{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.999244142101285,
  "eval_steps": 200,
  "global_step": 661,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "KL/chosen_KL_mean": 0.02867889404296875,
      "KL/mean": 0.029354453086853027,
      "KL/rejected_KL_mean": 0.030029296875,
      "KL/std": 0.2071000635623932,
      "epoch": 0.0015117157974300832,
      "fcm_dpo/beta": 0.10407507419586182,
      "fcm_dpo/delta": 0.19971171021461487,
      "fcm_dpo/margin": -0.0013532638549804688,
      "fcm_dpo/q_t": 0.5000380277633667,
      "grad_norm": 29.368080139160156,
      "learning_rate": 0.0,
      "logits/chosen": 0.13337239623069763,
      "logits/rejected": 0.12492949515581131,
      "logps/chosen": -64.5841293334961,
      "logps/ref_chosen": -64.61280822753906,
      "logps/ref_rejected": -64.17195129394531,
      "logps/rejected": -64.14192199707031,
      "loss": 1.3866,
      "margin_dpo/margin_mean": -0.0013527870178222656,
      "margin_dpo/margin_std": 0.2561596930027008,
      "step": 1
    },
    {
      "KL/chosen_KL_mean": -0.00289154052734375,
      "KL/mean": -0.021616414189338684,
      "KL/rejected_KL_mean": -0.04033660888671875,
      "KL/std": 0.19624735414981842,
      "epoch": 0.0030234315948601664,
      "fcm_dpo/beta": 0.10614679753780365,
      "fcm_dpo/delta": 0.19520045816898346,
      "fcm_dpo/margin": 0.037450045347213745,
      "fcm_dpo/q_t": 0.49902579188346863,
      "grad_norm": 29.559764862060547,
      "learning_rate": 7.462686567164179e-09,
      "logits/chosen": 0.09414851665496826,
      "logits/rejected": 0.07363267242908478,
      "logps/chosen": -56.101890563964844,
      "logps/ref_chosen": -56.0989990234375,
      "logps/ref_rejected": -66.59971618652344,
      "logps/rejected": -66.64006042480469,
      "loss": 1.3824,
      "margin_dpo/margin_mean": 0.03744968771934509,
      "margin_dpo/margin_std": 0.27811938524246216,
      "step": 2
    },
    {
      "KL/chosen_KL_mean": -0.014759063720703125,
      "KL/mean": 0.016700327396392822,
      "KL/rejected_KL_mean": 0.04815673828125,
      "KL/std": 0.2423822283744812,
      "epoch": 0.0045351473922902496,
      "fcm_dpo/beta": 0.10821852833032608,
      "fcm_dpo/delta": 0.0,
      "fcm_dpo/margin": -0.06291508674621582,
      "fcm_dpo/q_t": 0.5017015337944031,
      "grad_norm": 33.8881950378418,
      "learning_rate": 1.4925373134328357e-08,
      "logits/chosen": 0.09933048486709595,
      "logits/rejected": 0.061286427080631256,
      "logps/chosen": -65.47201538085938,
      "logps/ref_chosen": -65.45726013183594,
      "logps/ref_rejected": -90.82853698730469,
      "logps/rejected": -90.78038024902344,
      "loss": 1.3934,
      "margin_dpo/margin_mean": -0.06291499733924866,
      "margin_dpo/margin_std": 0.32749801874160767,
      "step": 3
    },
    {
      "KL/chosen_KL_mean": 0.0326690673828125,
      "KL/mean": 0.010438680648803711,
      "KL/rejected_KL_mean": -0.011798858642578125,
      "KL/std": 0.2497587949037552,
      "epoch": 0.006046863189720333,
      "fcm_dpo/beta": 0.11485374718904495,
      "fcm_dpo/delta": 0.39507198333740234,
      "fcm_dpo/margin": 0.04446643590927124,
      "fcm_dpo/q_t": 0.4987691640853882,
      "grad_norm": 38.972347259521484,
      "learning_rate": 2.2388059701492534e-08,
      "logits/chosen": 0.10525606572628021,
      "logits/rejected": 0.08916881680488586,
      "logps/chosen": -76.8275146484375,
      "logps/ref_chosen": -76.86018371582031,
      "logps/ref_rejected": -79.91523742675781,
      "logps/rejected": -79.92703247070312,
      "loss": 1.3815,
      "margin_dpo/margin_mean": 0.044465720653533936,
      "margin_dpo/margin_std": 0.3332005739212036,
      "step": 4
    },
    {
      "KL/chosen_KL_mean": 0.00232696533203125,
      "KL/mean": -0.004125773906707764,
      "KL/rejected_KL_mean": -0.0105743408203125,
      "KL/std": 0.21744784712791443,
      "epoch": 0.007558578987150416,
      "fcm_dpo/beta": 0.11947908252477646,
      "fcm_dpo/delta": 0.19778770208358765,
      "fcm_dpo/margin": 0.012903928756713867,
      "fcm_dpo/q_t": 0.4996222257614136,
      "grad_norm": 35.309425354003906,
      "learning_rate": 2.9850746268656714e-08,
      "logits/chosen": 0.08232656121253967,
      "logits/rejected": 0.043468791991472244,
      "logps/chosen": -62.96901321411133,
      "logps/ref_chosen": -62.97134017944336,
      "logps/ref_rejected": -79.9192123413086,
      "logps/rejected": -79.9297866821289,
      "loss": 1.3851,
      "margin_dpo/margin_mean": 0.012904614210128784,
      "margin_dpo/margin_std": 0.3334115147590637,
      "step": 5
    },
    {
      "KL/chosen_KL_mean": -0.06997871398925781,
      "KL/mean": -0.020912393927574158,
      "KL/rejected_KL_mean": 0.028154373168945312,
      "KL/std": 0.25535884499549866,
      "epoch": 0.009070294784580499,
      "fcm_dpo/beta": 0.1218419224023819,
      "fcm_dpo/delta": 0.0,
      "fcm_dpo/margin": -0.0981263816356659,
      "fcm_dpo/q_t": 0.5029854774475098,
      "grad_norm": 36.648460388183594,
      "learning_rate": 3.731343283582089e-08,
      "logits/chosen": 0.16564564406871796,
      "logits/rejected": 0.12472124397754669,
      "logps/chosen": -51.377342224121094,
      "logps/ref_chosen": -51.30736541748047,
      "logps/ref_rejected": -82.77239227294922,
      "logps/rejected": -82.7442398071289,
      "loss": 1.3989,
      "margin_dpo/margin_mean": -0.09812664985656738,
      "margin_dpo/margin_std": 0.4009813666343689,
      "step": 6
    },
    {
      "KL/chosen_KL_mean": 0.015993118286132812,
      "KL/mean": 0.01990698277950287,
      "KL/rejected_KL_mean": 0.0238189697265625,
      "KL/std": 0.1881713569164276,
      "epoch": 0.010582010582010581,
      "fcm_dpo/beta": 0.1268007904291153,
      "fcm_dpo/delta": 0.1994636505842209,
      "fcm_dpo/margin": -0.007827132940292358,
      "fcm_dpo/q_t": 0.5002532005310059,
      "grad_norm": 34.524017333984375,
      "learning_rate": 4.477611940298507e-08,
      "logits/chosen": 0.00693818274885416,
      "logits/rejected": -0.03615080937743187,
      "logps/chosen": -51.44342041015625,
      "logps/ref_chosen": -51.45941162109375,
      "logps/ref_rejected": -66.3828125,
      "logps/rejected": -66.35899353027344,
      "loss": 1.3875,
      "margin_dpo/margin_mean": -0.007827192544937134,
      "margin_dpo/margin_std": 0.2439565658569336,
      "step": 7
    },
    {
      "KL/chosen_KL_mean": 0.019863128662109375,
      "KL/mean": -0.010127946734428406,
      "KL/rejected_KL_mean": -0.040111541748046875,
      "KL/std": 0.24710315465927124,
      "epoch": 0.012093726379440665,
      "fcm_dpo/beta": 0.13449162244796753,
      "fcm_dpo/delta": 0.3923080265522003,
      "fcm_dpo/margin": 0.059976786375045776,
      "fcm_dpo/q_t": 0.49807968735694885,
      "grad_norm": 38.144073486328125,
      "learning_rate": 5.223880597014925e-08,
      "logits/chosen": 0.07271274924278259,
      "logits/rejected": 0.05057502165436745,
      "logps/chosen": -62.17768096923828,
      "logps/ref_chosen": -62.197547912597656,
      "logps/ref_rejected": -74.66180419921875,
      "logps/rejected": -74.70191955566406,
      "loss": 1.3789,
      "margin_dpo/margin_mean": 0.0599762499332428,
      "margin_dpo/margin_std": 0.34887492656707764,
      "step": 8
    },
    {
      "KL/chosen_KL_mean": -0.031890869140625,
      "KL/mean": -0.02231958508491516,
      "KL/rejected_KL_mean": -0.012752532958984375,
      "KL/std": 0.23765209317207336,
      "epoch": 0.013605442176870748,
      "fcm_dpo/beta": 0.1427447646856308,
      "fcm_dpo/delta": 0.199895977973938,
      "fcm_dpo/margin": -0.019134581089019775,
      "fcm_dpo/q_t": 0.5006834268569946,
      "grad_norm": 45.233642578125,
      "learning_rate": 5.970149253731343e-08,
      "logits/chosen": 0.16762521862983704,
      "logits/rejected": 0.10846076905727386,
      "logps/chosen": -55.66161346435547,
      "logps/ref_chosen": -55.629722595214844,
      "logps/ref_rejected": -86.21221923828125,
      "logps/rejected": -86.2249755859375,
      "loss": 1.3897,
      "margin_dpo/margin_mean": -0.019134342670440674,
      "margin_dpo/margin_std": 0.3501278758049011,
      "step": 9
    },
    {
      "KL/chosen_KL_mean": -0.02667236328125,
      "KL/mean": -0.001292884349822998,
      "KL/rejected_KL_mean": 0.024082183837890625,
      "KL/std": 0.2424069046974182,
      "epoch": 0.015117157974300832,
      "fcm_dpo/beta": 0.1427447646856308,
      "fcm_dpo/delta": 0.0,
      "fcm_dpo/margin": -0.05076020956039429,
      "fcm_dpo/q_t": 0.5018121004104614,
      "grad_norm": 42.52825927734375,
      "learning_rate": 6.71641791044776e-08,
      "logits/chosen": 0.1346004456281662,
      "logits/rejected": 0.10349121689796448,
      "logps/chosen": -62.7172737121582,
      "logps/ref_chosen": -62.69060134887695,
      "logps/ref_rejected": -90.610107421875,
      "logps/rejected": -90.58602905273438,
      "loss": 1.3942,
      "margin_dpo/margin_mean": -0.05076006054878235,
      "margin_dpo/margin_std": 0.36290353536605835,
      "step": 10
    },
    {
      "KL/chosen_KL_mean": -0.0164031982421875,
      "KL/mean": -0.019405141472816467,
      "KL/rejected_KL_mean": -0.022411346435546875,
      "KL/std": 0.19960369169712067,
      "epoch": 0.016628873771730914,
      "fcm_dpo/beta": 0.1484941840171814,
      "fcm_dpo/delta": 0.19743794202804565,
      "fcm_dpo/margin": 0.006010264158248901,
      "fcm_dpo/q_t": 0.4998025596141815,
      "grad_norm": 43.63276672363281,
      "learning_rate": 7.462686567164178e-08,
      "logits/chosen": 0.11837198585271835,
      "logits/rejected": 0.11128821223974228,
      "logps/chosen": -65.78352355957031,
      "logps/ref_chosen": -65.76712036132812,
      "logps/ref_rejected": -72.4764633178711,
      "logps/rejected": -72.49887084960938,
      "loss": 1.3859,
      "margin_dpo/margin_mean": 0.0060103535652160645,
      "margin_dpo/margin_std": 0.3046831786632538,
      "step": 11
    },
    {
      "KL/chosen_KL_mean": -0.016712188720703125,
      "KL/mean": -0.02420225739479065,
      "KL/rejected_KL_mean": -0.0316925048828125,
      "KL/std": 0.19888459146022797,
      "epoch": 0.018140589569160998,
      "fcm_dpo/beta": 0.15447428822517395,
      "fcm_dpo/delta": 0.19740913808345795,
      "fcm_dpo/margin": 0.014980614185333252,
      "fcm_dpo/q_t": 0.499448299407959,
      "grad_norm": 43.49526596069336,
      "learning_rate": 8.208955223880596e-08,
      "logits/chosen": 0.04357253015041351,
      "logits/rejected": 0.02697162888944149,
      "logps/chosen": -60.72160339355469,
      "logps/ref_chosen": -60.704891204833984,
      "logps/ref_rejected": -69.41564178466797,
      "logps/rejected": -69.44733428955078,
      "loss": 1.3844,
      "margin_dpo/margin_mean": 0.014980673789978027,
      "margin_dpo/margin_std": 0.2739675045013428,
      "step": 12
    },
    {
      "KL/chosen_KL_mean": 0.03954315185546875,
      "KL/mean": 0.020595967769622803,
      "KL/rejected_KL_mean": 0.00165557861328125,
      "KL/std": 0.19518539309501648,
      "epoch": 0.019652305366591082,
      "fcm_dpo/beta": 0.15752489864826202,
      "fcm_dpo/delta": 0.19368299841880798,
      "fcm_dpo/margin": 0.037889957427978516,
      "fcm_dpo/q_t": 0.4985363185405731,
      "grad_norm": 45.36277770996094,
      "learning_rate": 8.955223880597014e-08,
      "logits/chosen": 0.09215953946113586,
      "logits/rejected": 0.031327713280916214,
      "logps/chosen": -49.869712829589844,
      "logps/ref_chosen": -49.90925598144531,
      "logps/ref_rejected": -92.37818145751953,
      "logps/rejected": -92.37652587890625,
      "loss": 1.3806,
      "margin_dpo/margin_mean": 0.03788921236991882,
      "margin_dpo/margin_std": 0.25264811515808105,
      "step": 13
    },
    {
      "KL/chosen_KL_mean": -0.005786895751953125,
      "KL/mean": -0.01808251440525055,
      "KL/rejected_KL_mean": -0.030376434326171875,
      "KL/std": 0.1731014847755432,
      "epoch": 0.021164021164021163,
      "fcm_dpo/beta": 0.17045539617538452,
      "fcm_dpo/delta": 0.39592236280441284,
      "fcm_dpo/margin": 0.024587482213974,
      "fcm_dpo/q_t": 0.49898165464401245,
      "grad_norm": 50.24029541015625,
      "learning_rate": 9.701492537313432e-08,
      "logits/chosen": 0.09761021286249161,
      "logits/rejected": 0.07951986789703369,
      "logps/chosen": -60.62458038330078,
      "logps/ref_chosen": -60.61879348754883,
      "logps/ref_rejected": -71.79306030273438,
      "logps/rejected": -71.82344055175781,
      "loss": 1.3825,
      "margin_dpo/margin_mean": 0.024587631225585938,
      "margin_dpo/margin_std": 0.23990672826766968,
      "step": 14
    },
    {
      "KL/chosen_KL_mean": -0.026651382446289062,
      "KL/mean": -0.031851932406425476,
      "KL/rejected_KL_mean": -0.037044525146484375,
      "KL/std": 0.2409576177597046,
      "epoch": 0.022675736961451247,
      "fcm_dpo/beta": 0.18080593645572662,
      "fcm_dpo/delta": 0.19737780094146729,
      "fcm_dpo/margin": 0.010393649339675903,
      "fcm_dpo/q_t": 0.49955692887306213,
      "grad_norm": 60.253578186035156,
      "learning_rate": 1.044776119402985e-07,
      "logits/chosen": 0.09280131757259369,
      "logits/rejected": 0.04811263829469681,
      "logps/chosen": -63.49618911743164,
      "logps/ref_chosen": -63.46953582763672,
      "logps/ref_rejected": -88.88951110839844,
      "logps/rejected": -88.92655944824219,
      "loss": 1.3853,
      "margin_dpo/margin_mean": 0.010393589735031128,
      "margin_dpo/margin_std": 0.32523539662361145,
      "step": 15
    },
    {
      "KL/chosen_KL_mean": 0.0018177032470703125,
      "KL/mean": -0.00604432076215744,
      "KL/rejected_KL_mean": -0.01390838623046875,
      "KL/std": 0.22720113396644592,
      "epoch": 0.02418745275888133,
      "fcm_dpo/beta": 0.18434685468673706,
      "fcm_dpo/delta": 0.19210314750671387,
      "fcm_dpo/margin": 0.01572038233280182,
      "fcm_dpo/q_t": 0.49929431080818176,
      "grad_norm": 49.553138732910156,
      "learning_rate": 1.1194029850746268e-07,
      "logits/chosen": 0.13143643736839294,
      "logits/rejected": 0.09331878274679184,
      "logps/chosen": -46.530479431152344,
      "logps/ref_chosen": -46.53229904174805,
      "logps/ref_rejected": -74.27533721923828,
      "logps/rejected": -74.28924560546875,
      "loss": 1.3842,
      "margin_dpo/margin_mean": 0.01572073996067047,
      "margin_dpo/margin_std": 0.3393726944923401,
      "step": 16
    },
    {
      "KL/chosen_KL_mean": -0.023601531982421875,
      "KL/mean": -0.020472899079322815,
      "KL/rejected_KL_mean": -0.01734161376953125,
      "KL/std": 0.20937620103359222,
      "epoch": 0.025699168556311415,
      "fcm_dpo/beta": 0.1953100562095642,
      "fcm_dpo/delta": 0.19371725618839264,
      "fcm_dpo/margin": -0.006255239248275757,
      "fcm_dpo/q_t": 0.5003689527511597,
      "grad_norm": 64.56086730957031,
      "learning_rate": 1.1940298507462686e-07,
      "logits/chosen": 0.056045085191726685,
      "logits/rejected": 0.03731035813689232,
      "logps/chosen": -64.10143280029297,
      "logps/ref_chosen": -64.07783508300781,
      "logps/ref_rejected": -86.40876770019531,
      "logps/rejected": -86.42610931396484,
      "loss": 1.3884,
      "margin_dpo/margin_mean": -0.006255537271499634,
      "margin_dpo/margin_std": 0.2892005145549774,
      "step": 17
    },
    {
      "KL/chosen_KL_mean": 0.037494659423828125,
      "KL/mean": 0.004080817103385925,
      "KL/rejected_KL_mean": -0.029333114624023438,
      "KL/std": 0.19442421197891235,
      "epoch": 0.027210884353741496,
      "fcm_dpo/beta": 0.20706477761268616,
      "fcm_dpo/delta": 0.386584997177124,
      "fcm_dpo/margin": 0.06682577729225159,
      "fcm_dpo/q_t": 0.49664899706840515,
      "grad_norm": 57.38557434082031,
      "learning_rate": 1.2686567164179106e-07,
      "logits/chosen": 0.08666019141674042,
      "logits/rejected": 0.04165633022785187,
      "logps/chosen": -44.83684158325195,
      "logps/ref_chosen": -44.87433624267578,
      "logps/ref_rejected": -70.97604370117188,
      "logps/rejected": -71.00537872314453,
      "loss": 1.3732,
      "margin_dpo/margin_mean": 0.06682571768760681,
      "margin_dpo/margin_std": 0.26390185952186584,
      "step": 18
    },
    {
      "KL/chosen_KL_mean": 0.015768051147460938,
      "KL/mean": 0.011087000370025635,
      "KL/rejected_KL_mean": 0.006404876708984375,
      "KL/std": 0.2442140132188797,
      "epoch": 0.02872260015117158,
      "fcm_dpo/beta": 0.22403427958488464,
      "fcm_dpo/delta": 0.3979843854904175,
      "fcm_dpo/margin": 0.00936955213546753,
      "fcm_dpo/q_t": 0.4994935691356659,
      "grad_norm": 69.56298065185547,
      "learning_rate": 1.343283582089552e-07,
      "logits/chosen": 0.07334445416927338,
      "logits/rejected": 0.05974186211824417,
      "logps/chosen": -68.14404296875,
      "logps/ref_chosen": -68.1598129272461,
      "logps/ref_rejected": -81.17138671875,
      "logps/rejected": -81.16498565673828,
      "loss": 1.3856,
      "margin_dpo/margin_mean": 0.009370237588882446,
      "margin_dpo/margin_std": 0.33312809467315674,
      "step": 19
    },
    {
      "KL/chosen_KL_mean": -0.0043697357177734375,
      "KL/mean": 0.011444643139839172,
      "KL/rejected_KL_mean": 0.027263641357421875,
      "KL/std": 0.21824133396148682,
      "epoch": 0.030234315948601664,
      "fcm_dpo/beta": 0.23310576379299164,
      "fcm_dpo/delta": 0.19794167578220367,
      "fcm_dpo/margin": -0.03163328766822815,
      "fcm_dpo/q_t": 0.5018023252487183,
      "grad_norm": 68.3042984008789,
      "learning_rate": 1.4179104477611938e-07,
      "logits/chosen": 0.1361723095178604,
      "logits/rejected": 0.1131054013967514,
      "logps/chosen": -53.68292999267578,
      "logps/ref_chosen": -53.67856216430664,
      "logps/ref_rejected": -74.16911315917969,
      "logps/rejected": -74.14185333251953,
      "loss": 1.3948,
      "margin_dpo/margin_mean": -0.031633466482162476,
      "margin_dpo/margin_std": 0.3028011620044708,
      "step": 20
    },
    {
      "KL/chosen_KL_mean": 0.022541046142578125,
      "KL/mean": 0.02938457578420639,
      "KL/rejected_KL_mean": 0.036224365234375,
      "KL/std": 0.23280031979084015,
      "epoch": 0.031746031746031744,
      "fcm_dpo/beta": 0.24255003035068512,
      "fcm_dpo/delta": 0.19919079542160034,
      "fcm_dpo/margin": -0.01367737352848053,
      "fcm_dpo/q_t": 0.5008128881454468,
      "grad_norm": 72.05040740966797,
      "learning_rate": 1.4925373134328355e-07,
      "logits/chosen": 0.09666962176561356,
      "logits/rejected": 0.0715550035238266,
      "logps/chosen": -64.67901611328125,
      "logps/ref_chosen": -64.70155334472656,
      "logps/ref_rejected": -81.02095031738281,
      "logps/rejected": -80.98472595214844,
      "loss": 1.3912,
      "margin_dpo/margin_mean": -0.013677477836608887,
      "margin_dpo/margin_std": 0.3375088572502136,
      "step": 21
    },
    {
      "KL/chosen_KL_mean": -0.03160858154296875,
      "KL/mean": -0.027323782444000244,
      "KL/rejected_KL_mean": -0.023040771484375,
      "KL/std": 0.21388794481754303,
      "epoch": 0.03325774754346183,
      "fcm_dpo/beta": 0.2573118507862091,
      "fcm_dpo/delta": 0.19680048525333405,
      "fcm_dpo/margin": -0.008568942546844482,
      "fcm_dpo/q_t": 0.5005735158920288,
      "grad_norm": 75.60509490966797,
      "learning_rate": 1.5671641791044775e-07,
      "logits/chosen": 0.006399281322956085,
      "logits/rejected": -0.0146299097687006,
      "logps/chosen": -58.06760025024414,
      "logps/ref_chosen": -58.03599166870117,
      "logps/ref_rejected": -80.72721862792969,
      "logps/rejected": -80.75025939941406,
      "loss": 1.3901,
      "margin_dpo/margin_mean": -0.008568882942199707,
      "margin_dpo/margin_std": 0.3066539466381073,
      "step": 22
    },
    {
      "KL/chosen_KL_mean": 0.014617919921875,
      "KL/mean": 0.009073600172996521,
      "KL/rejected_KL_mean": 0.00353240966796875,
      "KL/std": 0.2195100635290146,
      "epoch": 0.03476946334089191,
      "fcm_dpo/beta": 0.26226556301116943,
      "fcm_dpo/delta": 0.1889042854309082,
      "fcm_dpo/margin": 0.011082977056503296,
      "fcm_dpo/q_t": 0.4992814064025879,
      "grad_norm": 87.26220703125,
      "learning_rate": 1.6417910447761193e-07,
      "logits/chosen": 0.11942335218191147,
      "logits/rejected": 0.09495219588279724,
      "logps/chosen": -66.34146881103516,
      "logps/ref_chosen": -66.35608673095703,
      "logps/ref_rejected": -93.02769470214844,
      "logps/rejected": -93.02415466308594,
      "loss": 1.3846,
      "margin_dpo/margin_mean": 0.011082679033279419,
      "margin_dpo/margin_std": 0.29644426703453064,
      "step": 23
    },
    {
      "KL/chosen_KL_mean": 0.0018310546875,
      "KL/mean": 0.010483011603355408,
      "KL/rejected_KL_mean": 0.019138336181640625,
      "KL/std": 0.19121773540973663,
      "epoch": 0.036281179138321996,
      "fcm_dpo/beta": 0.2780904471874237,
      "fcm_dpo/delta": 0.19938357174396515,
      "fcm_dpo/margin": -0.017304211854934692,
      "fcm_dpo/q_t": 0.5012048482894897,
      "grad_norm": 73.68643188476562,
      "learning_rate": 1.716417910447761e-07,
      "logits/chosen": 0.10710248351097107,
      "logits/rejected": 0.07515440881252289,
      "logps/chosen": -54.459407806396484,
      "logps/ref_chosen": -54.461238861083984,
      "logps/ref_rejected": -68.33817291259766,
      "logps/rejected": -68.31903839111328,
      "loss": 1.3924,
      "margin_dpo/margin_mean": -0.017304420471191406,
      "margin_dpo/margin_std": 0.25890859961509705,
      "step": 24
    },
    {
      "KL/chosen_KL_mean": -0.031110763549804688,
      "KL/mean": -0.03343985974788666,
      "KL/rejected_KL_mean": -0.035770416259765625,
      "KL/std": 0.2234145700931549,
      "epoch": 0.03779289493575208,
      "fcm_dpo/beta": 0.289237916469574,
      "fcm_dpo/delta": 0.19651614129543304,
      "fcm_dpo/margin": 0.004654437303543091,
      "fcm_dpo/q_t": 0.49969354271888733,
      "grad_norm": 85.54322052001953,
      "learning_rate": 1.7910447761194027e-07,
      "logits/chosen": 0.10172505676746368,
      "logits/rejected": 0.05041360855102539,
      "logps/chosen": -60.035316467285156,
      "logps/ref_chosen": -60.00420379638672,
      "logps/ref_rejected": -90.47376251220703,
      "logps/rejected": -90.50953674316406,
      "loss": 1.3872,
      "margin_dpo/margin_mean": 0.004654288291931152,
      "margin_dpo/margin_std": 0.32953381538391113,
      "step": 25
    },
    {
      "KL/chosen_KL_mean": -0.03518486022949219,
      "KL/mean": -0.012967824935913086,
      "KL/rejected_KL_mean": 0.009246826171875,
      "KL/std": 0.20040880143642426,
      "epoch": 0.039304610733182165,
      "fcm_dpo/beta": 0.289237916469574,
      "fcm_dpo/delta": 0.0,
      "fcm_dpo/margin": -0.04443258047103882,
      "fcm_dpo/q_t": 0.5032066106796265,
      "grad_norm": 87.03143310546875,
      "learning_rate": 1.8656716417910447e-07,
      "logits/chosen": 0.11158916354179382,
      "logits/rejected": 0.09309649467468262,
      "logps/chosen": -56.854339599609375,
      "logps/ref_chosen": -56.81915283203125,
      "logps/ref_rejected": -77.84333038330078,
      "logps/rejected": -77.8340835571289,
      "loss": 1.4012,
      "margin_dpo/margin_mean": -0.04443329572677612,
      "margin_dpo/margin_std": 0.31203919649124146,
      "step": 26
    },
    {
      "KL/chosen_KL_mean": -0.0073394775390625,
      "KL/mean": 0.002377226948738098,
      "KL/rejected_KL_mean": 0.012094497680664062,
      "KL/std": 0.24115195870399475,
      "epoch": 0.04081632653061224,
      "fcm_dpo/beta": 0.289237916469574,
      "fcm_dpo/delta": 0.0,
      "fcm_dpo/margin": -0.019435137510299683,
      "fcm_dpo/q_t": 0.5014033317565918,
      "grad_norm": 83.701904296875,
      "learning_rate": 1.9402985074626865e-07,
      "logits/chosen": 0.12982380390167236,
      "logits/rejected": 0.10382634401321411,
      "logps/chosen": -62.88436508178711,
      "logps/ref_chosen": -62.87702560424805,
      "logps/ref_rejected": -71.34437561035156,
      "logps/rejected": -71.33228302001953,
      "loss": 1.3939,
      "margin_dpo/margin_mean": -0.019435018301010132,
      "margin_dpo/margin_std": 0.3067646324634552,
      "step": 27
    },
    {
      "KL/chosen_KL_mean": -0.024747848510742188,
      "KL/mean": -0.01658952236175537,
      "KL/rejected_KL_mean": -0.008432388305664062,
      "KL/std": 0.2040315568447113,
      "epoch": 0.042328042328042326,
      "fcm_dpo/beta": 0.3005712330341339,
      "fcm_dpo/delta": 0.19217611849308014,
      "fcm_dpo/margin": -0.01631641387939453,
      "fcm_dpo/q_t": 0.5012955665588379,
      "grad_norm": 83.30447387695312,
      "learning_rate": 2.0149253731343282e-07,
      "logits/chosen": 0.0741531178355217,
      "logits/rejected": 0.06495097279548645,
      "logps/chosen": -59.858123779296875,
      "logps/ref_chosen": -59.8333740234375,
      "logps/ref_rejected": -70.39804077148438,
      "logps/rejected": -70.4064712524414,
      "loss": 1.393,
      "margin_dpo/margin_mean": -0.01631644368171692,
      "margin_dpo/margin_std": 0.2692261338233948,
      "step": 28
    },
    {
      "KL/chosen_KL_mean": -0.01944732666015625,
      "KL/mean": -0.053507544100284576,
      "KL/rejected_KL_mean": -0.08756256103515625,
      "KL/std": 0.23613537847995758,
      "epoch": 0.04383975812547241,
      "fcm_dpo/beta": 0.3183423578739166,
      "fcm_dpo/delta": 0.37896400690078735,
      "fcm_dpo/margin": 0.0681174248456955,
      "fcm_dpo/q_t": 0.494759202003479,
      "grad_norm": 102.14923095703125,
      "learning_rate": 2.08955223880597e-07,
      "logits/chosen": 0.15540289878845215,
      "logits/rejected": 0.1371532380580902,
      "logps/chosen": -74.1396484375,
      "logps/ref_chosen": -74.12020111083984,
      "logps/ref_rejected": -83.33099365234375,
      "logps/rejected": -83.41854858398438,
      "loss": 1.3673,
      "margin_dpo/margin_mean": 0.06811761856079102,
      "margin_dpo/margin_std": 0.32512563467025757,
      "step": 29
    },
    {
      "KL/chosen_KL_mean": -0.025737762451171875,
      "KL/mean": -0.03448393940925598,
      "KL/rejected_KL_mean": -0.0432281494140625,
      "KL/std": 0.1987377107143402,
      "epoch": 0.045351473922902494,
      "fcm_dpo/beta": 0.344091534614563,
      "fcm_dpo/delta": 0.3941713571548462,
      "fcm_dpo/margin": 0.017492294311523438,
      "fcm_dpo/q_t": 0.49854251742362976,
      "grad_norm": 103.57805633544922,
      "learning_rate": 2.1641791044776117e-07,
      "logits/chosen": 0.11821296066045761,
      "logits/rejected": 0.06425055116415024,
      "logps/chosen": -50.77702713012695,
      "logps/ref_chosen": -50.75128936767578,
      "logps/ref_rejected": -89.29063415527344,
      "logps/rejected": -89.3338623046875,
      "loss": 1.3828,
      "margin_dpo/margin_mean": 0.017491638660430908,
      "margin_dpo/margin_std": 0.2868330478668213,
      "step": 30
    },
    {
      "KL/chosen_KL_mean": -0.024738311767578125,
      "KL/mean": -0.04767066240310669,
      "KL/rejected_KL_mean": -0.0706024169921875,
      "KL/std": 0.2183002531528473,
      "epoch": 0.04686318972033258,
      "fcm_dpo/beta": 0.3570849895477295,
      "fcm_dpo/delta": 0.17508627474308014,
      "fcm_dpo/margin": 0.04586535692214966,
      "fcm_dpo/q_t": 0.49598830938339233,
      "grad_norm": 122.07454681396484,
      "learning_rate": 2.2388059701492537e-07,
      "logits/chosen": 0.10629313439130783,
      "logits/rejected": 0.06015738844871521,
      "logps/chosen": -65.36149597167969,
      "logps/ref_chosen": -65.33675384521484,
      "logps/ref_rejected": -100.76666259765625,
      "logps/rejected": -100.83726501464844,
      "loss": 1.3727,
      "margin_dpo/margin_mean": 0.04586589336395264,
      "margin_dpo/margin_std": 0.3037866950035095,
      "step": 31
    },
    {
      "KL/chosen_KL_mean": -0.001934051513671875,
      "KL/mean": -0.027067288756370544,
      "KL/rejected_KL_mean": -0.052204132080078125,
      "KL/std": 0.21004009246826172,
      "epoch": 0.04837490551776266,
      "fcm_dpo/beta": 0.38511091470718384,
      "fcm_dpo/delta": 0.38101980090141296,
      "fcm_dpo/margin": 0.050266534090042114,
      "fcm_dpo/q_t": 0.49527406692504883,
      "grad_norm": 115.42655181884766,
      "learning_rate": 2.3134328358208954e-07,
      "logits/chosen": 0.09014316648244858,
      "logits/rejected": 0.08219875395298004,
      "logps/chosen": -67.18526458740234,
      "logps/ref_chosen": -67.18333435058594,
      "logps/ref_rejected": -82.80763244628906,
      "logps/rejected": -82.8598403930664,
      "loss": 1.3699,
      "margin_dpo/margin_mean": 0.05026650428771973,
      "margin_dpo/margin_std": 0.2893450856208801,
      "step": 32
    },
    {
      "KL/chosen_KL_mean": -0.0030364990234375,
      "KL/mean": -0.04211585968732834,
      "KL/rejected_KL_mean": -0.08119583129882812,
      "KL/std": 0.24913763999938965,
      "epoch": 0.049886621315192746,
      "fcm_dpo/beta": 0.41478103399276733,
      "fcm_dpo/delta": 0.3684595823287964,
      "fcm_dpo/margin": 0.07815925776958466,
      "fcm_dpo/q_t": 0.492145836353302,
      "grad_norm": 127.29991149902344,
      "learning_rate": 2.388059701492537e-07,
      "logits/chosen": 0.02003825642168522,
      "logits/rejected": -0.005349243059754372,
      "logps/chosen": -64.04251861572266,
      "logps/ref_chosen": -64.03948211669922,
      "logps/ref_rejected": -75.68357849121094,
      "logps/rejected": -75.76477813720703,
      "loss": 1.3588,
      "margin_dpo/margin_mean": 0.07816009223461151,
      "margin_dpo/margin_std": 0.33464300632476807,
      "step": 33
    },
    {
      "KL/chosen_KL_mean": -0.04259300231933594,
      "KL/mean": -0.06921710819005966,
      "KL/rejected_KL_mean": -0.0958404541015625,
      "KL/std": 0.20878559350967407,
      "epoch": 0.05139833711262283,
      "fcm_dpo/beta": 0.4466947317123413,
      "fcm_dpo/delta": 0.37713927030563354,
      "fcm_dpo/margin": 0.053250834345817566,
      "fcm_dpo/q_t": 0.49428844451904297,
      "grad_norm": 126.29956817626953,
      "learning_rate": 2.4626865671641786e-07,
      "logits/chosen": 0.1000206470489502,
      "logits/rejected": 0.07004327327013016,
      "logps/chosen": -53.706886291503906,
      "logps/ref_chosen": -53.6642951965332,
      "logps/ref_rejected": -65.77989959716797,
      "logps/rejected": -65.87574005126953,
      "loss": 1.3668,
      "margin_dpo/margin_mean": 0.053250715136528015,
      "margin_dpo/margin_std": 0.2848781645298004,
      "step": 34
    },
    {
      "KL/chosen_KL_mean": -0.08517646789550781,
      "KL/mean": -0.09704282879829407,
      "KL/rejected_KL_mean": -0.10890960693359375,
      "KL/std": 0.22471538186073303,
      "epoch": 0.05291005291005291,
      "fcm_dpo/beta": 0.472166508436203,
      "fcm_dpo/delta": 0.1833934485912323,
      "fcm_dpo/margin": 0.02373194694519043,
      "fcm_dpo/q_t": 0.49740830063819885,
      "grad_norm": 130.8272247314453,
      "learning_rate": 2.537313432835821e-07,
      "logits/chosen": 0.05964568629860878,
      "logits/rejected": 0.037086836993694305,
      "logps/chosen": -61.1020393371582,
      "logps/ref_chosen": -61.01686096191406,
      "logps/ref_rejected": -72.78598022460938,
      "logps/rejected": -72.8948974609375,
      "loss": 1.3811,
      "margin_dpo/margin_mean": 0.023731738328933716,
      "margin_dpo/margin_std": 0.321732759475708,
      "step": 35
    },
    {
      "KL/chosen_KL_mean": -0.08330917358398438,
      "KL/mean": -0.09451384842395782,
      "KL/rejected_KL_mean": -0.10572052001953125,
      "KL/std": 0.24240775406360626,
      "epoch": 0.05442176870748299,
      "fcm_dpo/beta": 0.4811955988407135,
      "fcm_dpo/delta": 0.18766069412231445,
      "fcm_dpo/margin": 0.022410601377487183,
      "fcm_dpo/q_t": 0.4973487854003906,
      "grad_norm": 138.20062255859375,
      "learning_rate": 2.611940298507462e-07,
      "logits/chosen": 0.09503352642059326,
      "logits/rejected": 0.04270517826080322,
      "logps/chosen": -50.620670318603516,
      "logps/ref_chosen": -50.53736114501953,
      "logps/ref_rejected": -78.11678314208984,
      "logps/rejected": -78.22250366210938,
      "loss": 1.3824,
      "margin_dpo/margin_mean": 0.02241101861000061,
      "margin_dpo/margin_std": 0.35139748454093933,
      "step": 36
    },
    {
      "KL/chosen_KL_mean": -0.01982879638671875,
      "KL/mean": -0.08735629916191101,
      "KL/rejected_KL_mean": -0.15488815307617188,
      "KL/std": 0.3164428472518921,
      "epoch": 0.055933484504913075,
      "fcm_dpo/beta": 0.5152612328529358,
      "fcm_dpo/delta": 0.33278989791870117,
      "fcm_dpo/margin": 0.13505524396896362,
      "fcm_dpo/q_t": 0.48369383811950684,
      "grad_norm": 183.72947692871094,
      "learning_rate": 2.686567164179104e-07,
      "logits/chosen": 0.10293924808502197,
      "logits/rejected": 0.023218315094709396,
      "logps/chosen": -59.57377624511719,
      "logps/ref_chosen": -59.55394744873047,
      "logps/ref_rejected": -108.27702331542969,
      "logps/rejected": -108.43191528320312,
      "loss": 1.3308,
      "margin_dpo/margin_mean": 0.13505515456199646,
      "margin_dpo/margin_std": 0.4417878985404968,
      "step": 37
    },
    {
      "KL/chosen_KL_mean": -0.09215354919433594,
      "KL/mean": -0.10593746602535248,
      "KL/rejected_KL_mean": -0.11971855163574219,
      "KL/std": 0.28186145424842834,
      "epoch": 0.05744520030234316,
      "fcm_dpo/beta": 0.5423237681388855,
      "fcm_dpo/delta": 0.17220675945281982,
      "fcm_dpo/margin": 0.027556508779525757,
      "fcm_dpo/q_t": 0.496552973985672,
      "grad_norm": 161.56407165527344,
      "learning_rate": 2.761194029850746e-07,
      "logits/chosen": 0.05953256040811539,
      "logits/rejected": 0.045650843530893326,
      "logps/chosen": -65.88050842285156,
      "logps/ref_chosen": -65.78836059570312,
      "logps/ref_rejected": -76.1619873046875,
      "logps/rejected": -76.28170776367188,
      "loss": 1.3839,
      "margin_dpo/margin_mean": 0.02755632996559143,
      "margin_dpo/margin_std": 0.40284162759780884,
      "step": 38
    },
    {
      "KL/chosen_KL_mean": -0.079742431640625,
      "KL/mean": -0.1202605664730072,
      "KL/rejected_KL_mean": -0.16077423095703125,
      "KL/std": 0.25623682141304016,
      "epoch": 0.05895691609977324,
      "fcm_dpo/beta": 0.5728709697723389,
      "fcm_dpo/delta": 0.35464733839035034,
      "fcm_dpo/margin": 0.08103156089782715,
      "fcm_dpo/q_t": 0.4886939823627472,
      "grad_norm": 166.70558166503906,
      "learning_rate": 2.8358208955223876e-07,
      "logits/chosen": 0.15896514058113098,
      "logits/rejected": 0.13215456902980804,
      "logps/chosen": -57.256553649902344,
      "logps/ref_chosen": -57.17681121826172,
      "logps/ref_rejected": -79.486328125,
      "logps/rejected": -79.64710998535156,
      "loss": 1.3508,
      "margin_dpo/margin_mean": 0.08103141188621521,
      "margin_dpo/margin_std": 0.35928279161453247,
      "step": 39
    },
    {
      "KL/chosen_KL_mean": -0.11428260803222656,
      "KL/mean": -0.11064784228801727,
      "KL/rejected_KL_mean": -0.10701751708984375,
      "KL/std": 0.23384666442871094,
      "epoch": 0.06046863189720333,
      "fcm_dpo/beta": 0.6046038269996643,
      "fcm_dpo/delta": 0.18890410661697388,
      "fcm_dpo/margin": -0.007268577814102173,
      "fcm_dpo/q_t": 0.5011410117149353,
      "grad_norm": 200.42584228515625,
      "learning_rate": 2.9104477611940296e-07,
      "logits/chosen": 0.12039332836866379,
      "logits/rejected": 0.07013371586799622,
      "logps/chosen": -61.448448181152344,
      "logps/ref_chosen": -61.33416748046875,
      "logps/ref_rejected": -79.10697174072266,
      "logps/rejected": -79.2139892578125,
      "loss": 1.402,
      "margin_dpo/margin_mean": -0.007268846035003662,
      "margin_dpo/margin_std": 0.35015004873275757,
      "step": 40
    },
    {
      "KL/chosen_KL_mean": -0.122222900390625,
      "KL/mean": -0.16071859002113342,
      "KL/rejected_KL_mean": -0.19921493530273438,
      "KL/std": 0.26297086477279663,
      "epoch": 0.06198034769463341,
      "fcm_dpo/beta": 0.6379947662353516,
      "fcm_dpo/delta": 0.3523167371749878,
      "fcm_dpo/margin": 0.0769854187965393,
      "fcm_dpo/q_t": 0.48818713426589966,
      "grad_norm": 190.40330505371094,
      "learning_rate": 2.985074626865671e-07,
      "logits/chosen": 0.05450894311070442,
      "logits/rejected": 0.03441740572452545,
      "logps/chosen": -67.66895294189453,
      "logps/ref_chosen": -67.5467300415039,
      "logps/ref_rejected": -83.87788391113281,
      "logps/rejected": -84.07710266113281,
      "loss": 1.3483,
      "margin_dpo/margin_mean": 0.07698613405227661,
      "margin_dpo/margin_std": 0.3233705461025238,
      "step": 41
    },
    {
      "KL/chosen_KL_mean": -0.14746665954589844,
      "KL/mean": -0.15399429202079773,
      "KL/rejected_KL_mean": -0.1605224609375,
      "KL/std": 0.28739774227142334,
      "epoch": 0.06349206349206349,
      "fcm_dpo/beta": 0.6700236201286316,
      "fcm_dpo/delta": 0.16138070821762085,
      "fcm_dpo/margin": 0.013055786490440369,
      "fcm_dpo/q_t": 0.498212993144989,
      "grad_norm": 194.78236389160156,
      "learning_rate": 3.059701492537313e-07,
      "logits/chosen": 0.07204203307628632,
      "logits/rejected": 0.0498071014881134,
      "logps/chosen": -61.412322998046875,
      "logps/ref_chosen": -61.26485824584961,
      "logps/ref_rejected": -76.3629150390625,
      "logps/rejected": -76.5234375,
      "loss": 1.3974,
      "margin_dpo/margin_mean": 0.013056054711341858,
      "margin_dpo/margin_std": 0.40859484672546387,
      "step": 42
    },
    {
      "KL/chosen_KL_mean": -0.13036727905273438,
      "KL/mean": -0.14713996648788452,
      "KL/rejected_KL_mean": -0.16391372680664062,
      "KL/std": 0.3216056525707245,
      "epoch": 0.06500377928949358,
      "fcm_dpo/beta": 0.7093794345855713,
      "fcm_dpo/delta": 0.3769518733024597,
      "fcm_dpo/margin": 0.03354591131210327,
      "fcm_dpo/q_t": 0.4946945309638977,
      "grad_norm": 253.1779327392578,
      "learning_rate": 3.134328358208955e-07,
      "logits/chosen": 0.08703877031803131,
      "logits/rejected": 0.07603750377893448,
      "logps/chosen": -71.93939208984375,
      "logps/ref_chosen": -71.80902862548828,
      "logps/ref_rejected": -81.12464141845703,
      "logps/rejected": -81.28855895996094,
      "loss": 1.3871,
      "margin_dpo/margin_mean": 0.033546000719070435,
      "margin_dpo/margin_std": 0.4398638606071472,
      "step": 43
    },
    {
      "KL/chosen_KL_mean": -0.14861679077148438,
      "KL/mean": -0.15190964937210083,
      "KL/rejected_KL_mean": -0.15520477294921875,
      "KL/std": 0.27536773681640625,
      "epoch": 0.06651549508692366,
      "fcm_dpo/beta": 0.7668402791023254,
      "fcm_dpo/delta": 0.39510226249694824,
      "fcm_dpo/margin": 0.006583750247955322,
      "fcm_dpo/q_t": 0.49894002079963684,
      "grad_norm": 248.98779296875,
      "learning_rate": 3.2089552238805965e-07,
      "logits/chosen": 0.06032150238752365,
      "logits/rejected": 0.029356716200709343,
      "logps/chosen": -66.69905090332031,
      "logps/ref_chosen": -66.55043029785156,
      "logps/ref_rejected": -85.06198120117188,
      "logps/rejected": -85.2171859741211,
      "loss": 1.404,
      "margin_dpo/margin_mean": 0.006583571434020996,
      "margin_dpo/margin_std": 0.3948115110397339,
      "step": 44
    },
    {
      "KL/chosen_KL_mean": -0.16165733337402344,
      "KL/mean": -0.22152617573738098,
      "KL/rejected_KL_mean": -0.28139495849609375,
      "KL/std": 0.27924102544784546,
      "epoch": 0.06802721088435375,
      "fcm_dpo/beta": 0.8197171688079834,
      "fcm_dpo/delta": 0.30430763959884644,
      "fcm_dpo/margin": 0.11974488198757172,
      "fcm_dpo/q_t": 0.4763890504837036,
      "grad_norm": 252.16053771972656,
      "learning_rate": 3.2835820895522385e-07,
      "logits/chosen": 0.13100901246070862,
      "logits/rejected": 0.0773434191942215,
      "logps/chosen": -62.40550994873047,
      "logps/ref_chosen": -62.24385452270508,
      "logps/ref_rejected": -92.96665954589844,
      "logps/rejected": -93.24806213378906,
      "loss": 1.3108,
      "margin_dpo/margin_mean": 0.11974579095840454,
      "margin_dpo/margin_std": 0.3502202332019806,
      "step": 45
    },
    {
      "KL/chosen_KL_mean": -0.0877227783203125,
      "KL/mean": -0.1594286412000656,
      "KL/rejected_KL_mean": -0.23113632202148438,
      "KL/std": 0.3115689754486084,
      "epoch": 0.06953892668178382,
      "fcm_dpo/beta": 0.8677189350128174,
      "fcm_dpo/delta": 0.2785709500312805,
      "fcm_dpo/margin": 0.14340588450431824,
      "fcm_dpo/q_t": 0.4707869291305542,
      "grad_norm": 243.6422119140625,
      "learning_rate": 3.3582089552238805e-07,
      "logits/chosen": 0.14529743790626526,
      "logits/rejected": 0.09837373346090317,
      "logps/chosen": -61.58662796020508,
      "logps/ref_chosen": -61.498905181884766,
      "logps/ref_rejected": -78.91172790527344,
      "logps/rejected": -79.14286804199219,
      "loss": 1.2945,
      "margin_dpo/margin_mean": 0.14340543746948242,
      "margin_dpo/margin_std": 0.39542055130004883,
      "step": 46
    },
    {
      "KL/chosen_KL_mean": -0.14699363708496094,
      "KL/mean": -0.20896488428115845,
      "KL/rejected_KL_mean": -0.2709388732910156,
      "KL/std": 0.2617807388305664,
      "epoch": 0.0710506424792139,
      "fcm_dpo/beta": 0.9190011024475098,
      "fcm_dpo/delta": 0.2888580858707428,
      "fcm_dpo/margin": 0.12394267320632935,
      "fcm_dpo/q_t": 0.4727897047996521,
      "grad_norm": 253.87966918945312,
      "learning_rate": 3.432835820895522e-07,
      "logits/chosen": 0.026472479104995728,
      "logits/rejected": -0.016051139682531357,
      "logps/chosen": -51.725341796875,
      "logps/ref_chosen": -51.578346252441406,
      "logps/ref_rejected": -68.2215576171875,
      "logps/rejected": -68.49249267578125,
      "loss": 1.3024,
      "margin_dpo/margin_mean": 0.12394288182258606,
      "margin_dpo/margin_std": 0.35938262939453125,
      "step": 47
    },
    {
      "KL/chosen_KL_mean": -0.229888916015625,
      "KL/mean": -0.2503516674041748,
      "KL/rejected_KL_mean": -0.27081871032714844,
      "KL/std": 0.30007147789001465,
      "epoch": 0.07256235827664399,
      "fcm_dpo/beta": 0.9456481337547302,
      "fcm_dpo/delta": 0.15676452219486237,
      "fcm_dpo/margin": 0.04092279076576233,
      "fcm_dpo/q_t": 0.4911009967327118,
      "grad_norm": 254.16644287109375,
      "learning_rate": 3.507462686567164e-07,
      "logits/chosen": 0.12581509351730347,
      "logits/rejected": 0.09670676290988922,
      "logps/chosen": -52.02354049682617,
      "logps/ref_chosen": -51.79365158081055,
      "logps/ref_rejected": -64.22503662109375,
      "logps/rejected": -64.49585723876953,
      "loss": 1.3864,
      "margin_dpo/margin_mean": 0.04092302918434143,
      "margin_dpo/margin_std": 0.41956618428230286,
      "step": 48
    },
    {
      "KL/chosen_KL_mean": -0.19176101684570312,
      "KL/mean": -0.2539128065109253,
      "KL/rejected_KL_mean": -0.31606292724609375,
      "KL/std": 0.3132970929145813,
      "epoch": 0.07407407407407407,
      "fcm_dpo/beta": 1.0028908252716064,
      "fcm_dpo/delta": 0.27836254239082336,
      "fcm_dpo/margin": 0.12430019676685333,
      "fcm_dpo/q_t": 0.4709841310977936,
      "grad_norm": 258.2995910644531,
      "learning_rate": 3.5820895522388055e-07,
      "logits/chosen": 0.0265614353120327,
      "logits/rejected": 0.0051910169422626495,
      "logps/chosen": -58.32636260986328,
      "logps/ref_chosen": -58.13460159301758,
      "logps/ref_rejected": -64.63206481933594,
      "logps/rejected": -64.94813537597656,
      "loss": 1.3091,
      "margin_dpo/margin_mean": 0.12430053949356079,
      "margin_dpo/margin_std": 0.4250980615615845,
      "step": 49
    },
    {
      "KL/chosen_KL_mean": -0.24220848083496094,
      "KL/mean": -0.32140886783599854,
      "KL/rejected_KL_mean": -0.40061187744140625,
      "KL/std": 0.35282331705093384,
      "epoch": 0.07558578987150416,
      "fcm_dpo/beta": 1.0499261617660522,
      "fcm_dpo/delta": 0.2380904257297516,
      "fcm_dpo/margin": 0.15839967131614685,
      "fcm_dpo/q_t": 0.4610481262207031,
      "grad_norm": 269.3840026855469,
      "learning_rate": 3.6567164179104475e-07,
      "logits/chosen": 0.08331212401390076,
      "logits/rejected": 0.05470721051096916,
      "logps/chosen": -53.09864807128906,
      "logps/ref_chosen": -52.85643768310547,
      "logps/ref_rejected": -72.17460632324219,
      "logps/rejected": -72.57521057128906,
      "loss": 1.2754,
      "margin_dpo/margin_mean": 0.15839970111846924,
      "margin_dpo/margin_std": 0.4189513921737671,
      "step": 50
    },
    {
      "KL/chosen_KL_mean": -0.25580406188964844,
      "KL/mean": -0.35211101174354553,
      "KL/rejected_KL_mean": -0.44841766357421875,
      "KL/std": 0.3892095685005188,
      "epoch": 0.07709750566893424,
      "fcm_dpo/beta": 1.0947012901306152,
      "fcm_dpo/delta": 0.1934581845998764,
      "fcm_dpo/margin": 0.1926138997077942,
      "fcm_dpo/q_t": 0.45432114601135254,
      "grad_norm": 290.8935852050781,
      "learning_rate": 3.7313432835820895e-07,
      "logits/chosen": 0.08420804142951965,
      "logits/rejected": 0.056402526795864105,
      "logps/chosen": -63.91224670410156,
      "logps/ref_chosen": -63.65644073486328,
      "logps/ref_rejected": -86.13229370117188,
      "logps/rejected": -86.58071899414062,
      "loss": 1.2721,
      "margin_dpo/margin_mean": 0.1926141083240509,
      "margin_dpo/margin_std": 0.547376275062561,
      "step": 51
    },
    {
      "KL/chosen_KL_mean": -0.30968475341796875,
      "KL/mean": -0.39266496896743774,
      "KL/rejected_KL_mean": -0.47564697265625,
      "KL/std": 0.37696564197540283,
      "epoch": 0.07860922146636433,
      "fcm_dpo/beta": 1.1430740356445312,
      "fcm_dpo/delta": 0.21437308192253113,
      "fcm_dpo/margin": 0.16595804691314697,
      "fcm_dpo/q_t": 0.45774364471435547,
      "grad_norm": 346.4855041503906,
      "learning_rate": 3.805970149253731e-07,
      "logits/chosen": 0.09279091656208038,
      "logits/rejected": 0.04256044328212738,
      "logps/chosen": -68.14990234375,
      "logps/ref_chosen": -67.8402099609375,
      "logps/ref_rejected": -96.97090911865234,
      "logps/rejected": -97.4465560913086,
      "loss": 1.2806,
      "margin_dpo/margin_mean": 0.165958434343338,
      "margin_dpo/margin_std": 0.49165236949920654,
      "step": 52
    },
    {
      "KL/chosen_KL_mean": -0.31960487365722656,
      "KL/mean": -0.4031504988670349,
      "KL/rejected_KL_mean": -0.4866962432861328,
      "KL/std": 0.3649352788925171,
      "epoch": 0.0801209372637944,
      "fcm_dpo/beta": 1.184340238571167,
      "fcm_dpo/delta": 0.20668236911296844,
      "fcm_dpo/margin": 0.16709211468696594,
      "fcm_dpo/q_t": 0.45340898633003235,
      "grad_norm": 301.8042907714844,
      "learning_rate": 3.880597014925373e-07,
      "logits/chosen": 0.06423115730285645,
      "logits/rejected": 0.05389813706278801,
      "logps/chosen": -57.19774627685547,
      "logps/ref_chosen": -56.87813949584961,
      "logps/ref_rejected": -60.75569152832031,
      "logps/rejected": -61.24238967895508,
      "loss": 1.2681,
      "margin_dpo/margin_mean": 0.16709229350090027,
      "margin_dpo/margin_std": 0.4409226179122925,
      "step": 53
    },
    {
      "KL/chosen_KL_mean": -0.3518028259277344,
      "KL/mean": -0.4071524739265442,
      "KL/rejected_KL_mean": -0.46249961853027344,
      "KL/std": 0.3316037654876709,
      "epoch": 0.08163265306122448,
      "fcm_dpo/beta": 1.2483994960784912,
      "fcm_dpo/delta": 0.26621297001838684,
      "fcm_dpo/margin": 0.11069938540458679,
      "fcm_dpo/q_t": 0.46753621101379395,
      "grad_norm": 328.0696716308594,
      "learning_rate": 3.9552238805970144e-07,
      "logits/chosen": 0.0451444610953331,
      "logits/rejected": 0.029947971925139427,
      "logps/chosen": -47.61872100830078,
      "logps/ref_chosen": -47.26692199707031,
      "logps/ref_rejected": -62.19426727294922,
      "logps/rejected": -62.65676498413086,
      "loss": 1.3134,
      "margin_dpo/margin_mean": 0.11069947481155396,
      "margin_dpo/margin_std": 0.3909778594970703,
      "step": 54
    },
    {
      "KL/chosen_KL_mean": -0.33339691162109375,
      "KL/mean": -0.4660353660583496,
      "KL/rejected_KL_mean": -0.5986747741699219,
      "KL/std": 0.46117573976516724,
      "epoch": 0.08314436885865457,
      "fcm_dpo/beta": 1.2880147695541382,
      "fcm_dpo/delta": 0.059539683163166046,
      "fcm_dpo/margin": 0.265281081199646,
      "fcm_dpo/q_t": 0.4293213486671448,
      "grad_norm": 368.900634765625,
      "learning_rate": 4.0298507462686564e-07,
      "logits/chosen": 0.03194599226117134,
      "logits/rejected": -0.044348280876874924,
      "logps/chosen": -50.65958786010742,
      "logps/ref_chosen": -50.32619094848633,
      "logps/ref_rejected": -92.44389343261719,
      "logps/rejected": -93.04256439208984,
      "loss": 1.2,
      "margin_dpo/margin_mean": 0.2652807831764221,
      "margin_dpo/margin_std": 0.5791739225387573,
      "step": 55
    },
    {
      "KL/chosen_KL_mean": -0.30806732177734375,
      "KL/mean": -0.40079766511917114,
      "KL/rejected_KL_mean": -0.49352455139160156,
      "KL/std": 0.39821040630340576,
      "epoch": 0.08465608465608465,
      "fcm_dpo/beta": 1.314407229423523,
      "fcm_dpo/delta": 0.1603386551141739,
      "fcm_dpo/margin": 0.18546155095100403,
      "fcm_dpo/q_t": 0.4450131058692932,
      "grad_norm": 353.42315673828125,
      "learning_rate": 4.1044776119402984e-07,
      "logits/chosen": 0.12745994329452515,
      "logits/rejected": 0.10495086014270782,
      "logps/chosen": -57.07503890991211,
      "logps/ref_chosen": -56.766971588134766,
      "logps/ref_rejected": -66.30504608154297,
      "logps/rejected": -66.79856872558594,
      "loss": 1.2562,
      "margin_dpo/margin_mean": 0.1854616105556488,
      "margin_dpo/margin_std": 0.4858906865119934,
      "step": 56
    },
    {
      "KL/chosen_KL_mean": -0.3892097473144531,
      "KL/mean": -0.5379469394683838,
      "KL/rejected_KL_mean": -0.6866836547851562,
      "KL/std": 0.4577900171279907,
      "epoch": 0.08616780045351474,
      "fcm_dpo/beta": 1.320029377937317,
      "fcm_dpo/delta": 0.006817173212766647,
      "fcm_dpo/margin": 0.29747310280799866,
      "fcm_dpo/q_t": 0.412424236536026,
      "grad_norm": 309.8792724609375,
      "learning_rate": 4.17910447761194e-07,
      "logits/chosen": 0.08991461247205734,
      "logits/rejected": 0.026150904595851898,
      "logps/chosen": -58.156951904296875,
      "logps/ref_chosen": -57.76774597167969,
      "logps/ref_rejected": -82.75698852539062,
      "logps/rejected": -83.44367218017578,
      "loss": 1.1477,
      "margin_dpo/margin_mean": 0.297473281621933,
      "margin_dpo/margin_std": 0.5286021828651428,
      "step": 57
    },
    {
      "KL/chosen_KL_mean": -0.43707275390625,
      "KL/mean": -0.5622156262397766,
      "KL/rejected_KL_mean": -0.6873550415039062,
      "KL/std": 0.5071960687637329,
      "epoch": 0.08767951625094482,
      "fcm_dpo/beta": 1.3271738290786743,
      "fcm_dpo/delta": 0.06818326562643051,
      "fcm_dpo/margin": 0.25028324127197266,
      "fcm_dpo/q_t": 0.43941134214401245,
      "grad_norm": 388.10906982421875,
      "learning_rate": 4.253731343283582e-07,
      "logits/chosen": 0.05029626190662384,
      "logits/rejected": 0.03480858355760574,
      "logps/chosen": -73.20115661621094,
      "logps/ref_chosen": -72.76408386230469,
      "logps/ref_rejected": -84.49275207519531,
      "logps/rejected": -85.18010711669922,
      "loss": 1.2563,
      "margin_dpo/margin_mean": 0.2502829134464264,
      "margin_dpo/margin_std": 0.6835585832595825,
      "step": 58
    },
    {
      "KL/chosen_KL_mean": -0.4274005889892578,
      "KL/mean": -0.531050980091095,
      "KL/rejected_KL_mean": -0.6347026824951172,
      "KL/std": 0.42696622014045715,
      "epoch": 0.08919123204837491,
      "fcm_dpo/beta": 1.3373160362243652,
      "fcm_dpo/delta": -0.07443632930517197,
      "fcm_dpo/margin": 0.20729583501815796,
      "fcm_dpo/q_t": 0.44409847259521484,
      "grad_norm": 324.54107666015625,
      "learning_rate": 4.3283582089552234e-07,
      "logits/chosen": 0.11193365603685379,
      "logits/rejected": 0.04682060331106186,
      "logps/chosen": -50.24817657470703,
      "logps/ref_chosen": -49.820777893066406,
      "logps/ref_rejected": -77.14368438720703,
      "logps/rejected": -77.77838897705078,
      "loss": 1.261,
      "margin_dpo/margin_mean": 0.20729640126228333,
      "margin_dpo/margin_std": 0.5395167469978333,
      "step": 59
    },
    {
      "KL/chosen_KL_mean": -0.45989227294921875,
      "KL/mean": -0.5035088062286377,
      "KL/rejected_KL_mean": -0.5471286773681641,
      "KL/std": 0.44240349531173706,
      "epoch": 0.09070294784580499,
      "fcm_dpo/beta": 1.377070665359497,
      "fcm_dpo/delta": 0.2842600643634796,
      "fcm_dpo/margin": 0.08723104000091553,
      "fcm_dpo/q_t": 0.4692135453224182,
      "grad_norm": 475.13653564453125,
      "learning_rate": 4.4029850746268654e-07,
      "logits/chosen": 0.11263720691204071,
      "logits/rejected": 0.11117991805076599,
      "logps/chosen": -63.68466567993164,
      "logps/ref_chosen": -63.22477340698242,
      "logps/ref_rejected": -61.360477447509766,
      "logps/rejected": -61.90760803222656,
      "loss": 1.4386,
      "margin_dpo/margin_mean": 0.08723104000091553,
      "margin_dpo/margin_std": 0.6071930527687073,
      "step": 60
    },
    {
      "KL/chosen_KL_mean": -0.5582065582275391,
      "KL/mean": -0.6111811399459839,
      "KL/rejected_KL_mean": -0.6641578674316406,
      "KL/std": 0.5173063278198242,
      "epoch": 0.09221466364323508,
      "fcm_dpo/beta": 1.4540350437164307,
      "fcm_dpo/delta": 0.25040096044540405,
      "fcm_dpo/margin": 0.10594852268695831,
      "fcm_dpo/q_t": 0.4708403944969177,
      "grad_norm": 456.5729064941406,
      "learning_rate": 4.4776119402985074e-07,
      "logits/chosen": 0.1330236792564392,
      "logits/rejected": 0.10015955567359924,
      "logps/chosen": -49.57500457763672,
      "logps/ref_chosen": -49.01679992675781,
      "logps/ref_rejected": -74.90817260742188,
      "logps/rejected": -75.57232666015625,
      "loss": 1.4719,
      "margin_dpo/margin_mean": 0.10594844818115234,
      "margin_dpo/margin_std": 0.699350118637085,
      "step": 61
    },
    {
      "KL/chosen_KL_mean": -0.5634193420410156,
      "KL/mean": -0.6708869338035583,
      "KL/rejected_KL_mean": -0.7783546447753906,
      "KL/std": 0.4782254099845886,
      "epoch": 0.09372637944066516,
      "fcm_dpo/beta": 1.5046186447143555,
      "fcm_dpo/delta": 0.07768788933753967,
      "fcm_dpo/margin": 0.2149357795715332,
      "fcm_dpo/q_t": 0.4257014989852905,
      "grad_norm": 442.23895263671875,
      "learning_rate": 4.552238805970149e-07,
      "logits/chosen": 0.08146628737449646,
      "logits/rejected": 0.04323853552341461,
      "logps/chosen": -63.31528854370117,
      "logps/ref_chosen": -62.751869201660156,
      "logps/ref_rejected": -78.93360900878906,
      "logps/rejected": -79.71195983886719,
      "loss": 1.3105,
      "margin_dpo/margin_mean": 0.21493586897850037,
      "margin_dpo/margin_std": 0.6780645847320557,
      "step": 62
    },
    {
      "KL/chosen_KL_mean": -0.4312572479248047,
      "KL/mean": -0.6460943818092346,
      "KL/rejected_KL_mean": -0.860931396484375,
      "KL/std": 0.46772170066833496,
      "epoch": 0.09523809523809523,
      "fcm_dpo/beta": 1.4502391815185547,
      "fcm_dpo/delta": -0.23768939077854156,
      "fcm_dpo/margin": 0.42967158555984497,
      "fcm_dpo/q_t": 0.36431044340133667,
      "grad_norm": 334.40960693359375,
      "learning_rate": 4.626865671641791e-07,
      "logits/chosen": 0.19986870884895325,
      "logits/rejected": 0.17387685179710388,
      "logps/chosen": -60.94650650024414,
      "logps/ref_chosen": -60.51525115966797,
      "logps/ref_rejected": -85.11021423339844,
      "logps/rejected": -85.97114562988281,
      "loss": 0.9869,
      "margin_dpo/margin_mean": 0.4296703636646271,
      "margin_dpo/margin_std": 0.5399540662765503,
      "step": 63
    },
    {
      "KL/chosen_KL_mean": -0.6149120330810547,
      "KL/mean": -0.6417762041091919,
      "KL/rejected_KL_mean": -0.6686439514160156,
      "KL/std": 0.4757128357887268,
      "epoch": 0.09674981103552532,
      "fcm_dpo/beta": 1.4943134784698486,
      "fcm_dpo/delta": 0.32342731952667236,
      "fcm_dpo/margin": 0.05373135209083557,
      "fcm_dpo/q_t": 0.4820947051048279,
      "grad_norm": 445.0806884765625,
      "learning_rate": 4.701492537313433e-07,
      "logits/chosen": 0.08707167208194733,
      "logits/rejected": 0.06187276542186737,
      "logps/chosen": -51.82175827026367,
      "logps/ref_chosen": -51.20684814453125,
      "logps/ref_rejected": -66.93081665039062,
      "logps/rejected": -67.5994644165039,
      "loss": 1.4673,
      "margin_dpo/margin_mean": 0.05373141169548035,
      "margin_dpo/margin_std": 0.5616258382797241,
      "step": 64
    },
    {
      "KL/chosen_KL_mean": -0.5739707946777344,
      "KL/mean": -0.7809446454048157,
      "KL/rejected_KL_mean": -0.9879226684570312,
      "KL/std": 0.5529853701591492,
      "epoch": 0.0982615268329554,
      "fcm_dpo/beta": 1.475801944732666,
      "fcm_dpo/delta": -0.22470661997795105,
      "fcm_dpo/margin": 0.41395121812820435,
      "fcm_dpo/q_t": 0.3666613698005676,
      "grad_norm": 357.36529541015625,
      "learning_rate": 4.776119402985074e-07,
      "logits/chosen": 0.19512414932250977,
      "logits/rejected": 0.1650513857603073,
      "logps/chosen": -67.86266326904297,
      "logps/ref_chosen": -67.2886962890625,
      "logps/ref_rejected": -74.44281005859375,
      "logps/rejected": -75.43072509765625,
      "loss": 1.0443,
      "margin_dpo/margin_mean": 0.4139505922794342,
      "margin_dpo/margin_std": 0.6147488355636597,
      "step": 65
    },
    {
      "KL/chosen_KL_mean": -0.5845527648925781,
      "KL/mean": -0.6933289766311646,
      "KL/rejected_KL_mean": -0.8021011352539062,
      "KL/std": 0.5075341463088989,
      "epoch": 0.09977324263038549,
      "fcm_dpo/beta": 1.4587290287017822,
      "fcm_dpo/delta": 0.08307772874832153,
      "fcm_dpo/margin": 0.21754857897758484,
      "fcm_dpo/q_t": 0.43842756748199463,
      "grad_norm": 425.11907958984375,
      "learning_rate": 4.850746268656717e-07,
      "logits/chosen": 0.08232487738132477,
      "logits/rejected": 0.058401718735694885,
      "logps/chosen": -71.32796478271484,
      "logps/ref_chosen": -70.743408203125,
      "logps/ref_rejected": -77.26499938964844,
      "logps/rejected": -78.06710815429688,
      "loss": 1.2968,
      "margin_dpo/margin_mean": 0.21754851937294006,
      "margin_dpo/margin_std": 0.6472800970077515,
      "step": 66
    },
    {
      "KL/chosen_KL_mean": -0.48465538024902344,
      "KL/mean": -0.5979866981506348,
      "KL/rejected_KL_mean": -0.7113189697265625,
      "KL/std": 0.5123389959335327,
      "epoch": 0.10128495842781557,
      "fcm_dpo/beta": 1.4985034465789795,
      "fcm_dpo/delta": 0.06241083890199661,
      "fcm_dpo/margin": 0.2266659140586853,
      "fcm_dpo/q_t": 0.4364425837993622,
      "grad_norm": 406.1687927246094,
      "learning_rate": 4.925373134328357e-07,
      "logits/chosen": 0.08061225712299347,
      "logits/rejected": 0.02438163198530674,
      "logps/chosen": -61.087257385253906,
      "logps/ref_chosen": -60.60260009765625,
      "logps/ref_rejected": -75.22235870361328,
      "logps/rejected": -75.93367767333984,
      "loss": 1.2757,
      "margin_dpo/margin_mean": 0.22666651010513306,
      "margin_dpo/margin_std": 0.6498265862464905,
      "step": 67
    },
    {
      "KL/chosen_KL_mean": -0.6951408386230469,
      "KL/mean": -0.8220087289810181,
      "KL/rejected_KL_mean": -0.9488754272460938,
      "KL/std": 0.558422327041626,
      "epoch": 0.10279667422524566,
      "fcm_dpo/beta": 1.494733214378357,
      "fcm_dpo/delta": 0.020372621715068817,
      "fcm_dpo/margin": 0.2537391185760498,
      "fcm_dpo/q_t": 0.42070841789245605,
      "grad_norm": 446.0521545410156,
      "learning_rate": 5e-07,
      "logits/chosen": 0.051371585577726364,
      "logits/rejected": 0.0212722048163414,
      "logps/chosen": -78.2235107421875,
      "logps/ref_chosen": -77.52836608886719,
      "logps/ref_rejected": -93.17778015136719,
      "logps/rejected": -94.12664794921875,
      "loss": 1.2602,
      "margin_dpo/margin_mean": 0.2537391185760498,
      "margin_dpo/margin_std": 0.6757462024688721,
      "step": 68
    },
    {
      "KL/chosen_KL_mean": -0.6366329193115234,
      "KL/mean": -0.8170256614685059,
      "KL/rejected_KL_mean": -0.9974136352539062,
      "KL/std": 0.5379056930541992,
      "epoch": 0.10430839002267574,
      "fcm_dpo/beta": 1.4895339012145996,
      "fcm_dpo/delta": -0.14553788304328918,
      "fcm_dpo/margin": 0.3607826232910156,
      "fcm_dpo/q_t": 0.39012402296066284,
      "grad_norm": 371.2175598144531,
      "learning_rate": 4.999965034812934e-07,
      "logits/chosen": 0.06672338396310806,
      "logits/rejected": 0.024326715618371964,
      "logps/chosen": -66.57968139648438,
      "logps/ref_chosen": -65.94305419921875,
      "logps/ref_rejected": -89.7735595703125,
      "logps/rejected": -90.7709732055664,
      "loss": 1.1082,
      "margin_dpo/margin_mean": 0.36078232526779175,
      "margin_dpo/margin_std": 0.6062155365943909,
      "step": 69
    },
    {
      "KL/chosen_KL_mean": -0.6548347473144531,
      "KL/mean": -0.7635809779167175,
      "KL/rejected_KL_mean": -0.8723239898681641,
      "KL/std": 0.4746021628379822,
      "epoch": 0.10582010582010581,
      "fcm_dpo/beta": 1.4715263843536377,
      "fcm_dpo/delta": 0.08196324110031128,
      "fcm_dpo/margin": 0.21749544143676758,
      "fcm_dpo/q_t": 0.42556124925613403,
      "grad_norm": 401.7762756347656,
      "learning_rate": 4.999860140229787e-07,
      "logits/chosen": 0.12438544631004333,
      "logits/rejected": 0.10102519392967224,
      "logps/chosen": -62.61274719238281,
      "logps/ref_chosen": -61.95791244506836,
      "logps/ref_rejected": -75.80945587158203,
      "logps/rejected": -76.68177795410156,
      "loss": 1.2699,
      "margin_dpo/margin_mean": 0.2174956500530243,
      "margin_dpo/margin_std": 0.5834441184997559,
      "step": 70
    },
    {
      "KL/chosen_KL_mean": -0.7265167236328125,
      "KL/mean": -0.8052693605422974,
      "KL/rejected_KL_mean": -0.8840236663818359,
      "KL/std": 0.5016933679580688,
      "epoch": 0.1073318216175359,
      "fcm_dpo/beta": 1.532849907875061,
      "fcm_dpo/delta": 0.1612829566001892,
      "fcm_dpo/margin": 0.15750658512115479,
      "fcm_dpo/q_t": 0.4435678720474243,
      "grad_norm": 440.24615478515625,
      "learning_rate": 4.999685319184688e-07,
      "logits/chosen": 0.09006579220294952,
      "logits/rejected": 0.07397979497909546,
      "logps/chosen": -64.0740966796875,
      "logps/ref_chosen": -63.34757995605469,
      "logps/ref_rejected": -67.49658203125,
      "logps/rejected": -68.38059997558594,
      "loss": 1.3472,
      "margin_dpo/margin_mean": 0.15750646591186523,
      "margin_dpo/margin_std": 0.601201057434082,
      "step": 71
    },
    {
      "KL/chosen_KL_mean": -0.6553325653076172,
      "KL/mean": -0.8184918165206909,
      "KL/rejected_KL_mean": -0.9816474914550781,
      "KL/std": 0.5385796427726746,
      "epoch": 0.10884353741496598,
      "fcm_dpo/beta": 1.5117855072021484,
      "fcm_dpo/delta": -0.09856587648391724,
      "fcm_dpo/margin": 0.3263169527053833,
      "fcm_dpo/q_t": 0.3978579342365265,
      "grad_norm": 408.5451354980469,
      "learning_rate": 4.999440576567755e-07,
      "logits/chosen": 0.08719843626022339,
      "logits/rejected": 0.02425486594438553,
      "logps/chosen": -56.51463317871094,
      "logps/ref_chosen": -55.85929870605469,
      "logps/ref_rejected": -68.45423889160156,
      "logps/rejected": -69.4358901977539,
      "loss": 1.1726,
      "margin_dpo/margin_mean": 0.3263167440891266,
      "margin_dpo/margin_std": 0.6592986583709717,
      "step": 72
    },
    {
      "KL/chosen_KL_mean": -0.9284744262695312,
      "KL/mean": -0.994033694267273,
      "KL/rejected_KL_mean": -1.0595951080322266,
      "KL/std": 0.6347705125808716,
      "epoch": 0.11035525321239607,
      "fcm_dpo/beta": 1.561780333518982,
      "fcm_dpo/delta": 0.19763168692588806,
      "fcm_dpo/margin": 0.1311270296573639,
      "fcm_dpo/q_t": 0.4644252061843872,
      "grad_norm": 534.00927734375,
      "learning_rate": 4.999125919224965e-07,
      "logits/chosen": 0.07961059361696243,
      "logits/rejected": 0.06521774083375931,
      "logps/chosen": -70.06727600097656,
      "logps/ref_chosen": -69.13880920410156,
      "logps/ref_rejected": -79.04586791992188,
      "logps/rejected": -80.10546875,
      "loss": 1.4815,
      "margin_dpo/margin_mean": 0.13112786412239075,
      "margin_dpo/margin_std": 0.736503005027771,
      "step": 73
    },
    {
      "KL/chosen_KL_mean": -0.6928367614746094,
      "KL/mean": -0.9160832166671753,
      "KL/rejected_KL_mean": -1.1393318176269531,
      "KL/std": 0.5641649961471558,
      "epoch": 0.11186696900982615,
      "fcm_dpo/beta": 1.5258018970489502,
      "fcm_dpo/delta": -0.3024459183216095,
      "fcm_dpo/margin": 0.4464932680130005,
      "fcm_dpo/q_t": 0.36906197667121887,
      "grad_norm": 340.6932373046875,
      "learning_rate": 4.998741355957963e-07,
      "logits/chosen": 0.10730817168951035,
      "logits/rejected": 0.05475199967622757,
      "logps/chosen": -50.616573333740234,
      "logps/ref_chosen": -49.923736572265625,
      "logps/ref_rejected": -81.73213958740234,
      "logps/rejected": -82.87147521972656,
      "loss": 1.0628,
      "margin_dpo/margin_mean": 0.4464929401874542,
      "margin_dpo/margin_std": 0.6866585612297058,
      "step": 74
    },
    {
      "KL/chosen_KL_mean": -0.7750892639160156,
      "KL/mean": -0.9859875440597534,
      "KL/rejected_KL_mean": -1.1968860626220703,
      "KL/std": 0.6085139513015747,
      "epoch": 0.11337868480725624,
      "fcm_dpo/beta": 1.419792652130127,
      "fcm_dpo/delta": -0.21434250473976135,
      "fcm_dpo/margin": 0.4217972755432129,
      "fcm_dpo/q_t": 0.3794889450073242,
      "grad_norm": 300.08807373046875,
      "learning_rate": 4.998286897523808e-07,
      "logits/chosen": 0.07163898646831512,
      "logits/rejected": 0.040832214057445526,
      "logps/chosen": -46.843841552734375,
      "logps/ref_chosen": -46.06875228881836,
      "logps/ref_rejected": -66.1181411743164,
      "logps/rejected": -67.31503295898438,
      "loss": 1.0958,
      "margin_dpo/margin_mean": 0.42179712653160095,
      "margin_dpo/margin_std": 0.7199387550354004,
      "step": 75
    },
    {
      "KL/chosen_KL_mean": -0.7950630187988281,
      "KL/mean": -0.9406583309173584,
      "KL/rejected_KL_mean": -1.0862560272216797,
      "KL/std": 0.6709940433502197,
      "epoch": 0.11489040060468632,
      "fcm_dpo/beta": 1.4085371494293213,
      "fcm_dpo/delta": -0.010789202526211739,
      "fcm_dpo/margin": 0.2911915183067322,
      "fcm_dpo/q_t": 0.42344167828559875,
      "grad_norm": 378.4490966796875,
      "learning_rate": 4.997762556634679e-07,
      "logits/chosen": 0.09013240784406662,
      "logits/rejected": 0.04698844254016876,
      "logps/chosen": -54.857810974121094,
      "logps/ref_chosen": -54.06275177001953,
      "logps/ref_rejected": -74.87464141845703,
      "logps/rejected": -75.96089935302734,
      "loss": 1.2479,
      "margin_dpo/margin_mean": 0.2911912798881531,
      "margin_dpo/margin_std": 0.7412244081497192,
      "step": 76
    },
    {
      "KL/chosen_KL_mean": -0.9427947998046875,
      "KL/mean": -1.0837860107421875,
      "KL/rejected_KL_mean": -1.2247734069824219,
      "KL/std": 0.5913428068161011,
      "epoch": 0.1164021164021164,
      "fcm_dpo/beta": 1.3965743780136108,
      "fcm_dpo/delta": 0.003795936703681946,
      "fcm_dpo/margin": 0.2819800078868866,
      "fcm_dpo/q_t": 0.41928815841674805,
      "grad_norm": 390.9847412109375,
      "learning_rate": 4.99716834795752e-07,
      "logits/chosen": 0.12903091311454773,
      "logits/rejected": 0.08758988976478577,
      "logps/chosen": -54.01888656616211,
      "logps/ref_chosen": -53.07609176635742,
      "logps/ref_rejected": -74.45601654052734,
      "logps/rejected": -75.6807861328125,
      "loss": 1.2229,
      "margin_dpo/margin_mean": 0.28198015689849854,
      "margin_dpo/margin_std": 0.6389899253845215,
      "step": 77
    },
    {
      "KL/chosen_KL_mean": -0.8560752868652344,
      "KL/mean": -0.9924242496490479,
      "KL/rejected_KL_mean": -1.1287765502929688,
      "KL/std": 0.5523614287376404,
      "epoch": 0.11791383219954649,
      "fcm_dpo/beta": 1.4131134748458862,
      "fcm_dpo/delta": 0.015169057063758373,
      "fcm_dpo/margin": 0.27270209789276123,
      "fcm_dpo/q_t": 0.42009180784225464,
      "grad_norm": 419.772216796875,
      "learning_rate": 4.996504288113623e-07,
      "logits/chosen": 0.07726676762104034,
      "logits/rejected": 0.05717945471405983,
      "logps/chosen": -68.58149719238281,
      "logps/ref_chosen": -67.72541809082031,
      "logps/ref_rejected": -79.03926849365234,
      "logps/rejected": -80.16804504394531,
      "loss": 1.2435,
      "margin_dpo/margin_mean": 0.27270200848579407,
      "margin_dpo/margin_std": 0.673001766204834,
      "step": 78
    },
    {
      "KL/chosen_KL_mean": -0.9067211151123047,
      "KL/mean": -1.1093153953552246,
      "KL/rejected_KL_mean": -1.3119125366210938,
      "KL/std": 0.602173924446106,
      "epoch": 0.11942554799697656,
      "fcm_dpo/beta": 1.38477623462677,
      "fcm_dpo/delta": -0.17040225863456726,
      "fcm_dpo/margin": 0.4051881730556488,
      "fcm_dpo/q_t": 0.37906450033187866,
      "grad_norm": 295.46978759765625,
      "learning_rate": 4.995770395678171e-07,
      "logits/chosen": 0.13399448990821838,
      "logits/rejected": 0.07460729777812958,
      "logps/chosen": -53.06736755371094,
      "logps/ref_chosen": -52.16064453125,
      "logps/ref_rejected": -83.31062316894531,
      "logps/rejected": -84.62252807617188,
      "loss": 1.0786,
      "margin_dpo/margin_mean": 0.4051882326602936,
      "margin_dpo/margin_std": 0.6499325037002563,
      "step": 79
    },
    {
      "KL/chosen_KL_mean": -0.9616222381591797,
      "KL/mean": -1.1492910385131836,
      "KL/rejected_KL_mean": -1.3369598388671875,
      "KL/std": 0.6497531533241272,
      "epoch": 0.12093726379440665,
      "fcm_dpo/beta": 1.3587586879730225,
      "fcm_dpo/delta": -0.11657415330410004,
      "fcm_dpo/margin": 0.3753415048122406,
      "fcm_dpo/q_t": 0.39891305565834045,
      "grad_norm": 363.4054260253906,
      "learning_rate": 4.994966691179711e-07,
      "logits/chosen": 0.12045355886220932,
      "logits/rejected": 0.06095130741596222,
      "logps/chosen": -62.37218475341797,
      "logps/ref_chosen": -61.410560607910156,
      "logps/ref_rejected": -78.66004943847656,
      "logps/rejected": -79.99700927734375,
      "loss": 1.146,
      "margin_dpo/margin_mean": 0.37534135580062866,
      "margin_dpo/margin_std": 0.7054092884063721,
      "step": 80
    },
    {
      "KL/chosen_KL_mean": -1.0198345184326172,
      "KL/mean": -1.2168505191802979,
      "KL/rejected_KL_mean": -1.4138603210449219,
      "KL/std": 0.6422163248062134,
      "epoch": 0.12244897959183673,
      "fcm_dpo/beta": 1.3078601360321045,
      "fcm_dpo/delta": -0.12261784076690674,
      "fcm_dpo/margin": 0.39402908086776733,
      "fcm_dpo/q_t": 0.3908241391181946,
      "grad_norm": 330.7373962402344,
      "learning_rate": 4.994093197099587e-07,
      "logits/chosen": 0.07247719168663025,
      "logits/rejected": 0.03978278860449791,
      "logps/chosen": -64.82420349121094,
      "logps/ref_chosen": -63.80437088012695,
      "logps/ref_rejected": -79.3484115600586,
      "logps/rejected": -80.76226806640625,
      "loss": 1.0932,
      "margin_dpo/margin_mean": 0.3940291702747345,
      "margin_dpo/margin_std": 0.64724200963974,
      "step": 81
    },
    {
      "KL/chosen_KL_mean": -0.8772220611572266,
      "KL/mean": -1.1485819816589355,
      "KL/rejected_KL_mean": -1.419942855834961,
      "KL/std": 0.6433102488517761,
      "epoch": 0.12396069538926682,
      "fcm_dpo/beta": 1.2471857070922852,
      "fcm_dpo/delta": -0.2975814938545227,
      "fcm_dpo/margin": 0.542713463306427,
      "fcm_dpo/q_t": 0.34925180673599243,
      "grad_norm": 278.6336975097656,
      "learning_rate": 4.993149937871306e-07,
      "logits/chosen": 0.05593370646238327,
      "logits/rejected": -0.0052494630217552185,
      "logps/chosen": -49.69511413574219,
      "logps/ref_chosen": -48.817893981933594,
      "logps/ref_rejected": -70.31497955322266,
      "logps/rejected": -71.73492431640625,
      "loss": 0.964,
      "margin_dpo/margin_mean": 0.5427138209342957,
      "margin_dpo/margin_std": 0.651484489440918,
      "step": 82
    },
    {
