
      "rewards/rejected": -0.9458876848220825,
      "step": 515
    },
    {
      "epoch": 0.7577092511013216,
      "epsilon_dpo/beta": 0.0016963566886261106,
      "epsilon_dpo/beta_margin_grad_mean": -0.3456820547580719,
      "epsilon_dpo/beta_margin_grad_std": 0.12934663891792297,
      "epsilon_dpo/beta_margin_mean": 0.6971822381019592,
      "epsilon_dpo/beta_margin_std": 0.6565114259719849,
      "epsilon_dpo/loss_margin_mean": 412.7080993652344,
      "grad_norm": 35.10322952270508,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0017084666760638356,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 8.540489660386064e-08,
      "logits/chosen": -2.9307332038879395,
      "logits/rejected": -3.505221366882324,
      "logps/chosen": -244.57504272460938,
      "logps/ref_chosen": -48.20185852050781,
      "logps/ref_rejected": -119.76963806152344,
      "logps/rejected": -728.8509521484375,
      "loss": 0.8964,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.33476826548576355,
      "rewards/margins": 0.6971822381019592,
      "rewards/rejected": -1.0319504737854004,
      "step": 516
    },
    {
      "epoch": 0.7591776798825257,
      "epsilon_dpo/beta": 0.0016842512413859367,
      "epsilon_dpo/beta_margin_grad_mean": -0.3475935757160187,
      "epsilon_dpo/beta_margin_grad_std": 0.12928926944732666,
      "epsilon_dpo/beta_margin_mean": 0.6708550453186035,
      "epsilon_dpo/beta_margin_std": 0.5967744588851929,
      "epsilon_dpo/loss_margin_mean": 400.1361083984375,
      "grad_norm": 38.30036163330078,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0016962747322395444,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 8.444112552711752e-08,
      "logits/chosen": -2.9165990352630615,
      "logits/rejected": -3.483713150024414,
      "logps/chosen": -237.81472778320312,
      "logps/ref_chosen": -40.42920684814453,
      "logps/ref_rejected": -98.80826568603516,
      "logps/rejected": -696.3298950195312,
      "loss": 0.9048,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.33436378836631775,
      "rewards/margins": 0.6708550453186035,
      "rewards/rejected": -1.0052188634872437,
      "step": 517
    },
    {
      "epoch": 0.7606461086637298,
      "epsilon_dpo/beta": 0.0016727583715692163,
      "epsilon_dpo/beta_margin_grad_mean": -0.35805386304855347,
      "epsilon_dpo/beta_margin_grad_std": 0.1211152896285057,
      "epsilon_dpo/beta_margin_mean": 0.6288034915924072,
      "epsilon_dpo/beta_margin_std": 0.5658082962036133,
      "epsilon_dpo/loss_margin_mean": 377.4985046386719,
      "grad_norm": 43.182456970214844,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0016841697506606579,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 8.348171708068747e-08,
      "logits/chosen": -2.878077983856201,
      "logits/rejected": -3.482807159423828,
      "logps/chosen": -228.24789428710938,
      "logps/ref_chosen": -37.2679557800293,
      "logps/ref_rejected": -90.86311340332031,
      "logps/rejected": -659.341552734375,
      "loss": 0.9249,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.3209276795387268,
      "rewards/margins": 0.6288034915924072,
      "rewards/rejected": -0.9497312307357788,
      "step": 518
    },
    {
      "epoch": 0.762114537444934,
      "epsilon_dpo/beta": 0.0016592455795034766,
      "epsilon_dpo/beta_margin_grad_mean": -0.35770776867866516,
      "epsilon_dpo/beta_margin_grad_std": 0.13007250428199768,
      "epsilon_dpo/beta_margin_mean": 0.644246518611908,
      "epsilon_dpo/beta_margin_std": 0.639655590057373,
      "epsilon_dpo/loss_margin_mean": 389.47509765625,
      "grad_norm": 37.42156982421875,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0016726701287552714,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 8.25266965458755e-08,
      "logits/chosen": -2.999945878982544,
      "logits/rejected": -3.565791606903076,
      "logps/chosen": -240.69334411621094,
      "logps/ref_chosen": -47.603675842285156,
      "logps/ref_rejected": -110.82022094726562,
      "logps/rejected": -693.385009765625,
      "loss": 0.9302,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.32136669754981995,
      "rewards/margins": 0.6442465782165527,
      "rewards/rejected": -0.9656132459640503,
      "step": 519
    },
    {
      "epoch": 0.7635829662261381,
      "epsilon_dpo/beta": 0.0016458729514852166,
      "epsilon_dpo/beta_margin_grad_mean": -0.3453068435192108,
      "epsilon_dpo/beta_margin_grad_std": 0.10654862225055695,
      "epsilon_dpo/beta_margin_mean": 0.685132622718811,
      "epsilon_dpo/beta_margin_std": 0.5305820107460022,
      "epsilon_dpo/loss_margin_mean": 417.23046875,
      "grad_norm": 30.94179916381836,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0016591892344877124,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 8.15760890883607e-08,
      "logits/chosen": -2.9207892417907715,
      "logits/rejected": -3.5509166717529297,
      "logps/chosen": -229.61695861816406,
      "logps/ref_chosen": -43.462921142578125,
      "logps/ref_rejected": -109.22837829589844,
      "logps/rejected": -712.6129150390625,
      "loss": 0.8746,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.3073316216468811,
      "rewards/margins": 0.685132622718811,
      "rewards/rejected": -0.9924643039703369,
      "step": 520
    },
    {
      "epoch": 0.7650513950073421,
      "epsilon_dpo/beta": 0.001633122330531478,
      "epsilon_dpo/beta_margin_grad_mean": -0.3637232482433319,
      "epsilon_dpo/beta_margin_grad_std": 0.10722323507070541,
      "epsilon_dpo/beta_margin_mean": 0.5909576416015625,
      "epsilon_dpo/beta_margin_std": 0.4859505593776703,
      "epsilon_dpo/loss_margin_mean": 362.88031005859375,
      "grad_norm": 37.41933059692383,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0016458169557154179,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 8.062991975753378e-08,
      "logits/chosen": -2.949124336242676,
      "logits/rejected": -3.4791650772094727,
      "logps/chosen": -218.08642578125,
      "logps/ref_chosen": -39.79935836791992,
      "logps/ref_rejected": -89.2203140258789,
      "logps/rejected": -630.3876953125,
      "loss": 0.9344,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.2921355366706848,
      "rewards/margins": 0.5909576416015625,
      "rewards/rejected": -0.8830931782722473,
      "step": 521
    },
    {
      "epoch": 0.7665198237885462,
      "epsilon_dpo/beta": 0.0016209728782996535,
      "epsilon_dpo/beta_margin_grad_mean": -0.38271984457969666,
      "epsilon_dpo/beta_margin_grad_std": 0.12090186774730682,
      "epsilon_dpo/beta_margin_mean": 0.511188805103302,
      "epsilon_dpo/beta_margin_std": 0.5572635531425476,
      "epsilon_dpo/loss_margin_mean": 316.7029724121094,
      "grad_norm": 34.424713134765625,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0016330587677657604,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 7.968821348583643e-08,
      "logits/chosen": -2.922499895095825,
      "logits/rejected": -3.3887128829956055,
      "logps/chosen": -223.6859893798828,
      "logps/ref_chosen": -32.023380279541016,
      "logps/ref_rejected": -72.67029571533203,
      "logps/rejected": -581.035888671875,
      "loss": 1.0096,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.3123043179512024,
      "rewards/margins": 0.511188805103302,
      "rewards/rejected": -0.8234931230545044,
      "step": 522
    },
    {
      "epoch": 0.7679882525697503,
      "epsilon_dpo/beta": 0.0016109324060380459,
      "epsilon_dpo/beta_margin_grad_mean": -0.36688175797462463,
      "epsilon_dpo/beta_margin_grad_std": 0.12298644334077835,
      "epsilon_dpo/beta_margin_mean": 0.5938741564750671,
      "epsilon_dpo/beta_margin_std": 0.5870850682258606,
      "epsilon_dpo/loss_margin_mean": 370.4969482421875,
      "grad_norm": 34.04196548461914,
      "kl/avg_steps": 0.625,
      "kl/beta": 0.001620901981368661,
      "kl/n_epsilon_steps": 0.1875,
      "kl/p_epsilon_steps": 0.8125,
      "learning_rate": 7.875099508810484e-08,
      "logits/chosen": -2.9228086471557617,
      "logits/rejected": -3.478931427001953,
      "logps/chosen": -218.07928466796875,
      "logps/ref_chosen": -44.56959533691406,
      "logps/ref_rejected": -93.40034484863281,
      "logps/rejected": -637.406982421875,
      "loss": 0.9538,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.2807372212409973,
      "rewards/margins": 0.5938742160797119,
      "rewards/rejected": -0.8746113777160645,
      "step": 523
    },
    {
      "epoch": 0.7694566813509545,
      "epsilon_dpo/beta": 0.001596899121068418,
      "epsilon_dpo/beta_margin_grad_mean": -0.3489263653755188,
      "epsilon_dpo/beta_margin_grad_std": 0.09371962398290634,
      "epsilon_dpo/beta_margin_mean": 0.6547738313674927,
      "epsilon_dpo/beta_margin_std": 0.4421177804470062,
      "epsilon_dpo/loss_margin_mean": 410.6026916503906,
      "grad_norm": 34.071510314941406,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.0016108342679217458,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 7.781828926091535e-08,
      "logits/chosen": -3.0543670654296875,
      "logits/rejected": -3.656444549560547,
      "logps/chosen": -233.17068481445312,
      "logps/ref_chosen": -47.35930252075195,
      "logps/ref_rejected": -87.55693054199219,
      "logps/rejected": -683.9710083007812,
      "loss": 0.8791,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.296951562166214,
      "rewards/margins": 0.6547738313674927,
      "rewards/rejected": -0.9517253637313843,
      "step": 524
    },
    {
      "epoch": 0.7709251101321586,
      "epsilon_dpo/beta": 0.001583546632900834,
      "epsilon_dpo/beta_margin_grad_mean": -0.33597642183303833,
      "epsilon_dpo/beta_margin_grad_std": 0.11459914594888687,
      "epsilon_dpo/beta_margin_mean": 0.7328599691390991,
      "epsilon_dpo/beta_margin_std": 0.5668560862541199,
      "epsilon_dpo/loss_margin_mean": 463.7811584472656,
      "grad_norm": 33.28862380981445,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.0015968617517501116,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 7.689012058193384e-08,
      "logits/chosen": -2.9925098419189453,
      "logits/rejected": -3.6460928916931152,
      "logps/chosen": -205.41815185546875,
      "logps/ref_chosen": -29.801021575927734,
      "logps/ref_rejected": -107.73757934570312,
      "logps/rejected": -747.1358642578125,
      "loss": 0.8517,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.27915799617767334,
      "rewards/margins": 0.7328599691390991,
      "rewards/rejected": -1.0120179653167725,
      "step": 525
    },
    {
      "epoch": 0.7723935389133627,
      "epsilon_dpo/beta": 0.0015698021743446589,
      "epsilon_dpo/beta_margin_grad_mean": -0.3435947597026825,
      "epsilon_dpo/beta_margin_grad_std": 0.0916726216673851,
      "epsilon_dpo/beta_margin_mean": 0.6770256757736206,
      "epsilon_dpo/beta_margin_std": 0.4311753809452057,
      "epsilon_dpo/loss_margin_mean": 431.86358642578125,
      "grad_norm": 33.44818115234375,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.001583500881679356,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 7.596651350926836e-08,
      "logits/chosen": -2.96604061126709,
      "logits/rejected": -3.718456983566284,
      "logps/chosen": -219.9847412109375,
      "logps/ref_chosen": -37.456085205078125,
      "logps/ref_rejected": -93.045654296875,
      "logps/rejected": -707.4378662109375,
      "loss": 0.8619,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.2869272232055664,
      "rewards/margins": 0.6770256757736206,
      "rewards/rejected": -0.963952898979187,
      "step": 526
    },
    {
      "epoch": 0.7738619676945668,
      "epsilon_dpo/beta": 0.0015561856562271714,
      "epsilon_dpo/beta_margin_grad_mean": -0.35842540860176086,
      "epsilon_dpo/beta_margin_grad_std": 0.09837586432695389,
      "epsilon_dpo/beta_margin_mean": 0.6056324243545532,
      "epsilon_dpo/beta_margin_std": 0.44141340255737305,
      "epsilon_dpo/loss_margin_mean": 389.9037170410156,
      "grad_norm": 33.482730865478516,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.0015697655035182834,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 7.504749238082414e-08,
      "logits/chosen": -3.048184871673584,
      "logits/rejected": -3.7029123306274414,
      "logps/chosen": -219.21990966796875,
      "logps/ref_chosen": -43.55718231201172,
      "logps/ref_rejected": -83.77217102050781,
      "logps/rejected": -649.338623046875,
      "loss": 0.9152,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.27390962839126587,
      "rewards/margins": 0.6056324243545532,
      "rewards/rejected": -0.8795421123504639,
      "step": 527
    },
    {
      "epoch": 0.775330396475771,
      "epsilon_dpo/beta": 0.0015451188664883375,
      "epsilon_dpo/beta_margin_grad_mean": -0.35285684466362,
      "epsilon_dpo/beta_margin_grad_std": 0.11853712797164917,
      "epsilon_dpo/beta_margin_mean": 0.6550965309143066,
      "epsilon_dpo/beta_margin_std": 0.5716320276260376,
      "epsilon_dpo/loss_margin_mean": 425.4979248046875,
      "grad_norm": 36.37355041503906,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0015561492182314396,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 7.413308141366254e-08,
      "logits/chosen": -3.052718162536621,
      "logits/rejected": -3.7526984214782715,
      "logps/chosen": -250.51231384277344,
      "logps/ref_chosen": -46.319541931152344,
      "logps/ref_rejected": -101.05429077148438,
      "logps/rejected": -730.7449951171875,
      "loss": 0.9063,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.3164703845977783,
      "rewards/margins": 0.6550965309143066,
      "rewards/rejected": -0.971566915512085,
      "step": 528
    },
    {
      "epoch": 0.7767988252569751,
      "epsilon_dpo/beta": 0.0015350582543760538,
      "epsilon_dpo/beta_margin_grad_mean": -0.3771320581436157,
      "epsilon_dpo/beta_margin_grad_std": 0.11737023293972015,
      "epsilon_dpo/beta_margin_mean": 0.531406581401825,
      "epsilon_dpo/beta_margin_std": 0.5209585428237915,
      "epsilon_dpo/loss_margin_mean": 347.9163513183594,
      "grad_norm": 36.469417572021484,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.0015450441278517246,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 7.322330470336313e-08,
      "logits/chosen": -3.203193187713623,
      "logits/rejected": -3.8966808319091797,
      "logps/chosen": -311.4798889160156,
      "logps/ref_chosen": -39.26038360595703,
      "logps/ref_rejected": -94.54532623291016,
      "logps/rejected": -714.68115234375,
      "loss": 0.987,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.41968581080436707,
      "rewards/margins": 0.531406581401825,
      "rewards/rejected": -0.9510923624038696,
      "step": 529
    },
    {
      "epoch": 0.7782672540381792,
      "epsilon_dpo/beta": 0.0015236110193654895,
      "epsilon_dpo/beta_margin_grad_mean": -0.3523128926753998,
      "epsilon_dpo/beta_margin_grad_std": 0.11749440431594849,
      "epsilon_dpo/beta_margin_mean": 0.665305495262146,
      "epsilon_dpo/beta_margin_std": 0.6024741530418396,
      "epsilon_dpo/loss_margin_mean": 438.0719299316406,
      "grad_norm": 38.925601959228516,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0015349709428846836,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 7.231818622338822e-08,
      "logits/chosen": -3.1173648834228516,
      "logits/rejected": -3.8287353515625,
      "logps/chosen": -200.4312744140625,
      "logps/ref_chosen": -27.232646942138672,
      "logps/ref_rejected": -92.85646057128906,
      "logps/rejected": -704.1270751953125,
      "loss": 0.9029,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.264974981546402,
      "rewards/margins": 0.665305495262146,
      "rewards/rejected": -0.9302804470062256,
      "step": 530
    },
    {
      "epoch": 0.7797356828193832,
      "epsilon_dpo/beta": 0.0015113166300579906,
      "epsilon_dpo/beta_margin_grad_mean": -0.36579129099845886,
      "epsilon_dpo/beta_margin_grad_std": 0.10368164628744125,
      "epsilon_dpo/beta_margin_mean": 0.5806282758712769,
      "epsilon_dpo/beta_margin_std": 0.4778251051902771,
      "epsilon_dpo/loss_margin_mean": 385.13055419921875,
      "grad_norm": 39.48409652709961,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0015235443133860826,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 7.141774982445147e-08,
      "logits/chosen": -3.1506361961364746,
      "logits/rejected": -3.9508585929870605,
      "logps/chosen": -226.21939086914062,
      "logps/ref_chosen": -29.433807373046875,
      "logps/ref_rejected": -77.6962890625,
      "logps/rejected": -659.6124267578125,
      "loss": 0.9395,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.2982354760169983,
      "rewards/margins": 0.5806282758712769,
      "rewards/rejected": -0.8788637518882751,
      "step": 531
    },
    {
      "epoch": 0.7812041116005873,
      "epsilon_dpo/beta": 0.0015024424064904451,
      "epsilon_dpo/beta_margin_grad_mean": -0.36811915040016174,
      "epsilon_dpo/beta_margin_grad_std": 0.13502071797847748,
      "epsilon_dpo/beta_margin_mean": 0.5831210017204285,
      "epsilon_dpo/beta_margin_std": 0.6058715581893921,
      "epsilon_dpo/loss_margin_mean": 390.57232666015625,
      "grad_norm": 41.38243103027344,
      "kl/avg_steps": 0.59375,
      "kl/beta": 0.0015112652909010649,
      "kl/n_epsilon_steps": 0.203125,
      "kl/p_epsilon_steps": 0.796875,
      "learning_rate": 7.052201923388953e-08,
      "logits/chosen": -3.245467185974121,
      "logits/rejected": -3.924285888671875,
      "logps/chosen": -292.22418212890625,
      "logps/ref_chosen": -49.116477966308594,
      "logps/ref_rejected": -93.35752868652344,
      "logps/rejected": -727.0375366210938,
      "loss": 0.9702,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": -0.3671477138996124,
      "rewards/margins": 0.5831209421157837,
      "rewards/rejected": -0.9502686858177185,
      "step": 532
    },
    {
      "epoch": 0.7826725403817915,
      "epsilon_dpo/beta": 0.0014921656111255288,
      "epsilon_dpo/beta_margin_grad_mean": -0.3761315643787384,
      "epsilon_dpo/beta_margin_grad_std": 0.12172158807516098,
      "epsilon_dpo/beta_margin_mean": 0.5404428243637085,
      "epsilon_dpo/beta_margin_std": 0.5591951012611389,
      "epsilon_dpo/loss_margin_mean": 364.0026550292969,
      "grad_norm": 38.90617752075195,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0015023450832813978,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 6.963101805503646e-08,
      "logits/chosen": -3.1959118843078613,
      "logits/rejected": -3.9355404376983643,
      "logps/chosen": -277.5054016113281,
      "logps/ref_chosen": -39.03050994873047,
      "logps/ref_rejected": -81.61624145507812,
      "logps/rejected": -684.0938110351562,
      "loss": 0.9883,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.35740402340888977,
      "rewards/margins": 0.5404428243637085,
      "rewards/rejected": -0.8978468179702759,
      "step": 533
    },
    {
      "epoch": 0.7841409691629956,
      "epsilon_dpo/beta": 0.0014812753070145845,
      "epsilon_dpo/beta_margin_grad_mean": -0.35220903158187866,
      "epsilon_dpo/beta_margin_grad_std": 0.10367898643016815,
      "epsilon_dpo/beta_margin_mean": 0.6407039165496826,
      "epsilon_dpo/beta_margin_std": 0.46845534443855286,
      "epsilon_dpo/loss_margin_mean": 433.8296813964844,
      "grad_norm": 43.0300178527832,
      "kl/avg_steps": 0.734375,
      "kl/beta": 0.0014920870307832956,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 6.874476976660184e-08,
      "logits/chosen": -3.1139631271362305,
      "logits/rejected": -3.9010872840881348,
      "logps/chosen": -213.8875732421875,
      "logps/ref_chosen": -33.198890686035156,
      "logps/ref_rejected": -85.75593566894531,
      "logps/rejected": -700.2742919921875,
      "loss": 0.8956,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.26844221353530884,
      "rewards/margins": 0.6407039165496826,
      "rewards/rejected": -0.9091461896896362,
      "step": 534
    },
    {
      "epoch": 0.7856093979441997,
      "epsilon_dpo/beta": 0.0014679327141493559,
      "epsilon_dpo/beta_margin_grad_mean": -0.34026262164115906,
      "epsilon_dpo/beta_margin_grad_std": 0.11611247062683105,
      "epsilon_dpo/beta_margin_mean": 0.7134142518043518,
      "epsilon_dpo/beta_margin_std": 0.5855581164360046,
      "epsilon_dpo/loss_margin_mean": 486.8583068847656,
      "grad_norm": 40.018455505371094,
      "kl/avg_steps": 0.90625,
      "kl/beta": 0.0014812094159424305,
      "kl/n_epsilon_steps": 0.046875,
      "kl/p_epsilon_steps": 0.953125,
      "learning_rate": 6.786329772205246e-08,
      "logits/chosen": -3.2459397315979004,
      "logits/rejected": -4.044425010681152,
      "logps/chosen": -248.54649353027344,
      "logps/ref_chosen": -39.410072326660156,
      "logps/ref_rejected": -95.90589904785156,
      "logps/rejected": -791.900634765625,
      "loss": 0.8681,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.3082220256328583,
      "rewards/margins": 0.7134143114089966,
      "rewards/rejected": -1.0216362476348877,
      "step": 535
    },
    {
      "epoch": 0.7870778267254038,
      "epsilon_dpo/beta": 0.001455666613765061,
      "epsilon_dpo/beta_margin_grad_mean": -0.34118029475212097,
      "epsilon_dpo/beta_margin_grad_std": 0.1087789535522461,
      "epsilon_dpo/beta_margin_mean": 0.7139739990234375,
      "epsilon_dpo/beta_margin_std": 0.577499270439148,
      "epsilon_dpo/loss_margin_mean": 491.39007568359375,
      "grad_norm": 43.79270553588867,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.0014679065207019448,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 6.698662514899638e-08,
      "logits/chosen": -3.22434139251709,
      "logits/rejected": -4.060914039611816,
      "logps/chosen": -262.8885498046875,
      "logps/ref_chosen": -31.54595184326172,
      "logps/ref_rejected": -95.49956512451172,
      "logps/rejected": -818.2322387695312,
      "loss": 0.8622,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.33760565519332886,
      "rewards/margins": 0.7139739990234375,
      "rewards/rejected": -1.0515797138214111,
      "step": 536
    },
    {
      "epoch": 0.788546255506608,
      "epsilon_dpo/beta": 0.0014448519796133041,
