{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.989247311827957,
  "eval_steps": 500,
  "global_step": 1300,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "completion_length": 25.40625,
      "epoch": 0.0030721966205837174,
      "grad_norm": 9.998714447021484,
      "kl": 0.0,
      "learning_rate": 1e-06,
      "loss": -0.0262,
      "reward": 0.34854064881801605,
      "reward_std": 0.30738791823387146,
      "reward_std_all": 0.3601270765066147,
      "rewards-max/forensics_iou_reward": 0.8472031652927399,
      "rewards-max/forensics_strict_edge_reward": 0.011221585165912984,
      "rewards-max/forensics_tiou_reward": 0.7952301800251007,
      "rewards/forensics_iou_reward": 0.20877424627542496,
      "rewards/forensics_strict_edge_reward": 0.0008495347346979543,
      "rewards/forensics_tiou_reward": 0.16708185523748398,
      "step": 1
    },
    {
      "completion_length": 22.953125,
      "epoch": 0.006144393241167435,
      "grad_norm": 5.674647808074951,
      "kl": 0.0004215240478515625,
      "learning_rate": 9.992307692307693e-07,
      "loss": 0.0399,
      "reward": 0.36352086067199707,
      "reward_std": 0.26079244166612625,
      "reward_std_all": 0.3302724212408066,
      "rewards-max/forensics_iou_reward": 0.8569064736366272,
      "rewards-max/forensics_strict_edge_reward": 0.032440694980323315,
      "rewards-max/forensics_tiou_reward": 0.8201793730258942,
      "rewards/forensics_iou_reward": 0.22867927700281143,
      "rewards/forensics_strict_edge_reward": 0.0019462452619336545,
      "rewards/forensics_tiou_reward": 0.20209452509880066,
      "step": 2
    },
    {
      "completion_length": 23.59375,
      "epoch": 0.009216589861751152,
      "grad_norm": 7.2366437911987305,
      "kl": 0.002201080322265625,
      "learning_rate": 9.984615384615384e-07,
      "loss": 0.005,
      "reward": 0.3793722242116928,
      "reward_std": 0.29952429234981537,
      "reward_std_all": 0.34204794466495514,
      "rewards-max/forensics_iou_reward": 0.5651625990867615,
      "rewards-max/forensics_strict_edge_reward": 0.08590756636112928,
      "rewards-max/forensics_tiou_reward": 0.5410027205944061,
      "rewards/forensics_iou_reward": 0.19228937476873398,
      "rewards/forensics_strict_edge_reward": 0.0037238119984976947,
      "rewards/forensics_tiou_reward": 0.15490787476301193,
      "step": 3
    },
    {
      "completion_length": 22.734375,
      "epoch": 0.01228878648233487,
      "grad_norm": 6.951991081237793,
      "kl": 0.0040130615234375,
      "learning_rate": 9.976923076923077e-07,
      "loss": 0.0296,
      "reward": 0.3800809532403946,
      "reward_std": 0.26619958877563477,
      "reward_std_all": 0.32871294021606445,
      "rewards-max/forensics_iou_reward": 0.7383649349212646,
      "rewards-max/forensics_strict_edge_reward": 0.004980368539690971,
      "rewards-max/forensics_tiou_reward": 0.6672729849815369,
      "rewards/forensics_iou_reward": 0.21005283296108246,
      "rewards/forensics_strict_edge_reward": 0.00023377501929644495,
      "rewards/forensics_tiou_reward": 0.16772546619176865,
      "step": 4
    },
    {
      "completion_length": 21.3125,
      "epoch": 0.015360983102918587,
      "grad_norm": 7.519707202911377,
      "kl": 0.03204345703125,
      "learning_rate": 9.969230769230768e-07,
      "loss": -0.136,
      "reward": 0.36412930488586426,
      "reward_std": 0.3081128001213074,
      "reward_std_all": 0.35968051850795746,
      "rewards-max/forensics_iou_reward": 0.8253263831138611,
      "rewards-max/forensics_strict_edge_reward": 0.04566293887910433,
      "rewards-max/forensics_tiou_reward": 0.773260086774826,
      "rewards/forensics_iou_reward": 0.20235224068164825,
      "rewards/forensics_strict_edge_reward": 0.0019222513019485632,
      "rewards/forensics_tiou_reward": 0.16490577161312103,
      "step": 5
    },
    {
      "completion_length": 25.640625,
      "epoch": 0.018433179723502304,
      "grad_norm": 6.704076290130615,
      "kl": 0.02862548828125,
      "learning_rate": 9.961538461538461e-07,
      "loss": -0.0067,
      "reward": 0.4272548109292984,
      "reward_std": 0.24349217116832733,
      "reward_std_all": 0.2952973544597626,
      "rewards-max/forensics_iou_reward": 0.7385699450969696,
      "rewards-max/forensics_strict_edge_reward": 0.20745692681521177,
      "rewards-max/forensics_tiou_reward": 0.6947783827781677,
      "rewards/forensics_iou_reward": 0.2331426739692688,
      "rewards/forensics_strict_edge_reward": 0.00874503079103306,
      "rewards/forensics_tiou_reward": 0.17916946858167648,
      "step": 6
    },
    {
      "completion_length": 24.15625,
      "epoch": 0.021505376344086023,
      "grad_norm": 4.776968955993652,
      "kl": 0.04931640625,
      "learning_rate": 9.953846153846154e-07,
      "loss": -0.1021,
      "reward": 0.404163122177124,
      "reward_std": 0.28148137778043747,
      "reward_std_all": 0.32820095121860504,
      "rewards-max/forensics_iou_reward": 0.5818742513656616,
      "rewards-max/forensics_strict_edge_reward": 0.04513066511026409,
      "rewards-max/forensics_tiou_reward": 0.533827930688858,
      "rewards/forensics_iou_reward": 0.21353518962860107,
      "rewards/forensics_strict_edge_reward": 0.001886434443122198,
      "rewards/forensics_tiou_reward": 0.15956486761569977,
      "step": 7
    },
    {
      "completion_length": 24.15625,
      "epoch": 0.02457757296466974,
      "grad_norm": 5.0625834465026855,
      "kl": 0.0694580078125,
      "learning_rate": 9.946153846153846e-07,
      "loss": -0.0482,
      "reward": 0.3515145778656006,
      "reward_std": 0.2437221184372902,
      "reward_std_all": 0.32863758504390717,
      "rewards-max/forensics_iou_reward": 0.8060979247093201,
      "rewards-max/forensics_strict_edge_reward": 0.018061709124594927,
      "rewards-max/forensics_tiou_reward": 0.7583832740783691,
      "rewards/forensics_iou_reward": 0.19160261005163193,
      "rewards/forensics_strict_edge_reward": 0.000944649102166295,
      "rewards/forensics_tiou_reward": 0.14947060495615005,
      "step": 8
    },
    {
      "completion_length": 23.796875,
      "epoch": 0.027649769585253458,
      "grad_norm": 5.438048839569092,
      "kl": 0.06103515625,
      "learning_rate": 9.938461538461539e-07,
      "loss": 0.0583,
      "reward": 0.4217386692762375,
      "reward_std": 0.2560502886772156,
      "reward_std_all": 0.30506452918052673,
      "rewards-max/forensics_iou_reward": 0.7309165000915527,
      "rewards-max/forensics_strict_edge_reward": 0.16978277917951345,
      "rewards-max/forensics_tiou_reward": 0.6313729286193848,
      "rewards/forensics_iou_reward": 0.24938172847032547,
      "rewards/forensics_strict_edge_reward": 0.007452385674696416,
      "rewards/forensics_tiou_reward": 0.19300135225057602,
      "step": 9
    },
    {
      "completion_length": 24.5625,
      "epoch": 0.030721966205837174,
      "grad_norm": 13.576247215270996,
      "kl": 0.4853515625,
      "learning_rate": 9.93076923076923e-07,
      "loss": 0.1321,
      "reward": 0.38312408328056335,
      "reward_std": 0.26421723514795303,
      "reward_std_all": 0.32720591127872467,
      "rewards-max/forensics_iou_reward": 0.7919995486736298,
      "rewards-max/forensics_strict_edge_reward": 0.022035018977476284,
      "rewards-max/forensics_tiou_reward": 0.690963864326477,
      "rewards/forensics_iou_reward": 0.21645431220531464,
      "rewards/forensics_strict_edge_reward": 0.0014485135097856983,
      "rewards/forensics_tiou_reward": 0.15618479996919632,
      "step": 10
    },
    {
      "completion_length": 25.4375,
      "epoch": 0.03379416282642089,
      "grad_norm": 4.60784387588501,
      "kl": 0.1845703125,
      "learning_rate": 9.923076923076923e-07,
      "loss": 0.0185,
      "reward": 0.5167852640151978,
      "reward_std": 0.2723900377750397,
      "reward_std_all": 0.27168888598680496,
      "rewards-max/forensics_iou_reward": 0.8868924677371979,
