{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.0,
  "eval_steps": 100,
  "global_step": 340,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "beta_dpo/beta_used": 0.10121209919452667,
      "beta_dpo/beta_used_raw": 0.10121209919452667,
      "beta_dpo/gap_mean": -0.0009442940354347229,
      "beta_dpo/gap_std": 0.03691839799284935,
      "beta_dpo/mask_keep_frac": 0.9375,
      "epoch": 0.0029411764705882353,
      "grad_norm": 23.302410125732422,
      "learning_rate": 0.0,
      "logits/chosen": -0.4739703834056854,
      "logits/rejected": -0.44689586758613586,
      "loss": 0.6919,
      "step": 1
    },
    {
      "beta_dpo/beta_used": 0.10032124072313309,
      "beta_dpo/beta_used_raw": 0.10032124072313309,
      "beta_dpo/gap_mean": -0.0016960372449830174,
      "beta_dpo/gap_std": 0.1151522547006607,
      "beta_dpo/mask_keep_frac": 0.765625,
      "epoch": 0.014705882352941176,
      "grad_norm": 24.834075927734375,
      "learning_rate": 5.88235294117647e-08,
      "logits/chosen": -0.49943581223487854,
      "logits/rejected": -0.4934660494327545,
      "loss": 0.693,
      "step": 5
    },
    {
      "beta_dpo/beta_used": 0.101251520216465,
      "beta_dpo/beta_used_raw": 0.101251520216465,
      "beta_dpo/gap_mean": 0.0030363830737769604,
      "beta_dpo/gap_std": 0.2163175642490387,
      "beta_dpo/mask_keep_frac": 0.824999988079071,
      "epoch": 0.029411764705882353,
      "grad_norm": 21.942047119140625,
      "learning_rate": 1.3235294117647057e-07,
      "logits/chosen": -0.5174359083175659,
      "logits/rejected": -0.5005401968955994,
      "loss": 0.692,
      "step": 10
    },
    {
      "beta_dpo/beta_used": 0.10108586400747299,
      "beta_dpo/beta_used_raw": 0.10108586400747299,
      "beta_dpo/gap_mean": 0.024518460035324097,
      "beta_dpo/gap_std": 0.2784799039363861,
      "beta_dpo/mask_keep_frac": 0.824999988079071,
      "epoch": 0.04411764705882353,
      "grad_norm": 28.207460403442383,
      "learning_rate": 2.0588235294117645e-07,
      "logits/chosen": -0.5348216891288757,
      "logits/rejected": -0.5156930088996887,
      "loss": 0.6911,
      "step": 15
    },
    {
      "beta_dpo/beta_used": 0.10244777053594589,
      "beta_dpo/beta_used_raw": 0.10244777053594589,
      "beta_dpo/gap_mean": 0.0749056339263916,
      "beta_dpo/gap_std": 0.33879655599594116,
      "beta_dpo/mask_keep_frac": 0.8125,
      "epoch": 0.058823529411764705,
      "grad_norm": 19.902040481567383,
      "learning_rate": 2.7941176470588235e-07,
      "logits/chosen": -0.5660465955734253,
      "logits/rejected": -0.5419166088104248,
      "loss": 0.6874,
      "step": 20
    },
    {
      "beta_dpo/beta_used": 0.10622622072696686,
      "beta_dpo/beta_used_raw": 0.10622622072696686,
      "beta_dpo/gap_mean": 0.20916345715522766,
      "beta_dpo/gap_std": 0.456662118434906,
      "beta_dpo/mask_keep_frac": 0.887499988079071,
      "epoch": 0.07352941176470588,
      "grad_norm": 22.522640228271484,
      "learning_rate": 3.529411764705882e-07,
      "logits/chosen": -0.5110368132591248,
      "logits/rejected": -0.5050845146179199,
      "loss": 0.6769,
      "step": 25
    },
    {
      "beta_dpo/beta_used": 0.10997174680233002,
      "beta_dpo/beta_used_raw": 0.10997174680233002,
      "beta_dpo/gap_mean": 0.5209842920303345,
      "beta_dpo/gap_std": 0.7702666521072388,
      "beta_dpo/mask_keep_frac": 0.762499988079071,
      "epoch": 0.08823529411764706,
      "grad_norm": 19.37394142150879,
      "learning_rate": 4.264705882352941e-07,
      "logits/chosen": -0.5535926222801208,
      "logits/rejected": -0.5316442251205444,
      "loss": 0.6574,
      "step": 30
    },
    {
      "beta_dpo/beta_used": 0.11611036211252213,
      "beta_dpo/beta_used_raw": 0.11611036211252213,
      "beta_dpo/gap_mean": 0.9489548802375793,
      "beta_dpo/gap_std": 1.3326656818389893,
      "beta_dpo/mask_keep_frac": 0.6625000238418579,
      "epoch": 0.10294117647058823,
      "grad_norm": 23.465280532836914,
      "learning_rate": 5e-07,
      "logits/chosen": -0.5605362057685852,
      "logits/rejected": -0.5497816801071167,
      "loss": 0.6265,
      "step": 35
    },
    {
      "beta_dpo/beta_used": 0.1326821744441986,
      "beta_dpo/beta_used_raw": 0.1326821744441986,
      "beta_dpo/gap_mean": 1.789758324623108,
      "beta_dpo/gap_std": 2.447655200958252,
      "beta_dpo/mask_keep_frac": 0.7749999761581421,
      "epoch": 0.11764705882352941,
      "grad_norm": 20.867738723754883,
      "learning_rate": 4.996706849759452e-07,
      "logits/chosen": -0.6393685340881348,
      "logits/rejected": -0.6073721051216125,