      "epsilon_dpo/beta_margin_grad_mean": -0.3592894971370697,
      "epsilon_dpo/beta_margin_grad_std": 0.10450321435928345,
      "epsilon_dpo/beta_margin_mean": 0.6133064031600952,
      "epsilon_dpo/beta_margin_std": 0.48698103427886963,
      "epsilon_dpo/loss_margin_mean": 425.6983947753906,
      "grad_norm": 37.07543182373047,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0014556247042492032,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 6.611477514857114e-08,
      "logits/chosen": -3.1620330810546875,
      "logits/rejected": -3.9779064655303955,
      "logps/chosen": -205.30322265625,
      "logps/ref_chosen": -29.482723236083984,
      "logps/ref_rejected": -78.68902587890625,
      "logps/rejected": -680.2078857421875,
      "loss": 0.9176,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.2549804449081421,
      "rewards/margins": 0.6133064031600952,
      "rewards/rejected": -0.8682868480682373,
      "step": 537
    },
    {
      "epoch": 0.7900146842878121,
      "epsilon_dpo/beta": 0.0014349992852658033,
      "epsilon_dpo/beta_margin_grad_mean": -0.34261569380760193,
      "epsilon_dpo/beta_margin_grad_std": 0.132565438747406,
      "epsilon_dpo/beta_margin_mean": 0.7123464941978455,
      "epsilon_dpo/beta_margin_std": 0.6496385931968689,
      "epsilon_dpo/loss_margin_mean": 498.54559326171875,
      "grad_norm": 43.92093276977539,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0014447887660935521,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 6.524777069483525e-08,
      "logits/chosen": -3.295565128326416,
      "logits/rejected": -4.157257080078125,
      "logps/chosen": -342.5325622558594,
      "logps/ref_chosen": -46.485496520996094,
      "logps/ref_rejected": -91.08615112304688,
      "logps/rejected": -885.6788330078125,
      "loss": 0.8867,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.4264609217643738,
      "rewards/margins": 0.7123464345932007,
      "rewards/rejected": -1.1388074159622192,
      "step": 538
    },
    {
      "epoch": 0.7914831130690162,
      "epsilon_dpo/beta": 0.001425649388693273,
      "epsilon_dpo/beta_margin_grad_mean": -0.34682291746139526,
      "epsilon_dpo/beta_margin_grad_std": 0.12398696690797806,
      "epsilon_dpo/beta_margin_mean": 0.6886243224143982,
      "epsilon_dpo/beta_margin_std": 0.60282963514328,
      "epsilon_dpo/loss_margin_mean": 485.098388671875,
      "grad_norm": 45.51395034790039,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.0014349236153066158,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 6.438563463416221e-08,
      "logits/chosen": -3.2940754890441895,
      "logits/rejected": -4.077099800109863,
      "logps/chosen": -323.3985595703125,
      "logps/ref_chosen": -39.210296630859375,
      "logps/ref_rejected": -95.66792297363281,
      "logps/rejected": -864.95458984375,
      "loss": 0.8907,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.4068090319633484,
      "rewards/margins": 0.688624382019043,
      "rewards/rejected": -1.0954333543777466,
      "step": 539
    },
    {
      "epoch": 0.7929515418502202,
      "epsilon_dpo/beta": 0.0014163546729832888,
      "epsilon_dpo/beta_margin_grad_mean": -0.32467931509017944,
      "epsilon_dpo/beta_margin_grad_std": 0.1530696600675583,
      "epsilon_dpo/beta_margin_mean": 0.8351730108261108,
      "epsilon_dpo/beta_margin_std": 0.7946527004241943,
      "epsilon_dpo/loss_margin_mean": 592.330078125,
      "grad_norm": 49.0323486328125,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.0014255683636292815,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 6.352838968463919e-08,
      "logits/chosen": -3.369414806365967,
      "logits/rejected": -4.240825176239014,
      "logps/chosen": -385.478759765625,
      "logps/ref_chosen": -50.91047668457031,
      "logps/ref_rejected": -125.03810119628906,
      "logps/rejected": -1051.9365234375,
      "loss": 0.8461,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.47576338052749634,
      "rewards/margins": 0.8351730108261108,
      "rewards/rejected": -1.310936450958252,
      "step": 540
    },
    {
      "epoch": 0.7944199706314243,
      "epsilon_dpo/beta": 0.0014062352711334825,
      "epsilon_dpo/beta_margin_grad_mean": -0.3354036808013916,
      "epsilon_dpo/beta_margin_grad_std": 0.13668367266654968,
      "epsilon_dpo/beta_margin_mean": 0.7499971985816956,
      "epsilon_dpo/beta_margin_std": 0.6641453504562378,
      "epsilon_dpo/loss_margin_mean": 535.408447265625,
      "grad_norm": 54.5290412902832,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0014162741135805845,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 6.267605843546767e-08,
      "logits/chosen": -3.3453049659729004,
      "logits/rejected": -4.137938499450684,
      "logps/chosen": -380.36578369140625,
      "logps/ref_chosen": -48.1887321472168,
      "logps/ref_rejected": -110.07643127441406,
      "logps/rejected": -977.6619873046875,
      "loss": 0.8668,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.4690607190132141,
      "rewards/margins": 0.7499971985816956,
      "rewards/rejected": -1.2190579175949097,
      "step": 541
    },
    {
      "epoch": 0.7958883994126285,
      "epsilon_dpo/beta": 0.0013966395054012537,
      "epsilon_dpo/beta_margin_grad_mean": -0.31941473484039307,
      "epsilon_dpo/beta_margin_grad_std": 0.1469242125749588,
      "epsilon_dpo/beta_margin_mean": 0.8643131852149963,
      "epsilon_dpo/beta_margin_std": 0.7869569063186646,
      "epsilon_dpo/loss_margin_mean": 621.3475952148438,
      "grad_norm": 42.56067657470703,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0014061672845855355,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 6.182866334636888e-08,
      "logits/chosen": -3.2921829223632812,
      "logits/rejected": -4.272438049316406,
      "logps/chosen": -294.858642578125,
      "logps/ref_chosen": -34.308189392089844,
      "logps/ref_rejected": -104.48388671875,
      "logps/rejected": -986.3819580078125,
      "loss": 0.8216,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.3652591109275818,
      "rewards/margins": 0.8643131852149963,
      "rewards/rejected": -1.2295722961425781,
      "step": 542
    },
    {
      "epoch": 0.7973568281938326,
      "epsilon_dpo/beta": 0.0013875396689400077,
      "epsilon_dpo/beta_margin_grad_mean": -0.32512134313583374,
      "epsilon_dpo/beta_margin_grad_std": 0.160647913813591,
      "epsilon_dpo/beta_margin_mean": 0.8715978860855103,
      "epsilon_dpo/beta_margin_std": 0.9173281192779541,
      "epsilon_dpo/loss_margin_mean": 631.060791015625,
      "grad_norm": 50.612884521484375,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.0013965658145025373,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 6.098622674699147e-08,
      "logits/chosen": -3.523017168045044,
      "logits/rejected": -4.250926971435547,
      "logps/chosen": -422.712646484375,
      "logps/ref_chosen": -43.612937927246094,
      "logps/ref_rejected": -111.56673431396484,
      "logps/rejected": -1121.7271728515625,
      "loss": 0.8507,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.527320921421051,
      "rewards/margins": 0.8715978860855103,
      "rewards/rejected": -1.398918867111206,
      "step": 543
    },
    {
      "epoch": 0.7988252569750367,
      "epsilon_dpo/beta": 0.001378059620037675,
      "epsilon_dpo/beta_margin_grad_mean": -0.3185425102710724,
      "epsilon_dpo/beta_margin_grad_std": 0.15885071456432343,
      "epsilon_dpo/beta_margin_mean": 0.9396871328353882,
      "epsilon_dpo/beta_margin_std": 1.0003536939620972,
      "epsilon_dpo/loss_margin_mean": 684.6524047851562,
      "grad_norm": 46.56395721435547,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0013874606229364872,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 6.01487708363232e-08,
      "logits/chosen": -3.383917808532715,
      "logits/rejected": -4.392078876495361,
      "logps/chosen": -419.0536193847656,
      "logps/ref_chosen": -39.47812271118164,
      "logps/ref_rejected": -106.17308044433594,
      "logps/rejected": -1170.40087890625,
      "loss": 0.8257,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.5249900221824646,
      "rewards/margins": 0.9396871328353882,
      "rewards/rejected": -1.464677095413208,
      "step": 544
    },
    {
      "epoch": 0.8002936857562408,
      "epsilon_dpo/beta": 0.001366927521303296,
      "epsilon_dpo/beta_margin_grad_mean": -0.3136303424835205,
      "epsilon_dpo/beta_margin_grad_std": 0.1529344618320465,
      "epsilon_dpo/beta_margin_mean": 0.9094871282577515,
      "epsilon_dpo/beta_margin_std": 0.8563925623893738,
      "epsilon_dpo/loss_margin_mean": 667.2245483398438,
      "grad_norm": 44.38985061645508,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.001377986976876855,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 5.9316317682106294e-08,
      "logits/chosen": -3.4539644718170166,
      "logits/rejected": -4.4015960693359375,
      "logps/chosen": -428.8343505859375,
      "logps/ref_chosen": -42.28656768798828,
      "logps/ref_rejected": -100.58802032470703,
      "logps/rejected": -1154.3603515625,
      "loss": 0.8128,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.5297784805297852,
      "rewards/margins": 0.9094871282577515,
      "rewards/rejected": -1.4392656087875366,
      "step": 545
    },
    {
      "epoch": 0.801762114537445,
      "epsilon_dpo/beta": 0.0013584739062935114,
      "epsilon_dpo/beta_margin_grad_mean": -0.33587032556533813,
      "epsilon_dpo/beta_margin_grad_std": 0.16217297315597534,
      "epsilon_dpo/beta_margin_mean": 0.7780935764312744,
      "epsilon_dpo/beta_margin_std": 0.8156005144119263,
      "epsilon_dpo/loss_margin_mean": 576.030517578125,
      "grad_norm": 47.50184631347656,
      "kl/avg_steps": 0.625,
      "kl/beta": 0.0013668810715898871,
      "kl/n_epsilon_steps": 0.1875,
      "kl/p_epsilon_steps": 0.8125,
      "learning_rate": 5.848888922025552e-08,
      "logits/chosen": -3.3195810317993164,
      "logits/rejected": -4.246737957000732,
      "logps/chosen": -418.70166015625,
      "logps/ref_chosen": -35.013492584228516,
      "logps/ref_rejected": -86.37733459472656,
      "logps/rejected": -1046.0960693359375,
      "loss": 0.8927,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": -0.5232120156288147,
      "rewards/margins": 0.7780935764312744,
      "rewards/rejected": -1.3013055324554443,
      "step": 546
    },
    {
      "epoch": 0.8032305433186491,
      "epsilon_dpo/beta": 0.0013491871068254113,
      "epsilon_dpo/beta_margin_grad_mean": -0.3124694228172302,
      "epsilon_dpo/beta_margin_grad_std": 0.15798072516918182,
      "epsilon_dpo/beta_margin_mean": 0.9112512469291687,
      "epsilon_dpo/beta_margin_std": 0.8325828909873962,
      "epsilon_dpo/loss_margin_mean": 678.1919555664062,
      "grad_norm": 46.05626678466797,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0013583911349996924,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 5.7666507254280265e-08,
      "logits/chosen": -3.386657238006592,
      "logits/rejected": -4.366371154785156,
      "logps/chosen": -437.0877380371094,
      "logps/ref_chosen": -45.810516357421875,
      "logps/ref_rejected": -95.15408325195312,
      "logps/rejected": -1164.623291015625,
      "loss": 0.8106,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.5293705463409424,
      "rewards/margins": 0.9112513065338135,
      "rewards/rejected": -1.4406218528747559,
      "step": 547
    },
    {
      "epoch": 0.8046989720998532,
      "epsilon_dpo/beta": 0.0013399748131632805,
      "epsilon_dpo/beta_margin_grad_mean": -0.3127719461917877,
      "epsilon_dpo/beta_margin_grad_std": 0.14488740265369415,
      "epsilon_dpo/beta_margin_mean": 0.946838915348053,
      "epsilon_dpo/beta_margin_std": 0.9356114864349365,
      "epsilon_dpo/loss_margin_mean": 709.183349609375,
      "grad_norm": 44.01782989501953,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.001349115977063775,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 5.684919345471029e-08,
      "logits/chosen": -3.3794431686401367,
      "logits/rejected": -4.389235496520996,
      "logps/chosen": -411.8974914550781,
      "logps/ref_chosen": -40.29045486450195,
      "logps/ref_rejected": -98.83418273925781,
      "logps/rejected": -1179.62451171875,
      "loss": 0.7973,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.4989817142486572,
      "rewards/margins": 0.9468388557434082,
      "rewards/rejected": -1.4458205699920654,
      "step": 548
    },
    {
      "epoch": 0.8061674008810573,
      "epsilon_dpo/beta": 0.0013320817379280925,
      "epsilon_dpo/beta_margin_grad_mean": -0.35267123579978943,
      "epsilon_dpo/beta_margin_grad_std": 0.17075052857398987,
      "epsilon_dpo/beta_margin_mean": 0.7173733115196228,
      "epsilon_dpo/beta_margin_std": 0.8851867318153381,
      "epsilon_dpo/loss_margin_mean": 541.94580078125,
      "grad_norm": 81.65141296386719,
      "kl/avg_steps": 0.59375,
      "kl/beta": 0.0013399041490629315,
      "kl/n_epsilon_steps": 0.203125,
      "kl/p_epsilon_steps": 0.796875,
      "learning_rate": 5.603696935852426e-08,
      "logits/chosen": -3.2905972003936768,
      "logits/rejected": -4.312599182128906,
      "logps/chosen": -439.40911865234375,
      "logps/ref_chosen": -33.00281524658203,
      "logps/ref_rejected": -81.72914123535156,
      "logps/rejected": -1030.0811767578125,
      "loss": 0.9535,
      "rewards/accuracies": 0.828125,
      "rewards/chosen": -0.5441075563430786,
      "rewards/margins": 0.717373251914978,
      "rewards/rejected": -1.2614808082580566,
      "step": 549
    },
    {
      "epoch": 0.8076358296622613,
      "epsilon_dpo/beta": 0.0013221375411376357,
      "epsilon_dpo/beta_margin_grad_mean": -0.30787116289138794,
      "epsilon_dpo/beta_margin_grad_std": 0.15403428673744202,
      "epsilon_dpo/beta_margin_mean": 0.9514063596725464,
      "epsilon_dpo/beta_margin_std": 0.8637392520904541,
      "epsilon_dpo/loss_margin_mean": 721.8765869140625,
      "grad_norm": 47.79182052612305,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.001331995357759297,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 5.5229856368582376e-08,
      "logits/chosen": -3.2790517807006836,
      "logits/rejected": -4.2515106201171875,
      "logps/chosen": -415.656982421875,
      "logps/ref_chosen": -33.51374053955078,
      "logps/ref_rejected": -101.08897399902344,
      "logps/rejected": -1205.10888671875,
      "loss": 0.79,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.5064237713813782,
      "rewards/margins": 0.9514063596725464,
      "rewards/rejected": -1.4578301906585693,
      "step": 550
    },
    {
      "epoch": 0.8091042584434655,
      "epsilon_dpo/beta": 0.0013135350309312344,
      "epsilon_dpo/beta_margin_grad_mean": -0.28273874521255493,
      "epsilon_dpo/beta_margin_grad_std": 0.17567141354084015,
      "epsilon_dpo/beta_margin_mean": 1.0942435264587402,
      "epsilon_dpo/beta_margin_std": 0.9498324394226074,
      "epsilon_dpo/loss_margin_mean": 836.8770751953125,
      "grad_norm": 45.95981979370117,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.0013220797991380095,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 5.4427875753062734e-08,
      "logits/chosen": -3.347069501876831,
      "logits/rejected": -4.404940605163574,
      "logps/chosen": -488.1016845703125,
      "logps/ref_chosen": -42.42958068847656,
      "logps/ref_rejected": -117.51499938964844,
      "logps/rejected": -1400.064208984375,
      "loss": 0.7468,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.5882077217102051,
      "rewards/margins": 1.0942435264587402,
      "rewards/rejected": -1.6824512481689453,
      "step": 551
    },
    {
      "epoch": 0.8105726872246696,
      "epsilon_dpo/beta": 0.0013033291324973106,
      "epsilon_dpo/beta_margin_grad_mean": -0.27695924043655396,
      "epsilon_dpo/beta_margin_grad_std": 0.17922824621200562,
      "epsilon_dpo/beta_margin_mean": 1.1912535429000854,
      "epsilon_dpo/beta_margin_std": 1.1001869440078735,
      "epsilon_dpo/loss_margin_mean": 916.8844604492188,
      "grad_norm": 40.460792541503906,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0013134601758792996,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 5.363104864490034e-08,
      "logits/chosen": -3.3478827476501465,
      "logits/rejected": -4.524975776672363,
      "logps/chosen": -496.78289794921875,
      "logps/ref_chosen": -40.86379623413086,
      "logps/ref_rejected": -112.38937377929688,
      "logps/rejected": -1485.19287109375,
      "loss": 0.7312,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.5959450006484985,
      "rewards/margins": 1.1912535429000854,
      "rewards/rejected": -1.787198543548584,
      "step": 552
    },
    {
      "epoch": 0.8120411160058737,
      "epsilon_dpo/beta": 0.001297299051657319,
      "epsilon_dpo/beta_margin_grad_mean": -0.3533886671066284,
      "epsilon_dpo/beta_margin_grad_std": 0.17754045128822327,
      "epsilon_dpo/beta_margin_mean": 0.7519553303718567,
      "epsilon_dpo/beta_margin_std": 1.001431941986084,
      "epsilon_dpo/loss_margin_mean": 583.842041015625,
      "grad_norm": 88.0752182006836,
      "kl/avg_steps": 0.46875,
      "kl/beta": 0.001303278375416994,
      "kl/n_epsilon_steps": 0.265625,
      "kl/p_epsilon_steps": 0.734375,
      "learning_rate": 5.2839396041230415e-08,
      "logits/chosen": -3.496969699859619,
      "logits/rejected": -4.206221580505371,
      "logps/chosen": -561.1559448242188,
      "logps/ref_chosen": -51.35152816772461,
      "logps/ref_rejected": -103.84159851074219,
      "logps/rejected": -1197.488037109375,
      "loss": 0.9607,
      "rewards/accuracies": 0.796875,
      "rewards/chosen": -0.6639860272407532,
      "rewards/margins": 0.7519553303718567,
      "rewards/rejected": -1.4159413576126099,
      "step": 553
    },
    {
      "epoch": 0.8135095447870778,
      "epsilon_dpo/beta": 0.001288408413529396,
      "epsilon_dpo/beta_margin_grad_mean": -0.31087225675582886,
      "epsilon_dpo/beta_margin_grad_std": 0.16114763915538788,
      "epsilon_dpo/beta_margin_mean": 0.9212867021560669,
      "epsilon_dpo/beta_margin_std": 0.8502605557441711,
      "epsilon_dpo/loss_margin_mean": 718.2112426757812,
      "grad_norm": 48.403167724609375,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0012971977703273296,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 5.205293880283551e-08,
      "logits/chosen": -3.321976661682129,
      "logits/rejected": -4.445281982421875,
      "logps/chosen": -478.4771728515625,
      "logps/ref_chosen": -46.820640563964844,
      "logps/ref_rejected": -94.6780014038086,
      "logps/rejected": -1244.5457763671875,
      "loss": 0.8105,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.5585451126098633,
      "rewards/margins": 0.9212867021560669,
      "rewards/rejected": -1.4798319339752197,
      "step": 554
    },
    {
      "epoch": 0.8149779735682819,
      "epsilon_dpo/beta": 0.001279208343476057,
      "epsilon_dpo/beta_margin_grad_mean": -0.27689388394355774,
      "epsilon_dpo/beta_margin_grad_std": 0.16369499266147614,
      "epsilon_dpo/beta_margin_mean": 1.1780664920806885,
      "epsilon_dpo/beta_margin_std": 1.0308868885040283,
      "epsilon_dpo/loss_margin_mean": 923.9918212890625,
      "grad_norm": 45.80141830444336,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0012883404269814491,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 5.127169765359515e-08,
      "logits/chosen": -3.4040637016296387,
      "logits/rejected": -4.443061828613281,
      "logps/chosen": -423.08660888671875,
      "logps/ref_chosen": -33.03218460083008,
      "logps/ref_rejected": -115.6031265258789,
      "logps/rejected": -1429.6494140625,
      "loss": 0.7069,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.49974268674850464,
      "rewards/margins": 1.1780664920806885,
      "rewards/rejected": -1.677809238433838,
      "step": 555
    },
    {
      "epoch": 0.8164464023494861,
      "epsilon_dpo/beta": 0.001270079636014998,
      "epsilon_dpo/beta_margin_grad_mean": -0.3221823573112488,
      "epsilon_dpo/beta_margin_grad_std": 0.15197478234767914,
      "epsilon_dpo/beta_margin_mean": 0.8850791454315186,
      "epsilon_dpo/beta_margin_std": 0.8993016481399536,
      "epsilon_dpo/loss_margin_mean": 699.4880981445312,
      "grad_norm": 50.429317474365234,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.001279146526940167,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 5.049569317994012e-08,
      "logits/chosen": -3.259387493133545,
      "logits/rejected": -4.264613151550293,
      "logps/chosen": -447.8577575683594,
      "logps/ref_chosen": -32.514793395996094,
      "logps/ref_rejected": -106.99200439453125,
      "logps/rejected": -1221.822998046875,
      "loss": 0.8334,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.5285704135894775,
      "rewards/margins": 0.8850791454315186,
      "rewards/rejected": -1.413649559020996,
      "step": 556
    },
    {
      "epoch": 0.8179148311306902,
      "epsilon_dpo/beta": 0.00126141298096627,
      "epsilon_dpo/beta_margin_grad_mean": -0.3104838728904724,
      "epsilon_dpo/beta_margin_grad_std": 0.16874933242797852,
      "epsilon_dpo/beta_margin_mean": 0.9584492444992065,
      "epsilon_dpo/beta_margin_std": 0.9464617371559143,
      "epsilon_dpo/loss_margin_mean": 763.000244140625,
      "grad_norm": 49.268924713134766,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0012700182851403952,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 4.9724945830310144e-08,
      "logits/chosen": -3.3845365047454834,
      "logits/rejected": -4.363818645477295,
      "logps/chosen": -543.5952758789062,
      "logps/ref_chosen": -51.166099548339844,
      "logps/ref_rejected": -117.06163024902344,
      "logps/rejected": -1372.490966796875,
      "loss": 0.8139,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.6233251094818115,
      "rewards/margins": 0.9584492444992065,
      "rewards/rejected": -1.581774353981018,
      "step": 557
    },
    {
      "epoch": 0.8193832599118943,
      "epsilon_dpo/beta": 0.0012516174465417862,
      "epsilon_dpo/beta_margin_grad_mean": -0.27005571126937866,
      "epsilon_dpo/beta_margin_grad_std": 0.14844605326652527,
      "epsilon_dpo/beta_margin_mean": 1.16920006275177,