      "rewards-max/forensics_strict_edge_reward": 0.07099082693457603,
      "rewards-max/forensics_tiou_reward": 0.834943950176239,
      "rewards/forensics_iou_reward": 0.34276455640792847,
      "rewards/forensics_strict_edge_reward": 0.003053610387723893,
      "rewards/forensics_tiou_reward": 0.27520522475242615,
      "step": 11
    },
    {
      "completion_length": 28.484375,
      "epoch": 0.03686635944700461,
      "grad_norm": 3.7938594818115234,
      "kl": 0.2041015625,
      "learning_rate": 9.915384615384616e-07,
      "loss": 0.1153,
      "reward": 0.37516339123249054,
      "reward_std": 0.26202812045812607,
      "reward_std_all": 0.33373111486434937,
      "rewards-max/forensics_iou_reward": 0.7144283056259155,
      "rewards-max/forensics_strict_edge_reward": 0.04490538500249386,
      "rewards-max/forensics_tiou_reward": 0.6853466928005219,
      "rewards/forensics_iou_reward": 0.20626381784677505,
      "rewards/forensics_strict_edge_reward": 0.0019192470354028046,
      "rewards/forensics_tiou_reward": 0.165696382522583,
      "step": 12
    },
    {
      "completion_length": 25.546875,
      "epoch": 0.039938556067588324,
      "grad_norm": 4.68487548828125,
      "kl": 0.205078125,
      "learning_rate": 9.907692307692307e-07,
      "loss": 0.1166,
      "reward": 0.37449951469898224,
      "reward_std": 0.2809426933526993,
      "reward_std_all": 0.33997488021850586,
      "rewards-max/forensics_iou_reward": 0.7461948394775391,
      "rewards-max/forensics_strict_edge_reward": 0.1881035380065441,
      "rewards-max/forensics_tiou_reward": 0.7174101769924164,
      "rewards/forensics_iou_reward": 0.2182220071554184,
      "rewards/forensics_strict_edge_reward": 0.01544873439706862,
      "rewards/forensics_tiou_reward": 0.16625650227069855,
      "step": 13
    },
    {
      "completion_length": 31.46875,
      "epoch": 0.043010752688172046,
      "grad_norm": 4.3282318115234375,
      "kl": 0.2060546875,
      "learning_rate": 9.9e-07,
      "loss": -0.0839,
      "reward": 0.4780963063240051,
      "reward_std": 0.2659761756658554,
      "reward_std_all": 0.2821805328130722,
      "rewards-max/forensics_iou_reward": 0.7545108795166016,
      "rewards-max/forensics_strict_edge_reward": 0.030957683047745377,
      "rewards-max/forensics_tiou_reward": 0.6707252562046051,
      "rewards/forensics_iou_reward": 0.272745780646801,
      "rewards/forensics_strict_edge_reward": 0.002094523842970375,
      "rewards/forensics_tiou_reward": 0.208356112241745,
      "step": 14
    },
    {
      "completion_length": 26.59375,
      "epoch": 0.04608294930875576,
      "grad_norm": 5.530483245849609,
      "kl": 0.28076171875,
      "learning_rate": 9.892307692307692e-07,
      "loss": 0.1085,
      "reward": 0.415879487991333,
      "reward_std": 0.26944295316934586,
      "reward_std_all": 0.31775547564029694,
      "rewards-max/forensics_iou_reward": 0.7596935629844666,
      "rewards-max/forensics_strict_edge_reward": 0.04622473428025842,
      "rewards-max/forensics_tiou_reward": 0.7004691064357758,
      "rewards/forensics_iou_reward": 0.2414160668849945,
      "rewards/forensics_strict_edge_reward": 0.0019503840194374789,
      "rewards/forensics_tiou_reward": 0.19694123417139053,
      "step": 15
    },
    {
      "completion_length": 25.515625,
      "epoch": 0.04915514592933948,
      "grad_norm": 4.12813663482666,
      "kl": 0.18994140625,
      "learning_rate": 9.884615384615385e-07,
      "loss": 0.2903,
      "reward": 0.35289862751960754,
      "reward_std": 0.19176820665597916,
      "reward_std_all": 0.3059534877538681,
      "rewards-max/forensics_iou_reward": 0.8960985243320465,
      "rewards-max/forensics_strict_edge_reward": 0.20333411917090416,
      "rewards-max/forensics_tiou_reward": 0.8574626743793488,
      "rewards/forensics_iou_reward": 0.19105707108974457,
      "rewards/forensics_strict_edge_reward": 0.008733419235795736,
      "rewards/forensics_tiou_reward": 0.1450788378715515,
      "step": 16
    },
    {
      "completion_length": 28.75,
      "epoch": 0.05222734254992319,
      "grad_norm": 4.073248386383057,
      "kl": 0.22802734375,
      "learning_rate": 9.876923076923076e-07,
      "loss": 0.0678,
      "reward": 0.3583739697933197,
      "reward_std": 0.29430030286312103,
      "reward_std_all": 0.34108804166316986,
      "rewards-max/forensics_iou_reward": 0.8465403616428375,
      "rewards-max/forensics_strict_edge_reward": 0.011595690186368302,
      "rewards-max/forensics_tiou_reward": 0.81321120262146,
      "rewards/forensics_iou_reward": 0.197285495698452,
      "rewards/forensics_strict_edge_reward": 0.0004967353143001674,
      "rewards/forensics_tiou_reward": 0.1516408547759056,
      "step": 17
    },
    {
      "completion_length": 25.609375,
      "epoch": 0.055299539170506916,
      "grad_norm": 3.9082412719726562,
      "kl": 0.19384765625,
      "learning_rate": 9.86923076923077e-07,
      "loss": 0.0429,
      "reward": 0.4690662920475006,
      "reward_std": 0.28888630867004395,
      "reward_std_all": 0.29821908473968506,
      "rewards-max/forensics_iou_reward": 0.7844501435756683,
      "rewards-max/forensics_strict_edge_reward": 0.22709545493125916,
      "rewards-max/forensics_tiou_reward": 0.7714246511459351,
      "rewards/forensics_iou_reward": 0.302011176943779,
      "rewards/forensics_strict_edge_reward": 0.012521807802841067,
      "rewards/forensics_tiou_reward": 0.23131420463323593,
      "step": 18
    },
    {
      "completion_length": 26.1875,
      "epoch": 0.05837173579109063,
      "grad_norm": 3.8344690799713135,
      "kl": 0.1728515625,
      "learning_rate": 9.86153846153846e-07,
      "loss": -0.0338,
      "reward": 0.4206703305244446,
      "reward_std": 0.31416594982147217,
      "reward_std_all": 0.3296797424554825,
      "rewards-max/forensics_iou_reward": 0.796935498714447,
      "rewards-max/forensics_strict_edge_reward": 0.11244974657893181,
      "rewards-max/forensics_tiou_reward": 0.7242712378501892,
      "rewards/forensics_iou_reward": 0.24904713779687881,
      "rewards/forensics_strict_edge_reward": 0.005113692546728998,
      "rewards/forensics_tiou_reward": 0.2029663547873497,
      "step": 19
    },
    {
      "completion_length": 27.953125,
      "epoch": 0.06144393241167435,
      "grad_norm": 4.53551721572876,
      "kl": 0.18408203125,
      "learning_rate": 9.853846153846153e-07,
      "loss": 0.212,
      "reward": 0.439445897936821,
      "reward_std": 0.17775078862905502,
      "reward_std_all": 0.25827303528785706,
      "rewards-max/forensics_iou_reward": 0.9324679374694824,
      "rewards-max/forensics_strict_edge_reward": 0.24377141520380974,
      "rewards-max/forensics_tiou_reward": 0.8916313052177429,
      "rewards/forensics_iou_reward": 0.2781684845685959,
      "rewards/forensics_strict_edge_reward": 0.010885965544730425,
      "rewards/forensics_tiou_reward": 0.22379657626152039,
      "step": 20
    },
    {
      "completion_length": 23.84375,
      "epoch": 0.06451612903225806,
      "grad_norm": 4.913842678070068,
      "kl": 0.224609375,
      "learning_rate": 9.846153846153847e-07,
      "loss": 0.0432,
      "reward": 0.4523569792509079,
      "reward_std": 0.3050268515944481,
      "reward_std_all": 0.3117339313030243,
      "rewards-max/forensics_iou_reward": 0.8614645004272461,
      "rewards-max/forensics_strict_edge_reward": 0.0180649240501225,
      "rewards-max/forensics_tiou_reward": 0.7626045644283295,
      "rewards/forensics_iou_reward": 0.2858768254518509,
      "rewards/forensics_strict_edge_reward": 0.0008175098337233067,
      "rewards/forensics_tiou_reward": 0.2277614176273346,
      "step": 21
    },
    {
      "completion_length": 23.6875,
      "epoch": 0.06758832565284179,
      "grad_norm": 4.227017402648926,
      "kl": 0.18359375,
      "learning_rate": 9.838461538461538e-07,
      "loss": 0.028,