      "KL/chosen_KL_mean": -0.9735927581787109,
      "KL/mean": -1.2231380939483643,
      "KL/rejected_KL_mean": -1.472686767578125,
      "KL/std": 0.6370868682861328,
      "epoch": 0.1254724111866969,
      "fcm_dpo/beta": 1.1937222480773926,
      "fcm_dpo/delta": -0.2080550491809845,
      "fcm_dpo/margin": 0.4990951120853424,
      "fcm_dpo/q_t": 0.37340906262397766,
      "grad_norm": 277.4635925292969,
      "learning_rate": 4.992136939879856e-07,
      "logits/chosen": 0.1766347587108612,
      "logits/rejected": 0.12540043890476227,
      "logps/chosen": -58.124366760253906,
      "logps/ref_chosen": -57.15077209472656,
      "logps/ref_rejected": -75.1710205078125,
      "logps/rejected": -76.64370727539062,
      "loss": 1.0473,
      "margin_dpo/margin_mean": 0.49909481406211853,
      "margin_dpo/margin_std": 0.7604484558105469,
      "step": 83
    },
    {
      "KL/chosen_KL_mean": -1.1488761901855469,
      "KL/mean": -1.290728211402893,
      "KL/rejected_KL_mean": -1.4325942993164062,
      "KL/std": 0.730034589767456,
      "epoch": 0.12698412698412698,
      "fcm_dpo/beta": 1.1920582056045532,
      "fcm_dpo/delta": 0.06391885131597519,
      "fcm_dpo/margin": 0.28371578454971313,
      "fcm_dpo/q_t": 0.42771369218826294,
      "grad_norm": 378.65191650390625,
      "learning_rate": 4.991054231460969e-07,
      "logits/chosen": 0.11678484082221985,
      "logits/rejected": 0.0760355293750763,
      "logps/chosen": -65.92617797851562,
      "logps/ref_chosen": -64.77729797363281,
      "logps/ref_rejected": -84.71949768066406,
      "logps/rejected": -86.15209197998047,
      "loss": 1.2627,
      "margin_dpo/margin_mean": 0.2837153375148773,
      "margin_dpo/margin_std": 0.7518688440322876,
      "step": 84
    },
    {
      "KL/chosen_KL_mean": -0.9664173126220703,
      "KL/mean": -1.2492319345474243,
      "KL/rejected_KL_mean": -1.53204345703125,
      "KL/std": 0.746864914894104,
      "epoch": 0.12849584278155707,
      "fcm_dpo/beta": 1.1534860134124756,
      "fcm_dpo/delta": -0.26930180191993713,
      "fcm_dpo/margin": 0.5656286478042603,
      "fcm_dpo/q_t": 0.3612124025821686,
      "grad_norm": 289.289794921875,
      "learning_rate": 4.989901842900325e-07,
      "logits/chosen": 0.09598151594400406,
      "logits/rejected": 0.05402902513742447,
      "logps/chosen": -51.21811294555664,
      "logps/ref_chosen": -50.25169372558594,
      "logps/ref_rejected": -66.55439758300781,
      "logps/rejected": -68.08644104003906,
      "loss": 1.0391,
      "margin_dpo/margin_mean": 0.5656284093856812,
      "margin_dpo/margin_std": 0.8260301351547241,
      "step": 85
    },
    {
      "KL/chosen_KL_mean": -1.0875301361083984,
      "KL/mean": -1.324817419052124,
      "KL/rejected_KL_mean": -1.562103271484375,
      "KL/std": 0.7085767388343811,
      "epoch": 0.13000755857898716,
      "fcm_dpo/beta": 1.1074466705322266,
      "fcm_dpo/delta": -0.13303929567337036,
      "fcm_dpo/margin": 0.47456830739974976,
      "fcm_dpo/q_t": 0.3882756233215332,
      "grad_norm": 245.15232849121094,
      "learning_rate": 4.988679806432711e-07,
      "logits/chosen": 0.14248552918434143,
      "logits/rejected": 0.12411345541477203,
      "logps/chosen": -61.81671142578125,
      "logps/ref_chosen": -60.72917938232422,
      "logps/ref_rejected": -72.30961608886719,
      "logps/rejected": -73.87171936035156,
      "loss": 1.0649,
      "margin_dpo/margin_mean": 0.47456836700439453,
      "margin_dpo/margin_std": 0.7193993330001831,
      "step": 86
    },
    {
      "KL/chosen_KL_mean": -1.125885009765625,
      "KL/mean": -1.3568286895751953,
      "KL/rejected_KL_mean": -1.5877647399902344,
      "KL/std": 0.8056492805480957,
      "epoch": 0.13151927437641722,
      "fcm_dpo/beta": 1.076570987701416,
      "fcm_dpo/delta": -0.1047540009021759,
      "fcm_dpo/margin": 0.46188533306121826,
      "fcm_dpo/q_t": 0.3923509120941162,
      "grad_norm": 332.5335388183594,
      "learning_rate": 4.987388156241114e-07,
      "logits/chosen": 0.10630130767822266,
      "logits/rejected": 0.04966554418206215,
      "logps/chosen": -66.88385009765625,
      "logps/ref_chosen": -65.75796508789062,
      "logps/ref_rejected": -84.81159973144531,
      "logps/rejected": -86.39936065673828,
      "loss": 1.1763,
      "margin_dpo/margin_mean": 0.4618854522705078,
      "margin_dpo/margin_std": 0.9339560270309448,
      "step": 87
    },
    {
      "KL/chosen_KL_mean": -1.1211109161376953,
      "KL/mean": -1.355789303779602,
      "KL/rejected_KL_mean": -1.5904655456542969,
      "KL/std": 0.7843002676963806,
      "epoch": 0.1330309901738473,
      "fcm_dpo/beta": 1.0574965476989746,
      "fcm_dpo/delta": -0.10266615450382233,
      "fcm_dpo/margin": 0.46935614943504333,
      "fcm_dpo/q_t": 0.3971374034881592,
      "grad_norm": 295.2401428222656,
      "learning_rate": 4.986026928455767e-07,
      "logits/chosen": 0.14652103185653687,
      "logits/rejected": 0.12086239457130432,
      "logps/chosen": -63.94513702392578,
      "logps/ref_chosen": -62.82402801513672,
      "logps/ref_rejected": -74.9607162475586,
      "logps/rejected": -76.55117797851562,
      "loss": 1.1424,
      "margin_dpo/margin_mean": 0.4693562090396881,
      "margin_dpo/margin_std": 0.8767245411872864,
      "step": 88
    },
    {
      "KL/chosen_KL_mean": -0.9972019195556641,
      "KL/mean": -1.2437442541122437,
      "KL/rejected_KL_mean": -1.4902877807617188,
      "KL/std": 0.8000632524490356,
      "epoch": 0.1345427059712774,
      "fcm_dpo/beta": 1.0466407537460327,
      "fcm_dpo/delta": -0.12258999794721603,
      "fcm_dpo/margin": 0.49308302998542786,
      "fcm_dpo/q_t": 0.38694024085998535,
      "grad_norm": 262.8199157714844,
      "learning_rate": 4.984596161153135e-07,
      "logits/chosen": 0.2086452841758728,
      "logits/rejected": 0.12626194953918457,
      "logps/chosen": -42.188636779785156,
      "logps/ref_chosen": -41.191436767578125,
      "logps/ref_rejected": -85.44769287109375,
      "logps/rejected": -86.93798828125,
      "loss": 1.1202,
      "margin_dpo/margin_mean": 0.4930838346481323,
      "margin_dpo/margin_std": 0.8570024967193604,
      "step": 89
    },
    {
      "KL/chosen_KL_mean": -1.1763248443603516,
      "KL/mean": -1.383636713027954,
      "KL/rejected_KL_mean": -1.5909461975097656,
      "KL/std": 0.7591300010681152,
      "epoch": 0.1360544217687075,
      "fcm_dpo/beta": 1.0314490795135498,
      "fcm_dpo/delta": -0.0290251225233078,
      "fcm_dpo/margin": 0.4146195352077484,
      "fcm_dpo/q_t": 0.4088814854621887,
      "grad_norm": 280.2547302246094,
      "learning_rate": 4.983095894354857e-07,
      "logits/chosen": 0.07813448458909988,
      "logits/rejected": 0.02608073130249977,
      "logps/chosen": -57.760231018066406,
      "logps/ref_chosen": -56.58390808105469,
      "logps/ref_rejected": -86.86978149414062,
      "logps/rejected": -88.46072387695312,
      "loss": 1.1623,
      "margin_dpo/margin_mean": 0.4146192967891693,
      "margin_dpo/margin_std": 0.8090165853500366,
      "step": 90
    },
    {
      "KL/chosen_KL_mean": -1.1240406036376953,
      "KL/mean": -1.4229687452316284,
      "KL/rejected_KL_mean": -1.7218971252441406,
      "KL/std": 0.8534637093544006,
      "epoch": 0.13756613756613756,
      "fcm_dpo/beta": 0.9964852333068848,
      "fcm_dpo/delta": -0.207942932844162,
      "fcm_dpo/margin": 0.5978548526763916,
      "fcm_dpo/q_t": 0.37541231513023376,
      "grad_norm": 219.49380493164062,
      "learning_rate": 4.98152617002662e-07,
      "logits/chosen": 0.10854806005954742,
      "logits/rejected": 0.06545954942703247,
      "logps/chosen": -53.506385803222656,
      "logps/ref_chosen": -52.38234329223633,
      "logps/ref_rejected": -72.17642211914062,
      "logps/rejected": -73.8983154296875,
      "loss": 1.0439,
      "margin_dpo/margin_mean": 0.5978548526763916,
      "margin_dpo/margin_std": 0.8873600959777832,
      "step": 91
    },
    {
      "KL/chosen_KL_mean": -1.245452880859375,
      "KL/mean": -1.4888556003570557,
      "KL/rejected_KL_mean": -1.7322654724121094,
      "KL/std": 0.8518181443214417,
      "epoch": 0.13907785336356765,
      "fcm_dpo/beta": 0.9819341897964478,
      "fcm_dpo/delta": -0.08298197388648987,
      "fcm_dpo/margin": 0.4868069291114807,
      "fcm_dpo/q_t": 0.3972088098526001,
      "grad_norm": 227.3011016845703,
      "learning_rate": 4.979887032076988e-07,
      "logits/chosen": 0.1596505492925644,
      "logits/rejected": 0.12032721936702728,
      "logps/chosen": -54.254154205322266,
      "logps/ref_chosen": -53.00870132446289,
      "logps/ref_rejected": -79.77812957763672,
      "logps/rejected": -81.51039123535156,
      "loss": 1.1362,
      "margin_dpo/margin_mean": 0.4868068993091583,
      "margin_dpo/margin_std": 0.8662936687469482,
      "step": 92
    },
    {
      "KL/chosen_KL_mean": -1.30902099609375,
      "KL/mean": -1.531526803970337,
      "KL/rejected_KL_mean": -1.7540321350097656,
      "KL/std": 0.8664084672927856,
      "epoch": 0.14058956916099774,
      "fcm_dpo/beta": 0.9646643400192261,
      "fcm_dpo/delta": -0.030587686225771904,
      "fcm_dpo/margin": 0.4450099766254425,
      "fcm_dpo/q_t": 0.4068792164325714,
      "grad_norm": 209.89547729492188,
      "learning_rate": 4.978178526356172e-07,
      "logits/chosen": 0.10871215164661407,
      "logits/rejected": 0.08188057690858841,
      "logps/chosen": -46.21607208251953,
      "logps/ref_chosen": -44.90705108642578,
      "logps/ref_rejected": -58.7879524230957,
      "logps/rejected": -60.54198455810547,
      "loss": 1.1758,
      "margin_dpo/margin_mean": 0.44500961899757385,
      "margin_dpo/margin_std": 0.9137963056564331,
      "step": 93
    },
    {
      "KL/chosen_KL_mean": -1.0369033813476562,
      "KL/mean": -1.324483871459961,
      "KL/rejected_KL_mean": -1.612060546875,
      "KL/std": 0.8918026685714722,
      "epoch": 0.1421012849584278,
      "fcm_dpo/beta": 0.9319857358932495,
      "fcm_dpo/delta": -0.147476464509964,
      "fcm_dpo/margin": 0.5751581192016602,
      "fcm_dpo/q_t": 0.38574111461639404,
      "grad_norm": 235.57188415527344,
      "learning_rate": 4.976400700654751e-07,
      "logits/chosen": 0.16680672764778137,
      "logits/rejected": 0.1295737624168396,
      "logps/chosen": -60.974674224853516,
      "logps/ref_chosen": -59.93777084350586,
      "logps/ref_rejected": -79.3138427734375,
      "logps/rejected": -80.9259033203125,
      "loss": 1.1527,
      "margin_dpo/margin_mean": 0.5751579999923706,
      "margin_dpo/margin_std": 1.1100279092788696,
      "step": 94
    },
    {
      "KL/chosen_KL_mean": -1.3213310241699219,
      "KL/mean": -1.6439839601516724,
      "KL/rejected_KL_mean": -1.9666366577148438,
      "KL/std": 0.838896632194519,
      "epoch": 0.1436130007558579,
      "fcm_dpo/beta": 0.9028377532958984,
      "fcm_dpo/delta": -0.1956050544977188,
      "fcm_dpo/margin": 0.6453028917312622,
      "fcm_dpo/q_t": 0.37172389030456543,
      "grad_norm": 220.64878845214844,
      "learning_rate": 4.974553604702332e-07,
      "logits/chosen": 0.07326029241085052,
      "logits/rejected": 0.012177206575870514,
      "logps/chosen": -61.48982238769531,
      "logps/ref_chosen": -60.168487548828125,
      "logps/ref_rejected": -90.73665618896484,
      "logps/rejected": -92.70329284667969,
      "loss": 1.0313,
      "margin_dpo/margin_mean": 0.6453026533126831,
      "margin_dpo/margin_std": 0.8951278328895569,
      "step": 95
    },
    {
      "KL/chosen_KL_mean": -1.2811088562011719,
      "KL/mean": -1.5985305309295654,
      "KL/rejected_KL_mean": -1.91595458984375,
      "KL/std": 0.8724809885025024,
      "epoch": 0.14512471655328799,
      "fcm_dpo/beta": 0.8822429180145264,
      "fcm_dpo/delta": -0.16936419904232025,
      "fcm_dpo/margin": 0.634851336479187,
      "fcm_dpo/q_t": 0.37748467922210693,
      "grad_norm": 206.07525634765625,
      "learning_rate": 4.972637290166157e-07,
      "logits/chosen": 0.1217946782708168,
      "logits/rejected": 0.07989707589149475,
      "logps/chosen": -61.949886322021484,
      "logps/ref_chosen": -60.66877746582031,
      "logps/ref_rejected": -88.30673217773438,
      "logps/rejected": -90.22268676757812,
      "loss": 1.0672,
      "margin_dpo/margin_mean": 0.6348508596420288,
      "margin_dpo/margin_std": 0.949616551399231,
      "step": 96
    },
    {
      "KL/chosen_KL_mean": -1.5132274627685547,
      "KL/mean": -1.7518854141235352,
      "KL/rejected_KL_mean": -1.9905433654785156,
      "KL/std": 0.8738954067230225,
      "epoch": 0.14663643235071808,
      "fcm_dpo/beta": 0.8598639965057373,
      "fcm_dpo/delta": -0.012458987534046173,
      "fcm_dpo/margin": 0.4773203432559967,
      "fcm_dpo/q_t": 0.4190227687358856,
      "grad_norm": 245.85345458984375,
      "learning_rate": 4.970651810649666e-07,
      "logits/chosen": 0.06072793900966644,
      "logits/rejected": 0.01822689175605774,
      "logps/chosen": -66.5573501586914,
      "logps/ref_chosen": -65.04412078857422,
      "logps/ref_rejected": -78.42092895507812,
      "logps/rejected": -80.41146850585938,
      "loss": 1.192,
      "margin_dpo/margin_mean": 0.47732043266296387,
      "margin_dpo/margin_std": 1.0347280502319336,
      "step": 97
    },
    {
      "KL/chosen_KL_mean": -1.2721519470214844,
      "KL/mean": -1.4518141746520996,
      "KL/rejected_KL_mean": -1.6314773559570312,
      "KL/std": 0.8410429954528809,
      "epoch": 0.14814814814814814,
      "fcm_dpo/beta": 0.8739485144615173,
      "fcm_dpo/delta": 0.08870487660169601,
      "fcm_dpo/margin": 0.3593263030052185,
      "fcm_dpo/q_t": 0.43300604820251465,
      "grad_norm": 235.42140197753906,
      "learning_rate": 4.968597221690985e-07,
      "logits/chosen": 0.1716204583644867,
      "logits/rejected": 0.14437055587768555,
      "logps/chosen": -56.77538299560547,
      "logps/ref_chosen": -55.503231048583984,
      "logps/ref_rejected": -72.81553649902344,
      "logps/rejected": -74.447021484375,
      "loss": 1.2292,
      "margin_dpo/margin_mean": 0.35932624340057373,
      "margin_dpo/margin_std": 0.8575361967086792,
      "step": 98
    },
    {
      "KL/chosen_KL_mean": -1.261861801147461,
      "KL/mean": -1.5936691761016846,
      "KL/rejected_KL_mean": -1.92547607421875,
      "KL/std": 1.0141127109527588,
      "epoch": 0.14965986394557823,
      "fcm_dpo/beta": 0.8584821224212646,
      "fcm_dpo/delta": -0.1800466775894165,
      "fcm_dpo/margin": 0.663617730140686,
      "fcm_dpo/q_t": 0.38134729862213135,
      "grad_norm": 223.2831268310547,
      "learning_rate": 4.966473580761389e-07,
      "logits/chosen": 0.15894950926303864,
      "logits/rejected": 0.12229165434837341,
      "logps/chosen": -59.837501525878906,
      "logps/ref_chosen": -58.57563781738281,
      "logps/ref_rejected": -78.693603515625,
      "logps/rejected": -80.61907958984375,
      "loss": 1.0597,
      "margin_dpo/margin_mean": 0.6636172533035278,
      "margin_dpo/margin_std": 1.0391426086425781,
      "step": 99
    },
    {
      "KL/chosen_KL_mean": -1.4188766479492188,
      "KL/mean": -1.6849563121795654,
      "KL/rejected_KL_mean": -1.9510383605957031,
      "KL/std": 0.9670500755310059,
      "epoch": 0.15117157974300832,
      "fcm_dpo/beta": 0.8525861501693726,
      "fcm_dpo/delta": -0.05790305882692337,
      "fcm_dpo/margin": 0.5321630239486694,
      "fcm_dpo/q_t": 0.4141319692134857,
      "grad_norm": 253.90597534179688,
      "learning_rate": 4.964280947263676e-07,
      "logits/chosen": 0.12935097515583038,
      "logits/rejected": 0.12235041707754135,
      "logps/chosen": -81.00231170654297,
      "logps/ref_chosen": -79.58343505859375,
      "logps/ref_rejected": -92.152587890625,
      "logps/rejected": -94.10362243652344,
      "loss": 1.2366,
      "margin_dpo/margin_mean": 0.5321632623672485,
      "margin_dpo/margin_std": 1.2714905738830566,
      "step": 100
    },
    {
      "KL/chosen_KL_mean": -1.2464981079101562,
      "KL/mean": -1.6342787742614746,
      "KL/rejected_KL_mean": -2.0220565795898438,
      "KL/std": 0.988823652267456,
      "epoch": 0.15268329554043839,
      "fcm_dpo/beta": 0.8176276683807373,
      "fcm_dpo/delta": -0.2493351250886917,
      "fcm_dpo/margin": 0.7755611538887024,
      "fcm_dpo/q_t": 0.36288192868232727,
      "grad_norm": 168.5565948486328,
      "learning_rate": 4.96201938253052e-07,
      "logits/chosen": 0.15569889545440674,
      "logits/rejected": 0.11886613070964813,
      "logps/chosen": -53.57928466796875,
      "logps/ref_chosen": -52.332786560058594,
      "logps/ref_rejected": -69.55589294433594,
      "logps/rejected": -71.57795715332031,
      "loss": 0.9804,
      "margin_dpo/margin_mean": 0.7755616903305054,
      "margin_dpo/margin_std": 0.9270931482315063,
      "step": 101
    },
    {
      "KL/chosen_KL_mean": -1.4936847686767578,
      "KL/mean": -1.764020323753357,
      "KL/rejected_KL_mean": -2.03436279296875,
      "KL/std": 0.9787428379058838,
      "epoch": 0.15419501133786848,
      "fcm_dpo/beta": 0.7976725101470947,
      "fcm_dpo/delta": -0.03269146382808685,
      "fcm_dpo/margin": 0.5406721234321594,
      "fcm_dpo/q_t": 0.40772902965545654,
      "grad_norm": 206.413818359375,
      "learning_rate": 4.959688949822748e-07,
      "logits/chosen": 0.0799262598156929,
      "logits/rejected": 0.04157480597496033,
      "logps/chosen": -66.23716735839844,
      "logps/ref_chosen": -64.74348449707031,
      "logps/ref_rejected": -69.06132507324219,
      "logps/rejected": -71.09568786621094,
      "loss": 1.1749,
      "margin_dpo/margin_mean": 0.5406715869903564,
      "margin_dpo/margin_std": 1.1025452613830566,
      "step": 102
    },
    {
      "KL/chosen_KL_mean": -1.4423542022705078,
      "KL/mean": -1.7633211612701416,
      "KL/rejected_KL_mean": -2.08428955078125,
      "KL/std": 1.0820667743682861,
      "epoch": 0.15570672713529857,
      "fcm_dpo/beta": 0.7843052744865417,
      "fcm_dpo/delta": -0.10881097614765167,
      "fcm_dpo/margin": 0.6419306993484497,
      "fcm_dpo/q_t": 0.3946912884712219,
      "grad_norm": 215.6351318359375,
      "learning_rate": 4.957289714327572e-07,
      "logits/chosen": 0.16433322429656982,
      "logits/rejected": 0.13270039856433868,
      "logps/chosen": -65.27899932861328,
      "logps/ref_chosen": -63.83664321899414,
      "logps/ref_rejected": -79.32362365722656,
      "logps/rejected": -81.40791320800781,
      "loss": 1.1226,
      "margin_dpo/margin_mean": 0.6419302821159363,
      "margin_dpo/margin_std": 1.135719656944275,
      "step": 103
    },
    {
      "KL/chosen_KL_mean": -1.387624740600586,
      "KL/mean": -1.7114492654800415,
      "KL/rejected_KL_mean": -2.0352783203125,
      "KL/std": 1.1588332653045654,
      "epoch": 0.15721844293272866,
      "fcm_dpo/beta": 0.7690836787223816,
      "fcm_dpo/delta": -0.10314887762069702,
      "fcm_dpo/margin": 0.6476492285728455,
      "fcm_dpo/q_t": 0.39995285868644714,
      "grad_norm": 215.02183532714844,
      "learning_rate": 4.954821743156767e-07,
      "logits/chosen": 0.17702050507068634,
      "logits/rejected": 0.09312883764505386,
      "logps/chosen": -62.38683319091797,
      "logps/ref_chosen": -60.99920654296875,
      "logps/ref_rejected": -98.84645080566406,
      "logps/rejected": -100.88172912597656,
      "loss": 1.1088,
      "margin_dpo/margin_mean": 0.6476488709449768,
      "margin_dpo/margin_std": 1.122774600982666,
      "step": 104
    },
    {
      "KL/chosen_KL_mean": -1.458160400390625,
      "KL/mean": -1.7229704856872559,
      "KL/rejected_KL_mean": -1.9877815246582031,
      "KL/std": 1.0482707023620605,
      "epoch": 0.15873015873015872,
      "fcm_dpo/beta": 0.7652902007102966,
      "fcm_dpo/delta": -0.006055701524019241,
      "fcm_dpo/margin": 0.529621958732605,
      "fcm_dpo/q_t": 0.4152873754501343,
      "grad_norm": 237.42147827148438,
      "learning_rate": 4.952285105344791e-07,
      "logits/chosen": 0.09127002209424973,
      "logits/rejected": 0.04044020175933838,
      "logps/chosen": -72.40843200683594,
      "logps/ref_chosen": -70.95027160644531,
      "logps/ref_rejected": -87.88340759277344,
      "logps/rejected": -89.8711929321289,
      "loss": 1.1959,
      "margin_dpo/margin_mean": 0.5296220779418945,
      "margin_dpo/margin_std": 1.151416301727295,
      "step": 105
    },
    {
      "KL/chosen_KL_mean": -1.3451671600341797,
      "KL/mean": -1.6426818370819092,
      "KL/rejected_KL_mean": -1.9401969909667969,
      "KL/std": 0.9987841844558716,
      "epoch": 0.1602418745275888,
      "fcm_dpo/beta": 0.7533386945724487,
      "fcm_dpo/delta": -0.05080245062708855,
      "fcm_dpo/margin": 0.5950255393981934,
      "fcm_dpo/q_t": 0.4004518985748291,
      "grad_norm": 204.8391571044922,
      "learning_rate": 4.949679871846857e-07,
      "logits/chosen": 0.16151435673236847,
      "logits/rejected": 0.14808031916618347,
      "logps/chosen": -63.80449676513672,
      "logps/ref_chosen": -62.45933151245117,
      "logps/ref_rejected": -67.00595092773438,
      "logps/rejected": -68.9461441040039,
      "loss": 1.1499,
      "margin_dpo/margin_mean": 0.5950250029563904,
      "margin_dpo/margin_std": 1.1285878419876099,
      "step": 106
    },
    {
      "KL/chosen_KL_mean": -1.5086212158203125,
      "KL/mean": -1.7033252716064453,
      "KL/rejected_KL_mean": -1.8980293273925781,
      "KL/std": 1.1155552864074707,
      "epoch": 0.1617535903250189,
      "fcm_dpo/beta": 0.7590333819389343,
      "fcm_dpo/delta": 0.10701896250247955,
      "fcm_dpo/margin": 0.3894113004207611,
      "fcm_dpo/q_t": 0.4394727349281311,
      "grad_norm": 257.46246337890625,
      "learning_rate": 4.947006115536947e-07,
      "logits/chosen": 0.10400848835706711,
      "logits/rejected": 0.0829182118177414,
      "logps/chosen": -77.34658813476562,
      "logps/ref_chosen": -75.83796691894531,
      "logps/ref_rejected": -87.74038696289062,
      "logps/rejected": -89.63841247558594,
      "loss": 1.3187,
      "margin_dpo/margin_mean": 0.3894110918045044,
      "margin_dpo/margin_std": 1.2437622547149658,
      "step": 107
    },
    {
      "KL/chosen_KL_mean": -1.3229484558105469,
      "KL/mean": -1.6645491123199463,
      "KL/rejected_KL_mean": -2.0061492919921875,
      "KL/std": 0.9608972668647766,
      "epoch": 0.16326530612244897,
      "fcm_dpo/beta": 0.7589484453201294,
      "fcm_dpo/delta": -0.1248454749584198,
      "fcm_dpo/margin": 0.6832001805305481,
      "fcm_dpo/q_t": 0.39446821808815,
      "grad_norm": 196.87106323242188,
      "learning_rate": 4.944263911205772e-07,
      "logits/chosen": 0.0801883414387703,
      "logits/rejected": 0.053226158022880554,
      "logps/chosen": -69.7161865234375,
      "logps/ref_chosen": -68.39323425292969,
      "logps/ref_rejected": -83.24267578125,
      "logps/rejected": -85.24882507324219,
      "loss": 1.1301,
      "margin_dpo/margin_mean": 0.6831998229026794,
      "margin_dpo/margin_std": 1.2444127798080444,
      "step": 108
    },
    {
      "KL/chosen_KL_mean": -1.3410778045654297,
      "KL/mean": -1.7414236068725586,
      "KL/rejected_KL_mean": -2.1417694091796875,
      "KL/std": 1.0900359153747559,
      "epoch": 0.16477702191987906,
      "fcm_dpo/beta": 0.729695200920105,
      "fcm_dpo/delta": -0.19558557868003845,
      "fcm_dpo/margin": 0.8006910085678101,
      "fcm_dpo/q_t": 0.38231924176216125,
      "grad_norm": 165.07757568359375,
      "learning_rate": 4.941453335558681e-07,
      "logits/chosen": 0.12404434382915497,
      "logits/rejected": 0.07289332896471024,
      "logps/chosen": -56.868560791015625,
      "logps/ref_chosen": -55.52748107910156,
      "logps/ref_rejected": -83.55218505859375,
      "logps/rejected": -85.69395446777344,
      "loss": 1.0315,
      "margin_dpo/margin_mean": 0.8006904125213623,
      "margin_dpo/margin_std": 1.1890592575073242,
      "step": 109
    },
    {
      "KL/chosen_KL_mean": -1.5522308349609375,
      "KL/mean": -1.7875815629959106,
      "KL/rejected_KL_mean": -2.022930145263672,
      "KL/std": 1.0237528085708618,
      "epoch": 0.16628873771730915,
      "fcm_dpo/beta": 0.7312818169593811,
      "fcm_dpo/delta": 0.05713893473148346,
      "fcm_dpo/margin": 0.4706960916519165,
      "fcm_dpo/q_t": 0.4263302683830261,
      "grad_norm": 223.92262268066406,
      "learning_rate": 4.938574467213517e-07,
      "logits/chosen": 0.06877541542053223,
      "logits/rejected": 0.07652991265058517,
      "logps/chosen": -82.71097564697266,
      "logps/ref_chosen": -81.15874481201172,
      "logps/ref_rejected": -72.56021118164062,
      "logps/rejected": -74.58314514160156,
      "loss": 1.2187,
      "margin_dpo/margin_mean": 0.47069618105888367,
      "margin_dpo/margin_std": 1.1055645942687988,
      "step": 110
    },
    {
      "KL/chosen_KL_mean": -1.7109088897705078,
      "KL/mean": -1.9763455390930176,
      "KL/rejected_KL_mean": -2.2417869567871094,
      "KL/std": 1.0940556526184082,
      "epoch": 0.16780045351473924,
      "fcm_dpo/beta": 0.7348920106887817,
      "fcm_dpo/delta": 0.009718731045722961,
      "fcm_dpo/margin": 0.5308776497840881,
      "fcm_dpo/q_t": 0.41322413086891174,
      "grad_norm": 196.10678100585938,
      "learning_rate": 4.935627386698418e-07,
      "logits/chosen": 0.1737518310546875,
      "logits/rejected": 0.139468714594841,
      "logps/chosen": -54.06989288330078,
      "logps/ref_chosen": -52.358985900878906,
      "logps/ref_rejected": -77.06150817871094,
      "logps/rejected": -79.30329132080078,
      "loss": 1.2238,
      "margin_dpo/margin_mean": 0.5308778285980225,
      "margin_dpo/margin_std": 1.2132712602615356,
      "step": 111
    },
    {
      "KL/chosen_KL_mean": -1.4015159606933594,
      "KL/mean": -1.761267066001892,
      "KL/rejected_KL_mean": -2.1210174560546875,
      "KL/std": 1.0548936128616333,
      "epoch": 0.1693121693121693,
      "fcm_dpo/beta": 0.7160321474075317,
      "fcm_dpo/delta": -0.12210012227296829,
      "fcm_dpo/margin": 0.7195063829421997,
      "fcm_dpo/q_t": 0.3860316574573517,
      "grad_norm": 200.53457641601562,
      "learning_rate": 4.932612176449559e-07,
      "logits/chosen": 0.08839970827102661,
      "logits/rejected": 0.030455876141786575,
      "logps/chosen": -64.42158508300781,
      "logps/ref_chosen": -63.02006530761719,
      "logps/ref_rejected": -111.36941528320312,
      "logps/rejected": -113.49043273925781,
      "loss": 1.0908,
      "margin_dpo/margin_mean": 0.7195055484771729,
      "margin_dpo/margin_std": 1.1482014656066895,
      "step": 112
    },
    {
      "KL/chosen_KL_mean": -1.6727027893066406,
      "KL/mean": -1.985128402709961,
      "KL/rejected_KL_mean": -2.2975540161132812,
      "KL/std": 1.1449403762817383,
      "epoch": 0.1708238851095994,
      "fcm_dpo/beta": 0.7157949805259705,
      "fcm_dpo/delta": -0.05103008449077606,
      "fcm_dpo/margin": 0.6248476505279541,
      "fcm_dpo/q_t": 0.4022253751754761,
      "grad_norm": 205.1897735595703,
      "learning_rate": 4.929528920808854e-07,
      "logits/chosen": 0.10172772407531738,
      "logits/rejected": 0.06672029197216034,
      "logps/chosen": -57.48036575317383,
      "logps/ref_chosen": -55.80766296386719,
      "logps/ref_rejected": -69.84014129638672,
      "logps/rejected": -72.1376953125,
      "loss": 1.1852,
      "margin_dpo/margin_mean": 0.6248472929000854,
      "margin_dpo/margin_std": 1.253474235534668,
      "step": 113
    },
    {
      "KL/chosen_KL_mean": -1.373983383178711,
      "KL/mean": -1.8980087041854858,
      "KL/rejected_KL_mean": -2.422039031982422,
      "KL/std": 1.2073814868927002,
      "epoch": 0.17233560090702948,
      "fcm_dpo/beta": 0.6685370206832886,
      "fcm_dpo/delta": -0.3267349302768707,
      "fcm_dpo/margin": 1.0480527877807617,
      "fcm_dpo/q_t": 0.3540674149990082,
      "grad_norm": 143.26123046875,
      "learning_rate": 4.92637770602159e-07,
      "logits/chosen": 0.16866153478622437,
      "logits/rejected": 0.11124296486377716,
      "logps/chosen": -67.70675659179688,
      "logps/ref_chosen": -66.33277130126953,
      "logps/ref_rejected": -71.61489868164062,
      "logps/rejected": -74.03693389892578,
      "loss": 0.9664,
      "margin_dpo/margin_mean": 1.048052430152893,
      "margin_dpo/margin_std": 1.3153923749923706,
      "step": 114
    },
    {
      "KL/chosen_KL_mean": -1.6200141906738281,
      "KL/mean": -1.9981410503387451,
      "KL/rejected_KL_mean": -2.3762683868408203,
      "KL/std": 1.1273773908615112,
      "epoch": 0.17384731670445955,
      "fcm_dpo/beta": 0.6535633206367493,
      "fcm_dpo/delta": -0.09904934465885162,
      "fcm_dpo/margin": 0.7562557458877563,
      "fcm_dpo/q_t": 0.39832305908203125,
      "grad_norm": 175.138427734375,
      "learning_rate": 4.923158620234019e-07,
      "logits/chosen": 0.14349928498268127,
      "logits/rejected": 0.08649460226297379,
      "logps/chosen": -57.36905288696289,
      "logps/ref_chosen": -55.74903869628906,
      "logps/ref_rejected": -79.59849548339844,
      "logps/rejected": -81.97476196289062,
      "loss": 1.102,
      "margin_dpo/margin_mean": 0.7562559843063354,
      "margin_dpo/margin_std": 1.2784497737884521,
      "step": 115
    },
    {
      "KL/chosen_KL_mean": -1.5154285430908203,
      "KL/mean": -1.9526760578155518,
      "KL/rejected_KL_mean": -2.3899269104003906,
      "KL/std": 1.1451892852783203,
      "epoch": 0.17535903250188964,
      "fcm_dpo/beta": 0.6347832679748535,
      "fcm_dpo/delta": -0.1639837920665741,
      "fcm_dpo/margin": 0.8745034337043762,
      "fcm_dpo/q_t": 0.37725359201431274,
      "grad_norm": 135.62338256835938,
      "learning_rate": 4.91987175349089e-07,
      "logits/chosen": 0.14225362241268158,
      "logits/rejected": 0.07968921959400177,
      "logps/chosen": -50.880592346191406,
      "logps/ref_chosen": -49.36516571044922,
      "logps/ref_rejected": -72.84671020507812,
      "logps/rejected": -75.23663330078125,
      "loss": 1.0137,
      "margin_dpo/margin_mean": 0.8745033740997314,
      "margin_dpo/margin_std": 1.0888726711273193,
      "step": 116
    },
    {
      "KL/chosen_KL_mean": -1.4024600982666016,
      "KL/mean": -1.747429609298706,
      "KL/rejected_KL_mean": -2.0924034118652344,
      "KL/std": 1.0587306022644043,
      "epoch": 0.17687074829931973,
      "fcm_dpo/beta": 0.6196062564849854,
      "fcm_dpo/delta": -0.030679665505886078,
      "fcm_dpo/margin": 0.6899378895759583,
      "fcm_dpo/q_t": 0.4040665626525879,
      "grad_norm": 151.564697265625,
      "learning_rate": 4.916517197732933e-07,
      "logits/chosen": 0.1525934934616089,
      "logits/rejected": 0.1185963824391365,
      "logps/chosen": -59.11335754394531,
      "logps/ref_chosen": -57.710899353027344,
      "logps/ref_rejected": -69.77253723144531,
      "logps/rejected": -71.86494445800781,
      "loss": 1.1369,
      "margin_dpo/margin_mean": 0.6899383068084717,
      "margin_dpo/margin_std": 1.2021629810333252,
      "step": 117
    },
    {
      "KL/chosen_KL_mean": -1.4043560028076172,
      "KL/mean": -1.8314993381500244,
      "KL/rejected_KL_mean": -2.258636474609375,
      "KL/std": 1.092620849609375,
      "epoch": 0.17838246409674982,
      "fcm_dpo/beta": 0.6096054315567017,
      "fcm_dpo/delta": -0.12845522165298462,
      "fcm_dpo/margin": 0.8542841672897339,
      "fcm_dpo/q_t": 0.38544875383377075,
      "grad_norm": 150.0379180908203,
      "learning_rate": 4.913095046794281e-07,
      "logits/chosen": 0.18141870200634003,
      "logits/rejected": 0.1441626399755478,
      "logps/chosen": -53.884254455566406,
      "logps/ref_chosen": -52.479896545410156,
      "logps/ref_rejected": -81.359130859375,
      "logps/rejected": -83.61776733398438,
      "loss": 1.0589,
      "margin_dpo/margin_mean": 0.8542840480804443,
      "margin_dpo/margin_std": 1.218125820159912,
      "step": 118
    },
    {
      "KL/chosen_KL_mean": -1.8466033935546875,
      "KL/mean": -2.1639060974121094,
      "KL/rejected_KL_mean": -2.481212615966797,
      "KL/std": 1.2136081457138062,
      "epoch": 0.17989417989417988,
      "fcm_dpo/beta": 0.6111596822738647,
      "fcm_dpo/delta": 0.012469585984945297,
      "fcm_dpo/margin": 0.6346088647842407,
      "fcm_dpo/q_t": 0.41613152623176575,
      "grad_norm": 163.7132568359375,
      "learning_rate": 4.909605396399855e-07,
      "logits/chosen": 0.1168576031923294,
      "logits/rejected": 0.08203059434890747,
      "logps/chosen": -63.20427703857422,
      "logps/ref_chosen": -61.35767364501953,
      "logps/ref_rejected": -75.71510314941406,
      "logps/rejected": -78.1963119506836,
      "loss": 1.1991,
      "margin_dpo/margin_mean": 0.6346089839935303,
      "margin_dpo/margin_std": 1.380927562713623,
      "step": 119
    },
    {
      "KL/chosen_KL_mean": -1.4735527038574219,
      "KL/mean": -1.9682976007461548,
      "KL/rejected_KL_mean": -2.463043212890625,
      "KL/std": 1.1643463373184204,
      "epoch": 0.18140589569160998,
      "fcm_dpo/beta": 0.5917296409606934,
      "fcm_dpo/delta": -0.19727984070777893,
      "fcm_dpo/margin": 0.9894887208938599,
      "fcm_dpo/q_t": 0.3721131980419159,
      "grad_norm": 135.58323669433594,
      "learning_rate": 4.906048344162676e-07,
      "logits/chosen": 0.1065509170293808,
      "logits/rejected": 0.05253705382347107,
      "logps/chosen": -61.38111877441406,
      "logps/ref_chosen": -59.907569885253906,
      "logps/ref_rejected": -79.6910629272461,
      "logps/rejected": -82.15410614013672,
      "loss": 0.9956,
      "margin_dpo/margin_mean": 0.989488959312439,
      "margin_dpo/margin_std": 1.2124404907226562,
      "step": 120
    },
    {
      "KL/chosen_KL_mean": -1.682455062866211,
      "KL/mean": -2.09917950630188,
      "KL/rejected_KL_mean": -2.5159072875976562,
      "KL/std": 1.2347018718719482,
      "epoch": 0.18291761148904007,
      "fcm_dpo/beta": 0.577370822429657,
      "fcm_dpo/delta": -0.08569201827049255,
      "fcm_dpo/margin": 0.8334502577781677,
      "fcm_dpo/q_t": 0.39638879895210266,
      "grad_norm": 128.85064697265625,
      "learning_rate": 4.902423989581143e-07,
      "logits/chosen": 0.21048860251903534,
      "logits/rejected": 0.133277028799057,
      "logps/chosen": -57.34849548339844,
      "logps/ref_chosen": -55.66604232788086,
      "logps/ref_rejected": -101.56233978271484,
      "logps/rejected": -104.0782470703125,
      "loss": 1.0754,
      "margin_dpo/margin_mean": 0.8334493637084961,
      "margin_dpo/margin_std": 1.2757973670959473,
      "step": 121
    },
    {
      "KL/chosen_KL_mean": -1.7021980285644531,
      "KL/mean": -2.229940414428711,
      "KL/rejected_KL_mean": -2.7576828002929688,
      "KL/std": 1.328824520111084,
      "epoch": 0.18442932728647016,
      "fcm_dpo/beta": 0.5614632368087769,
      "fcm_dpo/delta": -0.20425938069820404,
      "fcm_dpo/margin": 1.0554801225662231,
      "fcm_dpo/q_t": 0.374777227640152,
      "grad_norm": 151.41307067871094,
      "learning_rate": 4.898732434036243e-07,
      "logits/chosen": 0.11900610476732254,
      "logits/rejected": 0.08751888573169708,
      "logps/chosen": -65.03657531738281,
      "logps/ref_chosen": -63.334373474121094,
      "logps/ref_rejected": -73.67523193359375,
      "logps/rejected": -76.43292236328125,
      "loss": 1.0244,
      "margin_dpo/margin_mean": 1.05548095703125,
      "margin_dpo/margin_std": 1.4838553667068481,
      "step": 122
    },
    {
      "KL/chosen_KL_mean": -1.6267662048339844,
      "KL/mean": -2.0346689224243164,
      "KL/rejected_KL_mean": -2.4425735473632812,
      "KL/std": 1.209721326828003,
      "epoch": 0.18594104308390022,
      "fcm_dpo/beta": 0.5542995929718018,
      "fcm_dpo/delta": -0.05618344992399216,
      "fcm_dpo/margin": 0.8158098459243774,
      "fcm_dpo/q_t": 0.39761269092559814,
      "grad_norm": 141.0292205810547,
      "learning_rate": 4.894973780788722e-07,
      "logits/chosen": 0.158452570438385,
      "logits/rejected": 0.11947640776634216,
      "logps/chosen": -58.525508880615234,
      "logps/ref_chosen": -56.89874267578125,
      "logps/ref_rejected": -78.97028350830078,
      "logps/rejected": -81.41285705566406,
      "loss": 1.1263,
      "margin_dpo/margin_mean": 0.8158098459243774,
      "margin_dpo/margin_std": 1.3711764812469482,
      "step": 123
    },
    {
      "KL/chosen_KL_mean": -1.748464584350586,
      "KL/mean": -2.3354647159576416,
      "KL/rejected_KL_mean": -2.922466278076172,
      "KL/std": 1.2677295207977295,
      "epoch": 0.1874527588813303,
      "fcm_dpo/beta": 0.5262091159820557,
      "fcm_dpo/delta": -0.23351916670799255,
      "fcm_dpo/margin": 1.1740025281906128,
      "fcm_dpo/q_t": 0.3634141683578491,
      "grad_norm": 114.40848541259766,
      "learning_rate": 4.89114813497619e-07,
      "logits/chosen": 0.16980503499507904,
      "logits/rejected": 0.11593275517225266,
      "logps/chosen": -58.86454772949219,
      "logps/ref_chosen": -57.116085052490234,
      "logps/ref_rejected": -87.93074035644531,
      "logps/rejected": -90.85321044921875,
      "loss": 0.9742,
      "margin_dpo/margin_mean": 1.1740025281906128,
      "margin_dpo/margin_std": 1.339629888534546,
      "step": 124
    },
    {
      "KL/chosen_KL_mean": -1.9098529815673828,
      "KL/mean": -2.3675875663757324,
      "KL/rejected_KL_mean": -2.8253173828125,
      "KL/std": 1.4183313846588135,
      "epoch": 0.1889644746787604,
      "fcm_dpo/beta": 0.5146192312240601,
      "fcm_dpo/delta": -0.07511819899082184,
      "fcm_dpo/margin": 0.9154636859893799,
      "fcm_dpo/q_t": 0.39758336544036865,
      "grad_norm": 131.99765014648438,
      "learning_rate": 4.887255603610184e-07,
      "logits/chosen": 0.19287623465061188,
      "logits/rejected": 0.13918162882328033,
      "logps/chosen": -67.61602783203125,
      "logps/ref_chosen": -65.7061767578125,
      "logps/ref_rejected": -91.72711944580078,
      "logps/rejected": -94.55243682861328,
      "loss": 1.0888,
      "margin_dpo/margin_mean": 0.9154644012451172,
      "margin_dpo/margin_std": 1.4167892932891846,
      "step": 125
    },
    {
      "KL/chosen_KL_mean": -1.6904373168945312,
      "KL/mean": -2.1172661781311035,
      "KL/rejected_KL_mean": -2.544099807739258,
      "KL/std": 1.5803725719451904,
      "epoch": 0.19047619047619047,
      "fcm_dpo/beta": 0.5104827880859375,
      "fcm_dpo/delta": -0.037569280713796616,
      "fcm_dpo/margin": 0.8536651134490967,
      "fcm_dpo/q_t": 0.4115992784500122,
      "grad_norm": 120.52528381347656,
      "learning_rate": 4.883296295573176e-07,
      "logits/chosen": 0.007485136389732361,
      "logits/rejected": 0.001452924683690071,
      "logps/chosen": -69.86652374267578,
      "logps/ref_chosen": -68.17608642578125,
      "logps/ref_rejected": -65.1175537109375,
      "logps/rejected": -67.66165161132812,
      "loss": 1.1464,
      "margin_dpo/margin_mean": 0.8536652326583862,
      "margin_dpo/margin_std": 1.6970319747924805,
      "step": 126
    },
    {
      "KL/chosen_KL_mean": -1.8688240051269531,
      "KL/mean": -2.4150619506835938,
      "KL/rejected_KL_mean": -2.961301803588867,
      "KL/std": 1.2444980144500732,
      "epoch": 0.19198790627362056,
      "fcm_dpo/beta": 0.4957229495048523,
      "fcm_dpo/delta": -0.15168866515159607,
      "fcm_dpo/margin": 1.0924748182296753,
      "fcm_dpo/q_t": 0.378802627325058,
      "grad_norm": 115.55281829833984,
      "learning_rate": 4.87927032161552e-07,
      "logits/chosen": 0.1286633312702179,
      "logits/rejected": 0.0994386374950409,
      "logps/chosen": -63.74905776977539,
      "logps/ref_chosen": -61.88023376464844,
      "logps/ref_rejected": -68.46012878417969,
      "logps/rejected": -71.42142486572266,
      "loss": 1.0108,
      "margin_dpo/margin_mean": 1.0924749374389648,
      "margin_dpo/margin_std": 1.278612494468689,
      "step": 127
    },
    {
      "KL/chosen_KL_mean": -2.003080368041992,
      "KL/mean": -2.453127384185791,
      "KL/rejected_KL_mean": -2.903179168701172,
      "KL/std": 1.4483566284179688,
      "epoch": 0.19349962207105065,
      "fcm_dpo/beta": 0.49017786979675293,
      "fcm_dpo/delta": -0.04357750713825226,
      "fcm_dpo/margin": 0.900090754032135,
      "fcm_dpo/q_t": 0.40476569533348083,
      "grad_norm": 127.00218963623047,
      "learning_rate": 4.875177794352363e-07,
      "logits/chosen": 0.16356945037841797,
      "logits/rejected": 0.11114731431007385,
      "logps/chosen": -68.71206665039062,
      "logps/ref_chosen": -66.708984375,
      "logps/ref_rejected": -94.97969055175781,
      "logps/rejected": -97.88287353515625,
      "loss": 1.1399,
      "margin_dpo/margin_mean": 0.9000910520553589,
      "margin_dpo/margin_std": 1.6488995552062988,
      "step": 128
    },
    {
      "KL/chosen_KL_mean": -2.2038707733154297,
      "KL/mean": -2.5676302909851074,
      "KL/rejected_KL_mean": -2.931396484375,
      "KL/std": 1.304401159286499,
      "epoch": 0.19501133786848074,
      "fcm_dpo/beta": 0.49330243468284607,
      "fcm_dpo/delta": 0.04265614598989487,
      "fcm_dpo/margin": 0.7275242805480957,
      "fcm_dpo/q_t": 0.4221858084201813,
      "grad_norm": 138.87521362304688,
      "learning_rate": 4.871018828260491e-07,
      "logits/chosen": 0.13030391931533813,
      "logits/rejected": 0.12313122302293777,
      "logps/chosen": -67.54269409179688,
      "logps/ref_chosen": -65.33882904052734,
      "logps/ref_rejected": -68.06109619140625,
      "logps/rejected": -70.99249267578125,
      "loss": 1.177,
      "margin_dpo/margin_mean": 0.7275242805480957,
      "margin_dpo/margin_std": 1.4568142890930176,
      "step": 129
    },
    {
      "KL/chosen_KL_mean": -2.025636672973633,
      "KL/mean": -2.517505645751953,
      "KL/rejected_KL_mean": -3.0093765258789062,
      "KL/std": 1.313607931137085,
      "epoch": 0.1965230536659108,
      "fcm_dpo/beta": 0.48430347442626953,
      "fcm_dpo/delta": -0.08306446671485901,
      "fcm_dpo/margin": 0.9837359189987183,
      "fcm_dpo/q_t": 0.3932478129863739,
      "grad_norm": 130.4871063232422,
      "learning_rate": 4.866793539675126e-07,
      "logits/chosen": 0.08757339417934418,
      "logits/rejected": 0.04383823275566101,
      "logps/chosen": -60.68638229370117,
      "logps/ref_chosen": -58.660743713378906,
      "logps/ref_rejected": -79.24510192871094,
      "logps/rejected": -82.25447845458984,
      "loss": 1.0553,
      "margin_dpo/margin_mean": 0.9837361574172974,
      "margin_dpo/margin_std": 1.2436227798461914,
      "step": 130
    },
    {
      "KL/chosen_KL_mean": -1.933258056640625,
      "KL/mean": -2.4867091178894043,
      "KL/rejected_KL_mean": -3.0401611328125,
      "KL/std": 1.5524529218673706,
      "epoch": 0.1980347694633409,
      "fcm_dpo/beta": 0.4773693382740021,
      "fcm_dpo/delta": -0.13555176556110382,
      "fcm_dpo/margin": 1.1069055795669556,
      "fcm_dpo/q_t": 0.3849940001964569,
      "grad_norm": 114.3026123046875,
      "learning_rate": 4.86250204678667e-07,
      "logits/chosen": 0.1033594086766243,
      "logits/rejected": 0.04820697754621506,
      "logps/chosen": -54.447792053222656,
      "logps/ref_chosen": -52.51453399658203,
      "logps/ref_rejected": -85.18299865722656,
      "logps/rejected": -88.22315979003906,
      "loss": 1.0762,
      "margin_dpo/margin_mean": 1.1069060564041138,
      "margin_dpo/margin_std": 1.702505111694336,
      "step": 131
    },
    {
      "KL/chosen_KL_mean": -2.1760196685791016,
      "KL/mean": -2.6583878993988037,
      "KL/rejected_KL_mean": -3.140756607055664,
      "KL/std": 1.49051034450531,
      "epoch": 0.19954648526077098,
      "fcm_dpo/beta": 0.4701109528541565,
      "fcm_dpo/delta": -0.05607073754072189,
      "fcm_dpo/margin": 0.9647371768951416,
      "fcm_dpo/q_t": 0.3987612724304199,
      "grad_norm": 122.06319427490234,
      "learning_rate": 4.858144469637408e-07,
      "logits/chosen": 0.21260422468185425,
      "logits/rejected": 0.18176573514938354,
      "logps/chosen": -67.86115264892578,
      "logps/ref_chosen": -65.68513488769531,
      "logps/ref_rejected": -69.54120635986328,
      "logps/rejected": -72.68196105957031,
      "loss": 1.1113,
      "margin_dpo/margin_mean": 0.9647375345230103,
      "margin_dpo/margin_std": 1.6063919067382812,
      "step": 132
    },
    {
      "KL/chosen_KL_mean": -2.1477317810058594,
      "KL/mean": -2.5903117656707764,
      "KL/rejected_KL_mean": -3.032886505126953,
      "KL/std": 1.4966492652893066,
      "epoch": 0.20105820105820105,
      "fcm_dpo/beta": 0.4674025774002075,
      "fcm_dpo/delta": -0.014334164559841156,
      "fcm_dpo/margin": 0.8851548433303833,
      "fcm_dpo/q_t": 0.4093969762325287,
      "grad_norm": 125.8607177734375,
      "learning_rate": 4.853720930118138e-07,
      "logits/chosen": 0.12384993582963943,
      "logits/rejected": 0.11437453329563141,
      "logps/chosen": -65.745849609375,
      "logps/ref_chosen": -63.598114013671875,
      "logps/ref_rejected": -73.72798156738281,
      "logps/rejected": -76.7608642578125,
      "loss": 1.1389,
      "margin_dpo/margin_mean": 0.8851544857025146,
      "margin_dpo/margin_std": 1.5839178562164307,
      "step": 133
    },
    {
      "KL/chosen_KL_mean": -2.069742202758789,
      "KL/mean": -2.7214608192443848,
      "KL/rejected_KL_mean": -3.373180389404297,
      "KL/std": 1.6641645431518555,
      "epoch": 0.20256991685563114,
      "fcm_dpo/beta": 0.45107993483543396,
      "fcm_dpo/delta": -0.20145797729492188,
      "fcm_dpo/margin": 1.3034403324127197,
      "fcm_dpo/q_t": 0.3716809153556824,
      "grad_norm": 105.60230255126953,
      "learning_rate": 4.849231551964771e-07,
      "logits/chosen": 0.21557536721229553,
      "logits/rejected": 0.16419634222984314,
      "logps/chosen": -55.86431121826172,
      "logps/ref_chosen": -53.79457092285156,
      "logps/ref_rejected": -74.16741943359375,
      "logps/rejected": -77.54060363769531,
      "loss": 1.0145,
      "margin_dpo/margin_mean": 1.303440809249878,
      "margin_dpo/margin_std": 1.7132940292358398,
      "step": 134
    },
    {
      "KL/chosen_KL_mean": -2.093151092529297,
      "KL/mean": -2.524364471435547,
      "KL/rejected_KL_mean": -2.9555816650390625,
      "KL/std": 1.5630788803100586,
      "epoch": 0.20408163265306123,
      "fcm_dpo/beta": 0.44986188411712646,
      "fcm_dpo/delta": 0.012506252154707909,
      "fcm_dpo/margin": 0.8624306321144104,
      "fcm_dpo/q_t": 0.4156408905982971,
      "grad_norm": 102.62089538574219,
      "learning_rate": 4.844676460754862e-07,
      "logits/chosen": 0.14737987518310547,
      "logits/rejected": 0.11628959327936172,
      "logps/chosen": -51.53422546386719,
      "logps/ref_chosen": -49.441078186035156,
      "logps/ref_rejected": -65.96878051757812,
      "logps/rejected": -68.92436218261719,
      "loss": 1.1536,
      "margin_dpo/margin_mean": 0.8624303340911865,
      "margin_dpo/margin_std": 1.610809564590454,
      "step": 135
    },
    {
      "KL/chosen_KL_mean": -2.437408447265625,
      "KL/mean": -2.967642307281494,
      "KL/rejected_KL_mean": -3.4978790283203125,
      "KL/std": 1.7607862949371338,
      "epoch": 0.20559334845049132,
      "fcm_dpo/beta": 0.4447017312049866,
      "fcm_dpo/delta": -0.07520163059234619,
      "fcm_dpo/margin": 1.0604674816131592,
      "fcm_dpo/q_t": 0.4090201258659363,
      "grad_norm": 136.33038330078125,
      "learning_rate": 4.840055783904106e-07,
      "logits/chosen": 0.1400383710861206,
      "logits/rejected": 0.0758211761713028,
      "logps/chosen": -69.19667053222656,
      "logps/ref_chosen": -66.75926208496094,
      "logps/ref_rejected": -94.61787414550781,
      "logps/rejected": -98.11575317382812,
      "loss": 1.1821,
      "margin_dpo/margin_mean": 1.0604681968688965,
      "margin_dpo/margin_std": 2.285677909851074,
      "step": 136
    },
    {
      "KL/chosen_KL_mean": -2.2930164337158203,
      "KL/mean": -2.846302032470703,
      "KL/rejected_KL_mean": -3.3995895385742188,
      "KL/std": 1.5753281116485596,
      "epoch": 0.20710506424792138,
      "fcm_dpo/beta": 0.4380854368209839,
      "fcm_dpo/delta": -0.08900899440050125,
      "fcm_dpo/margin": 1.1065716743469238,
      "fcm_dpo/q_t": 0.39505431056022644,
      "grad_norm": 101.92052459716797,
      "learning_rate": 4.835369650662767e-07,
      "logits/chosen": 0.17767782509326935,
      "logits/rejected": 0.15075279772281647,
      "logps/chosen": -59.076812744140625,
      "logps/ref_chosen": -56.78379821777344,
      "logps/ref_rejected": -69.89952087402344,
      "logps/rejected": -73.29911041259766,
      "loss": 1.098,
      "margin_dpo/margin_mean": 1.106572151184082,
      "margin_dpo/margin_std": 1.8102577924728394,
      "step": 137
    },
    {
      "KL/chosen_KL_mean": -2.701139450073242,
      "KL/mean": -3.1472229957580566,
      "KL/rejected_KL_mean": -3.5933074951171875,
      "KL/std": 1.609253168106079,
      "epoch": 0.20861678004535147,
      "fcm_dpo/beta": 0.4359511137008667,
      "fcm_dpo/delta": 0.01150511484593153,
      "fcm_dpo/margin": 0.8921665549278259,
      "fcm_dpo/q_t": 0.4128834009170532,
      "grad_norm": 108.2798080444336,
      "learning_rate": 4.830618192112065e-07,
      "logits/chosen": 0.16457317769527435,
      "logits/rejected": 0.13061930239200592,
      "logps/chosen": -61.4671516418457,
      "logps/ref_chosen": -58.766014099121094,
      "logps/ref_rejected": -68.12371826171875,
      "logps/rejected": -71.71702575683594,
      "loss": 1.1536,
      "margin_dpo/margin_mean": 0.8921666145324707,
      "margin_dpo/margin_std": 1.644634485244751,
      "step": 138
    },
    {
      "KL/chosen_KL_mean": -2.6400203704833984,
      "KL/mean": -3.2433128356933594,
      "KL/rejected_KL_mean": -3.846607208251953,
      "KL/std": 1.5791709423065186,
      "epoch": 0.21012849584278157,
      "fcm_dpo/beta": 0.4313918948173523,
      "fcm_dpo/delta": -0.1272524744272232,
      "fcm_dpo/margin": 1.2065834999084473,
      "fcm_dpo/q_t": 0.38509002327919006,
      "grad_norm": 110.69141387939453,
      "learning_rate": 4.825801541160509e-07,
      "logits/chosen": 0.1212557703256607,
      "logits/rejected": 0.09481331706047058,
      "logps/chosen": -73.86560821533203,
      "logps/ref_chosen": -71.2255859375,
      "logps/ref_rejected": -82.1834716796875,
      "logps/rejected": -86.03007507324219,
      "loss": 1.0402,
      "margin_dpo/margin_mean": 1.2065834999084473,
      "margin_dpo/margin_std": 1.5707457065582275,
      "step": 139
    },
    {
      "KL/chosen_KL_mean": -2.432710647583008,
      "KL/mean": -3.1976771354675293,
      "KL/rejected_KL_mean": -3.9626426696777344,
      "KL/std": 2.112020969390869,
      "epoch": 0.21164021164021163,
      "fcm_dpo/beta": 0.41016870737075806,
      "fcm_dpo/delta": -0.2429695725440979,
      "fcm_dpo/margin": 1.5299389362335205,
      "fcm_dpo/q_t": 0.3645162582397461,
      "grad_norm": 113.63577270507812,
      "learning_rate": 4.820919832540181e-07,
      "logits/chosen": 0.1273362636566162,
      "logits/rejected": 0.08577500283718109,
      "logps/chosen": -65.71037292480469,
      "logps/ref_chosen": -63.27766418457031,
      "logps/ref_rejected": -83.30647277832031,
      "logps/rejected": -87.26911926269531,
      "loss": 1.0491,
      "margin_dpo/margin_mean": 1.5299385786056519,
      "margin_dpo/margin_std": 2.3173608779907227,
      "step": 140
    },
    {
      "KL/chosen_KL_mean": -2.585247039794922,
      "KL/mean": -3.257610321044922,
      "KL/rejected_KL_mean": -3.929973602294922,
      "KL/std": 1.727674961090088,
      "epoch": 0.21315192743764172,
      "fcm_dpo/beta": 0.3947093188762665,
      "fcm_dpo/delta": -0.1396196335554123,
      "fcm_dpo/margin": 1.344720482826233,
      "fcm_dpo/q_t": 0.3882700204849243,
      "grad_norm": 102.85530853271484,
      "learning_rate": 4.815973202802966e-07,
      "logits/chosen": 0.15521973371505737,
      "logits/rejected": 0.11595935374498367,
      "logps/chosen": -64.35202026367188,
      "logps/ref_chosen": -61.76676940917969,
      "logps/ref_rejected": -88.60601806640625,
      "logps/rejected": -92.53599548339844,
      "loss": 1.0768,
      "margin_dpo/margin_mean": 1.3447209596633911,
      "margin_dpo/margin_std": 2.0976152420043945,
      "step": 141
    },
    {
      "KL/chosen_KL_mean": -2.634784698486328,
      "KL/mean": -3.140658378601074,
      "KL/rejected_KL_mean": -3.6465377807617188,
      "KL/std": 1.6917712688446045,
      "epoch": 0.2146636432350718,
      "fcm_dpo/beta": 0.39509785175323486,
      "fcm_dpo/delta": 0.0001626778393983841,
      "fcm_dpo/margin": 1.0117509365081787,
      "fcm_dpo/q_t": 0.4111095666885376,
      "grad_norm": 99.67555236816406,
      "learning_rate": 4.810961790316729e-07,
      "logits/chosen": 0.15274107456207275,
      "logits/rejected": 0.12943851947784424,
      "logps/chosen": -67.90956115722656,
      "logps/ref_chosen": -65.2747802734375,
      "logps/ref_rejected": -81.1378173828125,
      "logps/rejected": -84.78435516357422,
      "loss": 1.1272,
      "margin_dpo/margin_mean": 1.01175057888031,
      "margin_dpo/margin_std": 1.6819437742233276,
      "step": 142
    },
    {
      "KL/chosen_KL_mean": -2.6737728118896484,
      "KL/mean": -3.2099857330322266,
      "KL/rejected_KL_mean": -3.7461929321289062,
      "KL/std": 1.7688965797424316,
      "epoch": 0.2161753590325019,
      "fcm_dpo/beta": 0.391811341047287,
      "fcm_dpo/delta": -0.02130448818206787,
      "fcm_dpo/margin": 1.0724186897277832,
      "fcm_dpo/q_t": 0.40727514028549194,
      "grad_norm": 119.30868530273438,
      "learning_rate": 4.805885735261454e-07,
      "logits/chosen": 0.18711808323860168,
      "logits/rejected": 0.17115537822246552,
      "logps/chosen": -65.29159545898438,
      "logps/ref_chosen": -62.617828369140625,
      "logps/ref_rejected": -70.39239501953125,
      "logps/rejected": -74.13858032226562,
      "loss": 1.1568,
      "margin_dpo/margin_mean": 1.0724186897277832,
      "margin_dpo/margin_std": 2.059246063232422,
      "step": 143
    },
    {
      "KL/chosen_KL_mean": -2.8863258361816406,
      "KL/mean": -3.4791462421417236,
      "KL/rejected_KL_mean": -4.071956634521484,
      "KL/std": 1.829078197479248,
      "epoch": 0.21768707482993196,
      "fcm_dpo/beta": 0.3889148235321045,
      "fcm_dpo/delta": -0.06404763460159302,
      "fcm_dpo/margin": 1.1856343746185303,
      "fcm_dpo/q_t": 0.4013003408908844,
      "grad_norm": 105.5179443359375,
      "learning_rate": 4.800745179625307e-07,
      "logits/chosen": 0.1439233273267746,
      "logits/rejected": 0.11750271171331406,
      "logps/chosen": -63.68901443481445,
      "logps/ref_chosen": -60.80268859863281,
      "logps/ref_rejected": -79.07284545898438,
      "logps/rejected": -83.1447982788086,
      "loss": 1.1181,
      "margin_dpo/margin_mean": 1.1856343746185303,
      "margin_dpo/margin_std": 2.050663471221924,
      "step": 144
    },
    {
      "KL/chosen_KL_mean": -2.7375221252441406,
      "KL/mean": -3.34666109085083,
      "KL/rejected_KL_mean": -3.9557952880859375,
      "KL/std": 1.9480266571044922,
      "epoch": 0.21919879062736206,
      "fcm_dpo/beta": 0.3836725950241089,
      "fcm_dpo/delta": -0.07069326192140579,
      "fcm_dpo/margin": 1.2182751893997192,
      "fcm_dpo/q_t": 0.4001026749610901,
      "grad_norm": 115.86017608642578,
      "learning_rate": 4.795540267200686e-07,
      "logits/chosen": 0.11822667717933655,
      "logits/rejected": 0.13500581681728363,
      "logps/chosen": -77.34898376464844,
      "logps/ref_chosen": -74.61146545410156,
      "logps/ref_rejected": -83.24461364746094,
      "logps/rejected": -87.20040893554688,
      "loss": 1.1347,
      "margin_dpo/margin_mean": 1.2182748317718506,
      "margin_dpo/margin_std": 2.2440905570983887,
      "step": 145
    },
    {
      "KL/chosen_KL_mean": -2.5694961547851562,
      "KL/mean": -3.215273857116699,
      "KL/rejected_KL_mean": -3.8610496520996094,
      "KL/std": 1.8966618776321411,
      "epoch": 0.22071050642479215,
      "fcm_dpo/beta": 0.3777027726173401,
      "fcm_dpo/delta": -0.09222418069839478,
      "fcm_dpo/margin": 1.2915546894073486,
      "fcm_dpo/q_t": 0.39369216561317444,
      "grad_norm": 95.42322540283203,
      "learning_rate": 4.790271143580173e-07,
      "logits/chosen": 0.12121773511171341,
      "logits/rejected": 0.1055947095155716,
      "logps/chosen": -60.41048049926758,
      "logps/ref_chosen": -57.84098434448242,
      "logps/ref_rejected": -67.47422790527344,
      "logps/rejected": -71.33528137207031,
      "loss": 1.0812,
      "margin_dpo/margin_mean": 1.2915539741516113,
      "margin_dpo/margin_std": 1.969018578529358,
      "step": 146
    },
    {
      "KL/chosen_KL_mean": -3.021350860595703,
      "KL/mean": -3.513603687286377,
      "KL/rejected_KL_mean": -4.005867004394531,
      "KL/std": 1.8846187591552734,
      "epoch": 0.2222222222222222,
      "fcm_dpo/beta": 0.3774845004081726,
      "fcm_dpo/delta": 0.029318859800696373,
      "fcm_dpo/margin": 0.9845160245895386,
      "fcm_dpo/q_t": 0.417360782623291,
      "grad_norm": 120.66484069824219,
      "learning_rate": 4.784937956152489e-07,
      "logits/chosen": 0.12826010584831238,
      "logits/rejected": 0.08931194245815277,
      "logps/chosen": -69.83482360839844,
      "logps/ref_chosen": -66.81346893310547,
      "logps/ref_rejected": -81.1796875,
      "logps/rejected": -85.18556213378906,
      "loss": 1.1863,
      "margin_dpo/margin_mean": 0.9845158457756042,
      "margin_dpo/margin_std": 2.072244644165039,
      "step": 147
    },
    {
      "KL/chosen_KL_mean": -2.9220848083496094,
      "KL/mean": -3.633432388305664,
      "KL/rejected_KL_mean": -4.344776153564453,
      "KL/std": 1.9022397994995117,
      "epoch": 0.2237339380196523,
      "fcm_dpo/beta": 0.3716948628425598,
      "fcm_dpo/delta": -0.135844424366951,
      "fcm_dpo/margin": 1.4226953983306885,
      "fcm_dpo/q_t": 0.38612425327301025,
      "grad_norm": 80.1585693359375,
      "learning_rate": 4.779540854098347e-07,
      "logits/chosen": 0.2026696801185608,
      "logits/rejected": 0.13863450288772583,
      "logps/chosen": -51.609840393066406,
      "logps/ref_chosen": -48.6877555847168,
      "logps/ref_rejected": -67.50503540039062,
      "logps/rejected": -71.84980773925781,
      "loss": 1.0722,
      "margin_dpo/margin_mean": 1.4226953983306885,
      "margin_dpo/margin_std": 2.160921096801758,
      "step": 148
    },
    {
      "KL/chosen_KL_mean": -3.5425682067871094,
      "KL/mean": -4.333852767944336,
      "KL/rejected_KL_mean": -5.1251373291015625,
      "KL/std": 2.3124797344207764,
      "epoch": 0.2252456538170824,
      "fcm_dpo/beta": 0.3573852777481079,
      "fcm_dpo/delta": -0.1757103055715561,
      "fcm_dpo/margin": 1.5825666189193726,
      "fcm_dpo/q_t": 0.3796079754829407,
      "grad_norm": 85.42340850830078,
      "learning_rate": 4.774079988386296e-07,
      "logits/chosen": 0.11232784390449524,
      "logits/rejected": 0.06763314455747604,
      "logps/chosen": -58.68634796142578,
      "logps/ref_chosen": -55.143775939941406,
      "logps/ref_rejected": -64.79888916015625,
      "logps/rejected": -69.92402648925781,
      "loss": 1.0345,
      "margin_dpo/margin_mean": 1.582566261291504,
      "margin_dpo/margin_std": 2.227118968963623,
      "step": 149
    },
    {
      "KL/chosen_KL_mean": -2.5940914154052734,
      "KL/mean": -3.5496535301208496,
      "KL/rejected_KL_mean": -4.505218505859375,
      "KL/std": 2.0950212478637695,
      "epoch": 0.22675736961451248,
      "fcm_dpo/beta": 0.3405567407608032,
      "fcm_dpo/delta": -0.26843270659446716,
      "fcm_dpo/margin": 1.9111292362213135,
      "fcm_dpo/q_t": 0.3549882173538208,
      "grad_norm": 78.97344207763672,
      "learning_rate": 4.768555511768486e-07,
      "logits/chosen": 0.14000877737998962,
      "logits/rejected": 0.10108865052461624,
      "logps/chosen": -70.06483459472656,
      "logps/ref_chosen": -67.47074890136719,
      "logps/ref_rejected": -89.21170806884766,
      "logps/rejected": -93.71692657470703,
      "loss": 0.9473,
      "margin_dpo/margin_mean": 1.911129355430603,
      "margin_dpo/margin_std": 2.0432076454162598,
      "step": 150
    },
    {
      "KL/chosen_KL_mean": -2.626737594604492,
      "KL/mean": -3.621738910675049,
      "KL/rejected_KL_mean": -4.616741180419922,
      "KL/std": 2.104227066040039,
      "epoch": 0.22826908541194255,
      "fcm_dpo/beta": 0.3234705328941345,
      "fcm_dpo/delta": -0.26033222675323486,
      "fcm_dpo/margin": 1.9900016784667969,
      "fcm_dpo/q_t": 0.357607901096344,
      "grad_norm": 69.46760559082031,
      "learning_rate": 4.762967578776406e-07,
      "logits/chosen": 0.15124168992042542,
      "logits/rejected": 0.10280808806419373,
      "logps/chosen": -55.086280822753906,
      "logps/ref_chosen": -52.45954132080078,
      "logps/ref_rejected": -79.0630111694336,
      "logps/rejected": -83.67974853515625,
      "loss": 0.9495,
      "margin_dpo/margin_mean": 1.9900016784667969,
      "margin_dpo/margin_std": 2.164335012435913,
      "step": 151
    },
    {
      "KL/chosen_KL_mean": -3.3076038360595703,
      "KL/mean": -4.1723432540893555,
      "KL/rejected_KL_mean": -5.037086486816406,
      "KL/std": 2.2166404724121094,
      "epoch": 0.22978080120937264,
      "fcm_dpo/beta": 0.31334036588668823,
      "fcm_dpo/delta": -0.14978964626789093,
      "fcm_dpo/margin": 1.729482650756836,
      "fcm_dpo/q_t": 0.3832073211669922,
      "grad_norm": 76.83300018310547,
      "learning_rate": 4.757316345716553e-07,
      "logits/chosen": 0.22163252532482147,
      "logits/rejected": 0.1748337298631668,
      "logps/chosen": -59.86143493652344,
      "logps/ref_chosen": -56.5538330078125,
      "logps/ref_rejected": -76.55074310302734,
      "logps/rejected": -81.58782958984375,
      "loss": 1.0511,
      "margin_dpo/margin_mean": 1.729482650756836,
      "margin_dpo/margin_std": 2.4772982597351074,
      "step": 152
    },
    {
      "KL/chosen_KL_mean": -3.1301536560058594,
      "KL/mean": -3.992166042327881,
      "KL/rejected_KL_mean": -4.854175567626953,
      "KL/std": 2.3730661869049072,
      "epoch": 0.23129251700680273,
      "fcm_dpo/beta": 0.30294376611709595,
      "fcm_dpo/delta": -0.12941709160804749,
      "fcm_dpo/margin": 1.7240203619003296,
      "fcm_dpo/q_t": 0.3813447952270508,
      "grad_norm": 70.95280456542969,
      "learning_rate": 4.751601970666064e-07,
      "logits/chosen": 0.12408562004566193,
      "logits/rejected": 0.08880177140235901,
      "logps/chosen": -71.13704681396484,
      "logps/ref_chosen": -68.00689697265625,
      "logps/ref_rejected": -74.83482360839844,
      "logps/rejected": -79.68899536132812,
      "loss": 1.0188,
      "margin_dpo/margin_mean": 1.7240206003189087,
      "margin_dpo/margin_std": 2.048651933670044,
      "step": 153
    },
    {
      "KL/chosen_KL_mean": -3.7474632263183594,
      "KL/mean": -4.369100570678711,
      "KL/rejected_KL_mean": -4.990741729736328,
      "KL/std": 2.367417812347412,
      "epoch": 0.2328042328042328,
      "fcm_dpo/beta": 0.30299603939056396,
      "fcm_dpo/delta": 0.0241601150482893,
      "fcm_dpo/margin": 1.2432812452316284,
      "fcm_dpo/q_t": 0.4153980016708374,
      "grad_norm": 75.19510650634766,
      "learning_rate": 4.745824613468292e-07,
      "logits/chosen": 0.20469820499420166,
      "logits/rejected": 0.20125876367092133,
      "logps/chosen": -62.970001220703125,
      "logps/ref_chosen": -59.222537994384766,
      "logps/ref_rejected": -64.19131469726562,
      "logps/rejected": -69.18206024169922,
      "loss": 1.181,
      "margin_dpo/margin_mean": 1.243281364440918,
      "margin_dpo/margin_std": 2.5532095432281494,
      "step": 154
    },
    {
      "KL/chosen_KL_mean": -3.6322250366210938,
      "KL/mean": -4.443280220031738,
      "KL/rejected_KL_mean": -5.25433349609375,
      "KL/std": 2.3358683586120605,
      "epoch": 0.23431594860166288,
      "fcm_dpo/beta": 0.29752117395401,
      "fcm_dpo/delta": -0.08788089454174042,
      "fcm_dpo/margin": 1.6221065521240234,
      "fcm_dpo/q_t": 0.39678919315338135,
      "grad_norm": 78.17323303222656,
      "learning_rate": 4.7399844357283393e-07,
      "logits/chosen": 0.18790429830551147,
      "logits/rejected": 0.16967862844467163,
      "logps/chosen": -72.08692169189453,
      "logps/ref_chosen": -68.45469665527344,
      "logps/ref_rejected": -77.91763305664062,
      "logps/rejected": -83.17196655273438,
      "loss": 1.1194,
      "margin_dpo/margin_mean": 1.6221063137054443,
      "margin_dpo/margin_std": 2.804651975631714,
      "step": 155
    },
    {
      "KL/chosen_KL_mean": -3.6310901641845703,
      "KL/mean": -4.6390886306762695,
      "KL/rejected_KL_mean": -5.647087097167969,
      "KL/std": 2.6043336391448975,
      "epoch": 0.23582766439909297,
      "fcm_dpo/beta": 0.2896882891654968,
      "fcm_dpo/delta": -0.19520190358161926,
      "fcm_dpo/margin": 2.0159971714019775,
      "fcm_dpo/q_t": 0.37270045280456543,
      "grad_norm": 76.12146759033203,
      "learning_rate": 4.7340816008085305e-07,
      "logits/chosen": 0.14739282429218292,
      "logits/rejected": 0.10633807629346848,
      "logps/chosen": -70.90068817138672,
      "logps/ref_chosen": -67.26959991455078,
      "logps/ref_rejected": -86.95914459228516,
      "logps/rejected": -92.60623168945312,
      "loss": 0.9996,
      "margin_dpo/margin_mean": 2.0159971714019775,
      "margin_dpo/margin_std": 2.4939284324645996,
      "step": 156
    },
    {
      "KL/chosen_KL_mean": -3.4813175201416016,
      "KL/mean": -4.297900676727295,
      "KL/rejected_KL_mean": -5.114484786987305,
      "KL/std": 2.685899257659912,
      "epoch": 0.23733938019652306,
      "fcm_dpo/beta": 0.2811231017112732,
      "fcm_dpo/delta": -0.06421680748462677,
      "fcm_dpo/margin": 1.6331628561019897,
      "fcm_dpo/q_t": 0.39780938625335693,
      "grad_norm": 66.9979019165039,
      "learning_rate": 4.728116273823847e-07,
      "logits/chosen": 0.1526588797569275,
      "logits/rejected": 0.13304036855697632,
      "logps/chosen": -58.25419235229492,
      "logps/ref_chosen": -54.77287292480469,
      "logps/ref_rejected": -63.87866973876953,
      "logps/rejected": -68.99315643310547,
      "loss": 1.0859,
      "margin_dpo/margin_mean": 1.6331627368927002,
      "margin_dpo/margin_std": 2.385460376739502,
      "step": 157
    },
    {
      "KL/chosen_KL_mean": -3.7610626220703125,
      "KL/mean": -4.613132953643799,
      "KL/rejected_KL_mean": -5.465198516845703,
      "KL/std": 2.5535340309143066,
      "epoch": 0.23885109599395313,
      "fcm_dpo/beta": 0.2791249454021454,
      "fcm_dpo/delta": -0.07955023646354675,
      "fcm_dpo/margin": 1.7041363716125488,
      "fcm_dpo/q_t": 0.394135445356369,
      "grad_norm": 71.29205322265625,
      "learning_rate": 4.7220886216373085e-07,
      "logits/chosen": 0.17621684074401855,
      "logits/rejected": 0.14411872625350952,
      "logps/chosen": -68.68377685546875,
      "logps/ref_chosen": -64.92271423339844,
      "logps/ref_rejected": -82.23789978027344,
      "logps/rejected": -87.70309448242188,
      "loss": 1.0743,
      "margin_dpo/margin_mean": 1.7041361331939697,
      "margin_dpo/margin_std": 2.4774935245513916,
      "step": 158
    },
    {
      "KL/chosen_KL_mean": -3.9513187408447266,
      "KL/mean": -4.878116607666016,
      "KL/rejected_KL_mean": -5.804912567138672,
      "KL/std": 2.7326745986938477,
      "epoch": 0.24036281179138322,
      "fcm_dpo/beta": 0.2755487263202667,
      "fcm_dpo/delta": -0.11698535084724426,
      "fcm_dpo/margin": 1.8535940647125244,
      "fcm_dpo/q_t": 0.3866373896598816,
      "grad_norm": 79.30245208740234,
      "learning_rate": 4.715998812855304e-07,
      "logits/chosen": 0.18606290221214294,
      "logits/rejected": 0.1528480499982834,
      "logps/chosen": -60.998313903808594,
      "logps/ref_chosen": -57.046993255615234,
      "logps/ref_rejected": -73.32441711425781,
      "logps/rejected": -79.12933349609375,
      "loss": 1.0857,
      "margin_dpo/margin_mean": 1.853594183921814,
      "margin_dpo/margin_std": 2.899454116821289,
      "step": 159
    },
    {
      "KL/chosen_KL_mean": -4.420055389404297,
      "KL/mean": -5.301321983337402,
      "KL/rejected_KL_mean": -6.182594299316406,
      "KL/std": 2.5777411460876465,
      "epoch": 0.2418745275888133,
      "fcm_dpo/beta": 0.26921501755714417,
      "fcm_dpo/delta": -0.07813730090856552,
      "fcm_dpo/margin": 1.7625341415405273,
      "fcm_dpo/q_t": 0.399819016456604,
      "grad_norm": 63.23322677612305,
      "learning_rate": 4.7098470178228755e-07,
      "logits/chosen": 0.06095326691865921,
      "logits/rejected": 0.0238176379352808,
      "logps/chosen": -54.226966857910156,
      "logps/ref_chosen": -49.806915283203125,
      "logps/ref_rejected": -68.3370132446289,
      "logps/rejected": -74.51960754394531,
      "loss": 1.1081,
      "margin_dpo/margin_mean": 1.7625336647033691,
      "margin_dpo/margin_std": 2.9826478958129883,
      "step": 160
    },
    {
      "KL/chosen_KL_mean": -4.423643112182617,
      "KL/mean": -5.383838653564453,
      "KL/rejected_KL_mean": -6.344034194946289,
      "KL/std": 2.7854318618774414,
      "epoch": 0.24338624338624337,
      "fcm_dpo/beta": 0.2639704644680023,
      "fcm_dpo/delta": -0.11248427629470825,
      "fcm_dpo/margin": 1.9203957319259644,
      "fcm_dpo/q_t": 0.39128196239471436,
      "grad_norm": 61.89495086669922,
      "learning_rate": 4.703633408618955e-07,
      "logits/chosen": 0.165007084608078,
      "logits/rejected": 0.13076232373714447,
      "logps/chosen": -56.92413330078125,
      "logps/ref_chosen": -52.50048828125,
      "logps/ref_rejected": -66.04540252685547,
      "logps/rejected": -72.38943481445312,
      "loss": 1.0779,
      "margin_dpo/margin_mean": 1.920395851135254,
      "margin_dpo/margin_std": 2.972217559814453,
      "step": 161
    },
    {
      "KL/chosen_KL_mean": -4.803016662597656,
      "KL/mean": -6.140519142150879,
      "KL/rejected_KL_mean": -7.47802734375,
      "KL/std": 3.0558063983917236,
      "epoch": 0.24489795918367346,
      "fcm_dpo/beta": 0.24996890127658844,
      "fcm_dpo/delta": -0.2889677882194519,
      "fcm_dpo/margin": 2.675013542175293,
      "fcm_dpo/q_t": 0.3522527813911438,
      "grad_norm": 60.043983459472656,
      "learning_rate": 4.697358159051549e-07,
      "logits/chosen": 0.2536744773387909,
      "logits/rejected": 0.2079133689403534,
      "logps/chosen": -74.27220916748047,
      "logps/ref_chosen": -69.46919250488281,
      "logps/ref_rejected": -92.00952911376953,
      "logps/rejected": -99.48755645751953,
      "loss": 0.9402,
      "margin_dpo/margin_mean": 2.6750144958496094,
      "margin_dpo/margin_std": 2.9115896224975586,
      "step": 162
    },
    {
      "KL/chosen_KL_mean": -4.403450012207031,
      "KL/mean": -5.708391189575195,
      "KL/rejected_KL_mean": -7.013332366943359,
      "KL/std": 2.9234018325805664,
      "epoch": 0.24640967498110355,
      "fcm_dpo/beta": 0.24058547616004944,
      "fcm_dpo/delta": -0.24273084104061127,
      "fcm_dpo/margin": 2.6098759174346924,
      "fcm_dpo/q_t": 0.3646067976951599,
      "grad_norm": 57.50065994262695,
      "learning_rate": 4.691021444652876e-07,
      "logits/chosen": 0.18654045462608337,
      "logits/rejected": 0.1429648995399475,
      "logps/chosen": -55.01728439331055,
      "logps/ref_chosen": -50.613834381103516,
      "logps/ref_rejected": -74.62033081054688,
      "logps/rejected": -81.63365936279297,
      "loss": 1.0005,
      "margin_dpo/margin_mean": 2.6098763942718506,
      "margin_dpo/margin_std": 3.280284881591797,
      "step": 163
    },
    {
      "KL/chosen_KL_mean": -4.918027877807617,
      "KL/mean": -6.209537029266357,
      "KL/rejected_KL_mean": -7.501045227050781,
      "KL/std": 3.221226215362549,
      "epoch": 0.24792139077853365,
      "fcm_dpo/beta": 0.22847194969654083,
      "fcm_dpo/delta": -0.20196107029914856,
      "fcm_dpo/margin": 2.583019256591797,
      "fcm_dpo/q_t": 0.37327778339385986,
      "grad_norm": 54.232357025146484,
      "learning_rate": 4.6846234426744624e-07,
      "logits/chosen": 0.17183548212051392,
      "logits/rejected": 0.11392419040203094,
      "logps/chosen": -59.76613998413086,
      "logps/ref_chosen": -54.848114013671875,
      "logps/ref_rejected": -79.0630111694336,
      "logps/rejected": -86.56405639648438,
      "loss": 1.0215,
      "margin_dpo/margin_mean": 2.583019495010376,
      "margin_dpo/margin_std": 3.419358015060425,
      "step": 164
    },
    {
      "KL/chosen_KL_mean": -5.296621322631836,
      "KL/mean": -6.522599220275879,
      "KL/rejected_KL_mean": -7.7485809326171875,
      "KL/std": 3.3008265495300293,
      "epoch": 0.2494331065759637,
      "fcm_dpo/beta": 0.2219952940940857,
      "fcm_dpo/delta": -0.152376189827919,
      "fcm_dpo/margin": 2.451958417892456,
      "fcm_dpo/q_t": 0.37785401940345764,
      "grad_norm": 51.94645309448242,
      "learning_rate": 4.678164332082175e-07,
      "logits/chosen": 0.2255769670009613,
      "logits/rejected": 0.17380759119987488,
      "logps/chosen": -56.385833740234375,