      "epsilon_dpo/beta_margin_std": 0.9076313972473145,
      "epsilon_dpo/loss_margin_mean": 936.5662841796875,
      "grad_norm": 45.054534912109375,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0012613465078175068,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 4.8959475914614554e-08,
      "logits/chosen": -3.1819422245025635,
      "logits/rejected": -4.413593292236328,
      "logps/chosen": -387.129150390625,
      "logps/ref_chosen": -38.345916748046875,
      "logps/ref_rejected": -109.03158569335938,
      "logps/rejected": -1394.381103515625,
      "loss": 0.6774,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.4374838173389435,
      "rewards/margins": 1.16920006275177,
      "rewards/rejected": -1.6066837310791016,
      "step": 558
    },
    {
      "epoch": 0.8208516886930984,
      "epsilon_dpo/beta": 0.0012430885108187795,
      "epsilon_dpo/beta_margin_grad_mean": -0.31191176176071167,
      "epsilon_dpo/beta_margin_grad_std": 0.17650018632411957,
      "epsilon_dpo/beta_margin_mean": 0.9644886255264282,
      "epsilon_dpo/beta_margin_std": 1.0081322193145752,
      "epsilon_dpo/loss_margin_mean": 779.3812255859375,
      "grad_norm": 63.11558151245117,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0012515686685219407,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 4.8199303603697614e-08,
      "logits/chosen": -3.2931509017944336,
      "logits/rejected": -4.473822593688965,
      "logps/chosen": -542.97265625,
      "logps/ref_chosen": -42.82691955566406,
      "logps/ref_rejected": -98.61148071289062,
      "logps/rejected": -1378.138427734375,
      "loss": 0.8302,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.6229937672615051,
      "rewards/margins": 0.9644886255264282,
      "rewards/rejected": -1.5874824523925781,
      "step": 559
    },
    {
      "epoch": 0.8223201174743024,
      "epsilon_dpo/beta": 0.0012365429429337382,
      "epsilon_dpo/beta_margin_grad_mean": -0.34858644008636475,
      "epsilon_dpo/beta_margin_grad_std": 0.15379458665847778,
      "epsilon_dpo/beta_margin_mean": 0.7325094938278198,
      "epsilon_dpo/beta_margin_std": 0.8337624669075012,
      "epsilon_dpo/loss_margin_mean": 595.9330444335938,
      "grad_norm": 46.32260513305664,
      "kl/avg_steps": 0.53125,
      "kl/beta": 0.001243022852577269,
      "kl/n_epsilon_steps": 0.234375,
      "kl/p_epsilon_steps": 0.765625,
      "learning_rate": 4.7444448928806615e-08,
      "logits/chosen": -3.276106834411621,
      "logits/rejected": -4.303354740142822,
      "logps/chosen": -445.64739990234375,
      "logps/ref_chosen": -36.52899932861328,
      "logps/ref_rejected": -85.29887390136719,
      "logps/rejected": -1090.350341796875,
      "loss": 0.9186,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.5071742534637451,
      "rewards/margins": 0.7325094938278198,
      "rewards/rejected": -1.2396838665008545,
      "step": 560
    },
    {
      "epoch": 0.8237885462555066,
      "epsilon_dpo/beta": 0.0012296221684664488,
      "epsilon_dpo/beta_margin_grad_mean": -0.32985633611679077,
      "epsilon_dpo/beta_margin_grad_std": 0.16898517310619354,
      "epsilon_dpo/beta_margin_mean": 0.8398326635360718,
      "epsilon_dpo/beta_margin_std": 0.8977148532867432,
      "epsilon_dpo/loss_margin_mean": 686.9766845703125,
      "grad_norm": 51.50674057006836,
      "kl/avg_steps": 0.5625,
      "kl/beta": 0.001236454234458506,
      "kl/n_epsilon_steps": 0.21875,
      "kl/p_epsilon_steps": 0.78125,
      "learning_rate": 4.669493178106432e-08,
      "logits/chosen": -3.1779091358184814,
      "logits/rejected": -4.242127418518066,
      "logps/chosen": -401.35638427734375,
      "logps/ref_chosen": -33.39714050292969,
      "logps/ref_rejected": -104.86662292480469,
      "logps/rejected": -1159.802490234375,
      "loss": 0.8751,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.45402687788009644,
      "rewards/margins": 0.8398326635360718,
      "rewards/rejected": -1.293859601020813,
      "step": 561
    },
    {
      "epoch": 0.8252569750367107,
      "epsilon_dpo/beta": 0.001219669939018786,
      "epsilon_dpo/beta_margin_grad_mean": -0.31512200832366943,
      "epsilon_dpo/beta_margin_grad_std": 0.15188972651958466,
      "epsilon_dpo/beta_margin_mean": 0.9067568182945251,
      "epsilon_dpo/beta_margin_std": 0.8777819275856018,
      "epsilon_dpo/loss_margin_mean": 745.5482788085938,
      "grad_norm": 58.34850311279297,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.001229538000188768,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 4.5950771910944596e-08,
      "logits/chosen": -3.1661272048950195,
      "logits/rejected": -4.288407325744629,
      "logps/chosen": -410.8808898925781,
      "logps/ref_chosen": -37.05860900878906,
      "logps/ref_rejected": -104.74354553222656,
      "logps/rejected": -1224.1141357421875,
      "loss": 0.8169,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.4571970999240875,
      "rewards/margins": 0.9067568182945251,
      "rewards/rejected": -1.363953948020935,
      "step": 562
    },
    {
      "epoch": 0.8267254038179148,
      "epsilon_dpo/beta": 0.0012125081848353148,
      "epsilon_dpo/beta_margin_grad_mean": -0.33571353554725647,
      "epsilon_dpo/beta_margin_grad_std": 0.15231700241565704,
      "epsilon_dpo/beta_margin_mean": 0.7808871269226074,
      "epsilon_dpo/beta_margin_std": 0.7981037497520447,
      "epsilon_dpo/loss_margin_mean": 647.4954223632812,
      "grad_norm": 56.34610366821289,
      "kl/avg_steps": 0.59375,
      "kl/beta": 0.0012196284951642156,
      "kl/n_epsilon_steps": 0.203125,
      "kl/p_epsilon_steps": 0.796875,
      "learning_rate": 4.521198892775202e-08,
      "logits/chosen": -3.220041275024414,
      "logits/rejected": -4.239965915679932,
      "logps/chosen": -462.99176025390625,
      "logps/ref_chosen": -44.557411193847656,
      "logps/ref_rejected": -102.11149597167969,
      "logps/rejected": -1168.041259765625,
      "loss": 0.8805,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.5091303586959839,
      "rewards/margins": 0.7808871269226074,
      "rewards/rejected": -1.2900173664093018,
      "step": 563
    },
    {
      "epoch": 0.8281938325991189,
      "epsilon_dpo/beta": 0.0012034567771479487,
      "epsilon_dpo/beta_margin_grad_mean": -0.3051425516605377,
      "epsilon_dpo/beta_margin_grad_std": 0.15122495591640472,
      "epsilon_dpo/beta_margin_mean": 0.944663405418396,
      "epsilon_dpo/beta_margin_std": 0.820910632610321,
      "epsilon_dpo/loss_margin_mean": 787.502197265625,
      "grad_norm": 49.93864059448242,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0012124297209084034,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 4.447860229910544e-08,
      "logits/chosen": -3.3658523559570312,
      "logits/rejected": -4.48719596862793,
      "logps/chosen": -445.9974060058594,
      "logps/ref_chosen": -43.770145416259766,
      "logps/ref_rejected": -99.07379150390625,
      "logps/rejected": -1288.80322265625,
      "loss": 0.7846,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.4852016270160675,
      "rewards/margins": 0.944663405418396,
      "rewards/rejected": -1.4298651218414307,
      "step": 564
    },
    {
      "epoch": 0.8296622613803231,
      "epsilon_dpo/beta": 0.001195436459966004,
      "epsilon_dpo/beta_margin_grad_mean": -0.3250107765197754,
      "epsilon_dpo/beta_margin_grad_std": 0.1642923802137375,
      "epsilon_dpo/beta_margin_mean": 0.8570442795753479,
      "epsilon_dpo/beta_margin_std": 0.9114473462104797,
      "epsilon_dpo/loss_margin_mean": 720.2861328125,
      "grad_norm": 44.75363540649414,
      "kl/avg_steps": 0.671875,
      "kl/beta": 0.001203404157422483,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 4.375063135042445e-08,
      "logits/chosen": -3.1451213359832764,
      "logits/rejected": -4.4003705978393555,
      "logps/chosen": -405.4493103027344,
      "logps/ref_chosen": -35.76008605957031,
      "logps/ref_rejected": -92.45127868652344,
      "logps/rejected": -1182.4266357421875,
      "loss": 0.8657,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.4435133934020996,
      "rewards/margins": 0.8570442199707031,
      "rewards/rejected": -1.3005576133728027,
      "step": 565
    },
    {
      "epoch": 0.8311306901615272,
      "epsilon_dpo/beta": 0.0011870847083628178,
      "epsilon_dpo/beta_margin_grad_mean": -0.3126325309276581,
      "epsilon_dpo/beta_margin_grad_std": 0.1551150530576706,
      "epsilon_dpo/beta_margin_mean": 0.9413906335830688,
      "epsilon_dpo/beta_margin_std": 0.8944268226623535,
      "epsilon_dpo/loss_margin_mean": 795.7219848632812,
      "grad_norm": 53.6401481628418,
      "kl/avg_steps": 0.703125,
      "kl/beta": 0.0011953727807849646,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 4.3028095264420525e-08,
      "logits/chosen": -3.165487289428711,
      "logits/rejected": -4.408507347106934,
      "logps/chosen": -425.1823425292969,
      "logps/ref_chosen": -44.17033004760742,
      "logps/ref_rejected": -111.02940368652344,
      "logps/rejected": -1287.763427734375,
      "loss": 0.8029,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.4532804787158966,
      "rewards/margins": 0.9413906335830688,
      "rewards/rejected": -1.394671082496643,
      "step": 566
    },
    {
      "epoch": 0.8325991189427313,
      "epsilon_dpo/beta": 0.001178241684101522,
      "epsilon_dpo/beta_margin_grad_mean": -0.32961148023605347,
      "epsilon_dpo/beta_margin_grad_std": 0.1618032306432724,
      "epsilon_dpo/beta_margin_mean": 0.8316484093666077,
      "epsilon_dpo/beta_margin_std": 0.862678050994873,
      "epsilon_dpo/loss_margin_mean": 708.5885009765625,
      "grad_norm": 55.0069694519043,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0011870265007019043,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 4.231101308059165e-08,
      "logits/chosen": -3.2106685638427734,
      "logits/rejected": -4.49782657623291,
      "logps/chosen": -419.98046875,
      "logps/ref_chosen": -34.52465057373047,
      "logps/ref_rejected": -92.00111389160156,
      "logps/rejected": -1186.04541015625,
      "loss": 0.8671,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.4556201100349426,
      "rewards/margins": 0.8316484093666077,
      "rewards/rejected": -1.2872684001922607,
      "step": 567
    },
    {
      "epoch": 0.8340675477239354,
      "epsilon_dpo/beta": 0.0011691023828461766,
      "epsilon_dpo/beta_margin_grad_mean": -0.3085906505584717,
      "epsilon_dpo/beta_margin_grad_std": 0.14849402010440826,
      "epsilon_dpo/beta_margin_mean": 0.9396561980247498,
      "epsilon_dpo/beta_margin_std": 0.8285186886787415,
      "epsilon_dpo/loss_margin_mean": 805.8914794921875,
      "grad_norm": 51.46321487426758,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0011781901121139526,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 4.1599403694720145e-08,
      "logits/chosen": -3.223677158355713,
      "logits/rejected": -4.477365493774414,
      "logps/chosen": -342.3359375,
      "logps/ref_chosen": -20.8466854095459,
      "logps/ref_rejected": -96.66067504882812,
      "logps/rejected": -1224.04150390625,
      "loss": 0.7867,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.37704116106033325,
      "rewards/margins": 0.9396562576293945,
      "rewards/rejected": -1.316697359085083,
      "step": 568
    },
    {
      "epoch": 0.8355359765051396,
      "epsilon_dpo/beta": 0.0011607703054323792,
      "epsilon_dpo/beta_margin_grad_mean": -0.3284705877304077,
      "epsilon_dpo/beta_margin_grad_std": 0.159523144364357,
      "epsilon_dpo/beta_margin_mean": 0.8572857975959778,
      "epsilon_dpo/beta_margin_std": 0.9337847232818604,
      "epsilon_dpo/loss_margin_mean": 741.6339111328125,
      "grad_norm": 49.56915283203125,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0011690568644553423,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 4.089328585837512e-08,
      "logits/chosen": -3.329655647277832,
      "logits/rejected": -4.526782035827637,
      "logps/chosen": -403.1963195800781,
      "logps/ref_chosen": -44.985755920410156,
      "logps/ref_rejected": -84.71590423583984,
      "logps/rejected": -1184.5604248046875,
      "loss": 0.8626,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.41802793741226196,
      "rewards/margins": 0.857285737991333,
      "rewards/rejected": -1.2753137350082397,
      "step": 569
    },
    {
      "epoch": 0.8370044052863436,
      "epsilon_dpo/beta": 0.0011524868896231055,
      "epsilon_dpo/beta_margin_grad_mean": -0.3267635405063629,
      "epsilon_dpo/beta_margin_grad_std": 0.1530372053384781,
      "epsilon_dpo/beta_margin_mean": 0.8452085256576538,
      "epsilon_dpo/beta_margin_std": 0.8515332341194153,
      "epsilon_dpo/loss_margin_mean": 736.2088012695312,
      "grad_norm": 46.32669448852539,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0011607141932472587,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 4.019267817841834e-08,
      "logits/chosen": -3.3636648654937744,
      "logits/rejected": -4.605798721313477,
      "logps/chosen": -474.98382568359375,
      "logps/ref_chosen": -39.338470458984375,
      "logps/ref_rejected": -88.34091186523438,
      "logps/rejected": -1260.195068359375,
      "loss": 0.8506,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.5037564635276794,
      "rewards/margins": 0.8452085256576538,
      "rewards/rejected": -1.3489649295806885,
      "step": 570
    },
    {
      "epoch": 0.8384728340675477,
      "epsilon_dpo/beta": 0.001143542118370533,
      "epsilon_dpo/beta_margin_grad_mean": -0.2934305667877197,
      "epsilon_dpo/beta_margin_grad_std": 0.16104663908481598,
      "epsilon_dpo/beta_margin_mean": 1.0465232133865356,
      "epsilon_dpo/beta_margin_std": 0.9296081066131592,
      "epsilon_dpo/loss_margin_mean": 917.7656860351562,
      "grad_norm": 43.56798553466797,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0011524311266839504,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 3.9497599116513705e-08,
      "logits/chosen": -3.367818832397461,
      "logits/rejected": -4.655285835266113,
      "logps/chosen": -398.0816650390625,
      "logps/ref_chosen": -27.42264747619629,
      "logps/ref_rejected": -99.594482421875,
      "logps/rejected": -1388.0191650390625,
      "loss": 0.7563,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.4253048002719879,
      "rewards/margins": 1.0465232133865356,
      "rewards/rejected": -1.4718279838562012,
      "step": 571
    },
    {
      "epoch": 0.8399412628487518,
      "epsilon_dpo/beta": 0.001137536484748125,
      "epsilon_dpo/beta_margin_grad_mean": -0.32172301411628723,
      "epsilon_dpo/beta_margin_grad_std": 0.18587327003479004,
      "epsilon_dpo/beta_margin_mean": 0.8928326964378357,
      "epsilon_dpo/beta_margin_std": 0.9588112831115723,
      "epsilon_dpo/loss_margin_mean": 790.0717163085938,
      "grad_norm": 59.63137435913086,
      "kl/avg_steps": 0.53125,
      "kl/beta": 0.0011434975313022733,
      "kl/n_epsilon_steps": 0.234375,
      "kl/p_epsilon_steps": 0.765625,
      "learning_rate": 3.880806698864086e-08,
      "logits/chosen": -3.412736654281616,
      "logits/rejected": -4.5799455642700195,
      "logps/chosen": -437.66815185546875,
      "logps/ref_chosen": -28.469844818115234,
      "logps/ref_rejected": -90.8378677368164,
      "logps/rejected": -1290.10791015625,
      "loss": 0.8689,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": -0.4676932096481323,
      "rewards/margins": 0.8928326368331909,
      "rewards/rejected": -1.3605258464813232,
      "step": 572
    },
    {
      "epoch": 0.8414096916299559,
      "epsilon_dpo/beta": 0.00113116973079741,
      "epsilon_dpo/beta_margin_grad_mean": -0.3352082371711731,
      "epsilon_dpo/beta_margin_grad_std": 0.1706962138414383,
      "epsilon_dpo/beta_margin_mean": 0.8242603540420532,
      "epsilon_dpo/beta_margin_std": 0.914152204990387,
      "epsilon_dpo/loss_margin_mean": 732.9346923828125,
      "grad_norm": 58.05121612548828,
      "kl/avg_steps": 0.5625,
      "kl/beta": 0.0011374547611922026,
      "kl/n_epsilon_steps": 0.21875,
      "kl/p_epsilon_steps": 0.78125,
      "learning_rate": 3.812409996461275e-08,
      "logits/chosen": -3.55238676071167,
      "logits/rejected": -4.599684715270996,
      "logps/chosen": -550.3379516601562,
      "logps/ref_chosen": -35.605384826660156,
      "logps/ref_rejected": -91.72334289550781,
      "logps/rejected": -1339.390625,
      "loss": 0.8899,
      "rewards/accuracies": 0.796875,
      "rewards/chosen": -0.5845237970352173,
      "rewards/margins": 0.8242603540420532,
      "rewards/rejected": -1.4087841510772705,
      "step": 573
    },
    {
      "epoch": 0.8428781204111601,
      "epsilon_dpo/beta": 0.0011223680339753628,
      "epsilon_dpo/beta_margin_grad_mean": -0.2976383864879608,
      "epsilon_dpo/beta_margin_grad_std": 0.15246038138866425,
      "epsilon_dpo/beta_margin_mean": 1.003798246383667,
      "epsilon_dpo/beta_margin_std": 0.8582514524459839,
      "epsilon_dpo/loss_margin_mean": 896.7732543945312,
      "grad_norm": 47.21782302856445,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.001131092431023717,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 3.74457160675965e-08,
      "logits/chosen": -3.418008327484131,
      "logits/rejected": -4.645315170288086,
      "logps/chosen": -419.362060546875,
      "logps/ref_chosen": -31.43255615234375,
      "logps/ref_rejected": -98.70758056640625,
      "logps/rejected": -1383.410400390625,
      "loss": 0.7582,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.4360452890396118,
      "rewards/margins": 1.003798246383667,
      "rewards/rejected": -1.4398435354232788,
      "step": 574
    },
    {
      "epoch": 0.8443465491923642,
      "epsilon_dpo/beta": 0.0011150705395266414,
      "epsilon_dpo/beta_margin_grad_mean": -0.3144918978214264,
      "epsilon_dpo/beta_margin_grad_std": 0.16865824162960052,
      "epsilon_dpo/beta_margin_mean": 0.9695743918418884,
      "epsilon_dpo/beta_margin_std": 1.0260599851608276,
      "epsilon_dpo/loss_margin_mean": 873.2989501953125,
      "grad_norm": 52.89796447753906,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.0011223242618143559,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 3.677293317363864e-08,
      "logits/chosen": -3.360232353210449,
      "logits/rejected": -4.744388580322266,
      "logps/chosen": -476.3949279785156,
      "logps/ref_chosen": -40.592437744140625,
      "logps/ref_rejected": -102.49923706054688,
      "logps/rejected": -1411.6007080078125,
      "loss": 0.8223,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.4872722625732422,
      "rewards/margins": 0.9695743322372437,
      "rewards/rejected": -1.4568467140197754,
      "step": 575
    },
    {
      "epoch": 0.8458149779735683,
      "epsilon_dpo/beta": 0.001109194359742105,
      "epsilon_dpo/beta_margin_grad_mean": -0.3279215395450592,
      "epsilon_dpo/beta_margin_grad_std": 0.1897161304950714,
      "epsilon_dpo/beta_margin_mean": 0.9002985954284668,
      "epsilon_dpo/beta_margin_std": 1.0915855169296265,
      "epsilon_dpo/loss_margin_mean": 817.2003784179688,
      "grad_norm": 49.19211196899414,
      "kl/avg_steps": 0.53125,
      "kl/beta": 0.0011150069767609239,
      "kl/n_epsilon_steps": 0.234375,
      "kl/p_epsilon_steps": 0.765625,
      "learning_rate": 3.6105769011194224e-08,
      "logits/chosen": -3.442026376724243,
      "logits/rejected": -4.685531139373779,
      "logps/chosen": -509.7559814453125,
      "logps/ref_chosen": -34.52962875366211,
      "logps/ref_rejected": -107.82730102539062,
      "logps/rejected": -1400.2540283203125,
      "loss": 0.8987,
      "rewards/accuracies": 0.796875,
      "rewards/chosen": -0.5297691822052002,
      "rewards/margins": 0.9002985954284668,
      "rewards/rejected": -1.430067777633667,
      "step": 576
    },
    {
      "epoch": 0.8472834067547724,
      "epsilon_dpo/beta": 0.0011009065201506019,
      "epsilon_dpo/beta_margin_grad_mean": -0.3192014694213867,
      "epsilon_dpo/beta_margin_grad_std": 0.13799449801445007,
      "epsilon_dpo/beta_margin_mean": 0.8508294224739075,
      "epsilon_dpo/beta_margin_std": 0.7307493686676025,
      "epsilon_dpo/loss_margin_mean": 775.322509765625,
      "grad_norm": 47.395843505859375,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0011091148480772972,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 3.5444241160659304e-08,
      "logits/chosen": -3.390251636505127,
      "logits/rejected": -4.721263885498047,
      "logps/chosen": -472.676513671875,
      "logps/ref_chosen": -36.379669189453125,
      "logps/ref_rejected": -91.19670104980469,
      "logps/rejected": -1302.816162109375,
      "loss": 0.8153,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.4818669557571411,
      "rewards/margins": 0.8508294224739075,
      "rewards/rejected": -1.3326964378356934,
      "step": 577
    },
    {
      "epoch": 0.8487518355359766,
      "epsilon_dpo/beta": 0.0010930553544312716,
      "epsilon_dpo/beta_margin_grad_mean": -0.31814831495285034,
      "epsilon_dpo/beta_margin_grad_std": 0.1622655838727951,
      "epsilon_dpo/beta_margin_mean": 0.8751156330108643,
      "epsilon_dpo/beta_margin_std": 0.8696184754371643,
      "epsilon_dpo/loss_margin_mean": 803.7972412109375,
      "grad_norm": 56.689117431640625,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0011008584406226873,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 3.478836705390808e-08,
      "logits/chosen": -3.35703182220459,
      "logits/rejected": -4.623649597167969,
      "logps/chosen": -402.2279357910156,
      "logps/ref_chosen": -30.31075668334961,
      "logps/ref_rejected": -89.46212005615234,
      "logps/rejected": -1265.176513671875,
      "loss": 0.8456,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.4083082675933838,
      "rewards/margins": 0.8751156330108643,
      "rewards/rejected": -1.283423900604248,
      "step": 578
    },
    {
      "epoch": 0.8502202643171806,
      "epsilon_dpo/beta": 0.0010845718206837773,
      "epsilon_dpo/beta_margin_grad_mean": -0.3247964680194855,
      "epsilon_dpo/beta_margin_grad_std": 0.13387812674045563,
      "epsilon_dpo/beta_margin_mean": 0.8189929723739624,
      "epsilon_dpo/beta_margin_std": 0.7138299345970154,