      "reward": 0.47091686725616455,
      "reward_std": 0.19080167263746262,
      "reward_std_all": 0.260659396648407,
      "rewards-max/forensics_iou_reward": 0.7280048429965973,
      "rewards-max/forensics_strict_edge_reward": 0.30557074677199125,
      "rewards-max/forensics_tiou_reward": 0.6239878237247467,
      "rewards/forensics_iou_reward": 0.2867465019226074,
      "rewards/forensics_strict_edge_reward": 0.015081183897564188,
      "rewards/forensics_tiou_reward": 0.22121944278478622,
      "step": 22
    },
    {
      "completion_length": 23.390625,
      "epoch": 0.0706605222734255,
      "grad_norm": 6.1285786628723145,
      "kl": 0.1953125,
      "learning_rate": 9.83076923076923e-07,
      "loss": 0.0782,
      "reward": 0.5507396161556244,
      "reward_std": 0.243982195854187,
      "reward_std_all": 0.24769733846187592,
      "rewards-max/forensics_iou_reward": 0.8999203145503998,
      "rewards-max/forensics_strict_edge_reward": 0.21477015689015388,
      "rewards-max/forensics_tiou_reward": 0.858350545167923,
      "rewards/forensics_iou_reward": 0.36526498198509216,
      "rewards/forensics_strict_edge_reward": 0.01727563899476081,
      "rewards/forensics_tiou_reward": 0.28924982249736786,
      "step": 23
    },
    {
      "completion_length": 24.515625,
      "epoch": 0.07373271889400922,
      "grad_norm": 5.8261308670043945,
      "kl": 0.265625,
      "learning_rate": 9.823076923076922e-07,
      "loss": 0.0747,
      "reward": 0.4071965664625168,
      "reward_std": 0.2435106262564659,
      "reward_std_all": 0.3003208637237549,
      "rewards-max/forensics_iou_reward": 0.7166513502597809,
      "rewards-max/forensics_strict_edge_reward": 0.01093644043430686,
      "rewards-max/forensics_tiou_reward": 0.6743552982807159,
      "rewards/forensics_iou_reward": 0.2132006734609604,
      "rewards/forensics_strict_edge_reward": 0.0005893366469535977,
      "rewards/forensics_tiou_reward": 0.16961126774549484,
      "step": 24
    },
    {
      "completion_length": 23.859375,
      "epoch": 0.07680491551459294,
      "grad_norm": 4.762894153594971,
      "kl": 0.21826171875,
      "learning_rate": 9.815384615384615e-07,
      "loss": 0.0895,
      "reward": 0.39746661484241486,
      "reward_std": 0.24712259322404861,
      "reward_std_all": 0.3229939788579941,
      "rewards-max/forensics_iou_reward": 0.8028397560119629,
      "rewards-max/forensics_strict_edge_reward": 0.30556368827819824,
      "rewards-max/forensics_tiou_reward": 0.7243282794952393,
      "rewards/forensics_iou_reward": 0.2440699264407158,
      "rewards/forensics_strict_edge_reward": 0.014777558855712414,
      "rewards/forensics_tiou_reward": 0.20414910465478897,
      "step": 25
    },
    {
      "completion_length": 24.828125,
      "epoch": 0.07987711213517665,
      "grad_norm": 4.323911190032959,
      "kl": 0.21240234375,
      "learning_rate": 9.807692307692306e-07,
      "loss": 0.0537,
      "reward": 0.4498945474624634,
      "reward_std": 0.2371690422296524,
      "reward_std_all": 0.28268617391586304,
      "rewards-max/forensics_iou_reward": 0.9704712629318237,
      "rewards-max/forensics_strict_edge_reward": 0.326717309653759,
      "rewards-max/forensics_tiou_reward": 0.9565538465976715,
      "rewards/forensics_iou_reward": 0.2779495418071747,
      "rewards/forensics_strict_edge_reward": 0.016537233721464872,
      "rewards/forensics_tiou_reward": 0.2266450673341751,
      "step": 26
    },
    {
      "completion_length": 25.96875,
      "epoch": 0.08294930875576037,
      "grad_norm": 4.12412691116333,
      "kl": 0.22998046875,
      "learning_rate": 9.8e-07,
      "loss": 0.1197,
      "reward": 0.458485409617424,
      "reward_std": 0.27887389063835144,
      "reward_std_all": 0.29962804913520813,
      "rewards-max/forensics_iou_reward": 0.8711645007133484,
      "rewards-max/forensics_strict_edge_reward": 0.2090112827718258,
      "rewards-max/forensics_tiou_reward": 0.8422962725162506,
      "rewards/forensics_iou_reward": 0.28110405802726746,
      "rewards/forensics_strict_edge_reward": 0.011506080976687372,
      "rewards/forensics_tiou_reward": 0.22867287695407867,
      "step": 27
    },
    {
      "completion_length": 25.328125,
      "epoch": 0.08602150537634409,
      "grad_norm": 4.7042694091796875,
      "kl": 0.2822265625,
      "learning_rate": 9.792307692307693e-07,
      "loss": 0.0821,
      "reward": 0.4415077269077301,
      "reward_std": 0.2561313733458519,
      "reward_std_all": 0.30114397406578064,
      "rewards-max/forensics_iou_reward": 0.7714803218841553,
      "rewards-max/forensics_strict_edge_reward": 0.012477468408178538,
      "rewards-max/forensics_tiou_reward": 0.7113443315029144,
      "rewards/forensics_iou_reward": 0.23512758314609528,
      "rewards/forensics_strict_edge_reward": 0.0007169700984377414,
      "rewards/forensics_tiou_reward": 0.18087587505578995,
      "step": 28
    },
    {
      "completion_length": 24.4375,
      "epoch": 0.0890937019969278,
      "grad_norm": 4.724642753601074,
      "kl": 0.228515625,
      "learning_rate": 9.784615384615384e-07,
      "loss": 0.1506,
      "reward": 0.47276708483695984,
      "reward_std": 0.2816336452960968,
      "reward_std_all": 0.28890474140644073,
      "rewards-max/forensics_iou_reward": 0.7834965288639069,
      "rewards-max/forensics_strict_edge_reward": 0.015581437386572361,
      "rewards-max/forensics_tiou_reward": 0.7324159443378448,
      "rewards/forensics_iou_reward": 0.30985089391469955,
      "rewards/forensics_strict_edge_reward": 0.000794359395513311,
      "rewards/forensics_tiou_reward": 0.23950928449630737,
      "step": 29
    },
    {
      "completion_length": 30.125,
      "epoch": 0.09216589861751152,
      "grad_norm": 4.4085869789123535,
      "kl": 0.2626953125,
      "learning_rate": 9.776923076923077e-07,
      "loss": 0.1248,
      "reward": 0.5057809352874756,
      "reward_std": 0.246076337993145,
      "reward_std_all": 0.2615736573934555,
      "rewards-max/forensics_iou_reward": 0.7915416359901428,
      "rewards-max/forensics_strict_edge_reward": 0.13386431394610554,
      "rewards-max/forensics_tiou_reward": 0.7170830070972443,
      "rewards/forensics_iou_reward": 0.3125405013561249,
      "rewards/forensics_strict_edge_reward": 0.00610025244532153,
      "rewards/forensics_tiou_reward": 0.24948184192180634,
      "step": 30
    },
    {
      "completion_length": 25.75,
      "epoch": 0.09523809523809523,
      "grad_norm": 3.7393178939819336,
      "kl": 0.21240234375,
      "learning_rate": 9.769230769230768e-07,
      "loss": 0.0989,
      "reward": 0.4526165872812271,
      "reward_std": 0.24713073670864105,
      "reward_std_all": 0.2895912528038025,
      "rewards-max/forensics_iou_reward": 0.8366653323173523,
      "rewards-max/forensics_strict_edge_reward": 0.007092690524586942,
      "rewards-max/forensics_tiou_reward": 0.7613762319087982,
      "rewards/forensics_iou_reward": 0.26663103699684143,
      "rewards/forensics_strict_edge_reward": 0.00041393822243662726,
      "rewards/forensics_tiou_reward": 0.20442809909582138,
      "step": 31
    },
    {
      "completion_length": 27.375,
      "epoch": 0.09831029185867896,
      "grad_norm": 4.764737129211426,
      "kl": 0.2822265625,
      "learning_rate": 9.761538461538461e-07,
      "loss": 0.1386,
      "reward": 0.5300175547599792,
      "reward_std": 0.22703425586223602,
      "reward_std_all": 0.2403532862663269,
      "rewards-max/forensics_iou_reward": 0.7431769073009491,
      "rewards-max/forensics_strict_edge_reward": 0.09774231584742665,
      "rewards-max/forensics_tiou_reward": 0.6696097552776337,
      "rewards/forensics_iou_reward": 0.3190789073705673,
      "rewards/forensics_strict_edge_reward": 0.01150058934581466,
      "rewards/forensics_tiou_reward": 0.2399614453315735,
      "step": 32
    },
    {
      "completion_length": 29.546875,
      "epoch": 0.10138248847926268,