      "logps/ref_chosen": -51.089210510253906,
      "logps/ref_rejected": -71.23370361328125,
      "logps/rejected": -78.98228454589844,
      "loss": 1.0168,
      "margin_dpo/margin_mean": 2.451957941055298,
      "margin_dpo/margin_std": 2.9465808868408203,
      "step": 165
    },
    {
      "KL/chosen_KL_mean": -5.235536575317383,
      "KL/mean": -6.11887264251709,
      "KL/rejected_KL_mean": -7.0022125244140625,
      "KL/std": 2.914708375930786,
      "epoch": 0.2509448223733938,
      "fcm_dpo/beta": 0.21907421946525574,
      "fcm_dpo/delta": 0.013396350666880608,
      "fcm_dpo/margin": 1.7666780948638916,
      "fcm_dpo/q_t": 0.4149834215641022,
      "grad_norm": 61.127227783203125,
      "learning_rate": 4.6716442935512214e-07,
      "logits/chosen": 0.20381613075733185,
      "logits/rejected": 0.12277201563119888,
      "logps/chosen": -68.42634582519531,
      "logps/ref_chosen": -63.19081115722656,
      "logps/ref_rejected": -93.8402099609375,
      "logps/rejected": -100.84242248535156,
      "loss": 1.1348,
      "margin_dpo/margin_mean": 1.7666782140731812,
      "margin_dpo/margin_std": 2.991272449493408,
      "step": 166
    },
    {
      "KL/chosen_KL_mean": -4.86152458190918,
      "KL/mean": -6.267144203186035,
      "KL/rejected_KL_mean": -7.672763824462891,
      "KL/std": 3.211589813232422,
      "epoch": 0.25245653817082386,
      "fcm_dpo/beta": 0.2118726372718811,
      "fcm_dpo/delta": -0.20976994931697845,
      "fcm_dpo/margin": 2.81123685836792,
      "fcm_dpo/q_t": 0.36726197600364685,
      "grad_norm": 46.722129821777344,
      "learning_rate": 4.6650635094610966e-07,
      "logits/chosen": 0.15115973353385925,
      "logits/rejected": 0.11766296625137329,
      "logps/chosen": -63.785797119140625,
      "logps/ref_chosen": -58.92427062988281,
      "logps/ref_rejected": -72.97377014160156,
      "logps/rejected": -80.64653015136719,
      "loss": 0.9841,
      "margin_dpo/margin_mean": 2.811237335205078,
      "margin_dpo/margin_std": 3.2436766624450684,
      "step": 167
    },
    {
      "KL/chosen_KL_mean": -6.010761260986328,
      "KL/mean": -6.915679931640625,
      "KL/rejected_KL_mean": -7.8206024169921875,
      "KL/std": 3.365790367126465,
      "epoch": 0.25396825396825395,
      "fcm_dpo/beta": 0.21115413308143616,
      "fcm_dpo/delta": 0.018556442111730576,
      "fcm_dpo/margin": 1.809836983680725,
      "fcm_dpo/q_t": 0.4141741096973419,
      "grad_norm": 59.15978240966797,
      "learning_rate": 4.6584221638904767e-07,
      "logits/chosen": 0.16713829338550568,
      "logits/rejected": 0.13781222701072693,
      "logps/chosen": -71.66213989257812,
      "logps/ref_chosen": -65.65138244628906,
      "logps/ref_rejected": -79.71418762207031,
      "logps/rejected": -87.5347900390625,
      "loss": 1.1305,
      "margin_dpo/margin_mean": 1.8098368644714355,
      "margin_dpo/margin_std": 2.919257640838623,
      "step": 168
    },
    {
      "KL/chosen_KL_mean": -5.380914688110352,
      "KL/mean": -6.703707695007324,
      "KL/rejected_KL_mean": -8.026496887207031,
      "KL/std": 3.6164283752441406,
      "epoch": 0.25547996976568405,
      "fcm_dpo/beta": 0.20682457089424133,
      "fcm_dpo/delta": -0.15560072660446167,
      "fcm_dpo/margin": 2.645585060119629,
      "fcm_dpo/q_t": 0.3858036398887634,
      "grad_norm": 53.96177291870117,
      "learning_rate": 4.651720442612075e-07,
      "logits/chosen": 0.23225507140159607,
      "logits/rejected": 0.20093366503715515,
      "logps/chosen": -66.80677795410156,
      "logps/ref_chosen": -61.425865173339844,
      "logps/ref_rejected": -76.09590148925781,
      "logps/rejected": -84.12239074707031,
      "loss": 1.051,
      "margin_dpo/margin_mean": 2.645585060119629,
      "margin_dpo/margin_std": 3.978856086730957,
      "step": 169
    },
    {
      "KL/chosen_KL_mean": -5.6312408447265625,
      "KL/mean": -6.840090751647949,
      "KL/rejected_KL_mean": -8.04893684387207,
      "KL/std": 3.505796432495117,
      "epoch": 0.25699168556311414,
      "fcm_dpo/beta": 0.2026190310716629,
      "fcm_dpo/delta": -0.09439290314912796,
      "fcm_dpo/margin": 2.417698383331299,
      "fcm_dpo/q_t": 0.39143288135528564,
      "grad_norm": 46.08080291748047,
      "learning_rate": 4.6449585330874425e-07,
      "logits/chosen": 0.16252544522285461,
      "logits/rejected": 0.16037079691886902,
      "logps/chosen": -62.28443145751953,
      "logps/ref_chosen": -56.65319061279297,
      "logps/ref_rejected": -63.45965576171875,
      "logps/rejected": -71.50859069824219,
      "loss": 1.0905,
      "margin_dpo/margin_mean": 2.4176979064941406,
      "margin_dpo/margin_std": 3.8021955490112305,
      "step": 170
    },
    {
      "KL/chosen_KL_mean": -5.986204147338867,
      "KL/mean": -7.4430012702941895,
      "KL/rejected_KL_mean": -8.899799346923828,
      "KL/std": 4.016740798950195,
      "epoch": 0.2585034013605442,
      "fcm_dpo/beta": 0.19423140585422516,
      "fcm_dpo/delta": -0.17897388339042664,
      "fcm_dpo/margin": 2.91359806060791,
      "fcm_dpo/q_t": 0.3770707845687866,
      "grad_norm": 51.48481369018555,
      "learning_rate": 4.6381366244617224e-07,
      "logits/chosen": 0.23160839080810547,
      "logits/rejected": 0.18538227677345276,
      "logps/chosen": -69.72096252441406,
      "logps/ref_chosen": -63.73476028442383,
      "logps/ref_rejected": -78.50328063964844,
      "logps/rejected": -87.403076171875,
      "loss": 1.0544,
      "margin_dpo/margin_mean": 2.9135982990264893,
      "margin_dpo/margin_std": 4.2580180168151855,
      "step": 171
    },
    {
      "KL/chosen_KL_mean": -6.52869987487793,
      "KL/mean": -8.056679725646973,
      "KL/rejected_KL_mean": -9.58465576171875,
      "KL/std": 4.038125038146973,
      "epoch": 0.2600151171579743,
      "fcm_dpo/beta": 0.1900688111782074,
      "fcm_dpo/delta": -0.19167888164520264,
      "fcm_dpo/margin": 3.0559561252593994,
      "fcm_dpo/q_t": 0.37122875452041626,
      "grad_norm": 44.15302276611328,
      "learning_rate": 4.631254907558365e-07,
      "logits/chosen": 0.23724429309368134,
      "logits/rejected": 0.186652272939682,
      "logps/chosen": -58.73046112060547,
      "logps/ref_chosen": -52.201759338378906,
      "logps/ref_rejected": -82.85285949707031,
      "logps/rejected": -92.43751525878906,
      "loss": 1.0195,
      "margin_dpo/margin_mean": 3.055955410003662,
      "margin_dpo/margin_std": 3.9646401405334473,
      "step": 172
    },
    {
      "KL/chosen_KL_mean": -6.710458755493164,
      "KL/mean": -8.274261474609375,
      "KL/rejected_KL_mean": -9.838062286376953,
      "KL/std": 4.230109214782715,
      "epoch": 0.2615268329554044,
      "fcm_dpo/beta": 0.17975875735282898,
      "fcm_dpo/delta": -0.17714735865592957,
      "fcm_dpo/margin": 3.1276025772094727,
      "fcm_dpo/q_t": 0.38370195031166077,
      "grad_norm": 43.27824020385742,
      "learning_rate": 4.624313574873786e-07,
      "logits/chosen": 0.24271810054779053,
      "logits/rejected": 0.1620805412530899,
      "logps/chosen": -62.14518356323242,
      "logps/ref_chosen": -55.434722900390625,
      "logps/ref_rejected": -77.81967163085938,
      "logps/rejected": -87.6577377319336,
      "loss": 1.0919,
      "margin_dpo/margin_mean": 3.1276025772094727,
      "margin_dpo/margin_std": 5.053474426269531,
      "step": 173
    },
    {
      "KL/chosen_KL_mean": -7.342222213745117,
      "KL/mean": -9.029733657836914,
      "KL/rejected_KL_mean": -10.717247009277344,
      "KL/std": 4.268304824829102,
      "epoch": 0.26303854875283444,
      "fcm_dpo/beta": 0.1746232807636261,
      "fcm_dpo/delta": -0.20176327228546143,
      "fcm_dpo/margin": 3.375027894973755,
      "fcm_dpo/q_t": 0.3743577003479004,
      "grad_norm": 46.385498046875,
      "learning_rate": 4.61731282057198e-07,
      "logits/chosen": 0.232137069106102,
      "logits/rejected": 0.17018833756446838,
      "logps/chosen": -64.51417541503906,
      "logps/ref_chosen": -57.17195129394531,
      "logps/ref_rejected": -85.47578430175781,
      "logps/rejected": -96.19302368164062,
      "loss": 1.0447,
      "margin_dpo/margin_mean": 3.375028371810913,
      "margin_dpo/margin_std": 4.97684383392334,
      "step": 174
    },
    {
      "KL/chosen_KL_mean": -7.433553695678711,
      "KL/mean": -9.300992965698242,
      "KL/rejected_KL_mean": -11.168437957763672,
      "KL/std": 4.688220977783203,
      "epoch": 0.26455026455026454,
      "fcm_dpo/beta": 0.16730672121047974,
      "fcm_dpo/delta": -0.23966068029403687,
      "fcm_dpo/margin": 3.7348804473876953,
      "fcm_dpo/q_t": 0.37021753191947937,
      "grad_norm": 44.081607818603516,
      "learning_rate": 4.6102528404790965e-07,
      "logits/chosen": 0.2679125964641571,
      "logits/rejected": 0.23914484679698944,
      "logps/chosen": -75.09918212890625,
      "logps/ref_chosen": -67.6656265258789,
      "logps/ref_rejected": -84.36766815185547,
      "logps/rejected": -95.53610229492188,
      "loss": 1.0271,
      "margin_dpo/margin_mean": 3.7348804473876953,
      "margin_dpo/margin_std": 5.312939643859863,
      "step": 175
    },
    {
      "KL/chosen_KL_mean": -8.312698364257812,
      "KL/mean": -9.597501754760742,
      "KL/rejected_KL_mean": -10.882305145263672,
      "KL/std": 4.97230339050293,
      "epoch": 0.2660619803476946,
      "fcm_dpo/beta": 0.16365355253219604,
      "fcm_dpo/delta": -0.022150814533233643,
      "fcm_dpo/margin": 2.569603681564331,
      "fcm_dpo/q_t": 0.4120991826057434,
      "grad_norm": 52.76869583129883,
      "learning_rate": 4.603133832077953e-07,
      "logits/chosen": 0.21717190742492676,
      "logits/rejected": 0.19207710027694702,
      "logps/chosen": -86.17146301269531,
      "logps/ref_chosen": -77.8587646484375,
      "logps/ref_rejected": -81.08732604980469,
      "logps/rejected": -91.96963500976562,
      "loss": 1.1776,
      "margin_dpo/margin_mean": 2.569603681564331,
      "margin_dpo/margin_std": 5.330869197845459,
      "step": 176
    },
    {
      "KL/chosen_KL_mean": -7.614345550537109,
      "KL/mean": -10.125162124633789,
      "KL/rejected_KL_mean": -12.635982513427734,
      "KL/std": 5.118657112121582,
      "epoch": 0.2675736961451247,
      "fcm_dpo/beta": 0.15402725338935852,
      "fcm_dpo/delta": -0.4072016775608063,
      "fcm_dpo/margin": 5.021634101867676,
      "fcm_dpo/q_t": 0.33033889532089233,
      "grad_norm": 44.16107940673828,
      "learning_rate": 4.5959559945025183e-07,
      "logits/chosen": 0.31596648693084717,
      "logits/rejected": 0.22701580822467804,
      "logps/chosen": -62.83474349975586,
      "logps/ref_chosen": -55.22039794921875,
      "logps/ref_rejected": -92.54973602294922,
      "logps/rejected": -105.18571472167969,
      "loss": 0.8755,
      "margin_dpo/margin_mean": 5.021634101867676,
      "margin_dpo/margin_std": 4.907200813293457,
      "step": 177
    },
    {
      "KL/chosen_KL_mean": -8.229791641235352,
      "KL/mean": -9.763175964355469,
      "KL/rejected_KL_mean": -11.296558380126953,
      "KL/std": 4.665809631347656,
      "epoch": 0.2690854119425548,
      "fcm_dpo/beta": 0.14871251583099365,
      "fcm_dpo/delta": -0.06117831915616989,
      "fcm_dpo/margin": 3.0667619705200195,
      "fcm_dpo/q_t": 0.3978401720523834,
      "grad_norm": 40.795989990234375,
      "learning_rate": 4.588719528532341e-07,
      "logits/chosen": 0.2093246877193451,
      "logits/rejected": 0.1630682349205017,
      "logps/chosen": -69.040283203125,
      "logps/ref_chosen": -60.81049346923828,
      "logps/ref_rejected": -81.12973022460938,
      "logps/rejected": -92.42628479003906,
      "loss": 1.0854,
      "margin_dpo/margin_mean": 3.0667624473571777,
      "margin_dpo/margin_std": 4.368191242218018,
      "step": 178
    },
    {
      "KL/chosen_KL_mean": -9.152389526367188,
      "KL/mean": -10.899008750915527,
      "KL/rejected_KL_mean": -12.645626068115234,
      "KL/std": 4.8825554847717285,
      "epoch": 0.2705971277399849,
      "fcm_dpo/beta": 0.14718642830848694,
      "fcm_dpo/delta": -0.1202310174703598,
      "fcm_dpo/margin": 3.4932327270507812,
      "fcm_dpo/q_t": 0.38942965865135193,
      "grad_norm": 39.2275390625,
      "learning_rate": 4.581424636586928e-07,
      "logits/chosen": 0.25605344772338867,
      "logits/rejected": 0.23996365070343018,
      "logps/chosen": -74.82410430908203,
      "logps/ref_chosen": -65.67171478271484,
      "logps/ref_rejected": -75.32586669921875,
      "logps/rejected": -87.97149658203125,
      "loss": 1.0858,
      "margin_dpo/margin_mean": 3.4932332038879395,
      "margin_dpo/margin_std": 5.598725318908691,
      "step": 179
    },
    {
      "KL/chosen_KL_mean": -7.865631103515625,
      "KL/mean": -9.393498420715332,
      "KL/rejected_KL_mean": -10.921365737915039,
      "KL/std": 5.0374755859375,
      "epoch": 0.272108843537415,
      "fcm_dpo/beta": 0.14556020498275757,
      "fcm_dpo/delta": -0.04694035276770592,
      "fcm_dpo/margin": 3.0557327270507812,
      "fcm_dpo/q_t": 0.4066722095012665,
      "grad_norm": 39.35282897949219,
      "learning_rate": 4.5740715227200897e-07,
      "logits/chosen": 0.10142640769481659,
      "logits/rejected": 0.0828198567032814,
      "logps/chosen": -64.5484390258789,
      "logps/ref_chosen": -56.68280792236328,
      "logps/ref_rejected": -64.94414520263672,
      "logps/rejected": -75.86550903320312,
      "loss": 1.1508,
      "margin_dpo/margin_mean": 3.0557332038879395,
      "margin_dpo/margin_std": 5.835036754608154,
      "step": 180
    },
    {
      "KL/chosen_KL_mean": -7.593027114868164,
      "KL/mean": -10.005693435668945,
      "KL/rejected_KL_mean": -12.418354034423828,
      "KL/std": 5.6760430335998535,
      "epoch": 0.273620559334845,
      "fcm_dpo/beta": 0.13861241936683655,
      "fcm_dpo/delta": -0.2883051633834839,
      "fcm_dpo/margin": 4.825331687927246,
      "fcm_dpo/q_t": 0.3515399098396301,
      "grad_norm": 34.905521392822266,
      "learning_rate": 4.566660392614228e-07,
      "logits/chosen": 0.29494979977607727,
      "logits/rejected": 0.25655397772789,
      "logps/chosen": -68.36907958984375,
      "logps/ref_chosen": -60.77604675292969,
      "logps/ref_rejected": -83.98361206054688,
      "logps/rejected": -96.40196990966797,
      "loss": 0.9296,
      "margin_dpo/margin_mean": 4.825331211090088,
      "margin_dpo/margin_std": 5.00650691986084,
      "step": 181
    },
    {
      "KL/chosen_KL_mean": -8.961589813232422,
      "KL/mean": -11.407659530639648,
      "KL/rejected_KL_mean": -13.85372543334961,
      "KL/std": 5.829415321350098,
      "epoch": 0.2751322751322751,
      "fcm_dpo/beta": 0.13101884722709656,
      "fcm_dpo/delta": -0.25866827368736267,
      "fcm_dpo/margin": 4.892138957977295,
      "fcm_dpo/q_t": 0.36321139335632324,
      "grad_norm": 35.85899353027344,
      "learning_rate": 4.5591914535745817e-07,
      "logits/chosen": 0.24604353308677673,
      "logits/rejected": 0.1717510223388672,
      "logps/chosen": -69.21537780761719,
      "logps/ref_chosen": -60.2537841796875,
      "logps/ref_rejected": -89.7706298828125,
      "logps/rejected": -103.62435913085938,
      "loss": 1.0077,
      "margin_dpo/margin_mean": 4.892138957977295,
      "margin_dpo/margin_std": 6.573358535766602,
      "step": 182
    },
    {
      "KL/chosen_KL_mean": -10.759580612182617,
      "KL/mean": -11.767297744750977,
      "KL/rejected_KL_mean": -12.775016784667969,
      "KL/std": 5.797929763793945,
      "epoch": 0.2766439909297052,
      "fcm_dpo/beta": 0.13212950527668,
      "fcm_dpo/delta": 0.137569397687912,
      "fcm_dpo/margin": 2.015444278717041,
      "fcm_dpo/q_t": 0.44033879041671753,
      "grad_norm": 38.02180862426758,
      "learning_rate": 4.551664914523433e-07,
      "logits/chosen": 0.2160715013742447,
      "logits/rejected": 0.19661223888397217,
      "logps/chosen": -72.52099609375,
      "logps/ref_chosen": -61.76142120361328,
      "logps/ref_rejected": -72.54627990722656,
      "logps/rejected": -85.32130432128906,
      "loss": 1.2589,
      "margin_dpo/margin_mean": 2.015444278717041,
      "margin_dpo/margin_std": 5.434455394744873,
      "step": 183
    },
    {
      "KL/chosen_KL_mean": -8.564130783081055,
      "KL/mean": -10.616769790649414,
      "KL/rejected_KL_mean": -12.66940689086914,
      "KL/std": 5.086275100708008,
      "epoch": 0.2781557067271353,
      "fcm_dpo/beta": 0.1294369101524353,
      "fcm_dpo/delta": -0.1415943056344986,
      "fcm_dpo/margin": 4.105273246765137,
      "fcm_dpo/q_t": 0.3822307884693146,
      "grad_norm": 29.23053550720215,
      "learning_rate": 4.544080985994258e-07,
      "logits/chosen": 0.3366903066635132,
      "logits/rejected": 0.27632272243499756,
      "logps/chosen": -55.40485382080078,
      "logps/ref_chosen": -46.840721130371094,
      "logps/ref_rejected": -69.3609390258789,
      "logps/rejected": -82.03034973144531,
      "loss": 1.0213,
      "margin_dpo/margin_mean": 4.105273246765137,
      "margin_dpo/margin_std": 4.934863090515137,
      "step": 184
    },
    {
      "KL/chosen_KL_mean": -9.769927978515625,
      "KL/mean": -11.914457321166992,
      "KL/rejected_KL_mean": -14.058982849121094,
      "KL/std": 5.9276604652404785,
      "epoch": 0.2796674225245654,
      "fcm_dpo/beta": 0.1260417103767395,
      "fcm_dpo/delta": -0.15020085871219635,
      "fcm_dpo/margin": 4.289056777954102,
      "fcm_dpo/q_t": 0.3870917856693268,
      "grad_norm": 31.231266021728516,
      "learning_rate": 4.5364398801258394e-07,
      "logits/chosen": 0.25812608003616333,
      "logits/rejected": 0.21318121254444122,
      "logps/chosen": -62.091068267822266,
      "logps/ref_chosen": -52.32114028930664,
      "logps/ref_rejected": -68.3885726928711,
      "logps/rejected": -82.44755554199219,
      "loss": 1.1037,
      "margin_dpo/margin_mean": 4.289056777954102,
      "margin_dpo/margin_std": 7.260632514953613,
      "step": 185
    },
    {
      "KL/chosen_KL_mean": -9.387929916381836,
      "KL/mean": -11.70936393737793,
      "KL/rejected_KL_mean": -14.03079605102539,
      "KL/std": 5.995885848999023,
      "epoch": 0.2811791383219955,
      "fcm_dpo/beta": 0.12234297394752502,
      "fcm_dpo/delta": -0.1784307062625885,
      "fcm_dpo/margin": 4.642869472503662,
      "fcm_dpo/q_t": 0.3811233341693878,
      "grad_norm": 37.083396911621094,
      "learning_rate": 4.5287418106563354e-07,
      "logits/chosen": 0.20448735356330872,
      "logits/rejected": 0.16464738547801971,
      "logps/chosen": -76.80805969238281,
      "logps/ref_chosen": -67.42012786865234,
      "logps/ref_rejected": -82.50968933105469,
      "logps/rejected": -96.54048156738281,
      "loss": 1.0687,
      "margin_dpo/margin_mean": 4.64286994934082,
      "margin_dpo/margin_std": 7.29622745513916,
      "step": 186
    },
    {
      "KL/chosen_KL_mean": -10.648483276367188,
      "KL/mean": -12.794281005859375,
      "KL/rejected_KL_mean": -14.940082550048828,
      "KL/std": 6.210048675537109,
      "epoch": 0.28269085411942557,
      "fcm_dpo/beta": 0.11887460947036743,
      "fcm_dpo/delta": -0.11676714569330215,
      "fcm_dpo/margin": 4.291601657867432,
      "fcm_dpo/q_t": 0.3875843286514282,
      "grad_norm": 38.01297378540039,
      "learning_rate": 4.520986992917297e-07,
      "logits/chosen": 0.24502941966056824,
      "logits/rejected": 0.1913175731897354,
      "logps/chosen": -86.17398071289062,
      "logps/ref_chosen": -75.52549743652344,
      "logps/ref_rejected": -94.76289367675781,
      "logps/rejected": -109.70297241210938,
      "loss": 1.0866,
      "margin_dpo/margin_mean": 4.29160213470459,
      "margin_dpo/margin_std": 6.724270820617676,
      "step": 187
    },
    {
      "KL/chosen_KL_mean": -9.933109283447266,
      "KL/mean": -12.137847900390625,
      "KL/rejected_KL_mean": -14.342586517333984,
      "KL/std": 6.472745418548584,
      "epoch": 0.2842025699168556,
      "fcm_dpo/beta": 0.11644455790519714,
      "fcm_dpo/delta": -0.11966116726398468,
      "fcm_dpo/margin": 4.409476280212402,
      "fcm_dpo/q_t": 0.3881905674934387,
      "grad_norm": 36.36927032470703,
      "learning_rate": 4.5131756438276466e-07,
      "logits/chosen": 0.2666782736778259,
      "logits/rejected": 0.22476297616958618,
      "logps/chosen": -81.45643615722656,
      "logps/ref_chosen": -71.52333068847656,
      "logps/ref_rejected": -78.29949951171875,
      "logps/rejected": -92.64208984375,
      "loss": 1.0861,
      "margin_dpo/margin_mean": 4.409476280212402,
      "margin_dpo/margin_std": 7.109898567199707,
      "step": 188
    },
    {
      "KL/chosen_KL_mean": -9.555837631225586,
      "KL/mean": -11.801464080810547,
      "KL/rejected_KL_mean": -14.04709243774414,
      "KL/std": 6.3911943435668945,
      "epoch": 0.2857142857142857,
      "fcm_dpo/beta": 0.11252661049365997,
      "fcm_dpo/delta": -0.11500401049852371,
      "fcm_dpo/margin": 4.491254806518555,
      "fcm_dpo/q_t": 0.3868621587753296,
      "grad_norm": 33.63637161254883,
      "learning_rate": 4.5053079818876096e-07,
      "logits/chosen": 0.2828933596611023,
      "logits/rejected": 0.2940501868724823,
      "logps/chosen": -81.73210144042969,
      "logps/ref_chosen": -72.17626953125,
      "logps/ref_rejected": -75.26313781738281,
      "logps/rejected": -89.31022644042969,
      "loss": 1.074,
      "margin_dpo/margin_mean": 4.491255760192871,
      "margin_dpo/margin_std": 6.432233810424805,
      "step": 189
    },
    {
      "KL/chosen_KL_mean": -9.486690521240234,
      "KL/mean": -12.508905410766602,
      "KL/rejected_KL_mean": -15.531116485595703,
      "KL/std": 6.638813495635986,
      "epoch": 0.2872260015117158,
      "fcm_dpo/beta": 0.10864052176475525,
      "fcm_dpo/delta": -0.2747190594673157,
      "fcm_dpo/margin": 6.044425010681152,
      "fcm_dpo/q_t": 0.35702377557754517,
      "grad_norm": 32.1728401184082,
      "learning_rate": 4.4973842271726024e-07,
      "logits/chosen": 0.32548272609710693,
      "logits/rejected": 0.19106276333332062,
      "logps/chosen": -64.1109619140625,
      "logps/ref_chosen": -54.624271392822266,
      "logps/ref_rejected": -101.47068786621094,
      "logps/rejected": -117.00180053710938,
      "loss": 0.9636,
      "margin_dpo/margin_mean": 6.044425010681152,
      "margin_dpo/margin_std": 7.100013256072998,
      "step": 190
    },
    {
      "KL/chosen_KL_mean": -11.378448486328125,
      "KL/mean": -13.6748046875,
      "KL/rejected_KL_mean": -15.97116470336914,
      "KL/std": 6.738143444061279,
      "epoch": 0.2887377173091459,
      "fcm_dpo/beta": 0.10574862360954285,
      "fcm_dpo/delta": -0.0899963304400444,
      "fcm_dpo/margin": 4.592713832855225,
      "fcm_dpo/q_t": 0.3922455310821533,
      "grad_norm": 34.2210807800293,
      "learning_rate": 4.48940460132708e-07,
      "logits/chosen": 0.3366280794143677,
      "logits/rejected": 0.3104870915412903,
      "logps/chosen": -84.31095886230469,
      "logps/ref_chosen": -72.93251037597656,
      "logps/ref_rejected": -89.95103454589844,
      "logps/rejected": -105.92219543457031,
      "loss": 1.0981,
      "margin_dpo/margin_mean": 4.592713356018066,
      "margin_dpo/margin_std": 7.371159553527832,
      "step": 191
    },
    {
      "KL/chosen_KL_mean": -11.613393783569336,
      "KL/mean": -13.261909484863281,
      "KL/rejected_KL_mean": -14.910423278808594,
      "KL/std": 7.089072227478027,
      "epoch": 0.29024943310657597,
      "fcm_dpo/beta": 0.1063033863902092,
      "fcm_dpo/delta": 0.050992466509342194,
      "fcm_dpo/margin": 3.297028064727783,
      "fcm_dpo/q_t": 0.42166829109191895,
      "grad_norm": 27.335500717163086,
      "learning_rate": 4.481369327558329e-07,
      "logits/chosen": 0.30276888608932495,
      "logits/rejected": 0.2783222496509552,
      "logps/chosen": -65.61451721191406,
      "logps/ref_chosen": -54.001121520996094,
      "logps/ref_rejected": -63.531551361083984,
      "logps/rejected": -78.44197082519531,
      "loss": 1.1783,
      "margin_dpo/margin_mean": 3.297028064727783,
      "margin_dpo/margin_std": 6.561872482299805,
      "step": 192
    },
    {
      "KL/chosen_KL_mean": -10.664052963256836,
      "KL/mean": -13.527870178222656,
      "KL/rejected_KL_mean": -16.39168930053711,
      "KL/std": 6.945682525634766,
      "epoch": 0.29176114890400606,
      "fcm_dpo/beta": 0.10263273119926453,
      "fcm_dpo/delta": -0.20108722150325775,
      "fcm_dpo/margin": 5.727632522583008,
      "fcm_dpo/q_t": 0.3708486557006836,
      "grad_norm": 26.038829803466797,
      "learning_rate": 4.47327863063023e-07,
      "logits/chosen": 0.2456463873386383,
      "logits/rejected": 0.22312812507152557,
      "logps/chosen": -67.413330078125,
      "logps/ref_chosen": -56.74927520751953,
      "logps/ref_rejected": -58.80629348754883,
      "logps/rejected": -75.19798278808594,
      "loss": 0.9983,
      "margin_dpo/margin_mean": 5.727632522583008,
      "margin_dpo/margin_std": 7.0640764236450195,
      "step": 193
    },
    {
      "KL/chosen_KL_mean": -10.503494262695312,
      "KL/mean": -12.863252639770508,
      "KL/rejected_KL_mean": -15.223018646240234,
      "KL/std": 6.697719573974609,
      "epoch": 0.29327286470143615,
      "fcm_dpo/beta": 0.10034053027629852,
      "fcm_dpo/delta": -0.07828196883201599,
      "fcm_dpo/margin": 4.719526290893555,
      "fcm_dpo/q_t": 0.395353227853775,
      "grad_norm": 26.579702377319336,
      "learning_rate": 4.4651327368569684e-07,
      "logits/chosen": 0.31545290350914,
      "logits/rejected": 0.28684914112091064,
      "logps/chosen": -67.15293884277344,
      "logps/ref_chosen": -56.64944076538086,
      "logps/ref_rejected": -69.98954772949219,
      "logps/rejected": -85.21256256103516,
      "loss": 1.1333,
      "margin_dpo/margin_mean": 4.719526767730713,
      "margin_dpo/margin_std": 8.421168327331543,
      "step": 194
    },
    {
      "KL/chosen_KL_mean": -12.053577423095703,
      "KL/mean": -14.948898315429688,
      "KL/rejected_KL_mean": -17.844223022460938,
      "KL/std": 7.79929256439209,
      "epoch": 0.2947845804988662,
      "fcm_dpo/beta": 0.097338005900383,
      "fcm_dpo/delta": -0.175175741314888,
      "fcm_dpo/margin": 5.790646553039551,
      "fcm_dpo/q_t": 0.3771982192993164,
      "grad_norm": 27.508337020874023,
      "learning_rate": 4.4569318740967043e-07,
      "logits/chosen": 0.2035025656223297,
      "logits/rejected": 0.20420676469802856,
      "logps/chosen": -82.4633560180664,
      "logps/ref_chosen": -70.40977478027344,
      "logps/ref_rejected": -74.39448547363281,
      "logps/rejected": -92.23870849609375,
      "loss": 1.0269,
      "margin_dpo/margin_mean": 5.790646553039551,
      "margin_dpo/margin_std": 7.719186782836914,
      "step": 195
    },
    {
      "KL/chosen_KL_mean": -11.237726211547852,
      "KL/mean": -13.784488677978516,
      "KL/rejected_KL_mean": -16.331253051757812,
      "KL/std": 7.774652481079102,
      "epoch": 0.2962962962962963,
      "fcm_dpo/beta": 0.09619662165641785,
      "fcm_dpo/delta": -0.094697006046772,
      "fcm_dpo/margin": 5.093526840209961,
      "fcm_dpo/q_t": 0.3892650604248047,
      "grad_norm": 25.448198318481445,
      "learning_rate": 4.448676271745197e-07,
      "logits/chosen": 0.3075557351112366,
      "logits/rejected": 0.267707496881485,
      "logps/chosen": -70.46530151367188,
      "logps/ref_chosen": -59.227577209472656,
      "logps/ref_rejected": -83.54757690429688,
      "logps/rejected": -99.87882995605469,
      "loss": 1.0733,
      "margin_dpo/margin_mean": 5.093526840209961,
      "margin_dpo/margin_std": 7.3492279052734375,
      "step": 196
    },
    {
      "KL/chosen_KL_mean": -10.425508499145508,
      "KL/mean": -13.5262451171875,
      "KL/rejected_KL_mean": -16.626976013183594,
      "KL/std": 8.094280242919922,
      "epoch": 0.29780801209372637,
      "fcm_dpo/beta": 0.09323252737522125,
      "fcm_dpo/delta": -0.1887865662574768,
      "fcm_dpo/margin": 6.201464653015137,
      "fcm_dpo/q_t": 0.37647128105163574,
      "grad_norm": 25.889739990234375,
      "learning_rate": 4.440366160729392e-07,
      "logits/chosen": 0.3803238570690155,
      "logits/rejected": 0.3330530822277069,
      "logps/chosen": -61.95463943481445,
      "logps/ref_chosen": -51.52912902832031,
      "logps/ref_rejected": -73.70631408691406,
      "logps/rejected": -90.33328247070312,
      "loss": 1.0836,
      "margin_dpo/margin_mean": 6.201464653015137,
      "margin_dpo/margin_std": 9.826972007751465,
      "step": 197
    },
    {
      "KL/chosen_KL_mean": -10.89830207824707,
      "KL/mean": -14.291038513183594,
      "KL/rejected_KL_mean": -17.683780670166016,
      "KL/std": 7.223607063293457,
      "epoch": 0.29931972789115646,
      "fcm_dpo/beta": 0.08935528993606567,
      "fcm_dpo/delta": -0.21908718347549438,
      "fcm_dpo/margin": 6.7854743003845215,
      "fcm_dpo/q_t": 0.3636322617530823,
      "grad_norm": 24.653465270996094,
      "learning_rate": 4.432001773500957e-07,
      "logits/chosen": 0.34279125928878784,
      "logits/rejected": 0.3045985698699951,
      "logps/chosen": -70.68098449707031,
      "logps/ref_chosen": -59.78268051147461,
      "logps/ref_rejected": -72.24533081054688,
      "logps/rejected": -89.92910766601562,
      "loss": 0.9697,
      "margin_dpo/margin_mean": 6.7854743003845215,
      "margin_dpo/margin_std": 7.347314357757568,
      "step": 198
    },
    {
      "KL/chosen_KL_mean": -12.324453353881836,
      "KL/mean": -14.889604568481445,
      "KL/rejected_KL_mean": -17.454750061035156,
      "KL/std": 7.923635005950928,
      "epoch": 0.30083144368858655,
      "fcm_dpo/beta": 0.08797682076692581,
      "fcm_dpo/delta": -0.05551917850971222,
      "fcm_dpo/margin": 5.130303859710693,
      "fcm_dpo/q_t": 0.40116453170776367,
      "grad_norm": 26.78217124938965,
      "learning_rate": 4.4235833440297856e-07,
      "logits/chosen": 0.28896427154541016,
      "logits/rejected": 0.20546600222587585,
      "logps/chosen": -68.71122741699219,
      "logps/ref_chosen": -56.38677215576172,
      "logps/ref_rejected": -74.56779479980469,
      "logps/rejected": -92.02255249023438,
      "loss": 1.1612,
      "margin_dpo/margin_mean": 5.130303382873535,
      "margin_dpo/margin_std": 9.651971817016602,
      "step": 199
    },
    {
      "KL/chosen_KL_mean": -9.989532470703125,
      "KL/mean": -13.755962371826172,
      "KL/rejected_KL_mean": -17.52239990234375,
      "KL/std": 8.623006820678711,
      "epoch": 0.30234315948601664,
      "fcm_dpo/beta": 0.08316583186388016,
      "fcm_dpo/delta": -0.24403418600559235,
      "fcm_dpo/margin": 7.532864570617676,
      "fcm_dpo/q_t": 0.3652196526527405,
      "grad_norm": 24.068204879760742,
      "learning_rate": 4.415111107797445e-07,
      "logits/chosen": 0.3554863929748535,
      "logits/rejected": 0.28665488958358765,
      "logps/chosen": -67.81385803222656,
      "logps/ref_chosen": -57.82432556152344,
      "logps/ref_rejected": -89.28246307373047,
      "logps/rejected": -106.80486297607422,
      "loss": 1.017,
      "margin_dpo/margin_mean": 7.532864093780518,
      "margin_dpo/margin_std": 10.306327819824219,
      "step": 200
    },
    {
      "KL/chosen_KL_mean": -12.684650421142578,
      "KL/mean": -15.855332374572754,
      "KL/rejected_KL_mean": -19.026016235351562,
      "KL/std": 8.930841445922852,
      "epoch": 0.30385487528344673,
      "fcm_dpo/beta": 0.08092570304870605,
      "fcm_dpo/delta": -0.11991845071315765,
      "fcm_dpo/margin": 6.341360092163086,
      "fcm_dpo/q_t": 0.38912898302078247,
      "grad_norm": 25.944599151611328,
      "learning_rate": 4.4065853017905953e-07,
      "logits/chosen": 0.37400752305984497,
      "logits/rejected": 0.3305491805076599,
      "logps/chosen": -71.68441009521484,
      "logps/ref_chosen": -58.999759674072266,
      "logps/ref_rejected": -84.67575073242188,
      "logps/rejected": -103.70176696777344,
      "loss": 1.0722,
      "margin_dpo/margin_mean": 6.341360092163086,
      "margin_dpo/margin_std": 9.579675674438477,
      "step": 201
    },
    {
      "KL/chosen_KL_mean": -11.093196868896484,
      "KL/mean": -14.728687286376953,
      "KL/rejected_KL_mean": -18.364173889160156,
      "KL/std": 8.838526725769043,
      "epoch": 0.30536659108087677,
      "fcm_dpo/beta": 0.07887430489063263,
      "fcm_dpo/delta": -0.18363875150680542,
      "fcm_dpo/margin": 7.270977020263672,
      "fcm_dpo/q_t": 0.3720179796218872,
      "grad_norm": 22.277822494506836,
      "learning_rate": 4.3980061644943575e-07,
      "logits/chosen": 0.2807554602622986,
      "logits/rejected": 0.21378284692764282,
      "logps/chosen": -58.75384521484375,
      "logps/ref_chosen": -47.660648345947266,
      "logps/ref_rejected": -73.63249969482422,
      "logps/rejected": -91.99667358398438,
      "loss": 1.0233,
      "margin_dpo/margin_mean": 7.270977020263672,
      "margin_dpo/margin_std": 9.475584030151367,
      "step": 202
    },
    {
      "KL/chosen_KL_mean": -12.31509780883789,
      "KL/mean": -15.686324119567871,
      "KL/rejected_KL_mean": -19.057544708251953,
      "KL/std": 9.032432556152344,
      "epoch": 0.30687830687830686,
      "fcm_dpo/beta": 0.07658464461565018,
      "fcm_dpo/delta": -0.12252387404441833,
      "fcm_dpo/margin": 6.742450714111328,
      "fcm_dpo/q_t": 0.38822680711746216,
      "grad_norm": 25.437036514282227,
      "learning_rate": 4.3893739358856455e-07,
      "logits/chosen": 0.34383517503738403,
      "logits/rejected": 0.2762662172317505,
      "logps/chosen": -74.64063262939453,
      "logps/ref_chosen": -62.32553482055664,
      "logps/ref_rejected": -99.37226104736328,
      "logps/rejected": -118.4298095703125,
      "loss": 1.063,
      "margin_dpo/margin_mean": 6.742450714111328,
      "margin_dpo/margin_std": 9.920048713684082,
      "step": 203
    },
    {
      "KL/chosen_KL_mean": -12.230262756347656,
      "KL/mean": -15.652043342590332,
      "KL/rejected_KL_mean": -19.073822021484375,
      "KL/std": 9.692943572998047,
      "epoch": 0.30839002267573695,
      "fcm_dpo/beta": 0.07370705157518387,
      "fcm_dpo/delta": -0.11449645459651947,
      "fcm_dpo/margin": 6.843562126159668,
      "fcm_dpo/q_t": 0.3891705870628357,
      "grad_norm": 22.042787551879883,
      "learning_rate": 4.380688857426449e-07,
      "logits/chosen": 0.2842333912849426,
      "logits/rejected": 0.21995952725410461,
      "logps/chosen": -62.859580993652344,
      "logps/ref_chosen": -50.62931823730469,
      "logps/ref_rejected": -66.60475158691406,
      "logps/rejected": -85.67857360839844,
      "loss": 1.077,
      "margin_dpo/margin_mean": 6.843562602996826,
      "margin_dpo/margin_std": 9.982553482055664,
      "step": 204
    },
    {
      "KL/chosen_KL_mean": -13.222759246826172,
      "KL/mean": -16.707935333251953,
      "KL/rejected_KL_mean": -20.193115234375,
      "KL/std": 9.57691478729248,
      "epoch": 0.30990173847316704,
      "fcm_dpo/beta": 0.07268859446048737,
      "fcm_dpo/delta": -0.11301695555448532,
      "fcm_dpo/margin": 6.9703569412231445,
      "fcm_dpo/q_t": 0.39053136110305786,
      "grad_norm": 27.645692825317383,
      "learning_rate": 4.3719511720570814e-07,
      "logits/chosen": 0.3358631730079651,
      "logits/rejected": 0.2786390781402588,
      "logps/chosen": -83.57893371582031,
      "logps/ref_chosen": -70.3561782836914,
      "logps/ref_rejected": -93.39848327636719,
      "logps/rejected": -113.59159851074219,
      "loss": 1.0963,
      "margin_dpo/margin_mean": 6.9703569412231445,
      "margin_dpo/margin_std": 11.408649444580078,
      "step": 205
    },
    {
      "KL/chosen_KL_mean": -13.618179321289062,
      "KL/mean": -16.21625518798828,
      "KL/rejected_KL_mean": -18.814334869384766,
      "KL/std": 9.856538772583008,
      "epoch": 0.31141345427059713,
      "fcm_dpo/beta": 0.0731147825717926,
      "fcm_dpo/delta": 0.020249299705028534,
      "fcm_dpo/margin": 5.1961565017700195,
      "fcm_dpo/q_t": 0.4198933243751526,
      "grad_norm": 24.5611515045166,
      "learning_rate": 4.363161124189387e-07,
      "logits/chosen": 0.382521390914917,
      "logits/rejected": 0.36643290519714355,
      "logps/chosen": -81.26365661621094,
      "logps/ref_chosen": -67.64547729492188,
      "logps/ref_rejected": -79.89584350585938,
      "logps/rejected": -98.71017456054688,
      "loss": 1.2133,
      "margin_dpo/margin_mean": 5.196155548095703,
      "margin_dpo/margin_std": 11.782098770141602,
      "step": 206
    },
    {
      "KL/chosen_KL_mean": -14.933658599853516,
      "KL/mean": -18.79200553894043,
      "KL/rejected_KL_mean": -22.650352478027344,
      "KL/std": 10.600395202636719,
      "epoch": 0.3129251700680272,
      "fcm_dpo/beta": 0.07093209028244019,
      "fcm_dpo/delta": -0.1565505564212799,
      "fcm_dpo/margin": 7.716689586639404,
      "fcm_dpo/q_t": 0.38311976194381714,
      "grad_norm": 21.809812545776367,
      "learning_rate": 4.3543189596998986e-07,
      "logits/chosen": 0.2879766821861267,
      "logits/rejected": 0.2246440052986145,
      "logps/chosen": -82.59785461425781,
      "logps/ref_chosen": -67.66419219970703,
      "logps/ref_rejected": -85.10249328613281,
      "logps/rejected": -107.75283813476562,
      "loss": 1.0423,
      "margin_dpo/margin_mean": 7.716688632965088,
      "margin_dpo/margin_std": 10.914000511169434,
      "step": 207
    },
    {
      "KL/chosen_KL_mean": -12.039043426513672,
      "KL/mean": -14.267229080200195,
      "KL/rejected_KL_mean": -16.495407104492188,
      "KL/std": 9.71860122680664,
      "epoch": 0.3144368858654573,
      "fcm_dpo/beta": 0.07146717607975006,
      "fcm_dpo/delta": 0.08419809490442276,
      "fcm_dpo/margin": 4.456361770629883,
      "fcm_dpo/q_t": 0.43023359775543213,
      "grad_norm": 24.523517608642578,
      "learning_rate": 4.3454249259229664e-07,
      "logits/chosen": 0.31310391426086426,
      "logits/rejected": 0.28862231969833374,
      "logps/chosen": -69.770751953125,
      "logps/ref_chosen": -57.731712341308594,
      "logps/ref_rejected": -74.19276428222656,
      "logps/rejected": -90.68817138671875,
      "loss": 1.2268,
      "margin_dpo/margin_mean": 4.456361770629883,
      "margin_dpo/margin_std": 10.545194625854492,
      "step": 208
    },
    {
      "KL/chosen_KL_mean": -12.184415817260742,
      "KL/mean": -16.680660247802734,
      "KL/rejected_KL_mean": -21.176910400390625,
      "KL/std": 11.24008560180664,
      "epoch": 0.31594860166288735,
      "fcm_dpo/beta": 0.0693952813744545,
      "fcm_dpo/delta": -0.23861436545848846,
      "fcm_dpo/margin": 8.9924898147583,
      "fcm_dpo/q_t": 0.3661319613456726,
      "grad_norm": 24.027284622192383,
      "learning_rate": 4.336479271643833e-07,
      "logits/chosen": 0.28997889161109924,
      "logits/rejected": 0.23683319985866547,
      "logps/chosen": -80.7344970703125,
      "logps/ref_chosen": -68.55007934570312,
      "logps/ref_rejected": -87.90541076660156,
      "logps/rejected": -109.08232116699219,
      "loss": 1.0303,
      "margin_dpo/margin_mean": 8.992490768432617,
      "margin_dpo/margin_std": 12.858354568481445,
      "step": 209
    },
    {
      "KL/chosen_KL_mean": -12.258684158325195,
      "KL/mean": -16.604808807373047,
      "KL/rejected_KL_mean": -20.950939178466797,
      "KL/std": 11.92989730834961,
      "epoch": 0.31746031746031744,
      "fcm_dpo/beta": 0.06671866774559021,
      "fcm_dpo/delta": -0.19063086807727814,
      "fcm_dpo/margin": 8.692248344421387,
      "fcm_dpo/q_t": 0.37737345695495605,
      "grad_norm": 21.236785888671875,
      "learning_rate": 4.327482247091679e-07,
      "logits/chosen": 0.4041959047317505,
      "logits/rejected": 0.3123385012149811,
      "logps/chosen": -69.52696228027344,
      "logps/ref_chosen": -57.268272399902344,
      "logps/ref_rejected": -85.72807312011719,
      "logps/rejected": -106.67901611328125,
      "loss": 1.0505,
      "margin_dpo/margin_mean": 8.69224739074707,
      "margin_dpo/margin_std": 12.821832656860352,
      "step": 210
    },
    {
      "KL/chosen_KL_mean": -10.797460556030273,
      "KL/mean": -14.93586540222168,
      "KL/rejected_KL_mean": -19.07427978515625,
      "KL/std": 11.016217231750488,
      "epoch": 0.31897203325774753,
      "fcm_dpo/beta": 0.06479281187057495,
      "fcm_dpo/delta": -0.1437854766845703,
      "fcm_dpo/margin": 8.276815414428711,
      "fcm_dpo/q_t": 0.3854495584964752,
      "grad_norm": 23.455148696899414,
      "learning_rate": 4.3184341039326217e-07,
      "logits/chosen": 0.37784039974212646,
      "logits/rejected": 0.2924278676509857,
      "logps/chosen": -64.43817138671875,
      "logps/ref_chosen": -53.640708923339844,
      "logps/ref_rejected": -93.0387954711914,
      "logps/rejected": -112.11307525634766,
      "loss": 1.055,
      "margin_dpo/margin_mean": 8.276814460754395,
      "margin_dpo/margin_std": 11.989517211914062,
      "step": 211
    },
    {
      "KL/chosen_KL_mean": -12.629556655883789,
      "KL/mean": -17.281192779541016,
      "KL/rejected_KL_mean": -21.932838439941406,
      "KL/std": 11.690235137939453,
      "epoch": 0.3204837490551776,
      "fcm_dpo/beta": 0.06214489787817001,
      "fcm_dpo/delta": -0.18937504291534424,
      "fcm_dpo/margin": 9.303281784057617,
      "fcm_dpo/q_t": 0.37309345602989197,
      "grad_norm": 19.30466651916504,
      "learning_rate": 4.309335095262675e-07,
      "logits/chosen": 0.3716275095939636,
      "logits/rejected": 0.3046720027923584,
      "logps/chosen": -69.99629974365234,
      "logps/ref_chosen": -57.36674499511719,
      "logps/ref_rejected": -79.89643096923828,
      "logps/rejected": -101.82926940917969,
      "loss": 1.0357,
      "margin_dpo/margin_mean": 9.303279876708984,
      "margin_dpo/margin_std": 13.027929306030273,
      "step": 212
    },
    {
      "KL/chosen_KL_mean": -9.048931121826172,
      "KL/mean": -14.40077018737793,
      "KL/rejected_KL_mean": -19.752609252929688,
      "KL/std": 12.173477172851562,
      "epoch": 0.3219954648526077,
      "fcm_dpo/beta": 0.05909217149019241,
      "fcm_dpo/delta": -0.24991421401500702,
      "fcm_dpo/margin": 10.703678131103516,
      "fcm_dpo/q_t": 0.3626417815685272,
      "grad_norm": 17.545997619628906,
      "learning_rate": 4.3001854756006724e-07,
      "logits/chosen": 0.39142847061157227,
      "logits/rejected": 0.3677615523338318,
      "logps/chosen": -74.27005004882812,
      "logps/ref_chosen": -65.22111511230469,
      "logps/ref_rejected": -80.1810302734375,
      "logps/rejected": -99.93363952636719,
      "loss": 1.0005,
      "margin_dpo/margin_mean": 10.703678131103516,
      "margin_dpo/margin_std": 14.004202842712402,
      "step": 213
    },
    {
      "KL/chosen_KL_mean": -10.82396125793457,
      "KL/mean": -15.959489822387695,
      "KL/rejected_KL_mean": -21.095014572143555,
      "KL/std": 12.298513412475586,
      "epoch": 0.3235071806500378,
      "fcm_dpo/beta": 0.05724663287401199,
      "fcm_dpo/delta": -0.19926320016384125,
      "fcm_dpo/margin": 10.271051406860352,
      "fcm_dpo/q_t": 0.37353187799453735,
      "grad_norm": 21.21104621887207,
      "learning_rate": 4.290985500881143e-07,
      "logits/chosen": 0.2602912187576294,
      "logits/rejected": 0.23779809474945068,
      "logps/chosen": -72.11628723144531,
      "logps/ref_chosen": -61.292327880859375,
      "logps/ref_rejected": -67.69841003417969,
      "logps/rejected": -88.79342651367188,
      "loss": 1.0253,
      "margin_dpo/margin_mean": 10.271051406860352,
      "margin_dpo/margin_std": 13.87718391418457,
      "step": 214
    },
    {
      "KL/chosen_KL_mean": -13.394432067871094,
      "KL/mean": -18.717758178710938,
      "KL/rejected_KL_mean": -24.041080474853516,
      "KL/std": 12.819029808044434,
      "epoch": 0.3250188964474679,
      "fcm_dpo/beta": 0.055094048380851746,
      "fcm_dpo/delta": -0.19782397150993347,
      "fcm_dpo/margin": 10.646649360656738,
      "fcm_dpo/q_t": 0.37471532821655273,
      "grad_norm": 19.239511489868164,
      "learning_rate": 4.281735428447157e-07,
      "logits/chosen": 0.29680147767066956,
      "logits/rejected": 0.19680196046829224,
      "logps/chosen": -77.26356506347656,
      "logps/ref_chosen": -63.869136810302734,
      "logps/ref_rejected": -98.7657241821289,
      "logps/rejected": -122.80680847167969,
      "loss": 1.0339,
      "margin_dpo/margin_mean": 10.646649360656738,
      "margin_dpo/margin_std": 14.607285499572754,
      "step": 215
    },
    {
      "KL/chosen_KL_mean": -12.626960754394531,
      "KL/mean": -18.285587310791016,
      "KL/rejected_KL_mean": -23.9442138671875,
      "KL/std": 13.705946922302246,
      "epoch": 0.32653061224489793,
      "fcm_dpo/beta": 0.05256032198667526,
      "fcm_dpo/delta": -0.20697513222694397,
      "fcm_dpo/margin": 11.317251205444336,
      "fcm_dpo/q_t": 0.37171292304992676,
      "grad_norm": 21.164426803588867,
      "learning_rate": 4.2724355170431247e-07,
      "logits/chosen": 0.4230896234512329,
      "logits/rejected": 0.3399661183357239,
      "logps/chosen": -80.45191955566406,
      "logps/ref_chosen": -67.824951171875,
      "logps/ref_rejected": -96.40231323242188,
      "logps/rejected": -120.34652709960938,
      "loss": 1.013,
      "margin_dpo/margin_mean": 11.31725025177002,
      "margin_dpo/margin_std": 15.183269500732422,
      "step": 216
    },
    {
      "KL/chosen_KL_mean": -14.25343132019043,
      "KL/mean": -20.095951080322266,
      "KL/rejected_KL_mean": -25.938472747802734,
      "KL/std": 13.898407936096191,
      "epoch": 0.328042328042328,
      "fcm_dpo/beta": 0.05035218223929405,
      "fcm_dpo/delta": -0.2004796266555786,
      "fcm_dpo/margin": 11.685037612915039,
      "fcm_dpo/q_t": 0.37172871828079224,
      "grad_norm": 17.534391403198242,
      "learning_rate": 4.26308602680756e-07,
      "logits/chosen": 0.3500717282295227,
      "logits/rejected": 0.24914100766181946,
      "logps/chosen": -74.7584228515625,
      "logps/ref_chosen": -60.5049934387207,
      "logps/ref_rejected": -84.26618194580078,
      "logps/rejected": -110.20465087890625,
      "loss": 1.0056,
      "margin_dpo/margin_mean": 11.685037612915039,
      "margin_dpo/margin_std": 14.929855346679688,
      "step": 217
    },
    {
      "KL/chosen_KL_mean": -14.796497344970703,
      "KL/mean": -18.559711456298828,
      "KL/rejected_KL_mean": -22.322925567626953,
      "KL/std": 13.00196647644043,
      "epoch": 0.3295540438397581,
      "fcm_dpo/beta": 0.05051659792661667,
      "fcm_dpo/delta": 0.01861324906349182,
      "fcm_dpo/margin": 7.526429176330566,
      "fcm_dpo/q_t": 0.41731739044189453,
      "grad_norm": 20.19041633605957,
      "learning_rate": 4.253687219265803e-07,
      "logits/chosen": 0.2565411925315857,
      "logits/rejected": 0.25069519877433777,
      "logps/chosen": -85.39080810546875,
      "logps/ref_chosen": -70.59431457519531,
      "logps/ref_rejected": -73.89038848876953,
      "logps/rejected": -96.21331787109375,
      "loss": 1.2045,
      "margin_dpo/margin_mean": 7.526429176330566,
      "margin_dpo/margin_std": 16.088058471679688,
      "step": 218
    },
    {
      "KL/chosen_KL_mean": -13.562408447265625,
      "KL/mean": -17.571258544921875,
      "KL/rejected_KL_mean": -21.580114364624023,
      "KL/std": 14.031749725341797,
      "epoch": 0.3310657596371882,
      "fcm_dpo/beta": 0.04987781122326851,
      "fcm_dpo/delta": -0.0002720542252063751,
      "fcm_dpo/margin": 8.017708778381348,
      "fcm_dpo/q_t": 0.41187363862991333,
      "grad_norm": 19.11667823791504,
      "learning_rate": 4.2442393573227043e-07,
      "logits/chosen": 0.32280057668685913,
      "logits/rejected": 0.2829885184764862,
      "logps/chosen": -74.05335235595703,
      "logps/ref_chosen": -60.490943908691406,
      "logps/ref_rejected": -75.85001373291016,
      "logps/rejected": -97.43013000488281,
      "loss": 1.1407,
      "margin_dpo/margin_mean": 8.017709732055664,
      "margin_dpo/margin_std": 14.044523239135742,
      "step": 219
    },
    {
      "KL/chosen_KL_mean": -12.115495681762695,
      "KL/mean": -16.92727279663086,
      "KL/rejected_KL_mean": -21.739051818847656,
      "KL/std": 14.268013000488281,
      "epoch": 0.3325774754346183,
      "fcm_dpo/beta": 0.04924558848142624,
      "fcm_dpo/delta": -0.07858522981405258,
      "fcm_dpo/margin": 9.623554229736328,
      "fcm_dpo/q_t": 0.39898520708084106,
      "grad_norm": 16.392789840698242,
      "learning_rate": 4.234742705255272e-07,
      "logits/chosen": 0.4274730086326599,
      "logits/rejected": 0.36710894107818604,
      "logps/chosen": -57.1288948059082,
      "logps/ref_chosen": -45.013397216796875,
      "logps/ref_rejected": -70.49369812011719,
      "logps/rejected": -92.23275756835938,
      "loss": 1.1145,
      "margin_dpo/margin_mean": 9.623554229736328,
      "margin_dpo/margin_std": 16.26483154296875,
      "step": 220
    },
    {
      "KL/chosen_KL_mean": -11.787042617797852,
      "KL/mean": -17.021156311035156,
      "KL/rejected_KL_mean": -22.255271911621094,
      "KL/std": 14.119913101196289,
      "epoch": 0.3340891912320484,
      "fcm_dpo/beta": 0.04853355139493942,
      "fcm_dpo/delta": -0.11374709010124207,
      "fcm_dpo/margin": 10.468233108520508,
      "fcm_dpo/q_t": 0.39125609397888184,
      "grad_norm": 17.957706451416016,
      "learning_rate": 4.22519752870528e-07,
      "logits/chosen": 0.3971208333969116,
      "logits/rejected": 0.3293982148170471,
      "logps/chosen": -70.88288116455078,
      "logps/ref_chosen": -59.09584045410156,
      "logps/ref_rejected": -88.64388275146484,
      "logps/rejected": -110.89915466308594,
      "loss": 1.0807,
      "margin_dpo/margin_mean": 10.468233108520508,
      "margin_dpo/margin_std": 16.371990203857422,
      "step": 221
    },
    {
      "KL/chosen_KL_mean": -12.735214233398438,
      "KL/mean": -19.463855743408203,
      "KL/rejected_KL_mean": -26.19249725341797,
      "KL/std": 15.829582214355469,
      "epoch": 0.3356009070294785,
      "fcm_dpo/beta": 0.046448659151792526,
      "fcm_dpo/delta": -0.24044792354106903,
      "fcm_dpo/margin": 13.45728874206543,
      "fcm_dpo/q_t": 0.36255699396133423,
      "grad_norm": 18.125226974487305,
      "learning_rate": 4.2156040946718343e-07,
      "logits/chosen": 0.4456883668899536,
      "logits/rejected": 0.36282992362976074,
      "logps/chosen": -68.73291015625,
      "logps/ref_chosen": -55.9976921081543,
      "logps/ref_rejected": -111.94727325439453,
      "logps/rejected": -138.1397705078125,
      "loss": 0.9962,
      "margin_dpo/margin_mean": 13.457290649414062,
      "margin_dpo/margin_std": 16.974525451660156,
      "step": 222
    },
    {
      "KL/chosen_KL_mean": -14.5010986328125,
      "KL/mean": -20.88688850402832,
      "KL/rejected_KL_mean": -27.272682189941406,
      "KL/std": 15.840429306030273,
      "epoch": 0.3371126228269085,
      "fcm_dpo/beta": 0.04450047016143799,
      "fcm_dpo/delta": -0.1799127459526062,
      "fcm_dpo/margin": 12.771581649780273,
      "fcm_dpo/q_t": 0.37460440397262573,
      "grad_norm": 16.625930786132812,
      "learning_rate": 4.2059626715039065e-07,
      "logits/chosen": 0.4114035665988922,
      "logits/rejected": 0.357668936252594,
      "logps/chosen": -74.39251708984375,
      "logps/ref_chosen": -59.891422271728516,
      "logps/ref_rejected": -86.28954315185547,
      "logps/rejected": -113.56222534179688,
      "loss": 1.0008,
      "margin_dpo/margin_mean": 12.771581649780273,
      "margin_dpo/margin_std": 15.203545570373535,
      "step": 223
    },
    {
      "KL/chosen_KL_mean": -16.549915313720703,
      "KL/mean": -20.222265243530273,
      "KL/rejected_KL_mean": -23.894615173339844,
      "KL/std": 15.480034828186035,
      "epoch": 0.3386243386243386,
      "fcm_dpo/beta": 0.04471778869628906,
      "fcm_dpo/delta": 0.07405854016542435,
      "fcm_dpo/margin": 7.344696998596191,
      "fcm_dpo/q_t": 0.42939624190330505,
      "grad_norm": 21.553468704223633,
      "learning_rate": 4.1962735288928304e-07,
      "logits/chosen": 0.463707834482193,
      "logits/rejected": 0.44336050748825073,
      "logps/chosen": -80.59455108642578,
      "logps/ref_chosen": -64.04463195800781,
      "logps/ref_rejected": -75.05450439453125,
      "logps/rejected": -98.94911193847656,
      "loss": 1.1979,
      "margin_dpo/margin_mean": 7.344696998596191,
      "margin_dpo/margin_std": 15.690269470214844,
      "step": 224
    },
    {
      "KL/chosen_KL_mean": -15.249135971069336,
      "KL/mean": -21.75031280517578,
      "KL/rejected_KL_mean": -28.25149154663086,
      "KL/std": 17.294654846191406,
      "epoch": 0.3401360544217687,
      "fcm_dpo/beta": 0.04361790418624878,
      "fcm_dpo/delta": -0.17827551066875458,
      "fcm_dpo/margin": 13.002354621887207,
      "fcm_dpo/q_t": 0.37681907415390015,
      "grad_norm": 18.24221420288086,
      "learning_rate": 4.186536937864752e-07,
      "logits/chosen": 0.4368096590042114,
      "logits/rejected": 0.32517528533935547,
      "logps/chosen": -81.34495544433594,
      "logps/ref_chosen": -66.0958251953125,
      "logps/ref_rejected": -97.68675231933594,
      "logps/rejected": -125.93824005126953,
      "loss": 1.0228,
      "margin_dpo/margin_mean": 13.002354621887207,
      "margin_dpo/margin_std": 17.1579647064209,
      "step": 225
    },
    {
      "KL/chosen_KL_mean": -14.37647819519043,
      "KL/mean": -19.67775535583496,
      "KL/rejected_KL_mean": -24.97903060913086,
      "KL/std": 16.739892959594727,
      "epoch": 0.3416477702191988,
      "fcm_dpo/beta": 0.04320550337433815,
      "fcm_dpo/delta": -0.06102943420410156,
      "fcm_dpo/margin": 10.602548599243164,
      "fcm_dpo/q_t": 0.4007114768028259,
      "grad_norm": 15.894902229309082,
      "learning_rate": 4.176753170773052e-07,
      "logits/chosen": 0.49875736236572266,
      "logits/rejected": 0.45056477189064026,
      "logps/chosen": -65.79335021972656,
      "logps/ref_chosen": -51.4168701171875,
      "logps/ref_rejected": -66.30068969726562,
      "logps/rejected": -91.27971649169922,
      "loss": 1.1381,
      "margin_dpo/margin_mean": 10.602548599243164,
      "margin_dpo/margin_std": 19.16706085205078,
      "step": 226
    },
    {
      "KL/chosen_KL_mean": -15.519994735717773,
      "KL/mean": -21.301816940307617,
      "KL/rejected_KL_mean": -27.083641052246094,
      "KL/std": 17.973379135131836,
      "epoch": 0.3431594860166289,
      "fcm_dpo/beta": 0.042112961411476135,
      "fcm_dpo/delta": -0.09216802567243576,
      "fcm_dpo/margin": 11.563647270202637,
      "fcm_dpo/q_t": 0.3971030116081238,
      "grad_norm": 17.23468589782715,
      "learning_rate": 4.166922501290729e-07,
      "logits/chosen": 0.48868125677108765,
      "logits/rejected": 0.4502916932106018,
      "logps/chosen": -73.50977325439453,
      "logps/ref_chosen": -57.989776611328125,
      "logps/ref_rejected": -75.05464172363281,
      "logps/rejected": -102.13829040527344,
      "loss": 1.117,
      "margin_dpo/margin_mean": 11.563648223876953,
      "margin_dpo/margin_std": 19.933834075927734,
      "step": 227
    },
    {
      "KL/chosen_KL_mean": -17.620210647583008,
      "KL/mean": -23.31317901611328,
      "KL/rejected_KL_mean": -29.006141662597656,
      "KL/std": 17.29057502746582,
      "epoch": 0.34467120181405897,
      "fcm_dpo/beta": 0.0417354553937912,
      "fcm_dpo/delta": -0.07891019433736801,
      "fcm_dpo/margin": 11.385927200317383,
      "fcm_dpo/q_t": 0.3964489698410034,
      "grad_norm": 16.280597686767578,
      "learning_rate": 4.1570452044027405e-07,
      "logits/chosen": 0.46561571955680847,
      "logits/rejected": 0.3897852301597595,
      "logps/chosen": -73.17958068847656,
      "logps/ref_chosen": -55.55936813354492,
      "logps/ref_rejected": -77.02364349365234,
      "logps/rejected": -106.02978515625,
      "loss": 1.0928,
      "margin_dpo/margin_mean": 11.385927200317383,
      "margin_dpo/margin_std": 17.77695083618164,
      "step": 228
    },
    {
      "KL/chosen_KL_mean": -13.526348114013672,
      "KL/mean": -19.22211265563965,
      "KL/rejected_KL_mean": -24.917877197265625,
      "KL/std": 16.780139923095703,
      "epoch": 0.34618291761148906,
      "fcm_dpo/beta": 0.04094117134809494,
      "fcm_dpo/delta": -0.06977853924036026,
      "fcm_dpo/margin": 11.39152717590332,
      "fcm_dpo/q_t": 0.3969324827194214,
      "grad_norm": 33.38159942626953,
      "learning_rate": 4.147121556398312e-07,
      "logits/chosen": 0.5544859170913696,
      "logits/rejected": 0.4903777837753296,
      "logps/chosen": -64.32101440429688,
      "logps/ref_chosen": -50.79466247558594,
      "logps/ref_rejected": -78.4474105834961,
      "logps/rejected": -103.36528778076172,
      "loss": 1.1309,
      "margin_dpo/margin_mean": 11.39152717590332,
      "margin_dpo/margin_std": 20.368953704833984,
      "step": 229
    },
    {
      "KL/chosen_KL_mean": -15.6832275390625,
      "KL/mean": -22.08908462524414,
      "KL/rejected_KL_mean": -28.49494171142578,
      "KL/std": 18.338987350463867,
      "epoch": 0.3476946334089191,
      "fcm_dpo/beta": 0.040487367659807205,
      "fcm_dpo/delta": -0.12587293982505798,
      "fcm_dpo/margin": 12.811714172363281,
      "fcm_dpo/q_t": 0.38656824827194214,
      "grad_norm": 16.915529251098633,
      "learning_rate": 4.137151834863213e-07,
      "logits/chosen": 0.46070948243141174,
      "logits/rejected": 0.4588784873485565,
      "logps/chosen": -72.4124526977539,
      "logps/ref_chosen": -56.729225158691406,
      "logps/ref_rejected": -62.99180603027344,
      "logps/rejected": -91.48674774169922,
      "loss": 1.0587,
      "margin_dpo/margin_mean": 12.811714172363281,
      "margin_dpo/margin_std": 17.839630126953125,
      "step": 230
    },
    {
      "KL/chosen_KL_mean": -18.668453216552734,
      "KL/mean": -27.829383850097656,
      "KL/rejected_KL_mean": -36.990318298339844,
      "KL/std": 18.366313934326172,
      "epoch": 0.3492063492063492,
      "fcm_dpo/beta": 0.037801600992679596,
      "fcm_dpo/delta": -0.3168046176433563,
      "fcm_dpo/margin": 18.32186508178711,
      "fcm_dpo/q_t": 0.34421560168266296,
      "grad_norm": 17.75885581970215,
      "learning_rate": 4.1271363186719835e-07,
      "logits/chosen": 0.3724779784679413,
      "logits/rejected": 0.35940617322921753,
      "logps/chosen": -91.26554870605469,
      "logps/ref_chosen": -72.59709930419922,
      "logps/ref_rejected": -86.2322998046875,
      "logps/rejected": -123.22261810302734,
      "loss": 0.9318,
      "margin_dpo/margin_mean": 18.32186508178711,
      "margin_dpo/margin_std": 19.380470275878906,
      "step": 231
    },
    {
      "KL/chosen_KL_mean": -18.254302978515625,
      "KL/mean": -24.430221557617188,
      "KL/rejected_KL_mean": -30.606136322021484,
      "KL/std": 19.692710876464844,
      "epoch": 0.3507180650037793,
      "fcm_dpo/beta": 0.03702760860323906,
      "fcm_dpo/delta": -0.06017792224884033,
      "fcm_dpo/margin": 12.351831436157227,
      "fcm_dpo/q_t": 0.40328365564346313,
      "grad_norm": 17.288766860961914,
      "learning_rate": 4.1170752879801436e-07,
      "logits/chosen": 0.4320966601371765,
      "logits/rejected": 0.4029006063938141,
      "logps/chosen": -86.37283325195312,
      "logps/ref_chosen": -68.1185302734375,
      "logps/ref_rejected": -83.79415893554688,
      "logps/rejected": -114.40029907226562,
      "loss": 1.1292,
      "margin_dpo/margin_mean": 12.351832389831543,
      "margin_dpo/margin_std": 22.07088851928711,
      "step": 232
    },
    {
      "KL/chosen_KL_mean": -21.45623207092285,
      "KL/mean": -26.537189483642578,
      "KL/rejected_KL_mean": -31.618144989013672,
      "KL/std": 19.625057220458984,
      "epoch": 0.35222978080120937,
      "fcm_dpo/beta": 0.03667011857032776,
      "fcm_dpo/delta": 0.026034735143184662,
      "fcm_dpo/margin": 10.16191291809082,
      "fcm_dpo/q_t": 0.41923320293426514,
      "grad_norm": 16.044673919677734,
      "learning_rate": 4.106969024216348e-07,
      "logits/chosen": 0.47052958607673645,
      "logits/rejected": 0.4147252142429352,
      "logps/chosen": -76.52638244628906,
      "logps/ref_chosen": -55.070152282714844,
      "logps/ref_rejected": -66.61845397949219,
      "logps/rejected": -98.2365951538086,
      "loss": 1.1726,
      "margin_dpo/margin_mean": 10.16191291809082,
      "margin_dpo/margin_std": 18.987119674682617,
      "step": 233
    },
    {
      "KL/chosen_KL_mean": -19.42575454711914,
      "KL/mean": -24.680850982666016,
      "KL/rejected_KL_mean": -29.93594741821289,
      "KL/std": 18.938419342041016,
      "epoch": 0.35374149659863946,
      "fcm_dpo/beta": 0.0376385897397995,
      "fcm_dpo/delta": 0.0010745078325271606,
      "fcm_dpo/margin": 10.51019287109375,
      "fcm_dpo/q_t": 0.41605016589164734,
      "grad_norm": 19.10257339477539,
      "learning_rate": 4.09681781007452e-07,
      "logits/chosen": 0.4139387309551239,
      "logits/rejected": 0.4022018313407898,
      "logps/chosen": -75.35165405273438,
      "logps/ref_chosen": -55.92589569091797,
      "logps/ref_rejected": -51.11608123779297,
      "logps/rejected": -81.05203247070312,
      "loss": 1.1893,
      "margin_dpo/margin_mean": 10.51019287109375,
      "margin_dpo/margin_std": 20.87730598449707,
      "step": 234
    },
    {
      "KL/chosen_KL_mean": -15.336202621459961,
      "KL/mean": -24.05404281616211,
      "KL/rejected_KL_mean": -32.771888732910156,
      "KL/std": 20.12569808959961,
      "epoch": 0.35525321239606955,
      "fcm_dpo/beta": 0.03593028336763382,
      "fcm_dpo/delta": -0.24115046858787537,
      "fcm_dpo/margin": 17.435686111450195,
      "fcm_dpo/q_t": 0.3593088388442993,
      "grad_norm": 16.729801177978516,
      "learning_rate": 4.08662192950594e-07,
      "logits/chosen": 0.5166130661964417,
      "logits/rejected": 0.4991014003753662,
      "logps/chosen": -79.87593078613281,
      "logps/ref_chosen": -64.53972625732422,
      "logps/ref_rejected": -77.69151306152344,
      "logps/rejected": -110.46339416503906,
      "loss": 0.9662,
      "margin_dpo/margin_mean": 17.435686111450195,
      "margin_dpo/margin_std": 19.512123107910156,
      "step": 235
    },
    {
      "KL/chosen_KL_mean": -25.423974990844727,
      "KL/mean": -32.12845993041992,
      "KL/rejected_KL_mean": -38.83293914794922,
      "KL/std": 20.33449935913086,
      "epoch": 0.35676492819349964,
      "fcm_dpo/beta": 0.03493765741586685,
      "fcm_dpo/delta": -0.07212943583726883,
      "fcm_dpo/margin": 13.408965110778809,
      "fcm_dpo/q_t": 0.39774391055107117,
      "grad_norm": 16.289026260375977,
      "learning_rate": 4.076381667711306e-07,
      "logits/chosen": 0.4709205627441406,
      "logits/rejected": 0.457000195980072,
      "logps/chosen": -96.5787124633789,
      "logps/ref_chosen": -71.15473937988281,
      "logps/ref_rejected": -84.88541412353516,
      "logps/rejected": -123.71835327148438,
      "loss": 1.1192,
      "margin_dpo/margin_mean": 13.408964157104492,
      "margin_dpo/margin_std": 22.929529190063477,
      "step": 236
    },
    {
      "KL/chosen_KL_mean": -22.029151916503906,
      "KL/mean": -29.681533813476562,
      "KL/rejected_KL_mean": -37.33391189575195,
      "KL/std": 20.521644592285156,
      "epoch": 0.35827664399092973,
      "fcm_dpo/beta": 0.034239333122968674,
      "fcm_dpo/delta": -0.13081705570220947,
      "fcm_dpo/margin": 15.304759979248047,
      "fcm_dpo/q_t": 0.3851306438446045,
      "grad_norm": 18.769914627075195,
      "learning_rate": 4.066097311132753e-07,
      "logits/chosen": 0.5257338881492615,
      "logits/rejected": 0.5131624341011047,
      "logps/chosen": -98.1711654663086,
      "logps/ref_chosen": -76.14201354980469,
      "logps/ref_rejected": -80.88479614257812,
      "logps/rejected": -118.21870422363281,
      "loss": 1.0769,
      "margin_dpo/margin_mean": 15.304759979248047,
      "margin_dpo/margin_std": 23.187271118164062,
      "step": 237
    },
    {
      "KL/chosen_KL_mean": -19.780475616455078,
      "KL/mean": -27.63768768310547,
      "KL/rejected_KL_mean": -35.49489974975586,
      "KL/std": 22.138229370117188,
      "epoch": 0.35978835978835977,
      "fcm_dpo/beta": 0.033210694789886475,
      "fcm_dpo/delta": -0.12943503260612488,
      "fcm_dpo/margin": 15.714425086975098,
      "fcm_dpo/q_t": 0.3854626417160034,
      "grad_norm": 25.83220672607422,
      "learning_rate": 4.0557691474458414e-07,
      "logits/chosen": 0.4778020977973938,
      "logits/rejected": 0.4644400179386139,
      "logps/chosen": -88.66532897949219,
      "logps/ref_chosen": -68.88484954833984,
      "logps/ref_rejected": -75.8946304321289,
      "logps/rejected": -111.3895263671875,
      "loss": 1.0601,
      "margin_dpo/margin_mean": 15.714425086975098,
      "margin_dpo/margin_std": 22.588531494140625,
      "step": 238
    },
    {
      "KL/chosen_KL_mean": -24.469463348388672,
      "KL/mean": -32.38288116455078,
      "KL/rejected_KL_mean": -40.296295166015625,
      "KL/std": 23.43410873413086,
      "epoch": 0.36130007558578986,
      "fcm_dpo/beta": 0.03276742249727249,
      "fcm_dpo/delta": -0.12536010146141052,
      "fcm_dpo/margin": 15.826839447021484,
      "fcm_dpo/q_t": 0.3873627185821533,
      "grad_norm": 17.71445655822754,
      "learning_rate": 4.045397465551513e-07,
      "logits/chosen": 0.6087825894355774,
      "logits/rejected": 0.4834781587123871,
      "logps/chosen": -81.24128723144531,
      "logps/ref_chosen": -56.771827697753906,
      "logps/ref_rejected": -116.23050689697266,
      "logps/rejected": -156.52679443359375,
      "loss": 1.0763,
      "margin_dpo/margin_mean": 15.826841354370117,
      "margin_dpo/margin_std": 23.740345001220703,
      "step": 239
    },
    {
      "KL/chosen_KL_mean": -23.602027893066406,
      "KL/mean": -33.94664764404297,
      "KL/rejected_KL_mean": -44.29126739501953,
      "KL/std": 23.732091903686523,
      "epoch": 0.36281179138321995,
      "fcm_dpo/beta": 0.03111092746257782,
      "fcm_dpo/delta": -0.2601979374885559,
      "fcm_dpo/margin": 20.689245223999023,
      "fcm_dpo/q_t": 0.3577980101108551,
      "grad_norm": 13.626633644104004,
      "learning_rate": 4.0349825555680045e-07,
      "logits/chosen": 0.5065616369247437,
      "logits/rejected": 0.4162125587463379,
      "logps/chosen": -76.95613861083984,
      "logps/ref_chosen": -53.35411071777344,
      "logps/ref_rejected": -80.12019348144531,
      "logps/rejected": -124.41146850585938,
      "loss": 0.9648,
      "margin_dpo/margin_mean": 20.689245223999023,
      "margin_dpo/margin_std": 23.720182418823242,
      "step": 240
    },
    {
      "KL/chosen_KL_mean": -23.950889587402344,
      "KL/mean": -30.622859954833984,
      "KL/rejected_KL_mean": -37.29482650756836,
      "KL/std": 22.645187377929688,
      "epoch": 0.36432350718065004,
      "fcm_dpo/beta": 0.03072282299399376,
      "fcm_dpo/delta": -0.010602390393614769,
      "fcm_dpo/margin": 13.343942642211914,
      "fcm_dpo/q_t": 0.4106614291667938,
      "grad_norm": 17.857189178466797,
      "learning_rate": 4.0245247088227377e-07,
      "logits/chosen": 0.4656754732131958,
      "logits/rejected": 0.43102699518203735,
      "logps/chosen": -95.84629821777344,
      "logps/ref_chosen": -71.89541625976562,
      "logps/ref_rejected": -83.03492736816406,
      "logps/rejected": -120.32975769042969,
      "loss": 1.142,
      "margin_dpo/margin_mean": 13.343942642211914,
      "margin_dpo/margin_std": 23.82295799255371,
      "step": 241
    },
    {
      "KL/chosen_KL_mean": -24.637340545654297,
      "KL/mean": -34.245872497558594,
      "KL/rejected_KL_mean": -43.854408264160156,
      "KL/std": 24.97643280029297,
      "epoch": 0.36583522297808013,
      "fcm_dpo/beta": 0.029523026198148727,
      "fcm_dpo/delta": -0.18067717552185059,
      "fcm_dpo/margin": 19.217071533203125,
      "fcm_dpo/q_t": 0.3758378028869629,
      "grad_norm": 13.497591972351074,
      "learning_rate": 4.0140242178441665e-07,
      "logits/chosen": 0.46095046401023865,
      "logits/rejected": 0.4407215714454651,
      "logps/chosen": -82.56477355957031,
      "logps/ref_chosen": -57.927433013916016,
      "logps/ref_rejected": -67.838623046875,
      "logps/rejected": -111.69302368164062,
      "loss": 1.03,
      "margin_dpo/margin_mean": 19.217071533203125,
      "margin_dpo/margin_std": 25.75708770751953,
      "step": 242
    },
    {
      "KL/chosen_KL_mean": -25.040569305419922,
      "KL/mean": -33.38526916503906,
      "KL/rejected_KL_mean": -41.72996139526367,
      "KL/std": 24.873241424560547,
      "epoch": 0.3673469387755102,
      "fcm_dpo/beta": 0.0291166752576828,
      "fcm_dpo/delta": -0.09033676981925964,
      "fcm_dpo/margin": 16.68939208984375,
      "fcm_dpo/q_t": 0.39298439025878906,
      "grad_norm": 17.021177291870117,
      "learning_rate": 4.003481376353596e-07,
      "logits/chosen": 0.5095110535621643,
      "logits/rejected": 0.5118368864059448,
      "logps/chosen": -99.31724548339844,
      "logps/ref_chosen": -74.27667236328125,
      "logps/ref_rejected": -73.24340057373047,
      "logps/rejected": -114.97335815429688,
      "loss": 1.0805,
      "margin_dpo/margin_mean": 16.68939208984375,
      "margin_dpo/margin_std": 25.188854217529297,
      "step": 243
    },
    {
      "KL/chosen_KL_mean": -25.32716178894043,
      "KL/mean": -36.57783126831055,
      "KL/rejected_KL_mean": -47.82850646972656,
      "KL/std": 24.990251541137695,
      "epoch": 0.3688586545729403,
      "fcm_dpo/beta": 0.028054017573595047,
      "fcm_dpo/delta": -0.24619925022125244,
      "fcm_dpo/margin": 22.501346588134766,
      "fcm_dpo/q_t": 0.3596031367778778,
      "grad_norm": 15.71311092376709,
      "learning_rate": 3.9928964792569654e-07,
      "logits/chosen": 0.5257828831672668,
      "logits/rejected": 0.44374197721481323,
      "logps/chosen": -78.69107055664062,
      "logps/ref_chosen": -53.36390686035156,
      "logps/ref_rejected": -71.10276794433594,
      "logps/rejected": -118.9312744140625,
      "loss": 0.9604,
      "margin_dpo/margin_mean": 22.501344680786133,
      "margin_dpo/margin_std": 24.605850219726562,
      "step": 244
    },
    {
      "KL/chosen_KL_mean": -27.530160903930664,
      "KL/mean": -40.439083099365234,
      "KL/rejected_KL_mean": -53.34800720214844,
      "KL/std": 26.418277740478516,
      "epoch": 0.37037037037037035,
      "fcm_dpo/beta": 0.02635239064693451,
      "fcm_dpo/delta": -0.3012879490852356,
      "fcm_dpo/margin": 25.81784439086914,
      "fcm_dpo/q_t": 0.3481583893299103,
      "grad_norm": 20.990034103393555,
      "learning_rate": 3.982269822636601e-07,
      "logits/chosen": 0.5564769506454468,
      "logits/rejected": 0.5310531854629517,
      "logps/chosen": -98.72526550292969,
      "logps/ref_chosen": -71.19510650634766,
      "logps/ref_rejected": -80.76235961914062,
      "logps/rejected": -134.11036682128906,
      "loss": 0.9256,
      "margin_dpo/margin_mean": 25.817842483520508,
      "margin_dpo/margin_std": 26.52771759033203,
      "step": 245
    },
    {
      "KL/chosen_KL_mean": -33.67304229736328,
      "KL/mean": -44.59162521362305,
      "KL/rejected_KL_mean": -55.51020431518555,
      "KL/std": 27.117572784423828,
      "epoch": 0.37188208616780044,
      "fcm_dpo/beta": 0.02548816055059433,
      "fcm_dpo/delta": -0.1655275523662567,
      "fcm_dpo/margin": 21.837169647216797,
      "fcm_dpo/q_t": 0.37953728437423706,
      "grad_norm": 17.551103591918945,
      "learning_rate": 3.971601703742932e-07,
      "logits/chosen": 0.5693104863166809,
      "logits/rejected": 0.5100914239883423,
      "logps/chosen": -105.2940902709961,
      "logps/ref_chosen": -71.62104797363281,
      "logps/ref_rejected": -94.03392028808594,
      "logps/rejected": -149.54412841796875,
      "loss": 1.0613,
      "margin_dpo/margin_mean": 21.837169647216797,
      "margin_dpo/margin_std": 32.29937744140625,
      "step": 246
    },
    {
      "KL/chosen_KL_mean": -38.533203125,
      "KL/mean": -44.415794372558594,
      "KL/rejected_KL_mean": -50.29838562011719,
      "KL/std": 27.172008514404297,
      "epoch": 0.37339380196523053,
      "fcm_dpo/beta": 0.025636455044150352,
      "fcm_dpo/delta": 0.10047095268964767,
      "fcm_dpo/margin": 11.765192985534668,
      "fcm_dpo/q_t": 0.4332590401172638,
      "grad_norm": 18.015737533569336,
      "learning_rate": 3.960892420986177e-07,
      "logits/chosen": 0.5601674318313599,
      "logits/rejected": 0.5501619577407837,
      "logps/chosen": -118.55574798583984,
      "logps/ref_chosen": -80.02254486083984,
      "logps/ref_rejected": -89.22705841064453,
      "logps/rejected": -139.52545166015625,
      "loss": 1.2296,
      "margin_dpo/margin_mean": 11.765192985534668,
      "margin_dpo/margin_std": 27.990760803222656,
      "step": 247
    },
    {
      "KL/chosen_KL_mean": -32.911590576171875,
      "KL/mean": -44.313560485839844,
      "KL/rejected_KL_mean": -55.71553421020508,
      "KL/std": 30.605388641357422,
      "epoch": 0.3749055177626606,
      "fcm_dpo/beta": 0.02507779374718666,
      "fcm_dpo/delta": -0.1818971335887909,
      "fcm_dpo/margin": 22.803951263427734,