      "epsilon_dpo/loss_margin_mean": 757.1980590820312,
      "grad_norm": 47.10636520385742,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0010930025018751621,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 3.41381639738331e-08,
      "logits/chosen": -3.4386825561523438,
      "logits/rejected": -4.622516632080078,
      "logps/chosen": -481.96820068359375,
      "logps/ref_chosen": -38.872169494628906,
      "logps/ref_rejected": -102.441162109375,
      "logps/rejected": -1302.7352294921875,
      "loss": 0.8296,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.48166942596435547,
      "rewards/margins": 0.8189929723739624,
      "rewards/rejected": -1.3006623983383179,
      "step": 579
    },
    {
      "epoch": 0.8516886930983847,
      "epsilon_dpo/beta": 0.0010795537382364273,
      "epsilon_dpo/beta_margin_grad_mean": -0.3553590476512909,
      "epsilon_dpo/beta_margin_grad_std": 0.1680212914943695,
      "epsilon_dpo/beta_margin_mean": 0.7070707082748413,
      "epsilon_dpo/beta_margin_std": 0.8788829445838928,
      "epsilon_dpo/loss_margin_mean": 659.7229614257812,
      "grad_norm": 53.16489028930664,
      "kl/avg_steps": 0.46875,
      "kl/beta": 0.0010845295619219542,
      "kl/n_epsilon_steps": 0.265625,
      "kl/p_epsilon_steps": 0.734375,
      "learning_rate": 3.349364905389032e-08,
      "logits/chosen": -3.4820971488952637,
      "logits/rejected": -4.580034255981445,
      "logps/chosen": -516.2334594726562,
      "logps/ref_chosen": -43.879005432128906,
      "logps/ref_rejected": -87.93156433105469,
      "logps/rejected": -1220.0089111328125,
      "loss": 0.9567,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": -0.5118281841278076,
      "rewards/margins": 0.7070707082748413,
      "rewards/rejected": -1.2188990116119385,
      "step": 580
    },
    {
      "epoch": 0.8531571218795888,
      "epsilon_dpo/beta": 0.0010721554281190038,
      "epsilon_dpo/beta_margin_grad_mean": -0.32997187972068787,
      "epsilon_dpo/beta_margin_grad_std": 0.16221848130226135,
      "epsilon_dpo/beta_margin_mean": 0.8481745719909668,
      "epsilon_dpo/beta_margin_std": 0.9297420382499695,
      "epsilon_dpo/loss_margin_mean": 794.6920166015625,
      "grad_norm": 48.09032440185547,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0010794695699587464,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 3.285483927764726e-08,
      "logits/chosen": -3.527688503265381,
      "logits/rejected": -4.8321614265441895,
      "logps/chosen": -542.8037109375,
      "logps/ref_chosen": -43.27698516845703,
      "logps/ref_rejected": -97.41041564941406,
      "logps/rejected": -1391.629150390625,
      "loss": 0.8701,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.5371580123901367,
      "rewards/margins": 0.8481745719909668,
      "rewards/rejected": -1.3853325843811035,
      "step": 581
    },
    {
      "epoch": 0.8546255506607929,
      "epsilon_dpo/beta": 0.001064834650605917,
      "epsilon_dpo/beta_margin_grad_mean": -0.32155805826187134,
      "epsilon_dpo/beta_margin_grad_std": 0.18931035697460175,
      "epsilon_dpo/beta_margin_mean": 0.9292847514152527,
      "epsilon_dpo/beta_margin_std": 1.0593138933181763,
      "epsilon_dpo/loss_margin_mean": 877.0955200195312,
      "grad_norm": 56.42830276489258,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0010720988502725959,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 3.222175147833556e-08,
      "logits/chosen": -3.625027656555176,
      "logits/rejected": -4.7767333984375,
      "logps/chosen": -586.5724487304688,
      "logps/ref_chosen": -47.183502197265625,
      "logps/ref_rejected": -118.7525634765625,
      "logps/rejected": -1535.237060546875,
      "loss": 0.874,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.5769472718238831,
      "rewards/margins": 0.9292848110198975,
      "rewards/rejected": -1.5062320232391357,
      "step": 582
    },
    {
      "epoch": 0.856093979441997,
      "epsilon_dpo/beta": 0.0010598934022709727,
      "epsilon_dpo/beta_margin_grad_mean": -0.3548807203769684,
      "epsilon_dpo/beta_margin_grad_std": 0.19512853026390076,
      "epsilon_dpo/beta_margin_mean": 0.7214035391807556,
      "epsilon_dpo/beta_margin_std": 1.0699379444122314,
      "epsilon_dpo/loss_margin_mean": 686.6939697265625,
      "grad_norm": 76.16841888427734,
      "kl/avg_steps": 0.46875,
      "kl/beta": 0.0010647785384207964,
      "kl/n_epsilon_steps": 0.265625,
      "kl/p_epsilon_steps": 0.734375,
      "learning_rate": 3.159440233840763e-08,
      "logits/chosen": -3.5470635890960693,
      "logits/rejected": -4.620864391326904,
      "logps/chosen": -522.0079956054688,
      "logps/ref_chosen": -37.24636459350586,
      "logps/ref_rejected": -95.97724151611328,
      "logps/rejected": -1267.432861328125,
      "loss": 1.018,
      "rewards/accuracies": 0.75,
      "rewards/chosen": -0.5164851546287537,
      "rewards/margins": 0.7214035987854004,
      "rewards/rejected": -1.2378886938095093,
      "step": 583
    },
    {
      "epoch": 0.8575624082232012,
      "epsilon_dpo/beta": 0.0010506423423066735,
      "epsilon_dpo/beta_margin_grad_mean": -0.2919199466705322,
      "epsilon_dpo/beta_margin_grad_std": 0.1186746209859848,
      "epsilon_dpo/beta_margin_mean": 0.9708749651908875,
      "epsilon_dpo/beta_margin_std": 0.6488715410232544,
      "epsilon_dpo/loss_margin_mean": 925.4342041015625,
      "grad_norm": 45.21849060058594,
      "kl/avg_steps": 0.875,
      "kl/beta": 0.0010598106309771538,
      "kl/n_epsilon_steps": 0.0625,
      "kl/p_epsilon_steps": 0.9375,
      "learning_rate": 3.0972808389096635e-08,
      "logits/chosen": -3.285341262817383,
      "logits/rejected": -4.798908710479736,
      "logps/chosen": -381.122314453125,
      "logps/ref_chosen": -34.52136993408203,
      "logps/ref_rejected": -105.40646362304688,
      "logps/rejected": -1377.441650390625,
      "loss": 0.7209,
      "rewards/accuracies": 0.96875,
      "rewards/chosen": -0.3644617795944214,
      "rewards/margins": 0.9708749055862427,
      "rewards/rejected": -1.335336685180664,
      "step": 584
    },
    {
      "epoch": 0.8590308370044053,
      "epsilon_dpo/beta": 0.001042842399328947,
      "epsilon_dpo/beta_margin_grad_mean": -0.3161095976829529,
      "epsilon_dpo/beta_margin_grad_std": 0.1634971797466278,
      "epsilon_dpo/beta_margin_mean": 0.9011527895927429,
      "epsilon_dpo/beta_margin_std": 0.8913306593894958,
      "epsilon_dpo/loss_margin_mean": 867.5726928710938,
      "grad_norm": 49.96684265136719,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.001050617778673768,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 3.035698600998121e-08,
      "logits/chosen": -3.437217950820923,
      "logits/rejected": -4.798175811767578,
      "logps/chosen": -450.2691345214844,
      "logps/ref_chosen": -36.641475677490234,
      "logps/ref_rejected": -92.18727111816406,
      "logps/rejected": -1373.3876953125,
      "loss": 0.8314,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.4328582286834717,
      "rewards/margins": 0.9011527895927429,
      "rewards/rejected": -1.3340110778808594,
      "step": 585
    },
    {
      "epoch": 0.8604992657856094,
      "epsilon_dpo/beta": 0.0010367088252678514,
      "epsilon_dpo/beta_margin_grad_mean": -0.3300473392009735,
      "epsilon_dpo/beta_margin_grad_std": 0.16417358815670013,
      "epsilon_dpo/beta_margin_mean": 0.833533763885498,
      "epsilon_dpo/beta_margin_std": 0.8716885447502136,
      "epsilon_dpo/loss_margin_mean": 808.5335693359375,
      "grad_norm": 41.283748626708984,
      "kl/avg_steps": 0.59375,
      "kl/beta": 0.001042796764522791,
      "kl/n_epsilon_steps": 0.203125,
      "kl/p_epsilon_steps": 0.796875,
      "learning_rate": 2.974695142855388e-08,
      "logits/chosen": -3.5555217266082764,
      "logits/rejected": -4.804073333740234,
      "logps/chosen": -526.1253051757812,
      "logps/ref_chosen": -36.78833770751953,
      "logps/ref_rejected": -98.72267150878906,
      "logps/rejected": -1396.59326171875,
      "loss": 0.8691,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": -0.508251428604126,
      "rewards/margins": 0.833533763885498,
      "rewards/rejected": -1.341785192489624,
      "step": 586
    },
    {
      "epoch": 0.8619676945668135,
      "epsilon_dpo/beta": 0.0010302658192813396,
      "epsilon_dpo/beta_margin_grad_mean": -0.33965250849723816,
      "epsilon_dpo/beta_margin_grad_std": 0.15424403548240662,
      "epsilon_dpo/beta_margin_mean": 0.7954989075660706,
      "epsilon_dpo/beta_margin_std": 0.8702175617218018,
      "epsilon_dpo/loss_margin_mean": 775.6798706054688,
      "grad_norm": 43.92211151123047,
      "kl/avg_steps": 0.625,
      "kl/beta": 0.001036641770042479,
      "kl/n_epsilon_steps": 0.1875,
      "kl/p_epsilon_steps": 0.8125,
      "learning_rate": 2.9142720719793122e-08,
      "logits/chosen": -3.2840447425842285,
      "logits/rejected": -4.652719020843506,
      "logps/chosen": -320.67523193359375,
      "logps/ref_chosen": -27.06256103515625,
      "logps/ref_rejected": -87.31655883789062,
      "logps/rejected": -1156.609130859375,
      "loss": 0.8854,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": -0.30333322286605835,
      "rewards/margins": 0.7954989075660706,
      "rewards/rejected": -1.098832130432129,
      "step": 587
    },
    {
      "epoch": 0.8634361233480177,
      "epsilon_dpo/beta": 0.0010235445806756616,
      "epsilon_dpo/beta_margin_grad_mean": -0.3219730854034424,
      "epsilon_dpo/beta_margin_grad_std": 0.1715155392885208,
      "epsilon_dpo/beta_margin_mean": 0.8723087906837463,
      "epsilon_dpo/beta_margin_std": 0.9220552444458008,
      "epsilon_dpo/loss_margin_mean": 856.6797485351562,
      "grad_norm": 54.41194152832031,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.001030202955007553,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 2.8544309805740018e-08,
      "logits/chosen": -3.385420083999634,
      "logits/rejected": -4.716907978057861,
      "logps/chosen": -473.547607421875,
      "logps/ref_chosen": -32.258827209472656,
      "logps/ref_rejected": -112.76737976074219,
      "logps/rejected": -1410.73583984375,
      "loss": 0.8623,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.45345091819763184,
      "rewards/margins": 0.8723088502883911,
      "rewards/rejected": -1.3257596492767334,
      "step": 588
    },
    {
      "epoch": 0.8649045521292217,
      "epsilon_dpo/beta": 0.0010178310330957174,
      "epsilon_dpo/beta_margin_grad_mean": -0.3191934823989868,
      "epsilon_dpo/beta_margin_grad_std": 0.17622847855091095,
      "epsilon_dpo/beta_margin_mean": 0.9419023990631104,
      "epsilon_dpo/beta_margin_std": 1.028359293937683,
      "epsilon_dpo/loss_margin_mean": 930.8062744140625,
      "grad_norm": 57.298946380615234,
      "kl/avg_steps": 0.5625,
      "kl/beta": 0.001023486373014748,
      "kl/n_epsilon_steps": 0.21875,
      "kl/p_epsilon_steps": 0.78125,
      "learning_rate": 2.7951734455078786e-08,
      "logits/chosen": -3.4220762252807617,
      "logits/rejected": -4.901893138885498,
      "logps/chosen": -488.34820556640625,
      "logps/ref_chosen": -36.310951232910156,
      "logps/ref_rejected": -119.32230377197266,
      "logps/rejected": -1502.165771484375,
      "loss": 0.846,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": -0.46269693970680237,
      "rewards/margins": 0.9419023990631104,
      "rewards/rejected": -1.4045993089675903,
      "step": 589
    },
    {
      "epoch": 0.8663729809104258,
      "epsilon_dpo/beta": 0.0010111834853887558,
      "epsilon_dpo/beta_margin_grad_mean": -0.3087705075740814,
      "epsilon_dpo/beta_margin_grad_std": 0.15020431578159332,
      "epsilon_dpo/beta_margin_mean": 0.94552081823349,
      "epsilon_dpo/beta_margin_std": 0.8651568293571472,
      "epsilon_dpo/loss_margin_mean": 938.7587280273438,
      "grad_norm": 43.128292083740234,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.001017761416733265,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 2.736501028272095e-08,
      "logits/chosen": -3.2722253799438477,
      "logits/rejected": -4.7439470291137695,
      "logps/chosen": -384.21209716796875,
      "logps/ref_chosen": -31.54564094543457,
      "logps/ref_rejected": -112.30560302734375,
      "logps/rejected": -1403.730712890625,
      "loss": 0.7904,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.3574645221233368,
      "rewards/margins": 0.94552081823349,
      "rewards/rejected": -1.302985429763794,
      "step": 590
    },
    {
      "epoch": 0.8678414096916299,
      "epsilon_dpo/beta": 0.0010042748181149364,
      "epsilon_dpo/beta_margin_grad_mean": -0.3359038829803467,
      "epsilon_dpo/beta_margin_grad_std": 0.17261067032814026,
      "epsilon_dpo/beta_margin_mean": 0.8146455883979797,
      "epsilon_dpo/beta_margin_std": 0.9358741641044617,
      "epsilon_dpo/loss_margin_mean": 815.1807861328125,
      "grad_norm": 63.2326545715332,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0010111258598044515,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 2.678415274939408e-08,
      "logits/chosen": -3.32479190826416,
      "logits/rejected": -4.7041168212890625,
      "logps/chosen": -425.36102294921875,
      "logps/ref_chosen": -32.81976318359375,
      "logps/ref_rejected": -89.97746276855469,
      "logps/rejected": -1297.699462890625,
      "loss": 0.9023,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.39648881554603577,
      "rewards/margins": 0.814645528793335,
      "rewards/rejected": -1.211134433746338,
      "step": 591
    },
    {
      "epoch": 0.869309838472834,
      "epsilon_dpo/beta": 0.0009980452014133334,
      "epsilon_dpo/beta_margin_grad_mean": -0.33510512113571167,
      "epsilon_dpo/beta_margin_grad_std": 0.15517839789390564,
      "epsilon_dpo/beta_margin_mean": 0.7904956340789795,
      "epsilon_dpo/beta_margin_std": 0.8030670285224915,
      "epsilon_dpo/loss_margin_mean": 796.0484008789062,
      "grad_norm": 49.01154708862305,
      "kl/avg_steps": 0.625,
      "kl/beta": 0.001004221849143505,
      "kl/n_epsilon_steps": 0.1875,
      "kl/p_epsilon_steps": 0.8125,
      "learning_rate": 2.6209177161234442e-08,
      "logits/chosen": -3.4462766647338867,
      "logits/rejected": -4.8302812576293945,
      "logps/chosen": -423.3301696777344,
      "logps/ref_chosen": -31.739425659179688,
      "logps/ref_rejected": -81.3411865234375,
      "logps/rejected": -1268.98046875,
      "loss": 0.877,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.3929889500141144,
      "rewards/margins": 0.7904956340789795,
      "rewards/rejected": -1.1834845542907715,
      "step": 592
    },
    {
      "epoch": 0.8707782672540382,
      "epsilon_dpo/beta": 0.0009934057015925646,
      "epsilon_dpo/beta_margin_grad_mean": -0.3691524267196655,
      "epsilon_dpo/beta_margin_grad_std": 0.15739652514457703,
      "epsilon_dpo/beta_margin_mean": 0.611284613609314,
      "epsilon_dpo/beta_margin_std": 0.7558280825614929,
      "epsilon_dpo/loss_margin_mean": 619.8391723632812,
      "grad_norm": 50.16069412231445,
      "kl/avg_steps": 0.46875,
      "kl/beta": 0.0009979844326153398,
      "kl/n_epsilon_steps": 0.265625,
      "kl/p_epsilon_steps": 0.734375,
      "learning_rate": 2.564009866938349e-08,
      "logits/chosen": -3.442495822906494,
      "logits/rejected": -4.646967887878418,
      "logps/chosen": -388.3526611328125,
      "logps/ref_chosen": -33.75498962402344,
      "logps/ref_rejected": -66.24615478515625,
      "logps/rejected": -1040.68310546875,
      "loss": 0.9905,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": -0.35422414541244507,
      "rewards/margins": 0.611284613609314,
      "rewards/rejected": -0.965508759021759,
      "step": 593
    },
    {
      "epoch": 0.8722466960352423,
      "epsilon_dpo/beta": 0.0009878396522253752,
      "epsilon_dpo/beta_margin_grad_mean": -0.3453194499015808,
      "epsilon_dpo/beta_margin_grad_std": 0.15376316010951996,
      "epsilon_dpo/beta_margin_mean": 0.7299271821975708,
      "epsilon_dpo/beta_margin_std": 0.7822921276092529,
      "epsilon_dpo/loss_margin_mean": 743.34716796875,
      "grad_norm": 56.66592025756836,
      "kl/avg_steps": 0.5625,
      "kl/beta": 0.0009933282854035497,
      "kl/n_epsilon_steps": 0.21875,
      "kl/p_epsilon_steps": 0.78125,
      "learning_rate": 2.5076932269588708e-08,
      "logits/chosen": -3.383502960205078,
      "logits/rejected": -4.847668647766113,
      "logps/chosen": -447.0162353515625,
      "logps/ref_chosen": -34.56220245361328,
      "logps/ref_rejected": -92.779052734375,
      "logps/rejected": -1248.580322265625,
      "loss": 0.9111,
      "rewards/accuracies": 0.78125,
      "rewards/chosen": -0.40900394320487976,
      "rewards/margins": 0.7299271821975708,
      "rewards/rejected": -1.1389310359954834,
      "step": 594
    },
    {
      "epoch": 0.8737151248164464,
      "epsilon_dpo/beta": 0.0009810791816562414,
      "epsilon_dpo/beta_margin_grad_mean": -0.35142308473587036,
      "epsilon_dpo/beta_margin_grad_std": 0.14063841104507446,
      "epsilon_dpo/beta_margin_mean": 0.7219837307929993,
      "epsilon_dpo/beta_margin_std": 0.8028604388237,
      "epsilon_dpo/loss_margin_mean": 738.6736450195312,
      "grad_norm": 48.688880920410156,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0009877720149233937,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 2.451969280180849e-08,
      "logits/chosen": -3.31217098236084,
      "logits/rejected": -4.6105427742004395,
      "logps/chosen": -390.6729431152344,
      "logps/ref_chosen": -31.321365356445312,
      "logps/ref_rejected": -85.45933532714844,
      "logps/rejected": -1183.484619140625,
      "loss": 0.9122,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.3539902865886688,
      "rewards/margins": 0.7219836711883545,
      "rewards/rejected": -1.0759739875793457,
      "step": 595
    },
    {
      "epoch": 0.8751835535976505,
      "epsilon_dpo/beta": 0.0009740737732499838,
      "epsilon_dpo/beta_margin_grad_mean": -0.3495016098022461,
      "epsilon_dpo/beta_margin_grad_std": 0.13467830419540405,
      "epsilon_dpo/beta_margin_mean": 0.7045981287956238,
      "epsilon_dpo/beta_margin_std": 0.7173324227333069,
      "epsilon_dpo/loss_margin_mean": 726.050048828125,
      "grad_norm": 47.94287872314453,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0009810274932533503,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 2.396839494982103e-08,
      "logits/chosen": -3.158041000366211,
      "logits/rejected": -4.6905059814453125,
      "logps/chosen": -378.280517578125,
      "logps/ref_chosen": -30.307376861572266,
      "logps/ref_rejected": -88.15638732910156,
      "logps/rejected": -1162.1795654296875,
      "loss": 0.9049,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.3397039771080017,
      "rewards/margins": 0.7045981287956238,
      "rewards/rejected": -1.0443021059036255,
      "step": 596
    },
    {
      "epoch": 0.8766519823788547,
      "epsilon_dpo/beta": 0.0009677313501015306,
      "epsilon_dpo/beta_margin_grad_mean": -0.32157784700393677,
      "epsilon_dpo/beta_margin_grad_std": 0.17215141654014587,
      "epsilon_dpo/beta_margin_mean": 0.894204318523407,
      "epsilon_dpo/beta_margin_std": 0.9434449076652527,
      "epsilon_dpo/loss_margin_mean": 928.6572265625,
      "grad_norm": 41.356510162353516,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.0009740266250446439,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 2.3423053240837514e-08,
      "logits/chosen": -3.414703845977783,
      "logits/rejected": -4.802955627441406,
      "logps/chosen": -437.45770263671875,
      "logps/ref_chosen": -34.63069152832031,
      "logps/ref_rejected": -107.39651489257812,
      "logps/rejected": -1438.8807373046875,
      "loss": 0.8543,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.3919217586517334,
      "rewards/margins": 0.8942043781280518,
      "rewards/rejected": -1.2861260175704956,
      "step": 597
    },
    {
      "epoch": 0.8781204111600588,
      "epsilon_dpo/beta": 0.0009614220471121371,
      "epsilon_dpo/beta_margin_grad_mean": -0.34296542406082153,
      "epsilon_dpo/beta_margin_grad_std": 0.16591456532478333,
      "epsilon_dpo/beta_margin_mean": 0.7655530571937561,
      "epsilon_dpo/beta_margin_std": 0.8700756430625916,
      "epsilon_dpo/loss_margin_mean": 800.5443725585938,
      "grad_norm": 63.38301467895508,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.0009676762856543064,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 2.2883682045119062e-08,
      "logits/chosen": -3.349184989929199,
      "logits/rejected": -4.659879684448242,
      "logps/chosen": -458.43048095703125,
      "logps/ref_chosen": -38.996089935302734,
      "logps/ref_rejected": -97.60490417480469,
      "logps/rejected": -1317.583740234375,
      "loss": 0.9145,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.40632039308547974,
      "rewards/margins": 0.7655530571937561,
      "rewards/rejected": -1.1718735694885254,
      "step": 598
    },
    {
      "epoch": 0.8795888399412628,
      "epsilon_dpo/beta": 0.0009536515572108328,
      "epsilon_dpo/beta_margin_grad_mean": -0.34948039054870605,
      "epsilon_dpo/beta_margin_grad_std": 0.14211682975292206,
      "epsilon_dpo/beta_margin_mean": 0.6984655857086182,
      "epsilon_dpo/beta_margin_std": 0.7249807715415955,
      "epsilon_dpo/loss_margin_mean": 734.8526000976562,
      "grad_norm": 51.95116424560547,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0009613672737032175,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 2.2350295575598367e-08,
      "logits/chosen": -3.448948621749878,
      "logits/rejected": -4.653861045837402,
      "logps/chosen": -416.742431640625,
      "logps/ref_chosen": -33.18586730957031,
      "logps/ref_rejected": -88.57789611816406,
      "logps/rejected": -1206.987060546875,
      "loss": 0.915,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.3669903874397278,
      "rewards/margins": 0.6984655857086182,
      "rewards/rejected": -1.0654559135437012,
      "step": 599
    },
    {
      "epoch": 0.8810572687224669,
      "epsilon_dpo/beta": 0.0009474557591602206,
      "epsilon_dpo/beta_margin_grad_mean": -0.34580788016319275,
      "epsilon_dpo/beta_margin_grad_std": 0.13170255720615387,