      "grad_norm": 3.38571834564209,
      "kl": 0.21630859375,
      "learning_rate": 9.753846153846154e-07,
      "loss": 0.0978,
      "reward": 0.5062765032052994,
      "reward_std": 0.2423221319913864,
      "reward_std_all": 0.2561161369085312,
      "rewards-max/forensics_iou_reward": 0.803040474653244,
      "rewards-max/forensics_strict_edge_reward": 0.00011541305866558105,
      "rewards-max/forensics_tiou_reward": 0.7534773647785187,
      "rewards/forensics_iou_reward": 0.29372403025627136,
      "rewards/forensics_strict_edge_reward": 7.558242714367225e-06,
      "rewards/forensics_tiou_reward": 0.2258877083659172,
      "step": 33
    },
    {
      "completion_length": 28.34375,
      "epoch": 0.10445468509984639,
      "grad_norm": 3.8030240535736084,
      "kl": 0.21142578125,
      "learning_rate": 9.746153846153845e-07,
      "loss": 0.2903,
      "reward": 0.5510973036289215,
      "reward_std": 0.23219411075115204,
      "reward_std_all": 0.2349461168050766,
      "rewards-max/forensics_iou_reward": 0.873118668794632,
      "rewards-max/forensics_strict_edge_reward": 0.0766590116545558,
      "rewards-max/forensics_tiou_reward": 0.826191246509552,
      "rewards/forensics_iou_reward": 0.35964052379131317,
      "rewards/forensics_strict_edge_reward": 0.004138136631809175,
      "rewards/forensics_tiou_reward": 0.2778087854385376,
      "step": 34
    },
    {
      "completion_length": 31.453125,
      "epoch": 0.10752688172043011,
      "grad_norm": 3.408804178237915,
      "kl": 0.22900390625,
      "learning_rate": 9.738461538461539e-07,
      "loss": 0.1924,
      "reward": 0.518580973148346,
      "reward_std": 0.23122482746839523,
      "reward_std_all": 0.24190706759691238,
      "rewards-max/forensics_iou_reward": 0.8305939435958862,
      "rewards-max/forensics_strict_edge_reward": 0.3062147945165634,
      "rewards-max/forensics_tiou_reward": 0.802110493183136,
      "rewards/forensics_iou_reward": 0.2968219816684723,
      "rewards/forensics_strict_edge_reward": 0.014259975170716643,
      "rewards/forensics_tiou_reward": 0.2373257577419281,
      "step": 35
    },
    {
      "completion_length": 27.6875,
      "epoch": 0.11059907834101383,
      "grad_norm": 12.170392990112305,
      "kl": 0.50537109375,
      "learning_rate": 9.73076923076923e-07,
      "loss": 0.1309,
      "reward": 0.50149105489254,
      "reward_std": 0.19474755227565765,
      "reward_std_all": 0.2436426505446434,
      "rewards-max/forensics_iou_reward": 0.8409691452980042,
      "rewards-max/forensics_strict_edge_reward": 0.1601042900754237,
      "rewards-max/forensics_tiou_reward": 0.7925682961940765,
      "rewards/forensics_iou_reward": 0.31058959662914276,
      "rewards/forensics_strict_edge_reward": 0.0068506847064355725,
      "rewards/forensics_tiou_reward": 0.24273143708705902,
      "step": 36
    },
    {
      "completion_length": 28.8125,
      "epoch": 0.11367127496159754,
      "grad_norm": 9.608636856079102,
      "kl": 0.46484375,
      "learning_rate": 9.723076923076923e-07,
      "loss": 0.1016,
      "reward": 0.4551932066679001,
      "reward_std": 0.17208798974752426,
      "reward_std_all": 0.25292137265205383,
      "rewards-max/forensics_iou_reward": 0.7981210052967072,
      "rewards-max/forensics_strict_edge_reward": 0.24080939264968038,
      "rewards-max/forensics_tiou_reward": 0.7401867210865021,
      "rewards/forensics_iou_reward": 0.27523064613342285,
      "rewards/forensics_strict_edge_reward": 0.010075872494780924,
      "rewards/forensics_tiou_reward": 0.22107485681772232,
      "step": 37
    },
    {
      "completion_length": 26.9375,
      "epoch": 0.11674347158218126,
      "grad_norm": 4.121974945068359,
      "kl": 0.23681640625,
      "learning_rate": 9.715384615384616e-07,
      "loss": 0.1744,
      "reward": 0.4837314933538437,
      "reward_std": 0.2526382505893707,
      "reward_std_all": 0.2755091190338135,
      "rewards-max/forensics_iou_reward": 0.8624864816665649,
      "rewards-max/forensics_strict_edge_reward": 0.03724253084510565,
      "rewards-max/forensics_tiou_reward": 0.8100908994674683,
      "rewards/forensics_iou_reward": 0.29296939074993134,
      "rewards/forensics_strict_edge_reward": 0.0029773350106552243,
      "rewards/forensics_tiou_reward": 0.2341843843460083,
      "step": 38
    },
    {
      "completion_length": 27.96875,
      "epoch": 0.11981566820276497,
      "grad_norm": 8.158478736877441,
      "kl": 0.39404296875,
      "learning_rate": 9.707692307692307e-07,
      "loss": 0.2072,
      "reward": 0.5429139733314514,
      "reward_std": 0.22736480087041855,
      "reward_std_all": 0.2440131977200508,
      "rewards-max/forensics_iou_reward": 0.8600457310676575,
      "rewards-max/forensics_strict_edge_reward": 0.223982198163867,
      "rewards-max/forensics_tiou_reward": 0.8242935240268707,
      "rewards/forensics_iou_reward": 0.3879259079694748,
      "rewards/forensics_strict_edge_reward": 0.012694530363660306,
      "rewards/forensics_tiou_reward": 0.3210919052362442,
      "step": 39
    },
    {
      "completion_length": 29.875,
      "epoch": 0.1228878648233487,
      "grad_norm": 2.90103816986084,
      "kl": 0.203125,
      "learning_rate": 9.7e-07,
      "loss": 0.1237,
      "reward": 0.5418797731399536,
      "reward_std": 0.188985675573349,
      "reward_std_all": 0.21548254042863846,
      "rewards-max/forensics_iou_reward": 0.878109872341156,
      "rewards-max/forensics_strict_edge_reward": 0.041783928871154785,
      "rewards-max/forensics_tiou_reward": 0.8430311381816864,
      "rewards/forensics_iou_reward": 0.34033530950546265,
      "rewards/forensics_strict_edge_reward": 0.0018751409952528775,
      "rewards/forensics_tiou_reward": 0.2741905152797699,
      "step": 40
    },
    {
      "completion_length": 28.9375,
      "epoch": 0.1259600614439324,
      "grad_norm": 4.270599365234375,
      "kl": 0.21337890625,
      "learning_rate": 9.692307692307691e-07,
      "loss": 0.1376,
      "reward": 0.5145510882139206,
      "reward_std": 0.18283431231975555,
      "reward_std_all": 0.23682819306850433,
      "rewards-max/forensics_iou_reward": 0.7572788596153259,
      "rewards-max/forensics_strict_edge_reward": 0.11465803161263466,
      "rewards-max/forensics_tiou_reward": 0.7272694706916809,
      "rewards/forensics_iou_reward": 0.3166949674487114,
      "rewards/forensics_strict_edge_reward": 0.005705854156985879,
      "rewards/forensics_tiou_reward": 0.2652091383934021,
      "step": 41
    },
    {
      "completion_length": 25.859375,
      "epoch": 0.12903225806451613,
      "grad_norm": 3.4335408210754395,
      "kl": 0.2197265625,
      "learning_rate": 9.684615384615385e-07,
      "loss": 0.2753,
      "reward": 0.510451003909111,
      "reward_std": 0.2396893873810768,
      "reward_std_all": 0.2630775272846222,
      "rewards-max/forensics_iou_reward": 0.8898560404777527,
      "rewards-max/forensics_strict_edge_reward": 0.43792443454731256,
      "rewards-max/forensics_tiou_reward": 0.8394182324409485,
      "rewards/forensics_iou_reward": 0.3563801348209381,
      "rewards/forensics_strict_edge_reward": 0.01877146451442968,
      "rewards/forensics_tiou_reward": 0.29699182510375977,
      "step": 42
    },
    {
      "completion_length": 26.046875,
      "epoch": 0.13210445468509985,
      "grad_norm": 5.883197784423828,
      "kl": 0.1923828125,
      "learning_rate": 9.676923076923078e-07,
      "loss": 0.1291,
      "reward": 0.4755510240793228,
      "reward_std": 0.23135784268379211,
      "reward_std_all": 0.27449931204319,
      "rewards-max/forensics_iou_reward": 0.817508339881897,
      "rewards-max/forensics_strict_edge_reward": 0.12767722376156598,
      "rewards-max/forensics_tiou_reward": 0.7674700915813446,
      "rewards/forensics_iou_reward": 0.2940007448196411,
      "rewards/forensics_strict_edge_reward": 0.005778951400316146,