      "loss": 0.5663,
      "step": 40
    },
    {
      "beta_dpo/beta_used": 0.12091531604528427,
      "beta_dpo/beta_used_raw": 0.12091531604528427,
      "beta_dpo/gap_mean": 2.8082375526428223,
      "beta_dpo/gap_std": 4.084892272949219,
      "beta_dpo/mask_keep_frac": 0.7875000238418579,
      "epoch": 0.1323529411764706,
      "grad_norm": 19.109943389892578,
      "learning_rate": 4.986836074908615e-07,
      "logits/chosen": -0.6684064865112305,
      "logits/rejected": -0.6361075639724731,
      "loss": 0.5411,
      "step": 45
    },
    {
      "beta_dpo/beta_used": 0.11514081805944443,
      "beta_dpo/beta_used_raw": 0.11514081805944443,
      "beta_dpo/gap_mean": 3.5657267570495605,
      "beta_dpo/gap_std": 5.662721633911133,
      "beta_dpo/mask_keep_frac": 0.75,
      "epoch": 0.14705882352941177,
      "grad_norm": 22.54947853088379,
      "learning_rate": 4.970413680203148e-07,
      "logits/chosen": -0.6647295951843262,
      "logits/rejected": -0.6276803612709045,
      "loss": 0.536,
      "step": 50
    },
    {
      "beta_dpo/beta_used": 0.13137957453727722,
      "beta_dpo/beta_used_raw": 0.13137957453727722,
      "beta_dpo/gap_mean": 4.35926628112793,
      "beta_dpo/gap_std": 7.092940330505371,
      "beta_dpo/mask_keep_frac": 0.7250000238418579,
      "epoch": 0.16176470588235295,
      "grad_norm": 38.691123962402344,
      "learning_rate": 4.947482930773511e-07,
      "logits/chosen": -0.7010586261749268,
      "logits/rejected": -0.675391435623169,
      "loss": 0.493,
      "step": 55
    },
    {
      "beta_dpo/beta_used": 0.094205841422081,
      "beta_dpo/beta_used_raw": 0.094205841422081,
      "beta_dpo/gap_mean": 5.047989845275879,
      "beta_dpo/gap_std": 8.23731803894043,
      "beta_dpo/mask_keep_frac": 0.7250000238418579,
      "epoch": 0.17647058823529413,
      "grad_norm": 31.127901077270508,
      "learning_rate": 4.918104238142103e-07,
      "logits/chosen": -0.724422812461853,
      "logits/rejected": -0.6809322237968445,
      "loss": 0.5315,
      "step": 60
    },
    {
      "beta_dpo/beta_used": 0.11677428334951401,
      "beta_dpo/beta_used_raw": 0.11677428334951401,
      "beta_dpo/gap_mean": 5.827352523803711,
      "beta_dpo/gap_std": 8.861337661743164,
      "beta_dpo/mask_keep_frac": 0.8500000238418579,
      "epoch": 0.19117647058823528,
      "grad_norm": 20.432043075561523,
      "learning_rate": 4.882355001067891e-07,
      "logits/chosen": -0.6648474931716919,
      "logits/rejected": -0.637535572052002,
      "loss": 0.4741,
      "step": 65
    },
    {
      "beta_dpo/beta_used": 0.09036926180124283,
      "beta_dpo/beta_used_raw": 0.09036926180124283,
      "beta_dpo/gap_mean": 6.462141990661621,
      "beta_dpo/gap_std": 9.157753944396973,
      "beta_dpo/mask_keep_frac": 0.7875000238418579,
      "epoch": 0.20588235294117646,
      "grad_norm": 24.550621032714844,
      "learning_rate": 4.840329401637809e-07,
      "logits/chosen": -0.6986874341964722,
      "logits/rejected": -0.6637295484542847,
      "loss": 0.5026,
      "step": 70
    },
    {
      "beta_dpo/beta_used": 0.07552285492420197,
      "beta_dpo/beta_used_raw": 0.0741354450583458,
      "beta_dpo/gap_mean": 6.905457496643066,
      "beta_dpo/gap_std": 9.706171035766602,
      "beta_dpo/mask_keep_frac": 0.8125,
      "epoch": 0.22058823529411764,
      "grad_norm": 21.727449417114258,
      "learning_rate": 4.792138157142157e-07,
      "logits/chosen": -0.7469242215156555,
      "logits/rejected": -0.7223338484764099,
      "loss": 0.5172,
      "step": 75
    },
    {
      "beta_dpo/beta_used": 0.09862758219242096,
      "beta_dpo/beta_used_raw": 0.09059171378612518,
      "beta_dpo/gap_mean": 7.501389980316162,
      "beta_dpo/gap_std": 10.180580139160156,
      "beta_dpo/mask_keep_frac": 0.8374999761581421,
      "epoch": 0.23529411764705882,
      "grad_norm": 0.3042762279510498,
      "learning_rate": 4.737908228387656e-07,
      "logits/chosen": -0.7543509602546692,
      "logits/rejected": -0.7017374038696289,
      "loss": 0.4756,
      "step": 80
    },
    {
      "beta_dpo/beta_used": 0.06848205626010895,
      "beta_dpo/beta_used_raw": 0.05231575295329094,
      "beta_dpo/gap_mean": 7.9440507888793945,
      "beta_dpo/gap_std": 10.780364990234375,
      "beta_dpo/mask_keep_frac": 0.800000011920929,
      "epoch": 0.25,
      "grad_norm": 0.3057352602481842,
      "learning_rate": 4.6777824852166437e-07,
      "logits/chosen": -0.7203555107116699,
      "logits/rejected": -0.6912198066711426,
      "loss": 0.5345,
      "step": 85
    },
    {
      "beta_dpo/beta_used": 0.08253253251314163,