      "fcm_dpo/q_t": 0.3799017667770386,
      "grad_norm": 15.453660011291504,
      "learning_rate": 3.9501422739279953e-07,
      "logits/chosen": 0.5543103218078613,
      "logits/rejected": 0.5999008417129517,
      "logps/chosen": -98.28955078125,
      "logps/ref_chosen": -65.37796020507812,
      "logps/ref_rejected": -61.365787506103516,
      "logps/rejected": -117.0813217163086,
      "loss": 1.0584,
      "margin_dpo/margin_mean": 22.803951263427734,
      "margin_dpo/margin_std": 34.07537841796875,
      "step": 248
    },
    {
      "KL/chosen_KL_mean": -43.59640121459961,
      "KL/mean": -48.266441345214844,
      "KL/rejected_KL_mean": -52.936485290527344,
      "KL/std": 29.317392349243164,
      "epoch": 0.3764172335600907,
      "fcm_dpo/beta": 0.025514479726552963,
      "fcm_dpo/delta": 0.16352643072605133,
      "fcm_dpo/margin": 9.340082168579102,
      "fcm_dpo/q_t": 0.447033166885376,
      "grad_norm": 22.10836410522461,
      "learning_rate": 3.9393515632731094e-07,
      "logits/chosen": 0.546309232711792,
      "logits/rejected": 0.582387387752533,
      "logps/chosen": -118.19786071777344,
      "logps/ref_chosen": -74.60145568847656,
      "logps/ref_rejected": -63.79338455200195,
      "logps/rejected": -116.72987365722656,
      "loss": 1.3176,
      "margin_dpo/margin_mean": 9.340083122253418,
      "margin_dpo/margin_std": 31.282310485839844,
      "step": 249
    },
    {
      "KL/chosen_KL_mean": -37.384254455566406,
      "KL/mean": -49.02488708496094,
      "KL/rejected_KL_mean": -60.66551971435547,
      "KL/std": 30.749019622802734,
      "epoch": 0.3779289493575208,
      "fcm_dpo/beta": 0.02490631863474846,
      "fcm_dpo/delta": -0.19055120646953583,
      "fcm_dpo/margin": 23.281265258789062,
      "fcm_dpo/q_t": 0.37171927094459534,
      "grad_norm": 15.915420532226562,
      "learning_rate": 3.9285205908608934e-07,
      "logits/chosen": 0.6429681181907654,
      "logits/rejected": 0.6006037592887878,
      "logps/chosen": -99.32246398925781,
      "logps/ref_chosen": -61.938209533691406,
      "logps/ref_rejected": -72.21602630615234,
      "logps/rejected": -132.8815460205078,
      "loss": 1.0288,
      "margin_dpo/margin_mean": 23.281265258789062,
      "margin_dpo/margin_std": 31.80821418762207,
      "step": 250
    },
    {
      "KL/chosen_KL_mean": -42.976444244384766,
      "KL/mean": -50.41681671142578,
      "KL/rejected_KL_mean": -57.85719680786133,
      "KL/std": 29.169586181640625,
      "epoch": 0.3794406651549509,
      "fcm_dpo/beta": 0.024686751887202263,
      "fcm_dpo/delta": 0.03389812633395195,
      "fcm_dpo/margin": 14.880743980407715,
      "fcm_dpo/q_t": 0.4203200936317444,
      "grad_norm": 23.5372371673584,
      "learning_rate": 3.9176496596569265e-07,
      "logits/chosen": 0.6521850824356079,
      "logits/rejected": 0.6113159656524658,
      "logps/chosen": -109.83338928222656,
      "logps/ref_chosen": -66.85694885253906,
      "logps/ref_rejected": -84.83396911621094,
      "logps/rejected": -142.691162109375,
      "loss": 1.2014,
      "margin_dpo/margin_mean": 14.880743026733398,
      "margin_dpo/margin_std": 32.66436004638672,
      "step": 251
    },
    {
      "KL/chosen_KL_mean": -39.951820373535156,
      "KL/mean": -47.22810363769531,
      "KL/rejected_KL_mean": -54.50439453125,
      "KL/std": 31.698406219482422,
      "epoch": 0.38095238095238093,
      "fcm_dpo/beta": 0.025234483182430267,
      "fcm_dpo/delta": 0.02930210530757904,
      "fcm_dpo/margin": 14.55257511138916,
      "fcm_dpo/q_t": 0.42153429985046387,
      "grad_norm": 27.598159790039062,
      "learning_rate": 3.9067390737445254e-07,
      "logits/chosen": 0.5371665954589844,
      "logits/rejected": 0.4842592477798462,
      "logps/chosen": -96.17575073242188,
      "logps/ref_chosen": -56.22393035888672,
      "logps/ref_rejected": -77.1136245727539,
      "logps/rejected": -131.61801147460938,
      "loss": 1.2388,
      "margin_dpo/margin_mean": 14.552577018737793,
      "margin_dpo/margin_std": 33.91108703613281,
      "step": 252
    },
    {
      "KL/chosen_KL_mean": -42.02704620361328,
      "KL/mean": -50.181373596191406,
      "KL/rejected_KL_mean": -58.3357048034668,
      "KL/std": 31.003231048583984,
      "epoch": 0.382464096749811,
      "fcm_dpo/beta": 0.024656172841787338,
      "fcm_dpo/delta": -0.004082784056663513,
      "fcm_dpo/margin": 16.308658599853516,
      "fcm_dpo/q_t": 0.4137699604034424,
      "grad_norm": 19.795211791992188,
      "learning_rate": 3.8957891383162304e-07,
      "logits/chosen": 0.618567705154419,
      "logits/rejected": 0.5784329175949097,
      "logps/chosen": -94.237060546875,
      "logps/ref_chosen": -52.21001434326172,
      "logps/ref_rejected": -58.75764846801758,
      "logps/rejected": -117.09335327148438,
      "loss": 1.1667,
      "margin_dpo/margin_mean": 16.308656692504883,
      "margin_dpo/margin_std": 30.997234344482422,
      "step": 253
    },
    {
      "KL/chosen_KL_mean": -43.96406555175781,
      "KL/mean": -53.92585754394531,
      "KL/rejected_KL_mean": -63.88764190673828,
      "KL/std": 32.45115280151367,
      "epoch": 0.3839758125472411,
      "fcm_dpo/beta": 0.02441282756626606,
      "fcm_dpo/delta": -0.09193407744169235,
      "fcm_dpo/margin": 19.923580169677734,
      "fcm_dpo/q_t": 0.39709532260894775,
      "grad_norm": 16.3914737701416,
      "learning_rate": 3.884800159665276e-07,
      "logits/chosen": 0.5786576271057129,
      "logits/rejected": 0.5283867716789246,
      "logps/chosen": -109.60038757324219,
      "logps/ref_chosen": -65.63632202148438,
      "logps/ref_rejected": -82.34425354003906,
      "logps/rejected": -146.2318878173828,
      "loss": 1.1015,
      "margin_dpo/margin_mean": 19.923580169677734,
      "margin_dpo/margin_std": 32.648799896240234,
      "step": 254
    },
    {
      "KL/chosen_KL_mean": -42.81669998168945,
      "KL/mean": -53.882850646972656,
      "KL/rejected_KL_mean": -64.94898986816406,
      "KL/std": 33.449005126953125,
      "epoch": 0.3854875283446712,
      "fcm_dpo/beta": 0.023891150951385498,
      "fcm_dpo/delta": -0.13633979856967926,
      "fcm_dpo/margin": 22.132293701171875,
      "fcm_dpo/q_t": 0.38834911584854126,
      "grad_norm": 22.88119888305664,
      "learning_rate": 3.873772445177015e-07,
      "logits/chosen": 0.5677551627159119,
      "logits/rejected": 0.538213849067688,
      "logps/chosen": -110.72779083251953,
      "logps/ref_chosen": -67.91108703613281,
      "logps/ref_rejected": -83.89114379882812,
      "logps/rejected": -148.84014892578125,
      "loss": 1.0954,
      "margin_dpo/margin_mean": 22.132293701171875,
      "margin_dpo/margin_std": 36.88736343383789,
      "step": 255
    },
    {
      "KL/chosen_KL_mean": -49.34206008911133,
      "KL/mean": -59.96579360961914,
      "KL/rejected_KL_mean": -70.58953094482422,
      "KL/std": 30.069900512695312,
      "epoch": 0.3869992441421013,
      "fcm_dpo/beta": 0.023339707404375076,
      "fcm_dpo/delta": -0.10149209201335907,
      "fcm_dpo/margin": 21.247478485107422,
      "fcm_dpo/q_t": 0.39474910497665405,
      "grad_norm": 19.329296112060547,
      "learning_rate": 3.862706303320329e-07,
      "logits/chosen": 0.5633618831634521,
      "logits/rejected": 0.5043342113494873,
      "logps/chosen": -112.842041015625,
      "logps/ref_chosen": -63.49998474121094,
      "logps/ref_rejected": -90.77104187011719,
      "logps/rejected": -161.36056518554688,
      "loss": 1.1258,
      "margin_dpo/margin_mean": 21.247478485107422,
      "margin_dpo/margin_std": 37.96196746826172,
      "step": 256
    },
    {
      "KL/chosen_KL_mean": -49.642337799072266,
      "KL/mean": -61.384517669677734,
      "KL/rejected_KL_mean": -73.12670135498047,
      "KL/std": 34.05942916870117,
      "epoch": 0.3885109599395314,
      "fcm_dpo/beta": 0.022757742553949356,
      "fcm_dpo/delta": -0.1424705535173416,
      "fcm_dpo/margin": 23.484365463256836,
      "fcm_dpo/q_t": 0.38810110092163086,
      "grad_norm": 17.054840087890625,
      "learning_rate": 3.851602043638994e-07,
      "logits/chosen": 0.5620399713516235,
      "logits/rejected": 0.4995781183242798,
      "logps/chosen": -120.24298858642578,
      "logps/ref_chosen": -70.60064697265625,
      "logps/ref_rejected": -108.58313751220703,
      "logps/rejected": -181.7098388671875,
      "loss": 1.09,
      "margin_dpo/margin_mean": 23.484363555908203,
      "margin_dpo/margin_std": 38.58222198486328,
      "step": 257
    },
    {
      "KL/chosen_KL_mean": -48.516571044921875,
      "KL/mean": -59.06629943847656,
      "KL/rejected_KL_mean": -69.61604309082031,
      "KL/std": 30.164875030517578,
      "epoch": 0.3900226757369615,
      "fcm_dpo/beta": 0.022407684475183487,
      "fcm_dpo/delta": -0.07634987682104111,
      "fcm_dpo/margin": 21.099472045898438,
      "fcm_dpo/q_t": 0.39316412806510925,
      "grad_norm": 16.7133731842041,
      "learning_rate": 3.840459976743023e-07,
      "logits/chosen": 0.6331174373626709,
      "logits/rejected": 0.5832922458648682,
      "logps/chosen": -107.77073669433594,
      "logps/ref_chosen": -59.25416564941406,
      "logps/ref_rejected": -85.58709716796875,
      "logps/rejected": -155.203125,
      "loss": 1.0633,
      "margin_dpo/margin_mean": 21.099475860595703,
      "margin_dpo/margin_std": 28.931100845336914,
      "step": 258
    },
    {
      "KL/chosen_KL_mean": -39.62800216674805,
      "KL/mean": -55.97937774658203,
      "KL/rejected_KL_mean": -72.33076477050781,
      "KL/std": 34.773963928222656,
      "epoch": 0.3915343915343915,
      "fcm_dpo/beta": 0.02117306925356388,
      "fcm_dpo/delta": -0.3162829279899597,
      "fcm_dpo/margin": 32.70275115966797,
      "fcm_dpo/q_t": 0.3467109799385071,
      "grad_norm": 14.047627449035645,
      "learning_rate": 3.8292804142999796e-07,
      "logits/chosen": 0.5472550392150879,
      "logits/rejected": 0.4483594298362732,
      "logps/chosen": -105.06288146972656,
      "logps/ref_chosen": -65.43487548828125,
      "logps/ref_rejected": -95.41731262207031,
      "logps/rejected": -167.74807739257812,
      "loss": 0.9574,
      "margin_dpo/margin_mean": 32.70275115966797,
      "margin_dpo/margin_std": 38.047218322753906,
      "step": 259
    },
    {
      "KL/chosen_KL_mean": -40.95439910888672,
      "KL/mean": -53.225013732910156,
      "KL/rejected_KL_mean": -65.4956283569336,
      "KL/std": 32.90972900390625,
      "epoch": 0.3930461073318216,
      "fcm_dpo/beta": 0.02049969881772995,
      "fcm_dpo/delta": -0.1094924658536911,
      "fcm_dpo/margin": 24.54123306274414,
      "fcm_dpo/q_t": 0.3917416036128998,
      "grad_norm": 16.254623413085938,
      "learning_rate": 3.818063669026256e-07,
      "logits/chosen": 0.5976210832595825,
      "logits/rejected": 0.513933002948761,
      "logps/chosen": -90.04397583007812,
      "logps/ref_chosen": -49.08958435058594,
      "logps/ref_rejected": -79.01708221435547,
      "logps/rejected": -144.51271057128906,
      "loss": 1.0999,
      "margin_dpo/margin_mean": 24.54123306274414,
      "margin_dpo/margin_std": 40.160011291503906,
      "step": 260
    },
    {
      "KL/chosen_KL_mean": -42.14891815185547,
      "KL/mean": -52.500186920166016,
      "KL/rejected_KL_mean": -62.85145568847656,
      "KL/std": 31.833454132080078,
      "epoch": 0.3945578231292517,
      "fcm_dpo/beta": 0.02039218880236149,
      "fcm_dpo/delta": -0.023179400712251663,
      "fcm_dpo/margin": 20.702537536621094,
      "fcm_dpo/q_t": 0.407938152551651,
      "grad_norm": 17.459196090698242,
      "learning_rate": 3.806810054678331e-07,
      "logits/chosen": 0.47628605365753174,
      "logits/rejected": 0.5041095018386841,
      "logps/chosen": -113.02131652832031,
      "logps/ref_chosen": -70.87239074707031,
      "logps/ref_rejected": -65.01522064208984,
      "logps/rejected": -127.8666763305664,
      "loss": 1.1258,
      "margin_dpo/margin_mean": 20.702537536621094,
      "margin_dpo/margin_std": 35.16142272949219,
      "step": 261
    },
    {
      "KL/chosen_KL_mean": -41.45574951171875,
      "KL/mean": -53.318363189697266,
      "KL/rejected_KL_mean": -65.18097686767578,
      "KL/std": 30.827220916748047,
      "epoch": 0.3960695389266818,
      "fcm_dpo/beta": 0.020171720534563065,
      "fcm_dpo/delta": -0.0824466198682785,
      "fcm_dpo/margin": 23.725223541259766,
      "fcm_dpo/q_t": 0.3938416838645935,
      "grad_norm": 14.521263122558594,
      "learning_rate": 3.7955198860439887e-07,
      "logits/chosen": 0.6390868425369263,
      "logits/rejected": 0.5787541270256042,
      "logps/chosen": -109.32637786865234,
      "logps/ref_chosen": -67.8706283569336,
      "logps/ref_rejected": -88.7205810546875,
      "logps/rejected": -153.90155029296875,
      "loss": 1.0722,
      "margin_dpo/margin_mean": 23.725223541259766,
      "margin_dpo/margin_std": 34.007354736328125,
      "step": 262
    },
    {
      "KL/chosen_KL_mean": -40.47084045410156,
      "KL/mean": -50.60345458984375,
      "KL/rejected_KL_mean": -60.7360725402832,
      "KL/std": 32.10907745361328,
      "epoch": 0.3975812547241119,
      "fcm_dpo/beta": 0.020096808671951294,
      "fcm_dpo/delta": -0.007788047194480896,
      "fcm_dpo/margin": 20.265230178833008,
      "fcm_dpo/q_t": 0.40929681062698364,
      "grad_norm": 15.321991920471191,
      "learning_rate": 3.784193478933516e-07,
      "logits/chosen": 0.5773823261260986,
      "logits/rejected": 0.4729832410812378,
      "logps/chosen": -95.66542053222656,
      "logps/ref_chosen": -55.194583892822266,
      "logps/ref_rejected": -80.54048156738281,
      "logps/rejected": -141.27655029296875,
      "loss": 1.134,
      "margin_dpo/margin_mean": 20.265230178833008,
      "margin_dpo/margin_std": 34.80146789550781,
      "step": 263
    },
    {
      "KL/chosen_KL_mean": -41.95197296142578,
      "KL/mean": -53.48637390136719,
      "KL/rejected_KL_mean": -65.0207748413086,
      "KL/std": 33.68387222290039,
      "epoch": 0.39909297052154197,
      "fcm_dpo/beta": 0.019823521375656128,
      "fcm_dpo/delta": -0.06012331694364548,
      "fcm_dpo/margin": 23.068805694580078,
      "fcm_dpo/q_t": 0.3996187448501587,
      "grad_norm": 15.330732345581055,
      "learning_rate": 3.7728311501708674e-07,
      "logits/chosen": 0.5034226775169373,
      "logits/rejected": 0.4578082263469696,
      "logps/chosen": -125.1226577758789,
      "logps/ref_chosen": -83.17068481445312,
      "logps/ref_rejected": -88.33625793457031,
      "logps/rejected": -153.35702514648438,
      "loss": 1.1033,
      "margin_dpo/margin_mean": 23.068805694580078,
      "margin_dpo/margin_std": 37.05900573730469,
      "step": 264
    },
    {
      "KL/chosen_KL_mean": -38.964439392089844,
      "KL/mean": -52.6373291015625,
      "KL/rejected_KL_mean": -66.31021118164062,
      "KL/std": 35.45323944091797,
      "epoch": 0.40060468631897206,
      "fcm_dpo/beta": 0.01950901746749878,
      "fcm_dpo/delta": -0.14103645086288452,
      "fcm_dpo/margin": 27.345767974853516,
      "fcm_dpo/q_t": 0.38523828983306885,
      "grad_norm": 15.546408653259277,
      "learning_rate": 3.7614332175848027e-07,
      "logits/chosen": 0.678200900554657,
      "logits/rejected": 0.6123419404029846,
      "logps/chosen": -90.62728881835938,
      "logps/ref_chosen": -51.66284942626953,
      "logps/ref_rejected": -67.1720962524414,
      "logps/rejected": -133.48231506347656,
      "loss": 1.0887,
      "margin_dpo/margin_mean": 27.34576988220215,
      "margin_dpo/margin_std": 43.287872314453125,
      "step": 265
    },
    {
      "KL/chosen_KL_mean": -37.169830322265625,
      "KL/mean": -49.80873107910156,
      "KL/rejected_KL_mean": -62.447635650634766,
      "KL/std": 36.28041076660156,
      "epoch": 0.4021164021164021,
      "fcm_dpo/beta": 0.018952175974845886,
      "fcm_dpo/delta": -0.08332791179418564,
      "fcm_dpo/margin": 25.277809143066406,
      "fcm_dpo/q_t": 0.39448273181915283,
      "grad_norm": 14.954691886901855,
      "learning_rate": 3.75e-07,
      "logits/chosen": 0.5855690240859985,
      "logits/rejected": 0.5146670341491699,
      "logps/chosen": -94.62032318115234,
      "logps/ref_chosen": -57.45049285888672,
      "logps/ref_rejected": -77.60826110839844,
      "logps/rejected": -140.05589294433594,
      "loss": 1.0813,
      "margin_dpo/margin_mean": 25.277809143066406,
      "margin_dpo/margin_std": 38.019737243652344,
      "step": 266
    },
    {
      "KL/chosen_KL_mean": -32.33792495727539,
      "KL/mean": -41.74827575683594,
      "KL/rejected_KL_mean": -51.158626556396484,
      "KL/std": 32.253929138183594,
      "epoch": 0.4036281179138322,
      "fcm_dpo/beta": 0.019205017015337944,
      "fcm_dpo/delta": 0.03771677613258362,
      "fcm_dpo/margin": 18.82069969177246,
      "fcm_dpo/q_t": 0.4196903705596924,
      "grad_norm": 15.400888442993164,
      "learning_rate": 3.738531817228131e-07,
      "logits/chosen": 0.6113137006759644,
      "logits/rejected": 0.5943053960800171,
      "logps/chosen": -87.37327575683594,
      "logps/ref_chosen": -55.03535079956055,
      "logps/ref_rejected": -66.0953369140625,
      "logps/rejected": -117.25395965576172,
      "loss": 1.1901,
      "margin_dpo/margin_mean": 18.82069969177246,
      "margin_dpo/margin_std": 37.8029899597168,
      "step": 267
    },
    {
      "KL/chosen_KL_mean": -32.36951446533203,
      "KL/mean": -44.62208557128906,
      "KL/rejected_KL_mean": -56.874656677246094,
      "KL/std": 35.25495910644531,
      "epoch": 0.4051398337112623,
      "fcm_dpo/beta": 0.01891086995601654,
      "fcm_dpo/delta": -0.06644882261753082,
      "fcm_dpo/margin": 24.505144119262695,
      "fcm_dpo/q_t": 0.3967518210411072,
      "grad_norm": 13.332362174987793,
      "learning_rate": 3.7270289900589204e-07,
      "logits/chosen": 0.4891277849674225,
      "logits/rejected": 0.47358354926109314,
      "logps/chosen": -97.44125366210938,
      "logps/ref_chosen": -65.07174682617188,
      "logps/ref_rejected": -71.42485809326172,
      "logps/rejected": -128.2995147705078,
      "loss": 1.0643,
      "margin_dpo/margin_mean": 24.505144119262695,
      "margin_dpo/margin_std": 32.81787872314453,
      "step": 268
    },
    {
      "KL/chosen_KL_mean": -37.45969772338867,
      "KL/mean": -51.041168212890625,
      "KL/rejected_KL_mean": -64.62263488769531,
      "KL/std": 36.909271240234375,
      "epoch": 0.40665154950869237,
      "fcm_dpo/beta": 0.018644198775291443,
      "fcm_dpo/delta": -0.11254137754440308,
      "fcm_dpo/margin": 27.162946701049805,
      "fcm_dpo/q_t": 0.3878785967826843,
      "grad_norm": 14.419161796569824,
      "learning_rate": 3.7154918402511714e-07,
      "logits/chosen": 0.6810588836669922,
      "logits/rejected": 0.6340845823287964,
      "logps/chosen": -104.59590148925781,
      "logps/ref_chosen": -67.1362075805664,
      "logps/ref_rejected": -82.55778503417969,
      "logps/rejected": -147.180419921875,
      "loss": 1.0505,
      "margin_dpo/margin_mean": 27.162946701049805,
      "margin_dpo/margin_std": 35.8624382019043,
      "step": 269
    },
    {
      "KL/chosen_KL_mean": -38.30808639526367,
      "KL/mean": -48.73957061767578,
      "KL/rejected_KL_mean": -59.17105484008789,
      "KL/std": 36.70610046386719,
      "epoch": 0.40816326530612246,
      "fcm_dpo/beta": 0.01835768297314644,
      "fcm_dpo/delta": 0.017404425889253616,
      "fcm_dpo/margin": 20.86296844482422,
      "fcm_dpo/q_t": 0.4143076539039612,
      "grad_norm": 14.80038070678711,
      "learning_rate": 3.7039206905237656e-07,
      "logits/chosen": 0.6649652719497681,
      "logits/rejected": 0.5865774750709534,
      "logps/chosen": -104.99678039550781,
      "logps/ref_chosen": -66.6886978149414,
      "logps/ref_rejected": -85.16129302978516,
      "logps/rejected": -144.3323516845703,
      "loss": 1.1578,
      "margin_dpo/margin_mean": 20.86296844482422,
      "margin_dpo/margin_std": 38.87541198730469,
      "step": 270
    },
    {
      "KL/chosen_KL_mean": -36.99169921875,
      "KL/mean": -46.25218963623047,
      "KL/rejected_KL_mean": -55.51268005371094,
      "KL/std": 37.55319595336914,
      "epoch": 0.40967498110355255,
      "fcm_dpo/beta": 0.018530219793319702,
      "fcm_dpo/delta": 0.05868082866072655,
      "fcm_dpo/margin": 18.520978927612305,
      "fcm_dpo/q_t": 0.43027639389038086,
      "grad_norm": 16.319501876831055,
      "learning_rate": 3.692315864546635e-07,
      "logits/chosen": 0.6354309320449829,
      "logits/rejected": 0.5745704174041748,
      "logps/chosen": -109.39924621582031,
      "logps/ref_chosen": -72.40754699707031,
      "logps/ref_rejected": -92.06311798095703,
      "logps/rejected": -147.5758056640625,
      "loss": 1.2254,
      "margin_dpo/margin_mean": 18.520980834960938,
      "margin_dpo/margin_std": 44.24129104614258,
      "step": 271
    },
    {
      "KL/chosen_KL_mean": -29.994138717651367,
      "KL/mean": -48.404571533203125,
      "KL/rejected_KL_mean": -66.81500244140625,
      "KL/std": 36.747406005859375,
      "epoch": 0.41118669690098264,
      "fcm_dpo/beta": 0.017964601516723633,
      "fcm_dpo/delta": -0.2793579697608948,
      "fcm_dpo/margin": 36.820865631103516,
      "fcm_dpo/q_t": 0.35114264488220215,
      "grad_norm": 16.6761531829834,
      "learning_rate": 3.6806776869317067e-07,
      "logits/chosen": 0.6202855706214905,
      "logits/rejected": 0.6367502212524414,
      "logps/chosen": -96.59553527832031,
      "logps/ref_chosen": -66.60140228271484,
      "logps/ref_rejected": -67.74340057373047,
      "logps/rejected": -134.55841064453125,
      "loss": 0.9196,
      "margin_dpo/margin_mean": 36.82086181640625,
      "margin_dpo/margin_std": 35.04280090332031,
      "step": 272
    },
    {
      "KL/chosen_KL_mean": -44.28934097290039,
      "KL/mean": -56.83588790893555,
      "KL/rejected_KL_mean": -69.38243865966797,
      "KL/std": 39.201812744140625,
      "epoch": 0.4126984126984127,
      "fcm_dpo/beta": 0.017505496740341187,
      "fcm_dpo/delta": -0.04126958176493645,
      "fcm_dpo/margin": 25.093097686767578,
      "fcm_dpo/q_t": 0.4040055572986603,
      "grad_norm": 17.27816390991211,
      "learning_rate": 3.669006483223828e-07,
      "logits/chosen": 0.6422320604324341,
      "logits/rejected": 0.5763329267501831,
      "logps/chosen": -101.64421844482422,
      "logps/ref_chosen": -57.35487747192383,
      "logps/ref_rejected": -84.17168426513672,
      "logps/rejected": -153.5541229248047,
      "loss": 1.165,
      "margin_dpo/margin_mean": 25.093097686767578,
      "margin_dpo/margin_std": 49.38603973388672,
      "step": 273
    },
    {
      "KL/chosen_KL_mean": -43.80359649658203,
      "KL/mean": -57.957054138183594,
      "KL/rejected_KL_mean": -72.11050415039062,
      "KL/std": 40.29290771484375,
      "epoch": 0.41421012849584277,
      "fcm_dpo/beta": 0.0172972921282053,
      "fcm_dpo/delta": -0.09414344280958176,
      "fcm_dpo/margin": 28.306915283203125,
      "fcm_dpo/q_t": 0.39450687170028687,
      "grad_norm": 14.509082794189453,
      "learning_rate": 3.657302579891656e-07,
      "logits/chosen": 0.548915684223175,
      "logits/rejected": 0.5313537120819092,
      "logps/chosen": -103.4450912475586,
      "logps/ref_chosen": -59.64149475097656,
      "logps/ref_rejected": -68.29348754882812,
      "logps/rejected": -140.40399169921875,
      "loss": 1.1079,
      "margin_dpo/margin_mean": 28.306915283203125,
      "margin_dpo/margin_std": 47.45249557495117,
      "step": 274
    },
    {
      "KL/chosen_KL_mean": -42.47492980957031,
      "KL/mean": -57.8199462890625,
      "KL/rejected_KL_mean": -73.16497039794922,
      "KL/std": 39.48030090332031,
      "epoch": 0.41572184429327286,
      "fcm_dpo/beta": 0.016921717673540115,
      "fcm_dpo/delta": -0.1256747543811798,
      "fcm_dpo/margin": 30.690044403076172,
      "fcm_dpo/q_t": 0.38423293828964233,
      "grad_norm": 14.718132972717285,
      "learning_rate": 3.645566304318526e-07,
      "logits/chosen": 0.6262120008468628,
      "logits/rejected": 0.544623851776123,
      "logps/chosen": -95.7415771484375,
      "logps/ref_chosen": -53.26664352416992,
      "logps/ref_rejected": -73.84062194824219,
      "logps/rejected": -147.00558471679688,
      "loss": 1.049,
      "margin_dpo/margin_mean": 30.690044403076172,
      "margin_dpo/margin_std": 42.01318359375,
      "step": 275
    },
    {
      "KL/chosen_KL_mean": -40.822547912597656,
      "KL/mean": -54.688377380371094,
      "KL/rejected_KL_mean": -68.55419921875,
      "KL/std": 38.242088317871094,
      "epoch": 0.41723356009070295,
      "fcm_dpo/beta": 0.016595926135778427,
      "fcm_dpo/delta": -0.06314849853515625,
      "fcm_dpo/margin": 27.73164939880371,
      "fcm_dpo/q_t": 0.3971312642097473,
      "grad_norm": 16.296781539916992,
      "learning_rate": 3.633797984793294e-07,
      "logits/chosen": 0.5706058144569397,
      "logits/rejected": 0.538550615310669,
      "logps/chosen": -93.84334564208984,
      "logps/ref_chosen": -53.02079772949219,
      "logps/ref_rejected": -61.56678771972656,
      "logps/rejected": -130.12098693847656,
      "loss": 1.0897,
      "margin_dpo/margin_mean": 27.73164939880371,
      "margin_dpo/margin_std": 41.87763214111328,
      "step": 276
    },
    {
      "KL/chosen_KL_mean": -45.02290344238281,
      "KL/mean": -54.18437957763672,
      "KL/rejected_KL_mean": -63.345855712890625,
      "KL/std": 37.87938690185547,
      "epoch": 0.41874527588813304,
      "fcm_dpo/beta": 0.01680148020386696,
      "fcm_dpo/delta": 0.09492812305688858,
      "fcm_dpo/margin": 18.322956085205078,
      "fcm_dpo/q_t": 0.4322887063026428,
      "grad_norm": 18.618850708007812,
      "learning_rate": 3.6219979505011555e-07,
      "logits/chosen": 0.6885474920272827,
      "logits/rejected": 0.7152642607688904,
      "logps/chosen": -116.45588684082031,
      "logps/ref_chosen": -71.43299102783203,
      "logps/ref_rejected": -67.65852355957031,
      "logps/rejected": -131.00437927246094,
      "loss": 1.2154,
      "margin_dpo/margin_mean": 18.322956085205078,
      "margin_dpo/margin_std": 41.572853088378906,
      "step": 277
    },
    {
      "KL/chosen_KL_mean": -50.65361785888672,
      "KL/mean": -64.87567901611328,
      "KL/rejected_KL_mean": -79.09774780273438,
      "KL/std": 43.28910827636719,
      "epoch": 0.42025699168556313,
      "fcm_dpo/beta": 0.016770116984844208,
      "fcm_dpo/delta": -0.08121868222951889,
      "fcm_dpo/margin": 28.44412612915039,
      "fcm_dpo/q_t": 0.3952862620353699,
      "grad_norm": 23.123279571533203,
      "learning_rate": 3.6101665315144353e-07,
      "logits/chosen": 0.5624623894691467,
      "logits/rejected": 0.5138440132141113,
      "logps/chosen": -117.76438903808594,
      "logps/ref_chosen": -67.11076354980469,
      "logps/ref_rejected": -88.74851989746094,
      "logps/rejected": -167.8462677001953,
      "loss": 1.1019,
      "margin_dpo/margin_mean": 28.44412612915039,
      "margin_dpo/margin_std": 45.58794021606445,
      "step": 278
    },
    {
      "KL/chosen_KL_mean": -36.818626403808594,
      "KL/mean": -56.46868896484375,
      "KL/rejected_KL_mean": -76.11874389648438,
      "KL/std": 42.45269775390625,
      "epoch": 0.4217687074829932,
      "fcm_dpo/beta": 0.016032833606004715,
      "fcm_dpo/delta": -0.24514293670654297,
      "fcm_dpo/margin": 39.30011749267578,
      "fcm_dpo/q_t": 0.3587535619735718,
      "grad_norm": 18.80735206604004,
      "learning_rate": 3.5983040587833563e-07,
      "logits/chosen": 0.5911531448364258,
      "logits/rejected": 0.5552926063537598,
      "logps/chosen": -91.31611633300781,
      "logps/ref_chosen": -54.49748611450195,
      "logps/ref_rejected": -70.42373657226562,
      "logps/rejected": -146.54248046875,
      "loss": 0.9504,
      "margin_dpo/margin_mean": 39.30011749267578,
      "margin_dpo/margin_std": 40.607357025146484,
      "step": 279
    },
    {
      "KL/chosen_KL_mean": -40.82258224487305,
      "KL/mean": -61.65607452392578,
      "KL/rejected_KL_mean": -82.48957061767578,
      "KL/std": 44.68724822998047,
      "epoch": 0.42328042328042326,
      "fcm_dpo/beta": 0.015228111296892166,
      "fcm_dpo/delta": -0.2502999007701874,
      "fcm_dpo/margin": 41.6669921875,
      "fcm_dpo/q_t": 0.3581681251525879,
      "grad_norm": 11.828532218933105,
      "learning_rate": 3.586410864126781e-07,
      "logits/chosen": 0.6419994235038757,
      "logits/rejected": 0.6040011644363403,
      "logps/chosen": -101.2553939819336,
      "logps/ref_chosen": -60.43281173706055,
      "logps/ref_rejected": -78.39051818847656,
      "logps/rejected": -160.88009643554688,
      "loss": 0.9388,
      "margin_dpo/margin_mean": 41.6669921875,
      "margin_dpo/margin_std": 41.892921447753906,
      "step": 280
    },
    {
      "KL/chosen_KL_mean": -45.86112976074219,
      "KL/mean": -64.68388366699219,
      "KL/rejected_KL_mean": -83.50663757324219,
      "KL/std": 44.603004455566406,
      "epoch": 0.42479213907785335,
      "fcm_dpo/beta": 0.014688611030578613,
      "fcm_dpo/delta": -0.1617349535226822,
      "fcm_dpo/margin": 37.64551544189453,
      "fcm_dpo/q_t": 0.37709325551986694,
      "grad_norm": 12.660872459411621,
      "learning_rate": 3.574487280222929e-07,
      "logits/chosen": 0.5922428965568542,
      "logits/rejected": 0.6139553785324097,
      "logps/chosen": -106.14321899414062,
      "logps/ref_chosen": -60.2820930480957,
      "logps/ref_rejected": -62.04009246826172,
      "logps/rejected": -145.54673767089844,
      "loss": 1.0202,
      "margin_dpo/margin_mean": 37.64551544189453,
      "margin_dpo/margin_std": 47.75347137451172,
      "step": 281
    },
    {
      "KL/chosen_KL_mean": -52.35889434814453,
      "KL/mean": -68.90093994140625,
      "KL/rejected_KL_mean": -85.44297790527344,
      "KL/std": 44.76057434082031,
      "epoch": 0.42630385487528344,
      "fcm_dpo/beta": 0.014499800279736519,
      "fcm_dpo/delta": -0.08470294624567032,
      "fcm_dpo/margin": 33.084075927734375,
      "fcm_dpo/q_t": 0.39539259672164917,
      "grad_norm": 18.454259872436523,
      "learning_rate": 3.562533640600075e-07,
      "logits/chosen": 0.5745443105697632,
      "logits/rejected": 0.5282710790634155,
      "logps/chosen": -112.98281860351562,
      "logps/ref_chosen": -60.623924255371094,
      "logps/ref_rejected": -68.67400360107422,
      "logps/rejected": -154.11697387695312,
      "loss": 1.097,
      "margin_dpo/margin_mean": 33.084075927734375,
      "margin_dpo/margin_std": 51.07267379760742,
      "step": 282
    },
    {
      "KL/chosen_KL_mean": -53.396873474121094,
      "KL/mean": -68.42160034179688,
      "KL/rejected_KL_mean": -83.44633483886719,
      "KL/std": 43.950477600097656,
      "epoch": 0.42781557067271353,
      "fcm_dpo/beta": 0.014289552345871925,
      "fcm_dpo/delta": -0.030874045565724373,
      "fcm_dpo/margin": 30.049449920654297,
      "fcm_dpo/q_t": 0.4046875834465027,
      "grad_norm": 15.984442710876465,
      "learning_rate": 3.550550279627215e-07,
      "logits/chosen": 0.6232196688652039,
      "logits/rejected": 0.525344729423523,
      "logps/chosen": -121.04461669921875,
      "logps/ref_chosen": -67.64775085449219,
      "logps/ref_rejected": -99.96835327148438,
      "logps/rejected": -183.41468811035156,
      "loss": 1.121,
      "margin_dpo/margin_mean": 30.049449920654297,
      "margin_dpo/margin_std": 49.89061737060547,
      "step": 283
    },
    {
      "KL/chosen_KL_mean": -51.50770950317383,
      "KL/mean": -67.859619140625,
      "KL/rejected_KL_mean": -84.21153259277344,
      "KL/std": 47.82459259033203,
      "epoch": 0.4293272864701436,
      "fcm_dpo/beta": 0.01402394287288189,
      "fcm_dpo/delta": -0.06209135428071022,
      "fcm_dpo/margin": 32.703819274902344,
      "fcm_dpo/q_t": 0.3963380455970764,
      "grad_norm": 13.340622901916504,
      "learning_rate": 3.5385375325047163e-07,
      "logits/chosen": 0.697113037109375,
      "logits/rejected": 0.6345241069793701,
      "logps/chosen": -108.47513580322266,
      "logps/ref_chosen": -56.96742630004883,
      "logps/ref_rejected": -86.36236572265625,
      "logps/rejected": -170.5738983154297,
      "loss": 1.0736,
      "margin_dpo/margin_mean": 32.703819274902344,
      "margin_dpo/margin_std": 45.35961151123047,
      "step": 284
    },
    {
      "KL/chosen_KL_mean": -64.46051025390625,
      "KL/mean": -78.41537475585938,
      "KL/rejected_KL_mean": -92.3702392578125,
      "KL/std": 42.90559005737305,
      "epoch": 0.4308390022675737,
      "fcm_dpo/beta": 0.014100934378802776,
      "fcm_dpo/delta": 0.006449030712246895,
      "fcm_dpo/margin": 27.90973472595215,
      "fcm_dpo/q_t": 0.412301242351532,
      "grad_norm": 17.84235191345215,
      "learning_rate": 3.5264957352549375e-07,
      "logits/chosen": 0.686227560043335,
      "logits/rejected": 0.6604204177856445,
      "logps/chosen": -136.11660766601562,
      "logps/ref_chosen": -71.65611267089844,
      "logps/ref_rejected": -81.63829803466797,
      "logps/rejected": -174.008544921875,
      "loss": 1.1348,
      "margin_dpo/margin_mean": 27.909732818603516,
      "margin_dpo/margin_std": 47.20280456542969,
      "step": 285
    },
    {
      "KL/chosen_KL_mean": -58.55393600463867,
      "KL/mean": -80.9109878540039,
      "KL/rejected_KL_mean": -103.26805114746094,
      "KL/std": 46.35716247558594,
      "epoch": 0.4323507180650038,
      "fcm_dpo/beta": 0.013572895899415016,
      "fcm_dpo/delta": -0.2208898961544037,
      "fcm_dpo/margin": 44.71410369873047,
      "fcm_dpo/q_t": 0.3641088902950287,
      "grad_norm": 14.018209457397461,
      "learning_rate": 3.514425224712835e-07,
      "logits/chosen": 0.5620067715644836,
      "logits/rejected": 0.4744671583175659,
      "logps/chosen": -119.63345336914062,
      "logps/ref_chosen": -61.07952117919922,
      "logps/ref_rejected": -91.28128051757812,
      "logps/rejected": -194.54931640625,
      "loss": 0.9634,
      "margin_dpo/margin_mean": 44.71410369873047,
      "margin_dpo/margin_std": 48.03024673461914,
      "step": 286
    },
    {
      "KL/chosen_KL_mean": -47.898338317871094,
      "KL/mean": -70.18797302246094,
      "KL/rejected_KL_mean": -92.47761535644531,
      "KL/std": 49.48754119873047,
      "epoch": 0.43386243386243384,
      "fcm_dpo/beta": 0.01308109425008297,
      "fcm_dpo/delta": -0.19431273639202118,
      "fcm_dpo/margin": 44.57927703857422,
      "fcm_dpo/q_t": 0.3690979480743408,
      "grad_norm": 12.084219932556152,
      "learning_rate": 3.502326338516534e-07,
      "logits/chosen": 0.6664276719093323,
      "logits/rejected": 0.6316767930984497,
      "logps/chosen": -93.93412780761719,
      "logps/ref_chosen": -46.035789489746094,
      "logps/ref_rejected": -59.95293426513672,
      "logps/rejected": -152.4305419921875,
      "loss": 0.9841,
      "margin_dpo/margin_mean": 44.57927703857422,
      "margin_dpo/margin_std": 50.418670654296875,
      "step": 287
    },
    {
      "KL/chosen_KL_mean": -63.46906661987305,
      "KL/mean": -79.67216491699219,
      "KL/rejected_KL_mean": -95.87527465820312,
      "KL/std": 46.347686767578125,
      "epoch": 0.43537414965986393,
      "fcm_dpo/beta": 0.012921486049890518,
      "fcm_dpo/delta": -0.019581297412514687,
      "fcm_dpo/margin": 32.406211853027344,
      "fcm_dpo/q_t": 0.4060286283493042,
      "grad_norm": 14.524436950683594,
      "learning_rate": 3.490199415097892e-07,
      "logits/chosen": 0.5129345655441284,
      "logits/rejected": 0.4621993899345398,
      "logps/chosen": -128.8599090576172,
      "logps/ref_chosen": -65.3908462524414,
      "logps/ref_rejected": -88.53607940673828,
      "logps/rejected": -184.41134643554688,
      "loss": 1.1137,
      "margin_dpo/margin_mean": 32.40620803833008,
      "margin_dpo/margin_std": 51.43866729736328,
      "step": 288
    },
    {
      "KL/chosen_KL_mean": -63.998863220214844,
      "KL/mean": -78.66741943359375,
      "KL/rejected_KL_mean": -93.33598327636719,
      "KL/std": 46.610191345214844,
      "epoch": 0.436885865457294,
      "fcm_dpo/beta": 0.012985285371541977,
      "fcm_dpo/delta": 0.019250046461820602,
      "fcm_dpo/margin": 29.337120056152344,
      "fcm_dpo/q_t": 0.41725146770477295,
      "grad_norm": 16.7989444732666,
      "learning_rate": 3.4780447936730247e-07,
      "logits/chosen": 0.758312463760376,
      "logits/rejected": 0.7211253643035889,
      "logps/chosen": -118.59249114990234,
      "logps/ref_chosen": -54.5936279296875,
      "logps/ref_rejected": -67.20855712890625,
      "logps/rejected": -160.54452514648438,
      "loss": 1.1555,
      "margin_dpo/margin_mean": 29.337120056152344,
      "margin_dpo/margin_std": 53.40695571899414,
      "step": 289
    },
    {
      "KL/chosen_KL_mean": -73.34171295166016,
      "KL/mean": -91.50950622558594,
      "KL/rejected_KL_mean": -109.67730712890625,
      "KL/std": 45.4484977722168,
      "epoch": 0.4383975812547241,
      "fcm_dpo/beta": 0.01275827456265688,
      "fcm_dpo/delta": -0.06723187863826752,
      "fcm_dpo/margin": 36.33559036254883,
      "fcm_dpo/q_t": 0.396625816822052,
      "grad_norm": 16.9279727935791,
      "learning_rate": 3.465862814232821e-07,
      "logits/chosen": 0.7583713531494141,
      "logits/rejected": 0.6876074075698853,
      "logps/chosen": -134.72628784179688,
      "logps/ref_chosen": -61.38457489013672,
      "logps/ref_rejected": -91.92778015136719,
      "logps/rejected": -201.60507202148438,
      "loss": 1.0937,
      "margin_dpo/margin_mean": 36.33559036254883,
      "margin_dpo/margin_std": 56.050289154052734,
      "step": 290
    },
    {
      "KL/chosen_KL_mean": -70.3595962524414,
      "KL/mean": -89.24057006835938,
      "KL/rejected_KL_mean": -108.12155151367188,
      "KL/std": 45.235504150390625,
      "epoch": 0.4399092970521542,
      "fcm_dpo/beta": 0.012722047045826912,
      "fcm_dpo/delta": -0.08546677231788635,
      "fcm_dpo/margin": 37.76197052001953,
      "fcm_dpo/q_t": 0.3933485150337219,
      "grad_norm": 16.51990509033203,
      "learning_rate": 3.4536538175334343e-07,
      "logits/chosen": 0.8289006352424622,
      "logits/rejected": 0.7604660987854004,
      "logps/chosen": -121.2226333618164,
      "logps/ref_chosen": -50.863037109375,
      "logps/ref_rejected": -82.20868682861328,
      "logps/rejected": -190.33023071289062,
      "loss": 1.0671,
      "margin_dpo/margin_mean": 37.761966705322266,
      "margin_dpo/margin_std": 50.908905029296875,
      "step": 291
    },
    {
      "KL/chosen_KL_mean": -69.3774642944336,
      "KL/mean": -84.8953857421875,
      "KL/rejected_KL_mean": -100.41329956054688,
      "KL/std": 48.227535247802734,
      "epoch": 0.4414210128495843,
      "fcm_dpo/beta": 0.012567687779664993,
      "fcm_dpo/delta": 0.010354185476899147,
      "fcm_dpo/margin": 31.035823822021484,
      "fcm_dpo/q_t": 0.4122801125049591,
      "grad_norm": 15.849407196044922,
      "learning_rate": 3.4414181450867465e-07,
      "logits/chosen": 0.6781643033027649,
      "logits/rejected": 0.6310281753540039,
      "logps/chosen": -133.72634887695312,
      "logps/ref_chosen": -64.34888458251953,
      "logps/ref_rejected": -72.86434173583984,
      "logps/rejected": -173.27764892578125,
      "loss": 1.1477,
      "margin_dpo/margin_mean": 31.035825729370117,
      "margin_dpo/margin_std": 55.531700134277344,
      "step": 292
    },
    {
      "KL/chosen_KL_mean": -68.59029388427734,
      "KL/mean": -91.03254699707031,
      "KL/rejected_KL_mean": -113.47480773925781,
      "KL/std": 49.78746795654297,
      "epoch": 0.4429327286470144,
      "fcm_dpo/beta": 0.012334262020885944,
      "fcm_dpo/delta": -0.16231675446033478,
      "fcm_dpo/margin": 44.88451385498047,
      "fcm_dpo/q_t": 0.3771790862083435,
      "grad_norm": 11.519031524658203,
      "learning_rate": 3.4291561391508185e-07,
      "logits/chosen": 0.7449074983596802,
      "logits/rejected": 0.661964476108551,
      "logps/chosen": -123.45976257324219,
      "logps/ref_chosen": -54.869468688964844,
      "logps/ref_rejected": -81.858642578125,
      "logps/rejected": -195.3334503173828,
      "loss": 1.0391,
      "margin_dpo/margin_mean": 44.88451385498047,
      "margin_dpo/margin_std": 60.664588928222656,
      "step": 293
    },
    {
      "KL/chosen_KL_mean": -71.3341064453125,
      "KL/mean": -86.29549407958984,
      "KL/rejected_KL_mean": -101.25688171386719,
      "KL/std": 51.49919509887695,
      "epoch": 0.4444444444444444,
      "fcm_dpo/beta": 0.012190229259431362,
      "fcm_dpo/delta": 0.03639962524175644,
      "fcm_dpo/margin": 29.922760009765625,
      "fcm_dpo/q_t": 0.4193214476108551,
      "grad_norm": 12.29130744934082,
      "learning_rate": 3.4168681427203153e-07,
      "logits/chosen": 0.717154860496521,
      "logits/rejected": 0.6727120280265808,
      "logps/chosen": -128.0050048828125,
      "logps/ref_chosen": -56.670902252197266,
      "logps/ref_rejected": -70.32819366455078,
      "logps/rejected": -171.58506774902344,
      "loss": 1.1385,
      "margin_dpo/margin_mean": 29.922760009765625,
      "margin_dpo/margin_std": 49.23522186279297,
      "step": 294
    },
    {
      "KL/chosen_KL_mean": -77.55728149414062,
      "KL/mean": -91.69718933105469,
      "KL/rejected_KL_mean": -105.83709716796875,
      "KL/std": 49.87922668457031,
      "epoch": 0.4459561602418745,
      "fcm_dpo/beta": 0.012327713891863823,
      "fcm_dpo/delta": 0.05323711037635803,
      "fcm_dpo/margin": 28.27980613708496,
      "fcm_dpo/q_t": 0.42287230491638184,
      "grad_norm": 19.666093826293945,
      "learning_rate": 3.4045544995169125e-07,
      "logits/chosen": 0.7301595211029053,
      "logits/rejected": 0.6291660070419312,
      "logps/chosen": -127.95817565917969,
      "logps/ref_chosen": -50.40088653564453,
      "logps/ref_rejected": -83.43521881103516,
      "logps/rejected": -189.27230834960938,
      "loss": 1.1712,
      "margin_dpo/margin_mean": 28.27980613708496,
      "margin_dpo/margin_std": 53.96001434326172,
      "step": 295
    },
    {
      "KL/chosen_KL_mean": -76.57630920410156,
      "KL/mean": -94.01647186279297,
      "KL/rejected_KL_mean": -111.45663452148438,
      "KL/std": 50.745975494384766,
      "epoch": 0.4474678760393046,
      "fcm_dpo/beta": 0.0122788455337286,
      "fcm_dpo/delta": -0.030141662806272507,
      "fcm_dpo/margin": 34.88032150268555,
      "fcm_dpo/q_t": 0.40467900037765503,
      "grad_norm": 12.9859619140625,
      "learning_rate": 3.392215553979679e-07,
      "logits/chosen": 0.6649519205093384,
      "logits/rejected": 0.6230114698410034,
      "logps/chosen": -145.72665405273438,
      "logps/ref_chosen": -69.15034484863281,
      "logps/ref_rejected": -89.60166931152344,
      "logps/rejected": -201.0583038330078,
      "loss": 1.1129,
      "margin_dpo/margin_mean": 34.88031768798828,
      "margin_dpo/margin_std": 55.64665222167969,
      "step": 296
    },
    {
      "KL/chosen_KL_mean": -80.91482543945312,
      "KL/mean": -100.35860443115234,
      "KL/rejected_KL_mean": -119.80238342285156,
      "KL/std": 45.48234176635742,
      "epoch": 0.4489795918367347,
      "fcm_dpo/beta": 0.012235373258590698,
      "fcm_dpo/delta": -0.0797240287065506,
      "fcm_dpo/margin": 38.88756561279297,
      "fcm_dpo/q_t": 0.392116516828537,
      "grad_norm": 13.477424621582031,
      "learning_rate": 3.3798516512554485e-07,
      "logits/chosen": 0.6948250532150269,
      "logits/rejected": 0.638218343257904,
      "logps/chosen": -138.93113708496094,
      "logps/ref_chosen": -58.01630401611328,
      "logps/ref_rejected": -69.95780944824219,
      "logps/rejected": -189.76019287109375,
      "loss": 1.0511,
      "margin_dpo/margin_mean": 38.887569427490234,
      "margin_dpo/margin_std": 48.81712341308594,
      "step": 297
    },
    {
      "KL/chosen_KL_mean": -78.26363372802734,
      "KL/mean": -93.04943084716797,
      "KL/rejected_KL_mean": -107.83523559570312,
      "KL/std": 48.65501403808594,
      "epoch": 0.4504913076341648,
      "fcm_dpo/beta": 0.012198349460959435,
      "fcm_dpo/delta": 0.04073771834373474,
      "fcm_dpo/margin": 29.57158088684082,
      "fcm_dpo/q_t": 0.4210200309753418,
      "grad_norm": 13.820934295654297,
      "learning_rate": 3.367463137189156e-07,
      "logits/chosen": 0.8141465187072754,
      "logits/rejected": 0.7540981769561768,
      "logps/chosen": -134.43295288085938,
      "logps/ref_chosen": -56.1693115234375,
      "logps/ref_rejected": -68.55052185058594,
      "logps/rejected": -176.3857421875,
      "loss": 1.1832,
      "margin_dpo/margin_mean": 29.57158088684082,
      "margin_dpo/margin_std": 59.90074920654297,
      "step": 298
    },
    {
      "KL/chosen_KL_mean": -77.41154479980469,
      "KL/mean": -90.41871643066406,
      "KL/rejected_KL_mean": -103.42588806152344,
      "KL/std": 46.954566955566406,
      "epoch": 0.4520030234315949,
      "fcm_dpo/beta": 0.012318475171923637,
      "fcm_dpo/delta": 0.08214488625526428,
      "fcm_dpo/margin": 26.01433753967285,
      "fcm_dpo/q_t": 0.42924293875694275,
      "grad_norm": 18.75337791442871,
      "learning_rate": 3.355050358314172e-07,
      "logits/chosen": 0.6139011383056641,
      "logits/rejected": 0.586826503276825,
      "logps/chosen": -139.72933959960938,
      "logps/ref_chosen": -62.31780242919922,
      "logps/ref_rejected": -72.60028839111328,
      "logps/rejected": -176.02618408203125,
      "loss": 1.2215,
      "margin_dpo/margin_mean": 26.01433753967285,
      "margin_dpo/margin_std": 59.20307159423828,
      "step": 299
    },
    {
      "KL/chosen_KL_mean": -77.873291015625,
      "KL/mean": -94.65007019042969,
      "KL/rejected_KL_mean": -111.42684936523438,
      "KL/std": 47.93767547607422,
      "epoch": 0.45351473922902497,
      "fcm_dpo/beta": 0.012403897941112518,
      "fcm_dpo/delta": -0.016899462789297104,
      "fcm_dpo/margin": 33.553558349609375,
      "fcm_dpo/q_t": 0.40718337893486023,
      "grad_norm": 13.85849666595459,
      "learning_rate": 3.3426136618426043e-07,
      "logits/chosen": 0.7300994396209717,
      "logits/rejected": 0.6637899875640869,
      "logps/chosen": -138.2548828125,
      "logps/ref_chosen": -60.38157653808594,
      "logps/ref_rejected": -75.45442199707031,
      "logps/rejected": -186.8812713623047,
      "loss": 1.1346,
      "margin_dpo/margin_mean": 33.553558349609375,
      "margin_dpo/margin_std": 58.16008377075195,
      "step": 300
    },
    {
      "KL/chosen_KL_mean": -75.81689453125,
      "KL/mean": -90.96798706054688,
      "KL/rejected_KL_mean": -106.11907958984375,
      "KL/std": 47.327850341796875,
      "epoch": 0.455026455026455,
      "fcm_dpo/beta": 0.012391261756420135,
      "fcm_dpo/delta": 0.025361964479088783,
      "fcm_dpo/margin": 30.30219268798828,
      "fcm_dpo/q_t": 0.41580748558044434,
      "grad_norm": 15.147626876831055,
      "learning_rate": 3.3301533956555885e-07,
      "logits/chosen": 0.7389202117919922,
      "logits/rejected": 0.7118491530418396,
      "logps/chosen": -128.66778564453125,
      "logps/ref_chosen": -52.85089111328125,
      "logps/ref_rejected": -69.97584533691406,
      "logps/rejected": -176.0949249267578,
      "loss": 1.1545,
      "margin_dpo/margin_mean": 30.30219078063965,
      "margin_dpo/margin_std": 55.01183319091797,
      "step": 301
    },
    {
      "KL/chosen_KL_mean": -80.43342590332031,
      "KL/mean": -91.39155578613281,
      "KL/rejected_KL_mean": -102.34968566894531,
      "KL/std": 47.64586639404297,
      "epoch": 0.4565381708238851,
      "fcm_dpo/beta": 0.012671603821218014,
      "fcm_dpo/delta": 0.12586434185504913,
      "fcm_dpo/margin": 21.916255950927734,
      "fcm_dpo/q_t": 0.43761053681373596,
      "grad_norm": 18.956130981445312,
      "learning_rate": 3.317669908293554e-07,
      "logits/chosen": 0.5832021832466125,
      "logits/rejected": 0.5297772288322449,
      "logps/chosen": -147.39993286132812,
      "logps/ref_chosen": -66.96650695800781,
      "logps/ref_rejected": -88.09510803222656,
      "logps/rejected": -190.44479370117188,
      "loss": 1.2328,
      "margin_dpo/margin_mean": 21.916255950927734,
      "margin_dpo/margin_std": 52.96202850341797,
      "step": 302
    },
    {
      "KL/chosen_KL_mean": -69.9097900390625,
      "KL/mean": -90.5808334350586,
      "KL/rejected_KL_mean": -111.25189208984375,
      "KL/std": 48.5565185546875,
      "epoch": 0.4580498866213152,
      "fcm_dpo/beta": 0.012538776732981205,
      "fcm_dpo/delta": -0.12476019561290741,
      "fcm_dpo/margin": 41.34212112426758,
      "fcm_dpo/q_t": 0.38511401414871216,
      "grad_norm": 12.396648406982422,
      "learning_rate": 3.3051635489464793e-07,
      "logits/chosen": 0.6951050758361816,
      "logits/rejected": 0.6301414966583252,
      "logps/chosen": -132.03131103515625,
      "logps/ref_chosen": -62.12152862548828,
      "logps/ref_rejected": -90.31204223632812,
      "logps/rejected": -201.56393432617188,
      "loss": 1.0655,
      "margin_dpo/margin_mean": 41.342124938964844,
      "margin_dpo/margin_std": 60.36699295043945,
      "step": 303
    },
    {
      "KL/chosen_KL_mean": -60.231353759765625,
      "KL/mean": -80.84054565429688,
      "KL/rejected_KL_mean": -101.44972229003906,
      "KL/std": 51.155372619628906,
      "epoch": 0.4595616024187453,
      "fcm_dpo/beta": 0.0122285857796669,
      "fcm_dpo/delta": -0.10976764559745789,
      "fcm_dpo/margin": 41.21836853027344,
      "fcm_dpo/q_t": 0.38361793756484985,
      "grad_norm": 13.756430625915527,
      "learning_rate": 3.292634667444117e-07,
      "logits/chosen": 0.6984633207321167,
      "logits/rejected": 0.6454420685768127,
      "logps/chosen": -120.92644500732422,
      "logps/ref_chosen": -60.695091247558594,
      "logps/ref_rejected": -78.2525405883789,
      "logps/rejected": -179.7022705078125,
      "loss": 1.0208,
      "margin_dpo/margin_mean": 41.21836853027344,
      "margin_dpo/margin_std": 46.54623794555664,
      "step": 304
    },
    {
      "KL/chosen_KL_mean": -70.44589233398438,
      "KL/mean": -86.21815490722656,
      "KL/rejected_KL_mean": -101.99043273925781,
      "KL/std": 48.03346252441406,
      "epoch": 0.46107331821617537,
      "fcm_dpo/beta": 0.012153957970440388,
      "fcm_dpo/delta": 0.016938693821430206,
      "fcm_dpo/margin": 31.544536590576172,
      "fcm_dpo/q_t": 0.41375792026519775,
      "grad_norm": 13.166147232055664,
      "learning_rate": 3.280083614246217e-07,
      "logits/chosen": 0.6348699331283569,
      "logits/rejected": 0.6634021997451782,
      "logps/chosen": -143.14503479003906,
      "logps/ref_chosen": -72.69914245605469,
      "logps/ref_rejected": -65.65670776367188,
      "logps/rejected": -167.6471405029297,
      "loss": 1.155,
      "margin_dpo/margin_mean": 31.544536590576172,
      "margin_dpo/margin_std": 57.23649597167969,
      "step": 305
    },
    {
      "KL/chosen_KL_mean": -66.22614288330078,
      "KL/mean": -82.06704711914062,
      "KL/rejected_KL_mean": -97.9079360961914,
      "KL/std": 46.619834899902344,
      "epoch": 0.46258503401360546,
      "fcm_dpo/beta": 0.012171389535069466,
      "fcm_dpo/delta": 0.014306794852018356,
      "fcm_dpo/margin": 31.681785583496094,
      "fcm_dpo/q_t": 0.41203343868255615,
      "grad_norm": 13.922178268432617,
      "learning_rate": 3.267510740432719e-07,
      "logits/chosen": 0.7303614616394043,
      "logits/rejected": 0.6184151768684387,
      "logps/chosen": -120.19667053222656,
      "logps/ref_chosen": -53.97052764892578,
      "logps/ref_rejected": -71.02423095703125,
      "logps/rejected": -168.93215942382812,
      "loss": 1.1116,
      "margin_dpo/margin_mean": 31.681781768798828,
      "margin_dpo/margin_std": 45.20012283325195,
      "step": 306
    },
    {
      "KL/chosen_KL_mean": -60.205955505371094,
      "KL/mean": -68.00119018554688,
      "KL/rejected_KL_mean": -75.79643249511719,
      "KL/std": 45.12220001220703,
      "epoch": 0.46409674981103555,
      "fcm_dpo/beta": 0.012687662616372108,
      "fcm_dpo/delta": 0.20508906245231628,
      "fcm_dpo/margin": 15.590478897094727,
      "fcm_dpo/q_t": 0.457867294549942,
      "grad_norm": 17.767627716064453,
      "learning_rate": 3.2549163976939285e-07,
      "logits/chosen": 0.7361500859260559,
      "logits/rejected": 0.6878568530082703,
      "logps/chosen": -117.61906433105469,
      "logps/ref_chosen": -57.413108825683594,
      "logps/ref_rejected": -68.68010711669922,
      "logps/rejected": -144.47653198242188,
      "loss": 1.3222,
      "margin_dpo/margin_mean": 15.590479850769043,
      "margin_dpo/margin_std": 56.86015701293945,
      "step": 307
    },
    {
      "KL/chosen_KL_mean": -59.32799530029297,
      "KL/mean": -74.51152038574219,
      "KL/rejected_KL_mean": -89.69503784179688,
      "KL/std": 46.99913787841797,
      "epoch": 0.4656084656084656,
      "fcm_dpo/beta": 0.012745361775159836,
      "fcm_dpo/delta": 0.013256091624498367,
      "fcm_dpo/margin": 30.367046356201172,
      "fcm_dpo/q_t": 0.41266879439353943,
      "grad_norm": 11.760419845581055,
      "learning_rate": 3.2423009383206874e-07,
      "logits/chosen": 0.6760587096214294,
      "logits/rejected": 0.6648536324501038,
      "logps/chosen": -125.92678833007812,
      "logps/ref_chosen": -66.59879302978516,
      "logps/ref_rejected": -74.337158203125,
      "logps/rejected": -164.03219604492188,
      "loss": 1.1453,
      "margin_dpo/margin_mean": 30.367046356201172,
      "margin_dpo/margin_std": 53.325408935546875,
      "step": 308
    },
    {
      "KL/chosen_KL_mean": -68.26148986816406,
      "KL/mean": -84.21513366699219,
      "KL/rejected_KL_mean": -100.16877746582031,
      "KL/std": 44.402587890625,
      "epoch": 0.4671201814058957,
      "fcm_dpo/beta": 0.012739481404423714,
      "fcm_dpo/delta": -0.00707148015499115,
      "fcm_dpo/margin": 31.907291412353516,
      "fcm_dpo/q_t": 0.407754123210907,
      "grad_norm": 12.303749084472656,
      "learning_rate": 3.229664715194511e-07,
      "logits/chosen": 0.744890570640564,
      "logits/rejected": 0.6873106956481934,
      "logps/chosen": -133.65623474121094,
      "logps/ref_chosen": -65.39474487304688,
      "logps/ref_rejected": -75.70930480957031,
      "logps/rejected": -175.87808227539062,
      "loss": 1.1057,
      "margin_dpo/margin_mean": 31.90729331970215,
      "margin_dpo/margin_std": 46.92948913574219,
      "step": 309
    },
    {
      "KL/chosen_KL_mean": -68.58441162109375,
      "KL/mean": -78.29966735839844,
      "KL/rejected_KL_mean": -88.01490020751953,
      "KL/std": 41.900726318359375,
      "epoch": 0.46863189720332576,
      "fcm_dpo/beta": 0.01305387169122696,
      "fcm_dpo/delta": 0.1503462791442871,
      "fcm_dpo/margin": 19.430477142333984,
      "fcm_dpo/q_t": 0.4454615116119385,
      "grad_norm": 14.570418357849121,
      "learning_rate": 3.2170080817777257e-07,
      "logits/chosen": 0.6618015766143799,
      "logits/rejected": 0.6511035561561584,
      "logps/chosen": -143.252685546875,
      "logps/ref_chosen": -74.66827392578125,
      "logps/ref_rejected": -80.5689697265625,
      "logps/rejected": -168.5838623046875,
      "loss": 1.2538,
      "margin_dpo/margin_mean": 19.430479049682617,
      "margin_dpo/margin_std": 51.15632629394531,
      "step": 310
    },
    {
      "KL/chosen_KL_mean": -55.306392669677734,
      "KL/mean": -70.36241149902344,
      "KL/rejected_KL_mean": -85.41842651367188,
      "KL/std": 48.275238037109375,
      "epoch": 0.47014361300075586,
      "fcm_dpo/beta": 0.013104679062962532,
      "fcm_dpo/delta": 0.004920497536659241,
      "fcm_dpo/margin": 30.112035751342773,
      "fcm_dpo/q_t": 0.41163355112075806,
      "grad_norm": 15.74968433380127,
      "learning_rate": 3.204331392103574e-07,
      "logits/chosen": 0.5947407484054565,
      "logits/rejected": 0.46077167987823486,
      "logps/chosen": -115.04442596435547,
      "logps/ref_chosen": -59.738033294677734,
      "logps/ref_rejected": -93.60757446289062,
      "logps/rejected": -179.0260009765625,
      "loss": 1.1241,
      "margin_dpo/margin_mean": 30.112035751342773,
      "margin_dpo/margin_std": 47.993797302246094,
      "step": 311
    },
    {
      "KL/chosen_KL_mean": -57.70798873901367,
      "KL/mean": -77.60112762451172,
      "KL/rejected_KL_mean": -97.4942626953125,
      "KL/std": 46.28860092163086,
      "epoch": 0.47165532879818595,
      "fcm_dpo/beta": 0.012983372434973717,
      "fcm_dpo/delta": -0.12272703647613525,
      "fcm_dpo/margin": 39.786277770996094,
      "fcm_dpo/q_t": 0.3818385899066925,
      "grad_norm": 13.52045726776123,
      "learning_rate": 3.1916350007663176e-07,
      "logits/chosen": 0.7198551893234253,
      "logits/rejected": 0.6226128339767456,
      "logps/chosen": -111.52442169189453,
      "logps/ref_chosen": -53.816436767578125,
      "logps/ref_rejected": -68.6575698852539,
      "logps/rejected": -166.15182495117188,
      "loss": 1.0173,
      "margin_dpo/margin_mean": 39.786277770996094,
      "margin_dpo/margin_std": 45.88895797729492,
      "step": 312
    },
    {
      "KL/chosen_KL_mean": -58.545257568359375,
      "KL/mean": -69.0672836303711,
      "KL/rejected_KL_mean": -79.58930206298828,
      "KL/std": 44.045310974121094,
      "epoch": 0.47316704459561604,
      "fcm_dpo/beta": 0.013112520799040794,
      "fcm_dpo/delta": 0.12759315967559814,
      "fcm_dpo/margin": 21.044044494628906,
      "fcm_dpo/q_t": 0.44002604484558105,
      "grad_norm": 12.471875190734863,
      "learning_rate": 3.178919262911314e-07,
      "logits/chosen": 0.7591273784637451,
      "logits/rejected": 0.738348662853241,
      "logps/chosen": -118.50261688232422,
      "logps/ref_chosen": -59.957359313964844,
      "logps/ref_rejected": -69.31729888916016,
      "logps/rejected": -148.90660095214844,
      "loss": 1.2263,
      "margin_dpo/margin_mean": 21.044044494628906,
      "margin_dpo/margin_std": 49.037994384765625,
      "step": 313
    },
    {
      "KL/chosen_KL_mean": -55.58160400390625,
      "KL/mean": -76.64013671875,
      "KL/rejected_KL_mean": -97.69866180419922,
      "KL/std": 47.22520446777344,
      "epoch": 0.47467876039304613,
      "fcm_dpo/beta": 0.0128698218613863,
      "fcm_dpo/delta": -0.15077346563339233,
      "fcm_dpo/margin": 42.11705017089844,
      "fcm_dpo/q_t": 0.3795948624610901,
      "grad_norm": 12.830547332763672,
      "learning_rate": 3.166184534225087e-07,
      "logits/chosen": 0.6727685928344727,
      "logits/rejected": 0.7047948837280273,
      "logps/chosen": -125.84976196289062,
      "logps/ref_chosen": -70.26815795898438,
      "logps/ref_rejected": -69.23971557617188,
      "logps/rejected": -166.93838500976562,
      "loss": 1.0173,
      "margin_dpo/margin_mean": 42.11705017089844,
      "margin_dpo/margin_std": 51.619964599609375,
      "step": 314
    },
    {
      "KL/chosen_KL_mean": -61.01652145385742,
      "KL/mean": -77.28446960449219,
      "KL/rejected_KL_mean": -93.55242919921875,
      "KL/std": 46.021209716796875,
      "epoch": 0.47619047619047616,
      "fcm_dpo/beta": 0.012844046577811241,
      "fcm_dpo/delta": -0.01923982799053192,
      "fcm_dpo/margin": 32.535919189453125,
      "fcm_dpo/q_t": 0.40471553802490234,
      "grad_norm": 12.773975372314453,
      "learning_rate": 3.1534311709253723e-07,
      "logits/chosen": 0.6400498151779175,
      "logits/rejected": 0.6017441749572754,
      "logps/chosen": -128.81121826171875,
      "logps/ref_chosen": -67.79469299316406,
      "logps/ref_rejected": -74.55148315429688,
      "logps/rejected": -168.10391235351562,
      "loss": 1.1001,
      "margin_dpo/margin_mean": 32.535919189453125,
      "margin_dpo/margin_std": 47.00663757324219,
      "step": 315
    },
    {
      "KL/chosen_KL_mean": -56.92601776123047,
      "KL/mean": -77.34451293945312,
      "KL/rejected_KL_mean": -97.76300811767578,
      "KL/std": 47.411216735839844,
      "epoch": 0.47770219198790626,
      "fcm_dpo/beta": 0.012653250247240067,
      "fcm_dpo/delta": -0.12391935288906097,
      "fcm_dpo/margin": 40.83698654174805,
      "fcm_dpo/q_t": 0.38212883472442627,
      "grad_norm": 13.032649040222168,
      "learning_rate": 3.1406595297511564e-07,
      "logits/chosen": 0.5666062831878662,
      "logits/rejected": 0.4426162838935852,
      "logps/chosen": -112.2145004272461,
      "logps/ref_chosen": -55.288482666015625,
      "logps/ref_rejected": -96.15723419189453,
      "logps/rejected": -193.9202423095703,
      "loss": 1.0177,
      "margin_dpo/margin_mean": 40.83698272705078,
      "margin_dpo/margin_std": 44.17725372314453,
      "step": 316
    },
    {
      "KL/chosen_KL_mean": -54.153907775878906,
      "KL/mean": -75.89384460449219,
      "KL/rejected_KL_mean": -97.6337890625,
      "KL/std": 46.43383026123047,
      "epoch": 0.47921390778533635,
      "fcm_dpo/beta": 0.01215449906885624,
      "fcm_dpo/delta": -0.1363387107849121,
      "fcm_dpo/margin": 43.47986602783203,
      "fcm_dpo/q_t": 0.3797203302383423,
      "grad_norm": 17.4111328125,
      "learning_rate": 3.1278699679526975e-07,
      "logits/chosen": 0.7092440724372864,
      "logits/rejected": 0.6640247106552124,
      "logps/chosen": -108.73529052734375,
      "logps/ref_chosen": -54.58137512207031,
      "logps/ref_rejected": -72.77232360839844,
      "logps/rejected": -170.40611267089844,
      "loss": 1.0127,
      "margin_dpo/margin_mean": 43.47986602783203,
      "margin_dpo/margin_std": 50.48588562011719,
      "step": 317
    },
    {
      "KL/chosen_KL_mean": -62.91188049316406,
      "KL/mean": -77.77674865722656,
      "KL/rejected_KL_mean": -92.64160919189453,
      "KL/std": 49.21617889404297,
      "epoch": 0.48072562358276644,
      "fcm_dpo/beta": 0.012163069099187851,
      "fcm_dpo/delta": 0.03984696790575981,
      "fcm_dpo/margin": 29.729740142822266,
      "fcm_dpo/q_t": 0.4216643273830414,
      "grad_norm": 13.982091903686523,
      "learning_rate": 3.1150628432815336e-07,
      "logits/chosen": 0.7147679328918457,
      "logits/rejected": 0.6473454833030701,
      "logps/chosen": -115.80010986328125,
      "logps/ref_chosen": -52.88822937011719,
      "logps/ref_rejected": -80.63988494873047,
      "logps/rejected": -173.281494140625,
      "loss": 1.1905,
      "margin_dpo/margin_mean": 29.7297420501709,
      "margin_dpo/margin_std": 62.43902587890625,
      "step": 318
    },
    {
      "KL/chosen_KL_mean": -59.110321044921875,
      "KL/mean": -80.11236572265625,
      "KL/rejected_KL_mean": -101.11442565917969,
      "KL/std": 50.182960510253906,
      "epoch": 0.48223733938019653,
      "fcm_dpo/beta": 0.012012695893645287,
      "fcm_dpo/delta": -0.11011025309562683,
      "fcm_dpo/margin": 42.00410461425781,
      "fcm_dpo/q_t": 0.38811129331588745,
      "grad_norm": 14.462267875671387,
      "learning_rate": 3.1022385139804707e-07,
      "logits/chosen": 0.6171474456787109,
      "logits/rejected": 0.6034401655197144,
      "logps/chosen": -123.47364807128906,
      "logps/ref_chosen": -64.36333465576172,
      "logps/ref_rejected": -79.47296142578125,
      "logps/rejected": -180.58737182617188,
      "loss": 1.0538,
      "margin_dpo/margin_mean": 42.00410461425781,
      "margin_dpo/margin_std": 57.84809112548828,
      "step": 319
    },
    {
      "KL/chosen_KL_mean": -57.963951110839844,
      "KL/mean": -74.89822387695312,
      "KL/rejected_KL_mean": -91.83250427246094,
      "KL/std": 50.19336700439453,
      "epoch": 0.4837490551776266,
      "fcm_dpo/beta": 0.01207122951745987,
      "fcm_dpo/delta": -0.012276045978069305,
      "fcm_dpo/margin": 33.86854934692383,
      "fcm_dpo/q_t": 0.4102700352668762,
      "grad_norm": 15.46420955657959,
      "learning_rate": 3.0893973387735683e-07,
      "logits/chosen": 0.575273334980011,
      "logits/rejected": 0.5354819893836975,
      "logps/chosen": -107.52269744873047,
      "logps/ref_chosen": -49.558746337890625,
      "logps/ref_rejected": -71.23444366455078,
      "logps/rejected": -163.06695556640625,
      "loss": 1.1313,
      "margin_dpo/margin_mean": 33.86854934692383,
      "margin_dpo/margin_std": 54.04207229614258,
      "step": 320
    },
    {
      "KL/chosen_KL_mean": -62.31312561035156,
      "KL/mean": -81.66812896728516,
      "KL/rejected_KL_mean": -101.02313232421875,
      "KL/std": 49.185813903808594,
      "epoch": 0.4852607709750567,
      "fcm_dpo/beta": 0.011711912229657173,
      "fcm_dpo/delta": -0.057817112654447556,
      "fcm_dpo/margin": 38.71000671386719,
      "fcm_dpo/q_t": 0.3971693515777588,
      "grad_norm": 19.20660400390625,
      "learning_rate": 3.0765396768561004e-07,
      "logits/chosen": 0.6728286743164062,
      "logits/rejected": 0.6586112976074219,
      "logps/chosen": -114.39839172363281,
      "logps/ref_chosen": -52.08526611328125,
      "logps/ref_rejected": -55.58674621582031,
      "logps/rejected": -156.60987854003906,
      "loss": 1.0955,
      "margin_dpo/margin_mean": 38.71000671386719,
      "margin_dpo/margin_std": 57.03093719482422,
      "step": 321
    },
    {
      "KL/chosen_KL_mean": -73.02706146240234,
      "KL/mean": -96.4334716796875,
      "KL/rejected_KL_mean": -119.83988952636719,
      "KL/std": 53.84697723388672,
      "epoch": 0.48677248677248675,
      "fcm_dpo/beta": 0.011545605957508087,
      "fcm_dpo/delta": -0.14827696979045868,
      "fcm_dpo/margin": 46.81283187866211,
      "fcm_dpo/q_t": 0.3775472640991211,
      "grad_norm": 13.035255432128906,
      "learning_rate": 3.063665887884511e-07,
      "logits/chosen": 0.7389756441116333,
      "logits/rejected": 0.6568834781646729,
      "logps/chosen": -120.43116760253906,
      "logps/ref_chosen": -47.404109954833984,
      "logps/ref_rejected": -73.4260025024414,
      "logps/rejected": -193.26589965820312,
      "loss": 1.0112,
      "margin_dpo/margin_mean": 46.812828063964844,
      "margin_dpo/margin_std": 55.330101013183594,
      "step": 322
    },
    {
      "KL/chosen_KL_mean": -75.228515625,
      "KL/mean": -90.48625183105469,
      "KL/rejected_KL_mean": -105.7439956665039,
      "KL/std": 52.34074020385742,
      "epoch": 0.48828420256991684,
      "fcm_dpo/beta": 0.011556407436728477,
      "fcm_dpo/delta": 0.048632849007844925,
      "fcm_dpo/margin": 30.515487670898438,
      "fcm_dpo/q_t": 0.42401984333992004,
      "grad_norm": 14.605672836303711,
      "learning_rate": 3.0507763319663517e-07,
      "logits/chosen": 0.6270068883895874,
      "logits/rejected": 0.5478232502937317,
      "logps/chosen": -145.23480224609375,
      "logps/ref_chosen": -70.00630187988281,
      "logps/ref_rejected": -86.96690368652344,
      "logps/rejected": -192.71090698242188,
      "loss": 1.1979,
      "margin_dpo/margin_mean": 30.515487670898438,
      "margin_dpo/margin_std": 65.71580505371094,
      "step": 323
    },
    {
      "KL/chosen_KL_mean": -64.62179565429688,
      "KL/mean": -86.31397247314453,
      "KL/rejected_KL_mean": -108.00615692138672,
      "KL/std": 55.89674377441406,
      "epoch": 0.4897959183673469,
      "fcm_dpo/beta": 0.011341476812958717,
      "fcm_dpo/delta": -0.09731088578701019,
      "fcm_dpo/margin": 43.384361267089844,
      "fcm_dpo/q_t": 0.3883642852306366,
      "grad_norm": 18.400531768798828,
      "learning_rate": 3.0378713696502097e-07,
      "logits/chosen": 0.7231769561767578,
      "logits/rejected": 0.6682534217834473,
      "logps/chosen": -120.51061248779297,
      "logps/ref_chosen": -55.88882064819336,
      "logps/ref_rejected": -75.23088073730469,
      "logps/rejected": -183.23703002929688,
      "loss": 1.0393,
      "margin_dpo/margin_mean": 43.384361267089844,
      "margin_dpo/margin_std": 53.82892608642578,
      "step": 324
    },
    {
      "KL/chosen_KL_mean": -81.2076416015625,
      "KL/mean": -100.78556823730469,
      "KL/rejected_KL_mean": -120.36351013183594,
      "KL/std": 53.15364074707031,
      "epoch": 0.491307634164777,
      "fcm_dpo/beta": 0.011201553046703339,
      "fcm_dpo/delta": -0.040966056287288666,
      "fcm_dpo/margin": 39.15586853027344,
      "fcm_dpo/q_t": 0.4002138674259186,
      "grad_norm": 15.490608215332031,
      "learning_rate": 3.0249513619156206e-07,
      "logits/chosen": 0.6761427521705627,
      "logits/rejected": 0.6141278743743896,
      "logps/chosen": -145.3546600341797,
      "logps/ref_chosen": -64.14701843261719,
      "logps/ref_rejected": -79.91143798828125,
      "logps/rejected": -200.27493286132812,
      "loss": 1.1039,
      "margin_dpo/margin_mean": 39.15586471557617,
      "margin_dpo/margin_std": 60.82697296142578,
      "step": 325
    },
    {
      "KL/chosen_KL_mean": -95.6961898803711,
      "KL/mean": -105.70980834960938,
      "KL/rejected_KL_mean": -115.72342681884766,
      "KL/std": 55.04910659790039,
      "epoch": 0.4928193499622071,
      "fcm_dpo/beta": 0.011495251208543777,
      "fcm_dpo/delta": 0.17409512400627136,
      "fcm_dpo/margin": 20.02724266052246,
      "fcm_dpo/q_t": 0.4493880867958069,
      "grad_norm": 14.850746154785156,
      "learning_rate": 3.012016670162977e-07,
      "logits/chosen": 0.6151560544967651,
      "logits/rejected": 0.6213278770446777,
      "logps/chosen": -171.22750854492188,
      "logps/ref_chosen": -75.53131103515625,
      "logps/ref_rejected": -76.5898666381836,
      "logps/rejected": -192.31329345703125,
      "loss": 1.2868,
      "margin_dpo/margin_mean": 20.027244567871094,
      "margin_dpo/margin_std": 60.19707107543945,
      "step": 326
    },
    {
      "KL/chosen_KL_mean": -91.24993896484375,
      "KL/mean": -106.86180114746094,
      "KL/rejected_KL_mean": -122.47364807128906,
      "KL/std": 56.48907470703125,
      "epoch": 0.4943310657596372,
      "fcm_dpo/beta": 0.011725237593054771,
      "fcm_dpo/delta": 0.03484252840280533,
      "fcm_dpo/margin": 31.223709106445312,
      "fcm_dpo/q_t": 0.4198496341705322,
      "grad_norm": 17.629716873168945,
      "learning_rate": 2.99906765620341e-07,
      "logits/chosen": 0.5963134765625,
      "logits/rejected": 0.5632212162017822,
      "logps/chosen": -160.58709716796875,
      "logps/ref_chosen": -69.33717346191406,
      "logps/ref_rejected": -73.37751770019531,
      "logps/rejected": -195.85116577148438,
      "loss": 1.1785,
      "margin_dpo/margin_mean": 31.22370719909668,
      "margin_dpo/margin_std": 62.13964080810547,
      "step": 327
    },
    {
      "KL/chosen_KL_mean": -78.11366271972656,
      "KL/mean": -97.55792236328125,
      "KL/rejected_KL_mean": -117.00218200683594,
      "KL/std": 57.609336853027344,
      "epoch": 0.4958427815570673,
      "fcm_dpo/beta": 0.011624249629676342,
      "fcm_dpo/delta": -0.05448012053966522,
      "fcm_dpo/margin": 38.888511657714844,
      "fcm_dpo/q_t": 0.39957690238952637,
      "grad_norm": 13.897930145263672,
      "learning_rate": 2.9861046822486766e-07,
      "logits/chosen": 0.6323528289794922,
      "logits/rejected": 0.6021204590797424,
      "logps/chosen": -139.81988525390625,
      "logps/ref_chosen": -61.70623016357422,
      "logps/ref_rejected": -83.73808288574219,
      "logps/rejected": -200.74026489257812,
      "loss": 1.0849,
      "margin_dpo/margin_mean": 38.888511657714844,
      "margin_dpo/margin_std": 56.55249786376953,
      "step": 328
    },
    {
      "KL/chosen_KL_mean": -87.73931884765625,
      "KL/mean": -107.61006164550781,
      "KL/rejected_KL_mean": -127.48081970214844,
      "KL/std": 55.76243209838867,
      "epoch": 0.4973544973544973,
      "fcm_dpo/beta": 0.01148184947669506,
      "fcm_dpo/delta": -0.05898508056998253,
      "fcm_dpo/margin": 39.74150085449219,
      "fcm_dpo/q_t": 0.3992629051208496,
      "grad_norm": 18.06145477294922,
      "learning_rate": 2.9731281109010253e-07,
      "logits/chosen": 0.6991169452667236,
      "logits/rejected": 0.6458307504653931,
      "logps/chosen": -152.23773193359375,
      "logps/ref_chosen": -64.4984130859375,
      "logps/ref_rejected": -83.6591796875,
      "logps/rejected": -211.13998413085938,
      "loss": 1.0842,
      "margin_dpo/margin_mean": 39.74150085449219,
      "margin_dpo/margin_std": 58.48515701293945,
      "step": 329
    },
    {
      "KL/chosen_KL_mean": -74.7954330444336,
      "KL/mean": -96.87043762207031,
      "KL/rejected_KL_mean": -118.9454345703125,
      "KL/std": 56.645729064941406,
      "epoch": 0.4988662131519274,
      "fcm_dpo/beta": 0.01131785660982132,
      "fcm_dpo/delta": -0.10486404597759247,
      "fcm_dpo/margin": 44.15000534057617,
      "fcm_dpo/q_t": 0.3897658586502075,
      "grad_norm": 13.613192558288574,
      "learning_rate": 2.9601383051430505e-07,
      "logits/chosen": 0.7008596062660217,
      "logits/rejected": 0.6330828666687012,
      "logps/chosen": -129.60006713867188,
      "logps/ref_chosen": -54.80464172363281,
      "logps/ref_rejected": -75.3194351196289,
      "logps/rejected": -194.26486206054688,
      "loss": 1.0977,
      "margin_dpo/margin_mean": 44.15000534057617,
      "margin_dpo/margin_std": 70.77532196044922,
      "step": 330
    },
    {
      "KL/chosen_KL_mean": -81.3387451171875,
      "KL/mean": -108.13944244384766,
      "KL/rejected_KL_mean": -134.94012451171875,
      "KL/std": 59.597442626953125,