      "rewards/forensics_tiou_reward": 0.2352100908756256,
      "step": 43
    },
    {
      "completion_length": 25.515625,
      "epoch": 0.13517665130568357,
      "grad_norm": 4.114281177520752,
      "kl": 0.20361328125,
      "learning_rate": 9.669230769230769e-07,
      "loss": 0.1702,
      "reward": 0.5424610376358032,
      "reward_std": 0.25885358452796936,
      "reward_std_all": 0.25258690118789673,
      "rewards-max/forensics_iou_reward": 0.8952913284301758,
      "rewards-max/forensics_strict_edge_reward": 0.09256117418408394,
      "rewards-max/forensics_tiou_reward": 0.8622148036956787,
      "rewards/forensics_iou_reward": 0.36037030816078186,
      "rewards/forensics_strict_edge_reward": 0.006152243819087744,
      "rewards/forensics_tiou_reward": 0.3054259270429611,
      "step": 44
    },
    {
      "completion_length": 25.640625,
      "epoch": 0.1382488479262673,
      "grad_norm": 4.60336971282959,
      "kl": 0.19482421875,
      "learning_rate": 9.66153846153846e-07,
      "loss": 0.1728,
      "reward": 0.5431259274482727,
      "reward_std": 0.2660084515810013,
      "reward_std_all": 0.2633635252714157,
      "rewards-max/forensics_iou_reward": 0.9291431307792664,
      "rewards-max/forensics_strict_edge_reward": 0.05986521951854229,
      "rewards-max/forensics_tiou_reward": 0.8880851566791534,
      "rewards/forensics_iou_reward": 0.39051327109336853,
      "rewards/forensics_strict_edge_reward": 0.0030710132559761405,
      "rewards/forensics_tiou_reward": 0.32334689795970917,
      "step": 45
    },
    {
      "completion_length": 25.0,
      "epoch": 0.141321044546851,
      "grad_norm": 4.346749305725098,
      "kl": 0.21337890625,
      "learning_rate": 9.653846153846153e-07,
      "loss": -0.0498,
      "reward": 0.48744484782218933,
      "reward_std": 0.26174016296863556,
      "reward_std_all": 0.2804311662912369,
      "rewards-max/forensics_iou_reward": 0.8358815312385559,
      "rewards-max/forensics_strict_edge_reward": 0.1650601215660572,
      "rewards-max/forensics_tiou_reward": 0.7581008970737457,
      "rewards/forensics_iou_reward": 0.31047411262989044,
      "rewards/forensics_strict_edge_reward": 0.007274320116266608,
      "rewards/forensics_tiou_reward": 0.24511206150054932,
      "step": 46
    },
    {
      "completion_length": 27.078125,
      "epoch": 0.1443932411674347,
      "grad_norm": 4.892566680908203,
      "kl": 0.21875,
      "learning_rate": 9.646153846153846e-07,
      "loss": 0.093,
      "reward": 0.47427627444267273,
      "reward_std": 0.2797567844390869,
      "reward_std_all": 0.2888718247413635,
      "rewards-max/forensics_iou_reward": 0.86286261677742,
      "rewards-max/forensics_strict_edge_reward": 0.11573276855051517,
      "rewards-max/forensics_tiou_reward": 0.7949372231960297,
      "rewards/forensics_iou_reward": 0.30847562849521637,
      "rewards/forensics_strict_edge_reward": 0.005363430129364133,
      "rewards/forensics_tiou_reward": 0.25182221084833145,
      "step": 47
    },
    {
      "completion_length": 26.546875,
      "epoch": 0.14746543778801843,
      "grad_norm": 3.3581507205963135,
      "kl": 0.19580078125,
      "learning_rate": 9.638461538461537e-07,
      "loss": 0.1537,
      "reward": 0.4830833971500397,
      "reward_std": 0.2702941596508026,
      "reward_std_all": 0.28306227177381516,
      "rewards-max/forensics_iou_reward": 0.8016398549079895,
      "rewards-max/forensics_strict_edge_reward": 0.2570142336189747,
      "rewards-max/forensics_tiou_reward": 0.7424810826778412,
      "rewards/forensics_iou_reward": 0.2902292162179947,
      "rewards/forensics_strict_edge_reward": 0.012490018125390634,
      "rewards/forensics_tiou_reward": 0.23098241537809372,
      "step": 48
    },
    {
      "completion_length": 24.0,
      "epoch": 0.15053763440860216,
      "grad_norm": 4.178219795227051,
      "kl": 0.20458984375,
      "learning_rate": 9.63076923076923e-07,
      "loss": 0.2325,
      "reward": 0.4538935124874115,
      "reward_std": 0.29646943509578705,
      "reward_std_all": 0.30981361865997314,
      "rewards-max/forensics_iou_reward": 0.6718796789646149,
      "rewards-max/forensics_strict_edge_reward": 0.14394530840218067,
      "rewards-max/forensics_tiou_reward": 0.5894531309604645,
      "rewards/forensics_iou_reward": 0.24924372881650925,
      "rewards/forensics_strict_edge_reward": 0.00640818674582988,
      "rewards/forensics_tiou_reward": 0.2073345184326172,
      "step": 49
    },
    {
      "completion_length": 28.171875,
      "epoch": 0.15360983102918588,
      "grad_norm": 3.1438748836517334,
      "kl": 0.18701171875,
      "learning_rate": 9.623076923076922e-07,
      "loss": 0.1327,
      "reward": 0.5286864936351776,
      "reward_std": 0.22880710661411285,
      "reward_std_all": 0.24123717844486237,
      "rewards-max/forensics_iou_reward": 0.9382372796535492,
      "rewards-max/forensics_strict_edge_reward": 0.05414840579032898,
      "rewards-max/forensics_tiou_reward": 0.9146458804607391,
      "rewards/forensics_iou_reward": 0.33914993703365326,
      "rewards/forensics_strict_edge_reward": 0.002512746927095577,
      "rewards/forensics_tiou_reward": 0.28013429045677185,
      "step": 50
    },
    {
      "completion_length": 25.0625,
      "epoch": 0.15668202764976957,
      "grad_norm": 3.9948267936706543,
      "kl": 0.19775390625,
      "learning_rate": 9.615384615384615e-07,
      "loss": 0.0985,
      "reward": 0.48446235060691833,
      "reward_std": 0.17963039875030518,
      "reward_std_all": 0.24236033111810684,
      "rewards-max/forensics_iou_reward": 0.7364826500415802,
      "rewards-max/forensics_strict_edge_reward": 0.01130809729147586,
      "rewards-max/forensics_tiou_reward": 0.6595638394355774,
      "rewards/forensics_iou_reward": 0.2786768674850464,
      "rewards/forensics_strict_edge_reward": 0.0007066812074754125,
      "rewards/forensics_tiou_reward": 0.21069587022066116,
      "step": 51
    },
    {
      "completion_length": 27.828125,
      "epoch": 0.1597542242703533,
      "grad_norm": 5.137042999267578,
      "kl": 0.20654296875,
      "learning_rate": 9.607692307692308e-07,
      "loss": 0.2149,
      "reward": 0.569676399230957,
      "reward_std": 0.20082227885723114,
      "reward_std_all": 0.21535899490118027,
      "rewards-max/forensics_iou_reward": 0.8044843375682831,
      "rewards-max/forensics_strict_edge_reward": 0.328892782330513,
      "rewards-max/forensics_tiou_reward": 0.7422580122947693,
      "rewards/forensics_iou_reward": 0.3578084260225296,
      "rewards/forensics_strict_edge_reward": 0.01382074924185872,
      "rewards/forensics_tiou_reward": 0.28219202160835266,
      "step": 52
    },
    {
      "completion_length": 27.6875,
      "epoch": 0.16282642089093702,
      "grad_norm": 5.267909049987793,
      "kl": 0.21435546875,
      "learning_rate": 9.6e-07,
      "loss": 0.2428,
      "reward": 0.5626764893531799,
      "reward_std": 0.15076205134391785,
      "reward_std_all": 0.19142893701791763,
      "rewards-max/forensics_iou_reward": 0.9053874313831329,
      "rewards-max/forensics_strict_edge_reward": 0.2869211360812187,
      "rewards-max/forensics_tiou_reward": 0.868313729763031,
      "rewards/forensics_iou_reward": 0.3657614439725876,
      "rewards/forensics_strict_edge_reward": 0.012793406378477812,
      "rewards/forensics_tiou_reward": 0.310234934091568,
      "step": 53
    },
    {
      "completion_length": 23.875,
      "epoch": 0.16589861751152074,
      "grad_norm": 4.919012069702148,
      "kl": 0.22021484375,
      "learning_rate": 9.592307692307692e-07,
      "loss": 0.1589,
      "reward": 0.549897313117981,
      "reward_std": 0.27023182809352875,
      "reward_std_all": 0.25853170454502106,
      "rewards-max/forensics_iou_reward": 0.8511215448379517,
      "rewards-max/forensics_strict_edge_reward": 0.027924826834350824,
      "rewards-max/forensics_tiou_reward": 0.7890880703926086,