      "beta_dpo/beta_used_raw": 0.08253253251314163,
      "beta_dpo/gap_mean": 8.4508638381958,
      "beta_dpo/gap_std": 11.448507308959961,
      "beta_dpo/mask_keep_frac": 0.875,
      "epoch": 0.2647058823529412,
      "grad_norm": 21.29926300048828,
      "learning_rate": 4.611919330113591e-07,
      "logits/chosen": -0.6781951189041138,
      "logits/rejected": -0.6568866968154907,
      "loss": 0.4944,
      "step": 90
    },
    {
      "beta_dpo/beta_used": 0.05684714391827583,
      "beta_dpo/beta_used_raw": 0.05684714391827583,
      "beta_dpo/gap_mean": 8.875980377197266,
      "beta_dpo/gap_std": 12.020231246948242,
      "beta_dpo/mask_keep_frac": 0.7875000238418579,
      "epoch": 0.27941176470588236,
      "grad_norm": 15.502776145935059,
      "learning_rate": 4.5404922808905543e-07,
      "logits/chosen": -0.7086650729179382,
      "logits/rejected": -0.6651682257652283,
      "loss": 0.5335,
      "step": 95
    },
    {
      "beta_dpo/beta_used": 0.1179933100938797,
      "beta_dpo/beta_used_raw": 0.1179933100938797,
      "beta_dpo/gap_mean": 9.62360954284668,
      "beta_dpo/gap_std": 12.684171676635742,
      "beta_dpo/mask_keep_frac": 0.762499988079071,
      "epoch": 0.29411764705882354,
      "grad_norm": 34.129478454589844,
      "learning_rate": 4.4636895135509966e-07,
      "logits/chosen": -0.6843082904815674,
      "logits/rejected": -0.660758912563324,
      "loss": 0.4098,
      "step": 100
    },
    {
      "epoch": 0.29411764705882354,
      "eval_beta_dpo/beta_used": 0.04330332204699516,
      "eval_beta_dpo/beta_used_raw": 0.01594320312142372,
      "eval_beta_dpo/gap_mean": 7.997772216796875,
      "eval_beta_dpo/gap_std": 13.260690689086914,
      "eval_beta_dpo/mask_keep_frac": 1.0,
      "eval_logits/chosen": -0.6977978944778442,
      "eval_logits/rejected": -0.6668843626976013,
      "eval_loss": 0.6251118183135986,
      "eval_runtime": 20.4115,
      "eval_samples_per_second": 114.592,
      "eval_steps_per_second": 0.931,
      "step": 100
    },
    {
      "beta_dpo/beta_used": 0.15054886043071747,
      "beta_dpo/beta_used_raw": 0.15054886043071747,
      "beta_dpo/gap_mean": 7.8849334716796875,
      "beta_dpo/gap_std": 13.30543041229248,
      "beta_dpo/mask_keep_frac": 0.762499988079071,
      "epoch": 0.3088235294117647,
      "grad_norm": 15.665854454040527,
      "learning_rate": 4.381713366536311e-07,
      "logits/chosen": -0.7553393244743347,
      "logits/rejected": -0.710943341255188,
      "loss": 0.4061,
      "step": 105
    },
    {
      "beta_dpo/beta_used": 0.09847154468297958,
      "beta_dpo/beta_used_raw": 0.09847154468297958,
      "beta_dpo/gap_mean": 9.127924919128418,
      "beta_dpo/gap_std": 13.331835746765137,
      "beta_dpo/mask_keep_frac": 0.75,
      "epoch": 0.3235294117647059,
      "grad_norm": 7.982070446014404,
      "learning_rate": 4.2947798076611047e-07,
      "logits/chosen": -0.7284727692604065,
      "logits/rejected": -0.696673572063446,
      "loss": 0.4595,
      "step": 110
    },
    {
      "beta_dpo/beta_used": 0.12598751485347748,
      "beta_dpo/beta_used_raw": 0.12598751485347748,
      "beta_dpo/gap_mean": 9.942410469055176,
      "beta_dpo/gap_std": 13.166864395141602,
      "beta_dpo/mask_keep_frac": 0.762499988079071,
      "epoch": 0.3382352941176471,
      "grad_norm": 36.213523864746094,
      "learning_rate": 4.203117865141635e-07,
      "logits/chosen": -0.7145182490348816,
      "logits/rejected": -0.6985291242599487,
      "loss": 0.3778,
      "step": 115
    },
    {
      "beta_dpo/beta_used": 0.05509430170059204,
      "beta_dpo/beta_used_raw": 0.048566654324531555,
      "beta_dpo/gap_mean": 10.542096138000488,
      "beta_dpo/gap_std": 13.39216136932373,
      "beta_dpo/mask_keep_frac": 0.8125,
      "epoch": 0.35294117647058826,
      "grad_norm": 14.08332347869873,
      "learning_rate": 4.106969024216348e-07,
      "logits/chosen": -0.7127692103385925,
      "logits/rejected": -0.6740670204162598,
      "loss": 0.5271,
      "step": 120
    },
    {
      "beta_dpo/beta_used": 0.05550508573651314,
      "beta_dpo/beta_used_raw": 0.05064947530627251,
      "beta_dpo/gap_mean": 11.009790420532227,
      "beta_dpo/gap_std": 13.461648941040039,
      "beta_dpo/mask_keep_frac": 0.824999988079071,
      "epoch": 0.36764705882352944,
      "grad_norm": 1.2771987915039062,
      "learning_rate": 4.006586590948141e-07,
      "logits/chosen": -0.7180671691894531,
      "logits/rejected": -0.6869726777076721,
      "loss": 0.513,
      "step": 125
    },
    {
      "beta_dpo/beta_used": 0.05528440326452255,
      "beta_dpo/beta_used_raw": 0.05528440326452255,