      "epoch": 0.5003779289493575,
      "fcm_dpo/beta": 0.0109153613448143,
      "fcm_dpo/delta": -0.196200892329216,
      "fcm_dpo/margin": 53.60139465332031,
      "fcm_dpo/q_t": 0.368771493434906,
      "grad_norm": 13.300724029541016,
      "learning_rate": 2.947135628327544e-07,
      "logits/chosen": 0.8084205985069275,
      "logits/rejected": 0.7829192876815796,
      "logps/chosen": -140.58132934570312,
      "logps/ref_chosen": -59.242584228515625,
      "logps/ref_rejected": -69.87483215332031,
      "logps/rejected": -204.81497192382812,
      "loss": 1.0072,
      "margin_dpo/margin_mean": 53.60139465332031,
      "margin_dpo/margin_std": 66.9174575805664,
      "step": 331
    },
    {
      "KL/chosen_KL_mean": -82.68574523925781,
      "KL/mean": -104.79304504394531,
      "KL/rejected_KL_mean": -126.90032958984375,
      "KL/std": 56.367454528808594,
      "epoch": 0.5018896447467877,
      "fcm_dpo/beta": 0.010739924386143684,
      "fcm_dpo/delta": -0.07937489449977875,
      "fcm_dpo/margin": 44.21459197998047,
      "fcm_dpo/q_t": 0.393557608127594,
      "grad_norm": 14.145315170288086,
      "learning_rate": 2.934120444167326e-07,
      "logits/chosen": 0.6289262175559998,
      "logits/rejected": 0.5868539810180664,
      "logps/chosen": -149.79550170898438,
      "logps/ref_chosen": -67.10975646972656,
      "logps/ref_rejected": -77.11839294433594,
      "logps/rejected": -204.0187225341797,
      "loss": 1.0615,
      "margin_dpo/margin_mean": 44.21459197998047,
      "margin_dpo/margin_std": 57.61412048339844,
      "step": 332
    },
    {
      "KL/chosen_KL_mean": -87.93297576904297,
      "KL/mean": -111.47476196289062,
      "KL/rejected_KL_mean": -135.0165557861328,
      "KL/std": 56.53441619873047,
      "epoch": 0.5034013605442177,
      "fcm_dpo/beta": 0.010472394526004791,
      "fcm_dpo/delta": -0.09779446572065353,
      "fcm_dpo/margin": 47.08356475830078,
      "fcm_dpo/q_t": 0.3887941241264343,
      "grad_norm": 12.955265998840332,
      "learning_rate": 2.921093116725076e-07,
      "logits/chosen": 0.6959058046340942,
      "logits/rejected": 0.6248373985290527,
      "logps/chosen": -146.31411743164062,
      "logps/ref_chosen": -58.381134033203125,
      "logps/ref_rejected": -85.02839660644531,
      "logps/rejected": -220.04495239257812,
      "loss": 1.0398,
      "margin_dpo/margin_mean": 47.08356475830078,
      "margin_dpo/margin_std": 58.59512710571289,
      "step": 333
    },
    {
      "KL/chosen_KL_mean": -83.53094482421875,
      "KL/mean": -100.3437271118164,
      "KL/rejected_KL_mean": -117.15650177001953,
      "KL/std": 58.58315658569336,
      "epoch": 0.5049130763416477,
      "fcm_dpo/beta": 0.010494356974959373,
      "fcm_dpo/delta": 0.04870123788714409,
      "fcm_dpo/margin": 33.62556457519531,
      "fcm_dpo/q_t": 0.42179661989212036,
      "grad_norm": 13.383609771728516,
      "learning_rate": 2.9080540104031484e-07,
      "logits/chosen": 0.7296691536903381,
      "logits/rejected": 0.685336172580719,
      "logps/chosen": -150.42294311523438,
      "logps/ref_chosen": -66.89199829101562,
      "logps/ref_rejected": -91.83695220947266,
      "logps/rejected": -208.9934539794922,
      "loss": 1.1831,
      "margin_dpo/margin_mean": 33.62556457519531,
      "margin_dpo/margin_std": 68.05831146240234,
      "step": 334
    },
    {
      "KL/chosen_KL_mean": -85.08909606933594,
      "KL/mean": -104.37361145019531,
      "KL/rejected_KL_mean": -123.65814208984375,
      "KL/std": 57.88601303100586,
      "epoch": 0.5064247921390779,
      "fcm_dpo/beta": 0.010544189251959324,
      "fcm_dpo/delta": -0.007449910044670105,
      "fcm_dpo/margin": 38.56904220581055,
      "fcm_dpo/q_t": 0.41014015674591064,
      "grad_norm": 19.125410079956055,
      "learning_rate": 2.895003489933375e-07,
      "logits/chosen": 0.720893383026123,
      "logits/rejected": 0.6842397451400757,
      "logps/chosen": -146.60354614257812,
      "logps/ref_chosen": -61.51445770263672,
      "logps/ref_rejected": -75.68916320800781,
      "logps/rejected": -199.3472900390625,
      "loss": 1.131,
      "margin_dpo/margin_mean": 38.56904220581055,
      "margin_dpo/margin_std": 64.91070556640625,
      "step": 335
    },
    {
      "KL/chosen_KL_mean": -94.51179504394531,
      "KL/mean": -114.01209259033203,
      "KL/rejected_KL_mean": -133.51239013671875,
      "KL/std": 58.32542037963867,
      "epoch": 0.5079365079365079,
      "fcm_dpo/beta": 0.010401548817753792,
      "fcm_dpo/delta": -0.006995633244514465,
      "fcm_dpo/margin": 39.00059127807617,
      "fcm_dpo/q_t": 0.4102246165275574,
      "grad_norm": 12.244279861450195,
      "learning_rate": 2.8819419203668675e-07,
      "logits/chosen": 0.6266505122184753,
      "logits/rejected": 0.6072256565093994,
      "logps/chosen": -163.3618621826172,
      "logps/ref_chosen": -68.85006713867188,
      "logps/ref_rejected": -92.99603271484375,
      "logps/rejected": -226.5084228515625,
      "loss": 1.1214,
      "margin_dpo/margin_mean": 39.00059509277344,
      "margin_dpo/margin_std": 61.9708251953125,
      "step": 336
    },
    {
      "KL/chosen_KL_mean": -97.12661743164062,
      "KL/mean": -112.62835693359375,
      "KL/rejected_KL_mean": -128.13009643554688,
      "KL/std": 57.338382720947266,
      "epoch": 0.509448223733938,
      "fcm_dpo/beta": 0.010596680454909801,
      "fcm_dpo/delta": 0.07382012903690338,
      "fcm_dpo/margin": 31.003463745117188,
      "fcm_dpo/q_t": 0.4254772961139679,
      "grad_norm": 12.672013282775879,
      "learning_rate": 2.8688696670638053e-07,
      "logits/chosen": 0.5881543159484863,
      "logits/rejected": 0.5562857389450073,
      "logps/chosen": -170.314453125,
      "logps/ref_chosen": -73.18783569335938,
      "logps/ref_rejected": -86.89118957519531,
      "logps/rejected": -215.0212860107422,
      "loss": 1.1771,
      "margin_dpo/margin_mean": 31.003459930419922,
      "margin_dpo/margin_std": 59.29844284057617,
      "step": 337
    },
    {
      "KL/chosen_KL_mean": -92.96856689453125,
      "KL/mean": -109.99942016601562,
      "KL/rejected_KL_mean": -127.03028869628906,
      "KL/std": 57.12294006347656,
      "epoch": 0.5109599395313681,
      "fcm_dpo/beta": 0.010666831396520138,
      "fcm_dpo/delta": 0.038046009838581085,
      "fcm_dpo/margin": 34.061710357666016,
      "fcm_dpo/q_t": 0.41940271854400635,
      "grad_norm": 12.995147705078125,
      "learning_rate": 2.8557870956832133e-07,
      "logits/chosen": 0.6552461385726929,
      "logits/rejected": 0.6303938627243042,
      "logps/chosen": -156.90817260742188,
      "logps/ref_chosen": -63.939613342285156,
      "logps/ref_rejected": -75.34243774414062,
      "logps/rejected": -202.3727264404297,
      "loss": 1.1656,
      "margin_dpo/margin_mean": 34.061710357666016,
      "margin_dpo/margin_std": 64.4132308959961,
      "step": 338
    },
    {
      "KL/chosen_KL_mean": -76.82237243652344,
      "KL/mean": -95.09283447265625,
      "KL/rejected_KL_mean": -113.3633041381836,
      "KL/std": 55.23368835449219,
      "epoch": 0.5124716553287982,
      "fcm_dpo/beta": 0.010703757405281067,
      "fcm_dpo/delta": 0.009182396344840527,
      "fcm_dpo/margin": 36.54094696044922,
      "fcm_dpo/q_t": 0.4115958511829376,
      "grad_norm": 12.69047737121582,
      "learning_rate": 2.842694572172736e-07,
      "logits/chosen": 0.7880140542984009,
      "logits/rejected": 0.7043054103851318,
      "logps/chosen": -122.37149810791016,
      "logps/ref_chosen": -45.54913330078125,
      "logps/ref_rejected": -67.0482177734375,
      "logps/rejected": -180.41152954101562,
      "loss": 1.1228,
      "margin_dpo/margin_mean": 36.54094696044922,
      "margin_dpo/margin_std": 57.50127029418945,
      "step": 339
    },
    {
      "KL/chosen_KL_mean": -86.79705047607422,
      "KL/mean": -106.69656372070312,
      "KL/rejected_KL_mean": -126.59608459472656,
      "KL/std": 61.46400451660156,
      "epoch": 0.5139833711262283,
      "fcm_dpo/beta": 0.010734757408499718,
      "fcm_dpo/delta": -0.02878701686859131,
      "fcm_dpo/margin": 39.79903030395508,
      "fcm_dpo/q_t": 0.4072033762931824,
      "grad_norm": 12.821864128112793,
      "learning_rate": 2.8295924627584004e-07,
      "logits/chosen": 0.6960463523864746,
      "logits/rejected": 0.6766098141670227,
      "logps/chosen": -140.80270385742188,
      "logps/ref_chosen": -54.00564956665039,
      "logps/ref_rejected": -61.314430236816406,
      "logps/rejected": -187.9105224609375,
      "loss": 1.1353,
      "margin_dpo/margin_mean": 39.79903030395508,
      "margin_dpo/margin_std": 69.87727355957031,
      "step": 340
    },
    {
      "KL/chosen_KL_mean": -84.44651794433594,
      "KL/mean": -106.48483276367188,
      "KL/rejected_KL_mean": -128.52316284179688,
      "KL/std": 59.11941909790039,
      "epoch": 0.5154950869236583,
      "fcm_dpo/beta": 0.010431801900267601,
      "fcm_dpo/delta": -0.06666833162307739,
      "fcm_dpo/margin": 44.07664108276367,
      "fcm_dpo/q_t": 0.39677852392196655,
      "grad_norm": 13.550869941711426,
      "learning_rate": 2.816481133934373e-07,
      "logits/chosen": 0.7298463582992554,
      "logits/rejected": 0.6815722584724426,
      "logps/chosen": -147.84161376953125,
      "logps/ref_chosen": -63.39509582519531,
      "logps/ref_rejected": -76.20973205566406,
      "logps/rejected": -204.73287963867188,
      "loss": 1.0912,
      "margin_dpo/margin_mean": 44.07664108276367,
      "margin_dpo/margin_std": 62.568634033203125,
      "step": 341
    },
    {
      "KL/chosen_KL_mean": -83.20333099365234,
      "KL/mean": -106.3760986328125,
      "KL/rejected_KL_mean": -129.5488739013672,
      "KL/std": 61.28248596191406,
      "epoch": 0.5170068027210885,
      "fcm_dpo/beta": 0.010350107215344906,
      "fcm_dpo/delta": -0.08443048596382141,
      "fcm_dpo/margin": 46.345550537109375,
      "fcm_dpo/q_t": 0.39365118741989136,
      "grad_norm": 11.774492263793945,
      "learning_rate": 2.8033609524527046e-07,
      "logits/chosen": 0.7456991672515869,
      "logits/rejected": 0.7070844769477844,
      "logps/chosen": -136.2511444091797,
      "logps/ref_chosen": -53.047813415527344,
      "logps/ref_rejected": -68.2854232788086,
      "logps/rejected": -197.83428955078125,
      "loss": 1.0632,
      "margin_dpo/margin_mean": 46.345550537109375,
      "margin_dpo/margin_std": 63.2640495300293,
      "step": 342
    },
    {
      "KL/chosen_KL_mean": -79.01636505126953,
      "KL/mean": -93.48959350585938,
      "KL/rejected_KL_mean": -107.96282958984375,
      "KL/std": 57.215553283691406,
      "epoch": 0.5185185185185185,
      "fcm_dpo/beta": 0.010420860722661018,
      "fcm_dpo/delta": 0.10085594654083252,
      "fcm_dpo/margin": 28.94646644592285,
      "fcm_dpo/q_t": 0.43098893761634827,
      "grad_norm": 12.47730541229248,
      "learning_rate": 2.7902322853130753e-07,
      "logits/chosen": 0.6005036234855652,
      "logits/rejected": 0.5945264101028442,
      "logps/chosen": -149.5948944091797,
      "logps/ref_chosen": -70.57852935791016,
      "logps/ref_rejected": -84.73873901367188,
      "logps/rejected": -192.70156860351562,
      "loss": 1.1988,
      "margin_dpo/margin_mean": 28.946468353271484,
      "margin_dpo/margin_std": 57.34397888183594,
      "step": 343
    },
    {
      "KL/chosen_KL_mean": -86.4252700805664,
      "KL/mean": -109.1060791015625,
      "KL/rejected_KL_mean": -131.78689575195312,
      "KL/std": 58.18048095703125,
      "epoch": 0.5200302343159486,
      "fcm_dpo/beta": 0.010408244095742702,
      "fcm_dpo/delta": -0.07615945488214493,
      "fcm_dpo/margin": 45.36162567138672,
      "fcm_dpo/q_t": 0.3939076066017151,
      "grad_norm": 14.552360534667969,
      "learning_rate": 2.7770954997525274e-07,
      "logits/chosen": 0.715709388256073,
      "logits/rejected": 0.6526391506195068,
      "logps/chosen": -142.23626708984375,
      "logps/ref_chosen": -55.811004638671875,
      "logps/ref_rejected": -84.77637481689453,
      "logps/rejected": -216.56326293945312,
      "loss": 1.061,
      "margin_dpo/margin_mean": 45.36162567138672,
      "margin_dpo/margin_std": 60.71054458618164,
      "step": 344
    },
    {
      "KL/chosen_KL_mean": -67.61167907714844,
      "KL/mean": -87.00637817382812,
      "KL/rejected_KL_mean": -106.40106201171875,
      "KL/std": 51.96855545043945,
      "epoch": 0.5215419501133787,
      "fcm_dpo/beta": 0.01042355690151453,
      "fcm_dpo/delta": -0.004661137238144875,
      "fcm_dpo/margin": 38.78938674926758,
      "fcm_dpo/q_t": 0.40908119082450867,
      "grad_norm": 13.888686180114746,
      "learning_rate": 2.7639509632351927e-07,
      "logits/chosen": 0.7611949443817139,
      "logits/rejected": 0.7161680459976196,
      "logps/chosen": -125.39778137207031,
      "logps/ref_chosen": -57.78609848022461,
      "logps/ref_rejected": -78.91847229003906,
      "logps/rejected": -185.3195343017578,
      "loss": 1.1181,
      "margin_dpo/margin_mean": 38.78938674926758,
      "margin_dpo/margin_std": 61.487510681152344,
      "step": 345
    },
    {
      "KL/chosen_KL_mean": -75.70014953613281,
      "KL/mean": -98.07574462890625,
      "KL/rejected_KL_mean": -120.45133972167969,
      "KL/std": 59.482688903808594,
      "epoch": 0.5230536659108088,
      "fcm_dpo/beta": 0.010338194668293,
      "fcm_dpo/delta": -0.06582339853048325,
      "fcm_dpo/margin": 44.75118637084961,
      "fcm_dpo/q_t": 0.3961416482925415,
      "grad_norm": 14.283039093017578,
      "learning_rate": 2.7507990434420123e-07,
      "logits/chosen": 0.7346894145011902,
      "logits/rejected": 0.6535829305648804,
      "logps/chosen": -131.9852752685547,
      "logps/ref_chosen": -56.285125732421875,
      "logps/ref_rejected": -91.15303039550781,
      "logps/rejected": -211.6043701171875,
      "loss": 1.084,
      "margin_dpo/margin_mean": 44.751190185546875,
      "margin_dpo/margin_std": 64.32740020751953,
      "step": 346
    },
    {
      "KL/chosen_KL_mean": -82.76616668701172,
      "KL/mean": -99.7265625,
      "KL/rejected_KL_mean": -116.68695068359375,
      "KL/std": 58.51282501220703,
      "epoch": 0.5245653817082389,
      "fcm_dpo/beta": 0.01033782958984375,
      "fcm_dpo/delta": 0.05110158771276474,
      "fcm_dpo/margin": 33.920780181884766,
      "fcm_dpo/q_t": 0.4226782023906708,
      "grad_norm": 17.104528427124023,
      "learning_rate": 2.737640108260456e-07,
      "logits/chosen": 0.8117585182189941,
      "logits/rejected": 0.7625047564506531,
      "logps/chosen": -136.26571655273438,
      "logps/ref_chosen": -53.499542236328125,
      "logps/ref_rejected": -72.52565002441406,
      "logps/rejected": -189.21261596679688,
      "loss": 1.1607,
      "margin_dpo/margin_mean": 33.920780181884766,
      "margin_dpo/margin_std": 61.820030212402344,
      "step": 347
    },
    {
      "KL/chosen_KL_mean": -73.69622039794922,
      "KL/mean": -95.5750732421875,
      "KL/rejected_KL_mean": -117.45392608642578,
      "KL/std": 57.17124938964844,
      "epoch": 0.5260770975056689,
      "fcm_dpo/beta": 0.010252213105559349,
      "fcm_dpo/delta": -0.051215190440416336,
      "fcm_dpo/margin": 43.75770950317383,
      "fcm_dpo/q_t": 0.40280699729919434,
      "grad_norm": 12.792646408081055,
      "learning_rate": 2.724474525774229e-07,
      "logits/chosen": 0.816604495048523,
      "logits/rejected": 0.7879601716995239,
      "logps/chosen": -124.48307037353516,
      "logps/ref_chosen": -50.78684997558594,
      "logps/ref_rejected": -68.63732147216797,
      "logps/rejected": -186.09124755859375,
      "loss": 1.1036,
      "margin_dpo/margin_mean": 43.75770568847656,
      "margin_dpo/margin_std": 69.70893859863281,
      "step": 348
    },
    {
      "KL/chosen_KL_mean": -72.48434448242188,
      "KL/mean": -94.19596099853516,
      "KL/rejected_KL_mean": -115.9075927734375,
      "KL/std": 58.99268341064453,
      "epoch": 0.527588813303099,
      "fcm_dpo/beta": 0.010226529091596603,
      "fcm_dpo/delta": -0.046270184218883514,
      "fcm_dpo/margin": 43.42324447631836,
      "fcm_dpo/q_t": 0.4015776515007019,
      "grad_norm": 14.216320037841797,
      "learning_rate": 2.711302664252973e-07,
      "logits/chosen": 0.7372004985809326,
      "logits/rejected": 0.646864652633667,
      "logps/chosen": -125.80935668945312,
      "logps/ref_chosen": -53.325008392333984,
      "logps/ref_rejected": -83.21236419677734,
      "logps/rejected": -199.11996459960938,
      "loss": 1.09,
      "margin_dpo/margin_mean": 43.42324447631836,
      "margin_dpo/margin_std": 63.5445556640625,
      "step": 349
    },
    {
      "KL/chosen_KL_mean": -78.7496109008789,
      "KL/mean": -104.87985229492188,
      "KL/rejected_KL_mean": -131.0100860595703,
      "KL/std": 62.48160171508789,
      "epoch": 0.5291005291005291,
      "fcm_dpo/beta": 0.009954242035746574,
      "fcm_dpo/delta": -0.1270718276500702,
      "fcm_dpo/margin": 52.26048278808594,
      "fcm_dpo/q_t": 0.38217905163764954,
      "grad_norm": 15.84421157836914,
      "learning_rate": 2.698124892141971e-07,
      "logits/chosen": 0.6647555828094482,
      "logits/rejected": 0.5865734219551086,
      "logps/chosen": -140.37538146972656,
      "logps/ref_chosen": -61.625770568847656,
      "logps/ref_rejected": -87.63627624511719,
      "logps/rejected": -218.6463623046875,
      "loss": 1.0268,
      "margin_dpo/margin_mean": 52.26047897338867,
      "margin_dpo/margin_std": 64.52607727050781,
      "step": 350
    },
    {
      "KL/chosen_KL_mean": -74.44353485107422,
      "KL/mean": -96.31437683105469,
      "KL/rejected_KL_mean": -118.18523406982422,
      "KL/std": 56.73322677612305,
      "epoch": 0.5306122448979592,
      "fcm_dpo/beta": 0.009825417771935463,
      "fcm_dpo/delta": -0.03146745264530182,
      "fcm_dpo/margin": 43.74169921875,
      "fcm_dpo/q_t": 0.4013119041919708,
      "grad_norm": 13.427343368530273,
      "learning_rate": 2.6849415780518357e-07,
      "logits/chosen": 0.6526576280593872,
      "logits/rejected": 0.5777317881584167,
      "logps/chosen": -130.6998748779297,
      "logps/ref_chosen": -56.2563362121582,
      "logps/ref_rejected": -79.11589813232422,
      "logps/rejected": -197.30113220214844,
      "loss": 1.1167,
      "margin_dpo/margin_mean": 43.74169921875,
      "margin_dpo/margin_std": 70.45606231689453,
      "step": 351
    },
    {
      "KL/chosen_KL_mean": -72.97793579101562,
      "KL/mean": -95.73782348632812,
      "KL/rejected_KL_mean": -118.4977035522461,
      "KL/std": 58.78670883178711,
      "epoch": 0.5321239606953893,
      "fcm_dpo/beta": 0.009755909442901611,
      "fcm_dpo/delta": -0.04631410166621208,
      "fcm_dpo/margin": 45.51976776123047,
      "fcm_dpo/q_t": 0.3998575806617737,
      "grad_norm": 13.304686546325684,
      "learning_rate": 2.6717530907482024e-07,
      "logits/chosen": 0.7246947288513184,
      "logits/rejected": 0.672113835811615,
      "logps/chosen": -136.02987670898438,
      "logps/ref_chosen": -63.05195236206055,
      "logps/ref_rejected": -85.52035522460938,
      "logps/rejected": -204.01806640625,
      "loss": 1.0803,
      "margin_dpo/margin_mean": 45.51976776123047,
      "margin_dpo/margin_std": 63.94942855834961,
      "step": 352
    },
    {
      "KL/chosen_KL_mean": -70.28274536132812,
      "KL/mean": -93.31478881835938,
      "KL/rejected_KL_mean": -116.3468017578125,
      "KL/std": 58.474891662597656,
      "epoch": 0.5336356764928194,
      "fcm_dpo/beta": 0.009713239967823029,
      "fcm_dpo/delta": -0.04971469193696976,
      "fcm_dpo/margin": 46.06404495239258,
      "fcm_dpo/q_t": 0.39870840311050415,
      "grad_norm": 11.444036483764648,
      "learning_rate": 2.658559799141411e-07,
      "logits/chosen": 0.7031028270721436,
      "logits/rejected": 0.7094468474388123,
      "logps/chosen": -139.29193115234375,
      "logps/ref_chosen": -69.00918579101562,
      "logps/ref_rejected": -72.65840148925781,
      "logps/rejected": -189.0052032470703,
      "loss": 1.083,
      "margin_dpo/margin_mean": 46.064048767089844,
      "margin_dpo/margin_std": 65.06970977783203,
      "step": 353
    },
    {
      "KL/chosen_KL_mean": -74.60330963134766,
      "KL/mean": -99.59640502929688,
      "KL/rejected_KL_mean": -124.5895004272461,
      "KL/std": 57.67888641357422,
      "epoch": 0.5351473922902494,
      "fcm_dpo/beta": 0.009495913982391357,
      "fcm_dpo/delta": -0.07927623391151428,
      "fcm_dpo/margin": 49.98619079589844,
      "fcm_dpo/q_t": 0.3928118348121643,
      "grad_norm": 13.079569816589355,
      "learning_rate": 2.6453620722761895e-07,
      "logits/chosen": 0.7806761264801025,
      "logits/rejected": 0.6481184959411621,
      "logps/chosen": -114.39164733886719,
      "logps/ref_chosen": -39.78833770751953,
      "logps/ref_rejected": -69.56885528564453,
      "logps/rejected": -194.15835571289062,
      "loss": 1.0723,
      "margin_dpo/margin_mean": 49.98619079589844,
      "margin_dpo/margin_std": 69.9796142578125,
      "step": 354
    },
    {
      "KL/chosen_KL_mean": -78.99346923828125,
      "KL/mean": -104.57441711425781,
      "KL/rejected_KL_mean": -130.1553497314453,
      "KL/std": 62.086181640625,
      "epoch": 0.5366591080876795,
      "fcm_dpo/beta": 0.00941452570259571,
      "fcm_dpo/delta": -0.0857347846031189,
      "fcm_dpo/margin": 51.16187286376953,
      "fcm_dpo/q_t": 0.3914402723312378,
      "grad_norm": 15.902485847473145,
      "learning_rate": 2.632160279321328e-07,
      "logits/chosen": 0.7517096996307373,
      "logits/rejected": 0.6187185645103455,
      "logps/chosen": -125.24885559082031,
      "logps/ref_chosen": -46.25537872314453,
      "logps/ref_rejected": -78.20236206054688,
      "logps/rejected": -208.3577117919922,
      "loss": 1.0716,
      "margin_dpo/margin_mean": 51.16187286376953,
      "margin_dpo/margin_std": 72.24238586425781,
      "step": 355
    },
    {
      "KL/chosen_KL_mean": -73.79627990722656,
      "KL/mean": -94.85540771484375,
      "KL/rejected_KL_mean": -115.91452026367188,
      "KL/std": 61.34975051879883,
      "epoch": 0.5381708238851096,
      "fcm_dpo/beta": 0.00931059941649437,
      "fcm_dpo/delta": 0.007776662707328796,
      "fcm_dpo/margin": 42.118247985839844,
      "fcm_dpo/q_t": 0.4127386808395386,
      "grad_norm": 12.62276554107666,
      "learning_rate": 2.618954789559356e-07,
      "logits/chosen": 0.7203996181488037,
      "logits/rejected": 0.64216548204422,
      "logps/chosen": -121.70243835449219,
      "logps/ref_chosen": -47.906158447265625,
      "logps/ref_rejected": -74.29397583007812,
      "logps/rejected": -190.20849609375,
      "loss": 1.1675,
      "margin_dpo/margin_mean": 42.118247985839844,
      "margin_dpo/margin_std": 81.22731018066406,
      "step": 356
    },
    {
      "KL/chosen_KL_mean": -89.2645034790039,
      "KL/mean": -108.21392822265625,
      "KL/rejected_KL_mean": -127.16334533691406,
      "KL/std": 59.75215530395508,
      "epoch": 0.5396825396825397,
      "fcm_dpo/beta": 0.009327213279902935,
      "fcm_dpo/delta": 0.04655788093805313,
      "fcm_dpo/margin": 37.898834228515625,
      "fcm_dpo/q_t": 0.41907134652137756,
      "grad_norm": 12.688922882080078,
      "learning_rate": 2.6057459723762076e-07,
      "logits/chosen": 0.6992335319519043,
      "logits/rejected": 0.6748742461204529,
      "logps/chosen": -151.89950561523438,
      "logps/ref_chosen": -62.63500213623047,
      "logps/ref_rejected": -65.11399841308594,
      "logps/rejected": -192.27734375,
      "loss": 1.1569,
      "margin_dpo/margin_mean": 37.898834228515625,
      "margin_dpo/margin_std": 63.26298904418945,
      "step": 357
    },
    {
      "KL/chosen_KL_mean": -83.01390075683594,
      "KL/mean": -109.11406707763672,
      "KL/rejected_KL_mean": -135.2142333984375,
      "KL/std": 60.27477264404297,
      "epoch": 0.5411942554799698,
      "fcm_dpo/beta": 0.009361564181745052,
      "fcm_dpo/delta": -0.0932769775390625,
      "fcm_dpo/margin": 52.2003288269043,
      "fcm_dpo/q_t": 0.3902199864387512,
      "grad_norm": 15.763985633850098,
      "learning_rate": 2.5925341972508954e-07,
      "logits/chosen": 0.6433045864105225,
      "logits/rejected": 0.6595355272293091,
      "logps/chosen": -150.2235107421875,
      "logps/ref_chosen": -67.20960998535156,
      "logps/ref_rejected": -69.34715270996094,
      "logps/rejected": -204.5614013671875,
      "loss": 1.0522,
      "margin_dpo/margin_mean": 52.20033264160156,
      "margin_dpo/margin_std": 68.1099853515625,
      "step": 358
    },
    {
      "KL/chosen_KL_mean": -94.01116943359375,
      "KL/mean": -106.93074035644531,
      "KL/rejected_KL_mean": -119.85032653808594,
      "KL/std": 60.080284118652344,
      "epoch": 0.5427059712773998,
      "fcm_dpo/beta": 0.009553344920277596,
      "fcm_dpo/delta": 0.15497523546218872,
      "fcm_dpo/margin": 25.83917236328125,
      "fcm_dpo/q_t": 0.4445451498031616,
      "grad_norm": 13.971330642700195,
      "learning_rate": 2.579319833745169e-07,
      "logits/chosen": 0.6465896368026733,
      "logits/rejected": 0.6198573112487793,
      "logps/chosen": -156.53695678710938,
      "logps/ref_chosen": -62.52578353881836,
      "logps/ref_rejected": -76.63114929199219,
      "logps/rejected": -196.48147583007812,
      "loss": 1.2395,
      "margin_dpo/margin_mean": 25.83917236328125,
      "margin_dpo/margin_std": 60.63203430175781,
      "step": 359
    },
    {
      "KL/chosen_KL_mean": -92.86167907714844,
      "KL/mean": -114.65642547607422,
      "KL/rejected_KL_mean": -136.45118713378906,
      "KL/std": 63.91606140136719,
      "epoch": 0.54421768707483,
      "fcm_dpo/beta": 0.00952062476426363,
      "fcm_dpo/delta": -0.01576114259660244,
      "fcm_dpo/margin": 43.589508056640625,
      "fcm_dpo/q_t": 0.40740275382995605,
      "grad_norm": 12.668853759765625,
      "learning_rate": 2.5661032514931834e-07,
      "logits/chosen": 0.6336207985877991,
      "logits/rejected": 0.5403010249137878,
      "logps/chosen": -156.34939575195312,
      "logps/ref_chosen": -63.48772048950195,
      "logps/ref_rejected": -90.6891098022461,
      "logps/rejected": -227.14028930664062,
      "loss": 1.1058,
      "margin_dpo/margin_mean": 43.589508056640625,
      "margin_dpo/margin_std": 66.03152465820312,
      "step": 360
    },
    {
      "KL/chosen_KL_mean": -91.1150894165039,
      "KL/mean": -116.79319763183594,
      "KL/rejected_KL_mean": -142.4713134765625,
      "KL/std": 64.54615783691406,
      "epoch": 0.54572940287226,
      "fcm_dpo/beta": 0.009435942396521568,
      "fcm_dpo/delta": -0.08882280439138412,
      "fcm_dpo/margin": 51.35621643066406,
      "fcm_dpo/q_t": 0.3894497752189636,
      "grad_norm": 12.657031059265137,
      "learning_rate": 2.552884820191154e-07,
      "logits/chosen": 0.7684783935546875,
      "logits/rejected": 0.7189006805419922,
      "logps/chosen": -149.0322265625,
      "logps/ref_chosen": -57.917144775390625,
      "logps/ref_rejected": -72.39089965820312,
      "logps/rejected": -214.86221313476562,
      "loss": 1.0443,
      "margin_dpo/margin_mean": 51.35621643066406,
      "margin_dpo/margin_std": 63.71247863769531,
      "step": 361
    },
    {
      "KL/chosen_KL_mean": -92.40254211425781,
      "KL/mean": -118.06507873535156,
      "KL/rejected_KL_mean": -143.7276153564453,
      "KL/std": 65.0129623413086,
      "epoch": 0.54724111866969,
      "fcm_dpo/beta": 0.009297511540353298,
      "fcm_dpo/delta": -0.08115798979997635,
      "fcm_dpo/margin": 51.3250617980957,
      "fcm_dpo/q_t": 0.3938571810722351,
      "grad_norm": 14.403992652893066,
      "learning_rate": 2.53966490958702e-07,
      "logits/chosen": 0.7443888783454895,
      "logits/rejected": 0.637453556060791,
      "logps/chosen": -155.84600830078125,
      "logps/ref_chosen": -63.4434700012207,
      "logps/ref_rejected": -103.45516967773438,
      "logps/rejected": -247.1827850341797,
      "loss": 1.0805,
      "margin_dpo/margin_mean": 51.32505798339844,
      "margin_dpo/margin_std": 75.57747650146484,
      "step": 362
    },
    {
      "KL/chosen_KL_mean": -98.28819274902344,
      "KL/mean": -123.49717712402344,
      "KL/rejected_KL_mean": -148.7061767578125,
      "KL/std": 60.82764434814453,
      "epoch": 0.5487528344671202,
      "fcm_dpo/beta": 0.009109338745474815,
      "fcm_dpo/delta": -0.06234493851661682,
      "fcm_dpo/margin": 50.41798782348633,
      "fcm_dpo/q_t": 0.3953966796398163,
      "grad_norm": 15.891867637634277,
      "learning_rate": 2.526443889470099e-07,
      "logits/chosen": 0.7446722984313965,
      "logits/rejected": 0.6137604713439941,
      "logps/chosen": -146.9400177001953,
      "logps/ref_chosen": -48.65182876586914,
      "logps/ref_rejected": -88.65904235839844,
      "logps/rejected": -237.365234375,
      "loss": 1.0679,
      "margin_dpo/margin_mean": 50.41798782348633,
      "margin_dpo/margin_std": 68.12222290039062,
      "step": 363
    },
    {
      "KL/chosen_KL_mean": -88.94969177246094,
      "KL/mean": -116.70195007324219,
      "KL/rejected_KL_mean": -144.45420837402344,
      "KL/std": 67.17678833007812,
      "epoch": 0.5502645502645502,
      "fcm_dpo/beta": 0.008948586881160736,
      "fcm_dpo/delta": -0.10194439440965652,
      "fcm_dpo/margin": 55.50453186035156,
      "fcm_dpo/q_t": 0.38964322209358215,
      "grad_norm": 11.782469749450684,
      "learning_rate": 2.513222129660744e-07,
      "logits/chosen": 0.6123770475387573,
      "logits/rejected": 0.5227110981941223,
      "logps/chosen": -146.82077026367188,
      "logps/ref_chosen": -57.87107467651367,
      "logps/ref_rejected": -80.95503234863281,
      "logps/rejected": -225.40924072265625,
      "loss": 1.0762,
      "margin_dpo/margin_mean": 55.50453186035156,
      "margin_dpo/margin_std": 82.74433898925781,
      "step": 364
    },
    {
      "KL/chosen_KL_mean": -79.53478240966797,
      "KL/mean": -105.70310974121094,
      "KL/rejected_KL_mean": -131.87142944335938,
      "KL/std": 68.90129089355469,
      "epoch": 0.5517762660619804,
      "fcm_dpo/beta": 0.008822593837976456,
      "fcm_dpo/delta": -0.06480256468057632,
      "fcm_dpo/margin": 52.3366584777832,
      "fcm_dpo/q_t": 0.39349961280822754,
      "grad_norm": 11.822827339172363,
      "learning_rate": 2.5e-07,
      "logits/chosen": 0.7287126183509827,
      "logits/rejected": 0.7254834175109863,
      "logps/chosen": -144.47695922851562,
      "logps/ref_chosen": -64.94217681884766,
      "logps/ref_rejected": -74.8599853515625,
      "logps/rejected": -206.73141479492188,
      "loss": 1.0442,
      "margin_dpo/margin_mean": 52.3366584777832,
      "margin_dpo/margin_std": 60.6138801574707,
      "step": 365
    },
    {
      "KL/chosen_KL_mean": -80.14967346191406,
      "KL/mean": -102.59008026123047,
      "KL/rejected_KL_mean": -125.03048706054688,
      "KL/std": 61.42876434326172,
      "epoch": 0.5532879818594104,
      "fcm_dpo/beta": 0.008836669847369194,
      "fcm_dpo/delta": 0.003269568085670471,
      "fcm_dpo/margin": 44.88079833984375,
      "fcm_dpo/q_t": 0.41164153814315796,
      "grad_norm": 13.908421516418457,
      "learning_rate": 2.486777870339255e-07,
      "logits/chosen": 0.6701743602752686,
      "logits/rejected": 0.6558536291122437,
      "logps/chosen": -135.31565856933594,
      "logps/ref_chosen": -55.16598129272461,
      "logps/ref_rejected": -65.26121520996094,
      "logps/rejected": -190.29168701171875,
      "loss": 1.1444,
      "margin_dpo/margin_mean": 44.88079833984375,
      "margin_dpo/margin_std": 79.36189270019531,
      "step": 366
    },
    {
      "KL/chosen_KL_mean": -89.8109130859375,
      "KL/mean": -112.27853393554688,
      "KL/rejected_KL_mean": -134.7461700439453,
      "KL/std": 62.997825622558594,
      "epoch": 0.5547996976568406,
      "fcm_dpo/beta": 0.008791262283921242,
      "fcm_dpo/delta": 0.0050530992448329926,
      "fcm_dpo/margin": 44.93526077270508,
      "fcm_dpo/q_t": 0.4080847501754761,
      "grad_norm": 12.340958595275879,
      "learning_rate": 2.4735561105299014e-07,
      "logits/chosen": 0.7136199474334717,
      "logits/rejected": 0.6086731553077698,
      "logps/chosen": -145.82138061523438,
      "logps/ref_chosen": -56.01046371459961,
      "logps/ref_rejected": -77.31010437011719,
      "logps/rejected": -212.0562744140625,
      "loss": 1.1178,
      "margin_dpo/margin_mean": 44.935264587402344,
      "margin_dpo/margin_std": 69.36185455322266,
      "step": 367
    },
    {
      "KL/chosen_KL_mean": -99.1821517944336,
      "KL/mean": -120.2548828125,
      "KL/rejected_KL_mean": -141.32760620117188,
      "KL/std": 63.947425842285156,
      "epoch": 0.5563114134542706,
      "fcm_dpo/beta": 0.008854447863996029,
      "fcm_dpo/delta": 0.02785618230700493,
      "fcm_dpo/margin": 42.145442962646484,
      "fcm_dpo/q_t": 0.4143589735031128,
      "grad_norm": 12.742534637451172,
      "learning_rate": 2.46033509041298e-07,
      "logits/chosen": 0.5594249963760376,
      "logits/rejected": 0.5591206550598145,
      "logps/chosen": -174.0114288330078,
      "logps/ref_chosen": -74.82927703857422,
      "logps/ref_rejected": -76.11680603027344,
      "logps/rejected": -217.44439697265625,
      "loss": 1.1347,
      "margin_dpo/margin_mean": 42.14544677734375,
      "margin_dpo/margin_std": 68.57220458984375,
      "step": 368
    },
    {
      "KL/chosen_KL_mean": -93.78787231445312,
      "KL/mean": -112.06151580810547,
      "KL/rejected_KL_mean": -130.33514404296875,
      "KL/std": 63.32222366333008,
      "epoch": 0.5578231292517006,
      "fcm_dpo/beta": 0.009002558887004852,
      "fcm_dpo/delta": 0.0726294070482254,
      "fcm_dpo/margin": 36.547271728515625,
      "fcm_dpo/q_t": 0.42570725083351135,
      "grad_norm": 13.213028907775879,
      "learning_rate": 2.447115179808846e-07,
      "logits/chosen": 0.7237619757652283,
      "logits/rejected": 0.6711267232894897,
      "logps/chosen": -152.1140899658203,
      "logps/ref_chosen": -58.32621765136719,
      "logps/ref_rejected": -80.92183685302734,
      "logps/rejected": -211.25698852539062,
      "loss": 1.1729,
      "margin_dpo/margin_mean": 36.547271728515625,
      "margin_dpo/margin_std": 67.68240356445312,
      "step": 369
    },
    {
      "KL/chosen_KL_mean": -89.63414001464844,
      "KL/mean": -115.5050048828125,
      "KL/rejected_KL_mean": -141.37588500976562,
      "KL/std": 66.52014923095703,
      "epoch": 0.5593348450491308,
      "fcm_dpo/beta": 0.008897393941879272,
      "fcm_dpo/delta": -0.06340888887643814,
      "fcm_dpo/margin": 51.74176025390625,
      "fcm_dpo/q_t": 0.39567339420318604,
      "grad_norm": 13.42084789276123,
      "learning_rate": 2.4338967485068164e-07,
      "logits/chosen": 0.8113032579421997,
      "logits/rejected": 0.7431550025939941,
      "logps/chosen": -142.51785278320312,
      "logps/ref_chosen": -52.88372039794922,
      "logps/ref_rejected": -79.43692016601562,
      "logps/rejected": -220.81280517578125,
      "loss": 1.0872,
      "margin_dpo/margin_mean": 51.741764068603516,
      "margin_dpo/margin_std": 77.3583984375,
      "step": 370
    },
    {
      "KL/chosen_KL_mean": -89.46411895751953,
      "KL/mean": -112.49140930175781,
      "KL/rejected_KL_mean": -135.51870727539062,
      "KL/std": 65.15025329589844,
      "epoch": 0.5608465608465608,
      "fcm_dpo/beta": 0.00891664158552885,
      "fcm_dpo/delta": -0.011712536215782166,
      "fcm_dpo/margin": 46.05458068847656,
      "fcm_dpo/q_t": 0.40757519006729126,
      "grad_norm": 15.0881986618042,
      "learning_rate": 2.420680166254831e-07,
      "logits/chosen": 0.8154574632644653,
      "logits/rejected": 0.7834243774414062,
      "logps/chosen": -138.68832397460938,
      "logps/ref_chosen": -49.224212646484375,
      "logps/ref_rejected": -63.348472595214844,
      "logps/rejected": -198.86717224121094,
      "loss": 1.1103,
      "margin_dpo/margin_mean": 46.05458068847656,
      "margin_dpo/margin_std": 69.55319213867188,
      "step": 371
    },
    {
      "KL/chosen_KL_mean": -96.4041748046875,
      "KL/mean": -110.4830551147461,
      "KL/rejected_KL_mean": -124.56193542480469,
      "KL/std": 64.20169830322266,
      "epoch": 0.562358276643991,
      "fcm_dpo/beta": 0.009113049134612083,
      "fcm_dpo/delta": 0.1454039216041565,
      "fcm_dpo/margin": 28.15774917602539,
      "fcm_dpo/q_t": 0.443379282951355,
      "grad_norm": 15.851375579833984,
      "learning_rate": 2.4074658027491044e-07,
      "logits/chosen": 0.7606231570243835,
      "logits/rejected": 0.6628408432006836,
      "logps/chosen": -148.67373657226562,
      "logps/ref_chosen": -52.269554138183594,
      "logps/ref_rejected": -72.99522399902344,
      "logps/rejected": -197.55715942382812,
      "loss": 1.2817,
      "margin_dpo/margin_mean": 28.15774917602539,
      "margin_dpo/margin_std": 82.369384765625,
      "step": 372
    },
    {
      "KL/chosen_KL_mean": -105.25238800048828,
      "KL/mean": -124.43365478515625,
      "KL/rejected_KL_mean": -143.61492919921875,
      "KL/std": 64.15481567382812,
      "epoch": 0.563869992441421,
      "fcm_dpo/beta": 0.009200419299304485,
      "fcm_dpo/delta": 0.04866807162761688,
      "fcm_dpo/margin": 38.362518310546875,
      "fcm_dpo/q_t": 0.422132670879364,
      "grad_norm": 14.45687198638916,
      "learning_rate": 2.394254027623792e-07,
      "logits/chosen": 0.7406034469604492,
      "logits/rejected": 0.6692646741867065,
      "logps/chosen": -166.36538696289062,
      "logps/ref_chosen": -61.112998962402344,
      "logps/ref_rejected": -76.24851989746094,
      "logps/rejected": -219.86343383789062,
      "loss": 1.2032,
      "margin_dpo/margin_mean": 38.362518310546875,
      "margin_dpo/margin_std": 83.87307739257812,
      "step": 373
    },
    {
      "KL/chosen_KL_mean": -87.62055206298828,
      "KL/mean": -120.09354400634766,
      "KL/rejected_KL_mean": -152.56654357910156,
      "KL/std": 64.96662139892578,
      "epoch": 0.5653817082388511,
      "fcm_dpo/beta": 0.00893518328666687,
      "fcm_dpo/delta": -0.19185763597488403,
      "fcm_dpo/margin": 64.94598388671875,
      "fcm_dpo/q_t": 0.36740460991859436,
      "grad_norm": 15.2799072265625,
      "learning_rate": 2.381045210440644e-07,
      "logits/chosen": 0.6385786533355713,
      "logits/rejected": 0.6403213739395142,
      "logps/chosen": -160.28976440429688,
      "logps/ref_chosen": -72.66920471191406,
      "logps/ref_rejected": -76.83158874511719,
      "logps/rejected": -229.39813232421875,
      "loss": 0.9805,
      "margin_dpo/margin_mean": 64.94598388671875,
      "margin_dpo/margin_std": 71.19766998291016,
      "step": 374
    },
    {
      "KL/chosen_KL_mean": -86.73007202148438,
      "KL/mean": -108.32252502441406,
      "KL/rejected_KL_mean": -129.91497802734375,
      "KL/std": 67.20675659179688,
      "epoch": 0.5668934240362812,
      "fcm_dpo/beta": 0.008861735463142395,
      "fcm_dpo/delta": 0.01787165366113186,
      "fcm_dpo/margin": 43.18490982055664,
      "fcm_dpo/q_t": 0.41384047269821167,
      "grad_norm": 14.840850830078125,
      "learning_rate": 2.3678397206786715e-07,
      "logits/chosen": 0.7619016170501709,
      "logits/rejected": 0.7026609778404236,
      "logps/chosen": -144.4133758544922,
      "logps/ref_chosen": -57.68330383300781,
      "logps/ref_rejected": -79.34097290039062,
      "logps/rejected": -209.25595092773438,
      "loss": 1.1436,
      "margin_dpo/margin_mean": 43.184906005859375,
      "margin_dpo/margin_std": 74.84222412109375,
      "step": 375
    },
    {
      "KL/chosen_KL_mean": -94.26927185058594,
      "KL/mean": -121.34205627441406,
      "KL/rejected_KL_mean": -148.4148406982422,
      "KL/std": 67.80638122558594,
      "epoch": 0.5684051398337112,
      "fcm_dpo/beta": 0.008781617507338524,
      "fcm_dpo/delta": -0.07936884462833405,
      "fcm_dpo/margin": 54.14557647705078,
      "fcm_dpo/q_t": 0.3939359188079834,
      "grad_norm": 14.06691837310791,
      "learning_rate": 2.3546379277238103e-07,
      "logits/chosen": 0.7732566595077515,
      "logits/rejected": 0.7034204006195068,
      "logps/chosen": -145.94332885742188,
      "logps/ref_chosen": -51.674072265625,
      "logps/ref_rejected": -75.69713592529297,
      "logps/rejected": -224.11196899414062,
      "loss": 1.078,
      "margin_dpo/margin_mean": 54.14557647705078,
      "margin_dpo/margin_std": 79.33256530761719,
      "step": 376
    },
    {
      "KL/chosen_KL_mean": -98.88938903808594,
      "KL/mean": -118.69151306152344,
      "KL/rejected_KL_mean": -138.49365234375,
      "KL/std": 63.79244613647461,
      "epoch": 0.5699168556311414,
      "fcm_dpo/beta": 0.00886094756424427,
      "fcm_dpo/delta": 0.05022633075714111,
      "fcm_dpo/margin": 39.60426330566406,
      "fcm_dpo/q_t": 0.42025965452194214,
      "grad_norm": 14.110357284545898,
      "learning_rate": 2.3414402008585886e-07,
      "logits/chosen": 0.7264569997787476,
      "logits/rejected": 0.7055130004882812,
      "logps/chosen": -145.0679168701172,
      "logps/ref_chosen": -46.17853546142578,
      "logps/ref_rejected": -57.756500244140625,
      "logps/rejected": -196.25015258789062,
      "loss": 1.1649,
      "margin_dpo/margin_mean": 39.60426330566406,
      "margin_dpo/margin_std": 72.60446166992188,
      "step": 377
    },
    {
      "KL/chosen_KL_mean": -94.4163818359375,
      "KL/mean": -113.24252319335938,
      "KL/rejected_KL_mean": -132.0686798095703,
      "KL/std": 66.46366882324219,
      "epoch": 0.5714285714285714,
      "fcm_dpo/beta": 0.00897032581269741,
      "fcm_dpo/delta": 0.0636821836233139,
      "fcm_dpo/margin": 37.652305603027344,
      "fcm_dpo/q_t": 0.423561155796051,
      "grad_norm": 13.255199432373047,
      "learning_rate": 2.3282469092517977e-07,
      "logits/chosen": 0.7863258719444275,
      "logits/rejected": 0.7367914319038391,
      "logps/chosen": -153.63525390625,
      "logps/ref_chosen": -59.21887969970703,
      "logps/ref_rejected": -71.24818420410156,
      "logps/rejected": -203.31686401367188,
      "loss": 1.1713,
      "margin_dpo/margin_mean": 37.652305603027344,
      "margin_dpo/margin_std": 69.34526062011719,
      "step": 378
    },
    {
      "KL/chosen_KL_mean": -89.37190246582031,
      "KL/mean": -113.827880859375,
      "KL/rejected_KL_mean": -138.28387451171875,
      "KL/std": 65.70942687988281,
      "epoch": 0.5729402872260015,
      "fcm_dpo/beta": 0.008909247815608978,
      "fcm_dpo/delta": -0.03739330917596817,
      "fcm_dpo/margin": 48.91199493408203,
      "fcm_dpo/q_t": 0.40265560150146484,
      "grad_norm": 14.700591087341309,
      "learning_rate": 2.3150584219481643e-07,
      "logits/chosen": 0.7219586372375488,
      "logits/rejected": 0.6492105722427368,
      "logps/chosen": -165.6884765625,
      "logps/ref_chosen": -76.31658935546875,
      "logps/ref_rejected": -104.26200103759766,
      "logps/rejected": -242.54588317871094,
      "loss": 1.0938,
      "margin_dpo/margin_mean": 48.91199493408203,
      "margin_dpo/margin_std": 72.89404296875,
      "step": 379
    },
    {
      "KL/chosen_KL_mean": -79.29644012451172,
      "KL/mean": -110.41128540039062,
      "KL/rejected_KL_mean": -141.52613830566406,
      "KL/std": 63.3236083984375,
      "epoch": 0.5744520030234316,
      "fcm_dpo/beta": 0.008703596889972687,
      "fcm_dpo/delta": -0.14961189031600952,
      "fcm_dpo/margin": 62.229698181152344,
      "fcm_dpo/q_t": 0.3763843774795532,
      "grad_norm": 13.61119556427002,
      "learning_rate": 2.3018751078580283e-07,
      "logits/chosen": 0.7565820217132568,
      "logits/rejected": 0.7187171578407288,
      "logps/chosen": -140.57960510253906,
      "logps/ref_chosen": -61.283164978027344,
      "logps/ref_rejected": -72.38892364501953,
      "logps/rejected": -213.91505432128906,
      "loss": 1.0228,
      "margin_dpo/margin_mean": 62.229698181152344,
      "margin_dpo/margin_std": 77.56293487548828,
      "step": 380
    },
    {
      "KL/chosen_KL_mean": -97.59381866455078,
      "KL/mean": -109.89251708984375,
      "KL/rejected_KL_mean": -122.19123077392578,
      "KL/std": 63.37030029296875,
      "epoch": 0.5759637188208617,
      "fcm_dpo/beta": 0.008918230421841145,
      "fcm_dpo/delta": 0.18300682306289673,
      "fcm_dpo/margin": 24.597408294677734,
      "fcm_dpo/q_t": 0.4514863193035126,
      "grad_norm": 12.83474349975586,
      "learning_rate": 2.288697335747027e-07,
      "logits/chosen": 0.704852819442749,
      "logits/rejected": 0.6807745695114136,
      "logps/chosen": -155.80780029296875,
      "logps/ref_chosen": -58.2139892578125,
      "logps/ref_rejected": -60.78669357299805,
      "logps/rejected": -182.97792053222656,
      "loss": 1.2863,
      "margin_dpo/margin_mean": 24.597410202026367,
      "margin_dpo/margin_std": 74.75566101074219,
      "step": 381
    },
    {
      "KL/chosen_KL_mean": -98.49388122558594,
      "KL/mean": -119.5999526977539,
      "KL/rejected_KL_mean": -140.70602416992188,
      "KL/std": 65.70405578613281,
      "epoch": 0.5774754346182918,
      "fcm_dpo/beta": 0.009017659351229668,
      "fcm_dpo/delta": 0.019596684724092484,
      "fcm_dpo/margin": 42.21213150024414,
      "fcm_dpo/q_t": 0.4132706820964813,
      "grad_norm": 13.350465774536133,
      "learning_rate": 2.2755254742257706e-07,
      "logits/chosen": 0.7516306638717651,
      "logits/rejected": 0.6984708309173584,
      "logps/chosen": -160.3192138671875,
      "logps/ref_chosen": -61.82532501220703,
      "logps/ref_rejected": -83.0452880859375,
      "logps/rejected": -223.7512969970703,
      "loss": 1.1191,
      "margin_dpo/margin_mean": 42.212135314941406,
      "margin_dpo/margin_std": 62.93214416503906,
      "step": 382
    },
    {
      "KL/chosen_KL_mean": -95.25296783447266,
      "KL/mean": -118.4464111328125,
      "KL/rejected_KL_mean": -141.63986206054688,
      "KL/std": 65.69584655761719,
      "epoch": 0.5789871504157218,
      "fcm_dpo/beta": 0.008922882378101349,
      "fcm_dpo/delta": -0.015040282160043716,
      "fcm_dpo/margin": 46.386905670166016,
      "fcm_dpo/q_t": 0.4087279438972473,
      "grad_norm": 14.565018653869629,
      "learning_rate": 2.2623598917395436e-07,
      "logits/chosen": 0.5949560403823853,
      "logits/rejected": 0.6251660585403442,
      "logps/chosen": -175.81622314453125,
      "logps/ref_chosen": -80.56326293945312,
      "logps/ref_rejected": -74.62922668457031,
      "logps/rejected": -216.26910400390625,
      "loss": 1.1348,
      "margin_dpo/margin_mean": 46.386905670166016,
      "margin_dpo/margin_std": 80.02427673339844,
      "step": 383
    },
    {
      "KL/chosen_KL_mean": -95.48326110839844,
      "KL/mean": -117.88404083251953,
      "KL/rejected_KL_mean": -140.28482055664062,
      "KL/std": 65.55471801757812,
      "epoch": 0.5804988662131519,
      "fcm_dpo/beta": 0.00895443931221962,
      "fcm_dpo/delta": -0.001223698491230607,
      "fcm_dpo/margin": 44.801551818847656,
      "fcm_dpo/q_t": 0.4078155755996704,
      "grad_norm": 14.621257781982422,
      "learning_rate": 2.2492009565579875e-07,
      "logits/chosen": 0.733665943145752,
      "logits/rejected": 0.6879531145095825,
      "logps/chosen": -160.95840454101562,
      "logps/ref_chosen": -65.47514343261719,
      "logps/ref_rejected": -79.67378234863281,
      "logps/rejected": -219.95860290527344,
      "loss": 1.113,
      "margin_dpo/margin_mean": 44.801551818847656,
      "margin_dpo/margin_std": 68.81768798828125,
      "step": 384
    },
    {
      "KL/chosen_KL_mean": -93.10316467285156,
      "KL/mean": -122.23687744140625,
      "KL/rejected_KL_mean": -151.37060546875,
      "KL/std": 64.7177505493164,
      "epoch": 0.582010582010582,
      "fcm_dpo/beta": 0.008825909346342087,
      "fcm_dpo/delta": -0.12028289586305618,
      "fcm_dpo/margin": 58.26744842529297,
      "fcm_dpo/q_t": 0.38351568579673767,
      "grad_norm": 12.378807067871094,
      "learning_rate": 2.2360490367648084e-07,
      "logits/chosen": 0.6500192284584045,
      "logits/rejected": 0.6126409769058228,
      "logps/chosen": -159.15966796875,
      "logps/ref_chosen": -66.0565185546875,
      "logps/ref_rejected": -86.68023681640625,
      "logps/rejected": -238.05084228515625,
      "loss": 1.0231,
      "margin_dpo/margin_mean": 58.26744842529297,
      "margin_dpo/margin_std": 68.7908706665039,
      "step": 385
    },
    {
      "KL/chosen_KL_mean": -107.4451904296875,
      "KL/mean": -125.80526733398438,
      "KL/rejected_KL_mean": -144.16531372070312,
      "KL/std": 64.58238220214844,
      "epoch": 0.5835222978080121,
      "fcm_dpo/beta": 0.008834145963191986,
      "fcm_dpo/delta": 0.0782126784324646,
      "fcm_dpo/margin": 36.72013473510742,
      "fcm_dpo/q_t": 0.4248654842376709,
      "grad_norm": 14.023968696594238,
      "learning_rate": 2.2229045002474724e-07,
      "logits/chosen": 0.6312617659568787,
      "logits/rejected": 0.5722914934158325,
      "logps/chosen": -183.06884765625,
      "logps/ref_chosen": -75.6236572265625,
      "logps/ref_rejected": -92.62330627441406,
      "logps/rejected": -236.78863525390625,
      "loss": 1.1725,
      "margin_dpo/margin_mean": 36.72013473510742,
      "margin_dpo/margin_std": 67.48342895507812,
      "step": 386
    },
    {
      "KL/chosen_KL_mean": -95.84181213378906,
      "KL/mean": -123.29605102539062,
      "KL/rejected_KL_mean": -150.75030517578125,
      "KL/std": 63.237667083740234,
      "epoch": 0.5850340136054422,
      "fcm_dpo/beta": 0.008767899125814438,
      "fcm_dpo/delta": -0.08556218445301056,
      "fcm_dpo/margin": 54.908485412597656,
      "fcm_dpo/q_t": 0.39054542779922485,
      "grad_norm": 12.948434829711914,
      "learning_rate": 2.209767714686924e-07,
      "logits/chosen": 0.7226855754852295,
      "logits/rejected": 0.618818998336792,
      "logps/chosen": -143.06350708007812,
      "logps/ref_chosen": -47.22170639038086,
      "logps/ref_rejected": -87.338134765625,
      "logps/rejected": -238.08843994140625,
      "loss": 1.0379,
      "margin_dpo/margin_mean": 54.908485412597656,
      "margin_dpo/margin_std": 65.7297592163086,
      "step": 387
    },
    {
      "KL/chosen_KL_mean": -96.84806823730469,
      "KL/mean": -114.32777404785156,
      "KL/rejected_KL_mean": -131.8074951171875,
      "KL/std": 64.9911880493164,
      "epoch": 0.5865457294028723,
      "fcm_dpo/beta": 0.008810779079794884,
      "fcm_dpo/delta": 0.09486287087202072,
      "fcm_dpo/margin": 34.95942306518555,
      "fcm_dpo/q_t": 0.4320222735404968,
      "grad_norm": 12.77639102935791,
      "learning_rate": 2.1966390475472954e-07,
      "logits/chosen": 0.7181829810142517,
      "logits/rejected": 0.7101290225982666,
      "logps/chosen": -171.42755126953125,
      "logps/ref_chosen": -74.5794677734375,
      "logps/ref_rejected": -79.92558288574219,
      "logps/rejected": -211.7330780029297,
      "loss": 1.2113,
      "margin_dpo/margin_mean": 34.95942306518555,
      "margin_dpo/margin_std": 76.60311126708984,
      "step": 388
    },
    {
      "KL/chosen_KL_mean": -93.4125747680664,
      "KL/mean": -121.59010314941406,
      "KL/rejected_KL_mean": -149.76763916015625,
      "KL/std": 66.75245666503906,
      "epoch": 0.5880574452003023,
      "fcm_dpo/beta": 0.008777271024882793,
      "fcm_dpo/delta": -0.09948212653398514,
      "fcm_dpo/margin": 56.35505294799805,
      "fcm_dpo/q_t": 0.3881889581680298,
      "grad_norm": 28.938011169433594,
      "learning_rate": 2.1835188660656265e-07,
      "logits/chosen": 0.7351957559585571,
      "logits/rejected": 0.6990658044815063,
      "logps/chosen": -155.0369415283203,
      "logps/ref_chosen": -61.624366760253906,
      "logps/ref_rejected": -76.50978088378906,
      "logps/rejected": -226.2774200439453,
      "loss": 1.0491,
      "margin_dpo/margin_mean": 56.35505676269531,
      "margin_dpo/margin_std": 74.05264282226562,
      "step": 389
    },
    {
      "KL/chosen_KL_mean": -86.54637145996094,
      "KL/mean": -108.1448974609375,
      "KL/rejected_KL_mean": -129.74343872070312,
      "KL/std": 63.3463134765625,
      "epoch": 0.5895691609977324,
      "fcm_dpo/beta": 0.008739067241549492,
      "fcm_dpo/delta": 0.023366158828139305,
      "fcm_dpo/margin": 43.197052001953125,
      "fcm_dpo/q_t": 0.4136941432952881,
      "grad_norm": 10.710925102233887,
      "learning_rate": 2.170407537241599e-07,
      "logits/chosen": 0.7972933650016785,
      "logits/rejected": 0.7268404960632324,
      "logps/chosen": -132.41824340820312,
      "logps/ref_chosen": -45.871864318847656,
      "logps/ref_rejected": -61.305999755859375,
      "logps/rejected": -191.0494384765625,
      "loss": 1.1237,
      "margin_dpo/margin_mean": 43.197052001953125,
      "margin_dpo/margin_std": 66.48844146728516,
      "step": 390
    },
    {
      "KL/chosen_KL_mean": -93.65834045410156,
      "KL/mean": -119.96128845214844,
      "KL/rejected_KL_mean": -146.26422119140625,
      "KL/std": 64.77439880371094,
      "epoch": 0.5910808767951625,
      "fcm_dpo/beta": 0.008656751364469528,
      "fcm_dpo/delta": -0.05850217863917351,
      "fcm_dpo/margin": 52.605873107910156,
      "fcm_dpo/q_t": 0.3968993127346039,
      "grad_norm": 12.943575859069824,
      "learning_rate": 2.1573054278272636e-07,
      "logits/chosen": 0.728482723236084,
      "logits/rejected": 0.6600795984268188,
      "logps/chosen": -151.84535217285156,
      "logps/ref_chosen": -58.18701171875,
      "logps/ref_rejected": -83.63442993164062,
      "logps/rejected": -229.89866638183594,
      "loss": 1.0959,
      "margin_dpo/margin_mean": 52.605873107910156,
      "margin_dpo/margin_std": 80.5517807006836,
      "step": 391
    },
    {
      "KL/chosen_KL_mean": -80.34383392333984,
      "KL/mean": -109.28814697265625,
      "KL/rejected_KL_mean": -138.23245239257812,
      "KL/std": 66.96438598632812,
      "epoch": 0.5925925925925926,
      "fcm_dpo/beta": 0.008585932664573193,
      "fcm_dpo/delta": -0.10228624939918518,
      "fcm_dpo/margin": 57.88862991333008,
      "fcm_dpo/q_t": 0.3883962333202362,
      "grad_norm": 11.231030464172363,
      "learning_rate": 2.1442129043167873e-07,
      "logits/chosen": 0.7651580572128296,
      "logits/rejected": 0.7059136629104614,
      "logps/chosen": -150.08836364746094,
      "logps/ref_chosen": -69.7445297241211,
      "logps/ref_rejected": -94.05877685546875,
      "logps/rejected": -232.29122924804688,
      "loss": 1.0572,
      "margin_dpo/margin_mean": 57.88862609863281,
      "margin_dpo/margin_std": 78.10966491699219,
      "step": 392
    },
    {
      "KL/chosen_KL_mean": -94.92709350585938,
      "KL/mean": -123.78874206542969,
      "KL/rejected_KL_mean": -152.650390625,
      "KL/std": 66.06975555419922,
      "epoch": 0.5941043083900227,
      "fcm_dpo/beta": 0.008353622630238533,
      "fcm_dpo/delta": -0.08677142858505249,
      "fcm_dpo/margin": 57.723289489746094,
      "fcm_dpo/q_t": 0.3901318311691284,
      "grad_norm": 11.89721965789795,
      "learning_rate": 2.131130332936195e-07,
      "logits/chosen": 0.7530436515808105,
      "logits/rejected": 0.7150501012802124,
      "logps/chosen": -147.26199340820312,
      "logps/ref_chosen": -52.33489990234375,
      "logps/ref_rejected": -74.33809661865234,
      "logps/rejected": -226.98849487304688,
      "loss": 1.0421,
      "margin_dpo/margin_mean": 57.723289489746094,
      "margin_dpo/margin_std": 70.54959106445312,
      "step": 393
    },
    {
      "KL/chosen_KL_mean": -89.56831359863281,
      "KL/mean": -114.85767364501953,
      "KL/rejected_KL_mean": -140.14703369140625,
      "KL/std": 60.31583023071289,
      "epoch": 0.5956160241874527,
      "fcm_dpo/beta": 0.008336104452610016,
      "fcm_dpo/delta": -0.022687222808599472,
      "fcm_dpo/margin": 50.57872772216797,
      "fcm_dpo/q_t": 0.40135329961776733,
      "grad_norm": 12.148786544799805,
      "learning_rate": 2.1180580796331323e-07,
      "logits/chosen": 0.7515368461608887,
      "logits/rejected": 0.7221952676773071,
      "logps/chosen": -150.24444580078125,
      "logps/ref_chosen": -60.6761360168457,
      "logps/ref_rejected": -71.36074829101562,
      "logps/rejected": -211.50778198242188,
      "loss": 1.0731,
      "margin_dpo/margin_mean": 50.57872772216797,
      "margin_dpo/margin_std": 61.64701843261719,
      "step": 394
    },
    {
      "KL/chosen_KL_mean": -95.43479919433594,
      "KL/mean": -115.49359130859375,
      "KL/rejected_KL_mean": -135.55238342285156,
      "KL/std": 62.913185119628906,
      "epoch": 0.5971277399848829,
      "fcm_dpo/beta": 0.008418946526944637,
      "fcm_dpo/delta": 0.06374331563711166,
      "fcm_dpo/margin": 40.117591857910156,
      "fcm_dpo/q_t": 0.42396989464759827,
      "grad_norm": 15.213231086730957,
      "learning_rate": 2.104996510066625e-07,
      "logits/chosen": 0.7378535270690918,
      "logits/rejected": 0.6408002376556396,
      "logps/chosen": -146.03912353515625,
      "logps/ref_chosen": -50.60432434082031,
      "logps/ref_rejected": -77.08731079101562,
      "logps/rejected": -212.63970947265625,
      "loss": 1.1554,
      "margin_dpo/margin_mean": 40.117591857910156,
      "margin_dpo/margin_std": 67.76335906982422,
      "step": 395
    },
    {
      "KL/chosen_KL_mean": -89.58931732177734,
      "KL/mean": -113.90975952148438,
      "KL/rejected_KL_mean": -138.23020935058594,
      "KL/std": 70.15506744384766,
      "epoch": 0.5986394557823129,
      "fcm_dpo/beta": 0.008337254635989666,
      "fcm_dpo/delta": -0.006977491080760956,
      "fcm_dpo/margin": 48.64088439941406,
      "fcm_dpo/q_t": 0.40703338384628296,
      "grad_norm": 11.953834533691406,
      "learning_rate": 2.0919459895968517e-07,
      "logits/chosen": 0.7067517042160034,
      "logits/rejected": 0.6094527244567871,
      "logps/chosen": -140.9489288330078,
      "logps/ref_chosen": -51.35961151123047,
      "logps/ref_rejected": -79.89360046386719,
      "logps/rejected": -218.12380981445312,
      "loss": 1.091,
      "margin_dpo/margin_mean": 48.64088439941406,
      "margin_dpo/margin_std": 62.8603515625,
      "step": 396
    },
    {
      "KL/chosen_KL_mean": -103.04716491699219,
      "KL/mean": -116.64557647705078,
      "KL/rejected_KL_mean": -130.24398803710938,
      "KL/std": 64.92282104492188,
      "epoch": 0.600151171579743,
      "fcm_dpo/beta": 0.008626183494925499,
      "fcm_dpo/delta": 0.16903799772262573,
      "fcm_dpo/margin": 27.196826934814453,
      "fcm_dpo/q_t": 0.4464913010597229,
      "grad_norm": 13.257553100585938,
      "learning_rate": 2.078906883274924e-07,
      "logits/chosen": 0.6656841039657593,
      "logits/rejected": 0.6178501844406128,
      "logps/chosen": -169.50338745117188,
      "logps/ref_chosen": -66.45622253417969,
      "logps/ref_rejected": -85.74736785888672,
      "logps/rejected": -215.99136352539062,
      "loss": 1.2806,
      "margin_dpo/margin_mean": 27.196826934814453,
      "margin_dpo/margin_std": 80.3773193359375,
      "step": 397
    },
    {
      "KL/chosen_KL_mean": -88.44233703613281,
      "KL/mean": -118.74835205078125,
      "KL/rejected_KL_mean": -149.0543670654297,
      "KL/std": 67.16844177246094,
      "epoch": 0.6016628873771731,
      "fcm_dpo/beta": 0.008501345291733742,
      "fcm_dpo/delta": -0.1221761479973793,
      "fcm_dpo/margin": 60.61201477050781,
      "fcm_dpo/q_t": 0.38299331068992615,
      "grad_norm": 11.214990615844727,
      "learning_rate": 2.065879555832674e-07,
      "logits/chosen": 0.719185471534729,
      "logits/rejected": 0.6512874960899353,
      "logps/chosen": -137.68658447265625,
      "logps/ref_chosen": -49.244239807128906,
      "logps/ref_rejected": -75.18949127197266,
      "logps/rejected": -224.2438507080078,
      "loss": 1.0132,
      "margin_dpo/margin_mean": 60.61201477050781,
      "margin_dpo/margin_std": 68.35538482666016,
      "step": 398
    },
    {
      "KL/chosen_KL_mean": -103.34333801269531,
      "KL/mean": -136.01007080078125,
      "KL/rejected_KL_mean": -168.6768035888672,
      "KL/std": 69.25430297851562,
      "epoch": 0.6031746031746031,
      "fcm_dpo/beta": 0.008262135088443756,
      "fcm_dpo/delta": -0.1485450565814972,
      "fcm_dpo/margin": 65.33348083496094,
      "fcm_dpo/q_t": 0.37789463996887207,
      "grad_norm": 13.473124504089355,
      "learning_rate": 2.052864371672457e-07,
      "logits/chosen": 0.6423044204711914,
      "logits/rejected": 0.4999236464500427,
      "logps/chosen": -171.65013122558594,
      "logps/ref_chosen": -68.30679321289062,
      "logps/ref_rejected": -113.2708511352539,
      "logps/rejected": -281.9476623535156,
      "loss": 1.0073,
      "margin_dpo/margin_mean": 65.33348083496094,
      "margin_dpo/margin_std": 75.8927001953125,
      "step": 399
    },
    {
      "KL/chosen_KL_mean": -114.28192901611328,
      "KL/mean": -134.40957641601562,
      "KL/rejected_KL_mean": -154.5372314453125,
      "KL/std": 67.74653625488281,
      "epoch": 0.6046863189720333,
      "fcm_dpo/beta": 0.008252747356891632,
      "fcm_dpo/delta": 0.0695052295923233,
      "fcm_dpo/margin": 40.25529098510742,
      "fcm_dpo/q_t": 0.424492746591568,
      "grad_norm": 18.078105926513672,
      "learning_rate": 2.0398616948569493e-07,
      "logits/chosen": 0.6950872540473938,
      "logits/rejected": 0.636712908744812,
      "logps/chosen": -185.90841674804688,
      "logps/ref_chosen": -71.62649536132812,
      "logps/ref_rejected": -90.98765563964844,
      "logps/rejected": -245.52487182617188,
      "loss": 1.1625,
      "margin_dpo/margin_mean": 40.25529479980469,
      "margin_dpo/margin_std": 68.32823181152344,
      "step": 400
    },
    {
      "KL/chosen_KL_mean": -88.39434814453125,
      "KL/mean": -116.34048461914062,
      "KL/rejected_KL_mean": -144.28662109375,
      "KL/std": 71.89583587646484,
      "epoch": 0.6061980347694633,
      "fcm_dpo/beta": 0.008242808282375336,
      "fcm_dpo/delta": -0.06394520401954651,
      "fcm_dpo/margin": 55.89228057861328,
      "fcm_dpo/q_t": 0.39509570598602295,
      "grad_norm": 10.358372688293457,
      "learning_rate": 2.0268718890989752e-07,
      "logits/chosen": 0.7502726316452026,
      "logits/rejected": 0.6499309539794922,
      "logps/chosen": -142.11929321289062,
      "logps/ref_chosen": -53.72495651245117,
      "logps/ref_rejected": -75.06304931640625,
      "logps/rejected": -219.34967041015625,
      "loss": 1.0506,
      "margin_dpo/margin_mean": 55.89228057861328,
      "margin_dpo/margin_std": 67.9534683227539,
      "step": 401
    },
    {
      "KL/chosen_KL_mean": -95.33905029296875,
      "KL/mean": -119.19203186035156,
      "KL/rejected_KL_mean": -143.04501342773438,
      "KL/std": 64.60128021240234,
      "epoch": 0.6077097505668935,
      "fcm_dpo/beta": 0.008193889632821083,
      "fcm_dpo/delta": 0.008644327521324158,
      "fcm_dpo/margin": 47.705970764160156,
      "fcm_dpo/q_t": 0.41146695613861084,
      "grad_norm": 13.163501739501953,
      "learning_rate": 2.013895317751323e-07,
      "logits/chosen": 0.7267850637435913,
      "logits/rejected": 0.6951309442520142,
      "logps/chosen": -157.21298217773438,
      "logps/ref_chosen": -61.873931884765625,
      "logps/ref_rejected": -66.15198516845703,
      "logps/rejected": -209.19700622558594,
      "loss": 1.1294,
      "margin_dpo/margin_mean": 47.705970764160156,
      "margin_dpo/margin_std": 76.21807098388672,
      "step": 402
    },
    {
      "KL/chosen_KL_mean": -103.14884948730469,
      "KL/mean": -131.67453002929688,
      "KL/rejected_KL_mean": -160.20021057128906,
      "KL/std": 71.38824462890625,
      "epoch": 0.6092214663643235,
      "fcm_dpo/beta": 0.008181357756257057,
      "fcm_dpo/delta": -0.0699668824672699,
      "fcm_dpo/margin": 57.05136489868164,
      "fcm_dpo/q_t": 0.3950841426849365,
      "grad_norm": 11.315382957458496,
      "learning_rate": 2.0009323437965898e-07,
      "logits/chosen": 0.8247023820877075,
      "logits/rejected": 0.740180253982544,
      "logps/chosen": -154.47035217285156,
      "logps/ref_chosen": -51.321502685546875,
      "logps/ref_rejected": -86.54010772705078,
      "logps/rejected": -246.74032592773438,
      "loss": 1.0712,
      "margin_dpo/margin_mean": 57.05136489868164,
      "margin_dpo/margin_std": 78.94970703125,
      "step": 403
    },
    {
      "KL/chosen_KL_mean": -95.03086853027344,
      "KL/mean": -124.51020050048828,
      "KL/rejected_KL_mean": -153.98953247070312,
      "KL/std": 71.967529296875,
      "epoch": 0.6107331821617535,
      "fcm_dpo/beta": 0.00798763893544674,
      "fcm_dpo/delta": -0.07605341076850891,
      "fcm_dpo/margin": 58.95866012573242,
      "fcm_dpo/q_t": 0.39368003606796265,
      "grad_norm": 14.19412899017334,
      "learning_rate": 1.9879833298370237e-07,
      "logits/chosen": 0.6934102773666382,
      "logits/rejected": 0.5980790853500366,
      "logps/chosen": -157.29376220703125,
      "logps/ref_chosen": -62.26288604736328,
      "logps/ref_rejected": -95.19029998779297,
      "logps/rejected": -249.17982482910156,
      "loss": 1.0654,
      "margin_dpo/margin_mean": 58.958656311035156,
      "margin_dpo/margin_std": 77.97892761230469,
      "step": 404
    },
    {
      "KL/chosen_KL_mean": -99.53964233398438,
      "KL/mean": -122.60385131835938,
      "KL/rejected_KL_mean": -145.66806030273438,
      "KL/std": 66.06510925292969,
      "epoch": 0.6122448979591837,
      "fcm_dpo/beta": 0.007995378226041794,
      "fcm_dpo/delta": 0.03158438578248024,
      "fcm_dpo/margin": 46.12843322753906,
      "fcm_dpo/q_t": 0.4166746735572815,
      "grad_norm": 11.438131332397461,
      "learning_rate": 1.975048638084379e-07,
      "logits/chosen": 0.8034164905548096,
      "logits/rejected": 0.7549797296524048,
      "logps/chosen": -150.1239776611328,
      "logps/ref_chosen": -50.5843391418457,
      "logps/ref_rejected": -65.43156433105469,
      "logps/rejected": -211.09963989257812,
      "loss": 1.1317,
      "margin_dpo/margin_mean": 46.12843322753906,
      "margin_dpo/margin_std": 70.4107437133789,
      "step": 405
    },
    {
      "KL/chosen_KL_mean": -97.1700439453125,
      "KL/mean": -126.85040283203125,
      "KL/rejected_KL_mean": -156.53074645996094,
      "KL/std": 70.42720031738281,
      "epoch": 0.6137566137566137,
      "fcm_dpo/beta": 0.007977031171321869,
      "fcm_dpo/delta": -0.07714153826236725,
      "fcm_dpo/margin": 59.36070251464844,
      "fcm_dpo/q_t": 0.3917776644229889,
      "grad_norm": 14.39725112915039,
      "learning_rate": 1.9621286303497914e-07,
      "logits/chosen": 0.7903730869293213,
      "logits/rejected": 0.6238706111907959,
      "logps/chosen": -146.1656494140625,
      "logps/ref_chosen": -48.99560546875,
      "logps/ref_rejected": -92.47774505615234,
      "logps/rejected": -249.00848388671875,
      "loss": 1.0647,
      "margin_dpo/margin_mean": 59.36070251464844,
      "margin_dpo/margin_std": 80.22438049316406,
      "step": 406
    },
    {
      "KL/chosen_KL_mean": -117.63362884521484,
      "KL/mean": -140.78936767578125,
      "KL/rejected_KL_mean": -163.9451141357422,
      "KL/std": 74.67190551757812,
      "epoch": 0.6152683295540439,
      "fcm_dpo/beta": 0.007999084889888763,
      "fcm_dpo/delta": 0.030455123633146286,
      "fcm_dpo/margin": 46.31147003173828,
      "fcm_dpo/q_t": 0.4163338243961334,
      "grad_norm": 14.213495254516602,
      "learning_rate": 1.9492236680336483e-07,
      "logits/chosen": 0.606561541557312,
      "logits/rejected": 0.5306586027145386,
      "logps/chosen": -207.03419494628906,
      "logps/ref_chosen": -89.40056610107422,
      "logps/ref_rejected": -99.28775024414062,
      "logps/rejected": -263.23284912109375,
      "loss": 1.1412,
      "margin_dpo/margin_mean": 46.31147003173828,
      "margin_dpo/margin_std": 77.54383850097656,
      "step": 407
    },
    {
      "KL/chosen_KL_mean": -89.46073913574219,
      "KL/mean": -124.67997741699219,
      "KL/rejected_KL_mean": -159.8992156982422,
      "KL/std": 69.03326416015625,
      "epoch": 0.6167800453514739,
      "fcm_dpo/beta": 0.00785075593739748,
      "fcm_dpo/delta": -0.16175517439842224,
      "fcm_dpo/margin": 70.4384765625,
      "fcm_dpo/q_t": 0.37296241521835327,
      "grad_norm": 10.863518714904785,
      "learning_rate": 1.9363341121154895e-07,
      "logits/chosen": 0.7440523505210876,
      "logits/rejected": 0.6608445048332214,
      "logps/chosen": -144.16465759277344,
      "logps/ref_chosen": -54.70391845703125,
      "logps/ref_rejected": -73.98648834228516,
      "logps/rejected": -233.88571166992188,
      "loss": 0.9903,
      "margin_dpo/margin_mean": 70.4384765625,
      "margin_dpo/margin_std": 73.6658935546875,
      "step": 408
    },
    {
      "KL/chosen_KL_mean": -111.31812286376953,
      "KL/mean": -128.97467041015625,
      "KL/rejected_KL_mean": -146.6312255859375,
      "KL/std": 62.695716857910156,
      "epoch": 0.618291761148904,
      "fcm_dpo/beta": 0.007871621288359165,
      "fcm_dpo/delta": 0.12564970552921295,
      "fcm_dpo/margin": 35.31309509277344,
      "fcm_dpo/q_t": 0.4370152950286865,
      "grad_norm": 13.255595207214355,
      "learning_rate": 1.9234603231438994e-07,
      "logits/chosen": 0.7345675826072693,
      "logits/rejected": 0.743270754814148,
      "logps/chosen": -173.43634033203125,
      "logps/ref_chosen": -62.11822509765625,
      "logps/ref_rejected": -61.933509826660156,
      "logps/rejected": -208.56472778320312,
      "loss": 1.2051,
      "margin_dpo/margin_mean": 35.31309509277344,
      "margin_dpo/margin_std": 72.25135803222656,
      "step": 409
    },
    {
      "KL/chosen_KL_mean": -103.84161376953125,
      "KL/mean": -132.0266876220703,
      "KL/rejected_KL_mean": -160.21176147460938,
      "KL/std": 65.34846496582031,
      "epoch": 0.6198034769463341,
      "fcm_dpo/beta": 0.007834449410438538,
      "fcm_dpo/delta": -0.044498007744550705,
      "fcm_dpo/margin": 56.37013626098633,
      "fcm_dpo/q_t": 0.3971004784107208,
      "grad_norm": 11.596942901611328,
      "learning_rate": 1.9106026612264315e-07,
      "logits/chosen": 0.7368378639221191,
      "logits/rejected": 0.7123606204986572,
      "logps/chosen": -165.644287109375,
      "logps/ref_chosen": -61.80266189575195,
      "logps/ref_rejected": -76.60002136230469,
      "logps/rejected": -236.811767578125,
      "loss": 1.0501,
      "margin_dpo/margin_mean": 56.37013626098633,
      "margin_dpo/margin_std": 61.56016540527344,
      "step": 410
    },
    {
      "KL/chosen_KL_mean": -109.17242431640625,
      "KL/mean": -136.98342895507812,
      "KL/rejected_KL_mean": -164.79446411132812,
      "KL/std": 74.19166564941406,
      "epoch": 0.6213151927437641,
      "fcm_dpo/beta": 0.007825829088687897,
      "fcm_dpo/delta": -0.03687795624136925,
      "fcm_dpo/margin": 55.622039794921875,
      "fcm_dpo/q_t": 0.40110480785369873,
      "grad_norm": 10.238993644714355,
      "learning_rate": 1.8977614860195296e-07,
      "logits/chosen": 0.7125971913337708,
      "logits/rejected": 0.6526937484741211,
      "logps/chosen": -163.61781311035156,
      "logps/ref_chosen": -54.44539260864258,
      "logps/ref_rejected": -74.5650863647461,
      "logps/rejected": -239.3595428466797,
      "loss": 1.087,
      "margin_dpo/margin_mean": 55.622039794921875,
      "margin_dpo/margin_std": 79.11244201660156,
      "step": 411
    },
    {
      "KL/chosen_KL_mean": -115.19978332519531,
      "KL/mean": -141.18414306640625,
      "KL/rejected_KL_mean": -167.16848754882812,
      "KL/std": 65.96856689453125,
      "epoch": 0.6228269085411943,
      "fcm_dpo/beta": 0.007769760675728321,
      "fcm_dpo/delta": -0.004255469888448715,
      "fcm_dpo/margin": 51.968719482421875,
      "fcm_dpo/q_t": 0.40769216418266296,
      "grad_norm": 12.5255708694458,
      "learning_rate": 1.8849371567184662e-07,
      "logits/chosen": 0.7548251152038574,
      "logits/rejected": 0.6842193603515625,
      "logps/chosen": -170.4478759765625,
      "logps/ref_chosen": -55.248085021972656,
      "logps/ref_rejected": -68.96623229980469,
      "logps/rejected": -236.1347198486328,
      "loss": 1.0951,
      "margin_dpo/margin_mean": 51.96871566772461,
      "margin_dpo/margin_std": 70.8434829711914,
      "step": 412
    },
    {
      "KL/chosen_KL_mean": -124.92682647705078,
      "KL/mean": -146.77780151367188,
      "KL/rejected_KL_mean": -168.62875366210938,
      "KL/std": 68.0516357421875,
      "epoch": 0.6243386243386243,
      "fcm_dpo/beta": 0.007867778651416302,
      "fcm_dpo/delta": 0.05814486742019653,
      "fcm_dpo/margin": 43.7019157409668,
      "fcm_dpo/q_t": 0.4224427342414856,
      "grad_norm": 14.508427619934082,
      "learning_rate": 1.872130032047302e-07,
      "logits/chosen": 0.5644307136535645,
      "logits/rejected": 0.5286777019500732,
      "logps/chosen": -193.6475830078125,
      "logps/ref_chosen": -68.72074890136719,
      "logps/ref_rejected": -78.76539611816406,
      "logps/rejected": -247.39414978027344,
      "loss": 1.1851,
      "margin_dpo/margin_mean": 43.70191955566406,
      "margin_dpo/margin_std": 87.32107543945312,
      "step": 413
    },
    {
      "KL/chosen_KL_mean": -109.9261474609375,