      "epsilon_dpo/beta_margin_mean": 0.7138612270355225,
      "epsilon_dpo/beta_margin_std": 0.6741172075271606,
      "epsilon_dpo/loss_margin_mean": 756.361083984375,
      "grad_norm": 49.26176834106445,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.000953619135543704,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 2.1822907887504932e-08,
      "logits/chosen": -3.4161376953125,
      "logits/rejected": -4.741451263427734,
      "logps/chosen": -460.50970458984375,
      "logps/ref_chosen": -42.37248229980469,
      "logps/ref_rejected": -89.61239624023438,
      "logps/rejected": -1264.110595703125,
      "loss": 0.8895,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.39791154861450195,
      "rewards/margins": 0.7138612270355225,
      "rewards/rejected": -1.1117727756500244,
      "step": 600
    },
    {
      "epoch": 0.8810572687224669,
      "eval_epsilon_dpo/beta": 0.0009440449066460133,
      "eval_epsilon_dpo/beta_margin_grad_mean": -0.40722984075546265,
      "eval_epsilon_dpo/beta_margin_grad_std": 0.1766805797815323,
      "eval_epsilon_dpo/beta_margin_mean": 0.44606366753578186,
      "eval_epsilon_dpo/beta_margin_std": 0.8790761828422546,
      "eval_epsilon_dpo/loss_margin_mean": 478.8302001953125,
      "eval_kl/n_epsilon_steps": 0.31763699650764465,
      "eval_kl/p_epsilon_steps": 0.6819349527359009,
      "eval_logits/chosen": -3.8786203861236572,
      "eval_logits/rejected": -4.723351955413818,
      "eval_logps/chosen": -771.9093627929688,
      "eval_logps/ref_chosen": -68.29110717773438,
      "eval_logps/ref_rejected": -92.08038330078125,
      "eval_logps/rejected": -1274.5286865234375,
      "eval_loss": 0.5823308825492859,
      "eval_rewards/accuracies": 0.7037671208381653,
      "eval_rewards/chosen": -0.6674680709838867,
      "eval_rewards/margins": 0.44606366753578186,
      "eval_rewards/rejected": -1.1135317087173462,
      "eval_runtime": 38.382,
      "eval_samples_per_second": 60.94,
      "eval_steps_per_second": 1.928,
      "step": 600
    },
    {
      "epoch": 0.882525697503671,
      "epsilon_dpo/beta": 0.0009403903386555612,
      "epsilon_dpo/beta_margin_grad_mean": -0.32094281911849976,
      "epsilon_dpo/beta_margin_grad_std": 0.16856704652309418,
      "epsilon_dpo/beta_margin_mean": 0.8789842128753662,
      "epsilon_dpo/beta_margin_std": 0.901688814163208,
      "epsilon_dpo/loss_margin_mean": 938.4696044921875,
      "grad_norm": 48.359046936035156,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0009474018006585538,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 2.1301532877994742e-08,
      "logits/chosen": -3.238035202026367,
      "logits/rejected": -4.882007122039795,
      "logps/chosen": -406.67852783203125,
      "logps/ref_chosen": -34.16810607910156,
      "logps/ref_rejected": -99.91683959960938,
      "logps/rejected": -1410.8968505859375,
      "loss": 0.852,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.35252028703689575,
      "rewards/margins": 0.8789842128753662,
      "rewards/rejected": -1.2315044403076172,
      "step": 601
    },
    {
      "epoch": 0.8839941262848752,
      "epsilon_dpo/beta": 0.0009328021551482379,
      "epsilon_dpo/beta_margin_grad_mean": -0.3170680105686188,
      "epsilon_dpo/beta_margin_grad_std": 0.13237425684928894,
      "epsilon_dpo/beta_margin_mean": 0.8696146011352539,
      "epsilon_dpo/beta_margin_std": 0.728697657585144,
      "epsilon_dpo/loss_margin_mean": 934.039306640625,
      "grad_norm": 48.63134765625,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0009403491858392954,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 2.0786184285784298e-08,
      "logits/chosen": -3.2103443145751953,
      "logits/rejected": -4.763635635375977,
      "logps/chosen": -303.0639343261719,
      "logps/ref_chosen": -34.405120849609375,
      "logps/ref_rejected": -93.47988891601562,
      "logps/rejected": -1296.177978515625,
      "loss": 0.8003,
      "rewards/accuracies": 0.953125,
      "rewards/chosen": -0.25088733434677124,
      "rewards/margins": 0.8696146011352539,
      "rewards/rejected": -1.12050199508667,
      "step": 602
    },
    {
      "epoch": 0.8854625550660793,
      "epsilon_dpo/beta": 0.0009261587401852012,
      "epsilon_dpo/beta_margin_grad_mean": -0.3474172055721283,
      "epsilon_dpo/beta_margin_grad_std": 0.14739488065242767,
      "epsilon_dpo/beta_margin_mean": 0.7221988439559937,
      "epsilon_dpo/beta_margin_std": 0.7671189904212952,
      "epsilon_dpo/loss_margin_mean": 782.9607543945312,
      "grad_norm": 51.643531799316406,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0009327704319730401,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 2.0276875690788204e-08,
      "logits/chosen": -3.2944273948669434,
      "logits/rejected": -4.612957000732422,
      "logps/chosen": -470.8554992675781,
      "logps/ref_chosen": -53.07399368286133,
      "logps/ref_rejected": -107.52302551269531,
      "logps/rejected": -1308.265380859375,
      "loss": 0.9099,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.38833025097846985,
      "rewards/margins": 0.7221988439559937,
      "rewards/rejected": -1.1105290651321411,
      "step": 603
    },
    {
      "epoch": 0.8869309838472834,
      "epsilon_dpo/beta": 0.0009189706179313362,
      "epsilon_dpo/beta_margin_grad_mean": -0.3316219747066498,
      "epsilon_dpo/beta_margin_grad_std": 0.14688213169574738,
      "epsilon_dpo/beta_margin_mean": 0.8516996502876282,
      "epsilon_dpo/beta_margin_std": 0.9336607456207275,
      "epsilon_dpo/loss_margin_mean": 929.5543823242188,
      "grad_norm": 44.274085998535156,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0009261139784939587,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 1.977362051376158e-08,
      "logits/chosen": -3.4287848472595215,
      "logits/rejected": -4.8621602058410645,
      "logps/chosen": -422.5206298828125,
      "logps/ref_chosen": -32.21878433227539,
      "logps/ref_rejected": -99.47515106201172,
      "logps/rejected": -1419.331298828125,
      "loss": 0.8561,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.3594130277633667,
      "rewards/margins": 0.8516995906829834,
      "rewards/rejected": -1.2111127376556396,
      "step": 604
    },
    {
      "epoch": 0.8883994126284875,
      "epsilon_dpo/beta": 0.0009129956015385687,
      "epsilon_dpo/beta_margin_grad_mean": -0.33933547139167786,
      "epsilon_dpo/beta_margin_grad_std": 0.15050913393497467,
      "epsilon_dpo/beta_margin_mean": 0.7814961075782776,
      "epsilon_dpo/beta_margin_std": 0.8144083619117737,
      "epsilon_dpo/loss_margin_mean": 859.486328125,
      "grad_norm": 41.927146911621094,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.0009189348202198744,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 1.9276432015946446e-08,
      "logits/chosen": -3.3077027797698975,
      "logits/rejected": -4.713131904602051,
      "logps/chosen": -402.01544189453125,
      "logps/ref_chosen": -42.914276123046875,
      "logps/ref_rejected": -108.40269470214844,
      "logps/rejected": -1326.990234375,
      "loss": 0.882,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.3291019797325134,
      "rewards/margins": 0.7814960479736328,
      "rewards/rejected": -1.110598087310791,
      "step": 605
    },
    {
      "epoch": 0.8898678414096917,
      "epsilon_dpo/beta": 0.0009078990551643074,
      "epsilon_dpo/beta_margin_grad_mean": -0.3660072982311249,
      "epsilon_dpo/beta_margin_grad_std": 0.15605977177619934,
      "epsilon_dpo/beta_margin_mean": 0.618130087852478,
      "epsilon_dpo/beta_margin_std": 0.7797470688819885,
      "epsilon_dpo/loss_margin_mean": 685.6812133789062,
      "grad_norm": 57.995399475097656,
      "kl/avg_steps": 0.5625,
      "kl/beta": 0.000912943622097373,
      "kl/n_epsilon_steps": 0.21875,
      "kl/p_epsilon_steps": 0.78125,
      "learning_rate": 1.8785323298722093e-08,
      "logits/chosen": -3.486124277114868,
      "logits/rejected": -4.584151268005371,
      "logps/chosen": -506.50384521484375,
      "logps/ref_chosen": -37.19722366333008,
      "logps/ref_rejected": -102.87519836425781,
      "logps/rejected": -1257.863037109375,
      "loss": 0.9904,
      "rewards/accuracies": 0.828125,
      "rewards/chosen": -0.4276297986507416,
      "rewards/margins": 0.618130087852478,
      "rewards/rejected": -1.045759916305542,
      "step": 606
    },
    {
      "epoch": 0.8913362701908958,
      "epsilon_dpo/beta": 0.000903104490134865,
      "epsilon_dpo/beta_margin_grad_mean": -0.36472412943840027,
      "epsilon_dpo/beta_margin_grad_std": 0.16157901287078857,
      "epsilon_dpo/beta_margin_mean": 0.6311452984809875,
      "epsilon_dpo/beta_margin_std": 0.7819616198539734,
      "epsilon_dpo/loss_margin_mean": 704.2265014648438,
      "grad_norm": 55.2100830078125,
      "kl/avg_steps": 0.53125,
      "kl/beta": 0.0009078370640054345,
      "kl/n_epsilon_steps": 0.234375,
      "kl/p_epsilon_steps": 0.765625,
      "learning_rate": 1.8300307303259904e-08,
      "logits/chosen": -3.4703445434570312,
      "logits/rejected": -4.856854438781738,
      "logps/chosen": -499.9947509765625,
      "logps/ref_chosen": -43.06529235839844,
      "logps/ref_rejected": -84.84536743164062,
      "logps/rejected": -1246.0013427734375,
      "loss": 0.9843,
      "rewards/accuracies": 0.796875,
      "rewards/chosen": -0.4154084026813507,
      "rewards/margins": 0.6311452984809875,
      "rewards/rejected": -1.046553611755371,
      "step": 607
    },
    {
      "epoch": 0.8928046989720999,
      "epsilon_dpo/beta": 0.0008960742270573974,
      "epsilon_dpo/beta_margin_grad_mean": -0.3294691741466522,
      "epsilon_dpo/beta_margin_grad_std": 0.13619713485240936,
      "epsilon_dpo/beta_margin_mean": 0.8008559942245483,
      "epsilon_dpo/beta_margin_std": 0.7176367044448853,
      "epsilon_dpo/loss_margin_mean": 896.19287109375,
      "grad_norm": 60.484779357910156,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0009030396468006074,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 1.7821396810182437e-08,
      "logits/chosen": -3.1860623359680176,
      "logits/rejected": -4.672609806060791,
      "logps/chosen": -324.84686279296875,
      "logps/ref_chosen": -27.870777130126953,
      "logps/ref_rejected": -101.65553283691406,
      "logps/rejected": -1294.824462890625,
      "loss": 0.843,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.2668021619319916,
      "rewards/margins": 0.8008559942245483,
      "rewards/rejected": -1.0676581859588623,
      "step": 608
    },
    {
      "epoch": 0.8942731277533039,
      "epsilon_dpo/beta": 0.0008896880317479372,
      "epsilon_dpo/beta_margin_grad_mean": -0.3255755305290222,
      "epsilon_dpo/beta_margin_grad_std": 0.13147076964378357,
      "epsilon_dpo/beta_margin_mean": 0.8107938766479492,
      "epsilon_dpo/beta_margin_std": 0.689571738243103,
      "epsilon_dpo/loss_margin_mean": 914.2926025390625,
      "grad_norm": 52.808658599853516,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0008960393606685102,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 1.7348604439226617e-08,
      "logits/chosen": -3.266610860824585,
      "logits/rejected": -4.786360740661621,
      "logps/chosen": -432.412841796875,
      "logps/ref_chosen": -33.51665496826172,
      "logps/ref_rejected": -96.93180084228516,
      "logps/rejected": -1410.12060546875,
      "loss": 0.829,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.3561561107635498,
      "rewards/margins": 0.8107938766479492,
      "rewards/rejected": -1.166949987411499,
      "step": 609
    },
    {
      "epoch": 0.895741556534508,
      "epsilon_dpo/beta": 0.0008841731469146907,
      "epsilon_dpo/beta_margin_grad_mean": -0.34196192026138306,
      "epsilon_dpo/beta_margin_grad_std": 0.17104457318782806,
      "epsilon_dpo/beta_margin_mean": 0.7602141499519348,
      "epsilon_dpo/beta_margin_std": 0.892071008682251,
      "epsilon_dpo/loss_margin_mean": 865.0516967773438,
      "grad_norm": 50.597049713134766,
      "kl/avg_steps": 0.625,
      "kl/beta": 0.0008896450162865222,
      "kl/n_epsilon_steps": 0.1875,
      "kl/p_epsilon_steps": 0.8125,
      "learning_rate": 1.6881942648911074e-08,
      "logits/chosen": -3.319277286529541,
      "logits/rejected": -4.948149681091309,
      "logps/chosen": -423.406982421875,
      "logps/ref_chosen": -39.733856201171875,
      "logps/ref_rejected": -88.57766723632812,
      "logps/rejected": -1337.302490234375,
      "loss": 0.9283,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.34177446365356445,
      "rewards/margins": 0.7602142095565796,
      "rewards/rejected": -1.1019885540008545,
      "step": 610
    },
    {
      "epoch": 0.8972099853157122,
      "epsilon_dpo/beta": 0.0008777128532528877,
      "epsilon_dpo/beta_margin_grad_mean": -0.3441890478134155,
      "epsilon_dpo/beta_margin_grad_std": 0.1642007678747177,
      "epsilon_dpo/beta_margin_mean": 0.7517509460449219,
      "epsilon_dpo/beta_margin_std": 0.8516819477081299,
      "epsilon_dpo/loss_margin_mean": 860.3085327148438,
      "grad_norm": 54.84195327758789,
      "kl/avg_steps": 0.734375,
      "kl/beta": 0.0008841192466206849,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 1.6421423736208e-08,
      "logits/chosen": -3.4497299194335938,
      "logits/rejected": -4.820193290710449,
      "logps/chosen": -460.02740478515625,
      "logps/ref_chosen": -34.78019332885742,
      "logps/ref_rejected": -90.61834716796875,
      "logps/rejected": -1376.174072265625,
      "loss": 0.9188,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.3749154508113861,
      "rewards/margins": 0.7517509460449219,
      "rewards/rejected": -1.1266663074493408,
      "step": 611
    },
    {
      "epoch": 0.8986784140969163,
      "epsilon_dpo/beta": 0.0008717270102351904,
      "epsilon_dpo/beta_margin_grad_mean": -0.33847570419311523,
      "epsilon_dpo/beta_margin_grad_std": 0.15062828361988068,
      "epsilon_dpo/beta_margin_mean": 0.7811422348022461,
      "epsilon_dpo/beta_margin_std": 0.8147025108337402,
      "epsilon_dpo/loss_margin_mean": 899.7782592773438,
      "grad_norm": 41.552490234375,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0008776738541200757,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 1.5967059836219042e-08,
      "logits/chosen": -3.163311004638672,
      "logits/rejected": -4.886482238769531,
      "logps/chosen": -380.4996337890625,
      "logps/ref_chosen": -35.333831787109375,
      "logps/ref_rejected": -93.14432525634766,
      "logps/rejected": -1338.08837890625,
      "loss": 0.8826,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.3019838333129883,
      "rewards/margins": 0.7811422348022461,
      "rewards/rejected": -1.0831260681152344,
      "step": 612
    },
    {
      "epoch": 0.9001468428781204,
      "epsilon_dpo/beta": 0.0008652299293316901,
      "epsilon_dpo/beta_margin_grad_mean": -0.3142598271369934,
      "epsilon_dpo/beta_margin_grad_std": 0.12845391035079956,
      "epsilon_dpo/beta_margin_mean": 0.8559463620185852,
      "epsilon_dpo/beta_margin_std": 0.6566179990768433,
      "epsilon_dpo/loss_margin_mean": 992.203369140625,
      "grad_norm": 61.14183044433594,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0008716810261830688,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 1.551886292185553e-08,
      "logits/chosen": -3.458996057510376,
      "logits/rejected": -4.9003472328186035,
      "logps/chosen": -418.6805114746094,
      "logps/ref_chosen": -36.464019775390625,
      "logps/ref_rejected": -113.0091781616211,
      "logps/rejected": -1487.4290771484375,
      "loss": 0.7942,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.33134663105010986,
      "rewards/margins": 0.8559463620185852,
      "rewards/rejected": -1.1872930526733398,
      "step": 613
    },
    {
      "epoch": 0.9016152716593245,
      "epsilon_dpo/beta": 0.0008593298261985183,
      "epsilon_dpo/beta_margin_grad_mean": -0.3246811330318451,
      "epsilon_dpo/beta_margin_grad_std": 0.15554648637771606,
      "epsilon_dpo/beta_margin_mean": 0.8337700366973877,
      "epsilon_dpo/beta_margin_std": 0.7923649549484253,
      "epsilon_dpo/loss_margin_mean": 974.2940063476562,
      "grad_norm": 48.429771423339844,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0008651920943520963,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 1.507684480352292e-08,
      "logits/chosen": -3.3637969493865967,
      "logits/rejected": -4.854522705078125,
      "logps/chosen": -424.6036376953125,
      "logps/ref_chosen": -34.81976318359375,
      "logps/ref_rejected": -111.12577819824219,
      "logps/rejected": -1475.20361328125,
      "loss": 0.8482,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.3362823724746704,
      "rewards/margins": 0.8337700366973877,
      "rewards/rejected": -1.1700525283813477,
      "step": 614
    },
    {
      "epoch": 0.9030837004405287,
      "epsilon_dpo/beta": 0.0008531937492080033,
      "epsilon_dpo/beta_margin_grad_mean": -0.3395027816295624,
      "epsilon_dpo/beta_margin_grad_std": 0.15511348843574524,
      "epsilon_dpo/beta_margin_mean": 0.7793731689453125,
      "epsilon_dpo/beta_margin_std": 0.8427423238754272,
      "epsilon_dpo/loss_margin_mean": 917.2001342773438,
      "grad_norm": 52.556758880615234,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0008592845406383276,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 1.4641017128809801e-08,
      "logits/chosen": -3.465533971786499,
      "logits/rejected": -4.829683303833008,
      "logps/chosen": -447.3005676269531,
      "logps/ref_chosen": -41.42036819458008,
      "logps/ref_rejected": -101.49702453613281,
      "logps/rejected": -1424.577392578125,
      "loss": 0.8908,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.3474053144454956,
      "rewards/margins": 0.7793731689453125,
      "rewards/rejected": -1.126778483390808,
      "step": 615
    },
    {
      "epoch": 0.9045521292217328,
      "epsilon_dpo/beta": 0.000848971598315984,
      "epsilon_dpo/beta_margin_grad_mean": -0.3760168254375458,
      "epsilon_dpo/beta_margin_grad_std": 0.1653953492641449,
      "epsilon_dpo/beta_margin_mean": 0.5968384146690369,
      "epsilon_dpo/beta_margin_std": 0.8491699695587158,
      "epsilon_dpo/loss_margin_mean": 708.5220336914062,
      "grad_norm": 59.032936096191406,
      "kl/avg_steps": 0.5,
      "kl/beta": 0.0008531524799764156,
      "kl/n_epsilon_steps": 0.25,
      "kl/p_epsilon_steps": 0.75,
      "learning_rate": 1.4211391382180637e-08,
      "logits/chosen": -3.490535259246826,
      "logits/rejected": -4.813992500305176,
      "logps/chosen": -511.3292236328125,
      "logps/ref_chosen": -45.615753173828125,
      "logps/ref_rejected": -80.37959289550781,
      "logps/rejected": -1254.6151123046875,
      "loss": 1.0265,
      "rewards/accuracies": 0.828125,
      "rewards/chosen": -0.3976947069168091,
      "rewards/margins": 0.5968384146690369,
      "rewards/rejected": -0.9945331811904907,
      "step": 616
    },
    {
      "epoch": 0.9060205580029369,
      "epsilon_dpo/beta": 0.0008452784968540072,
      "epsilon_dpo/beta_margin_grad_mean": -0.3991197347640991,
      "epsilon_dpo/beta_margin_grad_std": 0.14801205694675446,
      "epsilon_dpo/beta_margin_mean": 0.4730072617530823,
      "epsilon_dpo/beta_margin_std": 0.7559821605682373,
      "epsilon_dpo/loss_margin_mean": 564.623291015625,
      "grad_norm": 69.23668670654297,
      "kl/avg_steps": 0.4375,
      "kl/beta": 0.0008489079191349447,
      "kl/n_epsilon_steps": 0.28125,
      "kl/p_epsilon_steps": 0.71875,
      "learning_rate": 1.378797888467345e-08,
      "logits/chosen": -3.5469822883605957,
      "logits/rejected": -4.843328475952148,
      "logps/chosen": -583.8492431640625,
      "logps/ref_chosen": -50.210060119628906,
      "logps/ref_rejected": -69.55174255371094,
      "logps/rejected": -1167.814208984375,
      "loss": 1.088,
      "rewards/accuracies": 0.78125,
      "rewards/chosen": -0.4543150067329407,
      "rewards/margins": 0.4730072617530823,
      "rewards/rejected": -0.927322268486023,
      "step": 617
    },
    {
      "epoch": 0.9074889867841409,
      "epsilon_dpo/beta": 0.0008396140765398741,
      "epsilon_dpo/beta_margin_grad_mean": -0.35606545209884644,
      "epsilon_dpo/beta_margin_grad_std": 0.17117975652217865,
      "epsilon_dpo/beta_margin_mean": 0.6664056181907654,
      "epsilon_dpo/beta_margin_std": 0.8637328147888184,
      "epsilon_dpo/loss_margin_mean": 798.8843383789062,
      "grad_norm": 48.00564193725586,
      "kl/avg_steps": 0.671875,
      "kl/beta": 0.0008452101610600948,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 1.3370790793601371e-08,
      "logits/chosen": -3.4216156005859375,
      "logits/rejected": -4.760526657104492,
      "logps/chosen": -533.0503540039062,
      "logps/ref_chosen": -43.185306549072266,
      "logps/ref_rejected": -98.49762725830078,
      "logps/rejected": -1387.2470703125,
      "loss": 0.9866,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.41405919194221497,
      "rewards/margins": 0.6664056181907654,
      "rewards/rejected": -1.0804648399353027,
      "step": 618
    },
    {
      "epoch": 0.908957415565345,
      "epsilon_dpo/beta": 0.0008336182800121605,
      "epsilon_dpo/beta_margin_grad_mean": -0.34246736764907837,
      "epsilon_dpo/beta_margin_grad_std": 0.14548242092132568,
      "epsilon_dpo/beta_margin_mean": 0.7482366561889648,
      "epsilon_dpo/beta_margin_std": 0.7654976844787598,
      "epsilon_dpo/loss_margin_mean": 900.8438110351562,
      "grad_norm": 55.89384460449219,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0008395693148486316,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 1.2959838102258535e-08,
      "logits/chosen": -3.394374370574951,
      "logits/rejected": -4.941725254058838,
      "logps/chosen": -482.2840576171875,
      "logps/ref_chosen": -33.69963836669922,
      "logps/ref_rejected": -101.2354736328125,
      "logps/rejected": -1450.6636962890625,
      "loss": 0.891,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.37517303228378296,
      "rewards/margins": 0.7482366561889648,
      "rewards/rejected": -1.1234097480773926,
      "step": 619
    },
    {
      "epoch": 0.9104258443465492,
      "epsilon_dpo/beta": 0.0008292324491776526,