      "beta_dpo/gap_mean": 11.50378704071045,
      "beta_dpo/gap_std": 14.039319038391113,
      "beta_dpo/mask_keep_frac": 0.7749999761581421,
      "epoch": 0.38235294117647056,
      "grad_norm": 5.609388828277588,
      "learning_rate": 3.9022350248844246e-07,
      "logits/chosen": -0.7099085450172424,
      "logits/rejected": -0.6715607643127441,
      "loss": 0.5068,
      "step": 130
    },
    {
      "beta_dpo/beta_used": 0.08324670791625977,
      "beta_dpo/beta_used_raw": 0.08324670791625977,
      "beta_dpo/gap_mean": 12.224153518676758,
      "beta_dpo/gap_std": 15.014795303344727,
      "beta_dpo/mask_keep_frac": 0.7749999761581421,
      "epoch": 0.39705882352941174,
      "grad_norm": 22.845937728881836,
      "learning_rate": 3.794189242333106e-07,
      "logits/chosen": -0.7034512758255005,
      "logits/rejected": -0.6600346565246582,
      "loss": 0.4231,
      "step": 135
    },
    {
      "beta_dpo/beta_used": 0.10875506699085236,
      "beta_dpo/beta_used_raw": 0.10875506699085236,
      "beta_dpo/gap_mean": 13.073277473449707,
      "beta_dpo/gap_std": 15.834657669067383,
      "beta_dpo/mask_keep_frac": 0.75,
      "epoch": 0.4117647058823529,
      "grad_norm": 31.461519241333008,
      "learning_rate": 3.6827338920900253e-07,
      "logits/chosen": -0.6739553213119507,
      "logits/rejected": -0.6366498470306396,
      "loss": 0.3902,
      "step": 140
    },
    {
      "beta_dpo/beta_used": 0.10356837511062622,
      "beta_dpo/beta_used_raw": 0.10356837511062622,
      "beta_dpo/gap_mean": 13.881492614746094,
      "beta_dpo/gap_std": 16.42840003967285,
      "beta_dpo/mask_keep_frac": 0.7250000238418579,
      "epoch": 0.4264705882352941,
      "grad_norm": 29.94761085510254,
      "learning_rate": 3.568162605525952e-07,
      "logits/chosen": -0.7220578193664551,
      "logits/rejected": -0.6809359788894653,
      "loss": 0.3819,
      "step": 145
    },
    {
      "beta_dpo/beta_used": 0.019714761525392532,
      "beta_dpo/beta_used_raw": -0.023198971524834633,
      "beta_dpo/gap_mean": 14.227258682250977,
      "beta_dpo/gap_std": 17.448183059692383,
      "beta_dpo/mask_keep_frac": 0.862500011920929,
      "epoch": 0.4411764705882353,
      "grad_norm": 5.527612209320068,
      "learning_rate": 3.4507772230088147e-07,
      "logits/chosen": -0.6326473355293274,
      "logits/rejected": -0.6032054424285889,
      "loss": 0.607,
      "step": 150
    },
    {
      "beta_dpo/beta_used": 0.07979521155357361,
      "beta_dpo/beta_used_raw": 0.06505511701107025,
      "beta_dpo/gap_mean": 14.670598983764648,
      "beta_dpo/gap_std": 18.554828643798828,
      "beta_dpo/mask_keep_frac": 0.699999988079071,
      "epoch": 0.45588235294117646,
      "grad_norm": 23.10860824584961,
      "learning_rate": 3.3308869986991487e-07,
      "logits/chosen": -0.7037164568901062,
      "logits/rejected": -0.6613154411315918,
      "loss": 0.4478,
      "step": 155
    },
    {
      "beta_dpo/beta_used": 0.06584476679563522,
      "beta_dpo/beta_used_raw": 0.040607184171676636,
      "beta_dpo/gap_mean": 15.389450073242188,
      "beta_dpo/gap_std": 19.081418991088867,
      "beta_dpo/mask_keep_frac": 0.8374999761581421,
      "epoch": 0.47058823529411764,
      "grad_norm": 31.426233291625977,
      "learning_rate": 3.208807785813777e-07,
      "logits/chosen": -0.6574662923812866,
      "logits/rejected": -0.630233883857727,
      "loss": 0.4758,
      "step": 160
    },
    {
      "beta_dpo/beta_used": 0.08810704201459885,
      "beta_dpo/beta_used_raw": 0.07671914994716644,
      "beta_dpo/gap_mean": 16.22821617126465,
      "beta_dpo/gap_std": 19.637792587280273,
      "beta_dpo/mask_keep_frac": 0.800000011920929,
      "epoch": 0.4852941176470588,
      "grad_norm": 73.11759948730469,
      "learning_rate": 3.084861204504122e-07,
      "logits/chosen": -0.6618590354919434,
      "logits/rejected": -0.6243924498558044,
      "loss": 0.4768,
      "step": 165
    },
    {
      "beta_dpo/beta_used": 0.02722100354731083,
      "beta_dpo/beta_used_raw": 0.02408101223409176,
      "beta_dpo/gap_mean": 17.07744598388672,
      "beta_dpo/gap_std": 20.277606964111328,
      "beta_dpo/mask_keep_frac": 0.7875000238418579,
      "epoch": 0.5,
      "grad_norm": 0.5254238247871399,
      "learning_rate": 2.959373794541426e-07,
      "logits/chosen": -0.6966148614883423,
      "logits/rejected": -0.666491687297821,
      "loss": 0.5686,
      "step": 170
    },
    {
      "beta_dpo/beta_used": 0.0664793998003006,
      "beta_dpo/beta_used_raw": 0.03722615912556648,
      "beta_dpo/gap_mean": 17.654155731201172,
      "beta_dpo/gap_std": 21.08226776123047,