      "KL/mean": -138.1259765625,
      "KL/rejected_KL_mean": -166.32582092285156,
      "KL/std": 74.28414916992188,
      "epoch": 0.6258503401360545,
      "fcm_dpo/beta": 0.00784690398722887,
      "fcm_dpo/delta": -0.04451046139001846,
      "fcm_dpo/margin": 56.39968490600586,
      "fcm_dpo/q_t": 0.39813709259033203,
      "grad_norm": 13.054259300231934,
      "learning_rate": 1.8593404702488436e-07,
      "logits/chosen": 0.7380519509315491,
      "logits/rejected": 0.674519419670105,
      "logps/chosen": -164.06436157226562,
      "logps/ref_chosen": -54.138214111328125,
      "logps/ref_rejected": -74.65741729736328,
      "logps/rejected": -240.98324584960938,
      "loss": 1.0685,
      "margin_dpo/margin_mean": 56.399681091308594,
      "margin_dpo/margin_std": 72.53543853759766,
      "step": 414
    },
    {
      "KL/chosen_KL_mean": -110.6787109375,
      "KL/mean": -135.13711547851562,
      "KL/rejected_KL_mean": -159.5955047607422,
      "KL/std": 72.54624938964844,
      "epoch": 0.6273620559334845,
      "fcm_dpo/beta": 0.007824394851922989,
      "fcm_dpo/delta": 0.0178983211517334,
      "fcm_dpo/margin": 48.91680145263672,
      "fcm_dpo/q_t": 0.4126567244529724,
      "grad_norm": 12.129854202270508,
      "learning_rate": 1.846568829074628e-07,
      "logits/chosen": 0.7589465379714966,
      "logits/rejected": 0.7433362603187561,
      "logps/chosen": -166.59725952148438,
      "logps/ref_chosen": -55.91856002807617,
      "logps/ref_rejected": -61.747703552246094,
      "logps/rejected": -221.34320068359375,
      "loss": 1.1291,
      "margin_dpo/margin_mean": 48.91680145263672,
      "margin_dpo/margin_std": 78.5286865234375,
      "step": 415
    },
    {
      "KL/chosen_KL_mean": -116.60771942138672,
      "KL/mean": -137.52841186523438,
      "KL/rejected_KL_mean": -158.44911193847656,
      "KL/std": 74.35231018066406,
      "epoch": 0.6288737717309146,
      "fcm_dpo/beta": 0.008027415722608566,
      "fcm_dpo/delta": 0.06183099001646042,
      "fcm_dpo/margin": 41.841392517089844,
      "fcm_dpo/q_t": 0.42444467544555664,
      "grad_norm": 14.416735649108887,
      "learning_rate": 1.8338154657749128e-07,
      "logits/chosen": 0.7013071775436401,
      "logits/rejected": 0.6499512195587158,
      "logps/chosen": -171.330810546875,
      "logps/ref_chosen": -54.72308349609375,
      "logps/ref_rejected": -69.17388916015625,
      "logps/rejected": -227.6230010986328,
      "loss": 1.1802,
      "margin_dpo/margin_mean": 41.841392517089844,
      "margin_dpo/margin_std": 74.81488037109375,
      "step": 416
    },
    {
      "KL/chosen_KL_mean": -119.80218505859375,
      "KL/mean": -148.10443115234375,
      "KL/rejected_KL_mean": -176.40670776367188,
      "KL/std": 70.20147705078125,
      "epoch": 0.6303854875283447,
      "fcm_dpo/beta": 0.007916878908872604,
      "fcm_dpo/delta": -0.05051497370004654,
      "fcm_dpo/margin": 56.604530334472656,
      "fcm_dpo/q_t": 0.39684611558914185,
      "grad_norm": 12.716817855834961,
      "learning_rate": 1.8210807370886849e-07,
      "logits/chosen": 0.8077340126037598,
      "logits/rejected": 0.7370149493217468,
      "logps/chosen": -176.59344482421875,
      "logps/ref_chosen": -56.791259765625,
      "logps/ref_rejected": -68.7791748046875,
      "logps/rejected": -245.18588256835938,
      "loss": 1.0905,
      "margin_dpo/margin_mean": 56.604530334472656,
      "margin_dpo/margin_std": 83.14535522460938,
      "step": 417
    },
    {
      "KL/chosen_KL_mean": -128.22242736816406,
      "KL/mean": -149.95968627929688,
      "KL/rejected_KL_mean": -171.69696044921875,
      "KL/std": 73.53202056884766,
      "epoch": 0.6318972033257747,
      "fcm_dpo/beta": 0.007869708351790905,
      "fcm_dpo/delta": 0.059024084359407425,
      "fcm_dpo/margin": 43.47452163696289,
      "fcm_dpo/q_t": 0.42358940839767456,
      "grad_norm": 13.894302368164062,
      "learning_rate": 1.8083649992336825e-07,
      "logits/chosen": 0.7667458057403564,
      "logits/rejected": 0.7746227979660034,
      "logps/chosen": -197.33041381835938,
      "logps/ref_chosen": -69.10798645019531,
      "logps/ref_rejected": -75.09132385253906,
      "logps/rejected": -246.7882843017578,
      "loss": 1.1614,
      "margin_dpo/margin_mean": 43.47452163696289,
      "margin_dpo/margin_std": 74.61412048339844,
      "step": 418
    },
    {
      "KL/chosen_KL_mean": -102.58634948730469,
      "KL/mean": -134.0736083984375,
      "KL/rejected_KL_mean": -165.56088256835938,
      "KL/std": 73.33294677734375,
      "epoch": 0.6334089191232048,
      "fcm_dpo/beta": 0.007823498919606209,
      "fcm_dpo/delta": -0.09785882383584976,
      "fcm_dpo/margin": 62.974517822265625,
      "fcm_dpo/q_t": 0.3892057538032532,
      "grad_norm": 12.586299896240234,
      "learning_rate": 1.7956686078964255e-07,
      "logits/chosen": 0.6310225129127502,
      "logits/rejected": 0.5767621994018555,
      "logps/chosen": -160.75811767578125,
      "logps/ref_chosen": -58.1717643737793,
      "logps/ref_rejected": -71.67066955566406,
      "logps/rejected": -237.23153686523438,
      "loss": 1.0468,
      "margin_dpo/margin_mean": 62.97452163696289,
      "margin_dpo/margin_std": 81.59907531738281,
      "step": 419
    },
    {
      "KL/chosen_KL_mean": -129.5718231201172,
      "KL/mean": -146.15481567382812,
      "KL/rejected_KL_mean": -162.73782348632812,
      "KL/std": 73.85808563232422,
      "epoch": 0.6349206349206349,
      "fcm_dpo/beta": 0.007942350581288338,
      "fcm_dpo/delta": 0.14052259922027588,
      "fcm_dpo/margin": 33.16600036621094,
      "fcm_dpo/q_t": 0.4420320391654968,
      "grad_norm": 13.626842498779297,
      "learning_rate": 1.782991918222275e-07,
      "logits/chosen": 0.7083577513694763,
      "logits/rejected": 0.6613192558288574,
      "logps/chosen": -186.62533569335938,
      "logps/ref_chosen": -57.05351257324219,
      "logps/ref_rejected": -62.670982360839844,
      "logps/rejected": -225.4088134765625,
      "loss": 1.2491,
      "margin_dpo/margin_mean": 33.16600036621094,
      "margin_dpo/margin_std": 84.56375122070312,
      "step": 420
    },
    {
      "KL/chosen_KL_mean": -121.5405502319336,
      "KL/mean": -145.28880310058594,
      "KL/rejected_KL_mean": -169.03704833984375,
      "KL/std": 73.51364135742188,
      "epoch": 0.636432350718065,
      "fcm_dpo/beta": 0.008013014681637287,
      "fcm_dpo/delta": 0.0198320671916008,
      "fcm_dpo/margin": 47.49650573730469,
      "fcm_dpo/q_t": 0.41550660133361816,
      "grad_norm": 13.676653861999512,
      "learning_rate": 1.7703352848054887e-07,
      "logits/chosen": 0.6900692582130432,
      "logits/rejected": 0.6295258402824402,
      "logps/chosen": -178.86380004882812,
      "logps/ref_chosen": -57.32324981689453,
      "logps/ref_rejected": -75.33782958984375,
      "logps/rejected": -244.3748779296875,
      "loss": 1.1733,
      "margin_dpo/margin_mean": 47.49650573730469,
      "margin_dpo/margin_std": 92.68571472167969,
      "step": 421
    },
    {
      "KL/chosen_KL_mean": -105.38725280761719,
      "KL/mean": -136.9675750732422,
      "KL/rejected_KL_mean": -168.5478973388672,
      "KL/std": 72.90692138671875,
      "epoch": 0.6379440665154951,
      "fcm_dpo/beta": 0.007948427461087704,
      "fcm_dpo/delta": -0.10728380084037781,
      "fcm_dpo/margin": 63.160648345947266,
      "fcm_dpo/q_t": 0.38589245080947876,
      "grad_norm": 14.917747497558594,
      "learning_rate": 1.7576990616793137e-07,
      "logits/chosen": 0.735478937625885,
      "logits/rejected": 0.7193405628204346,
      "logps/chosen": -172.44482421875,
      "logps/ref_chosen": -67.05757141113281,
      "logps/ref_rejected": -72.12803649902344,
      "logps/rejected": -240.67593383789062,
      "loss": 1.0318,
      "margin_dpo/margin_mean": 63.160648345947266,
      "margin_dpo/margin_std": 76.23689270019531,
      "step": 422
    },
    {
      "KL/chosen_KL_mean": -109.40867614746094,
      "KL/mean": -140.88455200195312,
      "KL/rejected_KL_mean": -172.3604278564453,
      "KL/std": 75.55366516113281,
      "epoch": 0.6394557823129252,
      "fcm_dpo/beta": 0.007762731984257698,
      "fcm_dpo/delta": -0.09350243210792542,
      "fcm_dpo/margin": 62.951759338378906,
      "fcm_dpo/q_t": 0.3896293640136719,
      "grad_norm": 11.912402153015137,
      "learning_rate": 1.745083602306071e-07,
      "logits/chosen": 0.7740094661712646,
      "logits/rejected": 0.700740396976471,
      "logps/chosen": -163.47035217285156,
      "logps/ref_chosen": -54.06167221069336,
      "logps/ref_rejected": -76.64092254638672,
      "logps/rejected": -249.00135803222656,
      "loss": 1.0444,
      "margin_dpo/margin_mean": 62.95175552368164,
      "margin_dpo/margin_std": 79.72633361816406,
      "step": 423
    },
    {
      "KL/chosen_KL_mean": -117.83436584472656,
      "KL/mean": -148.37539672851562,
      "KL/rejected_KL_mean": -178.91644287109375,
      "KL/std": 72.689697265625,
      "epoch": 0.6409674981103552,
      "fcm_dpo/beta": 0.007616098970174789,
      "fcm_dpo/delta": -0.0695209950208664,
      "fcm_dpo/margin": 61.08207702636719,
      "fcm_dpo/q_t": 0.39452964067459106,
      "grad_norm": 16.555137634277344,
      "learning_rate": 1.7324892595672804e-07,
      "logits/chosen": 0.660463809967041,
      "logits/rejected": 0.62034010887146,
      "logps/chosen": -171.4432373046875,
      "logps/ref_chosen": -53.60887145996094,
      "logps/ref_rejected": -79.2139892578125,
      "logps/rejected": -258.13043212890625,
      "loss": 1.0664,
      "margin_dpo/margin_mean": 61.08207702636719,
      "margin_dpo/margin_std": 81.5423583984375,
      "step": 424
    },
    {
      "KL/chosen_KL_mean": -114.80853271484375,
      "KL/mean": -139.91513061523438,
      "KL/rejected_KL_mean": -165.02171325683594,
      "KL/std": 71.188232421875,
      "epoch": 0.6424792139077853,
      "fcm_dpo/beta": 0.00764683960005641,
      "fcm_dpo/delta": 0.01666775718331337,
      "fcm_dpo/margin": 50.21317672729492,
      "fcm_dpo/q_t": 0.41330617666244507,
      "grad_norm": 14.110657691955566,
      "learning_rate": 1.7199163857537824e-07,
      "logits/chosen": 0.7706844806671143,
      "logits/rejected": 0.7416727542877197,
      "logps/chosen": -173.22320556640625,
      "logps/ref_chosen": -58.41468048095703,
      "logps/ref_rejected": -66.59054565429688,
      "logps/rejected": -231.6122589111328,
      "loss": 1.1279,
      "margin_dpo/margin_mean": 50.21317672729492,
      "margin_dpo/margin_std": 80.54307556152344,
      "step": 425
    },
    {
      "KL/chosen_KL_mean": -138.42042541503906,
      "KL/mean": -153.50100708007812,
      "KL/rejected_KL_mean": -168.58157348632812,
      "KL/std": 72.29527282714844,
      "epoch": 0.6439909297052154,
      "fcm_dpo/beta": 0.007851322181522846,
      "fcm_dpo/delta": 0.1672021448612213,
      "fcm_dpo/margin": 30.16112518310547,
      "fcm_dpo/q_t": 0.4458393454551697,
      "grad_norm": 15.771385192871094,
      "learning_rate": 1.7073653325558828e-07,
      "logits/chosen": 0.657271683216095,
      "logits/rejected": 0.6661276817321777,
      "logps/chosen": -210.128662109375,
      "logps/ref_chosen": -71.70822143554688,
      "logps/ref_rejected": -73.57725524902344,
      "logps/rejected": -242.1588134765625,
      "loss": 1.2797,
      "margin_dpo/margin_mean": 30.1611270904541,
      "margin_dpo/margin_std": 89.65312957763672,
      "step": 426
    },
    {
      "KL/chosen_KL_mean": -129.74212646484375,
      "KL/mean": -153.55258178710938,
      "KL/rejected_KL_mean": -177.36302185058594,
      "KL/std": 77.05906677246094,
      "epoch": 0.6455026455026455,
      "fcm_dpo/beta": 0.007935117930173874,
      "fcm_dpo/delta": 0.022978732362389565,
      "fcm_dpo/margin": 47.620872497558594,
      "fcm_dpo/q_t": 0.41612735390663147,
      "grad_norm": 14.509443283081055,
      "learning_rate": 1.6948364510535218e-07,
      "logits/chosen": 0.7272225618362427,
      "logits/rejected": 0.6681046485900879,
      "logps/chosen": -188.38490295410156,
      "logps/ref_chosen": -58.64276885986328,
      "logps/ref_rejected": -86.25437927246094,
      "logps/rejected": -263.6174011230469,
      "loss": 1.1563,
      "margin_dpo/margin_mean": 47.620872497558594,
      "margin_dpo/margin_std": 87.77416229248047,
      "step": 427
    },
    {
      "KL/chosen_KL_mean": -121.83946990966797,
      "KL/mean": -150.60853576660156,
      "KL/rejected_KL_mean": -179.37759399414062,
      "KL/std": 78.71310424804688,
      "epoch": 0.6470143613000756,
      "fcm_dpo/beta": 0.007923326455056667,
      "fcm_dpo/delta": -0.058716095983982086,
      "fcm_dpo/margin": 57.53812789916992,
      "fcm_dpo/q_t": 0.39860618114471436,
      "grad_norm": 12.580979347229004,
      "learning_rate": 1.6823300917064458e-07,
      "logits/chosen": 0.6929279565811157,
      "logits/rejected": 0.6461859941482544,
      "logps/chosen": -188.43551635742188,
      "logps/ref_chosen": -66.5960464477539,
      "logps/ref_rejected": -82.3941650390625,
      "logps/rejected": -261.7717590332031,
      "loss": 1.0884,
      "margin_dpo/margin_mean": 57.538124084472656,
      "margin_dpo/margin_std": 85.31922912597656,
      "step": 428
    },
    {
      "KL/chosen_KL_mean": -126.24545288085938,
      "KL/mean": -148.31695556640625,
      "KL/rejected_KL_mean": -170.38845825195312,
      "KL/std": 72.1160888671875,
      "epoch": 0.6485260770975056,
      "fcm_dpo/beta": 0.007961621508002281,
      "fcm_dpo/delta": 0.04956657066941261,
      "fcm_dpo/margin": 44.14300537109375,
      "fcm_dpo/q_t": 0.4205780029296875,
      "grad_norm": 14.980849266052246,
      "learning_rate": 1.669846604344412e-07,
      "logits/chosen": 0.6782246828079224,
      "logits/rejected": 0.6961945295333862,
      "logps/chosen": -183.25515747070312,
      "logps/ref_chosen": -57.00970458984375,
      "logps/ref_rejected": -59.86549377441406,
      "logps/rejected": -230.2539520263672,
      "loss": 1.1697,
      "margin_dpo/margin_mean": 44.14300537109375,
      "margin_dpo/margin_std": 82.05901336669922,
      "step": 429
    },
    {
      "KL/chosen_KL_mean": -113.78839111328125,
      "KL/mean": -149.21728515625,
      "KL/rejected_KL_mean": -184.64617919921875,
      "KL/std": 74.32037353515625,
      "epoch": 0.6500377928949358,
      "fcm_dpo/beta": 0.0077576362527906895,
      "fcm_dpo/delta": -0.15839435160160065,
      "fcm_dpo/margin": 70.8577880859375,
      "fcm_dpo/q_t": 0.37446311116218567,
      "grad_norm": 12.859457015991211,
      "learning_rate": 1.6573863381573954e-07,
      "logits/chosen": 0.6114371418952942,
      "logits/rejected": 0.6136624813079834,
      "logps/chosen": -173.35159301757812,
      "logps/ref_chosen": -59.563194274902344,
      "logps/ref_rejected": -70.52289581298828,
      "logps/rejected": -255.16908264160156,
      "loss": 1.0126,
      "margin_dpo/margin_mean": 70.85778045654297,
      "margin_dpo/margin_std": 84.95809936523438,
      "step": 430
    },
    {
      "KL/chosen_KL_mean": -110.34442901611328,
      "KL/mean": -134.9564971923828,
      "KL/rejected_KL_mean": -159.56854248046875,
      "KL/std": 74.60652160644531,
      "epoch": 0.6515495086923658,
      "fcm_dpo/beta": 0.00771299796178937,
      "fcm_dpo/delta": 0.021141374483704567,
      "fcm_dpo/margin": 49.22412872314453,
      "fcm_dpo/q_t": 0.41382941603660583,
      "grad_norm": 12.430391311645508,
      "learning_rate": 1.6449496416858282e-07,
      "logits/chosen": 0.7100223302841187,
      "logits/rejected": 0.6575336456298828,
      "logps/chosen": -160.5447540283203,
      "logps/ref_chosen": -50.20032501220703,
      "logps/ref_rejected": -77.81680297851562,
      "logps/rejected": -237.38534545898438,
      "loss": 1.1325,
      "margin_dpo/margin_mean": 49.224124908447266,
      "margin_dpo/margin_std": 80.1656494140625,
      "step": 431
    },
    {
      "KL/chosen_KL_mean": -115.60556030273438,
      "KL/mean": -142.0184326171875,
      "KL/rejected_KL_mean": -168.4313201904297,
      "KL/std": 71.81451416015625,
      "epoch": 0.6530612244897959,
      "fcm_dpo/beta": 0.007714337669312954,
      "fcm_dpo/delta": -0.00784326996654272,
      "fcm_dpo/margin": 52.82575607299805,
      "fcm_dpo/q_t": 0.40753287076950073,
      "grad_norm": 12.736228942871094,
      "learning_rate": 1.632536862810844e-07,
      "logits/chosen": 0.714253306388855,
      "logits/rejected": 0.6646943092346191,
      "logps/chosen": -177.268310546875,
      "logps/ref_chosen": -61.662757873535156,
      "logps/ref_rejected": -83.94496154785156,
      "logps/rejected": -252.37628173828125,
      "loss": 1.1177,
      "margin_dpo/margin_mean": 52.82575988769531,
      "margin_dpo/margin_std": 84.00270080566406,
      "step": 432
    },
    {
      "KL/chosen_KL_mean": -114.59454345703125,
      "KL/mean": -147.77920532226562,
      "KL/rejected_KL_mean": -180.96388244628906,
      "KL/std": 71.03900146484375,
      "epoch": 0.654572940287226,
      "fcm_dpo/beta": 0.007613149471580982,
      "fcm_dpo/delta": -0.1107371374964714,
      "fcm_dpo/margin": 66.36933135986328,
      "fcm_dpo/q_t": 0.3842705488204956,
      "grad_norm": 13.615633010864258,
      "learning_rate": 1.6201483487445515e-07,
      "logits/chosen": 0.8374870419502258,
      "logits/rejected": 0.8358460664749146,
      "logps/chosen": -178.32373046875,
      "logps/ref_chosen": -63.72917938232422,
      "logps/ref_rejected": -65.8391342163086,
      "logps/rejected": -246.80300903320312,
      "loss": 1.0352,
      "margin_dpo/margin_mean": 66.36933135986328,
      "margin_dpo/margin_std": 81.50929260253906,
      "step": 433
    },
    {
      "KL/chosen_KL_mean": -96.70651245117188,
      "KL/mean": -131.61221313476562,
      "KL/rejected_KL_mean": -166.51791381835938,
      "KL/std": 76.29074096679688,
      "epoch": 0.656084656084656,
      "fcm_dpo/beta": 0.007363257929682732,
      "fcm_dpo/delta": -0.12183529138565063,
      "fcm_dpo/margin": 69.8114013671875,
      "fcm_dpo/q_t": 0.3824422359466553,
      "grad_norm": 12.417044639587402,
      "learning_rate": 1.6077844460203204e-07,
      "logits/chosen": 0.8204172849655151,
      "logits/rejected": 0.7561012506484985,
      "logps/chosen": -144.67984008789062,
      "logps/ref_chosen": -47.97331619262695,
      "logps/ref_rejected": -72.51132202148438,
      "logps/rejected": -239.02923583984375,
      "loss": 1.0481,
      "margin_dpo/margin_mean": 69.81139373779297,
      "margin_dpo/margin_std": 92.68022155761719,
      "step": 434
    },
    {
      "KL/chosen_KL_mean": -118.66361999511719,
      "KL/mean": -146.527587890625,
      "KL/rejected_KL_mean": -174.39151000976562,
      "KL/std": 73.32241821289062,
      "epoch": 0.6575963718820862,
      "fcm_dpo/beta": 0.0073859672993421555,
      "fcm_dpo/delta": -0.01251184195280075,
      "fcm_dpo/margin": 55.7279052734375,
      "fcm_dpo/q_t": 0.4067944884300232,
      "grad_norm": 13.418008804321289,
      "learning_rate": 1.5954455004830878e-07,
      "logits/chosen": 0.8510400056838989,
      "logits/rejected": 0.8107461333274841,
      "logps/chosen": -175.723876953125,
      "logps/ref_chosen": -57.06024932861328,
      "logps/ref_rejected": -71.69146728515625,
      "logps/rejected": -246.08297729492188,
      "loss": 1.1112,
      "margin_dpo/margin_mean": 55.727901458740234,
      "margin_dpo/margin_std": 85.01725006103516,
      "step": 435
    },
    {
      "KL/chosen_KL_mean": -116.6640853881836,
      "KL/mean": -139.56753540039062,
      "KL/rejected_KL_mean": -162.47097778320312,
      "KL/std": 72.98526000976562,
      "epoch": 0.6591080876795162,
      "fcm_dpo/beta": 0.007416677661240101,
      "fcm_dpo/delta": 0.06236447021365166,
      "fcm_dpo/margin": 45.806915283203125,
      "fcm_dpo/q_t": 0.4232286810874939,
      "grad_norm": 14.484575271606445,
      "learning_rate": 1.5831318572796847e-07,
      "logits/chosen": 0.7485842108726501,
      "logits/rejected": 0.690552830696106,
      "logps/chosen": -172.8221435546875,
      "logps/ref_chosen": -56.158050537109375,
      "logps/ref_rejected": -67.63787841796875,
      "logps/rejected": -230.10885620117188,
      "loss": 1.1785,
      "margin_dpo/margin_mean": 45.80691146850586,
      "margin_dpo/margin_std": 89.24258422851562,
      "step": 436
    },
    {
      "KL/chosen_KL_mean": -123.96295928955078,
      "KL/mean": -149.2787322998047,
      "KL/rejected_KL_mean": -174.59451293945312,
      "KL/std": 75.40545654296875,
      "epoch": 0.6606198034769464,
      "fcm_dpo/beta": 0.0073969424702227116,
      "fcm_dpo/delta": 0.024935893714427948,
      "fcm_dpo/margin": 50.63155746459961,
      "fcm_dpo/q_t": 0.41527464985847473,
      "grad_norm": 16.458776473999023,
      "learning_rate": 1.5708438608491815e-07,
      "logits/chosen": 0.6856381893157959,
      "logits/rejected": 0.5573318004608154,
      "logps/chosen": -180.94873046875,
      "logps/ref_chosen": -56.98578643798828,
      "logps/ref_rejected": -85.61524963378906,
      "logps/rejected": -260.20977783203125,
      "loss": 1.1665,
      "margin_dpo/margin_mean": 50.63155746459961,
      "margin_dpo/margin_std": 93.0724868774414,
      "step": 437
    },
    {
      "KL/chosen_KL_mean": -106.18060302734375,
      "KL/mean": -141.49575805664062,
      "KL/rejected_KL_mean": -176.81088256835938,
      "KL/std": 79.54093933105469,
      "epoch": 0.6621315192743764,
      "fcm_dpo/beta": 0.007361456751823425,
      "fcm_dpo/delta": -0.12632791697978973,
      "fcm_dpo/margin": 70.63029479980469,
      "fcm_dpo/q_t": 0.38328486680984497,
      "grad_norm": 12.317891120910645,
      "learning_rate": 1.558581854913253e-07,
      "logits/chosen": 0.7801381349563599,
      "logits/rejected": 0.7134337425231934,
      "logps/chosen": -147.4583740234375,
      "logps/ref_chosen": -41.27777862548828,
      "logps/ref_rejected": -65.33840942382812,
      "logps/rejected": -242.1492919921875,
      "loss": 1.023,
      "margin_dpo/margin_mean": 70.63029479980469,
      "margin_dpo/margin_std": 84.64463806152344,
      "step": 438
    },
    {
      "KL/chosen_KL_mean": -119.46957397460938,
      "KL/mean": -148.65093994140625,
      "KL/rejected_KL_mean": -177.83230590820312,
      "KL/std": 80.0428237915039,
      "epoch": 0.6636432350718064,
      "fcm_dpo/beta": 0.0072980402037501335,
      "fcm_dpo/delta": -0.027525458484888077,
      "fcm_dpo/margin": 58.362728118896484,
      "fcm_dpo/q_t": 0.4030565023422241,
      "grad_norm": 12.534804344177246,
      "learning_rate": 1.5463461824665658e-07,
      "logits/chosen": 0.626327633857727,
      "logits/rejected": 0.5900809168815613,
      "logps/chosen": -200.88723754882812,
      "logps/ref_chosen": -81.41764831542969,
      "logps/ref_rejected": -94.72309875488281,
      "logps/rejected": -272.555419921875,
      "loss": 1.0883,
      "margin_dpo/margin_mean": 58.362728118896484,
      "margin_dpo/margin_std": 80.85906982421875,
      "step": 439
    },
    {
      "KL/chosen_KL_mean": -106.9936294555664,
      "KL/mean": -138.42535400390625,
      "KL/rejected_KL_mean": -169.8570556640625,
      "KL/std": 74.2397689819336,
      "epoch": 0.6651549508692366,
      "fcm_dpo/beta": 0.007225216832011938,
      "fcm_dpo/delta": -0.05703933909535408,
      "fcm_dpo/margin": 62.86344528198242,
      "fcm_dpo/q_t": 0.3970058262348175,
      "grad_norm": 21.790699005126953,
      "learning_rate": 1.534137185767178e-07,
      "logits/chosen": 0.7353362441062927,
      "logits/rejected": 0.6341665387153625,
      "logps/chosen": -149.5318145751953,
      "logps/ref_chosen": -42.538185119628906,
      "logps/ref_rejected": -69.78813934326172,
      "logps/rejected": -239.64520263671875,
      "loss": 1.0815,
      "margin_dpo/margin_mean": 62.863441467285156,
      "margin_dpo/margin_std": 88.48552703857422,
      "step": 440
    },
    {
      "KL/chosen_KL_mean": -104.67693328857422,
      "KL/mean": -137.52198791503906,
      "KL/rejected_KL_mean": -170.36705017089844,
      "KL/std": 76.4732666015625,
      "epoch": 0.6666666666666666,
      "fcm_dpo/beta": 0.007063678000122309,
      "fcm_dpo/delta": -0.06781375408172607,
      "fcm_dpo/margin": 65.69010925292969,
      "fcm_dpo/q_t": 0.3922446072101593,
      "grad_norm": 13.523235321044922,
      "learning_rate": 1.521955206326976e-07,
      "logits/chosen": 0.6944692134857178,
      "logits/rejected": 0.5998711585998535,
      "logps/chosen": -162.27015686035156,
      "logps/ref_chosen": -57.593223571777344,
      "logps/ref_rejected": -84.82878875732422,
      "logps/rejected": -255.19583129882812,
      "loss": 1.0366,
      "margin_dpo/margin_mean": 65.69010925292969,
      "margin_dpo/margin_std": 71.41874694824219,
      "step": 441
    },
    {
      "KL/chosen_KL_mean": -125.55252075195312,
      "KL/mean": -157.33255004882812,
      "KL/rejected_KL_mean": -189.11260986328125,
      "KL/std": 75.97675323486328,
      "epoch": 0.6681783824640968,
      "fcm_dpo/beta": 0.0070096999406814575,
      "fcm_dpo/delta": -0.04777270928025246,
      "fcm_dpo/margin": 63.56007766723633,
      "fcm_dpo/q_t": 0.3974016606807709,
      "grad_norm": 15.098261833190918,
      "learning_rate": 1.5098005849021078e-07,
      "logits/chosen": 0.7284977436065674,
      "logits/rejected": 0.6738515496253967,
      "logps/chosen": -193.01373291015625,
      "logps/ref_chosen": -67.46121978759766,
      "logps/ref_rejected": -89.0693588256836,
      "logps/rejected": -278.18194580078125,
      "loss": 1.0661,
      "margin_dpo/margin_mean": 63.56007385253906,
      "margin_dpo/margin_std": 81.50393676757812,
      "step": 442
    },
    {
      "KL/chosen_KL_mean": -104.5126953125,
      "KL/mean": -144.3497772216797,
      "KL/rejected_KL_mean": -184.18685913085938,
      "KL/std": 81.44729614257812,
      "epoch": 0.6696900982615268,
      "fcm_dpo/beta": 0.0068482570350170135,
      "fcm_dpo/delta": -0.15395689010620117,
      "fcm_dpo/margin": 79.67416381835938,
      "fcm_dpo/q_t": 0.3748539090156555,
      "grad_norm": 13.461181640625,
      "learning_rate": 1.4976736614834662e-07,
      "logits/chosen": 0.7253681421279907,
      "logits/rejected": 0.6550537943840027,
      "logps/chosen": -159.30880737304688,
      "logps/ref_chosen": -54.79610061645508,
      "logps/ref_rejected": -77.80781555175781,
      "logps/rejected": -261.99468994140625,
      "loss": 1.0049,
      "margin_dpo/margin_mean": 79.67416381835938,
      "margin_dpo/margin_std": 91.89553833007812,
      "step": 443
    },
    {
      "KL/chosen_KL_mean": -131.12448120117188,
      "KL/mean": -145.68206787109375,
      "KL/rejected_KL_mean": -160.23968505859375,
      "KL/std": 78.9239501953125,
      "epoch": 0.671201814058957,
      "fcm_dpo/beta": 0.006958891171962023,
      "fcm_dpo/delta": 0.20196352899074554,
      "fcm_dpo/margin": 29.115209579467773,
      "fcm_dpo/q_t": 0.45473960041999817,
      "grad_norm": 17.20579719543457,
      "learning_rate": 1.4855747752871654e-07,
      "logits/chosen": 0.737562894821167,
      "logits/rejected": 0.6410528421401978,
      "logps/chosen": -189.87353515625,
      "logps/ref_chosen": -58.749061584472656,
      "logps/ref_rejected": -86.87396240234375,
      "logps/rejected": -247.1136474609375,
      "loss": 1.2886,
      "margin_dpo/margin_mean": 29.11520767211914,
      "margin_dpo/margin_std": 87.74626159667969,
      "step": 444
    },
    {
      "KL/chosen_KL_mean": -123.03514099121094,
      "KL/mean": -157.38128662109375,
      "KL/rejected_KL_mean": -191.7274169921875,
      "KL/std": 77.95574951171875,
      "epoch": 0.672713529856387,
      "fcm_dpo/beta": 0.006988787557929754,
      "fcm_dpo/delta": -0.0840548425912857,
      "fcm_dpo/margin": 68.69226837158203,
      "fcm_dpo/q_t": 0.39024075865745544,
      "grad_norm": 15.367069244384766,
      "learning_rate": 1.473504264745062e-07,
      "logits/chosen": 0.7162766456604004,
      "logits/rejected": 0.7037680149078369,
      "logps/chosen": -183.95257568359375,
      "logps/ref_chosen": -60.91743850708008,
      "logps/ref_rejected": -71.5637435913086,
      "logps/rejected": -263.2911376953125,
      "loss": 1.0475,
      "margin_dpo/margin_mean": 68.6922607421875,
      "margin_dpo/margin_std": 86.09353637695312,
      "step": 445
    },
    {
      "KL/chosen_KL_mean": -111.1749267578125,
      "KL/mean": -146.70782470703125,
      "KL/rejected_KL_mean": -182.24072265625,
      "KL/std": 79.02342224121094,
      "epoch": 0.674225245653817,
      "fcm_dpo/beta": 0.0069199297577142715,
      "fcm_dpo/delta": -0.09750378131866455,
      "fcm_dpo/margin": 71.06580352783203,
      "fcm_dpo/q_t": 0.3867424726486206,
      "grad_norm": 11.384917259216309,
      "learning_rate": 1.461462467495284e-07,
      "logits/chosen": 0.7947849631309509,
      "logits/rejected": 0.7108563184738159,
      "logps/chosen": -159.97418212890625,
      "logps/ref_chosen": -48.79924774169922,
      "logps/ref_rejected": -71.8719482421875,
      "logps/rejected": -254.1126708984375,
      "loss": 1.0243,
      "margin_dpo/margin_mean": 71.06580352783203,
      "margin_dpo/margin_std": 73.57239532470703,
      "step": 446
    },
    {
      "KL/chosen_KL_mean": -109.34071350097656,
      "KL/mean": -149.2745361328125,
      "KL/rejected_KL_mean": -189.2083740234375,
      "KL/std": 78.32331848144531,
      "epoch": 0.6757369614512472,
      "fcm_dpo/beta": 0.006653862539678812,
      "fcm_dpo/delta": -0.13958829641342163,
      "fcm_dpo/margin": 79.86766052246094,
      "fcm_dpo/q_t": 0.37667399644851685,
      "grad_norm": 16.19961929321289,
      "learning_rate": 1.4494497203727843e-07,
      "logits/chosen": 0.677545428276062,
      "logits/rejected": 0.5737296342849731,
      "logps/chosen": -163.0234375,
      "logps/ref_chosen": -53.682716369628906,
      "logps/ref_rejected": -88.17315673828125,
      "logps/rejected": -277.38153076171875,
      "loss": 1.0133,
      "margin_dpo/margin_mean": 79.86766052246094,
      "margin_dpo/margin_std": 91.65660095214844,
      "step": 447
    },
    {
      "KL/chosen_KL_mean": -116.87925720214844,
      "KL/mean": -148.92613220214844,
      "KL/rejected_KL_mean": -180.9730224609375,
      "KL/std": 78.32139587402344,
      "epoch": 0.6772486772486772,
      "fcm_dpo/beta": 0.006604928523302078,
      "fcm_dpo/delta": -0.024372804909944534,
      "fcm_dpo/margin": 64.09376525878906,
      "fcm_dpo/q_t": 0.4019731283187866,
      "grad_norm": 11.611702919006348,
      "learning_rate": 1.4374663593999256e-07,
      "logits/chosen": 0.7389341592788696,
      "logits/rejected": 0.6849086880683899,
      "logps/chosen": -170.63050842285156,
      "logps/ref_chosen": -53.75125503540039,
      "logps/ref_rejected": -77.17623901367188,
      "logps/rejected": -258.1492614746094,
      "loss": 1.0813,
      "margin_dpo/margin_mean": 64.09376525878906,
      "margin_dpo/margin_std": 84.98629760742188,
      "step": 448
    },
    {
      "KL/chosen_KL_mean": -138.70216369628906,
      "KL/mean": -153.8754425048828,
      "KL/rejected_KL_mean": -169.04872131347656,
      "KL/std": 79.18728637695312,
      "epoch": 0.6787603930461074,
      "fcm_dpo/beta": 0.006783435121178627,
      "fcm_dpo/delta": 0.19855058193206787,
      "fcm_dpo/margin": 30.346546173095703,
      "fcm_dpo/q_t": 0.45337286591529846,
      "grad_norm": 18.612869262695312,
      "learning_rate": 1.4255127197770707e-07,
      "logits/chosen": 0.5823932886123657,
      "logits/rejected": 0.5809487104415894,
      "logps/chosen": -214.529541015625,
      "logps/ref_chosen": -75.82737731933594,
      "logps/ref_rejected": -82.20687866210938,
      "logps/rejected": -251.25558471679688,
      "loss": 1.2588,
      "margin_dpo/margin_mean": 30.346548080444336,
      "margin_dpo/margin_std": 77.39900970458984,
      "step": 449
    },
    {
      "KL/chosen_KL_mean": -117.12738037109375,
      "KL/mean": -141.69601440429688,
      "KL/rejected_KL_mean": -166.26466369628906,
      "KL/std": 79.31319427490234,
      "epoch": 0.6802721088435374,
      "fcm_dpo/beta": 0.006913709454238415,
      "fcm_dpo/delta": 0.06243385374546051,
      "fcm_dpo/margin": 49.137298583984375,
      "fcm_dpo/q_t": 0.4247365891933441,
      "grad_norm": 13.703746795654297,
      "learning_rate": 1.4135891358732205e-07,
      "logits/chosen": 0.8435397148132324,
      "logits/rejected": 0.7237097024917603,
      "logps/chosen": -164.24310302734375,
      "logps/ref_chosen": -47.11572265625,
      "logps/ref_rejected": -78.7546615600586,
      "logps/rejected": -245.0193328857422,
      "loss": 1.1683,
      "margin_dpo/margin_mean": 49.137298583984375,
      "margin_dpo/margin_std": 90.78799438476562,
      "step": 450
    },
    {
      "KL/chosen_KL_mean": -117.35845184326172,
      "KL/mean": -139.92562866210938,
      "KL/rejected_KL_mean": -162.49281311035156,
      "KL/std": 78.64613342285156,
      "epoch": 0.6817838246409675,
      "fcm_dpo/beta": 0.007054210640490055,
      "fcm_dpo/delta": 0.08403199911117554,
      "fcm_dpo/margin": 45.134361267089844,
      "fcm_dpo/q_t": 0.4276706576347351,
      "grad_norm": 13.453941345214844,
      "learning_rate": 1.4016959412166437e-07,
      "logits/chosen": 0.7027239799499512,
      "logits/rejected": 0.6480429172515869,
      "logps/chosen": -180.70889282226562,
      "logps/ref_chosen": -63.350440979003906,
      "logps/ref_rejected": -76.28530883789062,
      "logps/rejected": -238.77810668945312,
      "loss": 1.1803,
      "margin_dpo/margin_mean": 45.134361267089844,
      "margin_dpo/margin_std": 86.11540222167969,
      "step": 451
    },
    {
      "KL/chosen_KL_mean": -115.05659484863281,
      "KL/mean": -141.87234497070312,
      "KL/rejected_KL_mean": -168.6881103515625,
      "KL/std": 74.24472045898438,
      "epoch": 0.6832955404383976,
      "fcm_dpo/beta": 0.007099970243871212,
      "fcm_dpo/delta": 0.01991662010550499,
      "fcm_dpo/margin": 53.63151550292969,
      "fcm_dpo/q_t": 0.413438618183136,
      "grad_norm": 14.759235382080078,
      "learning_rate": 1.3898334684855645e-07,
      "logits/chosen": 0.685683012008667,
      "logits/rejected": 0.6011617183685303,
      "logps/chosen": -170.64242553710938,
      "logps/ref_chosen": -55.58583450317383,
      "logps/ref_rejected": -77.68738555908203,
      "logps/rejected": -246.37548828125,
      "loss": 1.1438,
      "margin_dpo/margin_mean": 53.63151550292969,
      "margin_dpo/margin_std": 92.3481674194336,
      "step": 452
    },
    {
      "KL/chosen_KL_mean": -111.09956359863281,
      "KL/mean": -140.02688598632812,
      "KL/rejected_KL_mean": -168.95423889160156,
      "KL/std": 77.94131469726562,
      "epoch": 0.6848072562358276,
      "fcm_dpo/beta": 0.007077625021338463,
      "fcm_dpo/delta": -0.009929983876645565,
      "fcm_dpo/margin": 57.85468292236328,
      "fcm_dpo/q_t": 0.4080149531364441,
      "grad_norm": 15.38828182220459,
      "learning_rate": 1.3780020494988445e-07,
      "logits/chosen": 0.6994049549102783,
      "logits/rejected": 0.6696133613586426,
      "logps/chosen": -172.8777618408203,
      "logps/ref_chosen": -61.778202056884766,
      "logps/ref_rejected": -71.51403045654297,
      "logps/rejected": -240.46826171875,
      "loss": 1.116,
      "margin_dpo/margin_mean": 57.85468292236328,
      "margin_dpo/margin_std": 91.03846740722656,
      "step": 453
    },
    {
      "KL/chosen_KL_mean": -104.55786895751953,
      "KL/mean": -136.8068084716797,
      "KL/rejected_KL_mean": -169.05575561523438,
      "KL/std": 79.33438110351562,
      "epoch": 0.6863189720332578,
      "fcm_dpo/beta": 0.007016636431217194,
      "fcm_dpo/delta": -0.05515647307038307,
      "fcm_dpo/margin": 64.49788665771484,
      "fcm_dpo/q_t": 0.3954145610332489,
      "grad_norm": 13.037610054016113,
      "learning_rate": 1.366202015206706e-07,
      "logits/chosen": 0.7593111991882324,
      "logits/rejected": 0.7165501117706299,
      "logps/chosen": -156.15301513671875,
      "logps/ref_chosen": -51.59515380859375,
      "logps/ref_rejected": -63.96732711791992,
      "logps/rejected": -233.0230712890625,
      "loss": 1.0717,
      "margin_dpo/margin_mean": 64.49789428710938,
      "margin_dpo/margin_std": 87.31886291503906,
      "step": 454
    },
    {
      "KL/chosen_KL_mean": -120.95310974121094,
      "KL/mean": -151.56578063964844,
      "KL/rejected_KL_mean": -182.17843627929688,
      "KL/std": 77.6214599609375,
      "epoch": 0.6878306878306878,
      "fcm_dpo/beta": 0.006965038366615772,
      "fcm_dpo/delta": -0.027724087238311768,
      "fcm_dpo/margin": 61.225337982177734,
      "fcm_dpo/q_t": 0.40344443917274475,
      "grad_norm": 12.801860809326172,
      "learning_rate": 1.354433695681474e-07,
      "logits/chosen": 0.6048588156700134,
      "logits/rejected": 0.570784330368042,
      "logps/chosen": -191.6048126220703,
      "logps/ref_chosen": -70.65170288085938,
      "logps/ref_rejected": -77.44276428222656,
      "logps/rejected": -259.6212158203125,
      "loss": 1.0845,
      "margin_dpo/margin_mean": 61.22533416748047,
      "margin_dpo/margin_std": 84.29511260986328,
      "step": 455
    },
    {
      "KL/chosen_KL_mean": -121.42324829101562,
      "KL/mean": -147.92445373535156,
      "KL/rejected_KL_mean": -174.42567443847656,
      "KL/std": 77.14093017578125,
      "epoch": 0.6893424036281179,
      "fcm_dpo/beta": 0.006989809684455395,
      "fcm_dpo/delta": 0.0306635070592165,
      "fcm_dpo/margin": 53.002410888671875,
      "fcm_dpo/q_t": 0.4164670705795288,
      "grad_norm": 15.369559288024902,
      "learning_rate": 1.3426974201083439e-07,
      "logits/chosen": 0.6348047852516174,
      "logits/rejected": 0.5682092905044556,
      "logps/chosen": -177.821533203125,
      "logps/ref_chosen": -56.398284912109375,
      "logps/ref_rejected": -82.61642456054688,
      "logps/rejected": -257.0420837402344,
      "loss": 1.1384,
      "margin_dpo/margin_mean": 53.002410888671875,
      "margin_dpo/margin_std": 87.64601135253906,
      "step": 456
    },
    {
      "KL/chosen_KL_mean": -118.45246887207031,
      "KL/mean": -147.9862060546875,
      "KL/rejected_KL_mean": -177.5199737548828,
      "KL/std": 78.8670883178711,
      "epoch": 0.690854119425548,
      "fcm_dpo/beta": 0.0069747138768434525,
      "fcm_dpo/delta": -0.012739144265651703,
      "fcm_dpo/margin": 59.06749725341797,
      "fcm_dpo/q_t": 0.40447694063186646,
      "grad_norm": 11.998747825622559,
      "learning_rate": 1.3309935167761717e-07,
      "logits/chosen": 0.808464765548706,
      "logits/rejected": 0.7306153774261475,
      "logps/chosen": -163.17303466796875,
      "logps/ref_chosen": -44.72057342529297,
      "logps/ref_rejected": -68.1158676147461,
      "logps/rejected": -245.63583374023438,
      "loss": 1.0837,
      "margin_dpo/margin_mean": 59.06749725341797,
      "margin_dpo/margin_std": 76.19584655761719,
      "step": 457
    },
    {
      "KL/chosen_KL_mean": -114.44857788085938,
      "KL/mean": -144.8822479248047,
      "KL/rejected_KL_mean": -175.31591796875,
      "KL/std": 81.75318908691406,
      "epoch": 0.6923658352229781,
      "fcm_dpo/beta": 0.0069658150896430016,
      "fcm_dpo/delta": -0.02505449205636978,
      "fcm_dpo/margin": 60.867340087890625,
      "fcm_dpo/q_t": 0.4031820595264435,
      "grad_norm": 13.551817893981934,
      "learning_rate": 1.3193223130682936e-07,
      "logits/chosen": 0.7497356534004211,
      "logits/rejected": 0.6261578798294067,
      "logps/chosen": -164.4542694091797,
      "logps/ref_chosen": -50.00569152832031,
      "logps/ref_rejected": -87.50015258789062,
      "logps/rejected": -262.8160705566406,
      "loss": 1.1009,
      "margin_dpo/margin_mean": 60.867340087890625,
      "margin_dpo/margin_std": 91.40193176269531,
      "step": 458
    },
    {
      "KL/chosen_KL_mean": -106.50845336914062,
      "KL/mean": -144.07794189453125,
      "KL/rejected_KL_mean": -181.64743041992188,
      "KL/std": 86.97746276855469,
      "epoch": 0.6938775510204082,
      "fcm_dpo/beta": 0.006888561882078648,
      "fcm_dpo/delta": -0.12443508207798004,
      "fcm_dpo/margin": 75.13898468017578,
      "fcm_dpo/q_t": 0.38124239444732666,
      "grad_norm": 12.490287780761719,
      "learning_rate": 1.3076841354533658e-07,
      "logits/chosen": 0.7258360981941223,
      "logits/rejected": 0.690863847732544,
      "logps/chosen": -171.8863983154297,
      "logps/ref_chosen": -65.37794494628906,
      "logps/ref_rejected": -88.19244384765625,
      "logps/rejected": -269.8398742675781,
      "loss": 1.0209,
      "margin_dpo/margin_mean": 75.13898468017578,
      "margin_dpo/margin_std": 84.60592651367188,
      "step": 459
    },
    {
      "KL/chosen_KL_mean": -120.4828109741211,
      "KL/mean": -156.8914794921875,
      "KL/rejected_KL_mean": -193.3001708984375,
      "KL/std": 80.62806701660156,
      "epoch": 0.6953892668178382,
      "fcm_dpo/beta": 0.006671781651675701,
      "fcm_dpo/delta": -0.09077716618776321,
      "fcm_dpo/margin": 72.81735229492188,
      "fcm_dpo/q_t": 0.38928472995758057,
      "grad_norm": 12.194050788879395,
      "learning_rate": 1.2960793094762345e-07,
      "logits/chosen": 0.7333745956420898,
      "logits/rejected": 0.6062943935394287,
      "logps/chosen": -185.0444793701172,
      "logps/ref_chosen": -64.5616683959961,
      "logps/ref_rejected": -88.67890167236328,
      "logps/rejected": -281.97906494140625,
      "loss": 1.0364,
      "margin_dpo/margin_mean": 72.81735229492188,
      "margin_dpo/margin_std": 86.62645721435547,
      "step": 460
    },
    {
      "KL/chosen_KL_mean": -99.09426879882812,
      "KL/mean": -136.2252960205078,
      "KL/rejected_KL_mean": -173.35629272460938,
      "KL/std": 74.94339752197266,
      "epoch": 0.6969009826152683,
      "fcm_dpo/beta": 0.006514571141451597,
      "fcm_dpo/delta": -0.0906326100230217,
      "fcm_dpo/margin": 74.26202392578125,
      "fcm_dpo/q_t": 0.3882749676704407,
      "grad_norm": 12.746190071105957,
      "learning_rate": 1.2845081597488286e-07,
      "logits/chosen": 0.8655481338500977,
      "logits/rejected": 0.7720522880554199,
      "logps/chosen": -148.5721893310547,
      "logps/ref_chosen": -49.4779167175293,
      "logps/ref_rejected": -72.65262603759766,
      "logps/rejected": -246.00892639160156,
      "loss": 1.0371,
      "margin_dpo/margin_mean": 74.26202392578125,
      "margin_dpo/margin_std": 83.59939575195312,
      "step": 461
    },
    {
      "KL/chosen_KL_mean": -111.19660949707031,
      "KL/mean": -149.9906463623047,
      "KL/rejected_KL_mean": -188.78466796875,
      "KL/std": 79.32613372802734,
      "epoch": 0.6984126984126984,
      "fcm_dpo/beta": 0.006418838165700436,
      "fcm_dpo/delta": -0.10389072448015213,
      "fcm_dpo/margin": 77.58805847167969,
      "fcm_dpo/q_t": 0.3840404748916626,
      "grad_norm": 12.104856491088867,
      "learning_rate": 1.27297100994108e-07,
      "logits/chosen": 0.7282806634902954,
      "logits/rejected": 0.6704307198524475,
      "logps/chosen": -171.6917266845703,
      "logps/ref_chosen": -60.4951171875,
      "logps/ref_rejected": -74.82136535644531,
      "logps/rejected": -263.6060485839844,
      "loss": 1.0245,
      "margin_dpo/margin_mean": 77.58805847167969,
      "margin_dpo/margin_std": 87.69475555419922,
      "step": 462
    },
    {
      "KL/chosen_KL_mean": -130.89987182617188,
      "KL/mean": -155.0953369140625,
      "KL/rejected_KL_mean": -179.29078674316406,
      "KL/std": 75.73387145996094,
      "epoch": 0.6999244142101285,
      "fcm_dpo/beta": 0.00646938756108284,
      "fcm_dpo/delta": 0.08976850658655167,
      "fcm_dpo/margin": 48.39093017578125,
      "fcm_dpo/q_t": 0.4274500608444214,
      "grad_norm": 15.210125923156738,
      "learning_rate": 1.2614681827718695e-07,
      "logits/chosen": 0.7191846966743469,
      "logits/rejected": 0.7206935882568359,
      "logps/chosen": -198.58499145507812,
      "logps/ref_chosen": -67.68511962890625,
      "logps/ref_rejected": -71.32196044921875,
      "logps/rejected": -250.6127471923828,
      "loss": 1.1651,
      "margin_dpo/margin_mean": 48.39093017578125,
      "margin_dpo/margin_std": 80.35902404785156,
      "step": 463
    },
    {
      "KL/chosen_KL_mean": -121.09156799316406,
      "KL/mean": -155.86016845703125,
      "KL/rejected_KL_mean": -190.62875366210938,
      "KL/std": 81.31057739257812,
      "epoch": 0.7014361300075586,
      "fcm_dpo/beta": 0.006499043665826321,
      "fcm_dpo/delta": -0.05448608100414276,
      "fcm_dpo/margin": 69.53720092773438,
      "fcm_dpo/q_t": 0.39638495445251465,
      "grad_norm": 11.573769569396973,
      "learning_rate": 1.2500000000000005e-07,
      "logits/chosen": 0.7025905847549438,
      "logits/rejected": 0.6747890710830688,
      "logps/chosen": -180.2572021484375,
      "logps/ref_chosen": -59.16564178466797,
      "logps/ref_rejected": -69.56146240234375,
      "logps/rejected": -260.1902160644531,
      "loss": 1.0773,
      "margin_dpo/margin_mean": 69.53720092773438,
      "margin_dpo/margin_std": 94.30844116210938,
      "step": 464
    },
    {
      "KL/chosen_KL_mean": -127.25249481201172,
      "KL/mean": -157.2509765625,
      "KL/rejected_KL_mean": -187.2494659423828,
      "KL/std": 79.7821044921875,
      "epoch": 0.7029478458049887,
      "fcm_dpo/beta": 0.006456965580582619,
      "fcm_dpo/delta": 0.013073013164103031,
      "fcm_dpo/margin": 59.9969596862793,
      "fcm_dpo/q_t": 0.41156071424484253,
      "grad_norm": 13.067460060119629,
      "learning_rate": 1.238566782415197e-07,
      "logits/chosen": 0.8062239289283752,
      "logits/rejected": 0.7412049770355225,
      "logps/chosen": -185.76617431640625,
      "logps/ref_chosen": -58.513671875,
      "logps/ref_rejected": -84.31745910644531,
      "logps/rejected": -271.5669250488281,
      "loss": 1.1205,
      "margin_dpo/margin_mean": 59.99695587158203,
      "margin_dpo/margin_std": 92.42967224121094,
      "step": 465
    },
    {
      "KL/chosen_KL_mean": -137.55921936035156,
      "KL/mean": -156.19586181640625,
      "KL/rejected_KL_mean": -174.83248901367188,
      "KL/std": 84.31153869628906,
      "epoch": 0.7044595616024187,
      "fcm_dpo/beta": 0.0066132983192801476,
      "fcm_dpo/delta": 0.1576167196035385,
      "fcm_dpo/margin": 37.27326583862305,
      "fcm_dpo/q_t": 0.4440735876560211,
      "grad_norm": 17.67957878112793,
      "learning_rate": 1.2271688498291334e-07,
      "logits/chosen": 0.7369842529296875,
      "logits/rejected": 0.7448593378067017,
      "logps/chosen": -210.82504272460938,
      "logps/ref_chosen": -73.26580810546875,
      "logps/ref_rejected": -74.83621215820312,
      "logps/rejected": -249.668701171875,
      "loss": 1.2244,
      "margin_dpo/margin_mean": 37.27326965332031,
      "margin_dpo/margin_std": 81.10662841796875,
      "step": 466
    },
    {
      "KL/chosen_KL_mean": -123.79527282714844,
      "KL/mean": -149.83084106445312,
      "KL/rejected_KL_mean": -175.86642456054688,
      "KL/std": 81.82411193847656,
      "epoch": 0.7059712773998488,
      "fcm_dpo/beta": 0.006732389330863953,
      "fcm_dpo/delta": 0.05121883749961853,
      "fcm_dpo/margin": 52.0711555480957,
      "fcm_dpo/q_t": 0.42000895738601685,
      "grad_norm": 12.780661582946777,
      "learning_rate": 1.2158065210664848e-07,
      "logits/chosen": 0.7948445081710815,
      "logits/rejected": 0.6336175203323364,
      "logps/chosen": -171.374755859375,
      "logps/ref_chosen": -47.57947540283203,
      "logps/ref_rejected": -78.68522644042969,
      "logps/rejected": -254.5516357421875,
      "loss": 1.1342,
      "margin_dpo/margin_mean": 52.07115936279297,
      "margin_dpo/margin_std": 80.35679626464844,
      "step": 467
    },
    {
      "KL/chosen_KL_mean": -115.81289672851562,
      "KL/mean": -152.6653289794922,
      "KL/rejected_KL_mean": -189.51779174804688,
      "KL/std": 81.38355255126953,
      "epoch": 0.7074829931972789,
      "fcm_dpo/beta": 0.006664079613983631,
      "fcm_dpo/delta": -0.09578202664852142,
      "fcm_dpo/margin": 73.70487976074219,
      "fcm_dpo/q_t": 0.38689950108528137,
      "grad_norm": 15.063311576843262,
      "learning_rate": 1.204480113956011e-07,
      "logits/chosen": 0.6837078928947449,
      "logits/rejected": 0.6759430170059204,
      "logps/chosen": -179.74069213867188,
      "logps/ref_chosen": -63.92778778076172,
      "logps/ref_rejected": -76.51626586914062,
      "logps/rejected": -266.0340576171875,
      "loss": 1.0428,
      "margin_dpo/margin_mean": 73.70487976074219,
      "margin_dpo/margin_std": 91.98550415039062,
      "step": 468
    },
    {
      "KL/chosen_KL_mean": -117.41605377197266,
      "KL/mean": -152.40176391601562,
      "KL/rejected_KL_mean": -187.38748168945312,
      "KL/std": 80.2814712524414,
      "epoch": 0.708994708994709,
      "fcm_dpo/beta": 0.006520968861877918,
      "fcm_dpo/delta": -0.060121070593595505,
      "fcm_dpo/margin": 69.971435546875,
      "fcm_dpo/q_t": 0.3936420977115631,
      "grad_norm": 12.434749603271484,
      "learning_rate": 1.1931899453216697e-07,
      "logits/chosen": 0.8080254793167114,
      "logits/rejected": 0.7981134653091431,
      "logps/chosen": -176.4742431640625,
      "logps/ref_chosen": -59.05818176269531,
      "logps/ref_rejected": -75.67672729492188,
      "logps/rejected": -263.064208984375,
      "loss": 1.0419,
      "margin_dpo/margin_mean": 69.971435546875,
      "margin_dpo/margin_std": 75.70211029052734,
      "step": 469
    },
    {
      "KL/chosen_KL_mean": -114.1371841430664,
      "KL/mean": -146.21981811523438,
      "KL/rejected_KL_mean": -178.30245971679688,
      "KL/std": 78.92847442626953,
      "epoch": 0.7105064247921391,
      "fcm_dpo/beta": 0.006540331523865461,
      "fcm_dpo/delta": -0.020640280097723007,
      "fcm_dpo/margin": 64.16529846191406,
      "fcm_dpo/q_t": 0.4032108783721924,
      "grad_norm": 12.303239822387695,
      "learning_rate": 1.1819363309737438e-07,
      "logits/chosen": 0.773910641670227,
      "logits/rejected": 0.6982936859130859,
      "logps/chosen": -162.00462341308594,
      "logps/ref_chosen": -47.86743927001953,
      "logps/ref_rejected": -65.96859741210938,
      "logps/rejected": -244.27105712890625,
      "loss": 1.0885,
      "margin_dpo/margin_mean": 64.16529846191406,
      "margin_dpo/margin_std": 87.60211181640625,
      "step": 470
    },
    {
      "KL/chosen_KL_mean": -108.57453155517578,
      "KL/mean": -146.01129150390625,
      "KL/rejected_KL_mean": -183.44805908203125,
      "KL/std": 79.43025207519531,
      "epoch": 0.7120181405895691,
      "fcm_dpo/beta": 0.006422456353902817,
      "fcm_dpo/delta": -0.08513358235359192,
      "fcm_dpo/margin": 74.87351989746094,
      "fcm_dpo/q_t": 0.38841187953948975,
      "grad_norm": 11.945752143859863,
      "learning_rate": 1.1707195857000215e-07,
      "logits/chosen": 0.7363976240158081,
      "logits/rejected": 0.6791332364082336,
      "logps/chosen": -166.35238647460938,
      "logps/ref_chosen": -57.777854919433594,
      "logps/ref_rejected": -73.81172180175781,
      "logps/rejected": -257.259765625,
      "loss": 1.0388,
      "margin_dpo/margin_mean": 74.87351989746094,
      "margin_dpo/margin_std": 88.61801147460938,
      "step": 471
    },
    {
      "KL/chosen_KL_mean": -111.45721435546875,
      "KL/mean": -140.34298706054688,
      "KL/rejected_KL_mean": -169.22872924804688,
      "KL/std": 80.24115753173828,
      "epoch": 0.7135298563869993,
      "fcm_dpo/beta": 0.006452543195337057,
      "fcm_dpo/delta": 0.027957899495959282,
      "fcm_dpo/margin": 57.77153015136719,
      "fcm_dpo/q_t": 0.41539645195007324,
      "grad_norm": 12.780068397521973,
      "learning_rate": 1.1595400232569768e-07,
      "logits/chosen": 0.7212907075881958,
      "logits/rejected": 0.6750373840332031,
      "logps/chosen": -167.36587524414062,
      "logps/ref_chosen": -55.908668518066406,
      "logps/ref_rejected": -74.70294189453125,
      "logps/rejected": -243.93167114257812,
      "loss": 1.149,
      "margin_dpo/margin_mean": 57.771522521972656,
      "margin_dpo/margin_std": 101.25234985351562,
      "step": 472
    },
    {
      "KL/chosen_KL_mean": -113.79630279541016,
      "KL/mean": -143.42703247070312,
      "KL/rejected_KL_mean": -173.05772399902344,
      "KL/std": 83.41293334960938,
      "epoch": 0.7150415721844293,
      "fcm_dpo/beta": 0.006456817500293255,
      "fcm_dpo/delta": 0.018051929771900177,
      "fcm_dpo/margin": 59.261444091796875,
      "fcm_dpo/q_t": 0.4138728678226471,
      "grad_norm": 14.013310432434082,
      "learning_rate": 1.1483979563610069e-07,
      "logits/chosen": 0.837611198425293,
      "logits/rejected": 0.7160617113113403,
      "logps/chosen": -167.95718383789062,
      "logps/ref_chosen": -54.16088104248047,
      "logps/ref_rejected": -92.76789855957031,
      "logps/rejected": -265.82562255859375,
      "loss": 1.1456,
      "margin_dpo/margin_mean": 59.261444091796875,
      "margin_dpo/margin_std": 103.79183959960938,
      "step": 473
    },
    {
      "KL/chosen_KL_mean": -118.40287017822266,
      "KL/mean": -146.54998779296875,
      "KL/rejected_KL_mean": -174.69711303710938,
      "KL/std": 81.75066375732422,
      "epoch": 0.7165532879818595,
      "fcm_dpo/beta": 0.006502171978354454,
      "fcm_dpo/delta": 0.03521895781159401,
      "fcm_dpo/margin": 56.29423904418945,
      "fcm_dpo/q_t": 0.41749250888824463,
      "grad_norm": 16.796932220458984,
      "learning_rate": 1.1372936966796709e-07,
      "logits/chosen": 0.7968182563781738,
      "logits/rejected": 0.7175832986831665,
      "logps/chosen": -165.0885772705078,
      "logps/ref_chosen": -46.685707092285156,
      "logps/ref_rejected": -71.44731903076172,
      "logps/rejected": -246.14443969726562,
      "loss": 1.1515,
      "margin_dpo/margin_mean": 56.29423904418945,
      "margin_dpo/margin_std": 99.43895721435547,
      "step": 474
    },
    {
      "KL/chosen_KL_mean": -110.75035095214844,
      "KL/mean": -152.51861572265625,
      "KL/rejected_KL_mean": -194.28683471679688,
      "KL/std": 83.92522430419922,
      "epoch": 0.7180650037792895,
      "fcm_dpo/beta": 0.006350798532366753,
      "fcm_dpo/delta": -0.13872043788433075,
      "fcm_dpo/margin": 83.5364990234375,
      "fcm_dpo/q_t": 0.37842634320259094,
      "grad_norm": 9.905010223388672,
      "learning_rate": 1.126227554822985e-07,
      "logits/chosen": 0.7357660531997681,
      "logits/rejected": 0.684442400932312,
      "logps/chosen": -169.23765563964844,
      "logps/ref_chosen": -58.4873046875,
      "logps/ref_rejected": -87.00187683105469,
      "logps/rejected": -281.2887268066406,
      "loss": 1.0003,
      "margin_dpo/margin_mean": 83.5364990234375,
      "margin_dpo/margin_std": 89.66531372070312,
      "step": 475
    },
    {
      "KL/chosen_KL_mean": -134.1099853515625,
      "KL/mean": -162.94261169433594,
      "KL/rejected_KL_mean": -191.7752227783203,
      "KL/std": 84.37803649902344,
      "epoch": 0.7195767195767195,
      "fcm_dpo/beta": 0.006389183923602104,
      "fcm_dpo/delta": 0.03224918246269226,
      "fcm_dpo/margin": 57.665252685546875,
      "fcm_dpo/q_t": 0.4159722924232483,
      "grad_norm": 12.826769828796387,
      "learning_rate": 1.1151998403347243e-07,
      "logits/chosen": 0.6389992237091064,
      "logits/rejected": 0.6379086971282959,
      "logps/chosen": -209.49160766601562,
      "logps/ref_chosen": -75.38162231445312,
      "logps/ref_rejected": -76.99822235107422,
      "logps/rejected": -268.7734375,
      "loss": 1.1346,
      "margin_dpo/margin_mean": 57.665252685546875,
      "margin_dpo/margin_std": 91.91641235351562,
      "step": 476
    },
    {
      "KL/chosen_KL_mean": -139.5823974609375,
      "KL/mean": -166.55809020996094,
      "KL/rejected_KL_mean": -193.53378295898438,
      "KL/std": 83.39258575439453,
      "epoch": 0.7210884353741497,
      "fcm_dpo/beta": 0.0064208912663161755,
      "fcm_dpo/delta": 0.055533844977617264,
      "fcm_dpo/margin": 53.95138931274414,
      "fcm_dpo/q_t": 0.4213593006134033,
      "grad_norm": 15.228546142578125,
      "learning_rate": 1.1042108616837692e-07,
      "logits/chosen": 0.7572512626647949,
      "logits/rejected": 0.7053213119506836,
      "logps/chosen": -200.65579223632812,
      "logps/ref_chosen": -61.073387145996094,
      "logps/ref_rejected": -81.34375,
      "logps/rejected": -274.8775329589844,
      "loss": 1.1896,
      "margin_dpo/margin_mean": 53.95138931274414,
      "margin_dpo/margin_std": 110.94090270996094,
      "step": 477
    },
    {
      "KL/chosen_KL_mean": -122.52293395996094,
      "KL/mean": -146.7134246826172,
      "KL/rejected_KL_mean": -170.9039306640625,
      "KL/std": 79.75495910644531,
      "epoch": 0.7226001511715797,
      "fcm_dpo/beta": 0.0065117040649056435,
      "fcm_dpo/delta": 0.08781292289495468,
      "fcm_dpo/margin": 48.3809814453125,
      "fcm_dpo/q_t": 0.42833536863327026,
      "grad_norm": 15.279951095581055,
      "learning_rate": 1.0932609262554746e-07,
      "logits/chosen": 0.68839031457901,
      "logits/rejected": 0.698960542678833,
      "logps/chosen": -179.69024658203125,
      "logps/ref_chosen": -57.16731643676758,
      "logps/ref_rejected": -53.30917739868164,
      "logps/rejected": -224.21310424804688,
      "loss": 1.1919,
      "margin_dpo/margin_mean": 48.3809814453125,
      "margin_dpo/margin_std": 97.17034912109375,
      "step": 478
    },
    {
      "KL/chosen_KL_mean": -127.67341613769531,
      "KL/mean": -149.62149047851562,
      "KL/rejected_KL_mean": -171.56954956054688,
      "KL/std": 76.23765563964844,
      "epoch": 0.7241118669690099,
      "fcm_dpo/beta": 0.006642760708928108,
      "fcm_dpo/delta": 0.11179714649915695,
      "fcm_dpo/margin": 43.89613342285156,
      "fcm_dpo/q_t": 0.43348538875579834,
      "grad_norm": 14.566923141479492,
      "learning_rate": 1.0823503403430734e-07,
      "logits/chosen": 0.7113995552062988,
      "logits/rejected": 0.661020815372467,
      "logps/chosen": -186.58673095703125,
      "logps/ref_chosen": -58.91331481933594,
      "logps/ref_rejected": -63.7403450012207,
      "logps/rejected": -235.3098907470703,
      "loss": 1.2029,
      "margin_dpo/margin_mean": 43.89613342285156,
      "margin_dpo/margin_std": 90.19206237792969,
      "step": 479
    },
    {
      "KL/chosen_KL_mean": -127.95087432861328,
      "KL/mean": -161.46563720703125,
      "KL/rejected_KL_mean": -194.98040771484375,
      "KL/std": 79.78359985351562,
      "epoch": 0.7256235827664399,
      "fcm_dpo/beta": 0.006699123419821262,
      "fcm_dpo/delta": -0.05192215368151665,
      "fcm_dpo/margin": 67.02953338623047,
      "fcm_dpo/q_t": 0.39704233407974243,
      "grad_norm": 15.194873809814453,
      "learning_rate": 1.0714794091391072e-07,
      "logits/chosen": 0.6825853586196899,
      "logits/rejected": 0.6737313270568848,
      "logps/chosen": -190.75149536132812,
      "logps/ref_chosen": -62.80061340332031,
      "logps/ref_rejected": -67.58859252929688,
      "logps/rejected": -262.5690002441406,
      "loss": 1.084,
      "margin_dpo/margin_mean": 67.02952575683594,
      "margin_dpo/margin_std": 93.52059173583984,
      "step": 480
    },
    {
      "KL/chosen_KL_mean": -125.33125305175781,
      "KL/mean": -154.91146850585938,
      "KL/rejected_KL_mean": -184.49171447753906,
      "KL/std": 81.761474609375,
      "epoch": 0.72713529856387,
      "fcm_dpo/beta": 0.006634542718529701,
      "fcm_dpo/delta": 0.007773694582283497,
      "fcm_dpo/margin": 59.16046142578125,
      "fcm_dpo/q_t": 0.41102468967437744,
      "grad_norm": 14.069964408874512,
      "learning_rate": 1.0606484367268906e-07,
      "logits/chosen": 0.6635995507240295,
      "logits/rejected": 0.6619523167610168,
      "logps/chosen": -190.61773681640625,
      "logps/ref_chosen": -65.28649139404297,
      "logps/ref_rejected": -70.78668212890625,
      "logps/rejected": -255.2783966064453,
      "loss": 1.1262,
      "margin_dpo/margin_mean": 59.16046142578125,
      "margin_dpo/margin_std": 95.72927856445312,
      "step": 481
    },
    {
      "KL/chosen_KL_mean": -140.9495391845703,
      "KL/mean": -169.96041870117188,
      "KL/rejected_KL_mean": -198.9713134765625,
      "KL/std": 82.55261993408203,
      "epoch": 0.7286470143613001,
      "fcm_dpo/beta": 0.0066745951771736145,
      "fcm_dpo/delta": 0.01310792751610279,
      "fcm_dpo/margin": 58.02178192138672,
      "fcm_dpo/q_t": 0.41360071301460266,
      "grad_norm": 14.600235939025879,
      "learning_rate": 1.0498577260720048e-07,
      "logits/chosen": 0.6595815420150757,
      "logits/rejected": 0.5047454833984375,
      "logps/chosen": -201.855712890625,
      "logps/ref_chosen": -60.906185150146484,
      "logps/ref_rejected": -103.44656372070312,
      "logps/rejected": -302.4178771972656,
      "loss": 1.1531,
      "margin_dpo/margin_mean": 58.02178192138672,
      "margin_dpo/margin_std": 106.09500122070312,
      "step": 482
    },
    {
      "KL/chosen_KL_mean": -120.04824829101562,
      "KL/mean": -156.00111389160156,
      "KL/rejected_KL_mean": -191.9539794921875,
      "KL/std": 79.58131408691406,
      "epoch": 0.7301587301587301,
      "fcm_dpo/beta": 0.006596090272068977,
      "fcm_dpo/delta": -0.07792834937572479,
      "fcm_dpo/margin": 71.90570831298828,
      "fcm_dpo/q_t": 0.3909834921360016,
      "grad_norm": 12.167679786682129,
      "learning_rate": 1.0391075790138232e-07,
      "logits/chosen": 0.8162363767623901,
      "logits/rejected": 0.7045830488204956,
      "logps/chosen": -173.24026489257812,
      "logps/ref_chosen": -53.192012786865234,
      "logps/ref_rejected": -81.83927154541016,
      "logps/rejected": -273.7932434082031,
      "loss": 1.0491,
      "margin_dpo/margin_mean": 71.90570831298828,
      "margin_dpo/margin_std": 89.77267456054688,
      "step": 483
    },
    {
      "KL/chosen_KL_mean": -122.33198547363281,
      "KL/mean": -147.74710083007812,
      "KL/rejected_KL_mean": -173.16220092773438,
      "KL/std": 76.41363525390625,
      "epoch": 0.7316704459561603,
      "fcm_dpo/beta": 0.006655214354395866,
      "fcm_dpo/delta": 0.06308972835540771,
      "fcm_dpo/margin": 50.830196380615234,
      "fcm_dpo/q_t": 0.42113012075424194,
      "grad_norm": 17.223285675048828,
      "learning_rate": 1.0283982962570681e-07,
      "logits/chosen": 0.7866549491882324,
      "logits/rejected": 0.7504280805587769,
      "logps/chosen": -180.1014404296875,
      "logps/ref_chosen": -57.76945877075195,
      "logps/ref_rejected": -71.6829833984375,
      "logps/rejected": -244.84518432617188,
      "loss": 1.1308,
      "margin_dpo/margin_mean": 50.830196380615234,
      "margin_dpo/margin_std": 70.53138732910156,
      "step": 484
    },
    {
      "KL/chosen_KL_mean": -128.30880737304688,
      "KL/mean": -154.93826293945312,
      "KL/rejected_KL_mean": -181.56771850585938,
      "KL/std": 81.91331481933594,
      "epoch": 0.7331821617535903,
      "fcm_dpo/beta": 0.0066419942304492,
      "fcm_dpo/delta": 0.047177840024232864,
      "fcm_dpo/margin": 53.25891876220703,
      "fcm_dpo/q_t": 0.41883209347724915,
      "grad_norm": 14.847817420959473,
      "learning_rate": 1.0177301773633992e-07,
      "logits/chosen": 0.7805214524269104,
      "logits/rejected": 0.7562474012374878,
      "logps/chosen": -184.94467163085938,
      "logps/ref_chosen": -56.63584899902344,
      "logps/ref_rejected": -70.85614013671875,
      "logps/rejected": -252.42385864257812,
      "loss": 1.1349,
      "margin_dpo/margin_mean": 53.258914947509766,
      "margin_dpo/margin_std": 79.47978210449219,
      "step": 485
    },
    {
      "KL/chosen_KL_mean": -146.35873413085938,
      "KL/mean": -170.71510314941406,
      "KL/rejected_KL_mean": -195.07147216796875,
      "KL/std": 87.13774871826172,
      "epoch": 0.7346938775510204,
      "fcm_dpo/beta": 0.006780010648071766,
      "fcm_dpo/delta": 0.07210691273212433,
      "fcm_dpo/margin": 48.712730407714844,
      "fcm_dpo/q_t": 0.42639386653900146,
      "grad_norm": 13.277361869812012,
      "learning_rate": 1.007103520743035e-07,
      "logits/chosen": 0.7420072555541992,
      "logits/rejected": 0.6164480447769165,
      "logps/chosen": -202.70574951171875,
      "logps/ref_chosen": -56.347023010253906,
      "logps/ref_rejected": -85.97221374511719,
      "logps/rejected": -281.043701171875,
      "loss": 1.1858,
      "margin_dpo/margin_mean": 48.712730407714844,
      "margin_dpo/margin_std": 98.05647277832031,
      "step": 486
    },
    {
      "KL/chosen_KL_mean": -130.673095703125,
      "KL/mean": -160.9501953125,
      "KL/rejected_KL_mean": -191.227294921875,
      "KL/std": 78.9404296875,
      "epoch": 0.7362055933484505,
      "fcm_dpo/beta": 0.006787487305700779,
      "fcm_dpo/delta": -0.011505922302603722,
      "fcm_dpo/margin": 60.55419921875,
      "fcm_dpo/q_t": 0.4070599675178528,
      "grad_norm": 16.90460205078125,
      "learning_rate": 9.965186236464046e-08,
      "logits/chosen": 0.7969303727149963,
      "logits/rejected": 0.7338086366653442,
      "logps/chosen": -191.29031372070312,
      "logps/ref_chosen": -60.617218017578125,
      "logps/ref_rejected": -82.50975036621094,
      "logps/rejected": -273.7370300292969,
      "loss": 1.1044,
      "margin_dpo/margin_mean": 60.55419921875,
      "margin_dpo/margin_std": 90.86151123046875,
      "step": 487
    },
    {
      "KL/chosen_KL_mean": -124.23251342773438,
      "KL/mean": -156.62188720703125,
      "KL/rejected_KL_mean": -189.01123046875,
      "KL/std": 79.33134460449219,
      "epoch": 0.7377173091458806,
      "fcm_dpo/beta": 0.0067710187286138535,
      "fcm_dpo/delta": -0.04048357903957367,
      "fcm_dpo/margin": 64.77871704101562,
      "fcm_dpo/q_t": 0.4005877673625946,
      "grad_norm": 17.137691497802734,
      "learning_rate": 9.859757821558337e-08,
      "logits/chosen": 0.7359976768493652,
      "logits/rejected": 0.669157862663269,
      "logps/chosen": -187.34156799316406,
      "logps/ref_chosen": -63.10905075073242,
      "logps/ref_rejected": -82.49348449707031,
      "logps/rejected": -271.50469970703125,
      "loss": 1.0807,
      "margin_dpo/margin_mean": 64.77871704101562,
      "margin_dpo/margin_std": 89.09386444091797,
      "step": 488
    },
    {
      "KL/chosen_KL_mean": -142.72219848632812,
      "KL/mean": -163.11502075195312,
      "KL/rejected_KL_mean": -183.50782775878906,
      "KL/std": 85.70088958740234,
      "epoch": 0.7392290249433107,
      "fcm_dpo/beta": 0.0068509941920638084,
      "fcm_dpo/delta": 0.12417726218700409,
      "fcm_dpo/margin": 40.78562927246094,
      "fcm_dpo/q_t": 0.4372069239616394,
      "grad_norm": 12.837416648864746,
      "learning_rate": 9.754752911772615e-08,
      "logits/chosen": 0.7282401323318481,
      "logits/rejected": 0.6781407594680786,
      "logps/chosen": -207.71116638183594,
      "logps/ref_chosen": -64.98896026611328,
      "logps/ref_rejected": -84.39607238769531,
      "logps/rejected": -267.9039001464844,
      "loss": 1.2316,
      "margin_dpo/margin_mean": 40.78562927246094,
      "margin_dpo/margin_std": 97.32069396972656,
      "step": 489
    },
    {
      "KL/chosen_KL_mean": -127.41735076904297,
      "KL/mean": -152.37875366210938,
      "KL/rejected_KL_mean": -177.34014892578125,
      "KL/std": 86.41929626464844,
      "epoch": 0.7407407407407407,
      "fcm_dpo/beta": 0.00692601315677166,
      "fcm_dpo/delta": 0.055926259607076645,
      "fcm_dpo/margin": 49.92280578613281,
      "fcm_dpo/q_t": 0.4228578209877014,
      "grad_norm": 12.514721870422363,
      "learning_rate": 9.650174444319956e-08,
      "logits/chosen": 0.8083858489990234,
      "logits/rejected": 0.7824463844299316,
      "logps/chosen": -189.3260955810547,
      "logps/ref_chosen": -61.90874481201172,
      "logps/ref_rejected": -70.58566284179688,
      "logps/rejected": -247.92581176757812,
      "loss": 1.2091,
      "margin_dpo/margin_mean": 49.92280578613281,
      "margin_dpo/margin_std": 110.34571838378906,
      "step": 490
    },
    {
      "KL/chosen_KL_mean": -123.9417724609375,
      "KL/mean": -152.5548095703125,
      "KL/rejected_KL_mean": -181.16781616210938,
      "KL/std": 81.78946685791016,
      "epoch": 0.7422524565381708,
      "fcm_dpo/beta": 0.006939752493053675,
      "fcm_dpo/delta": 0.0021833814680576324,
      "fcm_dpo/margin": 57.226051330566406,
      "fcm_dpo/q_t": 0.409621000289917,
      "grad_norm": 13.716757774353027,
      "learning_rate": 9.546025344484868e-08,
      "logits/chosen": 0.6482164859771729,
      "logits/rejected": 0.5923037528991699,
      "logps/chosen": -179.41746520996094,
      "logps/ref_chosen": -55.47570037841797,
      "logps/ref_rejected": -78.70318603515625,
      "logps/rejected": -259.8710021972656,
      "loss": 1.1124,
      "margin_dpo/margin_mean": 57.226051330566406,
      "margin_dpo/margin_std": 84.64307403564453,
      "step": 491
    },
    {
      "KL/chosen_KL_mean": -147.8551025390625,
      "KL/mean": -169.573486328125,
      "KL/rejected_KL_mean": -191.29185485839844,
      "KL/std": 85.98609924316406,
      "epoch": 0.7437641723356009,
      "fcm_dpo/beta": 0.00711585208773613,
      "fcm_dpo/delta": 0.09289870411157608,
      "fcm_dpo/margin": 43.43674850463867,
      "fcm_dpo/q_t": 0.43024128675460815,
      "grad_norm": 16.301864624023438,
      "learning_rate": 9.442308525541589e-08,
      "logits/chosen": 0.7306917309761047,
      "logits/rejected": 0.6532546281814575,
      "logps/chosen": -215.1414794921875,
      "logps/ref_chosen": -67.28638458251953,
      "logps/ref_rejected": -82.78628540039062,
      "logps/rejected": -274.0781555175781,
      "loss": 1.2146,
      "margin_dpo/margin_mean": 43.43675231933594,
      "margin_dpo/margin_std": 95.23860931396484,
      "step": 492
    },
    {
      "KL/chosen_KL_mean": -120.97908020019531,
      "KL/mean": -154.72320556640625,
      "KL/rejected_KL_mean": -188.4673309326172,
      "KL/std": 84.16322326660156,
      "epoch": 0.745275888133031,
      "fcm_dpo/beta": 0.007075267843902111,
      "fcm_dpo/delta": -0.08170966804027557,
      "fcm_dpo/margin": 67.48824310302734,
      "fcm_dpo/q_t": 0.39180272817611694,
      "grad_norm": 13.524187088012695,
      "learning_rate": 9.339026888672468e-08,
      "logits/chosen": 0.6676149368286133,
      "logits/rejected": 0.5849226117134094,
      "logps/chosen": -176.90658569335938,
      "logps/ref_chosen": -55.92750549316406,
      "logps/ref_rejected": -79.12149810791016,
      "logps/rejected": -267.5888366699219,
      "loss": 1.0693,
      "margin_dpo/margin_mean": 67.48824310302734,
      "margin_dpo/margin_std": 92.7247085571289,
      "step": 493
    },
    {
      "KL/chosen_KL_mean": -126.17925262451172,
      "KL/mean": -154.39076232910156,
      "KL/rejected_KL_mean": -182.602294921875,
      "KL/std": 85.36856079101562,
      "epoch": 0.7467876039304611,
      "fcm_dpo/beta": 0.006994927302002907,
      "fcm_dpo/delta": 0.005503425374627113,
      "fcm_dpo/margin": 56.423038482666016,
      "fcm_dpo/q_t": 0.4114704728126526,
      "grad_norm": 16.0062198638916,
      "learning_rate": 9.236183322886945e-08,
      "logits/chosen": 0.6324170827865601,
      "logits/rejected": 0.5757017135620117,
      "logps/chosen": -194.13336181640625,
      "logps/ref_chosen": -67.95410919189453,
      "logps/ref_rejected": -90.50865173339844,
      "logps/rejected": -273.1109619140625,
      "loss": 1.1543,
      "margin_dpo/margin_mean": 56.423038482666016,
      "margin_dpo/margin_std": 104.7546615600586,
      "step": 494
    },
    {
      "KL/chosen_KL_mean": -123.937744140625,
      "KL/mean": -147.29507446289062,
      "KL/rejected_KL_mean": -170.6524200439453,
      "KL/std": 83.14198303222656,
      "epoch": 0.7482993197278912,
      "fcm_dpo/beta": 0.00710693746805191,
      "fcm_dpo/delta": 0.0698116272687912,
      "fcm_dpo/margin": 46.71466064453125,
      "fcm_dpo/q_t": 0.4281880259513855,
      "grad_norm": 17.973764419555664,
      "learning_rate": 9.133780704940594e-08,
      "logits/chosen": 0.8091416358947754,
      "logits/rejected": 0.7367887496948242,
      "logps/chosen": -176.56320190429688,
      "logps/ref_chosen": -52.62546157836914,
      "logps/ref_rejected": -72.06781005859375,
      "logps/rejected": -242.72021484375,
      "loss": 1.1976,
      "margin_dpo/margin_mean": 46.71466064453125,
      "margin_dpo/margin_std": 99.67835998535156,
      "step": 495
    },
    {
      "KL/chosen_KL_mean": -138.32159423828125,
      "KL/mean": -166.4814453125,
      "KL/rejected_KL_mean": -194.64129638671875,
      "KL/std": 85.53994750976562,
      "epoch": 0.7498110355253212,
      "fcm_dpo/beta": 0.007061575539410114,
      "fcm_dpo/delta": 0.0015222802758216858,
      "fcm_dpo/margin": 56.319705963134766,
      "fcm_dpo/q_t": 0.41440892219543457,
      "grad_norm": 14.250332832336426,
      "learning_rate": 9.031821899254797e-08,
      "logits/chosen": 0.724806547164917,
      "logits/rejected": 0.606552004814148,
      "logps/chosen": -195.91891479492188,
      "logps/ref_chosen": -57.597320556640625,
      "logps/ref_rejected": -94.36127471923828,
      "logps/rejected": -289.0025634765625,