      "epsilon_dpo/beta_margin_grad_mean": -0.37484031915664673,
      "epsilon_dpo/beta_margin_grad_std": 0.15843062102794647,
      "epsilon_dpo/beta_margin_mean": 0.5776193737983704,
      "epsilon_dpo/beta_margin_std": 0.8028265833854675,
      "epsilon_dpo/loss_margin_mean": 701.974365234375,
      "grad_norm": 51.175804138183594,
      "kl/avg_steps": 0.53125,
      "kl/beta": 0.0008335779421031475,
      "kl/n_epsilon_steps": 0.234375,
      "kl/p_epsilon_steps": 0.765625,
      "learning_rate": 1.2555131639630567e-08,
      "logits/chosen": -3.558187961578369,
      "logits/rejected": -4.871739387512207,
      "logps/chosen": -579.7201538085938,
      "logps/ref_chosen": -42.774513244628906,
      "logps/ref_rejected": -84.47439575195312,
      "logps/rejected": -1323.3944091796875,
      "loss": 1.0273,
      "rewards/accuracies": 0.796875,
      "rewards/chosen": -0.4475584030151367,
      "rewards/margins": 0.5776193737983704,
      "rewards/rejected": -1.0251778364181519,
      "step": 620
    },
    {
      "epoch": 0.9118942731277533,
      "epsilon_dpo/beta": 0.0008243321790359914,
      "epsilon_dpo/beta_margin_grad_mean": -0.3368692100048065,
      "epsilon_dpo/beta_margin_grad_std": 0.16272993385791779,
      "epsilon_dpo/beta_margin_mean": 0.794694185256958,
      "epsilon_dpo/beta_margin_std": 0.8520438075065613,
      "epsilon_dpo/loss_margin_mean": 969.4013061523438,
      "grad_norm": 52.058837890625,
      "kl/avg_steps": 0.59375,
      "kl/beta": 0.000829172960948199,
      "kl/n_epsilon_steps": 0.203125,
      "kl/p_epsilon_steps": 0.796875,
      "learning_rate": 1.2156682070109086e-08,
      "logits/chosen": -3.53486967086792,
      "logits/rejected": -5.024101257324219,
      "logps/chosen": -459.840576171875,
      "logps/ref_chosen": -37.82067108154297,
      "logps/ref_rejected": -94.49537658691406,
      "logps/rejected": -1485.9166259765625,
      "loss": 0.8891,
      "rewards/accuracies": 0.828125,
      "rewards/chosen": -0.3498554825782776,
      "rewards/margins": 0.794694185256958,
      "rewards/rejected": -1.1445496082305908,
      "step": 621
    },
    {
      "epoch": 0.9133627019089574,
      "epsilon_dpo/beta": 0.0008184361504390836,
      "epsilon_dpo/beta_margin_grad_mean": -0.3743629455566406,
      "epsilon_dpo/beta_margin_grad_std": 0.13544589281082153,
      "epsilon_dpo/beta_margin_mean": 0.581274688243866,
      "epsilon_dpo/beta_margin_std": 0.6987016797065735,
      "epsilon_dpo/loss_margin_mean": 713.4944458007812,
      "grad_norm": 49.092411041259766,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0008242788026109338,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 1.1764499893210878e-08,
      "logits/chosen": -3.374379873275757,
      "logits/rejected": -4.7842888832092285,
      "logps/chosen": -553.1505126953125,
      "logps/ref_chosen": -39.961334228515625,
      "logps/ref_rejected": -92.28267669677734,
      "logps/rejected": -1318.96630859375,
      "loss": 0.9872,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.4212498664855957,
      "rewards/margins": 0.5812746286392212,
      "rewards/rejected": -1.0025246143341064,
      "step": 622
    },
    {
      "epoch": 0.9148311306901615,
      "epsilon_dpo/beta": 0.000813874474260956,
      "epsilon_dpo/beta_margin_grad_mean": -0.3725954592227936,
      "epsilon_dpo/beta_margin_grad_std": 0.1656135767698288,
      "epsilon_dpo/beta_margin_mean": 0.5889464020729065,
      "epsilon_dpo/beta_margin_std": 0.8087703585624695,
      "epsilon_dpo/loss_margin_mean": 729.1351928710938,
      "grad_norm": 50.80109405517578,
      "kl/avg_steps": 0.5625,
      "kl/beta": 0.0008183965692296624,
      "kl/n_epsilon_steps": 0.21875,
      "kl/p_epsilon_steps": 0.78125,
      "learning_rate": 1.1378595443300998e-08,
      "logits/chosen": -3.542423963546753,
      "logits/rejected": -4.91584587097168,
      "logps/chosen": -577.177001953125,
      "logps/ref_chosen": -49.0926513671875,
      "logps/ref_rejected": -91.09358215332031,
      "logps/rejected": -1348.3131103515625,
      "loss": 1.0243,
      "rewards/accuracies": 0.78125,
      "rewards/chosen": -0.432176798582077,
      "rewards/margins": 0.5889463424682617,
      "rewards/rejected": -1.0211231708526611,
      "step": 623
    },
    {
      "epoch": 0.9162995594713657,
      "epsilon_dpo/beta": 0.000808813376352191,
      "epsilon_dpo/beta_margin_grad_mean": -0.34141990542411804,
      "epsilon_dpo/beta_margin_grad_std": 0.16724152863025665,
      "epsilon_dpo/beta_margin_mean": 0.7709789872169495,
      "epsilon_dpo/beta_margin_std": 0.8832123875617981,
      "epsilon_dpo/loss_margin_mean": 958.7295532226562,
      "grad_norm": 48.523738861083984,
      "kl/avg_steps": 0.625,
      "kl/beta": 0.0008138188859447837,
      "kl/n_epsilon_steps": 0.1875,
      "kl/p_epsilon_steps": 0.8125,
      "learning_rate": 1.0998978889320582e-08,
      "logits/chosen": -3.3430838584899902,
      "logits/rejected": -4.918627738952637,
      "logps/chosen": -515.30517578125,
      "logps/ref_chosen": -46.57392501831055,
      "logps/ref_rejected": -105.08536529541016,
      "logps/rejected": -1532.546142578125,
      "loss": 0.9153,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.3814919590950012,
      "rewards/margins": 0.7709789872169495,
      "rewards/rejected": -1.1524710655212402,
      "step": 624
    },
    {
      "epoch": 0.9177679882525698,
      "epsilon_dpo/beta": 0.0008035369100980461,
      "epsilon_dpo/beta_margin_grad_mean": -0.34749388694763184,
      "epsilon_dpo/beta_margin_grad_std": 0.13651293516159058,
      "epsilon_dpo/beta_margin_mean": 0.6988623738288879,
      "epsilon_dpo/beta_margin_std": 0.6729094386100769,
      "epsilon_dpo/loss_margin_mean": 873.6240844726562,
      "grad_norm": 49.30683135986328,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.0008087640744633973,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 1.0625660234518913e-08,
      "logits/chosen": -3.506286144256592,
      "logits/rejected": -4.933239936828613,
      "logps/chosen": -515.078857421875,
      "logps/ref_chosen": -43.60509490966797,
      "logps/ref_rejected": -92.33833312988281,
      "logps/rejected": -1437.4361572265625,
      "loss": 0.902,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.38019776344299316,
      "rewards/margins": 0.6988623142242432,
      "rewards/rejected": -1.0790600776672363,
      "step": 625
    },
    {
      "epoch": 0.9192364170337739,
      "epsilon_dpo/beta": 0.0007988003198988736,
      "epsilon_dpo/beta_margin_grad_mean": -0.3574766218662262,
      "epsilon_dpo/beta_margin_grad_std": 0.16741888225078583,
      "epsilon_dpo/beta_margin_mean": 0.6738518476486206,
      "epsilon_dpo/beta_margin_std": 0.8334734439849854,
      "epsilon_dpo/loss_margin_mean": 849.4379272460938,
      "grad_norm": 50.068603515625,
      "kl/avg_steps": 0.59375,
      "kl/beta": 0.0008034911588765681,
      "kl/n_epsilon_steps": 0.203125,
      "kl/p_epsilon_steps": 0.796875,
      "learning_rate": 1.0258649316189721e-08,
      "logits/chosen": -3.526883363723755,
      "logits/rejected": -4.905704021453857,
      "logps/chosen": -593.0721435546875,
      "logps/ref_chosen": -50.95122528076172,
      "logps/ref_rejected": -103.29271697998047,
      "logps/rejected": -1494.8515625,
      "loss": 0.9691,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": -0.43541914224624634,
      "rewards/margins": 0.6738518476486206,
      "rewards/rejected": -1.1092710494995117,
      "step": 626
    },
    {
      "epoch": 0.920704845814978,
      "epsilon_dpo/beta": 0.00079333659959957,
      "epsilon_dpo/beta_margin_grad_mean": -0.32665178179740906,
      "epsilon_dpo/beta_margin_grad_std": 0.14166025817394257,
      "epsilon_dpo/beta_margin_mean": 0.8340927958488464,
      "epsilon_dpo/beta_margin_std": 0.7881202101707458,
      "epsilon_dpo/loss_margin_mean": 1055.364501953125,
      "grad_norm": 55.83228302001953,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0007987486314959824,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 9.897955805412e-09,
      "logits/chosen": -3.3846006393432617,
      "logits/rejected": -4.88566780090332,
      "logps/chosen": -337.388427734375,
      "logps/ref_chosen": -28.80577850341797,
      "logps/ref_rejected": -114.96311950683594,
      "logps/rejected": -1478.910400390625,
      "loss": 0.8381,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.24587570130825043,
      "rewards/margins": 0.8340927362442017,
      "rewards/rejected": -1.0799684524536133,
      "step": 627
    },
    {
      "epoch": 0.922173274596182,
      "epsilon_dpo/beta": 0.0007881674682721496,
      "epsilon_dpo/beta_margin_grad_mean": -0.34786510467529297,
      "epsilon_dpo/beta_margin_grad_std": 0.15347820520401,
      "epsilon_dpo/beta_margin_mean": 0.7319678068161011,
      "epsilon_dpo/beta_margin_std": 0.8163183331489563,
      "epsilon_dpo/loss_margin_mean": 933.1701049804688,
      "grad_norm": 49.38209915161133,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.0007932946900837123,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 9.543589206795238e-09,
      "logits/chosen": -3.572722911834717,
      "logits/rejected": -4.90597677230835,
      "logps/chosen": -581.771484375,
      "logps/ref_chosen": -45.28186798095703,
      "logps/ref_rejected": -108.524169921875,
      "logps/rejected": -1578.183837890625,
      "loss": 0.9161,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.42482954263687134,
      "rewards/margins": 0.7319678068161011,
      "rewards/rejected": -1.1567974090576172,
      "step": 628
    },
    {
      "epoch": 0.9236417033773862,
      "epsilon_dpo/beta": 0.00078401411883533,
      "epsilon_dpo/beta_margin_grad_mean": -0.34917938709259033,
      "epsilon_dpo/beta_margin_grad_std": 0.14244189858436584,
      "epsilon_dpo/beta_margin_mean": 0.7041851878166199,
      "epsilon_dpo/beta_margin_std": 0.7174273133277893,
      "epsilon_dpo/loss_margin_mean": 903.3053588867188,
      "grad_norm": 46.65474319458008,
      "kl/avg_steps": 0.53125,
      "kl/beta": 0.0007881226483732462,
      "kl/n_epsilon_steps": 0.234375,
      "kl/p_epsilon_steps": 0.765625,
      "learning_rate": 9.19555885822887e-09,
      "logits/chosen": -3.379412889480591,
      "logits/rejected": -4.922481536865234,
      "logps/chosen": -457.57720947265625,
      "logps/ref_chosen": -41.636070251464844,
      "logps/ref_rejected": -96.60995483398438,
      "logps/rejected": -1415.8564453125,
      "loss": 0.9092,
      "rewards/accuracies": 0.796875,
      "rewards/chosen": -0.3272828459739685,
      "rewards/margins": 0.7041851282119751,
      "rewards/rejected": -1.0314680337905884,
      "step": 629
    },
    {
      "epoch": 0.9251101321585903,
      "epsilon_dpo/beta": 0.0007786460919305682,
      "epsilon_dpo/beta_margin_grad_mean": -0.3540729284286499,
      "epsilon_dpo/beta_margin_grad_std": 0.13038687407970428,
      "epsilon_dpo/beta_margin_mean": 0.6700939536094666,
      "epsilon_dpo/beta_margin_std": 0.6649956703186035,
      "epsilon_dpo/loss_margin_mean": 864.0321655273438,
      "grad_norm": 40.582828521728516,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0007839578902348876,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 8.85387393063622e-09,
      "logits/chosen": -3.2850284576416016,
      "logits/rejected": -4.819733619689941,
      "logps/chosen": -374.94354248046875,
      "logps/ref_chosen": -31.366878509521484,
      "logps/ref_rejected": -90.5899658203125,
      "logps/rejected": -1298.19873046875,
      "loss": 0.9169,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.26877981424331665,
      "rewards/margins": 0.6700939536094666,
      "rewards/rejected": -0.9388737678527832,
      "step": 630
    },
    {
      "epoch": 0.9265785609397944,
      "epsilon_dpo/beta": 0.0007730860379524529,
      "epsilon_dpo/beta_margin_grad_mean": -0.3612503111362457,
      "epsilon_dpo/beta_margin_grad_std": 0.14340829849243164,
      "epsilon_dpo/beta_margin_mean": 0.6428188681602478,
      "epsilon_dpo/beta_margin_std": 0.7184677124023438,
      "epsilon_dpo/loss_margin_mean": 835.1441040039062,
      "grad_norm": 46.966888427734375,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0007786049391143024,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 8.518543427732949e-09,
      "logits/chosen": -3.4160025119781494,
      "logits/rejected": -4.869636535644531,
      "logps/chosen": -483.23895263671875,
      "logps/ref_chosen": -44.379119873046875,
      "logps/ref_rejected": -86.64693450927734,
      "logps/rejected": -1360.65087890625,
      "loss": 0.9527,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.34111732244491577,
      "rewards/margins": 0.642818808555603,
      "rewards/rejected": -0.9839361906051636,
      "step": 631
    },
    {
      "epoch": 0.9280469897209985,
      "epsilon_dpo/beta": 0.0007680523558519781,
      "epsilon_dpo/beta_margin_grad_mean": -0.3662717640399933,
      "epsilon_dpo/beta_margin_grad_std": 0.13854503631591797,
      "epsilon_dpo/beta_margin_mean": 0.6167137026786804,
      "epsilon_dpo/beta_margin_std": 0.6938005685806274,
      "epsilon_dpo/loss_margin_mean": 806.8697509765625,
      "grad_norm": 54.08295440673828,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.0007730486686341465,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 8.189576185789637e-09,
      "logits/chosen": -3.6578102111816406,
      "logits/rejected": -5.017969131469727,
      "logps/chosen": -500.601318359375,
      "logps/ref_chosen": -43.92643737792969,
      "logps/ref_rejected": -90.67631530761719,
      "logps/rejected": -1354.220947265625,
      "loss": 0.9639,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.35203373432159424,
      "rewards/margins": 0.6167136430740356,
      "rewards/rejected": -0.9687473773956299,
      "step": 632
    },
    {
      "epoch": 0.9295154185022027,
      "epsilon_dpo/beta": 0.0007630449254065752,
      "epsilon_dpo/beta_margin_grad_mean": -0.36804521083831787,
      "epsilon_dpo/beta_margin_grad_std": 0.1329408884048462,
      "epsilon_dpo/beta_margin_mean": 0.5877854824066162,
      "epsilon_dpo/beta_margin_std": 0.6214975118637085,
      "epsilon_dpo/loss_margin_mean": 774.373046875,
      "grad_norm": 49.735836029052734,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.000768008641898632,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 7.866980873399015e-09,
      "logits/chosen": -3.5298781394958496,
      "logits/rejected": -4.785070419311523,
      "logps/chosen": -524.2335205078125,
      "logps/ref_chosen": -42.23455047607422,
      "logps/ref_rejected": -100.14579772949219,
      "logps/rejected": -1356.517822265625,
      "loss": 0.9689,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.3694680333137512,
      "rewards/margins": 0.5877854824066162,
      "rewards/rejected": -0.9572535157203674,
      "step": 633
    },
    {
      "epoch": 0.9309838472834068,
      "epsilon_dpo/beta": 0.0007587854051962495,
      "epsilon_dpo/beta_margin_grad_mean": -0.3558174967765808,
      "epsilon_dpo/beta_margin_grad_std": 0.14750294387340546,
      "epsilon_dpo/beta_margin_mean": 0.6918417811393738,
      "epsilon_dpo/beta_margin_std": 0.782641589641571,
      "epsilon_dpo/loss_margin_mean": 916.8782958984375,
      "grad_norm": 43.45764923095703,
      "kl/avg_steps": 0.5625,
      "kl/beta": 0.0007630014442838728,
      "kl/n_epsilon_steps": 0.21875,
      "kl/p_epsilon_steps": 0.78125,
      "learning_rate": 7.550765991247654e-09,
      "logits/chosen": -3.4708662033081055,
      "logits/rejected": -4.9770402908325195,
      "logps/chosen": -502.7445068359375,
      "logps/ref_chosen": -39.36439895629883,
      "logps/ref_rejected": -113.15769958496094,
      "logps/rejected": -1493.4161376953125,
      "loss": 0.933,
      "rewards/accuracies": 0.796875,
      "rewards/chosen": -0.35321560502052307,
      "rewards/margins": 0.6918417811393738,
      "rewards/rejected": -1.0450574159622192,
      "step": 634
    },
    {
      "epoch": 0.9324522760646109,
      "epsilon_dpo/beta": 0.0007538297213613987,
      "epsilon_dpo/beta_margin_grad_mean": -0.39024749398231506,
      "epsilon_dpo/beta_margin_grad_std": 0.13376963138580322,
      "epsilon_dpo/beta_margin_mean": 0.5062357783317566,
      "epsilon_dpo/beta_margin_std": 0.6855813264846802,
      "epsilon_dpo/loss_margin_mean": 675.380859375,
      "grad_norm": 51.91278839111328,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.0007587335421703756,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 7.240939871891699e-09,
      "logits/chosen": -3.523716926574707,
      "logits/rejected": -4.783617973327637,
      "logps/chosen": -562.4381103515625,
      "logps/ref_chosen": -49.88642120361328,
      "logps/ref_rejected": -89.69390869140625,
      "logps/rejected": -1277.62646484375,
      "loss": 1.0411,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.38917070627212524,
      "rewards/margins": 0.5062357783317566,
      "rewards/rejected": -0.8954064846038818,
      "step": 635
    },
    {
      "epoch": 0.933920704845815,
      "epsilon_dpo/beta": 0.0007493861485272646,
      "epsilon_dpo/beta_margin_grad_mean": -0.3881148397922516,
      "epsilon_dpo/beta_margin_grad_std": 0.1402616947889328,
      "epsilon_dpo/beta_margin_mean": 0.5030276775360107,
      "epsilon_dpo/beta_margin_std": 0.6626613140106201,
      "epsilon_dpo/loss_margin_mean": 675.861083984375,
      "grad_norm": 52.333683013916016,
      "kl/avg_steps": 0.59375,
      "kl/beta": 0.0007537868223153055,
      "kl/n_epsilon_steps": 0.203125,
      "kl/p_epsilon_steps": 0.796875,
      "learning_rate": 6.937510679537628e-09,
      "logits/chosen": -3.715489387512207,
      "logits/rejected": -4.909304618835449,
      "logps/chosen": -566.8447875976562,
      "logps/ref_chosen": -46.58656692504883,
      "logps/ref_rejected": -86.21536254882812,
      "logps/rejected": -1282.334716796875,
      "loss": 1.043,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": -0.39245936274528503,
      "rewards/margins": 0.5030276775360107,
      "rewards/rejected": -0.8954870104789734,
      "step": 636
    },
    {
      "epoch": 0.9353891336270191,
      "epsilon_dpo/beta": 0.0007449628901667893,
      "epsilon_dpo/beta_margin_grad_mean": -0.3551463186740875,
      "epsilon_dpo/beta_margin_grad_std": 0.14750678837299347,
      "epsilon_dpo/beta_margin_mean": 0.6876585483551025,
      "epsilon_dpo/beta_margin_std": 0.7672451138496399,
      "epsilon_dpo/loss_margin_mean": 928.1746215820312,
      "grad_norm": 41.60480880737305,
      "kl/avg_steps": 0.59375,
      "kl/beta": 0.0007493376033380628,
      "kl/n_epsilon_steps": 0.203125,
      "kl/p_epsilon_steps": 0.796875,
      "learning_rate": 6.640486409826785e-09,
      "logits/chosen": -3.4943830966949463,
      "logits/rejected": -4.821347236633301,
      "logps/chosen": -483.88134765625,
      "logps/ref_chosen": -37.54460144042969,
      "logps/ref_rejected": -103.94780731201172,
      "logps/rejected": -1478.459228515625,
      "loss": 0.9332,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.33350521326065063,
      "rewards/margins": 0.6876585483551025,
      "rewards/rejected": -1.0211637020111084,
      "step": 637
    },
    {
      "epoch": 0.9368575624082232,
      "epsilon_dpo/beta": 0.0007394017884507775,
      "epsilon_dpo/beta_margin_grad_mean": -0.3611353039741516,
      "epsilon_dpo/beta_margin_grad_std": 0.13404038548469543,
      "epsilon_dpo/beta_margin_mean": 0.6473937034606934,
      "epsilon_dpo/beta_margin_std": 0.7178838849067688,
      "epsilon_dpo/loss_margin_mean": 878.7179565429688,
      "grad_norm": 38.22664260864258,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0007449146942235529,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 6.349874889624962e-09,
      "logits/chosen": -3.2752623558044434,
      "logits/rejected": -5.041168212890625,
      "logps/chosen": -439.9283447265625,
      "logps/ref_chosen": -35.51661682128906,
      "logps/ref_rejected": -85.09121704101562,
      "logps/rejected": -1368.220947265625,
      "loss": 0.9433,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.30027681589126587,
      "rewards/margins": 0.6473937034606934,
      "rewards/rejected": -0.9476705193519592,
      "step": 638
    },
    {
      "epoch": 0.9383259911894273,
      "epsilon_dpo/beta": 0.0007355150883086026,
      "epsilon_dpo/beta_margin_grad_mean": -0.400001585483551,
      "epsilon_dpo/beta_margin_grad_std": 0.13325192034244537,
      "epsilon_dpo/beta_margin_mean": 0.4458373486995697,
      "epsilon_dpo/beta_margin_std": 0.6368516087532043,
      "epsilon_dpo/loss_margin_mean": 610.8491821289062,
      "grad_norm": 46.616573333740234,
      "kl/avg_steps": 0.53125,
      "kl/beta": 0.0007393694249913096,
      "kl/n_epsilon_steps": 0.234375,
      "kl/p_epsilon_steps": 0.765625,
      "learning_rate": 6.065683776815933e-09,
      "logits/chosen": -3.443708896636963,
      "logits/rejected": -4.675836563110352,
      "logps/chosen": -551.689208984375,
      "logps/ref_chosen": -44.109619140625,
      "logps/ref_rejected": -81.57601928710938,
      "logps/rejected": -1200.0047607421875,
      "loss": 1.0788,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": -0.3750152587890625,
      "rewards/margins": 0.4458373486995697,
      "rewards/rejected": -0.8208526372909546,
      "step": 639
    },
    {
      "epoch": 0.9397944199706314,
      "epsilon_dpo/beta": 0.0007302492158487439,
      "epsilon_dpo/beta_margin_grad_mean": -0.34541165828704834,
      "epsilon_dpo/beta_margin_grad_std": 0.12778370082378387,
      "epsilon_dpo/beta_margin_mean": 0.7259718775749207,
      "epsilon_dpo/beta_margin_std": 0.7215431928634644,
      "epsilon_dpo/loss_margin_mean": 997.45458984375,
      "grad_norm": 47.58893966674805,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0007354622939601541,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 5.7879205600998296e-09,
      "logits/chosen": -3.3290483951568604,
      "logits/rejected": -4.9194793701171875,
      "logps/chosen": -445.7485656738281,
      "logps/ref_chosen": -40.14595413208008,
      "logps/ref_rejected": -114.68016815185547,
      "logps/rejected": -1517.7373046875,
      "loss": 0.8862,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.2974758744239807,
      "rewards/margins": 0.7259718179702759,