      "beta_dpo/mask_keep_frac": 0.8125,
      "epoch": 0.5147058823529411,
      "grad_norm": 0.5429490804672241,
      "learning_rate": 2.8326761550411346e-07,
      "logits/chosen": -0.6848665475845337,
      "logits/rejected": -0.6613831520080566,
      "loss": 0.4753,
      "step": 175
    },
    {
      "beta_dpo/beta_used": 0.024588093161582947,
      "beta_dpo/beta_used_raw": -0.008070843294262886,
      "beta_dpo/gap_mean": 17.7331485748291,
      "beta_dpo/gap_std": 22.08762550354004,
      "beta_dpo/mask_keep_frac": 0.824999988079071,
      "epoch": 0.5294117647058824,
      "grad_norm": 15.737401962280273,
      "learning_rate": 2.7051020734928443e-07,
      "logits/chosen": -0.6688377261161804,
      "logits/rejected": -0.6460214853286743,
      "loss": 0.5579,
      "step": 180
    },
    {
      "beta_dpo/beta_used": 0.08162590861320496,
      "beta_dpo/beta_used_raw": 0.05483890324831009,
      "beta_dpo/gap_mean": 18.408456802368164,
      "beta_dpo/gap_std": 22.61962890625,
      "beta_dpo/mask_keep_frac": 0.862500011920929,
      "epoch": 0.5441176470588235,
      "grad_norm": 0.506800651550293,
      "learning_rate": 2.5769876463904263e-07,
      "logits/chosen": -0.7215656042098999,
      "logits/rejected": -0.6699239611625671,
      "loss": 0.4855,
      "step": 185
    },
    {
      "beta_dpo/beta_used": 0.05661209672689438,
      "beta_dpo/beta_used_raw": 0.027242619544267654,
      "beta_dpo/gap_mean": 18.814666748046875,
      "beta_dpo/gap_std": 22.990680694580078,
      "beta_dpo/mask_keep_frac": 0.824999988079071,
      "epoch": 0.5588235294117647,
      "grad_norm": 26.899166107177734,
      "learning_rate": 2.4486703937790243e-07,
      "logits/chosen": -0.7019311785697937,
      "logits/rejected": -0.6498968005180359,
      "loss": 0.4949,
      "step": 190
    },
    {
      "beta_dpo/beta_used": 0.02816765382885933,
      "beta_dpo/beta_used_raw": 0.001962479902431369,
      "beta_dpo/gap_mean": 19.533567428588867,
      "beta_dpo/gap_std": 23.629451751708984,
      "beta_dpo/mask_keep_frac": 0.824999988079071,
      "epoch": 0.5735294117647058,
      "grad_norm": 11.008431434631348,
      "learning_rate": 2.320488370051681e-07,
      "logits/chosen": -0.7254117727279663,
      "logits/rejected": -0.6765154004096985,
      "loss": 0.5286,
      "step": 195
    },
    {
      "beta_dpo/beta_used": 0.0643467828631401,
      "beta_dpo/beta_used_raw": -0.0015767127042636275,
      "beta_dpo/gap_mean": 20.06936264038086,
      "beta_dpo/gap_std": 24.53436851501465,
      "beta_dpo/mask_keep_frac": 0.862500011920929,
      "epoch": 0.5882352941176471,
      "grad_norm": 78.41595458984375,
      "learning_rate": 2.192779273338215e-07,
      "logits/chosen": -0.6875912547111511,
      "logits/rejected": -0.6458339095115662,
      "loss": 0.5527,
      "step": 200
    },
    {
      "epoch": 0.5882352941176471,
      "eval_beta_dpo/beta_used": 0.03733323514461517,
      "eval_beta_dpo/beta_used_raw": -0.0800742357969284,
      "eval_beta_dpo/gap_mean": 17.105911254882812,
      "eval_beta_dpo/gap_std": 25.945871353149414,
      "eval_beta_dpo/mask_keep_frac": 1.0,
      "eval_logits/chosen": -0.6982784271240234,
      "eval_logits/rejected": -0.6586927771568298,
      "eval_loss": 0.6420564651489258,
      "eval_runtime": 20.3208,
      "eval_samples_per_second": 115.104,
      "eval_steps_per_second": 0.935,
      "step": 200
    },
    {
      "beta_dpo/beta_used": 0.18078216910362244,
      "beta_dpo/beta_used_raw": 0.18078216910362244,
      "beta_dpo/gap_mean": 16.670331954956055,
      "beta_dpo/gap_std": 27.034832000732422,
      "beta_dpo/mask_keep_frac": 0.887499988079071,
      "epoch": 0.6029411764705882,
      "grad_norm": 76.40715789794922,
      "learning_rate": 2.065879555832674e-07,
      "logits/chosen": -0.6358317136764526,
      "logits/rejected": -0.5890509486198425,
      "loss": 0.4449,
      "step": 205
    },
    {
      "beta_dpo/beta_used": 0.187847301363945,
      "beta_dpo/beta_used_raw": 0.187847301363945,
      "beta_dpo/gap_mean": 18.74222183227539,
      "beta_dpo/gap_std": 27.3233642578125,
      "beta_dpo/mask_keep_frac": 0.7875000238418579,
      "epoch": 0.6176470588235294,
      "grad_norm": 39.985557556152344,
      "learning_rate": 1.9401235374032425e-07,
      "logits/chosen": -0.6837745308876038,
      "logits/rejected": -0.6255474090576172,
      "loss": 0.4006,
      "step": 210
    },
    {
      "beta_dpo/beta_used": 0.0615837462246418,
      "beta_dpo/beta_used_raw": 0.053963758051395416,
      "beta_dpo/gap_mean": 20.168214797973633,
      "beta_dpo/gap_std": 27.281606674194336,