      "loss": 1.1541,
      "margin_dpo/margin_mean": 56.319705963134766,
      "margin_dpo/margin_std": 104.30931091308594,
      "step": 496
    },
    {
      "KL/chosen_KL_mean": -136.56988525390625,
      "KL/mean": -170.1834716796875,
      "KL/rejected_KL_mean": -203.79702758789062,
      "KL/std": 85.25959014892578,
      "epoch": 0.7513227513227513,
      "fcm_dpo/beta": 0.007018047384917736,
      "fcm_dpo/delta": -0.07552605867385864,
      "fcm_dpo/margin": 67.22714233398438,
      "fcm_dpo/q_t": 0.3920058012008667,
      "grad_norm": 11.70369815826416,
      "learning_rate": 8.930309757836516e-08,
      "logits/chosen": 0.7534537315368652,
      "logits/rejected": 0.7208569049835205,
      "logps/chosen": -209.35983276367188,
      "logps/ref_chosen": -72.78994750976562,
      "logps/ref_rejected": -89.48483276367188,
      "logps/rejected": -293.2818603515625,
      "loss": 1.0679,
      "margin_dpo/margin_mean": 67.22714233398438,
      "margin_dpo/margin_std": 92.52470397949219,
      "step": 497
    },
    {
      "KL/chosen_KL_mean": -123.65620422363281,
      "KL/mean": -155.9206085205078,
      "KL/rejected_KL_mean": -188.1850128173828,
      "KL/std": 81.950439453125,
      "epoch": 0.7528344671201814,
      "fcm_dpo/beta": 0.006946182809770107,
      "fcm_dpo/delta": -0.05051981657743454,
      "fcm_dpo/margin": 64.52880859375,
      "fcm_dpo/q_t": 0.3971995711326599,
      "grad_norm": 16.324182510375977,
      "learning_rate": 8.829247120198563e-08,
      "logits/chosen": 0.7444478869438171,
      "logits/rejected": 0.7110167145729065,
      "logps/chosen": -192.02191162109375,
      "logps/ref_chosen": -68.36572265625,
      "logps/ref_rejected": -71.28846740722656,
      "logps/rejected": -259.4734802246094,
      "loss": 1.069,
      "margin_dpo/margin_mean": 64.52880096435547,
      "margin_dpo/margin_std": 85.4512939453125,
      "step": 498
    },
    {
      "KL/chosen_KL_mean": -121.36831665039062,
      "KL/mean": -152.87835693359375,
      "KL/rejected_KL_mean": -184.38839721679688,
      "KL/std": 83.8720474243164,
      "epoch": 0.7543461829176115,
      "fcm_dpo/beta": 0.006907904986292124,
      "fcm_dpo/delta": -0.03695734590291977,
      "fcm_dpo/margin": 63.02008819580078,
      "fcm_dpo/q_t": 0.40249934792518616,
      "grad_norm": 14.695865631103516,
      "learning_rate": 8.728636813280163e-08,
      "logits/chosen": 0.7439185380935669,
      "logits/rejected": 0.6724633574485779,
      "logps/chosen": -183.2771453857422,
      "logps/ref_chosen": -61.90882873535156,
      "logps/ref_rejected": -91.9411392211914,
      "logps/rejected": -276.32952880859375,
      "loss": 1.1306,
      "margin_dpo/margin_mean": 63.02008819580078,
      "margin_dpo/margin_std": 108.50146484375,
      "step": 499
    },
    {
      "KL/chosen_KL_mean": -128.1096954345703,
      "KL/mean": -157.46087646484375,
      "KL/rejected_KL_mean": -186.81202697753906,
      "KL/std": 79.26026916503906,
      "epoch": 0.7558578987150416,
      "fcm_dpo/beta": 0.00685786921530962,
      "fcm_dpo/delta": -0.0028558000922203064,
      "fcm_dpo/margin": 58.70232391357422,
      "fcm_dpo/q_t": 0.40761202573776245,
      "grad_norm": 16.214675903320312,
      "learning_rate": 8.628481651367875e-08,
      "logits/chosen": 0.6643021702766418,
      "logits/rejected": 0.6609218120574951,
      "logps/chosen": -198.3355255126953,
      "logps/ref_chosen": -70.225830078125,
      "logps/ref_rejected": -71.72203063964844,
      "logps/rejected": -258.5340576171875,
      "loss": 1.1489,
      "margin_dpo/margin_mean": 58.70232391357422,
      "margin_dpo/margin_std": 106.12753295898438,
      "step": 500
    },
    {
      "KL/chosen_KL_mean": -128.66282653808594,
      "KL/mean": -154.97549438476562,
      "KL/rejected_KL_mean": -181.28819274902344,
      "KL/std": 83.97608947753906,
      "epoch": 0.7573696145124716,
      "fcm_dpo/beta": 0.006935385055840015,
      "fcm_dpo/delta": 0.03603249043226242,
      "fcm_dpo/margin": 52.62535858154297,
      "fcm_dpo/q_t": 0.41577181220054626,
      "grad_norm": 13.582621574401855,
      "learning_rate": 8.528784436016878e-08,
      "logits/chosen": 0.7301015853881836,
      "logits/rejected": 0.7303687334060669,
      "logps/chosen": -193.26162719726562,
      "logps/ref_chosen": -64.59880828857422,
      "logps/ref_rejected": -70.59329223632812,
      "logps/rejected": -251.8814697265625,
      "loss": 1.1147,
      "margin_dpo/margin_mean": 52.6253547668457,
      "margin_dpo/margin_std": 71.9767074584961,
      "step": 501
    },
    {
      "KL/chosen_KL_mean": -126.94548034667969,
      "KL/mean": -154.24295043945312,
      "KL/rejected_KL_mean": -181.54042053222656,
      "KL/std": 85.87179565429688,
      "epoch": 0.7588813303099018,
      "fcm_dpo/beta": 0.0069640884175896645,
      "fcm_dpo/delta": 0.02041056379675865,
      "fcm_dpo/margin": 54.59494400024414,
      "fcm_dpo/q_t": 0.4131236672401428,
      "grad_norm": 14.177937507629395,
      "learning_rate": 8.4295479559726e-08,
      "logits/chosen": 0.7359659075737,
      "logits/rejected": 0.68549644947052,
      "logps/chosen": -192.412109375,
      "logps/ref_chosen": -65.46662902832031,
      "logps/ref_rejected": -90.22233581542969,
      "logps/rejected": -271.76275634765625,
      "loss": 1.1209,
      "margin_dpo/margin_mean": 54.59494400024414,
      "margin_dpo/margin_std": 83.03755187988281,
      "step": 502
    },
    {
      "KL/chosen_KL_mean": -117.69458770751953,
      "KL/mean": -147.38555908203125,
      "KL/rejected_KL_mean": -177.07650756835938,
      "KL/std": 81.83952331542969,
      "epoch": 0.7603930461073318,
      "fcm_dpo/beta": 0.006930957548320293,
      "fcm_dpo/delta": -0.012141318060457706,
      "fcm_dpo/margin": 59.381935119628906,
      "fcm_dpo/q_t": 0.4062880873680115,
      "grad_norm": 12.284950256347656,
      "learning_rate": 8.330774987092712e-08,
      "logits/chosen": 0.7286855578422546,
      "logits/rejected": 0.7282934188842773,
      "logps/chosen": -169.52935791015625,
      "logps/ref_chosen": -51.83476257324219,
      "logps/ref_rejected": -57.62522506713867,
      "logps/rejected": -234.70175170898438,
      "loss": 1.117,
      "margin_dpo/margin_mean": 59.381935119628906,
      "margin_dpo/margin_std": 93.7178726196289,
      "step": 503
    },
    {
      "KL/chosen_KL_mean": -119.32576751708984,
      "KL/mean": -157.84042358398438,
      "KL/rejected_KL_mean": -196.35508728027344,
      "KL/std": 80.00068664550781,
      "epoch": 0.7619047619047619,
      "fcm_dpo/beta": 0.006835811771452427,
      "fcm_dpo/delta": -0.13343694806098938,
      "fcm_dpo/margin": 77.02932739257812,
      "fcm_dpo/q_t": 0.3784865736961365,
      "grad_norm": 18.003236770629883,
      "learning_rate": 8.232468292269479e-08,
      "logits/chosen": 0.7067838311195374,
      "logits/rejected": 0.6834101676940918,
      "logps/chosen": -187.97695922851562,
      "logps/ref_chosen": -68.65119934082031,
      "logps/ref_rejected": -77.91394805908203,
      "logps/rejected": -274.26904296875,
      "loss": 1.0051,
      "margin_dpo/margin_mean": 77.0293197631836,
      "margin_dpo/margin_std": 82.89633178710938,
      "step": 504
    },
    {
      "KL/chosen_KL_mean": -125.51513671875,
      "KL/mean": -150.27340698242188,
      "KL/rejected_KL_mean": -175.0316925048828,
      "KL/std": 87.24829864501953,
      "epoch": 0.763416477702192,
      "fcm_dpo/beta": 0.006775484420359135,
      "fcm_dpo/delta": 0.0662289410829544,
      "fcm_dpo/margin": 49.51654815673828,
      "fcm_dpo/q_t": 0.4252125322818756,
      "grad_norm": 13.68161678314209,
      "learning_rate": 8.134630621352483e-08,
      "logits/chosen": 0.729232907295227,
      "logits/rejected": 0.6899206638336182,
      "logps/chosen": -185.51397705078125,
      "logps/ref_chosen": -59.99884796142578,
      "logps/ref_rejected": -76.88048553466797,
      "logps/rejected": -251.91217041015625,
      "loss": 1.1936,
      "margin_dpo/margin_mean": 49.51654815673828,
      "margin_dpo/margin_std": 100.8291244506836,
      "step": 505
    },
    {
      "KL/chosen_KL_mean": -121.28506469726562,
      "KL/mean": -148.55630493164062,
      "KL/rejected_KL_mean": -175.82757568359375,
      "KL/std": 79.10469818115234,
      "epoch": 0.764928193499622,
      "fcm_dpo/beta": 0.006854447070509195,
      "fcm_dpo/delta": 0.02708333171904087,
      "fcm_dpo/margin": 54.54250717163086,
      "fcm_dpo/q_t": 0.41415175795555115,
      "grad_norm": 15.952156066894531,
      "learning_rate": 8.037264711071698e-08,
      "logits/chosen": 0.72569739818573,
      "logits/rejected": 0.707362711429596,
      "logps/chosen": -191.35635375976562,
      "logps/ref_chosen": -70.07130432128906,
      "logps/ref_rejected": -82.03775024414062,
      "logps/rejected": -257.8653259277344,
      "loss": 1.158,
      "margin_dpo/margin_mean": 54.542503356933594,
      "margin_dpo/margin_std": 99.5510025024414,
      "step": 506
    },
    {
      "KL/chosen_KL_mean": -134.30670166015625,
      "KL/mean": -163.48135375976562,
      "KL/rejected_KL_mean": -192.656005859375,
      "KL/std": 85.48699951171875,
      "epoch": 0.7664399092970522,
      "fcm_dpo/beta": 0.006860400550067425,
      "fcm_dpo/delta": -0.000510435551404953,
      "fcm_dpo/margin": 58.34930419921875,
      "fcm_dpo/q_t": 0.4124026894569397,
      "grad_norm": 13.52925968170166,
      "learning_rate": 7.940373284960933e-08,
      "logits/chosen": 0.717583954334259,
      "logits/rejected": 0.6637930870056152,
      "logps/chosen": -206.31373596191406,
      "logps/ref_chosen": -72.00703430175781,
      "logps/ref_rejected": -93.94987487792969,
      "logps/rejected": -286.60589599609375,
      "loss": 1.1391,
      "margin_dpo/margin_mean": 58.349308013916016,
      "margin_dpo/margin_std": 101.02263641357422,
      "step": 507
    },
    {
      "KL/chosen_KL_mean": -122.50996398925781,
      "KL/mean": -154.63719177246094,
      "KL/rejected_KL_mean": -186.76441955566406,
      "KL/std": 86.9670181274414,
      "epoch": 0.7679516250944822,
      "fcm_dpo/beta": 0.006876353174448013,
      "fcm_dpo/delta": -0.04420164227485657,
      "fcm_dpo/margin": 64.25446319580078,
      "fcm_dpo/q_t": 0.4004051089286804,
      "grad_norm": 15.83190631866455,
      "learning_rate": 7.843959053281663e-08,
      "logits/chosen": 0.6986652612686157,
      "logits/rejected": 0.5631051659584045,
      "logps/chosen": -182.72988891601562,
      "logps/ref_chosen": -60.21992492675781,
      "logps/ref_rejected": -95.9200668334961,
      "logps/rejected": -282.6844787597656,
      "loss": 1.0958,
      "margin_dpo/margin_mean": 64.25446319580078,
      "margin_dpo/margin_std": 95.26860809326172,
      "step": 508
    },
    {
      "KL/chosen_KL_mean": -127.48722839355469,
      "KL/mean": -155.79217529296875,
      "KL/rejected_KL_mean": -184.09713745117188,
      "KL/std": 81.48516845703125,
      "epoch": 0.7694633408919124,
      "fcm_dpo/beta": 0.006818421185016632,
      "fcm_dpo/delta": 0.014477929100394249,
      "fcm_dpo/margin": 56.60991287231445,
      "fcm_dpo/q_t": 0.41192716360092163,
      "grad_norm": 16.497425079345703,
      "learning_rate": 7.748024712947204e-08,
      "logits/chosen": 0.6621476411819458,
      "logits/rejected": 0.639892578125,
      "logps/chosen": -193.7574005126953,
      "logps/ref_chosen": -66.27017211914062,
      "logps/ref_rejected": -71.73065185546875,
      "logps/rejected": -255.82778930664062,
      "loss": 1.1274,
      "margin_dpo/margin_mean": 56.60991287231445,
      "margin_dpo/margin_std": 90.43913269042969,
      "step": 509
    },
    {
      "KL/chosen_KL_mean": -127.44882202148438,
      "KL/mean": -160.54481506347656,
      "KL/rejected_KL_mean": -193.64077758789062,
      "KL/std": 85.18399047851562,
      "epoch": 0.7709750566893424,
      "fcm_dpo/beta": 0.006756870076060295,
      "fcm_dpo/delta": -0.05019165575504303,
      "fcm_dpo/margin": 66.19197845458984,
      "fcm_dpo/q_t": 0.40029990673065186,
      "grad_norm": 14.753068923950195,
      "learning_rate": 7.652572947447272e-08,
      "logits/chosen": 0.8324567675590515,
      "logits/rejected": 0.7259620428085327,
      "logps/chosen": -180.9936981201172,
      "logps/ref_chosen": -53.54487609863281,
      "logps/ref_rejected": -91.36648559570312,
      "logps/rejected": -285.00726318359375,
      "loss": 1.1059,
      "margin_dpo/margin_mean": 66.19198608398438,
      "margin_dpo/margin_std": 104.75790405273438,
      "step": 510
    },
    {
      "KL/chosen_KL_mean": -122.54195404052734,
      "KL/mean": -161.23318481445312,
      "KL/rejected_KL_mean": -199.9244384765625,
      "KL/std": 82.09759521484375,
      "epoch": 0.7724867724867724,
      "fcm_dpo/beta": 0.00666405912488699,
      "fcm_dpo/delta": -0.12181182205677032,
      "fcm_dpo/margin": 77.38249206542969,
      "fcm_dpo/q_t": 0.3824625611305237,
      "grad_norm": 16.800376892089844,
      "learning_rate": 7.557606426772961e-08,
      "logits/chosen": 0.7611442804336548,
      "logits/rejected": 0.6997200846672058,
      "logps/chosen": -178.38633728027344,
      "logps/ref_chosen": -55.844383239746094,
      "logps/ref_rejected": -86.49819946289062,
      "logps/rejected": -286.4226379394531,
      "loss": 1.0323,
      "margin_dpo/margin_mean": 77.38249206542969,
      "margin_dpo/margin_std": 95.76220703125,
      "step": 511
    },
    {
      "KL/chosen_KL_mean": -123.5075454711914,
      "KL/mean": -149.7601318359375,
      "KL/rejected_KL_mean": -176.01272583007812,
      "KL/std": 77.66827392578125,
      "epoch": 0.7739984882842026,
      "fcm_dpo/beta": 0.006653419695794582,
      "fcm_dpo/delta": 0.0525285080075264,
      "fcm_dpo/margin": 52.505191802978516,
      "fcm_dpo/q_t": 0.42007583379745483,
      "grad_norm": 18.692623138427734,
      "learning_rate": 7.463127807341966e-08,
      "logits/chosen": 0.6151683926582336,
      "logits/rejected": 0.6060619354248047,
      "logps/chosen": -185.16058349609375,
      "logps/ref_chosen": -61.653038024902344,
      "logps/ref_rejected": -72.83148193359375,
      "logps/rejected": -248.84420776367188,
      "loss": 1.168,
      "margin_dpo/margin_mean": 52.50518798828125,
      "margin_dpo/margin_std": 97.72168731689453,
      "step": 512
    },
    {
      "KL/chosen_KL_mean": -110.83531188964844,
      "KL/mean": -143.1207275390625,
      "KL/rejected_KL_mean": -175.4061279296875,
      "KL/std": 80.16166687011719,
      "epoch": 0.7755102040816326,
      "fcm_dpo/beta": 0.00662595871835947,
      "fcm_dpo/delta": -0.02946261130273342,
      "fcm_dpo/margin": 64.57081604003906,
      "fcm_dpo/q_t": 0.40187883377075195,
      "grad_norm": 12.440008163452148,
      "learning_rate": 7.369139731924401e-08,
      "logits/chosen": 0.842035174369812,
      "logits/rejected": 0.7855624556541443,
      "logps/chosen": -161.68788146972656,
      "logps/ref_chosen": -50.85256576538086,
      "logps/ref_rejected": -69.21754455566406,
      "logps/rejected": -244.62367248535156,
      "loss": 1.0747,
      "margin_dpo/margin_mean": 64.57081604003906,
      "margin_dpo/margin_std": 82.73863220214844,
      "step": 513
    },
    {
      "KL/chosen_KL_mean": -125.44157409667969,
      "KL/mean": -161.4381561279297,
      "KL/rejected_KL_mean": -197.43472290039062,
      "KL/std": 83.99867248535156,
      "epoch": 0.7770219198790628,
      "fcm_dpo/beta": 0.006588710471987724,
      "fcm_dpo/delta": -0.07804595679044724,
      "fcm_dpo/margin": 71.99317169189453,
      "fcm_dpo/q_t": 0.391022652387619,
      "grad_norm": 14.347710609436035,
      "learning_rate": 7.275644829568747e-08,
      "logits/chosen": 0.7848989963531494,
      "logits/rejected": 0.7469111680984497,
      "logps/chosen": -194.82650756835938,
      "logps/ref_chosen": -69.38493347167969,
      "logps/ref_rejected": -83.32447814941406,
      "logps/rejected": -280.75921630859375,
      "loss": 1.0615,
      "margin_dpo/margin_mean": 71.9931640625,
      "margin_dpo/margin_std": 95.76860046386719,
      "step": 514
    },
    {
      "KL/chosen_KL_mean": -134.65899658203125,
      "KL/mean": -163.10086059570312,
      "KL/rejected_KL_mean": -191.542724609375,
      "KL/std": 82.38731384277344,
      "epoch": 0.7785336356764928,
      "fcm_dpo/beta": 0.006560333073139191,
      "fcm_dpo/delta": 0.027873113751411438,
      "fcm_dpo/margin": 56.88371276855469,
      "fcm_dpo/q_t": 0.4149225950241089,
      "grad_norm": 17.680585861206055,
      "learning_rate": 7.182645715528435e-08,
      "logits/chosen": 0.7651482820510864,
      "logits/rejected": 0.6802647113800049,
      "logps/chosen": -188.34603881835938,
      "logps/ref_chosen": -53.687034606933594,
      "logps/ref_rejected": -83.59614562988281,
      "logps/rejected": -275.13885498046875,
      "loss": 1.1414,
      "margin_dpo/margin_mean": 56.88371276855469,
      "margin_dpo/margin_std": 96.1820297241211,
      "step": 515
    },
    {
      "KL/chosen_KL_mean": -115.13967895507812,
      "KL/mean": -142.64561462402344,
      "KL/rejected_KL_mean": -170.15155029296875,
      "KL/std": 81.29928588867188,
      "epoch": 0.780045351473923,
      "fcm_dpo/beta": 0.0066216206178069115,
      "fcm_dpo/delta": 0.0369550921022892,
      "fcm_dpo/margin": 55.011871337890625,
      "fcm_dpo/q_t": 0.41581177711486816,
      "grad_norm": 17.865516662597656,
      "learning_rate": 7.090144991188568e-08,
      "logits/chosen": 0.7032291889190674,
      "logits/rejected": 0.6626705527305603,
      "logps/chosen": -172.04139709472656,
      "logps/ref_chosen": -56.9017219543457,
      "logps/ref_rejected": -67.83477783203125,
      "logps/rejected": -237.986328125,
      "loss": 1.1504,
      "margin_dpo/margin_mean": 55.011878967285156,
      "margin_dpo/margin_std": 95.83123779296875,
      "step": 516
    },
    {
      "KL/chosen_KL_mean": -138.56973266601562,
      "KL/mean": -158.37887573242188,
      "KL/rejected_KL_mean": -178.18801879882812,
      "KL/std": 82.17344665527344,
      "epoch": 0.781557067271353,
      "fcm_dpo/beta": 0.006789041217416525,
      "fcm_dpo/delta": 0.13350741565227509,
      "fcm_dpo/margin": 39.6182861328125,
      "fcm_dpo/q_t": 0.4397648572921753,
      "grad_norm": 17.148391723632812,
      "learning_rate": 6.998145243993284e-08,
      "logits/chosen": 0.7460171580314636,
      "logits/rejected": 0.7420037984848022,
      "logps/chosen": -200.34487915039062,
      "logps/ref_chosen": -61.775142669677734,
      "logps/ref_rejected": -62.88270950317383,
      "logps/rejected": -241.07073974609375,
      "loss": 1.2232,
      "margin_dpo/margin_mean": 39.6182861328125,
      "margin_dpo/margin_std": 88.61754608154297,
      "step": 517
    },
    {
      "KL/chosen_KL_mean": -116.14431762695312,
      "KL/mean": -144.33358764648438,
      "KL/rejected_KL_mean": -172.52285766601562,
      "KL/std": 81.81043243408203,
      "epoch": 0.783068783068783,
      "fcm_dpo/beta": 0.006819076836109161,
      "fcm_dpo/delta": 0.016169242560863495,
      "fcm_dpo/margin": 56.3785400390625,
      "fcm_dpo/q_t": 0.41407328844070435,
      "grad_norm": 13.720296859741211,
      "learning_rate": 6.906649047373245e-08,
      "logits/chosen": 0.7310158014297485,
      "logits/rejected": 0.6800619959831238,
      "logps/chosen": -178.1695556640625,
      "logps/ref_chosen": -62.02523422241211,
      "logps/ref_rejected": -79.06085205078125,
      "logps/rejected": -251.58370971679688,
      "loss": 1.1275,
      "margin_dpo/margin_mean": 56.3785400390625,
      "margin_dpo/margin_std": 90.98041534423828,
      "step": 518
    },
    {
      "KL/chosen_KL_mean": -143.06752014160156,
      "KL/mean": -160.43930053710938,
      "KL/rejected_KL_mean": -177.81106567382812,
      "KL/std": 84.53593444824219,
      "epoch": 0.7845804988662132,
      "fcm_dpo/beta": 0.007009794004261494,
      "fcm_dpo/delta": 0.15968723595142365,
      "fcm_dpo/margin": 34.74354553222656,
      "fcm_dpo/q_t": 0.4461938738822937,
      "grad_norm": 23.412137985229492,
      "learning_rate": 6.815658960673781e-08,
      "logits/chosen": 0.769728422164917,
      "logits/rejected": 0.711004376411438,
      "logps/chosen": -204.67388916015625,
      "logps/ref_chosen": -61.60636901855469,
      "logps/ref_rejected": -74.50727844238281,
      "logps/rejected": -252.31834411621094,
      "loss": 1.3111,
      "margin_dpo/margin_mean": 34.74354553222656,
      "margin_dpo/margin_std": 114.71300506591797,
      "step": 519
    },
    {
      "KL/chosen_KL_mean": -130.02651977539062,
      "KL/mean": -154.45364379882812,
      "KL/rejected_KL_mean": -178.88076782226562,
      "KL/std": 83.43859100341797,
      "epoch": 0.7860922146636432,
      "fcm_dpo/beta": 0.007063240744173527,
      "fcm_dpo/delta": 0.05651885271072388,
      "fcm_dpo/margin": 48.85424041748047,
      "fcm_dpo/q_t": 0.42267489433288574,
      "grad_norm": 16.14238166809082,
      "learning_rate": 6.725177529083209e-08,
      "logits/chosen": 0.8081783056259155,
      "logits/rejected": 0.7461931109428406,
      "logps/chosen": -192.89996337890625,
      "logps/ref_chosen": -62.87343215942383,
      "logps/ref_rejected": -76.505615234375,
      "logps/rejected": -255.38636779785156,
      "loss": 1.1634,
      "margin_dpo/margin_mean": 48.85424041748047,
      "margin_dpo/margin_std": 87.38202667236328,
      "step": 520
    },
    {
      "KL/chosen_KL_mean": -128.45904541015625,
      "KL/mean": -164.06591796875,
      "KL/rejected_KL_mean": -199.67279052734375,
      "KL/std": 81.88728332519531,
      "epoch": 0.7876039304610734,
      "fcm_dpo/beta": 0.0070292167365550995,
      "fcm_dpo/delta": -0.1057443767786026,
      "fcm_dpo/margin": 71.21371459960938,
      "fcm_dpo/q_t": 0.38589245080947876,
      "grad_norm": 13.517394065856934,
      "learning_rate": 6.63520728356167e-08,
      "logits/chosen": 0.5954105257987976,
      "logits/rejected": 0.5143781304359436,
      "logps/chosen": -192.66574096679688,
      "logps/ref_chosen": -64.20668029785156,
      "logps/ref_rejected": -92.28083038330078,
      "logps/rejected": -291.95361328125,
      "loss": 1.042,
      "margin_dpo/margin_mean": 71.21371459960938,
      "margin_dpo/margin_std": 91.10281372070312,
      "step": 521
    },
    {
      "KL/chosen_KL_mean": -128.47425842285156,
      "KL/mean": -152.20489501953125,
      "KL/rejected_KL_mean": -175.93553161621094,
      "KL/std": 84.94644165039062,
      "epoch": 0.7891156462585034,
      "fcm_dpo/beta": 0.007029838860034943,
      "fcm_dpo/delta": 0.06868653744459152,
      "fcm_dpo/margin": 47.46127700805664,
      "fcm_dpo/q_t": 0.4257607161998749,
      "grad_norm": 16.1845703125,
      "learning_rate": 6.545750740770336e-08,
      "logits/chosen": 0.6975493431091309,
      "logits/rejected": 0.6909911632537842,
      "logps/chosen": -186.84397888183594,
      "logps/ref_chosen": -58.369720458984375,
      "logps/ref_rejected": -68.79248046875,
      "logps/rejected": -244.72801208496094,
      "loss": 1.217,
      "margin_dpo/margin_mean": 47.461273193359375,
      "margin_dpo/margin_std": 108.51187896728516,
      "step": 522
    },
    {
      "KL/chosen_KL_mean": -132.29612731933594,
      "KL/mean": -159.51329040527344,
      "KL/rejected_KL_mean": -186.73040771484375,
      "KL/std": 82.13084411621094,
      "epoch": 0.7906273620559335,
      "fcm_dpo/beta": 0.007071667816489935,
      "fcm_dpo/delta": 0.015576612204313278,
      "fcm_dpo/margin": 54.43429183959961,
      "fcm_dpo/q_t": 0.41066181659698486,
      "grad_norm": 19.963043212890625,
      "learning_rate": 6.456810403001012e-08,
      "logits/chosen": 0.748306393623352,
      "logits/rejected": 0.616180956363678,
      "logps/chosen": -198.00936889648438,
      "logps/ref_chosen": -65.71324157714844,
      "logps/ref_rejected": -91.98896789550781,
      "logps/rejected": -278.7193908691406,
      "loss": 1.1578,
      "margin_dpo/margin_mean": 54.434295654296875,
      "margin_dpo/margin_std": 100.6620101928711,
      "step": 523
    },
    {
      "KL/chosen_KL_mean": -112.74836730957031,
      "KL/mean": -141.45703125,
      "KL/rejected_KL_mean": -170.1656951904297,
      "KL/std": 79.19841766357422,
      "epoch": 0.7921390778533636,
      "fcm_dpo/beta": 0.007112534716725349,
      "fcm_dpo/delta": -0.009007304906845093,
      "fcm_dpo/margin": 57.417335510253906,
      "fcm_dpo/q_t": 0.4063315987586975,
      "grad_norm": 16.227237701416016,
      "learning_rate": 6.368388758106134e-08,
      "logits/chosen": 0.6562612056732178,
      "logits/rejected": 0.6327396631240845,
      "logps/chosen": -189.099609375,
      "logps/ref_chosen": -76.35124969482422,
      "logps/ref_rejected": -89.96072387695312,
      "logps/rejected": -260.1264343261719,
      "loss": 1.11,
      "margin_dpo/margin_mean": 57.417327880859375,
      "margin_dpo/margin_std": 86.96146392822266,
      "step": 524
    },
    {
      "KL/chosen_KL_mean": -126.07438659667969,
      "KL/mean": -150.48638916015625,
      "KL/rejected_KL_mean": -174.89840698242188,
      "KL/std": 80.91346740722656,
      "epoch": 0.7936507936507936,
      "fcm_dpo/beta": 0.0071428874507546425,
      "fcm_dpo/delta": 0.05306369066238403,
      "fcm_dpo/margin": 48.82402420043945,
      "fcm_dpo/q_t": 0.42222845554351807,
      "grad_norm": 18.715373992919922,
      "learning_rate": 6.280488279429185e-08,
      "logits/chosen": 0.5620905160903931,
      "logits/rejected": 0.5573893785476685,
      "logps/chosen": -201.5701904296875,
      "logps/ref_chosen": -75.49578857421875,
      "logps/ref_rejected": -84.04852294921875,
      "logps/rejected": -258.9469299316406,
      "loss": 1.1757,
      "margin_dpo/margin_mean": 48.82402038574219,
      "margin_dpo/margin_std": 95.36903381347656,
      "step": 525
    },
    {
      "KL/chosen_KL_mean": -137.1304168701172,
      "KL/mean": -157.73135375976562,
      "KL/rejected_KL_mean": -178.33230590820312,
      "KL/std": 81.97193908691406,
      "epoch": 0.7951625094482238,
      "fcm_dpo/beta": 0.007212967611849308,
      "fcm_dpo/delta": 0.1054811105132103,
      "fcm_dpo/margin": 41.20189666748047,
      "fcm_dpo/q_t": 0.43233275413513184,
      "grad_norm": 16.225221633911133,
      "learning_rate": 6.193111425735515e-08,
      "logits/chosen": 0.7405989170074463,
      "logits/rejected": 0.6656317710876465,
      "logps/chosen": -198.42282104492188,
      "logps/ref_chosen": -61.29241943359375,
      "logps/ref_rejected": -82.47763061523438,
      "logps/rejected": -260.8099365234375,
      "loss": 1.2147,
      "margin_dpo/margin_mean": 41.201900482177734,
      "margin_dpo/margin_std": 88.14698028564453,
      "step": 526
    },
    {
      "KL/chosen_KL_mean": -145.92190551757812,
      "KL/mean": -163.89610290527344,
      "KL/rejected_KL_mean": -181.87030029296875,
      "KL/std": 82.55867004394531,
      "epoch": 0.7966742252456538,
      "fcm_dpo/beta": 0.007491440512239933,
      "fcm_dpo/delta": 0.13326705992221832,
      "fcm_dpo/margin": 35.948394775390625,
      "fcm_dpo/q_t": 0.44097664952278137,
      "grad_norm": 16.771879196166992,
      "learning_rate": 6.106260641143546e-08,
      "logits/chosen": 0.8035439252853394,
      "logits/rejected": 0.7174798250198364,
      "logps/chosen": -207.39453125,
      "logps/ref_chosen": -61.472625732421875,
      "logps/ref_rejected": -90.52831268310547,
      "logps/rejected": -272.39862060546875,
      "loss": 1.2597,
      "margin_dpo/margin_mean": 35.948394775390625,
      "margin_dpo/margin_std": 96.81865692138672,
      "step": 527
    },
    {
      "KL/chosen_KL_mean": -129.02345275878906,
      "KL/mean": -149.21463012695312,
      "KL/rejected_KL_mean": -169.40582275390625,
      "KL/std": 82.30186462402344,
      "epoch": 0.7981859410430839,
      "fcm_dpo/beta": 0.007598328869789839,
      "fcm_dpo/delta": 0.0962243527173996,
      "fcm_dpo/margin": 40.38237762451172,
      "fcm_dpo/q_t": 0.43211063742637634,
      "grad_norm": 18.41050148010254,
      "learning_rate": 6.019938355056422e-08,
      "logits/chosen": 0.642044186592102,
      "logits/rejected": 0.5571998357772827,
      "logps/chosen": -187.81546020507812,
      "logps/ref_chosen": -58.792015075683594,
      "logps/ref_rejected": -71.82516479492188,
      "logps/rejected": -241.23098754882812,
      "loss": 1.248,
      "margin_dpo/margin_mean": 40.38237762451172,
      "margin_dpo/margin_std": 103.26986694335938,
      "step": 528
    },
    {
      "KL/chosen_KL_mean": -119.10054779052734,
      "KL/mean": -162.97311401367188,
      "KL/rejected_KL_mean": -206.845703125,
      "KL/std": 82.37154388427734,
      "epoch": 0.799697656840514,
      "fcm_dpo/beta": 0.007344953715801239,
      "fcm_dpo/delta": -0.26204729080200195,
      "fcm_dpo/margin": 87.74514770507812,
      "fcm_dpo/q_t": 0.3532130718231201,
      "grad_norm": 16.302799224853516,
      "learning_rate": 5.934146982094049e-08,
      "logits/chosen": 0.656600832939148,
      "logits/rejected": 0.6004124283790588,
      "logps/chosen": -174.1715087890625,
      "logps/ref_chosen": -55.070960998535156,
      "logps/ref_rejected": -75.44007873535156,
      "logps/rejected": -282.2857666015625,
      "loss": 0.9417,
      "margin_dpo/margin_mean": 87.74514770507812,
      "margin_dpo/margin_std": 89.89462280273438,
      "step": 529
    },
    {
      "KL/chosen_KL_mean": -123.7182846069336,
      "KL/mean": -149.55343627929688,
      "KL/rejected_KL_mean": -175.3886260986328,
      "KL/std": 82.55381774902344,
      "epoch": 0.8012093726379441,
      "fcm_dpo/beta": 0.007294186390936375,
      "fcm_dpo/delta": 0.023981863632798195,
      "fcm_dpo/margin": 51.670318603515625,
      "fcm_dpo/q_t": 0.41621410846710205,
      "grad_norm": 19.178972244262695,
      "learning_rate": 5.848888922025552e-08,
      "logits/chosen": 0.7491636276245117,
      "logits/rejected": 0.7024554014205933,
      "logps/chosen": -180.46209716796875,
      "logps/ref_chosen": -56.743812561035156,
      "logps/ref_rejected": -76.6692123413086,
      "logps/rejected": -252.05783081054688,
      "loss": 1.1437,
      "margin_dpo/margin_mean": 51.670318603515625,
      "margin_dpo/margin_std": 88.48190307617188,
      "step": 530
    },
    {
      "KL/chosen_KL_mean": -124.6670913696289,
      "KL/mean": -152.11793518066406,
      "KL/rejected_KL_mean": -179.56875610351562,
      "KL/std": 83.25489044189453,
      "epoch": 0.8027210884353742,
      "fcm_dpo/beta": 0.007327442988753319,
      "fcm_dpo/delta": -0.0027846992015838623,
      "fcm_dpo/margin": 54.901676177978516,
      "fcm_dpo/q_t": 0.4097214937210083,
      "grad_norm": 15.689215660095215,
      "learning_rate": 5.7641665597021435e-08,
      "logits/chosen": 0.7318727970123291,
      "logits/rejected": 0.6541762351989746,
      "logps/chosen": -175.78353881835938,
      "logps/ref_chosen": -51.116455078125,
      "logps/ref_rejected": -79.52884674072266,
      "logps/rejected": -259.09759521484375,
      "loss": 1.1231,
      "margin_dpo/margin_mean": 54.90167236328125,
      "margin_dpo/margin_std": 88.21559143066406,
      "step": 531
    },
    {
      "KL/chosen_KL_mean": -144.3057403564453,
      "KL/mean": -174.24826049804688,
      "KL/rejected_KL_mean": -204.19082641601562,
      "KL/std": 81.07133483886719,
      "epoch": 0.8042328042328042,
      "fcm_dpo/beta": 0.007251083850860596,
      "fcm_dpo/delta": -0.03580852597951889,
      "fcm_dpo/margin": 59.88507080078125,
      "fcm_dpo/q_t": 0.4021422564983368,
      "grad_norm": 16.10748863220215,
      "learning_rate": 5.679982264990424e-08,
      "logits/chosen": 0.6791332960128784,
      "logits/rejected": 0.6243463754653931,
      "logps/chosen": -202.585693359375,
      "logps/ref_chosen": -58.279945373535156,
      "logps/ref_rejected": -78.05426788330078,
      "logps/rejected": -282.2450866699219,
      "loss": 1.1035,
      "margin_dpo/margin_mean": 59.88507080078125,
      "margin_dpo/margin_std": 92.61534118652344,
      "step": 532
    },
    {
      "KL/chosen_KL_mean": -101.09480285644531,
      "KL/mean": -131.0142364501953,
      "KL/rejected_KL_mean": -160.9336700439453,
      "KL/std": 80.21917724609375,
      "epoch": 0.8057445200302343,
      "fcm_dpo/beta": 0.007215174846351147,
      "fcm_dpo/delta": -0.03317198529839516,
      "fcm_dpo/margin": 59.83887481689453,
      "fcm_dpo/q_t": 0.4015866816043854,
      "grad_norm": 15.353870391845703,
      "learning_rate": 5.596338392706076e-08,
      "logits/chosen": 0.8840050101280212,
      "logits/rejected": 0.8084684610366821,
      "logps/chosen": -157.5128173828125,
      "logps/ref_chosen": -56.41801071166992,
      "logps/ref_rejected": -73.89324951171875,
      "logps/rejected": -234.82691955566406,
      "loss": 1.0938,
      "margin_dpo/margin_mean": 59.83887481689453,
      "margin_dpo/margin_std": 86.97126007080078,
      "step": 533
    },
    {
      "KL/chosen_KL_mean": -125.39263153076172,
      "KL/mean": -154.99639892578125,
      "KL/rejected_KL_mean": -184.60012817382812,
      "KL/std": 83.15888977050781,
      "epoch": 0.8072562358276644,
      "fcm_dpo/beta": 0.007139452733099461,
      "fcm_dpo/delta": -0.024127241224050522,
      "fcm_dpo/margin": 59.20750427246094,
      "fcm_dpo/q_t": 0.4062623977661133,
      "grad_norm": 13.482405662536621,
      "learning_rate": 5.513237282548033e-08,
      "logits/chosen": 0.7019526958465576,
      "logits/rejected": 0.6646519899368286,
      "logps/chosen": -186.14132690429688,
      "logps/ref_chosen": -60.748687744140625,
      "logps/ref_rejected": -73.8623046875,
      "logps/rejected": -258.4624328613281,
      "loss": 1.1279,
      "margin_dpo/margin_mean": 59.20750427246094,
      "margin_dpo/margin_std": 100.45853424072266,
      "step": 534
    },
    {
      "KL/chosen_KL_mean": -137.643798828125,
      "KL/mean": -162.00775146484375,
      "KL/rejected_KL_mean": -186.37168884277344,
      "KL/std": 86.37908935546875,
      "epoch": 0.8087679516250945,
      "fcm_dpo/beta": 0.007220745086669922,
      "fcm_dpo/delta": 0.04981802776455879,
      "fcm_dpo/margin": 48.727882385253906,
      "fcm_dpo/q_t": 0.4213569462299347,
      "grad_norm": 17.222349166870117,
      "learning_rate": 5.430681259032957e-08,
      "logits/chosen": 0.6391861438751221,
      "logits/rejected": 0.5727677345275879,
      "logps/chosen": -199.28121948242188,
      "logps/ref_chosen": -61.637413024902344,
      "logps/ref_rejected": -80.93138885498047,
      "logps/rejected": -267.3031005859375,
      "loss": 1.1729,
      "margin_dpo/margin_mean": 48.727882385253906,
      "margin_dpo/margin_std": 93.55561828613281,
      "step": 535
    },
    {
      "KL/chosen_KL_mean": -118.49983978271484,
      "KL/mean": -157.49525451660156,
      "KL/rejected_KL_mean": -196.49066162109375,
      "KL/std": 87.45317077636719,
      "epoch": 0.8102796674225246,
      "fcm_dpo/beta": 0.007055111229419708,
      "fcm_dpo/delta": -0.15915080904960632,
      "fcm_dpo/margin": 77.99081420898438,
      "fcm_dpo/q_t": 0.37544578313827515,
      "grad_norm": 11.248884201049805,
      "learning_rate": 5.3486726314303175e-08,
      "logits/chosen": 0.7687609195709229,
      "logits/rejected": 0.6766946315765381,
      "logps/chosen": -170.38882446289062,
      "logps/ref_chosen": -51.88897705078125,
      "logps/ref_rejected": -73.34864044189453,
      "logps/rejected": -269.83929443359375,
      "loss": 0.9901,
      "margin_dpo/margin_mean": 77.99082946777344,
      "margin_dpo/margin_std": 84.41676330566406,
      "step": 536
    },
    {
      "KL/chosen_KL_mean": -135.27239990234375,
      "KL/mean": -164.75808715820312,
      "KL/rejected_KL_mean": -194.24378967285156,
      "KL/std": 88.089599609375,
      "epoch": 0.8117913832199547,
      "fcm_dpo/beta": 0.0069625237956643105,
      "fcm_dpo/delta": -0.011460717767477036,
      "fcm_dpo/margin": 58.971397399902344,
      "fcm_dpo/q_t": 0.40821582078933716,
      "grad_norm": 14.422006607055664,
      "learning_rate": 5.267213693697695e-08,
      "logits/chosen": 0.8400592803955078,
      "logits/rejected": 0.7391371726989746,
      "logps/chosen": -189.52101135253906,
      "logps/ref_chosen": -54.248619079589844,
      "logps/ref_rejected": -94.94343566894531,
      "logps/rejected": -289.1872253417969,
      "loss": 1.1192,
      "margin_dpo/margin_mean": 58.971397399902344,
      "margin_dpo/margin_std": 94.33164978027344,
      "step": 537
    },
    {
      "KL/chosen_KL_mean": -130.045654296875,
      "KL/mean": -163.341064453125,
      "KL/rejected_KL_mean": -196.636474609375,
      "KL/std": 84.741943359375,
      "epoch": 0.8133030990173847,
      "fcm_dpo/beta": 0.006943310610949993,
      "fcm_dpo/delta": -0.06539023667573929,
      "fcm_dpo/margin": 66.5908203125,
      "fcm_dpo/q_t": 0.3946533203125,
      "grad_norm": 13.937400817871094,
      "learning_rate": 5.1863067244167144e-08,
      "logits/chosen": 0.7151967883110046,
      "logits/rejected": 0.6850166320800781,
      "logps/chosen": -200.13919067382812,
      "logps/ref_chosen": -70.09353637695312,
      "logps/ref_rejected": -79.49833679199219,
      "logps/rejected": -276.13482666015625,
      "loss": 1.0624,
      "margin_dpo/margin_mean": 66.5908203125,
      "margin_dpo/margin_std": 87.32984924316406,
      "step": 538
    },
    {
      "KL/chosen_KL_mean": -140.43576049804688,
      "KL/mean": -165.48690795898438,
      "KL/rejected_KL_mean": -190.5380401611328,
      "KL/std": 83.2811508178711,
      "epoch": 0.8148148148148148,
      "fcm_dpo/beta": 0.006932198069989681,
      "fcm_dpo/delta": 0.05455028638243675,
      "fcm_dpo/margin": 50.1022834777832,
      "fcm_dpo/q_t": 0.4223223328590393,
      "grad_norm": 14.812973022460938,
      "learning_rate": 5.105953986729195e-08,
      "logits/chosen": 0.6730055809020996,
      "logits/rejected": 0.588666558265686,
      "logps/chosen": -202.36746215820312,
      "logps/ref_chosen": -61.93169403076172,
      "logps/ref_rejected": -84.08946228027344,
      "logps/rejected": -274.62750244140625,
      "loss": 1.1575,
      "margin_dpo/margin_mean": 50.10228729248047,
      "margin_dpo/margin_std": 88.67442321777344,
      "step": 539
    },
    {
      "KL/chosen_KL_mean": -127.05425262451172,
      "KL/mean": -163.42059326171875,
      "KL/rejected_KL_mean": -199.7869110107422,
      "KL/std": 91.1855697631836,
      "epoch": 0.8163265306122449,
      "fcm_dpo/beta": 0.006901265121996403,
      "fcm_dpo/delta": -0.10737490653991699,
      "fcm_dpo/margin": 72.73268127441406,
      "fcm_dpo/q_t": 0.38571393489837646,
      "grad_norm": 13.47286605834961,
      "learning_rate": 5.026157728273966e-08,
      "logits/chosen": 0.7549730539321899,
      "logits/rejected": 0.6512420177459717,
      "logps/chosen": -189.7584991455078,
      "logps/ref_chosen": -62.704254150390625,
      "logps/ref_rejected": -95.63597106933594,
      "logps/rejected": -295.4228820800781,
      "loss": 1.033,
      "margin_dpo/margin_mean": 72.73268127441406,
      "margin_dpo/margin_std": 87.08503723144531,
      "step": 540
    },
    {
      "KL/chosen_KL_mean": -124.87004852294922,
      "KL/mean": -158.8612060546875,
      "KL/rejected_KL_mean": -192.8523712158203,
      "KL/std": 85.35519409179688,
      "epoch": 0.817838246409675,
      "fcm_dpo/beta": 0.006723807659000158,
      "fcm_dpo/delta": -0.060890041291713715,
      "fcm_dpo/margin": 67.9823226928711,
      "fcm_dpo/q_t": 0.3946753144264221,
      "grad_norm": 13.895951271057129,
      "learning_rate": 4.9469201811239035e-08,
      "logits/chosen": 0.7421223521232605,
      "logits/rejected": 0.7663363218307495,
      "logps/chosen": -187.35089111328125,
      "logps/ref_chosen": -62.48084259033203,
      "logps/ref_rejected": -57.55541229248047,
      "logps/rejected": -250.40777587890625,
      "loss": 1.0618,
      "margin_dpo/margin_mean": 67.9823226928711,
      "margin_dpo/margin_std": 85.90786743164062,
      "step": 541
    },
    {
      "KL/chosen_KL_mean": -108.08326721191406,
      "KL/mean": -145.9094696044922,
      "KL/rejected_KL_mean": -183.73565673828125,
      "KL/std": 83.32110595703125,
      "epoch": 0.8193499622071051,
      "fcm_dpo/beta": 0.006626952439546585,
      "fcm_dpo/delta": -0.10687525570392609,
      "fcm_dpo/margin": 75.65240478515625,
      "fcm_dpo/q_t": 0.3857290744781494,
      "grad_norm": 13.757902145385742,
      "learning_rate": 4.868243561723534e-08,
      "logits/chosen": 0.7640467882156372,
      "logits/rejected": 0.7123850584030151,
      "logps/chosen": -157.53814697265625,
      "logps/ref_chosen": -49.454891204833984,
      "logps/ref_rejected": -65.33275604248047,
      "logps/rejected": -249.06842041015625,
      "loss": 1.0517,
      "margin_dpo/margin_mean": 75.65240478515625,
      "margin_dpo/margin_std": 101.577392578125,
      "step": 542
    },
    {
      "KL/chosen_KL_mean": -115.50267791748047,
      "KL/mean": -150.74710083007812,
      "KL/rejected_KL_mean": -185.9915313720703,
      "KL/std": 81.07098388671875,
      "epoch": 0.8208616780045351,
      "fcm_dpo/beta": 0.006557712331414223,
      "fcm_dpo/delta": -0.06527210772037506,
      "fcm_dpo/margin": 70.48886108398438,
      "fcm_dpo/q_t": 0.39285480976104736,
      "grad_norm": 11.07712459564209,
      "learning_rate": 4.790130070827028e-08,
      "logits/chosen": 0.7415127754211426,
      "logits/rejected": 0.6479545831680298,
      "logps/chosen": -166.60354614257812,
      "logps/ref_chosen": -51.100860595703125,
      "logps/ref_rejected": -76.06130981445312,
      "logps/rejected": -262.0528564453125,
      "loss": 1.0594,
      "margin_dpo/margin_mean": 70.48886108398438,
      "margin_dpo/margin_std": 89.37535095214844,
      "step": 543
    },
    {
      "KL/chosen_KL_mean": -127.9332504272461,
      "KL/mean": -165.9463653564453,
      "KL/rejected_KL_mean": -203.95947265625,
      "KL/std": 87.00491333007812,
      "epoch": 0.8223733938019653,
      "fcm_dpo/beta": 0.006429500877857208,
      "fcm_dpo/delta": -0.09331250935792923,
      "fcm_dpo/margin": 76.02620697021484,
      "fcm_dpo/q_t": 0.38860976696014404,
      "grad_norm": 14.734076499938965,
      "learning_rate": 4.7125818934366454e-08,
      "logits/chosen": 0.7422240972518921,
      "logits/rejected": 0.6599665880203247,
      "logps/chosen": -188.21047973632812,
      "logps/ref_chosen": -60.2772331237793,
      "logps/ref_rejected": -88.40553283691406,
      "logps/rejected": -292.364990234375,
      "loss": 1.051,
      "margin_dpo/margin_mean": 76.02620697021484,
      "margin_dpo/margin_std": 99.28297424316406,
      "step": 544
    },
    {
      "KL/chosen_KL_mean": -136.6294708251953,
      "KL/mean": -160.23928833007812,
      "KL/rejected_KL_mean": -183.84912109375,
      "KL/std": 84.00276184082031,
      "epoch": 0.8238851095993953,
      "fcm_dpo/beta": 0.006486566737294197,
      "fcm_dpo/delta": 0.09666649997234344,
      "fcm_dpo/margin": 47.219642639160156,
      "fcm_dpo/q_t": 0.4306223690509796,
      "grad_norm": 13.613739967346191,
      "learning_rate": 4.635601198741607e-08,
      "logits/chosen": 0.6614166498184204,
      "logits/rejected": 0.6035963296890259,
      "logps/chosen": -198.24472045898438,
      "logps/ref_chosen": -61.61524963378906,
      "logps/ref_rejected": -78.71266174316406,
      "logps/rejected": -262.561767578125,
      "loss": 1.1946,
      "margin_dpo/margin_mean": 47.219642639160156,
      "margin_dpo/margin_std": 95.0772705078125,
      "step": 545
    },
    {
      "KL/chosen_KL_mean": -129.49948120117188,
      "KL/mean": -156.82077026367188,
      "KL/rejected_KL_mean": -184.14207458496094,
      "KL/std": 81.37245178222656,
      "epoch": 0.8253968253968254,
      "fcm_dpo/beta": 0.006546933203935623,
      "fcm_dpo/delta": 0.04382575675845146,
      "fcm_dpo/margin": 54.64259338378906,
      "fcm_dpo/q_t": 0.41790369153022766,
      "grad_norm": 15.220459938049316,
      "learning_rate": 4.559190140057428e-08,
      "logits/chosen": 0.813021183013916,
      "logits/rejected": 0.8038173913955688,
      "logps/chosen": -188.812744140625,
      "logps/ref_chosen": -59.313262939453125,
      "logps/ref_rejected": -64.73631286621094,
      "logps/rejected": -248.87838745117188,
      "loss": 1.1598,
      "margin_dpo/margin_mean": 54.6425895690918,
      "margin_dpo/margin_std": 98.65786743164062,
      "step": 546
    },
    {
      "KL/chosen_KL_mean": -115.29865264892578,
      "KL/mean": -150.95513916015625,
      "KL/rejected_KL_mean": -186.6116180419922,
      "KL/std": 84.62432098388672,
      "epoch": 0.8269085411942555,
      "fcm_dpo/beta": 0.006497037131339312,
      "fcm_dpo/delta": -0.06663330644369125,
      "fcm_dpo/margin": 71.31297302246094,
      "fcm_dpo/q_t": 0.3933113217353821,
      "grad_norm": 13.908299446105957,
      "learning_rate": 4.483350854765672e-08,
      "logits/chosen": 0.6590763926506042,
      "logits/rejected": 0.5905526876449585,
      "logps/chosen": -170.275390625,
      "logps/ref_chosen": -54.97674560546875,
      "logps/ref_rejected": -75.35922241210938,
      "logps/rejected": -261.9708251953125,
      "loss": 1.0669,
      "margin_dpo/margin_mean": 71.31297302246094,
      "margin_dpo/margin_std": 96.31216430664062,
      "step": 547
    },
    {
      "KL/chosen_KL_mean": -132.67001342773438,
      "KL/mean": -156.9127197265625,
      "KL/rejected_KL_mean": -181.15542602539062,
      "KL/std": 84.14120483398438,
      "epoch": 0.8284202569916855,
      "fcm_dpo/beta": 0.0065902965143322945,
      "fcm_dpo/delta": 0.0826481357216835,
      "fcm_dpo/margin": 48.48541259765625,
      "fcm_dpo/q_t": 0.42783021926879883,
      "grad_norm": 16.948848724365234,
      "learning_rate": 4.4080854642541826e-08,
      "logits/chosen": 0.6547988653182983,
      "logits/rejected": 0.5863006711006165,
      "logps/chosen": -195.88067626953125,
      "logps/ref_chosen": -63.21067428588867,
      "logps/ref_rejected": -81.23347473144531,
      "logps/rejected": -262.388916015625,
      "loss": 1.1768,
      "margin_dpo/margin_mean": 48.48541259765625,
      "margin_dpo/margin_std": 90.33981323242188,
      "step": 548
    },
    {
      "KL/chosen_KL_mean": -128.98789978027344,
      "KL/mean": -156.6746826171875,
      "KL/rejected_KL_mean": -184.3614501953125,
      "KL/std": 84.68203735351562,
      "epoch": 0.8299319727891157,
      "fcm_dpo/beta": 0.006622135639190674,
      "fcm_dpo/delta": 0.03458665311336517,
      "fcm_dpo/margin": 55.373558044433594,
      "fcm_dpo/q_t": 0.41641145944595337,
      "grad_norm": 15.642621994018555,
      "learning_rate": 4.333396073857723e-08,
      "logits/chosen": 0.8145290017127991,
      "logits/rejected": 0.7393472194671631,
      "logps/chosen": -193.26141357421875,
      "logps/ref_chosen": -64.27351379394531,
      "logps/ref_rejected": -92.31663513183594,
      "logps/rejected": -276.6781005859375,
      "loss": 1.1508,
      "margin_dpo/margin_mean": 55.37355422973633,
      "margin_dpo/margin_std": 97.06788635253906,
      "step": 549
    },
    {
      "KL/chosen_KL_mean": -141.47198486328125,
      "KL/mean": -159.78488159179688,
      "KL/rejected_KL_mean": -178.09779357910156,
      "KL/std": 81.69731140136719,
      "epoch": 0.8314436885865457,
      "fcm_dpo/beta": 0.006767358630895615,
      "fcm_dpo/delta": 0.15628939867019653,
      "fcm_dpo/margin": 36.62580490112305,
      "fcm_dpo/q_t": 0.4437524974346161,
      "grad_norm": 17.119239807128906,
      "learning_rate": 4.259284772799099e-08,
      "logits/chosen": 0.7669326066970825,
      "logits/rejected": 0.7338714599609375,
      "logps/chosen": -197.70242309570312,
      "logps/ref_chosen": -56.230438232421875,
      "logps/ref_rejected": -62.59788513183594,
      "logps/rejected": -240.6956787109375,
      "loss": 1.2428,
      "margin_dpo/margin_mean": 36.62580108642578,
      "margin_dpo/margin_std": 88.4005126953125,
      "step": 550
    },
    {
      "KL/chosen_KL_mean": -139.54542541503906,
      "KL/mean": -162.89857482910156,
      "KL/rejected_KL_mean": -186.251708984375,
      "KL/std": 85.5566635131836,
      "epoch": 0.8329554043839759,
      "fcm_dpo/beta": 0.0069540319964289665,
      "fcm_dpo/delta": 0.07730760425329208,
      "fcm_dpo/margin": 46.70626449584961,
      "fcm_dpo/q_t": 0.4272865653038025,
      "grad_norm": 14.740558624267578,
      "learning_rate": 4.1857536341307176e-08,
      "logits/chosen": 0.728614330291748,
      "logits/rejected": 0.6926642656326294,
      "logps/chosen": -207.29263305664062,
      "logps/ref_chosen": -67.74720764160156,
      "logps/ref_rejected": -87.04285430908203,
      "logps/rejected": -273.2945556640625,
      "loss": 1.1645,
      "margin_dpo/margin_mean": 46.706260681152344,
      "margin_dpo/margin_std": 83.12886810302734,
      "step": 551
    },
    {
      "KL/chosen_KL_mean": -129.97451782226562,
      "KL/mean": -158.94329833984375,
      "KL/rejected_KL_mean": -187.91207885742188,
      "KL/std": 85.75680541992188,
      "epoch": 0.8344671201814059,
      "fcm_dpo/beta": 0.006984234321862459,
      "fcm_dpo/delta": -0.005182873457670212,
      "fcm_dpo/margin": 57.937564849853516,
      "fcm_dpo/q_t": 0.4047027826309204,
      "grad_norm": 14.647710800170898,
      "learning_rate": 4.112804714676593e-08,
      "logits/chosen": 0.7080692052841187,
      "logits/rejected": 0.6515108346939087,
      "logps/chosen": -192.90078735351562,
      "logps/ref_chosen": -62.92625427246094,
      "logps/ref_rejected": -82.98365783691406,
      "logps/rejected": -270.8957214355469,
      "loss": 1.1029,
      "margin_dpo/margin_mean": 57.937564849853516,
      "margin_dpo/margin_std": 82.95120239257812,
      "step": 552
    },
    {
      "KL/chosen_KL_mean": -140.59323120117188,
      "KL/mean": -167.88583374023438,
      "KL/rejected_KL_mean": -195.17840576171875,
      "KL/std": 81.97215270996094,
      "epoch": 0.8359788359788359,
      "fcm_dpo/beta": 0.0069772712886333466,
      "fcm_dpo/delta": 0.019887106493115425,
      "fcm_dpo/margin": 54.585182189941406,
      "fcm_dpo/q_t": 0.415116548538208,
      "grad_norm": 17.540098190307617,
      "learning_rate": 4.0404400549748144e-08,
      "logits/chosen": 0.6930921673774719,
      "logits/rejected": 0.5812926292419434,
      "logps/chosen": -196.63172912597656,
      "logps/ref_chosen": -56.038490295410156,
      "logps/ref_rejected": -84.48454284667969,
      "logps/rejected": -279.6629638671875,
      "loss": 1.1521,
      "margin_dpo/margin_mean": 54.585182189941406,
      "margin_dpo/margin_std": 99.14485931396484,
      "step": 553
    },
    {
      "KL/chosen_KL_mean": -126.84615325927734,
      "KL/mean": -159.65020751953125,
      "KL/rejected_KL_mean": -192.45425415039062,
      "KL/std": 82.10330200195312,
      "epoch": 0.8374905517762661,
      "fcm_dpo/beta": 0.006926331669092178,
      "fcm_dpo/delta": -0.05698609724640846,
      "fcm_dpo/margin": 65.60808563232422,
      "fcm_dpo/q_t": 0.3963993191719055,
      "grad_norm": 14.771784782409668,
      "learning_rate": 3.968661679220467e-08,
      "logits/chosen": 0.7141705751419067,
      "logits/rejected": 0.6928715705871582,
      "logps/chosen": -191.37673950195312,
      "logps/ref_chosen": -64.53059387207031,
      "logps/ref_rejected": -71.2155990600586,
      "logps/rejected": -263.66986083984375,
      "loss": 1.0791,
      "margin_dpo/margin_mean": 65.60808563232422,
      "margin_dpo/margin_std": 92.8201904296875,
      "step": 554
    },
    {
      "KL/chosen_KL_mean": -138.47244262695312,
      "KL/mean": -166.33506774902344,
      "KL/rejected_KL_mean": -194.19769287109375,
      "KL/std": 83.23756408691406,
      "epoch": 0.8390022675736961,
      "fcm_dpo/beta": 0.006969560869038105,
      "fcm_dpo/delta": 0.010640434920787811,
      "fcm_dpo/margin": 55.725242614746094,
      "fcm_dpo/q_t": 0.41070425510406494,
      "grad_norm": 15.12730598449707,
      "learning_rate": 3.89747159520904e-08,
      "logits/chosen": 0.7322822213172913,
      "logits/rejected": 0.703670859336853,
      "logps/chosen": -205.12435913085938,
      "logps/ref_chosen": -66.65191650390625,
      "logps/ref_rejected": -68.6667251586914,
      "logps/rejected": -262.8644104003906,
      "loss": 1.1592,
      "margin_dpo/margin_mean": 55.725242614746094,
      "margin_dpo/margin_std": 100.45161437988281,
      "step": 555
    },
    {
      "KL/chosen_KL_mean": -140.06524658203125,
      "KL/mean": -164.30087280273438,
      "KL/rejected_KL_mean": -188.53648376464844,
      "KL/std": 87.56776428222656,
      "epoch": 0.8405139833711263,
      "fcm_dpo/beta": 0.006966423708945513,
      "fcm_dpo/delta": 0.06454331427812576,
      "fcm_dpo/margin": 48.47126007080078,
      "fcm_dpo/q_t": 0.42569175362586975,
      "grad_norm": 13.4514799118042,
      "learning_rate": 3.826871794280192e-08,
      "logits/chosen": 0.7811669707298279,
      "logits/rejected": 0.7258546948432922,
      "logps/chosen": -192.89761352539062,
      "logps/ref_chosen": -52.832366943359375,
      "logps/ref_rejected": -64.49044036865234,
      "logps/rejected": -253.02691650390625,
      "loss": 1.1975,
      "margin_dpo/margin_mean": 48.47126007080078,
      "margin_dpo/margin_std": 102.2444076538086,
      "step": 556
    },
    {
      "KL/chosen_KL_mean": -137.24900817871094,
      "KL/mean": -174.63917541503906,
      "KL/rejected_KL_mean": -212.02932739257812,
      "KL/std": 84.4654541015625,
      "epoch": 0.8420256991685563,
      "fcm_dpo/beta": 0.00685353297740221,
      "fcm_dpo/delta": -0.11947400867938995,
      "fcm_dpo/margin": 74.78031158447266,
      "fcm_dpo/q_t": 0.3829210102558136,
      "grad_norm": 12.452169418334961,
      "learning_rate": 3.756864251262143e-08,
      "logits/chosen": 0.8675057888031006,
      "logits/rejected": 0.7827339172363281,
      "logps/chosen": -192.28500366210938,
      "logps/ref_chosen": -55.03598403930664,
      "logps/ref_rejected": -75.80644989013672,
      "logps/rejected": -287.8357849121094,
      "loss": 1.017,
      "margin_dpo/margin_mean": 74.78031921386719,
      "margin_dpo/margin_std": 84.29581451416016,
      "step": 557
    },
    {
      "KL/chosen_KL_mean": -123.99850463867188,
      "KL/mean": -163.4159393310547,
      "KL/rejected_KL_mean": -202.83338928222656,
      "KL/std": 89.908935546875,
      "epoch": 0.8435374149659864,
      "fcm_dpo/beta": 0.0066793859004974365,
      "fcm_dpo/delta": -0.1344417929649353,
      "fcm_dpo/margin": 78.83488464355469,
      "fcm_dpo/q_t": 0.38055527210235596,
      "grad_norm": 12.364710807800293,
      "learning_rate": 3.687450924416341e-08,
      "logits/chosen": 0.7823951840400696,
      "logits/rejected": 0.7239669561386108,
      "logps/chosen": -187.224853515625,
      "logps/ref_chosen": -63.226348876953125,
      "logps/ref_rejected": -91.46881866455078,
      "logps/rejected": -294.30218505859375,
      "loss": 1.0155,
      "margin_dpo/margin_mean": 78.83488464355469,
      "margin_dpo/margin_std": 92.04693603515625,
      "step": 558
    },
    {
      "KL/chosen_KL_mean": -129.73312377929688,
      "KL/mean": -163.43589782714844,
      "KL/rejected_KL_mean": -197.13868713378906,
      "KL/std": 86.80623626708984,
      "epoch": 0.8450491307634165,
      "fcm_dpo/beta": 0.00656571751460433,
      "fcm_dpo/delta": -0.04597283899784088,
      "fcm_dpo/margin": 67.40557861328125,
      "fcm_dpo/q_t": 0.4019114673137665,
      "grad_norm": 12.485381126403809,
      "learning_rate": 3.6186337553827743e-08,
      "logits/chosen": 0.713494062423706,
      "logits/rejected": 0.640967845916748,
      "logps/chosen": -191.2547607421875,
      "logps/ref_chosen": -61.521644592285156,
      "logps/ref_rejected": -82.83859252929688,
      "logps/rejected": -279.977294921875,
      "loss": 1.0991,
      "margin_dpo/margin_mean": 67.40557861328125,
      "margin_dpo/margin_std": 101.77870178222656,
      "step": 559
    },
    {
      "KL/chosen_KL_mean": -141.1266326904297,
      "KL/mean": -170.9273681640625,
      "KL/rejected_KL_mean": -200.7281036376953,
      "KL/std": 85.53338623046875,
      "epoch": 0.8465608465608465,
      "fcm_dpo/beta": 0.006634770426899195,
      "fcm_dpo/delta": 0.004244182258844376,
      "fcm_dpo/margin": 59.60146713256836,
      "fcm_dpo/q_t": 0.4095200300216675,
      "grad_norm": 14.990598678588867,
      "learning_rate": 3.550414669125573e-08,
      "logits/chosen": 0.7438457012176514,
      "logits/rejected": 0.7018010020256042,
      "logps/chosen": -201.76785278320312,
      "logps/ref_chosen": -60.64122009277344,
      "logps/ref_rejected": -78.75474548339844,
      "logps/rejected": -279.48284912109375,
      "loss": 1.1056,
      "margin_dpo/margin_mean": 59.60146713256836,
      "margin_dpo/margin_std": 84.04834747314453,
      "step": 560
    },
    {
      "KL/chosen_KL_mean": -128.27587890625,
      "KL/mean": -158.63803100585938,
      "KL/rejected_KL_mean": -189.0001983642578,
      "KL/std": 85.89263916015625,
      "epoch": 0.8480725623582767,
      "fcm_dpo/beta": 0.0065963175147771835,
      "fcm_dpo/delta": -0.000604395754635334,
      "fcm_dpo/margin": 60.72432327270508,
      "fcm_dpo/q_t": 0.40989649295806885,
      "grad_norm": 13.60486125946045,
      "learning_rate": 3.482795573879241e-08,
      "logits/chosen": 0.7049076557159424,
      "logits/rejected": 0.6703420877456665,
      "logps/chosen": -190.77447509765625,
      "logps/ref_chosen": -62.49859619140625,
      "logps/ref_rejected": -78.72064208984375,
      "logps/rejected": -267.7208251953125,
      "loss": 1.1156,
      "margin_dpo/margin_mean": 60.724327087402344,
      "margin_dpo/margin_std": 93.66244506835938,
      "step": 561
    },
    {
      "KL/chosen_KL_mean": -135.21826171875,
      "KL/mean": -172.11111450195312,
      "KL/rejected_KL_mean": -209.00396728515625,
      "KL/std": 90.7817611694336,
      "epoch": 0.8495842781557067,
      "fcm_dpo/beta": 0.006486967206001282,
      "fcm_dpo/delta": -0.08362391591072083,
      "fcm_dpo/margin": 73.78570556640625,
      "fcm_dpo/q_t": 0.3917215168476105,
      "grad_norm": 14.613893508911133,
      "learning_rate": 3.415778361095226e-08,
      "logits/chosen": 0.7305380702018738,
      "logits/rejected": 0.6867002248764038,
      "logps/chosen": -210.0,
      "logps/ref_chosen": -74.78173828125,
      "logps/ref_rejected": -92.63499450683594,
      "logps/rejected": -301.63897705078125,
      "loss": 1.0514,
      "margin_dpo/margin_mean": 73.78570556640625,
      "margin_dpo/margin_std": 93.51516723632812,
      "step": 562
    },
    {
      "KL/chosen_KL_mean": -116.42094421386719,
      "KL/mean": -149.75637817382812,
      "KL/rejected_KL_mean": -183.09182739257812,
      "KL/std": 79.57887268066406,
      "epoch": 0.8510959939531368,
      "fcm_dpo/beta": 0.006465439684689045,
      "fcm_dpo/delta": -0.03244488313794136,
      "fcm_dpo/margin": 66.67088317871094,
      "fcm_dpo/q_t": 0.40143883228302,
      "grad_norm": 17.642074584960938,
      "learning_rate": 3.349364905389032e-08,
      "logits/chosen": 0.8246127367019653,
      "logits/rejected": 0.7726570963859558,
      "logps/chosen": -166.61944580078125,
      "logps/ref_chosen": -50.19850158691406,
      "logps/ref_rejected": -66.76687622070312,
      "logps/rejected": -249.85870361328125,
      "loss": 1.1031,
      "margin_dpo/margin_mean": 66.67088317871094,
      "margin_dpo/margin_std": 101.82766723632812,
      "step": 563
    },
    {
      "KL/chosen_KL_mean": -118.33200073242188,
      "KL/mean": -159.535400390625,
      "KL/rejected_KL_mean": -200.73880004882812,
      "KL/std": 85.4729232788086,
      "epoch": 0.8526077097505669,
      "fcm_dpo/beta": 0.00632768590003252,
      "fcm_dpo/delta": -0.12817589938640594,
      "fcm_dpo/margin": 82.40682220458984,
      "fcm_dpo/q_t": 0.37969234585762024,
      "grad_norm": 13.583505630493164,
      "learning_rate": 3.283557064487785e-08,
      "logits/chosen": 0.7059386968612671,
      "logits/rejected": 0.6721379160881042,
      "logps/chosen": -174.07284545898438,
      "logps/ref_chosen": -55.7408447265625,
      "logps/ref_rejected": -74.82323455810547,
      "logps/rejected": -275.5620422363281,
      "loss": 1.0288,
      "margin_dpo/margin_mean": 82.40682220458984,
      "margin_dpo/margin_std": 102.0660171508789,
      "step": 564
    },
    {
      "KL/chosen_KL_mean": -145.9530487060547,
      "KL/mean": -173.38087463378906,
      "KL/rejected_KL_mean": -200.8087158203125,
      "KL/std": 84.87850952148438,
      "epoch": 0.854119425547997,
      "fcm_dpo/beta": 0.006371333263814449,
      "fcm_dpo/delta": 0.051446303725242615,
      "fcm_dpo/margin": 54.85563659667969,
      "fcm_dpo/q_t": 0.41960498690605164,
      "grad_norm": 15.257643699645996,
      "learning_rate": 3.218356679178252e-08,
      "logits/chosen": 0.7865220308303833,
      "logits/rejected": 0.7267142534255981,
      "logps/chosen": -204.29043579101562,
      "logps/ref_chosen": -58.33738327026367,
      "logps/ref_rejected": -78.31776428222656,
      "logps/rejected": -279.12646484375,
      "loss": 1.1414,
      "margin_dpo/margin_mean": 54.85563278198242,
      "margin_dpo/margin_std": 86.63621520996094,
      "step": 565
    },
    {
      "KL/chosen_KL_mean": -133.1039581298828,
      "KL/mean": -160.00714111328125,
      "KL/rejected_KL_mean": -186.9103240966797,
      "KL/std": 87.2846908569336,
      "epoch": 0.8556311413454271,
      "fcm_dpo/beta": 0.006431379355490208,
      "fcm_dpo/delta": 0.05535087734460831,
      "fcm_dpo/margin": 53.80636215209961,
      "fcm_dpo/q_t": 0.42274099588394165,
      "grad_norm": 16.522539138793945,
      "learning_rate": 3.1537655732553764e-08,
      "logits/chosen": 0.7471953630447388,
      "logits/rejected": 0.7294028997421265,
      "logps/chosen": -204.32769775390625,
      "logps/ref_chosen": -71.22373962402344,
      "logps/ref_rejected": -71.11601257324219,
      "logps/rejected": -258.0263366699219,
      "loss": 1.1891,
      "margin_dpo/margin_mean": 53.806365966796875,
      "margin_dpo/margin_std": 110.33029174804688,
      "step": 566
    },
    {
      "KL/chosen_KL_mean": -129.12673950195312,
      "KL/mean": -162.45062255859375,
      "KL/rejected_KL_mean": -195.7744903564453,
      "KL/std": 85.07839965820312,
      "epoch": 0.8571428571428571,
      "fcm_dpo/beta": 0.006362162530422211,
      "fcm_dpo/delta": -0.025937747210264206,
      "fcm_dpo/margin": 66.64776611328125,
      "fcm_dpo/q_t": 0.40236735343933105,
      "grad_norm": 12.060728073120117,
      "learning_rate": 3.089785553471233e-08,
      "logits/chosen": 0.7700002789497375,
      "logits/rejected": 0.6686394214630127,
      "logps/chosen": -181.79600524902344,
      "logps/ref_chosen": -52.669273376464844,
      "logps/ref_rejected": -74.34785461425781,
      "logps/rejected": -270.1223449707031,
      "loss": 1.0901,
      "margin_dpo/margin_mean": 66.64776611328125,
      "margin_dpo/margin_std": 91.76837158203125,
      "step": 567
    },
    {
      "KL/chosen_KL_mean": -118.26817321777344,
      "KL/mean": -159.29119873046875,
      "KL/rejected_KL_mean": -200.31422424316406,
      "KL/std": 87.94129180908203,
      "epoch": 0.8586545729402872,
      "fcm_dpo/beta": 0.006308428943157196,
      "fcm_dpo/delta": -0.12392584979534149,
      "fcm_dpo/margin": 82.04606628417969,
      "fcm_dpo/q_t": 0.3821907639503479,
      "grad_norm": 13.965873718261719,
      "learning_rate": 3.026418409484513e-08,
      "logits/chosen": 0.7885270714759827,
      "logits/rejected": 0.6974135637283325,
      "logps/chosen": -170.4461669921875,
      "logps/ref_chosen": -52.178001403808594,
      "logps/ref_rejected": -85.8277587890625,
      "logps/rejected": -286.1419677734375,
      "loss": 1.0129,
      "margin_dpo/margin_mean": 82.04606628417969,
      "margin_dpo/margin_std": 89.78453063964844,
      "step": 568
    },
    {
      "KL/chosen_KL_mean": -136.9675750732422,
      "KL/mean": -157.94886779785156,
      "KL/rejected_KL_mean": -178.93014526367188,
      "KL/std": 86.604248046875,
      "epoch": 0.8601662887377173,
      "fcm_dpo/beta": 0.006321226246654987,
      "fcm_dpo/delta": 0.13839997351169586,
      "fcm_dpo/margin": 41.96255874633789,
      "fcm_dpo/q_t": 0.43971410393714905,
      "grad_norm": 14.764012336730957,
      "learning_rate": 2.963665913810451e-08,
      "logits/chosen": 0.6662417650222778,
      "logits/rejected": 0.6308369636535645,
      "logps/chosen": -199.61683654785156,
      "logps/ref_chosen": -62.649261474609375,
      "logps/ref_rejected": -75.4298324584961,
      "logps/rejected": -254.3599853515625,
      "loss": 1.2198,
      "margin_dpo/margin_mean": 41.96255874633789,
      "margin_dpo/margin_std": 88.51194763183594,
      "step": 569
    },
    {
      "KL/chosen_KL_mean": -121.72870635986328,
      "KL/mean": -163.15060424804688,
      "KL/rejected_KL_mean": -204.57249450683594,
      "KL/std": 83.75516510009766,
      "epoch": 0.8616780045351474,
      "fcm_dpo/beta": 0.0063173240050673485,
      "fcm_dpo/delta": -0.13001130521297455,
      "fcm_dpo/margin": 82.84379577636719,
      "fcm_dpo/q_t": 0.3797076940536499,
      "grad_norm": 12.93687915802002,
      "learning_rate": 2.9015298217712453e-08,
      "logits/chosen": 0.7038021087646484,
      "logits/rejected": 0.617842435836792,
      "logps/chosen": -171.7705078125,
      "logps/ref_chosen": -50.04179382324219,
      "logps/ref_rejected": -78.27146911621094,
      "logps/rejected": -282.8439636230469,
      "loss": 1.0115,
      "margin_dpo/margin_mean": 82.84379577636719,
      "margin_dpo/margin_std": 92.3207015991211,
      "step": 570
    },
    {
      "KL/chosen_KL_mean": -133.497802734375,
      "KL/mean": -157.1429443359375,
      "KL/rejected_KL_mean": -180.78807067871094,
      "KL/std": 81.17963409423828,
      "epoch": 0.8631897203325775,
      "fcm_dpo/beta": 0.006368095986545086,
      "fcm_dpo/delta": 0.10106943547725677,
      "fcm_dpo/margin": 47.29026794433594,
      "fcm_dpo/q_t": 0.4300426244735718,
      "grad_norm": 13.396882057189941,
      "learning_rate": 2.840011871446962e-08,
      "logits/chosen": 0.7285163402557373,
      "logits/rejected": 0.695633053779602,
      "logps/chosen": -187.15460205078125,
      "logps/ref_chosen": -53.65681457519531,
      "logps/ref_rejected": -66.13298034667969,
      "logps/rejected": -246.92105102539062,
      "loss": 1.1986,
      "margin_dpo/margin_mean": 47.2902717590332,
      "margin_dpo/margin_std": 94.57589721679688,
      "step": 571
    },
    {
      "KL/chosen_KL_mean": -129.37582397460938,
      "KL/mean": -157.59616088867188,
      "KL/rejected_KL_mean": -185.8165283203125,
      "KL/std": 84.47772216796875,
      "epoch": 0.8647014361300076,
      "fcm_dpo/beta": 0.006431017071008682,
      "fcm_dpo/delta": 0.0380452424287796,
      "fcm_dpo/margin": 56.440704345703125,
      "fcm_dpo/q_t": 0.41607558727264404,
      "grad_norm": 14.078507423400879,
      "learning_rate": 2.7791137836269158e-08,
      "logits/chosen": 0.6906898617744446,
      "logits/rejected": 0.7365379333496094,
      "logps/chosen": -204.19374084472656,
      "logps/ref_chosen": -74.81792449951172,
      "logps/ref_rejected": -65.88681030273438,
      "logps/rejected": -251.7033233642578,
      "loss": 1.1253,
      "margin_dpo/margin_mean": 56.440704345703125,
      "margin_dpo/margin_std": 83.41856384277344,
      "step": 572
    },
    {
      "KL/chosen_KL_mean": -143.9004669189453,
      "KL/mean": -172.6072235107422,
      "KL/rejected_KL_mean": -201.31398010253906,
      "KL/std": 89.49702453613281,
      "epoch": 0.8662131519274376,
      "fcm_dpo/beta": 0.006447994615882635,
      "fcm_dpo/delta": 0.030953213572502136,
      "fcm_dpo/margin": 57.413509368896484,
      "fcm_dpo/q_t": 0.4167448580265045,
      "grad_norm": 14.49169921875,
      "learning_rate": 2.718837261761528e-08,
      "logits/chosen": 0.7309734225273132,
      "logits/rejected": 0.6812965273857117,
      "logps/chosen": -212.62611389160156,
      "logps/ref_chosen": -68.72564697265625,
      "logps/ref_rejected": -88.16201782226562,
      "logps/rejected": -289.47601318359375,
      "loss": 1.16,
      "margin_dpo/margin_mean": 57.413509368896484,
      "margin_dpo/margin_std": 106.84358978271484,
      "step": 573
    },
    {
      "KL/chosen_KL_mean": -128.6298065185547,
      "KL/mean": -164.86526489257812,
      "KL/rejected_KL_mean": -201.1007080078125,
      "KL/std": 81.8821792602539,
      "epoch": 0.8677248677248677,
      "fcm_dpo/beta": 0.006425968836992979,
      "fcm_dpo/delta": -0.06904555857181549,
      "fcm_dpo/margin": 72.47090911865234,
      "fcm_dpo/q_t": 0.3918231129646301,
      "grad_norm": 11.799493789672852,
      "learning_rate": 2.659183991914696e-08,
      "logits/chosen": 0.7962836623191833,
      "logits/rejected": 0.7224004864692688,
      "logps/chosen": -184.94320678710938,
      "logps/ref_chosen": -56.31340026855469,
      "logps/ref_rejected": -83.91553497314453,
      "logps/rejected": -285.0162353515625,
      "loss": 1.0329,
      "margin_dpo/margin_mean": 72.47090911865234,
      "margin_dpo/margin_std": 76.56863403320312,
      "step": 574
    },
    {
      "KL/chosen_KL_mean": -131.7310791015625,
      "KL/mean": -157.78036499023438,
      "KL/rejected_KL_mean": -183.82962036132812,
      "KL/std": 86.12600708007812,
      "epoch": 0.8692365835222978,
      "fcm_dpo/beta": 0.00639111315831542,
      "fcm_dpo/delta": 0.06874041259288788,
      "fcm_dpo/margin": 52.09855651855469,
      "fcm_dpo/q_t": 0.4262588620185852,
      "grad_norm": 14.336535453796387,
      "learning_rate": 2.600155642716606e-08,
      "logits/chosen": 0.7968418598175049,
      "logits/rejected": 0.7067606449127197,
      "logps/chosen": -196.31520080566406,
      "logps/ref_chosen": -64.5841293334961,
      "logps/ref_rejected": -93.47034454345703,
      "logps/rejected": -277.29998779296875,
      "loss": 1.1947,
      "margin_dpo/margin_mean": 52.09856414794922,
      "margin_dpo/margin_std": 105.56285095214844,
      "step": 575
    },
    {
      "KL/chosen_KL_mean": -116.5837631225586,
      "KL/mean": -153.09439086914062,
      "KL/rejected_KL_mean": -189.60504150390625,
      "KL/std": 85.46895599365234,
      "epoch": 0.8707482993197279,
      "fcm_dpo/beta": 0.006358052603900433,
      "fcm_dpo/delta": -0.06849232316017151,
      "fcm_dpo/margin": 73.02128601074219,
      "fcm_dpo/q_t": 0.39405471086502075,
      "grad_norm": 13.795515060424805,
      "learning_rate": 2.5417538653170754e-08,
      "logits/chosen": 0.7595860958099365,
      "logits/rejected": 0.6430308818817139,
      "logps/chosen": -169.86428833007812,
      "logps/ref_chosen": -53.28052520751953,
      "logps/ref_rejected": -84.2000503540039,
      "logps/rejected": -273.80511474609375,
      "loss": 1.0624,
      "margin_dpo/margin_mean": 73.02128601074219,
      "margin_dpo/margin_std": 95.01397705078125,
      "step": 576
    },
    {
      "KL/chosen_KL_mean": -130.9795379638672,
      "KL/mean": -157.20046997070312,
      "KL/rejected_KL_mean": -183.42141723632812,
      "KL/std": 84.2352523803711,
      "epoch": 0.872260015117158,
      "fcm_dpo/beta": 0.0064292605966329575,
      "fcm_dpo/delta": 0.06507161259651184,
      "fcm_dpo/margin": 52.44186782836914,
      "fcm_dpo/q_t": 0.4226120114326477,
      "grad_norm": 14.239356994628906,
      "learning_rate": 2.4839802933393607e-08,
      "logits/chosen": 0.7271114587783813,
      "logits/rejected": 0.7114219665527344,
      "logps/chosen": -193.30422973632812,
      "logps/ref_chosen": -62.32468795776367,
      "logps/ref_rejected": -67.300537109375,
      "logps/rejected": -250.72193908691406,
      "loss": 1.17,
      "margin_dpo/margin_mean": 52.441864013671875,
      "margin_dpo/margin_std": 95.66375732421875,
      "step": 577
    },
    {
      "KL/chosen_KL_mean": -125.42353820800781,
      "KL/mean": -150.02896118164062,
      "KL/rejected_KL_mean": -174.6343994140625,
      "KL/std": 85.88644409179688,
      "epoch": 0.873771730914588,
      "fcm_dpo/beta": 0.006528651341795921,
      "fcm_dpo/delta": 0.08139324188232422,
      "fcm_dpo/margin": 49.21087646484375,
      "fcm_dpo/q_t": 0.42820054292678833,
      "grad_norm": 16.858417510986328,
      "learning_rate": 2.4268365428344733e-08,
      "logits/chosen": 0.8017420172691345,
      "logits/rejected": 0.7788885831832886,
      "logps/chosen": -182.0791015625,
      "logps/ref_chosen": -56.65557861328125,
      "logps/ref_rejected": -68.21835327148438,
      "logps/rejected": -242.85275268554688,
      "loss": 1.1902,
      "margin_dpo/margin_mean": 49.21087646484375,
      "margin_dpo/margin_std": 99.24673461914062,
      "step": 578
    },
    {
      "KL/chosen_KL_mean": -131.38169860839844,
      "KL/mean": -167.9196014404297,
      "KL/rejected_KL_mean": -204.45751953125,
      "KL/std": 82.55770111083984,
      "epoch": 0.8752834467120182,
      "fcm_dpo/beta": 0.006479881703853607,
      "fcm_dpo/delta": -0.07728934288024902,
      "fcm_dpo/margin": 73.0758056640625,
      "fcm_dpo/q_t": 0.3899138271808624,
      "grad_norm": 13.836790084838867,
      "learning_rate": 2.3703242122359357e-08,
      "logits/chosen": 0.6977138519287109,
      "logits/rejected": 0.6707988381385803,
      "logps/chosen": -188.1913604736328,
      "logps/ref_chosen": -56.809661865234375,