      "rewards/forensics_iou_reward": 0.3720137029886246,
      "rewards/forensics_strict_edge_reward": 0.001543309772387147,
      "rewards/forensics_tiou_reward": 0.31218235194683075,
      "step": 54
    },
    {
      "completion_length": 31.265625,
      "epoch": 0.16897081413210446,
      "grad_norm": 4.147881507873535,
      "kl": 0.224609375,
      "learning_rate": 9.584615384615383e-07,
      "loss": 0.1819,
      "reward": 0.5429204255342484,
      "reward_std": 0.23377089202404022,
      "reward_std_all": 0.2350546047091484,
      "rewards-max/forensics_iou_reward": 0.7790199220180511,
      "rewards-max/forensics_strict_edge_reward": 0.06839179806411266,
      "rewards-max/forensics_tiou_reward": 0.70164954662323,
      "rewards/forensics_iou_reward": 0.31945639848709106,
      "rewards/forensics_strict_edge_reward": 0.003619754337705672,
      "rewards/forensics_tiou_reward": 0.25548599660396576,
      "step": 55
    },
    {
      "completion_length": 27.078125,
      "epoch": 0.17204301075268819,
      "grad_norm": 3.8615472316741943,
      "kl": 0.23291015625,
      "learning_rate": 9.576923076923077e-07,
      "loss": 0.1204,
      "reward": 0.47881025075912476,
      "reward_std": 0.2438114881515503,
      "reward_std_all": 0.2694903612136841,
      "rewards-max/forensics_iou_reward": 0.8049245178699493,
      "rewards-max/forensics_strict_edge_reward": 0.03350572892668424,
      "rewards-max/forensics_tiou_reward": 0.7301084101200104,
      "rewards/forensics_iou_reward": 0.30920808017253876,
      "rewards/forensics_strict_edge_reward": 0.0018750797780739958,
      "rewards/forensics_tiou_reward": 0.23392269015312195,
      "step": 56
    },
    {
      "completion_length": 25.5,
      "epoch": 0.17511520737327188,
      "grad_norm": 3.758517026901245,
      "kl": 0.19873046875,
      "learning_rate": 9.569230769230768e-07,
      "loss": 0.1709,
      "reward": 0.5187320411205292,
      "reward_std": 0.15151949226856232,
      "reward_std_all": 0.21395964920520782,
      "rewards-max/forensics_iou_reward": 0.8355538547039032,
      "rewards-max/forensics_strict_edge_reward": 0.03873904049396515,
      "rewards-max/forensics_tiou_reward": 0.8096478879451752,
      "rewards/forensics_iou_reward": 0.3267243057489395,
      "rewards/forensics_strict_edge_reward": 0.0025014046696014702,
      "rewards/forensics_tiou_reward": 0.2643943801522255,
      "step": 57
    },
    {
      "completion_length": 26.046875,
      "epoch": 0.1781874039938556,
      "grad_norm": 5.052969932556152,
      "kl": 0.1923828125,
      "learning_rate": 9.56153846153846e-07,
      "loss": 0.2157,
      "reward": 0.5419197678565979,
      "reward_std": 0.24166518449783325,
      "reward_std_all": 0.2433449774980545,
      "rewards-max/forensics_iou_reward": 0.9793746769428253,
      "rewards-max/forensics_strict_edge_reward": 0.1971503752283752,
      "rewards-max/forensics_tiou_reward": 0.9640296101570129,
      "rewards/forensics_iou_reward": 0.3538721650838852,
      "rewards/forensics_strict_edge_reward": 0.01428544957889244,
      "rewards/forensics_tiou_reward": 0.29027658700942993,
      "step": 58
    },
    {
      "completion_length": 27.015625,
      "epoch": 0.18125960061443933,
      "grad_norm": 5.272010326385498,
      "kl": 0.2392578125,
      "learning_rate": 9.553846153846154e-07,
      "loss": 0.0426,
      "reward": 0.43581318855285645,
      "reward_std": 0.2687651589512825,
      "reward_std_all": 0.30896909534931183,
      "rewards-max/forensics_iou_reward": 0.9452104568481445,
      "rewards-max/forensics_strict_edge_reward": 0.1969794470933266,
      "rewards-max/forensics_tiou_reward": 0.9057509005069733,
      "rewards/forensics_iou_reward": 0.26133888959884644,
      "rewards/forensics_strict_edge_reward": 0.008648361246741842,
      "rewards/forensics_tiou_reward": 0.20983487367630005,
      "step": 59
    },
    {
      "completion_length": 30.46875,
      "epoch": 0.18433179723502305,
      "grad_norm": 5.105294227600098,
      "kl": 0.21923828125,
      "learning_rate": 9.546153846153845e-07,
      "loss": 0.1727,
      "reward": 0.4706375151872635,
      "reward_std": 0.2503727748990059,
      "reward_std_all": 0.2811124324798584,
      "rewards-max/forensics_iou_reward": 0.8267965316772461,
      "rewards-max/forensics_strict_edge_reward": 0.04500552825629711,
      "rewards-max/forensics_tiou_reward": 0.7616594135761261,
      "rewards/forensics_iou_reward": 0.2931535914540291,
      "rewards/forensics_strict_edge_reward": 0.003187456983141601,
      "rewards/forensics_tiou_reward": 0.24522975087165833,
      "step": 60
    },
    {
      "completion_length": 31.890625,
      "epoch": 0.18740399385560677,
      "grad_norm": 4.119278907775879,
      "kl": 0.2734375,
      "learning_rate": 9.538461538461538e-07,
      "loss": 0.1658,
      "reward": 0.42558765411376953,
      "reward_std": 0.27067603170871735,
      "reward_std_all": 0.3107048124074936,
      "rewards-max/forensics_iou_reward": 0.6989990770816803,
      "rewards-max/forensics_strict_edge_reward": 0.17750053224153817,
      "rewards-max/forensics_tiou_reward": 0.6627191603183746,
      "rewards/forensics_iou_reward": 0.2433636635541916,
      "rewards/forensics_strict_edge_reward": 0.00842726099290303,
      "rewards/forensics_tiou_reward": 0.1994149535894394,
      "step": 61
    },
    {
      "completion_length": 28.59375,
      "epoch": 0.19047619047619047,
      "grad_norm": 3.539764642715454,
      "kl": 0.27734375,
      "learning_rate": 9.53076923076923e-07,
      "loss": 0.1288,
      "reward": 0.5334112346172333,
      "reward_std": 0.25485628843307495,
      "reward_std_all": 0.2531484365463257,
      "rewards-max/forensics_iou_reward": 0.8559071123600006,
      "rewards-max/forensics_strict_edge_reward": 0.013656849157996476,
      "rewards-max/forensics_tiou_reward": 0.8009251356124878,
      "rewards/forensics_iou_reward": 0.34104420244693756,
      "rewards/forensics_strict_edge_reward": 0.0011041094185202383,
      "rewards/forensics_tiou_reward": 0.2744106948375702,
      "step": 62
    },
    {
      "completion_length": 24.96875,
      "epoch": 0.1935483870967742,
      "grad_norm": 4.899261474609375,
      "kl": 0.232421875,
      "learning_rate": 9.523076923076923e-07,
      "loss": 0.2091,
      "reward": 0.4785761684179306,
      "reward_std": 0.21422485262155533,
      "reward_std_all": 0.2591053247451782,
      "rewards-max/forensics_iou_reward": 0.7060606181621552,
      "rewards-max/forensics_strict_edge_reward": 0.14275673811789602,
      "rewards-max/forensics_tiou_reward": 0.6479683816432953,
      "rewards/forensics_iou_reward": 0.27973103523254395,
      "rewards/forensics_strict_edge_reward": 0.00842179542814847,
      "rewards/forensics_tiou_reward": 0.2144213169813156,
      "step": 63
    },
    {
      "completion_length": 28.25,
      "epoch": 0.1966205837173579,
      "grad_norm": 2.697859287261963,
      "kl": 0.24365234375,
      "learning_rate": 9.515384615384615e-07,
      "loss": 0.1972,
      "reward": 0.5977026224136353,
      "reward_std": 0.21486081928014755,
      "reward_std_all": 0.2069316729903221,
      "rewards-max/forensics_iou_reward": 0.8983542919158936,
      "rewards-max/forensics_strict_edge_reward": 0.10126302251592278,
      "rewards-max/forensics_tiou_reward": 0.8439333438873291,
      "rewards/forensics_iou_reward": 0.4021283835172653,
      "rewards/forensics_strict_edge_reward": 0.010478436248376966,
      "rewards/forensics_tiou_reward": 0.33907996118068695,
      "step": 64
    },
    {
      "completion_length": 28.890625,
      "epoch": 0.19969278033794163,
      "grad_norm": 3.2922935485839844,
      "kl": 0.19384765625,
      "learning_rate": 9.507692307692308e-07,
      "loss": 0.1484,
      "reward": 0.658186137676239,
      "reward_std": 0.15366102010011673,
      "reward_std_all": 0.14857058227062225,