      "beta_dpo/mask_keep_frac": 0.862500011920929,
      "epoch": 0.6323529411764706,
      "grad_norm": 56.95214080810547,
      "learning_rate": 1.8158425248197928e-07,
      "logits/chosen": -0.5969057083129883,
      "logits/rejected": -0.5545859336853027,
      "loss": 0.5414,
      "step": 215
    },
    {
      "beta_dpo/beta_used": 0.031995899975299835,
      "beta_dpo/beta_used_raw": 0.008795802481472492,
      "beta_dpo/gap_mean": 20.327245712280273,
      "beta_dpo/gap_std": 26.49213218688965,
      "beta_dpo/mask_keep_frac": 0.8374999761581421,
      "epoch": 0.6470588235294118,
      "grad_norm": 0.543950080871582,
      "learning_rate": 1.6933639389195134e-07,
      "logits/chosen": -0.6841639280319214,
      "logits/rejected": -0.6511374711990356,
      "loss": 0.5464,
      "step": 220
    },
    {
      "beta_dpo/beta_used": 0.04347361996769905,
      "beta_dpo/beta_used_raw": 0.016606144607067108,
      "beta_dpo/gap_mean": 20.5634765625,
      "beta_dpo/gap_std": 25.834671020507812,
      "beta_dpo/mask_keep_frac": 0.7875000238418579,
      "epoch": 0.6617647058823529,
      "grad_norm": 29.658409118652344,
      "learning_rate": 1.573010452010098e-07,
      "logits/chosen": -0.6632441282272339,
      "logits/rejected": -0.6577039957046509,
      "loss": 0.4873,
      "step": 225
    },
    {
      "beta_dpo/beta_used": 0.06577815115451813,
      "beta_dpo/beta_used_raw": 0.05692853406071663,
      "beta_dpo/gap_mean": 21.10856819152832,
      "beta_dpo/gap_std": 25.58962059020996,
      "beta_dpo/mask_keep_frac": 0.737500011920929,
      "epoch": 0.6764705882352942,
      "grad_norm": 0.6241604685783386,
      "learning_rate": 1.4550991377830423e-07,
      "logits/chosen": -0.7060235738754272,
      "logits/rejected": -0.669354259967804,
      "loss": 0.4797,
      "step": 230
    },
    {
      "beta_dpo/beta_used": 0.06475953757762909,
      "beta_dpo/beta_used_raw": 0.023799167945981026,
      "beta_dpo/gap_mean": 21.435104370117188,
      "beta_dpo/gap_std": 25.414148330688477,
      "beta_dpo/mask_keep_frac": 0.800000011920929,
      "epoch": 0.6911764705882353,
      "grad_norm": 14.550406455993652,
      "learning_rate": 1.339940635976592e-07,
      "logits/chosen": -0.6889506578445435,
      "logits/rejected": -0.6716668009757996,
      "loss": 0.4357,
      "step": 235
    },
    {
      "beta_dpo/beta_used": 0.022588472813367844,
      "beta_dpo/beta_used_raw": 0.0025838587898761034,
      "beta_dpo/gap_mean": 21.869482040405273,
      "beta_dpo/gap_std": 25.504459381103516,
      "beta_dpo/mask_keep_frac": 0.737500011920929,
      "epoch": 0.7058823529411765,
      "grad_norm": 11.02522087097168,
      "learning_rate": 1.227838333989088e-07,
      "logits/chosen": -0.6110752820968628,
      "logits/rejected": -0.5741311311721802,
      "loss": 0.5265,
      "step": 240
    },
    {
      "beta_dpo/beta_used": 0.03154964745044708,
      "beta_dpo/beta_used_raw": -0.007365362253040075,
      "beta_dpo/gap_mean": 22.568851470947266,
      "beta_dpo/gap_std": 25.90200424194336,
      "beta_dpo/mask_keep_frac": 0.7875000238418579,
      "epoch": 0.7205882352941176,
      "grad_norm": 0.5983785390853882,
      "learning_rate": 1.1190875675987355e-07,
      "logits/chosen": -0.6300492286682129,
      "logits/rejected": -0.6109535098075867,
      "loss": 0.5448,
      "step": 245
    },
    {
      "beta_dpo/beta_used": 0.010828005149960518,
      "beta_dpo/beta_used_raw": -0.058729518204927444,
      "beta_dpo/gap_mean": 22.215688705444336,
      "beta_dpo/gap_std": 27.019912719726562,
      "beta_dpo/mask_keep_frac": 0.8374999761581421,
      "epoch": 0.7352941176470589,
      "grad_norm": 17.40310287475586,
      "learning_rate": 1.0139748428955333e-07,
      "logits/chosen": -0.6890392303466797,
      "logits/rejected": -0.629682183265686,
      "loss": 0.629,
      "step": 250
    },
    {
      "beta_dpo/beta_used": 0.09269052743911743,
      "beta_dpo/beta_used_raw": 0.08087030053138733,
      "beta_dpo/gap_mean": 22.695995330810547,
      "beta_dpo/gap_std": 27.621633529663086,
      "beta_dpo/mask_keep_frac": 0.8125,
      "epoch": 0.75,
      "grad_norm": 53.0207405090332,
      "learning_rate": 9.127770814751932e-08,
      "logits/chosen": -0.6670210361480713,
      "logits/rejected": -0.6118627786636353,
      "loss": 0.483,
      "step": 255
    },
    {
      "beta_dpo/beta_used": 0.06755250692367554,
      "beta_dpo/beta_used_raw": 0.03205912187695503,
      "beta_dpo/gap_mean": 23.055011749267578,
      "beta_dpo/gap_std": 28.25390625,
      "beta_dpo/mask_keep_frac": 0.762499988079071,
      "epoch": 0.7647058823529411,