      "logps/ref_rejected": -68.09613037109375,
      "logps/rejected": -272.55364990234375,
      "loss": 1.0382,
      "margin_dpo/margin_mean": 73.0758056640625,
      "margin_dpo/margin_std": 84.57794189453125,
      "step": 579
    },
    {
      "KL/chosen_KL_mean": -131.20523071289062,
      "KL/mean": -160.20413208007812,
      "KL/rejected_KL_mean": -189.20303344726562,
      "KL/std": 90.06289672851562,
      "epoch": 0.8767951625094482,
      "fcm_dpo/beta": 0.00650765048339963,
      "fcm_dpo/delta": 0.02309592440724373,
      "fcm_dpo/margin": 57.997798919677734,
      "fcm_dpo/q_t": 0.4151464104652405,
      "grad_norm": 13.468850135803223,
      "learning_rate": 2.3144448823151392e-08,
      "logits/chosen": 0.7023211717605591,
      "logits/rejected": 0.6415441632270813,
      "logps/chosen": -188.9053497314453,
      "logps/ref_chosen": -57.70011520385742,
      "logps/ref_rejected": -77.90664672851562,
      "logps/rejected": -267.10968017578125,
      "loss": 1.1525,
      "margin_dpo/margin_mean": 57.997802734375,
      "margin_dpo/margin_std": 103.89713287353516,
      "step": 580
    },
    {
      "KL/chosen_KL_mean": -142.56919860839844,
      "KL/mean": -172.34552001953125,
      "KL/rejected_KL_mean": -202.12184143066406,
      "KL/std": 82.92886352539062,
      "epoch": 0.8783068783068783,
      "fcm_dpo/beta": 0.006482171826064587,
      "fcm_dpo/delta": 0.014322012662887573,
      "fcm_dpo/margin": 59.552635192871094,
      "fcm_dpo/q_t": 0.4122830033302307,
      "grad_norm": 14.482420921325684,
      "learning_rate": 2.259200116137039e-08,
      "logits/chosen": 0.7224099636077881,
      "logits/rejected": 0.6515974998474121,
      "logps/chosen": -201.90155029296875,
      "logps/ref_chosen": -59.332359313964844,
      "logps/ref_rejected": -83.64482116699219,
      "logps/rejected": -285.76666259765625,
      "loss": 1.1249,
      "margin_dpo/margin_mean": 59.55262756347656,
      "margin_dpo/margin_std": 93.51412963867188,
      "step": 581
    },
    {
      "KL/chosen_KL_mean": -127.24820709228516,
      "KL/mean": -157.314453125,
      "KL/rejected_KL_mean": -187.38070678710938,
      "KL/std": 83.08796691894531,
      "epoch": 0.8798185941043084,
      "fcm_dpo/beta": 0.006529543083161116,
      "fcm_dpo/delta": 0.00757815595716238,
      "fcm_dpo/margin": 60.132503509521484,
      "fcm_dpo/q_t": 0.4105338454246521,
      "grad_norm": 11.920287132263184,
      "learning_rate": 2.204591459016525e-08,
      "logits/chosen": 0.7106418609619141,
      "logits/rejected": 0.7402459383010864,
      "logps/chosen": -191.41107177734375,
      "logps/ref_chosen": -64.16285705566406,
      "logps/ref_rejected": -58.632896423339844,
      "logps/rejected": -246.01361083984375,
      "loss": 1.121,
      "margin_dpo/margin_mean": 60.13249969482422,
      "margin_dpo/margin_std": 94.03811645507812,
      "step": 582
    },
    {
      "KL/chosen_KL_mean": -128.94577026367188,
      "KL/mean": -162.572998046875,
      "KL/rejected_KL_mean": -196.20022583007812,
      "KL/std": 88.48190307617188,
      "epoch": 0.8813303099017384,
      "fcm_dpo/beta": 0.006483516190201044,
      "fcm_dpo/delta": -0.03768821805715561,
      "fcm_dpo/margin": 67.25444793701172,
      "fcm_dpo/q_t": 0.40098488330841064,
      "grad_norm": 15.756782531738281,
      "learning_rate": 2.1506204384751064e-08,
      "logits/chosen": 0.8333076238632202,
      "logits/rejected": 0.7157968282699585,
      "logps/chosen": -180.81817626953125,
      "logps/ref_chosen": -51.87239456176758,
      "logps/ref_rejected": -83.86331176757812,
      "logps/rejected": -280.06353759765625,
      "loss": 1.107,
      "margin_dpo/margin_mean": 67.25444030761719,
      "margin_dpo/margin_std": 105.29559326171875,
      "step": 583
    },
    {
      "KL/chosen_KL_mean": -117.70571899414062,
      "KL/mean": -148.47128295898438,
      "KL/rejected_KL_mean": -179.23681640625,
      "KL/std": 83.452880859375,
      "epoch": 0.8828420256991686,
      "fcm_dpo/beta": 0.006448123138397932,
      "fcm_dpo/delta": 0.0030598118901252747,
      "fcm_dpo/margin": 61.531105041503906,
      "fcm_dpo/q_t": 0.4117351174354553,
      "grad_norm": 12.771221160888672,
      "learning_rate": 2.09728856419826e-08,
      "logits/chosen": 0.8767859935760498,
      "logits/rejected": 0.7617365717887878,
      "logps/chosen": -164.277099609375,
      "logps/ref_chosen": -46.571388244628906,
      "logps/ref_rejected": -80.67969512939453,
      "logps/rejected": -259.91650390625,
      "loss": 1.1436,
      "margin_dpo/margin_mean": 61.531105041503906,
      "margin_dpo/margin_std": 108.26390838623047,
      "step": 584
    },
    {
      "KL/chosen_KL_mean": -136.58163452148438,
      "KL/mean": -158.10357666015625,
      "KL/rejected_KL_mean": -179.6255340576172,
      "KL/std": 89.20777893066406,
      "epoch": 0.8843537414965986,
      "fcm_dpo/beta": 0.006614618003368378,
      "fcm_dpo/delta": 0.11793427169322968,
      "fcm_dpo/margin": 43.04389953613281,
      "fcm_dpo/q_t": 0.4342869818210602,
      "grad_norm": 13.342413902282715,
      "learning_rate": 2.044597327993153e-08,
      "logits/chosen": 0.7053936719894409,
      "logits/rejected": 0.6552038788795471,
      "logps/chosen": -194.7061767578125,
      "logps/ref_chosen": -58.124534606933594,
      "logps/ref_rejected": -79.00538635253906,
      "logps/rejected": -258.63092041015625,
      "loss": 1.2109,
      "margin_dpo/margin_mean": 43.04389953613281,
      "margin_dpo/margin_std": 91.91615295410156,
      "step": 585
    },
    {
      "KL/chosen_KL_mean": -128.1500244140625,
      "KL/mean": -160.58642578125,
      "KL/rejected_KL_mean": -193.02284240722656,
      "KL/std": 80.80349731445312,
      "epoch": 0.8858654572940288,
      "fcm_dpo/beta": 0.006581123918294907,
      "fcm_dpo/delta": -0.02841946855187416,
      "fcm_dpo/margin": 64.87281799316406,
      "fcm_dpo/q_t": 0.40084946155548096,
      "grad_norm": 17.90386199951172,
      "learning_rate": 1.9925482037469187e-08,
      "logits/chosen": 0.7673027515411377,
      "logits/rejected": 0.7177405953407288,
      "logps/chosen": -182.25164794921875,
      "logps/ref_chosen": -54.10163879394531,
      "logps/ref_rejected": -63.72113037109375,
      "logps/rejected": -256.74395751953125,
      "loss": 1.0596,
      "margin_dpo/margin_mean": 64.87281799316406,
      "margin_dpo/margin_std": 73.49497985839844,
      "step": 586
    },
    {
      "KL/chosen_KL_mean": -130.09133911132812,
      "KL/mean": -161.97439575195312,
      "KL/rejected_KL_mean": -193.85745239257812,
      "KL/std": 88.5246810913086,
      "epoch": 0.8873771730914588,
      "fcm_dpo/beta": 0.006571411155164242,
      "fcm_dpo/delta": -0.019869251176714897,
      "fcm_dpo/margin": 63.76612091064453,
      "fcm_dpo/q_t": 0.40272602438926697,
      "grad_norm": 14.561052322387695,
      "learning_rate": 1.9411426473854687e-08,
      "logits/chosen": 0.7784076929092407,
      "logits/rejected": 0.7722544074058533,
      "logps/chosen": -193.508544921875,
      "logps/ref_chosen": -63.41719436645508,
      "logps/ref_rejected": -63.47003936767578,
      "logps/rejected": -257.3275146484375,
      "loss": 1.1402,
      "margin_dpo/margin_mean": 63.76612091064453,
      "margin_dpo/margin_std": 114.6404800415039,
      "step": 587
    },
    {
      "KL/chosen_KL_mean": -132.00868225097656,
      "KL/mean": -167.29954528808594,
      "KL/rejected_KL_mean": -202.5904083251953,
      "KL/std": 87.38584899902344,
      "epoch": 0.8888888888888888,
      "fcm_dpo/beta": 0.006490037776529789,
      "fcm_dpo/delta": -0.06114605814218521,
      "fcm_dpo/margin": 70.58171844482422,
      "fcm_dpo/q_t": 0.3969864845275879,
      "grad_norm": 14.922515869140625,
      "learning_rate": 1.890382096832699e-08,
      "logits/chosen": 0.7928558588027954,
      "logits/rejected": 0.7458092570304871,
      "logps/chosen": -194.209716796875,
      "logps/ref_chosen": -62.20103454589844,
      "logps/ref_rejected": -82.10249328613281,
      "logps/rejected": -284.6929016113281,
      "loss": 1.0833,
      "margin_dpo/margin_mean": 70.58171844482422,
      "margin_dpo/margin_std": 103.22766876220703,
      "step": 588
    },
    {
      "KL/chosen_KL_mean": -125.28984069824219,
      "KL/mean": -159.66366577148438,
      "KL/rejected_KL_mean": -194.03750610351562,
      "KL/std": 86.01371002197266,
      "epoch": 0.890400604686319,
      "fcm_dpo/beta": 0.0064490921795368195,
      "fcm_dpo/delta": -0.04534650966525078,
      "fcm_dpo/margin": 68.74765014648438,
      "fcm_dpo/q_t": 0.39677882194519043,
      "grad_norm": 11.68237018585205,
      "learning_rate": 1.840267971970344e-08,
      "logits/chosen": 0.7326580286026001,
      "logits/rejected": 0.7025114297866821,
      "logps/chosen": -182.00344848632812,
      "logps/ref_chosen": -56.71361541748047,
      "logps/ref_rejected": -76.7366943359375,
      "logps/rejected": -270.774169921875,
      "loss": 1.0496,
      "margin_dpo/margin_mean": 68.7476577758789,
      "margin_dpo/margin_std": 77.22756958007812,
      "step": 589
    },
    {
      "KL/chosen_KL_mean": -139.61033630371094,
      "KL/mean": -172.72789001464844,
      "KL/rejected_KL_mean": -205.84544372558594,
      "KL/std": 81.31230163574219,
      "epoch": 0.891912320483749,
      "fcm_dpo/beta": 0.006417558528482914,
      "fcm_dpo/delta": -0.026233412325382233,
      "fcm_dpo/margin": 66.235107421875,
      "fcm_dpo/q_t": 0.40231773257255554,
      "grad_norm": 14.054452896118164,
      "learning_rate": 1.7908016745981856e-08,
      "logits/chosen": 0.6718609929084778,
      "logits/rejected": 0.6410232782363892,
      "logps/chosen": -206.1241455078125,
      "logps/ref_chosen": -66.5138168334961,
      "logps/ref_rejected": -85.70820617675781,
      "logps/rejected": -291.55364990234375,
      "loss": 1.0852,
      "margin_dpo/margin_mean": 66.235107421875,
      "margin_dpo/margin_std": 90.15684509277344,
      "step": 590
    },
    {
      "KL/chosen_KL_mean": -118.54312896728516,
      "KL/mean": -158.638916015625,
      "KL/rejected_KL_mean": -198.73472595214844,
      "KL/std": 91.9343490600586,
      "epoch": 0.8934240362811792,
      "fcm_dpo/beta": 0.00624478654935956,
      "fcm_dpo/delta": -0.10805796831846237,
      "fcm_dpo/margin": 80.19159698486328,
      "fcm_dpo/q_t": 0.3870072066783905,
      "grad_norm": 15.261886596679688,
      "learning_rate": 1.7419845883949098e-08,
      "logits/chosen": 0.788226842880249,
      "logits/rejected": 0.7209928035736084,
      "logps/chosen": -179.2403106689453,
      "logps/ref_chosen": -60.697181701660156,
      "logps/ref_rejected": -86.12278747558594,
      "logps/rejected": -284.8575134277344,
      "loss": 1.0635,
      "margin_dpo/margin_mean": 80.19159698486328,
      "margin_dpo/margin_std": 111.43367004394531,
      "step": 591
    },
    {
      "KL/chosen_KL_mean": -129.67877197265625,
      "KL/mean": -159.23971557617188,
      "KL/rejected_KL_mean": -188.8006591796875,
      "KL/std": 81.70418548583984,
      "epoch": 0.8949357520786092,
      "fcm_dpo/beta": 0.006255676969885826,
      "fcm_dpo/delta": 0.030980493873357773,
      "fcm_dpo/margin": 59.121883392333984,
      "fcm_dpo/q_t": 0.41672664880752563,
      "grad_norm": 15.047146797180176,
      "learning_rate": 1.6938180788793556e-08,
      "logits/chosen": 0.8125218152999878,
      "logits/rejected": 0.6916435956954956,
      "logps/chosen": -180.91610717773438,
      "logps/ref_chosen": -51.237327575683594,
      "logps/ref_rejected": -81.60242462158203,
      "logps/rejected": -270.403076171875,
      "loss": 1.1297,
      "margin_dpo/margin_mean": 59.12188720703125,
      "margin_dpo/margin_std": 91.02842712402344,
      "step": 592
    },
    {
      "KL/chosen_KL_mean": -109.8043441772461,
      "KL/mean": -141.08143615722656,
      "KL/rejected_KL_mean": -172.35853576660156,
      "KL/std": 84.07896423339844,
      "epoch": 0.8964474678760394,
      "fcm_dpo/beta": 0.006288354285061359,
      "fcm_dpo/delta": 0.006823260337114334,
      "fcm_dpo/margin": 62.55418395996094,
      "fcm_dpo/q_t": 0.4100233316421509,
      "grad_norm": 15.268369674682617,
      "learning_rate": 1.6463034933723336e-08,
      "logits/chosen": 0.7873135805130005,
      "logits/rejected": 0.6847280859947205,
      "logps/chosen": -151.88433837890625,
      "logps/ref_chosen": -42.08000183105469,
      "logps/ref_rejected": -68.47499084472656,
      "logps/rejected": -240.83352661132812,
      "loss": 1.1211,
      "margin_dpo/margin_mean": 62.55418395996094,
      "margin_dpo/margin_std": 97.87245178222656,
      "step": 593
    },
    {
      "KL/chosen_KL_mean": -137.69334411621094,
      "KL/mean": -166.93182373046875,
      "KL/rejected_KL_mean": -196.17031860351562,
      "KL/std": 82.94169616699219,
      "epoch": 0.8979591836734694,
      "fcm_dpo/beta": 0.006330487318336964,
      "fcm_dpo/delta": 0.03096877783536911,
      "fcm_dpo/margin": 58.47697067260742,
      "fcm_dpo/q_t": 0.4138473868370056,
      "grad_norm": 13.774144172668457,
      "learning_rate": 1.5994421609589385e-08,
      "logits/chosen": 0.680525541305542,
      "logits/rejected": 0.6660400629043579,
      "logps/chosen": -201.35202026367188,
      "logps/ref_chosen": -63.658668518066406,
      "logps/ref_rejected": -70.35597229003906,
      "logps/rejected": -266.52630615234375,
      "loss": 1.12,
      "margin_dpo/margin_mean": 58.476966857910156,
      "margin_dpo/margin_std": 85.59440612792969,
      "step": 594
    },
    {
      "KL/chosen_KL_mean": -124.25723266601562,
      "KL/mean": -162.75222778320312,
      "KL/rejected_KL_mean": -201.24722290039062,
      "KL/std": 87.26791381835938,
      "epoch": 0.8994708994708994,
      "fcm_dpo/beta": 0.006278345361351967,
      "fcm_dpo/delta": -0.08751600980758667,
      "fcm_dpo/margin": 76.98997497558594,
      "fcm_dpo/q_t": 0.3899436593055725,
      "grad_norm": 11.65858268737793,
      "learning_rate": 1.553235392451377e-08,
      "logits/chosen": 0.8071564435958862,
      "logits/rejected": 0.7116265296936035,
      "logps/chosen": -180.47598266601562,
      "logps/ref_chosen": -56.21875762939453,
      "logps/ref_rejected": -83.95773315429688,
      "logps/rejected": -285.2049560546875,
      "loss": 1.0656,
      "margin_dpo/margin_mean": 76.98997497558594,
      "margin_dpo/margin_std": 107.02861785888672,
      "step": 595
    },
    {
      "KL/chosen_KL_mean": -143.4345703125,
      "KL/mean": -161.5347900390625,
      "KL/rejected_KL_mean": -179.63502502441406,
      "KL/std": 82.03053283691406,
      "epoch": 0.9009826152683296,
      "fcm_dpo/beta": 0.00638097058981657,
      "fcm_dpo/delta": 0.1733156442642212,
      "fcm_dpo/margin": 36.20046615600586,
      "fcm_dpo/q_t": 0.44862306118011475,
      "grad_norm": 12.725683212280273,
      "learning_rate": 1.507684480352292e-08,
      "logits/chosen": 0.6726224422454834,
      "logits/rejected": 0.6943622827529907,
      "logps/chosen": -211.91543579101562,
      "logps/ref_chosen": -68.48088073730469,
      "logps/ref_rejected": -61.732967376708984,
      "logps/rejected": -241.36798095703125,
      "loss": 1.2479,
      "margin_dpo/margin_mean": 36.200462341308594,
      "margin_dpo/margin_std": 89.72453308105469,
      "step": 596
    },
    {
      "KL/chosen_KL_mean": -113.01495361328125,
      "KL/mean": -142.4587860107422,
      "KL/rejected_KL_mean": -171.9026336669922,
      "KL/std": 75.85586547851562,
      "epoch": 0.9024943310657596,
      "fcm_dpo/beta": 0.0064786700531840324,
      "fcm_dpo/delta": 0.019177531823515892,
      "fcm_dpo/margin": 58.8876838684082,
      "fcm_dpo/q_t": 0.41371387243270874,
      "grad_norm": 11.908020973205566,
      "learning_rate": 1.4627906988186111e-08,
      "logits/chosen": 0.7234436273574829,
      "logits/rejected": 0.7067715525627136,
      "logps/chosen": -161.87246704101562,
      "logps/ref_chosen": -48.85750961303711,
      "logps/ref_rejected": -55.068084716796875,
      "logps/rejected": -226.97071838378906,
      "loss": 1.1269,
      "margin_dpo/margin_mean": 58.8876838684082,
      "margin_dpo/margin_std": 93.37579345703125,
      "step": 597
    },
    {
      "KL/chosen_KL_mean": -147.4332733154297,
      "KL/mean": -165.73919677734375,
      "KL/rejected_KL_mean": -184.04510498046875,
      "KL/std": 84.73491668701172,
      "epoch": 0.9040060468631897,
      "fcm_dpo/beta": 0.006653393618762493,
      "fcm_dpo/delta": 0.15970611572265625,
      "fcm_dpo/margin": 36.611839294433594,
      "fcm_dpo/q_t": 0.44547873735427856,
      "grad_norm": 14.159972190856934,
      "learning_rate": 1.4185553036259095e-08,
      "logits/chosen": 0.7855315208435059,
      "logits/rejected": 0.6991105079650879,
      "logps/chosen": -206.3204345703125,
      "logps/ref_chosen": -58.88715362548828,
      "logps/ref_rejected": -81.43145751953125,
      "logps/rejected": -265.4765625,
      "loss": 1.2481,
      "margin_dpo/margin_mean": 36.611839294433594,
      "margin_dpo/margin_std": 92.60962677001953,
      "step": 598
    },
    {
      "KL/chosen_KL_mean": -148.88742065429688,
      "KL/mean": -171.52340698242188,
      "KL/rejected_KL_mean": -194.159423828125,
      "KL/std": 83.06340026855469,
      "epoch": 0.9055177626606198,
      "fcm_dpo/beta": 0.00680171512067318,
      "fcm_dpo/delta": 0.09460268169641495,
      "fcm_dpo/margin": 45.272010803222656,
      "fcm_dpo/q_t": 0.43099159002304077,
      "grad_norm": 16.28043556213379,
      "learning_rate": 1.3749795321332885e-08,
      "logits/chosen": 0.810226321220398,
      "logits/rejected": 0.7629626989364624,
      "logps/chosen": -206.49459838867188,
      "logps/ref_chosen": -57.60719299316406,
      "logps/ref_rejected": -71.80469512939453,
      "logps/rejected": -265.964111328125,
      "loss": 1.1985,
      "margin_dpo/margin_mean": 45.272010803222656,
      "margin_dpo/margin_std": 93.96842193603516,
      "step": 599
    },
    {
      "KL/chosen_KL_mean": -136.0979766845703,
      "KL/mean": -162.57826232910156,
      "KL/rejected_KL_mean": -189.0585479736328,
      "KL/std": 87.23487854003906,
      "epoch": 0.9070294784580499,
      "fcm_dpo/beta": 0.006898914463818073,
      "fcm_dpo/delta": 0.03474727272987366,
      "fcm_dpo/margin": 52.960575103759766,
      "fcm_dpo/q_t": 0.4179023802280426,
      "grad_norm": 15.556018829345703,
      "learning_rate": 1.3320646032487393e-08,
      "logits/chosen": 0.807528018951416,
      "logits/rejected": 0.7501277923583984,
      "logps/chosen": -194.540283203125,
      "logps/ref_chosen": -58.44231414794922,
      "logps/ref_rejected": -83.64639282226562,
      "logps/rejected": -272.7049560546875,
      "loss": 1.1522,
      "margin_dpo/margin_mean": 52.9605712890625,
      "margin_dpo/margin_std": 91.92424011230469,
      "step": 600
    },
    {
      "KL/chosen_KL_mean": -119.36679077148438,
      "KL/mean": -155.3492431640625,
      "KL/rejected_KL_mean": -191.33169555664062,
      "KL/std": 91.31021118164062,
      "epoch": 0.90854119425548,
      "fcm_dpo/beta": 0.006747450679540634,
      "fcm_dpo/delta": -0.09054695814847946,
      "fcm_dpo/margin": 71.96490478515625,
      "fcm_dpo/q_t": 0.39126312732696533,
      "grad_norm": 12.5004301071167,
      "learning_rate": 1.2898117173950868e-08,
      "logits/chosen": 0.7509094476699829,
      "logits/rejected": 0.6697346568107605,
      "logps/chosen": -174.96112060546875,
      "logps/ref_chosen": -55.59432601928711,
      "logps/ref_rejected": -83.68630981445312,
      "logps/rejected": -275.01800537109375,
      "loss": 1.0727,
      "margin_dpo/margin_mean": 71.96490478515625,
      "margin_dpo/margin_std": 103.93519592285156,
      "step": 601
    },
    {
      "KL/chosen_KL_mean": -107.9805679321289,
      "KL/mean": -143.06297302246094,
      "KL/rejected_KL_mean": -178.14535522460938,
      "KL/std": 81.46908569335938,
      "epoch": 0.91005291005291,
      "fcm_dpo/beta": 0.006687416695058346,
      "fcm_dpo/delta": -0.07258869707584381,
      "fcm_dpo/margin": 70.16480255126953,
      "fcm_dpo/q_t": 0.392657071352005,
      "grad_norm": 15.200639724731445,
      "learning_rate": 1.2482220564763667e-08,
      "logits/chosen": 0.7669535875320435,
      "logits/rejected": 0.7359898090362549,
      "logps/chosen": -164.3297576904297,
      "logps/ref_chosen": -56.349185943603516,
      "logps/ref_rejected": -71.9959716796875,
      "logps/rejected": -250.14132690429688,
      "loss": 1.0545,
      "margin_dpo/margin_mean": 70.16481018066406,
      "margin_dpo/margin_std": 88.97705841064453,
      "step": 602
    },
    {
      "KL/chosen_KL_mean": -121.89750671386719,
      "KL/mean": -154.79470825195312,
      "KL/rejected_KL_mean": -187.69189453125,
      "KL/std": 81.63348388671875,
      "epoch": 0.9115646258503401,
      "fcm_dpo/beta": 0.006597849540412426,
      "fcm_dpo/delta": -0.03575563803315163,
      "fcm_dpo/margin": 65.79436492919922,
      "fcm_dpo/q_t": 0.40124818682670593,
      "grad_norm": 15.745288848876953,
      "learning_rate": 1.2072967838448051e-08,
      "logits/chosen": 0.7055135369300842,
      "logits/rejected": 0.6456756591796875,
      "logps/chosen": -175.06588745117188,
      "logps/ref_chosen": -53.16838836669922,
      "logps/ref_rejected": -73.8604736328125,
      "logps/rejected": -261.5523681640625,
      "loss": 1.0937,
      "margin_dpo/margin_mean": 65.79436492919922,
      "margin_dpo/margin_std": 96.13505554199219,
      "step": 603
    },
    {
      "KL/chosen_KL_mean": -117.92399597167969,
      "KL/mean": -147.4550018310547,
      "KL/rejected_KL_mean": -176.9860076904297,
      "KL/std": 78.23771667480469,
      "epoch": 0.9130763416477702,
      "fcm_dpo/beta": 0.006615637801587582,
      "fcm_dpo/delta": 0.00958221685141325,
      "fcm_dpo/margin": 59.06201934814453,
      "fcm_dpo/q_t": 0.4119419753551483,
      "grad_norm": 15.086995124816895,
      "learning_rate": 1.1670370442682459e-08,
      "logits/chosen": 0.7008262872695923,
      "logits/rejected": 0.7061389684677124,
      "logps/chosen": -190.57342529296875,
      "logps/ref_chosen": -72.64942169189453,
      "logps/ref_rejected": -69.8792724609375,
      "logps/rejected": -246.8652801513672,
      "loss": 1.132,
      "margin_dpo/margin_mean": 59.06201934814453,
      "margin_dpo/margin_std": 98.15986633300781,
      "step": 604
    },
    {
      "KL/chosen_KL_mean": -136.82138061523438,
      "KL/mean": -165.74734497070312,
      "KL/rejected_KL_mean": -194.67333984375,
      "KL/std": 81.17981719970703,
      "epoch": 0.9145880574452003,
      "fcm_dpo/beta": 0.006611844524741173,
      "fcm_dpo/delta": 0.0181251373142004,
      "fcm_dpo/margin": 57.851959228515625,
      "fcm_dpo/q_t": 0.41242408752441406,
      "grad_norm": 14.910988807678223,
      "learning_rate": 1.1274439638981532e-08,
      "logits/chosen": 0.7845560312271118,
      "logits/rejected": 0.7260922193527222,
      "logps/chosen": -198.43421936035156,
      "logps/ref_chosen": -61.61284637451172,
      "logps/ref_rejected": -79.34398651123047,
      "logps/rejected": -274.017333984375,
      "loss": 1.1413,
      "margin_dpo/margin_mean": 57.851959228515625,
      "margin_dpo/margin_std": 98.74382019042969,
      "step": 605
    },
    {
      "KL/chosen_KL_mean": -119.74642944335938,
      "KL/mean": -154.47683715820312,
      "KL/rejected_KL_mean": -189.207275390625,
      "KL/std": 82.59077453613281,
      "epoch": 0.9160997732426304,
      "fcm_dpo/beta": 0.00659895408898592,
      "fcm_dpo/delta": -0.061252620071172714,
      "fcm_dpo/margin": 69.46084594726562,
      "fcm_dpo/q_t": 0.39719897508621216,
      "grad_norm": 16.641035079956055,
      "learning_rate": 1.0885186502381016e-08,
      "logits/chosen": 0.7329899072647095,
      "logits/rejected": 0.662407636642456,
      "logps/chosen": -174.21066284179688,
      "logps/ref_chosen": -54.46424102783203,
      "logps/ref_rejected": -79.62708282470703,
      "logps/rejected": -268.8343505859375,
      "loss": 1.0787,
      "margin_dpo/margin_mean": 69.46084594726562,
      "margin_dpo/margin_std": 98.36412048339844,
      "step": 606
    },
    {
      "KL/chosen_KL_mean": -134.2374725341797,
      "KL/mean": -165.53192138671875,
      "KL/rejected_KL_mean": -196.82638549804688,
      "KL/std": 83.826416015625,
      "epoch": 0.9176114890400605,
      "fcm_dpo/beta": 0.006492065265774727,
      "fcm_dpo/delta": -0.007983453571796417,
      "fcm_dpo/margin": 62.58891296386719,
      "fcm_dpo/q_t": 0.407237708568573,
      "grad_norm": 14.207694053649902,
      "learning_rate": 1.0502621921127774e-08,
      "logits/chosen": 0.6957761645317078,
      "logits/rejected": 0.6701861023902893,
      "logps/chosen": -197.09832763671875,
      "logps/ref_chosen": -62.86086654663086,
      "logps/ref_rejected": -72.5501937866211,
      "logps/rejected": -269.3765869140625,
      "loss": 1.1179,
      "margin_dpo/margin_mean": 62.58891296386719,
      "margin_dpo/margin_std": 95.68101501464844,
      "step": 607
    },
    {
      "KL/chosen_KL_mean": -135.67677307128906,
      "KL/mean": -166.17120361328125,
      "KL/rejected_KL_mean": -196.66561889648438,
      "KL/std": 85.26414489746094,
      "epoch": 0.9191232048374905,
      "fcm_dpo/beta": 0.006548555102199316,
      "fcm_dpo/delta": 0.0006168922409415245,
      "fcm_dpo/margin": 60.98883819580078,
      "fcm_dpo/q_t": 0.4089788794517517,
      "grad_norm": 14.347258567810059,
      "learning_rate": 1.0126756596375685e-08,
      "logits/chosen": 0.7177504301071167,
      "logits/rejected": 0.6351562738418579,
      "logps/chosen": -198.8574981689453,
      "logps/ref_chosen": -63.18071746826172,
      "logps/ref_rejected": -99.15888214111328,
      "logps/rejected": -295.82452392578125,
      "loss": 1.1048,
      "margin_dpo/margin_mean": 60.98883819580078,
      "margin_dpo/margin_std": 88.06265258789062,
      "step": 608
    },
    {
      "KL/chosen_KL_mean": -120.43633270263672,
      "KL/mean": -155.95809936523438,
      "KL/rejected_KL_mean": -191.4798583984375,
      "KL/std": 82.86610412597656,
      "epoch": 0.9206349206349206,
      "fcm_dpo/beta": 0.006455970928072929,
      "fcm_dpo/delta": -0.062065087258815765,
      "fcm_dpo/margin": 71.04353332519531,
      "fcm_dpo/q_t": 0.3932182192802429,
      "grad_norm": 12.620007514953613,
      "learning_rate": 9.757601041885694e-09,
      "logits/chosen": 0.7929271459579468,
      "logits/rejected": 0.7561884522438049,
      "logps/chosen": -169.05955505371094,
      "logps/ref_chosen": -48.62322235107422,
      "logps/ref_rejected": -68.28271484375,
      "logps/rejected": -259.7625732421875,
      "loss": 1.0536,
      "margin_dpo/margin_mean": 71.04353332519531,
      "margin_dpo/margin_std": 84.36087036132812,
      "step": 609
    },
    {
      "KL/chosen_KL_mean": -129.75857543945312,
      "KL/mean": -164.24969482421875,
      "KL/rejected_KL_mean": -198.74082946777344,
      "KL/std": 86.79969024658203,
      "epoch": 0.9221466364323507,
      "fcm_dpo/beta": 0.0063972556963562965,
      "fcm_dpo/delta": -0.04361305385828018,
      "fcm_dpo/margin": 68.98225402832031,
      "fcm_dpo/q_t": 0.40069544315338135,
      "grad_norm": 14.097638130187988,
      "learning_rate": 9.395165583732379e-09,
      "logits/chosen": 0.699439525604248,
      "logits/rejected": 0.696520209312439,
      "logps/chosen": -202.4237060546875,
      "logps/ref_chosen": -72.66513061523438,
      "logps/ref_rejected": -87.15310668945312,
      "logps/rejected": -285.8939208984375,
      "loss": 1.0843,
      "margin_dpo/margin_mean": 68.98225402832031,
      "margin_dpo/margin_std": 97.63822937011719,
      "step": 610
    },
    {
      "KL/chosen_KL_mean": -122.30868530273438,
      "KL/mean": -149.33563232421875,
      "KL/rejected_KL_mean": -176.3625946044922,
      "KL/std": 80.62203979492188,
      "epoch": 0.9236583522297808,
      "fcm_dpo/beta": 0.006440991070121527,
      "fcm_dpo/delta": 0.053715676069259644,
      "fcm_dpo/margin": 54.05390167236328,
      "fcm_dpo/q_t": 0.41938698291778564,
      "grad_norm": 14.460498809814453,
      "learning_rate": 9.03946036001449e-09,
      "logits/chosen": 0.7778803110122681,
      "logits/rejected": 0.7268798351287842,
      "logps/chosen": -170.6172637939453,
      "logps/ref_chosen": -48.30857849121094,
      "logps/ref_rejected": -70.6141128540039,
      "logps/rejected": -246.97671508789062,
      "loss": 1.1383,
      "margin_dpo/margin_mean": 54.05390167236328,
      "margin_dpo/margin_std": 83.94294738769531,
      "step": 611
    },
    {
      "KL/chosen_KL_mean": -132.1772003173828,
      "KL/mean": -170.843994140625,
      "KL/rejected_KL_mean": -209.51080322265625,
      "KL/std": 84.51614379882812,
      "epoch": 0.9251700680272109,
      "fcm_dpo/beta": 0.006384260021150112,
      "fcm_dpo/delta": -0.09853628277778625,
      "fcm_dpo/margin": 77.33361053466797,
      "fcm_dpo/q_t": 0.3858988583087921,
      "grad_norm": 12.366726875305176,
      "learning_rate": 8.690495320571839e-09,
      "logits/chosen": 0.6507830619812012,
      "logits/rejected": 0.5809448957443237,
      "logps/chosen": -193.40875244140625,
      "logps/ref_chosen": -61.23155975341797,
      "logps/ref_rejected": -94.37979888916016,
      "logps/rejected": -303.8905944824219,
      "loss": 1.0366,
      "margin_dpo/margin_mean": 77.33360290527344,
      "margin_dpo/margin_std": 94.39988708496094,
      "step": 612
    },
    {
      "KL/chosen_KL_mean": -112.42122650146484,
      "KL/mean": -149.91241455078125,
      "KL/rejected_KL_mean": -187.40362548828125,
      "KL/std": 86.88549041748047,
      "epoch": 0.926681783824641,
      "fcm_dpo/beta": 0.006296813488006592,
      "fcm_dpo/delta": -0.07568641006946564,
      "fcm_dpo/margin": 74.98237609863281,
      "fcm_dpo/q_t": 0.3908718228340149,
      "grad_norm": 11.499692916870117,
      "learning_rate": 8.348280226706722e-09,
      "logits/chosen": 0.7143006324768066,
      "logits/rejected": 0.7107223272323608,
      "logps/chosen": -166.40432739257812,
      "logps/ref_chosen": -53.98310852050781,
      "logps/ref_rejected": -58.32208251953125,
      "logps/rejected": -245.7257080078125,
      "loss": 1.0477,
      "margin_dpo/margin_mean": 74.98237609863281,
      "margin_dpo/margin_std": 91.57196044921875,
      "step": 613
    },
    {
      "KL/chosen_KL_mean": -132.50953674316406,
      "KL/mean": -166.3426971435547,
      "KL/rejected_KL_mean": -200.17584228515625,
      "KL/std": 77.45066833496094,
      "epoch": 0.9281934996220711,
      "fcm_dpo/beta": 0.0062566944397985935,
      "fcm_dpo/delta": -0.02466324158012867,
      "fcm_dpo/margin": 67.66629028320312,
      "fcm_dpo/q_t": 0.40125787258148193,
      "grad_norm": 14.503677368164062,
      "learning_rate": 8.012824650910937e-09,
      "logits/chosen": 0.7738802433013916,
      "logits/rejected": 0.7663772106170654,
      "logps/chosen": -192.75257873535156,
      "logps/ref_chosen": -60.24303436279297,
      "logps/ref_rejected": -72.26258850097656,
      "logps/rejected": -272.43841552734375,
      "loss": 1.0817,
      "margin_dpo/margin_mean": 67.66629028320312,
      "margin_dpo/margin_std": 88.4701156616211,
      "step": 614
    },
    {
      "KL/chosen_KL_mean": -138.55084228515625,
      "KL/mean": -171.46139526367188,
      "KL/rejected_KL_mean": -204.37196350097656,
      "KL/std": 94.35890197753906,
      "epoch": 0.9297052154195011,
      "fcm_dpo/beta": 0.006181590724736452,
      "fcm_dpo/delta": -0.007811099290847778,
      "fcm_dpo/margin": 65.82112121582031,
      "fcm_dpo/q_t": 0.4073028564453125,
      "grad_norm": 13.95335865020752,
      "learning_rate": 7.684137976598088e-09,
      "logits/chosen": 0.6753830909729004,
      "logits/rejected": 0.6221166849136353,
      "logps/chosen": -210.6455078125,
      "logps/ref_chosen": -72.09467315673828,
      "logps/ref_rejected": -104.02980041503906,
      "logps/rejected": -308.4017639160156,
      "loss": 1.1256,
      "margin_dpo/margin_mean": 65.82112121582031,
      "margin_dpo/margin_std": 106.96760559082031,
      "step": 615
    },
    {
      "KL/chosen_KL_mean": -130.0317840576172,
      "KL/mean": -160.46640014648438,
      "KL/rejected_KL_mean": -190.9010009765625,
      "KL/std": 87.95735168457031,
      "epoch": 0.9312169312169312,
      "fcm_dpo/beta": 0.006221260875463486,
      "fcm_dpo/delta": 0.02212386578321457,
      "fcm_dpo/margin": 60.869224548339844,
      "fcm_dpo/q_t": 0.41363078355789185,
      "grad_norm": 12.034595489501953,
      "learning_rate": 7.36222939784098e-09,
      "logits/chosen": 0.7884162068367004,
      "logits/rejected": 0.7024547457695007,
      "logps/chosen": -188.5625,
      "logps/ref_chosen": -58.530723571777344,
      "logps/ref_rejected": -75.48025512695312,
      "logps/rejected": -266.3812561035156,
      "loss": 1.1211,
      "margin_dpo/margin_mean": 60.869224548339844,
      "margin_dpo/margin_std": 92.23234558105469,
      "step": 616
    },
    {
      "KL/chosen_KL_mean": -144.70751953125,
      "KL/mean": -168.7492218017578,
      "KL/rejected_KL_mean": -192.79092407226562,
      "KL/std": 83.13668823242188,
      "epoch": 0.9327286470143613,
      "fcm_dpo/beta": 0.006288778968155384,
      "fcm_dpo/delta": 0.10031160712242126,
      "fcm_dpo/margin": 48.083404541015625,
      "fcm_dpo/q_t": 0.430725634098053,
      "grad_norm": 15.752001762390137,
      "learning_rate": 7.047107919114586e-09,
      "logits/chosen": 0.7634665966033936,
      "logits/rejected": 0.7092599868774414,
      "logps/chosen": -202.31619262695312,
      "logps/ref_chosen": -57.608673095703125,
      "logps/ref_rejected": -81.22109985351562,
      "logps/rejected": -274.01202392578125,
      "loss": 1.1847,
      "margin_dpo/margin_mean": 48.083404541015625,
      "margin_dpo/margin_std": 87.39576721191406,
      "step": 617
    },
    {
      "KL/chosen_KL_mean": -124.86834716796875,
      "KL/mean": -155.3878936767578,
      "KL/rejected_KL_mean": -185.90744018554688,
      "KL/std": 89.1173095703125,
      "epoch": 0.9342403628117913,
      "fcm_dpo/beta": 0.006361059378832579,
      "fcm_dpo/delta": 0.012086881324648857,
      "fcm_dpo/margin": 61.039093017578125,
      "fcm_dpo/q_t": 0.41178715229034424,
      "grad_norm": 16.19603729248047,
      "learning_rate": 6.738782355044048e-09,
      "logits/chosen": 0.7317014932632446,
      "logits/rejected": 0.6204960346221924,
      "logps/chosen": -181.5642852783203,
      "logps/ref_chosen": -56.69594192504883,
      "logps/ref_rejected": -85.92362976074219,
      "logps/rejected": -271.8310546875,
      "loss": 1.1099,
      "margin_dpo/margin_mean": 61.039093017578125,
      "margin_dpo/margin_std": 88.63739013671875,
      "step": 618
    },
    {
      "KL/chosen_KL_mean": -123.36766815185547,
      "KL/mean": -156.4314422607422,
      "KL/rejected_KL_mean": -189.4952392578125,
      "KL/std": 86.46694946289062,
      "epoch": 0.9357520786092215,
      "fcm_dpo/beta": 0.0063554528169333935,
      "fcm_dpo/delta": -0.021202079951763153,
      "fcm_dpo/margin": 66.1275634765625,
      "fcm_dpo/q_t": 0.4031738042831421,
      "grad_norm": 13.8381929397583,
      "learning_rate": 6.437261330158206e-09,
      "logits/chosen": 0.8231503963470459,
      "logits/rejected": 0.7432982921600342,
      "logps/chosen": -177.42608642578125,
      "logps/ref_chosen": -54.05841827392578,
      "logps/ref_rejected": -83.55493927001953,
      "logps/rejected": -273.0501708984375,
      "loss": 1.0951,
      "margin_dpo/margin_mean": 66.1275634765625,
      "margin_dpo/margin_std": 95.22664642333984,
      "step": 619
    },
    {
      "KL/chosen_KL_mean": -121.48786926269531,
      "KL/mean": -149.679443359375,
      "KL/rejected_KL_mean": -177.87103271484375,
      "KL/std": 78.28753662109375,
      "epoch": 0.9372637944066515,
      "fcm_dpo/beta": 0.006452272646129131,
      "fcm_dpo/delta": 0.03521667420864105,
      "fcm_dpo/margin": 56.383155822753906,
      "fcm_dpo/q_t": 0.4167051613330841,
      "grad_norm": 14.916093826293945,
      "learning_rate": 6.142553278648238e-09,
      "logits/chosen": 0.7305669188499451,
      "logits/rejected": 0.7315757274627686,
      "logps/chosen": -184.8575897216797,
      "logps/ref_chosen": -63.36971664428711,
      "logps/ref_rejected": -65.68269348144531,
      "logps/rejected": -243.5537109375,
      "loss": 1.14,
      "margin_dpo/margin_mean": 56.38315200805664,
      "margin_dpo/margin_std": 87.23406982421875,
      "step": 620
    },
    {
      "KL/chosen_KL_mean": -125.88284301757812,
      "KL/mean": -152.0810546875,
      "KL/rejected_KL_mean": -178.27926635742188,
      "KL/std": 81.47933959960938,
      "epoch": 0.9387755102040817,
      "fcm_dpo/beta": 0.006477131508290768,
      "fcm_dpo/delta": 0.06243708357214928,
      "fcm_dpo/margin": 52.39640426635742,
      "fcm_dpo/q_t": 0.4232029914855957,
      "grad_norm": 16.247522354125977,
      "learning_rate": 5.854666444131934e-09,
      "logits/chosen": 0.7784920334815979,
      "logits/rejected": 0.6623125672340393,
      "logps/chosen": -178.20407104492188,
      "logps/ref_chosen": -52.321224212646484,
      "logps/ref_rejected": -88.09001159667969,
      "logps/rejected": -266.3692626953125,
      "loss": 1.1679,
      "margin_dpo/margin_mean": 52.39640808105469,
      "margin_dpo/margin_std": 96.42157745361328,
      "step": 621
    },
    {
      "KL/chosen_KL_mean": -135.88552856445312,
      "KL/mean": -165.81393432617188,
      "KL/rejected_KL_mean": -195.74232482910156,
      "KL/std": 86.95231628417969,
      "epoch": 0.9402872260015117,
      "fcm_dpo/beta": 0.006505551747977734,
      "fcm_dpo/delta": 0.010898975655436516,
      "fcm_dpo/margin": 59.85680389404297,
      "fcm_dpo/q_t": 0.41105908155441284,
      "grad_norm": 15.815767288208008,
      "learning_rate": 5.573608879422875e-09,
      "logits/chosen": 0.7089018821716309,
      "logits/rejected": 0.6645277142524719,
      "logps/chosen": -195.7509765625,
      "logps/ref_chosen": -59.86545944213867,
      "logps/ref_rejected": -81.86668395996094,
      "logps/rejected": -277.6090087890625,
      "loss": 1.1207,
      "margin_dpo/margin_mean": 59.85680389404297,
      "margin_dpo/margin_std": 92.72256469726562,
      "step": 622
    },
    {
      "KL/chosen_KL_mean": -144.07167053222656,
      "KL/mean": -175.5872344970703,
      "KL/rejected_KL_mean": -207.102783203125,
      "KL/std": 89.51210021972656,
      "epoch": 0.9417989417989417,
      "fcm_dpo/beta": 0.006464548408985138,
      "fcm_dpo/delta": -0.008051551878452301,
      "fcm_dpo/margin": 63.0311164855957,
      "fcm_dpo/q_t": 0.40689247846603394,
      "grad_norm": 13.286748886108398,
      "learning_rate": 5.299388446305342e-09,
      "logits/chosen": 0.7314743399620056,
      "logits/rejected": 0.6700347661972046,
      "logps/chosen": -211.44012451171875,
      "logps/ref_chosen": -67.36846160888672,
      "logps/ref_rejected": -82.02733612060547,
      "logps/rejected": -289.130126953125,
      "loss": 1.1063,
      "margin_dpo/margin_mean": 63.03112030029297,
      "margin_dpo/margin_std": 93.41072082519531,
      "step": 623
    },
    {
      "KL/chosen_KL_mean": -121.00255584716797,
      "KL/mean": -156.09371948242188,
      "KL/rejected_KL_mean": -191.18487548828125,
      "KL/std": 88.24122619628906,
      "epoch": 0.9433106575963719,
      "fcm_dpo/beta": 0.006418424658477306,
      "fcm_dpo/delta": -0.05306580290198326,
      "fcm_dpo/margin": 70.18232727050781,
      "fcm_dpo/q_t": 0.39883553981781006,
      "grad_norm": 16.014657974243164,
      "learning_rate": 5.03201281531429e-09,
      "logits/chosen": 0.7183233499526978,
      "logits/rejected": 0.6201412677764893,
      "logps/chosen": -172.02911376953125,
      "logps/ref_chosen": -51.02655029296875,
      "logps/ref_rejected": -76.49203491210938,
      "logps/rejected": -267.6769104003906,
      "loss": 1.0823,
      "margin_dpo/margin_mean": 70.18232727050781,
      "margin_dpo/margin_std": 100.0351333618164,
      "step": 624
    },
    {
      "KL/chosen_KL_mean": -124.61507415771484,
      "KL/mean": -151.14584350585938,
      "KL/rejected_KL_mean": -177.6766357421875,
      "KL/std": 84.18926239013672,
      "epoch": 0.9448223733938019,
      "fcm_dpo/beta": 0.0064721377566456795,
      "fcm_dpo/delta": 0.05858816206455231,
      "fcm_dpo/margin": 53.061546325683594,
      "fcm_dpo/q_t": 0.4235745668411255,
      "grad_norm": 14.729056358337402,
      "learning_rate": 4.7714894655209174e-09,
      "logits/chosen": 0.8177133202552795,
      "logits/rejected": 0.7256171107292175,
      "logps/chosen": -178.82269287109375,
      "logps/ref_chosen": -54.20761489868164,
      "logps/ref_rejected": -84.93669128417969,
      "logps/rejected": -262.61334228515625,
      "loss": 1.1742,
      "margin_dpo/margin_mean": 53.061546325683594,
      "margin_dpo/margin_std": 102.11767578125,
      "step": 625
    },
    {
      "KL/chosen_KL_mean": -118.29896545410156,
      "KL/mean": -153.6566619873047,
      "KL/rejected_KL_mean": -189.01434326171875,
      "KL/std": 87.5848388671875,
      "epoch": 0.9463340891912321,
      "fcm_dpo/beta": 0.006424276158213615,
      "fcm_dpo/delta": -0.05705726146697998,
      "fcm_dpo/margin": 70.71540069580078,
      "fcm_dpo/q_t": 0.3995022177696228,
      "grad_norm": 13.807735443115234,
      "learning_rate": 4.517825684323323e-09,
      "logits/chosen": 0.8485870361328125,
      "logits/rejected": 0.7073635458946228,
      "logps/chosen": -163.36099243164062,
      "logps/ref_chosen": -45.06201934814453,
      "logps/ref_rejected": -89.66368103027344,
      "logps/rejected": -278.67803955078125,
      "loss": 1.0926,
      "margin_dpo/margin_mean": 70.71539306640625,
      "margin_dpo/margin_std": 107.69230651855469,
      "step": 626
    },
    {
      "KL/chosen_KL_mean": -133.79067993164062,
      "KL/mean": -171.92047119140625,
      "KL/rejected_KL_mean": -210.05026245117188,
      "KL/std": 89.4090347290039,
      "epoch": 0.9478458049886621,
      "fcm_dpo/beta": 0.006317706312984228,
      "fcm_dpo/delta": -0.08633463829755783,
      "fcm_dpo/margin": 76.25958251953125,
      "fcm_dpo/q_t": 0.38901180028915405,
      "grad_norm": 14.114999771118164,
      "learning_rate": 4.271028567242818e-09,
      "logits/chosen": 0.6920174360275269,
      "logits/rejected": 0.5662950277328491,
      "logps/chosen": -192.58172607421875,
      "logps/ref_chosen": -58.791053771972656,
      "logps/ref_rejected": -94.90802001953125,
      "logps/rejected": -304.9582824707031,
      "loss": 1.0548,
      "margin_dpo/margin_mean": 76.25958251953125,
      "margin_dpo/margin_std": 99.61177062988281,
      "step": 627
    },
    {
      "KL/chosen_KL_mean": -119.68309020996094,
      "KL/mean": -158.0455780029297,
      "KL/rejected_KL_mean": -196.40806579589844,
      "KL/std": 86.85939025878906,
      "epoch": 0.9493575207860923,
      "fcm_dpo/beta": 0.006282473914325237,
      "fcm_dpo/delta": -0.08701348304748535,
      "fcm_dpo/margin": 76.7249755859375,
      "fcm_dpo/q_t": 0.388563334941864,
      "grad_norm": 18.08017349243164,
      "learning_rate": 4.0311050177251895e-09,
      "logits/chosen": 0.734749436378479,
      "logits/rejected": 0.6959051489830017,
      "logps/chosen": -172.48666381835938,
      "logps/ref_chosen": -52.80357360839844,
      "logps/ref_rejected": -76.49468994140625,
      "logps/rejected": -272.90277099609375,
      "loss": 1.0688,
      "margin_dpo/margin_mean": 76.7249755859375,
      "margin_dpo/margin_std": 100.08718872070312,
      "step": 628
    },
    {
      "KL/chosen_KL_mean": -129.71226501464844,
      "KL/mean": -157.51455688476562,
      "KL/rejected_KL_mean": -185.31686401367188,
      "KL/std": 83.63412475585938,
      "epoch": 0.9508692365835223,
      "fcm_dpo/beta": 0.006266086827963591,
      "fcm_dpo/delta": 0.0531187504529953,
      "fcm_dpo/margin": 55.60459518432617,
      "fcm_dpo/q_t": 0.4206107258796692,
      "grad_norm": 11.870978355407715,
      "learning_rate": 3.798061746947995e-09,
      "logits/chosen": 0.7496498823165894,
      "logits/rejected": 0.7434634566307068,
      "logps/chosen": -200.42974853515625,
      "logps/ref_chosen": -70.71749877929688,
      "logps/ref_rejected": -78.96273803710938,
      "logps/rejected": -264.27960205078125,
      "loss": 1.1358,
      "margin_dpo/margin_mean": 55.604591369628906,
      "margin_dpo/margin_std": 84.9172134399414,
      "step": 629
    },
    {
      "KL/chosen_KL_mean": -122.6483154296875,
      "KL/mean": -159.9776611328125,
      "KL/rejected_KL_mean": -197.30702209472656,
      "KL/std": 89.26514434814453,
      "epoch": 0.9523809523809523,
      "fcm_dpo/beta": 0.006218547932803631,
      "fcm_dpo/delta": -0.0673595741391182,
      "fcm_dpo/margin": 74.65872192382812,
      "fcm_dpo/q_t": 0.3939681053161621,
      "grad_norm": 10.882546424865723,
      "learning_rate": 3.5719052736323806e-09,
      "logits/chosen": 0.7177826166152954,
      "logits/rejected": 0.671391487121582,
      "logps/chosen": -178.8497314453125,
      "logps/ref_chosen": -56.201412200927734,
      "logps/ref_rejected": -74.69807434082031,
      "logps/rejected": -272.0050964355469,
      "loss": 1.0578,
      "margin_dpo/margin_mean": 74.65872192382812,
      "margin_dpo/margin_std": 95.45807647705078,
      "step": 630
    },
    {
      "KL/chosen_KL_mean": -117.01614379882812,
      "KL/mean": -158.29913330078125,
      "KL/rejected_KL_mean": -199.58212280273438,
      "KL/std": 91.83998107910156,
      "epoch": 0.9538926681783825,
      "fcm_dpo/beta": 0.0060370988212525845,
      "fcm_dpo/delta": -0.1057269275188446,
      "fcm_dpo/margin": 82.56596374511719,
      "fcm_dpo/q_t": 0.38690662384033203,
      "grad_norm": 13.035811424255371,
      "learning_rate": 3.352641923861144e-09,
      "logits/chosen": 0.8515388369560242,
      "logits/rejected": 0.7325578331947327,
      "logps/chosen": -175.8367462158203,
      "logps/ref_chosen": -58.82059860229492,
      "logps/ref_rejected": -96.51437377929688,
      "logps/rejected": -296.09649658203125,
      "loss": 1.0445,
      "margin_dpo/margin_mean": 82.56597137451172,
      "margin_dpo/margin_std": 103.42387390136719,
      "step": 631
    },
    {
      "KL/chosen_KL_mean": -119.71453857421875,
      "KL/mean": -160.3410186767578,
      "KL/rejected_KL_mean": -200.96749877929688,
      "KL/std": 84.33313751220703,
      "epoch": 0.9554043839758125,
      "fcm_dpo/beta": 0.0059857675805687904,
      "fcm_dpo/delta": -0.09067746251821518,
      "fcm_dpo/margin": 81.25294494628906,
      "fcm_dpo/q_t": 0.3871810734272003,
      "grad_norm": 12.395666122436523,
      "learning_rate": 3.140277830901428e-09,
      "logits/chosen": 0.7373045682907104,
      "logits/rejected": 0.7179843187332153,
      "logps/chosen": -178.50059509277344,
      "logps/ref_chosen": -58.786048889160156,
      "logps/ref_rejected": -67.21923828125,
      "logps/rejected": -268.1867370605469,
      "loss": 1.0324,
      "margin_dpo/margin_mean": 81.25294494628906,
      "margin_dpo/margin_std": 93.44187927246094,
      "step": 632
    },
    {
      "KL/chosen_KL_mean": -112.61619567871094,
      "KL/mean": -143.66635131835938,
      "KL/rejected_KL_mean": -174.71653747558594,
      "KL/std": 80.20565032958984,
      "epoch": 0.9569160997732427,
      "fcm_dpo/beta": 0.005983942188322544,
      "fcm_dpo/delta": 0.029293209314346313,
      "fcm_dpo/margin": 62.1003303527832,
      "fcm_dpo/q_t": 0.41520097851753235,
      "grad_norm": 12.656014442443848,
      "learning_rate": 2.9348189350335007e-09,
      "logits/chosen": 0.7250348329544067,
      "logits/rejected": 0.6618514060974121,
      "logps/chosen": -164.74639892578125,
      "logps/ref_chosen": -52.13019561767578,
      "logps/ref_rejected": -67.23016357421875,
      "logps/rejected": -241.94668579101562,
      "loss": 1.1224,
      "margin_dpo/margin_mean": 62.1003303527832,
      "margin_dpo/margin_std": 92.83853149414062,
      "step": 633
    },
    {
      "KL/chosen_KL_mean": -141.18545532226562,
      "KL/mean": -153.7686767578125,
      "KL/rejected_KL_mean": -166.3519287109375,
      "KL/std": 83.00604248046875,
      "epoch": 0.9584278155706727,
      "fcm_dpo/beta": 0.006214272230863571,
      "fcm_dpo/delta": 0.24698540568351746,
      "fcm_dpo/margin": 25.166465759277344,
      "fcm_dpo/q_t": 0.46551811695098877,
      "grad_norm": 15.885687828063965,
      "learning_rate": 2.736270983384276e-09,
      "logits/chosen": 0.8349080085754395,
      "logits/rejected": 0.8475719690322876,
      "logps/chosen": -202.16525268554688,
      "logps/ref_chosen": -60.97979736328125,
      "logps/ref_rejected": -58.50825119018555,
      "logps/rejected": -224.86016845703125,
      "loss": 1.3209,
      "margin_dpo/margin_mean": 25.166465759277344,
      "margin_dpo/margin_std": 95.85585021972656,
      "step": 634
    },
    {
      "KL/chosen_KL_mean": -152.8594970703125,
      "KL/mean": -177.40724182128906,
      "KL/rejected_KL_mean": -201.95498657226562,
      "KL/std": 83.1712646484375,
      "epoch": 0.9599395313681028,
      "fcm_dpo/beta": 0.006378865800797939,
      "fcm_dpo/delta": 0.08934434503316879,
      "fcm_dpo/margin": 49.09549331665039,
      "fcm_dpo/q_t": 0.42987093329429626,
      "grad_norm": 14.362222671508789,
      "learning_rate": 2.5446395297668287e-09,
      "logits/chosen": 0.6214234828948975,
      "logits/rejected": 0.5565516948699951,
      "logps/chosen": -218.83251953125,
      "logps/ref_chosen": -65.9730224609375,
      "logps/ref_rejected": -85.61317443847656,
      "logps/rejected": -287.5681457519531,
      "loss": 1.2007,
      "margin_dpo/margin_mean": 49.095489501953125,
      "margin_dpo/margin_std": 103.11039733886719,
      "step": 635
    },
    {
      "KL/chosen_KL_mean": -120.62947082519531,
      "KL/mean": -154.38397216796875,
      "KL/rejected_KL_mean": -188.1384735107422,
      "KL/std": 84.78205871582031,
      "epoch": 0.9614512471655329,
      "fcm_dpo/beta": 0.006347968243062496,
      "fcm_dpo/delta": -0.030059967190027237,
      "fcm_dpo/margin": 67.50898742675781,
      "fcm_dpo/q_t": 0.4006476104259491,
      "grad_norm": 11.630902290344238,
      "learning_rate": 2.359929934524829e-09,
      "logits/chosen": 0.7302621006965637,
      "logits/rejected": 0.6309200525283813,
      "logps/chosen": -169.76963806152344,
      "logps/ref_chosen": -49.140167236328125,
      "logps/ref_rejected": -81.26971435546875,
      "logps/rejected": -269.4081726074219,
      "loss": 1.08,
      "margin_dpo/margin_mean": 67.50898742675781,
      "margin_dpo/margin_std": 89.56611633300781,
      "step": 636
    },
    {
      "KL/chosen_KL_mean": -157.49581909179688,
      "KL/mean": -182.8763427734375,
      "KL/rejected_KL_mean": -208.25686645507812,
      "KL/std": 80.46435546875,
      "epoch": 0.9629629629629629,
      "fcm_dpo/beta": 0.006448294967412949,
      "fcm_dpo/delta": 0.07466259598731995,
      "fcm_dpo/margin": 50.76104736328125,
      "fcm_dpo/q_t": 0.42705270648002625,
      "grad_norm": 15.661286354064941,
      "learning_rate": 2.1821473643827137e-09,
      "logits/chosen": 0.7203917503356934,
      "logits/rejected": 0.648820161819458,
      "logps/chosen": -231.19241333007812,
      "logps/ref_chosen": -73.69658660888672,
      "logps/ref_rejected": -83.01487731933594,
      "logps/rejected": -291.2717590332031,
      "loss": 1.1893,
      "margin_dpo/margin_mean": 50.76104736328125,
      "margin_dpo/margin_std": 103.28202819824219,
      "step": 637
    },
    {
      "KL/chosen_KL_mean": -138.66412353515625,
      "KL/mean": -169.12185668945312,
      "KL/rejected_KL_mean": -199.57955932617188,
      "KL/std": 83.98857116699219,
      "epoch": 0.9644746787603931,
      "fcm_dpo/beta": 0.006465583108365536,
      "fcm_dpo/delta": 0.0063573624938726425,
      "fcm_dpo/margin": 60.91543960571289,
      "fcm_dpo/q_t": 0.4109139144420624,
      "grad_norm": 14.0470609664917,
      "learning_rate": 2.0112967923011646e-09,
      "logits/chosen": 0.728748083114624,
      "logits/rejected": 0.6802102327346802,
      "logps/chosen": -201.44570922851562,
      "logps/ref_chosen": -62.78158187866211,
      "logps/ref_rejected": -85.40478515625,
      "logps/rejected": -284.9843444824219,
      "loss": 1.1145,
      "margin_dpo/margin_mean": 60.91543960571289,
      "margin_dpo/margin_std": 92.47264862060547,
      "step": 638
    },
    {
      "KL/chosen_KL_mean": -124.04972076416016,
      "KL/mean": -159.39202880859375,
      "KL/rejected_KL_mean": -194.73435974121094,
      "KL/std": 89.03472900390625,
      "epoch": 0.9659863945578231,
      "fcm_dpo/beta": 0.006414017640054226,
      "fcm_dpo/delta": -0.055869266390800476,
      "fcm_dpo/margin": 70.68461608886719,
      "fcm_dpo/q_t": 0.39816033840179443,
      "grad_norm": 14.064215660095215,
      "learning_rate": 1.847382997337943e-09,
      "logits/chosen": 0.7247157096862793,
      "logits/rejected": 0.6204769611358643,
      "logps/chosen": -177.81631469726562,
      "logps/ref_chosen": -53.76658630371094,
      "logps/ref_rejected": -72.30009460449219,
      "logps/rejected": -267.0344543457031,
      "loss": 1.0755,
      "margin_dpo/margin_mean": 70.68461608886719,
      "margin_dpo/margin_std": 98.21234893798828,
      "step": 639
    },
    {
      "KL/chosen_KL_mean": -128.9110870361328,
      "KL/mean": -161.59844970703125,
      "KL/rejected_KL_mean": -194.2858123779297,
      "KL/std": 86.63613891601562,
      "epoch": 0.9674981103552532,
      "fcm_dpo/beta": 0.00635608471930027,
      "fcm_dpo/delta": -0.01637229695916176,
      "fcm_dpo/margin": 65.37472534179688,
      "fcm_dpo/q_t": 0.40459680557250977,
      "grad_norm": 13.13107967376709,
      "learning_rate": 1.690410564514244e-09,
      "logits/chosen": 0.7759475708007812,
      "logits/rejected": 0.7121703624725342,
      "logps/chosen": -180.328857421875,
      "logps/ref_chosen": -51.41777801513672,
      "logps/ref_rejected": -77.27879333496094,
      "logps/rejected": -271.5646057128906,
      "loss": 1.099,
      "margin_dpo/margin_mean": 65.37472534179688,
      "margin_dpo/margin_std": 94.90809631347656,
      "step": 640
    },
    {
      "KL/chosen_KL_mean": -136.29029846191406,
      "KL/mean": -165.4718017578125,
      "KL/rejected_KL_mean": -194.65328979492188,
      "KL/std": 82.64945983886719,
      "epoch": 0.9690098261526833,
      "fcm_dpo/beta": 0.006381419487297535,
      "fcm_dpo/delta": 0.028604013845324516,
      "fcm_dpo/margin": 58.363006591796875,
      "fcm_dpo/q_t": 0.41331246495246887,
      "grad_norm": 13.78744125366211,
      "learning_rate": 1.5403838846864692e-09,
      "logits/chosen": 0.711683452129364,
      "logits/rejected": 0.6908361911773682,
      "logps/chosen": -207.344970703125,
      "logps/ref_chosen": -71.0546646118164,
      "logps/ref_rejected": -82.2440185546875,
      "logps/rejected": -276.8973083496094,
      "loss": 1.109,
      "margin_dpo/margin_mean": 58.363006591796875,
      "margin_dpo/margin_std": 79.44768524169922,
      "step": 641
    },
    {
      "KL/chosen_KL_mean": -146.58837890625,
      "KL/mean": -167.76840209960938,
      "KL/rejected_KL_mean": -188.9484100341797,
      "KL/std": 79.4186782836914,
      "epoch": 0.9705215419501134,
      "fcm_dpo/beta": 0.00648085493594408,
      "fcm_dpo/delta": 0.12887108325958252,
      "fcm_dpo/margin": 42.36002731323242,
      "fcm_dpo/q_t": 0.43818730115890503,
      "grad_norm": 16.881349563598633,
      "learning_rate": 1.3973071544233218e-09,
      "logits/chosen": 0.7018405795097351,
      "logits/rejected": 0.7198396921157837,
      "logps/chosen": -215.51766967773438,
      "logps/ref_chosen": -68.92927551269531,
      "logps/ref_rejected": -70.85682678222656,
      "logps/rejected": -259.80523681640625,
      "loss": 1.23,
      "margin_dpo/margin_mean": 42.36002731323242,
      "margin_dpo/margin_std": 95.72892761230469,
      "step": 642
    },
    {
      "KL/chosen_KL_mean": -130.13673400878906,
      "KL/mean": -163.25579833984375,
      "KL/rejected_KL_mean": -196.3748779296875,
      "KL/std": 84.13433837890625,
      "epoch": 0.9720332577475435,
      "fcm_dpo/beta": 0.0065269810147583485,
      "fcm_dpo/delta": -0.03390323370695114,
      "fcm_dpo/margin": 66.23814392089844,
      "fcm_dpo/q_t": 0.40193212032318115,
      "grad_norm": 20.258853912353516,
      "learning_rate": 1.261184375888541e-09,
      "logits/chosen": 0.6894593238830566,
      "logits/rejected": 0.5981128215789795,
      "logps/chosen": -195.44577026367188,
      "logps/ref_chosen": -65.30903625488281,
      "logps/ref_rejected": -83.61613464355469,
      "logps/rejected": -279.99102783203125,
      "loss": 1.0996,
      "margin_dpo/margin_mean": 66.23814392089844,
      "margin_dpo/margin_std": 99.40937805175781,
      "step": 643
    },
    {
      "KL/chosen_KL_mean": -121.13558959960938,
      "KL/mean": -146.59364318847656,
      "KL/rejected_KL_mean": -172.0517120361328,
      "KL/std": 79.79216003417969,
      "epoch": 0.9735449735449735,
      "fcm_dpo/beta": 0.006630894262343645,
      "fcm_dpo/delta": 0.06329631060361862,
      "fcm_dpo/margin": 50.91610336303711,
      "fcm_dpo/q_t": 0.4230857491493225,
      "grad_norm": 13.730220794677734,
      "learning_rate": 1.1320193567288527e-09,
      "logits/chosen": 0.8315029144287109,
      "logits/rejected": 0.7990458011627197,
      "logps/chosen": -172.13818359375,
      "logps/ref_chosen": -51.002601623535156,
      "logps/ref_rejected": -64.46372985839844,
      "logps/rejected": -236.51544189453125,
      "loss": 1.1862,
      "margin_dpo/margin_mean": 50.916099548339844,
      "margin_dpo/margin_std": 100.9294204711914,
      "step": 644
    },
    {
      "KL/chosen_KL_mean": -130.87391662597656,
      "KL/mean": -161.52090454101562,
      "KL/rejected_KL_mean": -192.16786193847656,
      "KL/std": 83.88922119140625,
      "epoch": 0.9750566893424036,
      "fcm_dpo/beta": 0.006598624866455793,
      "fcm_dpo/delta": -0.004665223415941,
      "fcm_dpo/margin": 61.29396057128906,
      "fcm_dpo/q_t": 0.406668096780777,
      "grad_norm": 14.988168716430664,
      "learning_rate": 1.0098157099674987e-09,
      "logits/chosen": 0.6676120758056641,
      "logits/rejected": 0.6508312225341797,
      "logps/chosen": -191.8373260498047,
      "logps/ref_chosen": -60.963409423828125,
      "logps/ref_rejected": -69.73353576660156,
      "logps/rejected": -261.9013977050781,
      "loss": 1.0955,
      "margin_dpo/margin_mean": 61.29396057128906,
      "margin_dpo/margin_std": 84.65945434570312,
      "step": 645
    },
    {
      "KL/chosen_KL_mean": -144.24465942382812,
      "KL/mean": -170.72772216796875,
      "KL/rejected_KL_mean": -197.21078491210938,
      "KL/std": 90.64221954345703,
      "epoch": 0.9765684051398337,
      "fcm_dpo/beta": 0.006654680240899324,
      "fcm_dpo/delta": 0.0492386557161808,
      "fcm_dpo/margin": 52.96611022949219,
      "fcm_dpo/q_t": 0.4206312298774719,
      "grad_norm": 13.949078559875488,
      "learning_rate": 8.945768539031783e-10,
      "logits/chosen": 0.7426609992980957,
      "logits/rejected": 0.6854439377784729,
      "logps/chosen": -206.53472900390625,
      "logps/ref_chosen": -62.290069580078125,
      "logps/ref_rejected": -85.54812622070312,
      "logps/rejected": -282.7589111328125,
      "loss": 1.165,
      "margin_dpo/margin_mean": 52.96611022949219,
      "margin_dpo/margin_std": 98.530029296875,
      "step": 646
    },
    {
      "KL/chosen_KL_mean": -140.38922119140625,
      "KL/mean": -183.11294555664062,
      "KL/rejected_KL_mean": -225.836669921875,
      "KL/std": 90.22059631347656,
      "epoch": 0.9780801209372638,
      "fcm_dpo/beta": 0.006521148607134819,
      "fcm_dpo/delta": -0.16621947288513184,
      "fcm_dpo/margin": 85.44746398925781,
      "fcm_dpo/q_t": 0.37209028005599976,
      "grad_norm": 13.996685028076172,
      "learning_rate": 7.863060120144316e-10,
      "logits/chosen": 0.7676196098327637,
      "logits/rejected": 0.668016254901886,
      "logps/chosen": -207.90509033203125,
      "logps/ref_chosen": -67.515869140625,
      "logps/ref_rejected": -101.50871276855469,
      "logps/rejected": -327.34539794921875,
      "loss": 0.9843,
      "margin_dpo/margin_mean": 85.44746398925781,
      "margin_dpo/margin_std": 89.50350952148438,
      "step": 647
    },
    {
      "KL/chosen_KL_mean": -139.56105041503906,
      "KL/mean": -169.70150756835938,
      "KL/rejected_KL_mean": -199.8419189453125,
      "KL/std": 80.48648071289062,
      "epoch": 0.9795918367346939,
      "fcm_dpo/beta": 0.006439392454922199,
      "fcm_dpo/delta": 0.012094831094145775,
      "fcm_dpo/margin": 60.28087615966797,
      "fcm_dpo/q_t": 0.4116312563419342,
      "grad_norm": 14.129268646240234,
      "learning_rate": 6.850062128694045e-10,
      "logits/chosen": 0.7082226276397705,
      "logits/rejected": 0.6414953470230103,
      "logps/chosen": -204.156982421875,
      "logps/ref_chosen": -64.59593963623047,
      "logps/ref_rejected": -83.384033203125,
      "logps/rejected": -283.2259521484375,
      "loss": 1.1374,
      "margin_dpo/margin_mean": 60.28087615966797,
      "margin_dpo/margin_std": 101.12339782714844,
      "step": 648
    },
    {
      "KL/chosen_KL_mean": -131.25221252441406,
      "KL/mean": -162.04953002929688,
      "KL/rejected_KL_mean": -192.84683227539062,
      "KL/std": 83.36114501953125,
      "epoch": 0.981103552532124,
      "fcm_dpo/beta": 0.0064369370229542255,
      "fcm_dpo/delta": 0.0031705833971500397,
      "fcm_dpo/margin": 61.59461212158203,
      "fcm_dpo/q_t": 0.40989232063293457,
      "grad_norm": 18.868831634521484,
      "learning_rate": 5.906802900412788e-10,
      "logits/chosen": 0.7710450887680054,
      "logits/rejected": 0.7096936702728271,
      "logps/chosen": -180.56185913085938,
      "logps/ref_chosen": -49.30964660644531,
      "logps/ref_rejected": -73.73710632324219,
      "logps/rejected": -266.5839538574219,
      "loss": 1.1323,
      "margin_dpo/margin_mean": 61.594608306884766,
      "margin_dpo/margin_std": 101.82295227050781,
      "step": 649
    },
    {
      "KL/chosen_KL_mean": -124.59217834472656,
      "KL/mean": -155.55430603027344,
      "KL/rejected_KL_mean": -186.5164337158203,
      "KL/std": 83.22000122070312,
      "epoch": 0.982615268329554,
      "fcm_dpo/beta": 0.006457272917032242,
      "fcm_dpo/delta": 6.21061772108078e-05,
      "fcm_dpo/margin": 61.924259185791016,
      "fcm_dpo/q_t": 0.4095534086227417,
      "grad_norm": 13.497651100158691,
      "learning_rate": 5.033308820289184e-10,
      "logits/chosen": 0.840035080909729,
      "logits/rejected": 0.769923210144043,
      "logps/chosen": -179.6554412841797,
      "logps/ref_chosen": -55.06325912475586,
      "logps/ref_rejected": -77.39610290527344,
      "logps/rejected": -263.91253662109375,
      "loss": 1.1324,
      "margin_dpo/margin_mean": 61.924259185791016,
      "margin_dpo/margin_std": 103.21102905273438,
      "step": 650
    },
    {
      "KL/chosen_KL_mean": -125.5594253540039,
      "KL/mean": -152.31875610351562,
      "KL/rejected_KL_mean": -179.0780792236328,
      "KL/std": 84.67516326904297,
      "epoch": 0.9841269841269841,
      "fcm_dpo/beta": 0.006532649975270033,
      "fcm_dpo/delta": 0.052041828632354736,
      "fcm_dpo/margin": 53.518653869628906,
      "fcm_dpo/q_t": 0.4217528998851776,
      "grad_norm": 13.061285018920898,
      "learning_rate": 4.2296043218295606e-10,
      "logits/chosen": 0.7976604104042053,
      "logits/rejected": 0.7193989753723145,
      "logps/chosen": -179.62457275390625,
      "logps/ref_chosen": -54.065162658691406,
      "logps/ref_rejected": -77.79080200195312,
      "logps/rejected": -256.868896484375,
      "loss": 1.1507,
      "margin_dpo/margin_mean": 53.518653869628906,
      "margin_dpo/margin_std": 91.1530990600586,
      "step": 651
    },
    {
      "KL/chosen_KL_mean": -147.7435760498047,
      "KL/mean": -173.71324157714844,
      "KL/rejected_KL_mean": -199.68289184570312,
      "KL/std": 87.17472076416016,
      "epoch": 0.9856386999244142,
      "fcm_dpo/beta": 0.0065635861828923225,
      "fcm_dpo/delta": 0.06094294786453247,
      "fcm_dpo/margin": 51.93932342529297,
      "fcm_dpo/q_t": 0.42427873611450195,
      "grad_norm": 15.52949333190918,
      "learning_rate": 3.4957118863768176e-10,
      "logits/chosen": 0.7149733304977417,
      "logits/rejected": 0.6630445718765259,
      "logps/chosen": -211.38388061523438,
      "logps/ref_chosen": -63.64030456542969,
      "logps/ref_rejected": -78.86882019042969,
      "logps/rejected": -278.55169677734375,
      "loss": 1.1832,
      "margin_dpo/margin_mean": 51.939327239990234,
      "margin_dpo/margin_std": 102.12509155273438,
      "step": 652
    },
    {
      "KL/chosen_KL_mean": -132.9952392578125,
      "KL/mean": -164.79788208007812,
      "KL/rejected_KL_mean": -196.60049438476562,
      "KL/std": 83.29473114013672,
      "epoch": 0.9871504157218443,
      "fcm_dpo/beta": 0.006613045930862427,
      "fcm_dpo/delta": -0.021599669009447098,
      "fcm_dpo/margin": 63.60527801513672,
      "fcm_dpo/q_t": 0.4043147563934326,
      "grad_norm": 14.842588424682617,
      "learning_rate": 2.831652042480093e-10,
      "logits/chosen": 0.7406543493270874,
      "logits/rejected": 0.6956373453140259,
      "logps/chosen": -194.66360473632812,
      "logps/ref_chosen": -61.668373107910156,
      "logps/ref_rejected": -73.83012390136719,
      "logps/rejected": -270.4306335449219,
      "loss": 1.1116,
      "margin_dpo/margin_mean": 63.60527801513672,
      "margin_dpo/margin_std": 99.63491821289062,
      "step": 653
    },
    {
      "KL/chosen_KL_mean": -137.654052734375,
      "KL/mean": -163.44244384765625,
      "KL/rejected_KL_mean": -189.2308349609375,
      "KL/std": 92.28964233398438,
      "epoch": 0.9886621315192744,
      "fcm_dpo/beta": 0.006587079726159573,
      "fcm_dpo/delta": 0.061099112033843994,
      "fcm_dpo/margin": 51.5767822265625,
      "fcm_dpo/q_t": 0.42364975810050964,
      "grad_norm": 13.761890411376953,
      "learning_rate": 2.2374433653205016e-10,
      "logits/chosen": 0.7290184497833252,
      "logits/rejected": 0.6216862201690674,
      "logps/chosen": -195.22232055664062,
      "logps/ref_chosen": -57.568267822265625,
      "logps/ref_rejected": -87.74789428710938,
      "logps/rejected": -276.9787292480469,
      "loss": 1.1653,
      "margin_dpo/margin_mean": 51.57678985595703,
      "margin_dpo/margin_std": 88.0830307006836,
      "step": 654
    },
    {
      "KL/chosen_KL_mean": -107.77822875976562,
      "KL/mean": -150.01248168945312,
      "KL/rejected_KL_mean": -192.24676513671875,
      "KL/std": 85.24494171142578,
      "epoch": 0.9901738473167044,
      "fcm_dpo/beta": 0.006456909701228142,
      "fcm_dpo/delta": -0.15699413418769836,
      "fcm_dpo/margin": 84.46852111816406,
      "fcm_dpo/q_t": 0.3735496699810028,
      "grad_norm": 12.970213890075684,
      "learning_rate": 1.7131024761923852e-10,
      "logits/chosen": 0.7261030673980713,
      "logits/rejected": 0.6327559947967529,
      "logps/chosen": -159.92538452148438,
      "logps/ref_chosen": -52.14714813232422,
      "logps/ref_rejected": -80.85014343261719,
      "logps/rejected": -273.0968933105469,
      "loss": 0.9833,
      "margin_dpo/margin_mean": 84.46852111816406,
      "margin_dpo/margin_std": 81.12989044189453,
      "step": 655
    },
    {
      "KL/chosen_KL_mean": -138.53622436523438,
      "KL/mean": -171.3187255859375,
      "KL/rejected_KL_mean": -204.10121154785156,
      "KL/std": 85.09347534179688,
      "epoch": 0.9916855631141346,
      "fcm_dpo/beta": 0.0064401887357234955,
      "fcm_dpo/delta": -0.02324036881327629,
      "fcm_dpo/margin": 65.56498718261719,
      "fcm_dpo/q_t": 0.40316781401634216,
      "grad_norm": 10.821041107177734,
      "learning_rate": 1.2586440420372934e-10,
      "logits/chosen": 0.6753008365631104,
      "logits/rejected": 0.6247135400772095,
      "logps/chosen": -211.79293823242188,
      "logps/ref_chosen": -73.25672912597656,
      "logps/ref_rejected": -85.35127258300781,
      "logps/rejected": -289.4524841308594,
      "loss": 1.0942,
      "margin_dpo/margin_mean": 65.56498718261719,
      "margin_dpo/margin_std": 94.33009338378906,
      "step": 656
    },
    {
      "KL/chosen_KL_mean": -126.80171966552734,
      "KL/mean": -163.933349609375,
      "KL/rejected_KL_mean": -201.06500244140625,
      "KL/std": 94.10905456542969,
      "epoch": 0.9931972789115646,
      "fcm_dpo/beta": 0.006361294537782669,
      "fcm_dpo/delta": -0.0760110542178154,
      "fcm_dpo/margin": 74.26329803466797,
      "fcm_dpo/q_t": 0.39246702194213867,
      "grad_norm": 11.380026817321777,
      "learning_rate": 8.740807750345913e-11,
      "logits/chosen": 0.8557542562484741,
      "logits/rejected": 0.7586342096328735,
      "logps/chosen": -176.52511596679688,
      "logps/ref_chosen": -49.72339630126953,
      "logps/ref_rejected": -75.1568603515625,
      "logps/rejected": -276.22186279296875,
      "loss": 1.0715,
      "margin_dpo/margin_mean": 74.26329040527344,
      "margin_dpo/margin_std": 104.2190933227539,
      "step": 657
    },
    {
      "KL/chosen_KL_mean": -133.42100524902344,
      "KL/mean": -162.8004150390625,
      "KL/rejected_KL_mean": -192.17982482910156,
      "KL/std": 91.8773193359375,
      "epoch": 0.9947089947089947,
      "fcm_dpo/beta": 0.006399291567504406,
      "fcm_dpo/delta": 0.024006865918636322,
      "fcm_dpo/margin": 58.758811950683594,
      "fcm_dpo/q_t": 0.4153459072113037,
      "grad_norm": 12.453079223632812,
      "learning_rate": 5.594234322453539e-11,
      "logits/chosen": 0.7587237358093262,
      "logits/rejected": 0.7135840654373169,
      "logps/chosen": -196.46734619140625,
      "logps/ref_chosen": -63.04634094238281,
      "logps/ref_rejected": -83.44963073730469,
      "logps/rejected": -275.62945556640625,
      "loss": 1.1623,
      "margin_dpo/margin_mean": 58.758811950683594,
      "margin_dpo/margin_std": 109.14431762695312,
      "step": 658
    },
    {
      "KL/chosen_KL_mean": -138.11993408203125,
      "KL/mean": -162.07423400878906,
      "KL/rejected_KL_mean": -186.02853393554688,
      "KL/std": 86.01627349853516,
      "epoch": 0.9962207105064248,
      "fcm_dpo/beta": 0.006414198316633701,
      "fcm_dpo/delta": 0.09530773013830185,
      "fcm_dpo/margin": 47.90861129760742,
      "fcm_dpo/q_t": 0.43013453483581543,
      "grad_norm": 17.540708541870117,
      "learning_rate": 3.146808153123293e-11,
      "logits/chosen": 0.8054675459861755,
      "logits/rejected": 0.7398273944854736,
      "logps/chosen": -193.20013427734375,
      "logps/ref_chosen": -55.0802001953125,
      "logps/ref_rejected": -71.91049194335938,
      "logps/rejected": -257.93902587890625,
      "loss": 1.2052,
      "margin_dpo/margin_mean": 47.90861129760742,
      "margin_dpo/margin_std": 99.61247253417969,
      "step": 659
    },
    {
      "KL/chosen_KL_mean": -130.65850830078125,
      "KL/mean": -166.74920654296875,
      "KL/rejected_KL_mean": -202.83990478515625,
      "KL/std": 88.41796875,
      "epoch": 0.9977324263038548,
      "fcm_dpo/beta": 0.006431464105844498,
      "fcm_dpo/delta": -0.06731559336185455,
      "fcm_dpo/margin": 72.1814193725586,
      "fcm_dpo/q_t": 0.3944994807243347,
      "grad_norm": 13.725423812866211,
      "learning_rate": 1.3985977021235829e-11,
      "logits/chosen": 0.8140730857849121,
      "logits/rejected": 0.7385152578353882,
      "logps/chosen": -185.18441772460938,
      "logps/ref_chosen": -54.525917053222656,
      "logps/ref_rejected": -81.23604583740234,
      "logps/rejected": -284.075927734375,
      "loss": 1.0573,
      "margin_dpo/margin_mean": 72.1814193725586,
      "margin_dpo/margin_std": 93.41641235351562,
      "step": 660
    },
    {
      "KL/chosen_KL_mean": -149.49850463867188,
      "KL/mean": -173.75375366210938,
      "KL/rejected_KL_mean": -198.00900268554688,
      "KL/std": 92.14109802246094,
      "epoch": 0.999244142101285,
      "fcm_dpo/beta": 0.0065015703439712524,
      "fcm_dpo/delta": 0.08707258105278015,
      "fcm_dpo/margin": 48.51050567626953,
      "fcm_dpo/q_t": 0.4291042387485504,
      "grad_norm": 14.858864784240723,
      "learning_rate": 3.4965187065971735e-12,
      "logits/chosen": 0.6810092329978943,
      "logits/rejected": 0.5987504720687866,
      "logps/chosen": -209.87115478515625,
      "logps/ref_chosen": -60.37263870239258,
      "logps/ref_rejected": -77.42874145507812,
      "logps/rejected": -275.437744140625,
      "loss": 1.2089,
      "margin_dpo/margin_mean": 48.51050567626953,
      "margin_dpo/margin_std": 106.95087432861328,
      "step": 661
    },
    {
      "epoch": 0.999244142101285,
      "step": 661,
      "total_flos": 0.0,
      "train_loss": 1.132411924234497,
      "train_runtime": 1654.1278,
      "train_samples_per_second": 25.594,
      "train_steps_per_second": 0.4
    }
  ],
  "logging_steps": 1,
  "max_steps": 661,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 50,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": false,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}