      "rewards-max/forensics_iou_reward": 0.9278949499130249,
      "rewards-max/forensics_strict_edge_reward": 0.010842306655831635,
      "rewards-max/forensics_tiou_reward": 0.9054783582687378,
      "rewards/forensics_iou_reward": 0.46541501581668854,
      "rewards/forensics_strict_edge_reward": 0.0009704139083623886,
      "rewards/forensics_tiou_reward": 0.37495872378349304,
      "step": 65
    },
    {
      "completion_length": 28.78125,
      "epoch": 0.20276497695852536,
      "grad_norm": 4.106528282165527,
      "kl": 0.19580078125,
      "learning_rate": 9.499999999999999e-07,
      "loss": 0.1368,
      "reward": 0.5199043452739716,
      "reward_std": 0.22372879832983017,
      "reward_std_all": 0.2426452487707138,
      "rewards-max/forensics_iou_reward": 0.9447087943553925,
      "rewards-max/forensics_strict_edge_reward": 0.01693317759782076,
      "rewards-max/forensics_tiou_reward": 0.912828803062439,
      "rewards/forensics_iou_reward": 0.34306108951568604,
      "rewards/forensics_strict_edge_reward": 0.0007727183983661234,
      "rewards/forensics_tiou_reward": 0.27681174874305725,
      "step": 66
    },
    {
      "completion_length": 27.765625,
      "epoch": 0.20583717357910905,
      "grad_norm": 3.7860054969787598,
      "kl": 0.279296875,
      "learning_rate": 9.492307692307691e-07,
      "loss": 0.1675,
      "reward": 0.4774613380432129,
      "reward_std": 0.26825499534606934,
      "reward_std_all": 0.28500938415527344,
      "rewards-max/forensics_iou_reward": 0.6838767826557159,
      "rewards-max/forensics_strict_edge_reward": 0.038740134332329035,
      "rewards-max/forensics_tiou_reward": 0.5672252029180527,
      "rewards/forensics_iou_reward": 0.2746768072247505,
      "rewards/forensics_strict_edge_reward": 0.0029393831464403775,
      "rewards/forensics_tiou_reward": 0.21539027988910675,
      "step": 67
    },
    {
      "completion_length": 31.671875,
      "epoch": 0.20890937019969277,
      "grad_norm": 5.297826290130615,
      "kl": 0.23193359375,
      "learning_rate": 9.484615384615384e-07,
      "loss": 0.1231,
      "reward": 0.5081624239683151,
      "reward_std": 0.18132459372282028,
      "reward_std_all": 0.2282658889889717,
      "rewards-max/forensics_iou_reward": 0.8685543537139893,
      "rewards-max/forensics_strict_edge_reward": 0.19488580897450447,
      "rewards-max/forensics_tiou_reward": 0.8134043216705322,
      "rewards/forensics_iou_reward": 0.3313516825437546,
      "rewards/forensics_strict_edge_reward": 0.009228024166077375,
      "rewards/forensics_tiou_reward": 0.2687537968158722,
      "step": 68
    },
    {
      "completion_length": 28.53125,
      "epoch": 0.2119815668202765,
      "grad_norm": 4.469012260437012,
      "kl": 0.2177734375,
      "learning_rate": 9.476923076923077e-07,
      "loss": 0.2268,
      "reward": 0.525408536195755,
      "reward_std": 0.20416267216205597,
      "reward_std_all": 0.23466383665800095,
      "rewards-max/forensics_iou_reward": 0.9035506248474121,
      "rewards-max/forensics_strict_edge_reward": 0.35038137435913086,
      "rewards-max/forensics_tiou_reward": 0.8741714954376221,
      "rewards/forensics_iou_reward": 0.31450314819812775,
      "rewards/forensics_strict_edge_reward": 0.024001482408493757,
      "rewards/forensics_tiou_reward": 0.2604968696832657,
      "step": 69
    },
    {
      "completion_length": 33.40625,
      "epoch": 0.21505376344086022,
      "grad_norm": 5.4001054763793945,
      "kl": 0.23681640625,
      "learning_rate": 9.46923076923077e-07,
      "loss": -0.0207,
      "reward": 0.5451056957244873,
      "reward_std": 0.18199308216571808,
      "reward_std_all": 0.21369104832410812,
      "rewards-max/forensics_iou_reward": 0.7488856315612793,
      "rewards-max/forensics_strict_edge_reward": 0.08962861122563481,
      "rewards-max/forensics_tiou_reward": 0.7152707576751709,
      "rewards/forensics_iou_reward": 0.3143090307712555,
      "rewards/forensics_strict_edge_reward": 0.0038224030286073685,
      "rewards/forensics_tiou_reward": 0.2641558349132538,
      "step": 70
    },
    {
      "completion_length": 28.125,
      "epoch": 0.21812596006144394,
      "grad_norm": 3.17753005027771,
      "kl": 0.1982421875,
      "learning_rate": 9.461538461538461e-07,
      "loss": 0.1665,
      "reward": 0.5099339485168457,
      "reward_std": 0.22034354507923126,
      "reward_std_all": 0.24600455909967422,
      "rewards-max/forensics_iou_reward": 0.9284546077251434,
      "rewards-max/forensics_strict_edge_reward": 0.1355353705585003,
      "rewards-max/forensics_tiou_reward": 0.9104205667972565,
      "rewards/forensics_iou_reward": 0.3381035476922989,
      "rewards/forensics_strict_edge_reward": 0.008155487710610032,
      "rewards/forensics_tiou_reward": 0.27337896823883057,
      "step": 71
    },
    {
      "completion_length": 26.984375,
      "epoch": 0.22119815668202766,
      "grad_norm": 4.026138782501221,
      "kl": 0.2392578125,
      "learning_rate": 9.453846153846153e-07,
      "loss": 0.1106,
      "reward": 0.46006689965724945,
      "reward_std": 0.23610955476760864,
      "reward_std_all": 0.2849913537502289,
      "rewards-max/forensics_iou_reward": 0.9277383685112,
      "rewards-max/forensics_strict_edge_reward": 0.4339800179004669,
      "rewards-max/forensics_tiou_reward": 0.9027507901191711,
      "rewards/forensics_iou_reward": 0.30342474579811096,
      "rewards/forensics_strict_edge_reward": 0.026007014326751232,
      "rewards/forensics_tiou_reward": 0.2546912059187889,
      "step": 72
    },
    {
      "completion_length": 27.875,
      "epoch": 0.22427035330261136,
      "grad_norm": 3.3030519485473633,
      "kl": 0.24560546875,
      "learning_rate": 9.446153846153846e-07,
      "loss": 0.1472,
      "reward": 0.5307677090167999,
      "reward_std": 0.24446120858192444,
      "reward_std_all": 0.25418075919151306,
      "rewards-max/forensics_iou_reward": 0.8999760150909424,
      "rewards-max/forensics_strict_edge_reward": 0.03120942134410143,
      "rewards-max/forensics_tiou_reward": 0.8481034934520721,
      "rewards/forensics_iou_reward": 0.3478623628616333,
      "rewards/forensics_strict_edge_reward": 0.003130662807961926,
      "rewards/forensics_tiou_reward": 0.28895701467990875,
      "step": 73
    },
    {
      "completion_length": 26.28125,
      "epoch": 0.22734254992319508,
      "grad_norm": 5.007814884185791,
      "kl": 0.2275390625,
      "learning_rate": 9.438461538461538e-07,
      "loss": 0.1747,
      "reward": 0.562582477927208,
      "reward_std": 0.20334972441196442,
      "reward_std_all": 0.22551849484443665,
      "rewards-max/forensics_iou_reward": 0.8376660645008087,
      "rewards-max/forensics_strict_edge_reward": 0.1240482572466135,
      "rewards-max/forensics_tiou_reward": 0.78209188580513,
      "rewards/forensics_iou_reward": 0.36202576756477356,
      "rewards/forensics_strict_edge_reward": 0.007275836076587439,
      "rewards/forensics_tiou_reward": 0.30045777559280396,
      "step": 74
    },
    {
      "completion_length": 25.46875,
      "epoch": 0.2304147465437788,
      "grad_norm": 4.450520992279053,
      "kl": 0.208984375,
      "learning_rate": 9.43076923076923e-07,
      "loss": 0.0861,
      "reward": 0.5045589506626129,
      "reward_std": 0.2656550407409668,
      "reward_std_all": 0.2810758650302887,
      "rewards-max/forensics_iou_reward": 0.8980752229690552,
      "rewards-max/forensics_strict_edge_reward": 0.33108527958393097,
      "rewards-max/forensics_tiou_reward": 0.8724885284900665,
      "rewards/forensics_iou_reward": 0.34555576741695404,
      "rewards/forensics_strict_edge_reward": 0.022416285704821348,
      "rewards/forensics_tiou_reward": 0.2848922908306122,
      "step": 75
    },
    {
      "completion_length": 25.484375,
      "epoch": 0.23348694316436253,
      "grad_norm": 3.6695258617401123,
      "kl": 0.1875,
      "learning_rate": 9.423076923076923e-07,