      "grad_norm": 0.6237814426422119,
      "learning_rate": 8.15760890883607e-08,
      "logits/chosen": -0.6667768359184265,
      "logits/rejected": -0.6239995956420898,
      "loss": 0.4968,
      "step": 260
    },
    {
      "beta_dpo/beta_used": 0.11417696624994278,
      "beta_dpo/beta_used_raw": 0.1062905341386795,
      "beta_dpo/gap_mean": 22.97963523864746,
      "beta_dpo/gap_std": 28.165149688720703,
      "beta_dpo/mask_keep_frac": 0.762499988079071,
      "epoch": 0.7794117647058824,
      "grad_norm": 93.24835205078125,
      "learning_rate": 7.231818622338822e-08,
      "logits/chosen": -0.6425198316574097,
      "logits/rejected": -0.6141684651374817,
      "loss": 0.4401,
      "step": 265
    },
    {
      "beta_dpo/beta_used": 0.07874588668346405,
      "beta_dpo/beta_used_raw": -0.0038310796953737736,
      "beta_dpo/gap_mean": 23.209665298461914,
      "beta_dpo/gap_std": 28.643651962280273,
      "beta_dpo/mask_keep_frac": 0.8125,
      "epoch": 0.7941176470588235,
      "grad_norm": 12.020166397094727,
      "learning_rate": 6.352838968463919e-08,
      "logits/chosen": -0.6789681911468506,
      "logits/rejected": -0.6184022426605225,
      "loss": 0.4832,
      "step": 270
    },
    {
      "beta_dpo/beta_used": 0.05330665037035942,
      "beta_dpo/beta_used_raw": -0.023246586322784424,
      "beta_dpo/gap_mean": 24.013660430908203,
      "beta_dpo/gap_std": 29.33469009399414,
      "beta_dpo/mask_keep_frac": 0.7250000238418579,
      "epoch": 0.8088235294117647,
      "grad_norm": 0.6346384882926941,
      "learning_rate": 5.5229856368582376e-08,
      "logits/chosen": -0.6784375905990601,
      "logits/rejected": -0.6448493599891663,
      "loss": 0.5253,
      "step": 275
    },
    {
      "beta_dpo/beta_used": 0.05292302370071411,
      "beta_dpo/beta_used_raw": -0.010663707740604877,
      "beta_dpo/gap_mean": 24.447540283203125,
      "beta_dpo/gap_std": 29.648815155029297,
      "beta_dpo/mask_keep_frac": 0.762499988079071,
      "epoch": 0.8235294117647058,
      "grad_norm": 0.6082450151443481,
      "learning_rate": 4.7444448928806615e-08,
      "logits/chosen": -0.6179937720298767,
      "logits/rejected": -0.5764154195785522,
      "loss": 0.53,
      "step": 280
    },
    {
      "beta_dpo/beta_used": 0.007934780791401863,
      "beta_dpo/beta_used_raw": -0.07739663124084473,
      "beta_dpo/gap_mean": 24.31735610961914,
      "beta_dpo/gap_std": 29.43593406677246,
      "beta_dpo/mask_keep_frac": 0.762499988079071,
      "epoch": 0.8382352941176471,
      "grad_norm": 0.6881201863288879,
      "learning_rate": 4.019267817841834e-08,
      "logits/chosen": -0.6771946549415588,
      "logits/rejected": -0.6086295247077942,
      "loss": 0.6357,
      "step": 285
    },
    {
      "beta_dpo/beta_used": 0.06074627488851547,
      "beta_dpo/beta_used_raw": 0.009315362200140953,
      "beta_dpo/gap_mean": 24.635099411010742,
      "beta_dpo/gap_std": 30.013864517211914,
      "beta_dpo/mask_keep_frac": 0.862500011920929,
      "epoch": 0.8529411764705882,
      "grad_norm": 2.793721914291382,
      "learning_rate": 3.349364905389032e-08,
      "logits/chosen": -0.6379111409187317,
      "logits/rejected": -0.5973175764083862,
      "loss": 0.5345,
      "step": 290
    },
    {
      "beta_dpo/beta_used": 0.03756168484687805,
      "beta_dpo/beta_used_raw": -0.01273317076265812,
      "beta_dpo/gap_mean": 24.830781936645508,
      "beta_dpo/gap_std": 30.81571388244629,
      "beta_dpo/mask_keep_frac": 0.7875000238418579,
      "epoch": 0.8676470588235294,
      "grad_norm": 50.0855598449707,
      "learning_rate": 2.736501028272095e-08,
      "logits/chosen": -0.617714524269104,
      "logits/rejected": -0.6301193237304688,
      "loss": 0.5441,
      "step": 295
    },
    {
      "beta_dpo/beta_used": 0.044209837913513184,
      "beta_dpo/beta_used_raw": 0.004354533273726702,
      "beta_dpo/gap_mean": 24.904342651367188,
      "beta_dpo/gap_std": 31.082351684570312,
      "beta_dpo/mask_keep_frac": 0.75,
      "epoch": 0.8823529411764706,
      "grad_norm": 0.6376844048500061,
      "learning_rate": 2.1822907887504932e-08,
      "logits/chosen": -0.5964897274971008,
      "logits/rejected": -0.6028931736946106,
      "loss": 0.5831,
      "step": 300
    },
    {
      "epoch": 0.8823529411764706,
      "eval_beta_dpo/beta_used": 0.031669970601797104,
      "eval_beta_dpo/beta_used_raw": -0.11691396683454514,
      "eval_beta_dpo/gap_mean": 20.08871841430664,
      "eval_beta_dpo/gap_std": 30.078739166259766,
      "eval_beta_dpo/mask_keep_frac": 1.0,
      "eval_logits/chosen": -0.6625580191612244,