      "rewards/rejected": -1.0234477519989014,
      "step": 640
    },
    {
      "epoch": 0.9412628487518355,
      "epsilon_dpo/beta": 0.0007254944066517055,
      "epsilon_dpo/beta_margin_grad_mean": -0.3516106605529785,
      "epsilon_dpo/beta_margin_grad_std": 0.14073027670383453,
      "epsilon_dpo/beta_margin_mean": 0.6929788589477539,
      "epsilon_dpo/beta_margin_std": 0.7391694188117981,
      "epsilon_dpo/loss_margin_mean": 959.70703125,
      "grad_norm": 43.9137077331543,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.0007302138837985694,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 5.516592558795746e-09,
      "logits/chosen": -3.348452568054199,
      "logits/rejected": -5.015802383422852,
      "logps/chosen": -465.90545654296875,
      "logps/ref_chosen": -39.17839050292969,
      "logps/ref_rejected": -94.15284729003906,
      "logps/rejected": -1480.5869140625,
      "loss": 0.9185,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.3105979263782501,
      "rewards/margins": 0.6929788589477539,
      "rewards/rejected": -1.0035767555236816,
      "step": 641
    },
    {
      "epoch": 0.9427312775330396,
      "epsilon_dpo/beta": 0.0007200841791927814,
      "epsilon_dpo/beta_margin_grad_mean": -0.34033486247062683,
      "epsilon_dpo/beta_margin_grad_std": 0.1520089954137802,
      "epsilon_dpo/beta_margin_mean": 0.7516418695449829,
      "epsilon_dpo/beta_margin_std": 0.7943530082702637,
      "epsilon_dpo/loss_margin_mean": 1047.7049560546875,
      "grad_norm": 42.03846740722656,
      "kl/avg_steps": 0.75,
      "kl/beta": 0.0007254530792124569,
      "kl/n_epsilon_steps": 0.125,
      "kl/p_epsilon_steps": 0.875,
      "learning_rate": 5.251706922648868e-09,
      "logits/chosen": -3.4749832153320312,
      "logits/rejected": -5.020747184753418,
      "logps/chosen": -480.06494140625,
      "logps/ref_chosen": -46.66090393066406,
      "logps/ref_rejected": -115.78807067871094,
      "logps/rejected": -1596.8970947265625,
      "loss": 0.8997,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.3134210705757141,
      "rewards/margins": 0.7516418695449829,
      "rewards/rejected": -1.0650629997253418,
      "step": 642
    },
    {
      "epoch": 0.9441997063142438,
      "epsilon_dpo/beta": 0.0007149488083086908,
      "epsilon_dpo/beta_margin_grad_mean": -0.3584926128387451,
      "epsilon_dpo/beta_margin_grad_std": 0.11874385923147202,
      "epsilon_dpo/beta_margin_mean": 0.6369244456291199,
      "epsilon_dpo/beta_margin_std": 0.594021201133728,
      "epsilon_dpo/loss_margin_mean": 893.6637573242188,
      "grad_norm": 39.312984466552734,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0007200526888482273,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 4.993270631642038e-09,
      "logits/chosen": -3.4138784408569336,
      "logits/rejected": -4.991301536560059,
      "logps/chosen": -453.1264343261719,
      "logps/ref_chosen": -35.49954605102539,
      "logps/ref_rejected": -92.612060546875,
      "logps/rejected": -1403.9027099609375,
      "loss": 0.9222,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.29885566234588623,
      "rewards/margins": 0.6369244456291199,
      "rewards/rejected": -0.9357801079750061,
      "step": 643
    },
    {
      "epoch": 0.9456681350954479,
      "epsilon_dpo/beta": 0.0007100701914168894,
      "epsilon_dpo/beta_margin_grad_mean": -0.37228909134864807,
      "epsilon_dpo/beta_margin_grad_std": 0.1232423335313797,
      "epsilon_dpo/beta_margin_mean": 0.5710337162017822,
      "epsilon_dpo/beta_margin_std": 0.5858953595161438,
      "epsilon_dpo/loss_margin_mean": 807.3773193359375,
      "grad_norm": 46.13814926147461,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0007149142329581082,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 4.741290495811873e-09,
      "logits/chosen": -3.398604154586792,
      "logits/rejected": -4.7896952629089355,
      "logps/chosen": -432.34515380859375,
      "logps/ref_chosen": -37.63022232055664,
      "logps/ref_rejected": -93.44629669189453,
      "logps/rejected": -1295.53857421875,
      "loss": 0.9704,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.28092193603515625,
      "rewards/margins": 0.5710337162017822,
      "rewards/rejected": -0.8519556522369385,
      "step": 644
    },
    {
      "epoch": 0.947136563876652,
      "epsilon_dpo/beta": 0.0007061094511300325,
      "epsilon_dpo/beta_margin_grad_mean": -0.38581857085227966,
      "epsilon_dpo/beta_margin_grad_std": 0.14517748355865479,
      "epsilon_dpo/beta_margin_mean": 0.5264767408370972,
      "epsilon_dpo/beta_margin_std": 0.6991389393806458,
      "epsilon_dpo/loss_margin_mean": 750.698974609375,
      "grad_norm": 42.622047424316406,
      "kl/avg_steps": 0.5625,
      "kl/beta": 0.0007100327638909221,
      "kl/n_epsilon_steps": 0.21875,
      "kl/p_epsilon_steps": 0.78125,
      "learning_rate": 4.495773155069299e-09,
      "logits/chosen": -3.5165293216705322,
      "logits/rejected": -4.699129104614258,
      "logps/chosen": -535.257568359375,
      "logps/ref_chosen": -37.85113525390625,
      "logps/ref_rejected": -105.40227508544922,
      "logps/rejected": -1353.5076904296875,
      "loss": 1.0346,
      "rewards/accuracies": 0.828125,
      "rewards/chosen": -0.35279667377471924,
      "rewards/margins": 0.5264767408370972,
      "rewards/rejected": -0.8792734146118164,
      "step": 645
    },
    {
      "epoch": 0.9486049926578561,
      "epsilon_dpo/beta": 0.0007019391632638872,
      "epsilon_dpo/beta_margin_grad_mean": -0.39241456985473633,
      "epsilon_dpo/beta_margin_grad_std": 0.129283145070076,
      "epsilon_dpo/beta_margin_mean": 0.4752976894378662,
      "epsilon_dpo/beta_margin_std": 0.5920587778091431,
      "epsilon_dpo/loss_margin_mean": 681.503662109375,
      "grad_norm": 47.977237701416016,
      "kl/avg_steps": 0.59375,
      "kl/beta": 0.0007060611969791353,
      "kl/n_epsilon_steps": 0.203125,
      "kl/p_epsilon_steps": 0.796875,
      "learning_rate": 4.256725079024553e-09,
      "logits/chosen": -3.479154109954834,
      "logits/rejected": -4.7768402099609375,
      "logps/chosen": -520.402099609375,
      "logps/ref_chosen": -41.30128860473633,
      "logps/ref_rejected": -82.82234954833984,
      "logps/rejected": -1243.4267578125,
      "loss": 1.0461,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": -0.33754080533981323,
      "rewards/margins": 0.4752976894378662,
      "rewards/rejected": -0.8128384947776794,
      "step": 646
    },
    {
      "epoch": 0.9500734214390602,
      "epsilon_dpo/beta": 0.0006964798085391521,
      "epsilon_dpo/beta_margin_grad_mean": -0.3713095486164093,
      "epsilon_dpo/beta_margin_grad_std": 0.11274945735931396,
      "epsilon_dpo/beta_margin_mean": 0.5578054785728455,
      "epsilon_dpo/beta_margin_std": 0.5175791382789612,
      "epsilon_dpo/loss_margin_mean": 803.4420776367188,
      "grad_norm": 48.550132751464844,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0007018937030807137,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 4.024152566816791e-09,
      "logits/chosen": -3.453265905380249,
      "logits/rejected": -4.754009246826172,
      "logps/chosen": -431.2979736328125,
      "logps/ref_chosen": -35.967567443847656,
      "logps/ref_rejected": -98.74945068359375,
      "logps/rejected": -1297.52197265625,
      "loss": 0.9654,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.2764018774032593,
      "rewards/margins": 0.5578055381774902,
      "rewards/rejected": -0.8342074155807495,
      "step": 647
    },
    {
      "epoch": 0.9515418502202643,
      "epsilon_dpo/beta": 0.0006910806987434626,
      "epsilon_dpo/beta_margin_grad_mean": -0.35618698596954346,
      "epsilon_dpo/beta_margin_grad_std": 0.14514584839344025,
      "epsilon_dpo/beta_margin_mean": 0.6508342027664185,
      "epsilon_dpo/beta_margin_std": 0.699810802936554,
      "epsilon_dpo/loss_margin_mean": 945.5726318359375,
      "grad_norm": 52.17890930175781,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0006964526255615056,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 3.798061746947995e-09,
      "logits/chosen": -3.3199658393859863,
      "logits/rejected": -4.965579986572266,
      "logps/chosen": -432.31951904296875,
      "logps/ref_chosen": -33.676727294921875,
      "logps/ref_rejected": -105.11663818359375,
      "logps/rejected": -1449.33203125,
      "loss": 0.9453,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.2765774726867676,
      "rewards/margins": 0.6508342027664185,
      "rewards/rejected": -0.927411675453186,
      "step": 648
    },
    {
      "epoch": 0.9530102790014684,
      "epsilon_dpo/beta": 0.0006873422535136342,
      "epsilon_dpo/beta_margin_grad_mean": -0.40578776597976685,
      "epsilon_dpo/beta_margin_grad_std": 0.1477607786655426,
      "epsilon_dpo/beta_margin_mean": 0.43915510177612305,
      "epsilon_dpo/beta_margin_std": 0.7278276085853577,
      "epsilon_dpo/loss_margin_mean": 644.4627685546875,
      "grad_norm": 40.87678146362305,
      "kl/avg_steps": 0.546875,
      "kl/beta": 0.0006910538068041205,
      "kl/n_epsilon_steps": 0.21875,
      "kl/p_epsilon_steps": 0.765625,
      "learning_rate": 3.5784585771215235e-09,
      "logits/chosen": -3.504033088684082,
      "logits/rejected": -4.718470573425293,
      "logps/chosen": -526.4951171875,
      "logps/ref_chosen": -45.06011199951172,
      "logps/ref_rejected": -86.40021514892578,
      "logps/rejected": -1212.298095703125,
      "loss": 1.1091,
      "rewards/accuracies": 0.78125,
      "rewards/chosen": -0.3326740860939026,
      "rewards/margins": 0.43915510177612305,
      "rewards/rejected": -0.7718292474746704,
      "step": 649
    },
    {
      "epoch": 0.9544787077826725,
      "epsilon_dpo/beta": 0.0006830678903497756,
      "epsilon_dpo/beta_margin_grad_mean": -0.3580322861671448,
      "epsilon_dpo/beta_margin_grad_std": 0.16048245131969452,
      "epsilon_dpo/beta_margin_mean": 0.6828190684318542,
      "epsilon_dpo/beta_margin_std": 0.8265655040740967,
      "epsilon_dpo/loss_margin_mean": 1005.4606323242188,
      "grad_norm": 45.862091064453125,
      "kl/avg_steps": 0.625,
      "kl/beta": 0.0006872951635159552,
      "kl/n_epsilon_steps": 0.1875,
      "kl/p_epsilon_steps": 0.8125,
      "learning_rate": 3.3653488440851253e-09,
      "logits/chosen": -3.553635835647583,
      "logits/rejected": -5.113847732543945,
      "logps/chosen": -577.81494140625,
      "logps/ref_chosen": -39.21210861206055,
      "logps/ref_rejected": -103.7435302734375,
      "logps/rejected": -1647.8070068359375,
      "loss": 0.957,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.36932051181793213,
      "rewards/margins": 0.682819128036499,
      "rewards/rejected": -1.0521395206451416,
      "step": 650
    },
    {
      "epoch": 0.9559471365638766,
      "epsilon_dpo/beta": 0.0006790386396460235,
      "epsilon_dpo/beta_margin_grad_mean": -0.3646124601364136,
      "epsilon_dpo/beta_margin_grad_std": 0.14354108273983002,
      "epsilon_dpo/beta_margin_mean": 0.6235911250114441,
      "epsilon_dpo/beta_margin_std": 0.7001570463180542,
      "epsilon_dpo/loss_margin_mean": 923.6541137695312,
      "grad_norm": 40.79838943481445,
      "kl/avg_steps": 0.59375,
      "kl/beta": 0.0006830262136645615,
      "kl/n_epsilon_steps": 0.203125,
      "kl/p_epsilon_steps": 0.796875,
      "learning_rate": 3.158738163478475e-09,
      "logits/chosen": -3.4843456745147705,
      "logits/rejected": -4.997735500335693,
      "logps/chosen": -481.7735595703125,
      "logps/ref_chosen": -34.16796875,
      "logps/ref_rejected": -105.96416473388672,
      "logps/rejected": -1477.223876953125,
      "loss": 0.9624,
      "rewards/accuracies": 0.828125,
      "rewards/chosen": -0.3056422173976898,
      "rewards/margins": 0.6235911250114441,
      "rewards/rejected": -0.9292333126068115,
      "step": 651
    },
    {
      "epoch": 0.9574155653450808,
      "epsilon_dpo/beta": 0.0006750306929461658,
      "epsilon_dpo/beta_margin_grad_mean": -0.36323583126068115,
      "epsilon_dpo/beta_margin_grad_std": 0.1515437662601471,
      "epsilon_dpo/beta_margin_mean": 0.6380264759063721,
      "epsilon_dpo/beta_margin_std": 0.7482752203941345,
      "epsilon_dpo/loss_margin_mean": 950.8151245117188,
      "grad_norm": 47.86579895019531,
      "kl/avg_steps": 0.59375,
      "kl/beta": 0.0006789946928620338,
      "kl/n_epsilon_steps": 0.203125,
      "kl/p_epsilon_steps": 0.796875,
      "learning_rate": 2.9586319796851555e-09,
      "logits/chosen": -3.3652729988098145,
      "logits/rejected": -4.884889602661133,
      "logps/chosen": -517.8300170898438,
      "logps/ref_chosen": -43.1708984375,
      "logps/ref_rejected": -119.19691467285156,
      "logps/rejected": -1544.671142578125,
      "loss": 0.9666,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": -0.32234591245651245,
      "rewards/margins": 0.6380264759063721,
      "rewards/rejected": -0.9603723883628845,
      "step": 652
    },
    {
      "epoch": 0.9588839941262849,
      "epsilon_dpo/beta": 0.0006693587056361139,
      "epsilon_dpo/beta_margin_grad_mean": -0.3340756595134735,
      "epsilon_dpo/beta_margin_grad_std": 0.13510021567344666,
      "epsilon_dpo/beta_margin_mean": 0.7706581950187683,
      "epsilon_dpo/beta_margin_std": 0.685066819190979,
      "epsilon_dpo/loss_margin_mean": 1153.705078125,
      "grad_norm": 48.55562973022461,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.0006749869789928198,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 2.7650355656892166e-09,
      "logits/chosen": -3.2305068969726562,
      "logits/rejected": -5.08480167388916,
      "logps/chosen": -467.5439453125,
      "logps/ref_chosen": -36.02939987182617,
      "logps/ref_rejected": -111.6745376586914,
      "logps/rejected": -1696.89404296875,
      "loss": 0.8559,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.28941768407821655,
      "rewards/margins": 0.7706582546234131,
      "rewards/rejected": -1.0600758790969849,
      "step": 653
    },
    {
      "epoch": 0.960352422907489,
      "epsilon_dpo/beta": 0.0006645949906669557,
      "epsilon_dpo/beta_margin_grad_mean": -0.36657387018203735,
      "epsilon_dpo/beta_margin_grad_std": 0.1234474629163742,
      "epsilon_dpo/beta_margin_mean": 0.605144739151001,
      "epsilon_dpo/beta_margin_std": 0.6342909932136536,
      "epsilon_dpo/loss_margin_mean": 914.0985717773438,
      "grad_norm": 41.65330505371094,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0006693393806926906,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 2.577954022936174e-09,
      "logits/chosen": -3.4308786392211914,
      "logits/rejected": -4.949398040771484,
      "logps/chosen": -473.72698974609375,
      "logps/ref_chosen": -38.32045364379883,
      "logps/ref_rejected": -105.32196807861328,
      "logps/rejected": -1454.8271484375,
      "loss": 0.9534,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.2903170585632324,
      "rewards/margins": 0.605144739151001,
      "rewards/rejected": -0.8954617977142334,
      "step": 654
    },
    {
      "epoch": 0.9618208516886931,
      "epsilon_dpo/beta": 0.000660475343465805,
      "epsilon_dpo/beta_margin_grad_mean": -0.3668808341026306,
      "epsilon_dpo/beta_margin_grad_std": 0.14767590165138245,
      "epsilon_dpo/beta_margin_mean": 0.6252740025520325,
      "epsilon_dpo/beta_margin_std": 0.7545697093009949,
      "epsilon_dpo/loss_margin_mean": 952.0883178710938,
      "grad_norm": 43.04616165161133,
      "kl/avg_steps": 0.625,
      "kl/beta": 0.0006645628600381315,
      "kl/n_epsilon_steps": 0.1875,
      "kl/p_epsilon_steps": 0.8125,
      "learning_rate": 2.397392281198729e-09,
      "logits/chosen": -3.374847173690796,
      "logits/rejected": -4.775622844696045,
      "logps/chosen": -389.5122985839844,
      "logps/ref_chosen": -29.801528930664062,
      "logps/ref_rejected": -104.75204467773438,
      "logps/rejected": -1416.551025390625,
      "loss": 0.9735,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": -0.23886506259441376,
      "rewards/margins": 0.6252740025520325,
      "rewards/rejected": -0.8641390800476074,
      "step": 655
    },
    {
      "epoch": 0.9632892804698973,
      "epsilon_dpo/beta": 0.0006549282115884125,
      "epsilon_dpo/beta_margin_grad_mean": -0.32215529680252075,
      "epsilon_dpo/beta_margin_grad_std": 0.1285041719675064,
      "epsilon_dpo/beta_margin_mean": 0.8166189193725586,
      "epsilon_dpo/beta_margin_std": 0.6401075720787048,
      "epsilon_dpo/loss_margin_mean": 1249.337646484375,
      "grad_norm": 51.54441833496094,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.0006604351219721138,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 2.223355098446622e-09,
      "logits/chosen": -3.3955602645874023,
      "logits/rejected": -5.103156089782715,
      "logps/chosen": -448.05487060546875,
      "logps/ref_chosen": -40.732295989990234,
      "logps/ref_rejected": -120.36123657226562,
      "logps/rejected": -1777.021484375,
      "loss": 0.8158,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.2671954035758972,
      "rewards/margins": 0.8166189193725586,
      "rewards/rejected": -1.0838143825531006,
      "step": 656
    },
    {
      "epoch": 0.9647577092511013,
      "epsilon_dpo/beta": 0.0006502671749331057,
      "epsilon_dpo/beta_margin_grad_mean": -0.3548159599304199,
      "epsilon_dpo/beta_margin_grad_std": 0.12233418226242065,
      "epsilon_dpo/beta_margin_mean": 0.6513071060180664,
      "epsilon_dpo/beta_margin_std": 0.5884512662887573,
      "epsilon_dpo/loss_margin_mean": 1004.9851684570312,
      "grad_norm": 43.996578216552734,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0006549093523062766,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 2.055847060721566e-09,
      "logits/chosen": -3.505584716796875,
      "logits/rejected": -5.049525260925293,
      "logps/chosen": -486.0547790527344,
      "logps/ref_chosen": -32.56511688232422,
      "logps/ref_rejected": -104.74242401123047,
      "logps/rejected": -1563.21728515625,
      "loss": 0.9133,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.29575812816619873,
      "rewards/margins": 0.6513071060180664,
      "rewards/rejected": -0.9470652341842651,
      "step": 657
    },
    {
      "epoch": 0.9662261380323054,
      "epsilon_dpo/beta": 0.0006452203379012644,
      "epsilon_dpo/beta_margin_grad_mean": -0.36330175399780273,
      "epsilon_dpo/beta_margin_grad_std": 0.11596217751502991,
      "epsilon_dpo/beta_margin_mean": 0.5977858304977417,
      "epsilon_dpo/beta_margin_std": 0.5393864512443542,
      "epsilon_dpo/loss_margin_mean": 929.5409545898438,
      "grad_norm": 35.85395431518555,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0006502358010038733,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 1.8948725820160662e-09,
      "logits/chosen": -3.3597793579101562,
      "logits/rejected": -5.021791458129883,
      "logps/chosen": -452.764404296875,
      "logps/ref_chosen": -32.16458511352539,
      "logps/ref_rejected": -100.98091125488281,
      "logps/rejected": -1451.1217041015625,
      "loss": 0.9409,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.2720116972923279,
      "rewards/margins": 0.5977858304977417,
      "rewards/rejected": -0.8697974681854248,
      "step": 658
    },
    {
      "epoch": 0.9676945668135095,
      "epsilon_dpo/beta": 0.0006414285162463784,
      "epsilon_dpo/beta_margin_grad_mean": -0.3757038414478302,
      "epsilon_dpo/beta_margin_grad_std": 0.12834706902503967,
      "epsilon_dpo/beta_margin_mean": 0.5647572875022888,
      "epsilon_dpo/beta_margin_std": 0.6336069703102112,
      "epsilon_dpo/loss_margin_mean": 885.123046875,
      "grad_norm": 40.669891357421875,
      "kl/avg_steps": 0.59375,
      "kl/beta": 0.0006451951921917498,
      "kl/n_epsilon_steps": 0.203125,
      "kl/p_epsilon_steps": 0.796875,
      "learning_rate": 1.7404359041573723e-09,
      "logits/chosen": -3.3426105976104736,
      "logits/rejected": -4.952641487121582,
      "logps/chosen": -490.76068115234375,
      "logps/ref_chosen": -44.455406188964844,
      "logps/ref_rejected": -93.29725646972656,
      "logps/rejected": -1424.7255859375,
      "loss": 0.9852,
      "rewards/accuracies": 0.828125,
      "rewards/chosen": -0.2872786521911621,
      "rewards/margins": 0.5647573471069336,
      "rewards/rejected": -0.8520359992980957,
      "step": 659
    },
    {
      "epoch": 0.9691629955947136,
      "epsilon_dpo/beta": 0.0006372415809892118,
      "epsilon_dpo/beta_margin_grad_mean": -0.36822497844696045,
      "epsilon_dpo/beta_margin_grad_std": 0.1204964891076088,
      "epsilon_dpo/beta_margin_mean": 0.5889629125595093,
      "epsilon_dpo/beta_margin_std": 0.5796295404434204,
      "epsilon_dpo/loss_margin_mean": 928.2510375976562,
      "grad_norm": 37.189697265625,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.0006413869559764862,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 1.592541096695571e-09,
      "logits/chosen": -3.402561664581299,
      "logits/rejected": -5.116570472717285,
      "logps/chosen": -454.3048095703125,
      "logps/ref_chosen": -41.76560974121094,
      "logps/ref_rejected": -82.32925415039062,
      "logps/rejected": -1423.1195068359375,
      "loss": 0.9552,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.26376718282699585,
      "rewards/margins": 0.5889629125595093,
      "rewards/rejected": -0.8527300953865051,
      "step": 660
    },
    {
      "epoch": 0.9706314243759178,
      "epsilon_dpo/beta": 0.0006320912507362664,
      "epsilon_dpo/beta_margin_grad_mean": -0.3780907392501831,
      "epsilon_dpo/beta_margin_grad_std": 0.11522994190454483,
      "epsilon_dpo/beta_margin_mean": 0.5599713921546936,
      "epsilon_dpo/beta_margin_std": 0.6463156938552856,
      "epsilon_dpo/loss_margin_mean": 888.2638549804688,
      "grad_norm": 32.038639068603516,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0006372053176164627,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 1.4511920567963908e-09,
      "logits/chosen": -3.5372705459594727,
      "logits/rejected": -4.98753547668457,
      "logps/chosen": -504.87115478515625,
      "logps/ref_chosen": -38.271453857421875,
      "logps/ref_rejected": -92.10589599609375,
      "logps/rejected": -1446.969482421875,