      "eval_logits/rejected": -0.6206780672073364,
      "eval_loss": 0.6427361965179443,
      "eval_runtime": 20.3459,
      "eval_samples_per_second": 114.962,
      "eval_steps_per_second": 0.934,
      "step": 300
    },
    {
      "beta_dpo/beta_used": 0.20903603732585907,
      "beta_dpo/beta_used_raw": 0.20903603732585907,
      "beta_dpo/gap_mean": 19.461414337158203,
      "beta_dpo/gap_std": 29.93111801147461,
      "beta_dpo/mask_keep_frac": 0.824999988079071,
      "epoch": 0.8970588235294118,
      "grad_norm": 89.4169692993164,
      "learning_rate": 1.6881942648911074e-08,
      "logits/chosen": -0.6793561577796936,
      "logits/rejected": -0.6282657384872437,
      "loss": 0.5221,
      "step": 305
    },
    {
      "beta_dpo/beta_used": 0.1465708315372467,
      "beta_dpo/beta_used_raw": 0.13552138209342957,
      "beta_dpo/gap_mean": 21.82315444946289,
      "beta_dpo/gap_std": 30.26885414123535,
      "beta_dpo/mask_keep_frac": 0.800000011920929,
      "epoch": 0.9117647058823529,
      "grad_norm": 0.6243640780448914,
      "learning_rate": 1.2555131639630567e-08,
      "logits/chosen": -0.5958537459373474,
      "logits/rejected": -0.5621305704116821,
      "loss": 0.4927,
      "step": 310
    },
    {
      "beta_dpo/beta_used": 0.06637457758188248,
      "beta_dpo/beta_used_raw": -0.010266167111694813,
      "beta_dpo/gap_mean": 23.013385772705078,
      "beta_dpo/gap_std": 30.935138702392578,
      "beta_dpo/mask_keep_frac": 0.7875000238418579,
      "epoch": 0.9264705882352942,
      "grad_norm": 0.5633993148803711,
      "learning_rate": 8.85387393063622e-09,
      "logits/chosen": -0.6413298845291138,
      "logits/rejected": -0.6052228808403015,
      "loss": 0.5593,
      "step": 315
    },
    {
      "beta_dpo/beta_used": 0.0010000000474974513,
      "beta_dpo/beta_used_raw": -0.06694652885198593,
      "beta_dpo/gap_mean": 22.677587509155273,
      "beta_dpo/gap_std": 31.181507110595703,
      "beta_dpo/mask_keep_frac": 0.875,
      "epoch": 0.9411764705882353,
      "grad_norm": 0.6485550999641418,
      "learning_rate": 5.7879205600998296e-09,
      "logits/chosen": -0.6589199304580688,
      "logits/rejected": -0.6012631058692932,
      "loss": 0.6832,
      "step": 320
    },
    {
      "beta_dpo/beta_used": 0.03462111949920654,
      "beta_dpo/beta_used_raw": -0.054052434861660004,
      "beta_dpo/gap_mean": 23.140369415283203,
      "beta_dpo/gap_std": 31.43625259399414,
      "beta_dpo/mask_keep_frac": 0.8500000238418579,
      "epoch": 0.9558823529411765,
      "grad_norm": 0.5991944670677185,
      "learning_rate": 3.3653488440851253e-09,
      "logits/chosen": -0.6647250652313232,
      "logits/rejected": -0.6088197231292725,
      "loss": 0.6068,
      "step": 325
    },
    {
      "beta_dpo/beta_used": 0.03790256381034851,
      "beta_dpo/beta_used_raw": 0.029860854148864746,
      "beta_dpo/gap_mean": 24.296361923217773,
      "beta_dpo/gap_std": 31.577083587646484,
      "beta_dpo/mask_keep_frac": 0.7124999761581421,
      "epoch": 0.9705882352941176,
      "grad_norm": 0.614765465259552,
      "learning_rate": 1.592541096695571e-09,
      "logits/chosen": -0.6624591946601868,
      "logits/rejected": -0.62751704454422,
      "loss": 0.4828,
      "step": 330
    },
    {
      "beta_dpo/beta_used": 0.02071220614016056,
      "beta_dpo/beta_used_raw": -0.03980039432644844,
      "beta_dpo/gap_mean": 24.669193267822266,
      "beta_dpo/gap_std": 30.901264190673828,
      "beta_dpo/mask_keep_frac": 0.8125,
      "epoch": 0.9852941176470589,
      "grad_norm": 25.446868896484375,
      "learning_rate": 4.741678157389739e-10,
      "logits/chosen": -0.6353505849838257,
      "logits/rejected": -0.590802788734436,
      "loss": 0.5653,
      "step": 335
    },
    {
      "beta_dpo/beta_used": 0.04662991315126419,
      "beta_dpo/beta_used_raw": -0.04880411922931671,
      "beta_dpo/gap_mean": 25.268396377563477,
      "beta_dpo/gap_std": 30.97623062133789,
      "beta_dpo/mask_keep_frac": 0.75,
      "epoch": 1.0,
      "grad_norm": 0.6792064309120178,
      "learning_rate": 1.31753782067201e-11,
      "logits/chosen": -0.6614812016487122,
      "logits/rejected": -0.6312215924263,
      "loss": 0.6244,
      "step": 340
    },
    {
      "epoch": 1.0,
      "step": 340,
      "total_flos": 0.0,
      "train_loss": 0.5267414394546958,
      "train_runtime": 1440.2657,
      "train_samples_per_second": 30.271,
      "train_steps_per_second": 0.236
    }
  ],
  "logging_steps": 5,
  "max_steps": 340,
  "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": 16,
  "trial_name": null,
  "trial_params": null
}