      "loss": 0.9829,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.29568618535995483,
      "rewards/margins": 0.5599713921546936,
      "rewards/rejected": -0.8556575775146484,
      "step": 661
    },
    {
      "epoch": 0.9720998531571219,
      "epsilon_dpo/beta": 0.0006279845256358385,
      "epsilon_dpo/beta_margin_grad_mean": -0.3694445788860321,
      "epsilon_dpo/beta_margin_grad_std": 0.1416751593351364,
      "epsilon_dpo/beta_margin_mean": 0.5916183590888977,
      "epsilon_dpo/beta_margin_std": 0.681064784526825,
      "epsilon_dpo/loss_margin_mean": 946.8544311523438,
      "grad_norm": 37.34529113769531,
      "kl/avg_steps": 0.65625,
      "kl/beta": 0.0006320697139017284,
      "kl/n_epsilon_steps": 0.171875,
      "kl/p_epsilon_steps": 0.828125,
      "learning_rate": 1.3163925091384532e-09,
      "logits/chosen": -3.4751718044281006,
      "logits/rejected": -5.172747611999512,
      "logps/chosen": -533.3956298828125,
      "logps/ref_chosen": -39.42928695678711,
      "logps/ref_rejected": -96.40357208251953,
      "logps/rejected": -1537.224365234375,
      "loss": 0.9818,
      "rewards/accuracies": 0.875,
      "rewards/chosen": -0.31171175837516785,
      "rewards/margins": 0.5916184186935425,
      "rewards/rejected": -0.9033301472663879,
      "step": 662
    },
    {
      "epoch": 0.973568281938326,
      "epsilon_dpo/beta": 0.0006246752454899251,
      "epsilon_dpo/beta_margin_grad_mean": -0.3684435486793518,
      "epsilon_dpo/beta_margin_grad_std": 0.14285063743591309,
      "epsilon_dpo/beta_margin_mean": 0.6072575449943542,
      "epsilon_dpo/beta_margin_std": 0.6948506832122803,
      "epsilon_dpo/loss_margin_mean": 978.2127075195312,
      "grad_norm": 46.52898025512695,
      "kl/avg_steps": 0.53125,
      "kl/beta": 0.0006279487861320376,
      "kl/n_epsilon_steps": 0.234375,
      "kl/p_epsilon_steps": 0.765625,
      "learning_rate": 1.1881460058152382e-09,
      "logits/chosen": -3.4985265731811523,
      "logits/rejected": -4.843735694885254,
      "logps/chosen": -518.152099609375,
      "logps/ref_chosen": -44.08625411987305,
      "logps/ref_rejected": -121.31452178955078,
      "logps/rejected": -1573.593017578125,
      "loss": 0.9729,
      "rewards/accuracies": 0.828125,
      "rewards/chosen": -0.2980585992336273,
      "rewards/margins": 0.6072574853897095,
      "rewards/rejected": -0.9053161144256592,
      "step": 663
    },
    {
      "epoch": 0.9750367107195301,
      "epsilon_dpo/beta": 0.0006194221205078065,
      "epsilon_dpo/beta_margin_grad_mean": -0.33726438879966736,
      "epsilon_dpo/beta_margin_grad_std": 0.1218588799238205,
      "epsilon_dpo/beta_margin_mean": 0.7527669072151184,
      "epsilon_dpo/beta_margin_std": 0.655570924282074,
      "epsilon_dpo/loss_margin_mean": 1217.3953857421875,
      "grad_norm": 43.377994537353516,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.0006246304837986827,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 1.066455926241383e-09,
      "logits/chosen": -3.1317358016967773,
      "logits/rejected": -5.002200126647949,
      "logps/chosen": -338.7730712890625,
      "logps/ref_chosen": -31.542118072509766,
      "logps/ref_rejected": -116.06498718261719,
      "logps/rejected": -1640.69140625,
      "loss": 0.8558,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.19074614346027374,
      "rewards/margins": 0.7527669072151184,
      "rewards/rejected": -0.943513035774231,
      "step": 664
    },
    {
      "epoch": 0.9765051395007343,
      "epsilon_dpo/beta": 0.0006142394850030541,
      "epsilon_dpo/beta_margin_grad_mean": -0.37084051966667175,
      "epsilon_dpo/beta_margin_grad_std": 0.11781810969114304,
      "epsilon_dpo/beta_margin_mean": 0.5667802691459656,
      "epsilon_dpo/beta_margin_std": 0.5540540218353271,
      "epsilon_dpo/loss_margin_mean": 925.40185546875,
      "grad_norm": 44.8764762878418,
      "kl/avg_steps": 0.84375,
      "kl/beta": 0.0006194042507559061,
      "kl/n_epsilon_steps": 0.078125,
      "kl/p_epsilon_steps": 0.921875,
      "learning_rate": 9.513254770636137e-10,
      "logits/chosen": -3.3389735221862793,
      "logits/rejected": -5.07022762298584,
      "logps/chosen": -454.4164123535156,
      "logps/ref_chosen": -36.933109283447266,
      "logps/ref_rejected": -89.9020767211914,
      "logps/rejected": -1432.787353515625,
      "loss": 0.9663,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.2572126090526581,
      "rewards/margins": 0.5667802095413208,
      "rewards/rejected": -0.8239928483963013,
      "step": 665
    },
    {
      "epoch": 0.9779735682819384,
      "epsilon_dpo/beta": 0.0006110197864472866,
      "epsilon_dpo/beta_margin_grad_mean": -0.39552050828933716,
      "epsilon_dpo/beta_margin_grad_std": 0.1250423640012741,
      "epsilon_dpo/beta_margin_mean": 0.46575719118118286,
      "epsilon_dpo/beta_margin_std": 0.5870808362960815,
      "epsilon_dpo/loss_margin_mean": 767.228515625,
      "grad_norm": 41.77029037475586,
      "kl/avg_steps": 0.53125,
      "kl/beta": 0.0006142217316664755,
      "kl/n_epsilon_steps": 0.234375,
      "kl/p_epsilon_steps": 0.765625,
      "learning_rate": 8.427576920763956e-10,
      "logits/chosen": -3.6731343269348145,
      "logits/rejected": -4.908390522003174,
      "logps/chosen": -623.1746826171875,
      "logps/ref_chosen": -47.59907913208008,
      "logps/ref_rejected": -102.33778381347656,
      "logps/rejected": -1445.141845703125,
      "loss": 1.0509,
      "rewards/accuracies": 0.796875,
      "rewards/chosen": -0.3538281321525574,
      "rewards/margins": 0.46575719118118286,
      "rewards/rejected": -0.8195853233337402,
      "step": 666
    },
    {
      "epoch": 0.9794419970631424,
      "epsilon_dpo/beta": 0.000606836169026792,
      "epsilon_dpo/beta_margin_grad_mean": -0.38951799273490906,
      "epsilon_dpo/beta_margin_grad_std": 0.12825334072113037,
      "epsilon_dpo/beta_margin_mean": 0.4804462492465973,
      "epsilon_dpo/beta_margin_std": 0.5962188839912415,
      "epsilon_dpo/loss_margin_mean": 796.4560546875,
      "grad_norm": 38.44181823730469,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0006109759560786188,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 7.407554321417764e-10,
      "logits/chosen": -3.463103771209717,
      "logits/rejected": -4.9513959884643555,
      "logps/chosen": -567.8262939453125,
      "logps/ref_chosen": -43.69598388671875,
      "logps/ref_rejected": -93.95926666259766,
      "logps/rejected": -1414.545654296875,
      "loss": 1.0431,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.320189893245697,
      "rewards/margins": 0.4804462790489197,
      "rewards/rejected": -0.8006361722946167,
      "step": 667
    },
    {
      "epoch": 0.9809104258443465,
      "epsilon_dpo/beta": 0.0006032615783624351,
      "epsilon_dpo/beta_margin_grad_mean": -0.3832368850708008,
      "epsilon_dpo/beta_margin_grad_std": 0.11170457303524017,
      "epsilon_dpo/beta_margin_mean": 0.5103982090950012,
      "epsilon_dpo/beta_margin_std": 0.517078161239624,
      "epsilon_dpo/loss_margin_mean": 850.3494262695312,
      "grad_norm": 37.77009963989258,
      "kl/avg_steps": 0.59375,
      "kl/beta": 0.0006068041548132896,
      "kl/n_epsilon_steps": 0.203125,
      "kl/p_epsilon_steps": 0.796875,
      "learning_rate": 6.453213851142225e-10,
      "logits/chosen": -3.5545034408569336,
      "logits/rejected": -4.857883453369141,
      "logps/chosen": -569.862548828125,
      "logps/ref_chosen": -48.83540344238281,
      "logps/ref_rejected": -108.24037170410156,
      "logps/rejected": -1479.616943359375,
      "loss": 0.9999,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.31520360708236694,
      "rewards/margins": 0.5103981494903564,
      "rewards/rejected": -0.8256018161773682,
      "step": 668
    },
    {
      "epoch": 0.9823788546255506,
      "epsilon_dpo/beta": 0.0005995123065076768,
      "epsilon_dpo/beta_margin_grad_mean": -0.3920701742172241,
      "epsilon_dpo/beta_margin_grad_std": 0.11787072569131851,
      "epsilon_dpo/beta_margin_mean": 0.4740796983242035,
      "epsilon_dpo/beta_margin_std": 0.5450995564460754,
      "epsilon_dpo/loss_margin_mean": 795.2932739257812,
      "grad_norm": 37.88663101196289,
      "kl/avg_steps": 0.625,
      "kl/beta": 0.0006032225210219622,
      "kl/n_epsilon_steps": 0.1875,
      "kl/p_epsilon_steps": 0.8125,
      "learning_rate": 5.564580657695939e-10,
      "logits/chosen": -3.4180922508239746,
      "logits/rejected": -4.825163841247559,
      "logps/chosen": -459.27081298828125,
      "logps/ref_chosen": -33.70474624633789,
      "logps/ref_rejected": -84.20474243164062,
      "logps/rejected": -1305.064208984375,
      "loss": 1.0343,
      "rewards/accuracies": 0.828125,
      "rewards/chosen": -0.2562495172023773,
      "rewards/margins": 0.4740796983242035,
      "rewards/rejected": -0.7303292751312256,
      "step": 669
    },
    {
      "epoch": 0.9838472834067548,
      "epsilon_dpo/beta": 0.0005952265928499401,
      "epsilon_dpo/beta_margin_grad_mean": -0.3791544735431671,
      "epsilon_dpo/beta_margin_grad_std": 0.1323607712984085,
      "epsilon_dpo/beta_margin_mean": 0.5472928881645203,
      "epsilon_dpo/beta_margin_std": 0.6459078788757324,
      "epsilon_dpo/loss_margin_mean": 923.514404296875,
      "grad_norm": 38.17559051513672,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0005994758103042841,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 4.741678157389739e-10,
      "logits/chosen": -3.493618965148926,
      "logits/rejected": -5.067645072937012,
      "logps/chosen": -505.29241943359375,
      "logps/ref_chosen": -41.06857681274414,
      "logps/ref_rejected": -103.42701721191406,
      "logps/rejected": -1491.165283203125,
      "loss": 1.0026,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.27749842405319214,
      "rewards/margins": 0.547292947769165,
      "rewards/rejected": -0.8247913122177124,
      "step": 670
    },
    {
      "epoch": 0.9853157121879589,
      "epsilon_dpo/beta": 0.000591909047216177,
      "epsilon_dpo/beta_margin_grad_mean": -0.38887372612953186,
      "epsilon_dpo/beta_margin_grad_std": 0.14615273475646973,
      "epsilon_dpo/beta_margin_mean": 0.5095086693763733,
      "epsilon_dpo/beta_margin_std": 0.7063105702400208,
      "epsilon_dpo/loss_margin_mean": 866.9005737304688,
      "grad_norm": 41.20299530029297,
      "kl/avg_steps": 0.5625,
      "kl/beta": 0.0005951978382654488,
      "kl/n_epsilon_steps": 0.21875,
      "kl/p_epsilon_steps": 0.78125,
      "learning_rate": 3.9845280344705245e-10,
      "logits/chosen": -3.6080431938171387,
      "logits/rejected": -5.071390151977539,
      "logps/chosen": -567.7605590820312,
      "logps/ref_chosen": -35.17292785644531,
      "logps/ref_rejected": -90.90328216552734,
      "logps/rejected": -1490.3914794921875,
      "loss": 1.0497,
      "rewards/accuracies": 0.796875,
      "rewards/chosen": -0.3173062801361084,
      "rewards/margins": 0.5095087289810181,
      "rewards/rejected": -0.8268150091171265,
      "step": 671
    },
    {
      "epoch": 0.986784140969163,
      "epsilon_dpo/beta": 0.000588598137255758,
      "epsilon_dpo/beta_margin_grad_mean": -0.3892599046230316,
      "epsilon_dpo/beta_margin_grad_std": 0.13194513320922852,
      "epsilon_dpo/beta_margin_mean": 0.491289347410202,
      "epsilon_dpo/beta_margin_std": 0.6089746356010437,
      "epsilon_dpo/loss_margin_mean": 840.4798583984375,
      "grad_norm": 42.78655242919922,
      "kl/avg_steps": 0.5625,
      "kl/beta": 0.0005918685346841812,
      "kl/n_epsilon_steps": 0.21875,
      "kl/p_epsilon_steps": 0.78125,
      "learning_rate": 3.293150240547549e-10,
      "logits/chosen": -3.515069007873535,
      "logits/rejected": -5.043622970581055,
      "logps/chosen": -580.21044921875,
      "logps/ref_chosen": -35.802879333496094,
      "logps/ref_rejected": -100.60040283203125,
      "logps/rejected": -1485.48779296875,
      "loss": 1.0379,
      "rewards/accuracies": 0.8125,
      "rewards/chosen": -0.32212620973587036,
      "rewards/margins": 0.491289347410202,
      "rewards/rejected": -0.8134155869483948,
      "step": 672
    },
    {
      "epoch": 0.9882525697503671,
      "epsilon_dpo/beta": 0.0005843861144967377,
      "epsilon_dpo/beta_margin_grad_mean": -0.3730253279209137,
      "epsilon_dpo/beta_margin_grad_std": 0.10759243369102478,
      "epsilon_dpo/beta_margin_mean": 0.5548798441886902,
      "epsilon_dpo/beta_margin_std": 0.5018904209136963,
      "epsilon_dpo/loss_margin_mean": 952.566162109375,
      "grad_norm": 38.58977508544922,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0005885579157620668,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 2.6675629940689504e-10,
      "logits/chosen": -3.419285297393799,
      "logits/rejected": -5.03121280670166,
      "logps/chosen": -399.2283935546875,
      "logps/ref_chosen": -26.926271438598633,
      "logps/ref_rejected": -91.83390808105469,
      "logps/rejected": -1416.7021484375,
      "loss": 0.963,
      "rewards/accuracies": 0.9375,
      "rewards/chosen": -0.2182503640651703,
      "rewards/margins": 0.5548798441886902,
      "rewards/rejected": -0.7731301784515381,
      "step": 673
    },
    {
      "epoch": 0.9897209985315712,
      "epsilon_dpo/beta": 0.0005809462745673954,
      "epsilon_dpo/beta_margin_grad_mean": -0.392082154750824,
      "epsilon_dpo/beta_margin_grad_std": 0.13613100349903107,
      "epsilon_dpo/beta_margin_mean": 0.47949162125587463,
      "epsilon_dpo/beta_margin_std": 0.6231409311294556,
      "epsilon_dpo/loss_margin_mean": 831.1757202148438,
      "grad_norm": 32.84586715698242,
      "kl/avg_steps": 0.59375,
      "kl/beta": 0.0005843578255735338,
      "kl/n_epsilon_steps": 0.203125,
      "kl/p_epsilon_steps": 0.796875,
      "learning_rate": 2.1077827798404725e-10,
      "logits/chosen": -3.5797486305236816,
      "logits/rejected": -5.058620452880859,
      "logps/chosen": -462.48150634765625,
      "logps/ref_chosen": -32.161155700683594,
      "logps/ref_rejected": -73.33118438720703,
      "logps/rejected": -1334.8271484375,
      "loss": 1.0513,
      "rewards/accuracies": 0.796875,
      "rewards/chosen": -0.25154006481170654,
      "rewards/margins": 0.479491651058197,
      "rewards/rejected": -0.7310316562652588,
      "step": 674
    },
    {
      "epoch": 0.9911894273127754,
      "epsilon_dpo/beta": 0.0005764279630966485,
      "epsilon_dpo/beta_margin_grad_mean": -0.37939831614494324,
      "epsilon_dpo/beta_margin_grad_std": 0.11590909957885742,
      "epsilon_dpo/beta_margin_mean": 0.5321040749549866,
      "epsilon_dpo/beta_margin_std": 0.5455756187438965,
      "epsilon_dpo/loss_margin_mean": 925.9025268554688,
      "grad_norm": 30.256996154785156,
      "kl/avg_steps": 0.78125,
      "kl/beta": 0.0005809086724184453,
      "kl/n_epsilon_steps": 0.109375,
      "kl/p_epsilon_steps": 0.890625,
      "learning_rate": 1.6138243485910863e-10,
      "logits/chosen": -3.4077110290527344,
      "logits/rejected": -5.058513641357422,
      "logps/chosen": -380.3410339355469,
      "logps/ref_chosen": -26.25046157836914,
      "logps/ref_rejected": -80.57308959960938,
      "logps/rejected": -1360.566162109375,
      "loss": 0.9896,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.20505669713020325,
      "rewards/margins": 0.5321040749549866,
      "rewards/rejected": -0.7371607422828674,
      "step": 675
    },
    {
      "epoch": 0.9926578560939795,
      "epsilon_dpo/beta": 0.0005732206045649946,
      "epsilon_dpo/beta_margin_grad_mean": -0.39650610089302063,
      "epsilon_dpo/beta_margin_grad_std": 0.13006651401519775,
      "epsilon_dpo/beta_margin_mean": 0.45457127690315247,
      "epsilon_dpo/beta_margin_std": 0.583877682685852,
      "epsilon_dpo/loss_margin_mean": 798.5032958984375,
      "grad_norm": 36.964908599853516,
      "kl/avg_steps": 0.5625,
      "kl/beta": 0.0005764055531471968,
      "kl/n_epsilon_steps": 0.21875,
      "kl/p_epsilon_steps": 0.78125,
      "learning_rate": 1.1857007165852472e-10,
      "logits/chosen": -3.637737512588501,
      "logits/rejected": -4.942368507385254,
      "logps/chosen": -621.6793823242188,
      "logps/ref_chosen": -44.59165954589844,
      "logps/ref_rejected": -96.4842300415039,
      "logps/rejected": -1472.0751953125,
      "loss": 1.0612,
      "rewards/accuracies": 0.828125,
      "rewards/chosen": -0.332570344209671,
      "rewards/margins": 0.4545712471008301,
      "rewards/rejected": -0.7871416211128235,
      "step": 676
    },
    {
      "epoch": 0.9941262848751835,
      "epsilon_dpo/beta": 0.0005685811047442257,
      "epsilon_dpo/beta_margin_grad_mean": -0.3787325918674469,
      "epsilon_dpo/beta_margin_grad_std": 0.11398439854383469,
      "epsilon_dpo/beta_margin_mean": 0.5294075012207031,
      "epsilon_dpo/beta_margin_std": 0.5406895279884338,
      "epsilon_dpo/loss_margin_mean": 933.8375244140625,
      "grad_norm": 35.77093505859375,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.0005731813726015389,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 8.23423165278725e-11,
      "logits/chosen": -3.3925907611846924,
      "logits/rejected": -5.077644348144531,
      "logps/chosen": -438.6134948730469,
      "logps/ref_chosen": -29.166034698486328,
      "logps/ref_rejected": -84.22445678710938,
      "logps/rejected": -1427.509521484375,
      "loss": 0.9904,
      "rewards/accuracies": 0.921875,
      "rewards/chosen": -0.23343393206596375,
      "rewards/margins": 0.5294075012207031,
      "rewards/rejected": -0.7628414630889893,
      "step": 677
    },
    {
      "epoch": 0.9955947136563876,
      "epsilon_dpo/beta": 0.0005647094221785665,
      "epsilon_dpo/beta_margin_grad_mean": -0.3867654800415039,
      "epsilon_dpo/beta_margin_grad_std": 0.12935031950473785,
      "epsilon_dpo/beta_margin_mean": 0.4952041506767273,
      "epsilon_dpo/beta_margin_std": 0.5945535898208618,
      "epsilon_dpo/loss_margin_mean": 881.53369140625,
      "grad_norm": 34.48164749145508,
      "kl/avg_steps": 0.6875,
      "kl/beta": 0.0005685618380084634,
      "kl/n_epsilon_steps": 0.15625,
      "kl/p_epsilon_steps": 0.84375,
      "learning_rate": 5.270012410216185e-11,
      "logits/chosen": -3.5082788467407227,
      "logits/rejected": -5.051633358001709,
      "logps/chosen": -463.51409912109375,
      "logps/ref_chosen": -31.562625885009766,
      "logps/ref_rejected": -86.20599365234375,
      "logps/rejected": -1399.691162109375,
      "loss": 1.0318,
      "rewards/accuracies": 0.859375,
      "rewards/chosen": -0.24528416991233826,
      "rewards/margins": 0.4952041506767273,
      "rewards/rejected": -0.7404882907867432,
      "step": 678
    },
    {
      "epoch": 0.9970631424375918,
      "epsilon_dpo/beta": 0.0005620888550765812,
      "epsilon_dpo/beta_margin_grad_mean": -0.4052530825138092,
      "epsilon_dpo/beta_margin_grad_std": 0.11750102043151855,
      "epsilon_dpo/beta_margin_mean": 0.41717761754989624,
      "epsilon_dpo/beta_margin_std": 0.5460913777351379,
      "epsilon_dpo/loss_margin_mean": 747.7733154296875,
      "grad_norm": 37.97880935668945,
      "kl/avg_steps": 0.46875,
      "kl/beta": 0.0005646796198561788,
      "kl/n_epsilon_steps": 0.265625,
      "kl/p_epsilon_steps": 0.734375,
      "learning_rate": 2.9644275480772416e-11,
      "logits/chosen": -3.43449068069458,
      "logits/rejected": -4.8752970695495605,
      "logps/chosen": -521.1033935546875,
      "logps/ref_chosen": -35.110084533691406,
      "logps/ref_rejected": -82.25491333007812,
      "logps/rejected": -1316.021484375,
      "loss": 1.0791,
      "rewards/accuracies": 0.796875,
      "rewards/chosen": -0.2744186222553253,
      "rewards/margins": 0.41717761754989624,
      "rewards/rejected": -0.691596269607544,
      "step": 679
    },
    {
      "epoch": 0.9985315712187959,
      "epsilon_dpo/beta": 0.0005580611759796739,
      "epsilon_dpo/beta_margin_grad_mean": -0.360996276140213,
      "epsilon_dpo/beta_margin_grad_std": 0.11674246937036514,
      "epsilon_dpo/beta_margin_mean": 0.621047854423523,
      "epsilon_dpo/beta_margin_std": 0.5679141879081726,
      "epsilon_dpo/loss_margin_mean": 1116.3912353515625,
      "grad_norm": 45.923912048339844,
      "kl/avg_steps": 0.71875,
      "kl/beta": 0.0005620450829155743,
      "kl/n_epsilon_steps": 0.140625,
      "kl/p_epsilon_steps": 0.859375,
      "learning_rate": 1.31753782067201e-11,
      "logits/chosen": -3.399845600128174,
      "logits/rejected": -4.949437618255615,
      "logps/chosen": -484.4205017089844,
      "logps/ref_chosen": -49.117393493652344,
      "logps/ref_rejected": -118.43990325927734,
      "logps/rejected": -1670.13427734375,
      "loss": 0.9288,
      "rewards/accuracies": 0.890625,
      "rewards/chosen": -0.24402347207069397,
      "rewards/margins": 0.621047854423523,
      "rewards/rejected": -0.8650712966918945,
      "step": 680
    },
    {
      "epoch": 1.0,
      "epsilon_dpo/beta": 0.0005535554955713451,
      "epsilon_dpo/beta_margin_grad_mean": -0.37510794401168823,
      "epsilon_dpo/beta_margin_grad_std": 0.11336696147918701,
      "epsilon_dpo/beta_margin_mean": 0.5475642085075378,
      "epsilon_dpo/beta_margin_std": 0.535968005657196,
      "epsilon_dpo/loss_margin_mean": 992.2255859375,
      "grad_norm": 39.61741638183594,
      "kl/avg_steps": 0.8125,
      "kl/beta": 0.00055803416762501,
      "kl/n_epsilon_steps": 0.09375,
      "kl/p_epsilon_steps": 0.90625,
      "learning_rate": 3.2938662507808745e-12,
      "logits/chosen": -3.4865760803222656,
      "logits/rejected": -5.149275302886963,
      "logps/chosen": -487.08197021484375,
      "logps/ref_chosen": -40.42382049560547,
      "logps/ref_rejected": -94.08821105957031,
      "logps/rejected": -1532.971923828125,
      "loss": 0.9757,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -0.2478674352169037,
      "rewards/margins": 0.5475642085075378,
      "rewards/rejected": -0.7954316735267639,
      "step": 681
    },
    {
      "epoch": 1.0,
      "step": 681,
      "total_flos": 0.0,
      "train_loss": 0.6833117403385398,
      "train_runtime": 3053.5655,
      "train_samples_per_second": 14.278,
      "train_steps_per_second": 0.223
    }
  ],
  "logging_steps": 1,
  "max_steps": 681,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 200,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}