      "loss": 0.1628,
      "reward": 0.48770514130592346,
      "reward_std": 0.24562441557645798,
      "reward_std_all": 0.2832365334033966,
      "rewards-max/forensics_iou_reward": 0.8631973266601562,
      "rewards-max/forensics_strict_edge_reward": 0.06710305158048868,
      "rewards-max/forensics_tiou_reward": 0.8414453268051147,
      "rewards/forensics_iou_reward": 0.33086898922920227,
      "rewards/forensics_strict_edge_reward": 0.005638211732730269,
      "rewards/forensics_tiou_reward": 0.2804080843925476,
      "step": 76
    },
    {
      "completion_length": 25.125,
      "epoch": 0.23655913978494625,
      "grad_norm": 4.318560600280762,
      "kl": 0.1943359375,
      "learning_rate": 9.415384615384615e-07,
      "loss": 0.0064,
      "reward": 0.4568541347980499,
      "reward_std": 0.2539641782641411,
      "reward_std_all": 0.2943035662174225,
      "rewards-max/forensics_iou_reward": 0.822197675704956,
      "rewards-max/forensics_strict_edge_reward": 0.16110089141875505,
      "rewards-max/forensics_tiou_reward": 0.7710871398448944,
      "rewards/forensics_iou_reward": 0.24462337791919708,
      "rewards/forensics_strict_edge_reward": 0.0069385909591801465,
      "rewards/forensics_tiou_reward": 0.19909656792879105,
      "step": 77
    },
    {
      "completion_length": 24.515625,
      "epoch": 0.23963133640552994,
      "grad_norm": 4.223361492156982,
      "kl": 0.2119140625,
      "learning_rate": 9.407692307692308e-07,
      "loss": 0.2206,
      "reward": 0.5021296590566635,
      "reward_std": 0.26244089007377625,
      "reward_std_all": 0.2714625969529152,
      "rewards-max/forensics_iou_reward": 0.8085032403469086,
      "rewards-max/forensics_strict_edge_reward": 0.003490357135888189,
      "rewards-max/forensics_tiou_reward": 0.7159744501113892,
      "rewards/forensics_iou_reward": 0.31615056097507477,
      "rewards/forensics_strict_edge_reward": 0.00021970788657199591,
      "rewards/forensics_tiou_reward": 0.24771136790513992,
      "step": 78
    },
    {
      "completion_length": 24.0625,
      "epoch": 0.24270353302611367,
      "grad_norm": 4.471465110778809,
      "kl": 0.2275390625,
      "learning_rate": 9.399999999999999e-07,
      "loss": 0.0278,
      "reward": 0.5546729862689972,
      "reward_std": 0.21083477139472961,
      "reward_std_all": 0.2269774079322815,
      "rewards-max/forensics_iou_reward": 0.850015789270401,
      "rewards-max/forensics_strict_edge_reward": 0.043757292442023754,
      "rewards-max/forensics_tiou_reward": 0.8035780191421509,
      "rewards/forensics_iou_reward": 0.3578891158103943,
      "rewards/forensics_strict_edge_reward": 0.0036304613895481452,
      "rewards/forensics_tiou_reward": 0.2915349304676056,
      "step": 79
    },
    {
      "completion_length": 24.328125,
      "epoch": 0.2457757296466974,
      "grad_norm": 3.5720362663269043,
      "kl": 0.22607421875,
      "learning_rate": 9.392307692307691e-07,
      "loss": 0.1947,
      "reward": 0.5619150102138519,
      "reward_std": 0.18084337562322617,
      "reward_std_all": 0.20838981866836548,
      "rewards-max/forensics_iou_reward": 0.9056389629840851,
      "rewards-max/forensics_strict_edge_reward": 0.1757798946928233,
      "rewards-max/forensics_tiou_reward": 0.8586551249027252,
      "rewards/forensics_iou_reward": 0.3839392066001892,
      "rewards/forensics_strict_edge_reward": 0.013423366748611443,
      "rewards/forensics_tiou_reward": 0.3220924735069275,
      "step": 80
    },
    {
      "completion_length": 25.484375,
      "epoch": 0.2488479262672811,
      "grad_norm": 4.635062217712402,
      "kl": 0.20166015625,
      "learning_rate": 9.384615384615384e-07,
      "loss": 0.1717,
      "reward": 0.5618854761123657,
      "reward_std": 0.23211556673049927,
      "reward_std_all": 0.23735354840755463,
      "rewards-max/forensics_iou_reward": 0.9602221250534058,
      "rewards-max/forensics_strict_edge_reward": 0.6939628571271896,
      "rewards-max/forensics_tiou_reward": 0.944243460893631,
      "rewards/forensics_iou_reward": 0.4078385829925537,
      "rewards/forensics_strict_edge_reward": 0.03331188764423132,
      "rewards/forensics_tiou_reward": 0.35160472989082336,
      "step": 81
    },
    {
      "completion_length": 28.546875,
      "epoch": 0.2519201228878648,
      "grad_norm": 3.4717533588409424,
      "kl": 0.23193359375,
      "learning_rate": 9.376923076923076e-07,
      "loss": 0.2748,
      "reward": 0.5939584374427795,
      "reward_std": 0.1525558978319168,
      "reward_std_all": 0.17914631962776184,
      "rewards-max/forensics_iou_reward": 0.898864358663559,
      "rewards-max/forensics_strict_edge_reward": 0.010251524858176708,
      "rewards-max/forensics_tiou_reward": 0.848728358745575,
      "rewards/forensics_iou_reward": 0.372407466173172,
      "rewards/forensics_strict_edge_reward": 0.000558264262508601,
      "rewards/forensics_tiou_reward": 0.30308660864830017,
      "step": 82
    },
    {
      "completion_length": 26.109375,
      "epoch": 0.25499231950844853,
      "grad_norm": 3.7874093055725098,
      "kl": 0.15673828125,
      "learning_rate": 9.36923076923077e-07,
      "loss": 0.0818,
      "reward": 0.5298178791999817,
      "reward_std": 0.20325736701488495,
      "reward_std_all": 0.22251388430595398,
      "rewards-max/forensics_iou_reward": 0.8632234036922455,
      "rewards-max/forensics_strict_edge_reward": 0.07732220240768584,
      "rewards-max/forensics_tiou_reward": 0.8249554932117462,
      "rewards/forensics_iou_reward": 0.3245372176170349,
      "rewards/forensics_strict_edge_reward": 0.004023567251451254,
      "rewards/forensics_tiou_reward": 0.2750927731394768,
      "step": 83
    },
    {
      "completion_length": 23.0,
      "epoch": 0.25806451612903225,
      "grad_norm": 5.397904396057129,
      "kl": 0.20751953125,
      "learning_rate": 9.361538461538461e-07,
      "loss": 0.1961,
      "reward": 0.49638062715530396,
      "reward_std": 0.23278315365314484,
      "reward_std_all": 0.26505880802869797,
      "rewards-max/forensics_iou_reward": 0.9621055722236633,
      "rewards-max/forensics_strict_edge_reward": 0.4525264501571655,
      "rewards-max/forensics_tiou_reward": 0.9383745491504669,
      "rewards/forensics_iou_reward": 0.3290621191263199,
      "rewards/forensics_strict_edge_reward": 0.020467102993279696,
      "rewards/forensics_tiou_reward": 0.27139126509428024,
      "step": 84
    },
    {
      "completion_length": 27.546875,
      "epoch": 0.261136712749616,
      "grad_norm": 6.267568588256836,
      "kl": 0.22705078125,
      "learning_rate": 9.353846153846153e-07,
      "loss": 0.2305,
      "reward": 0.5375814437866211,
      "reward_std": 0.1928371712565422,
      "reward_std_all": 0.22907938808202744,
      "rewards-max/forensics_iou_reward": 0.9160169064998627,
      "rewards-max/forensics_strict_edge_reward": 0.16246182844042778,
      "rewards-max/forensics_tiou_reward": 0.9028409719467163,
      "rewards/forensics_iou_reward": 0.34458573162555695,
      "rewards/forensics_strict_edge_reward": 0.007529463153332472,
      "rewards/forensics_tiou_reward": 0.28060542047023773,
      "step": 85
    },
    {
      "completion_length": 28.5,
      "epoch": 0.2642089093701997,
      "grad_norm": 2.8200483322143555,
      "kl": 0.205078125,
      "learning_rate": 9.346153846153846e-07,
      "loss": 0.3667,
      "reward": 0.5835021734237671,
      "reward_std": 0.15511813014745712,
      "reward_std_all": 0.1797860935330391,
      "rewards-max/forensics_iou_reward": 0.8771173655986786,
      "rewards-max/forensics_strict_edge_reward": 0.17412479780614376,
      "rewards-max/forensics_tiou_reward": 0.8521700203418732,
      "rewards/forensics_iou_reward": 0.38479849696159363,
      "rewards/forensics_strict_edge_reward": 0.008268859121017158,
      "rewards/forensics_tiou_reward": 0.30616986751556396,
      "step": 86
    },
    {
      "completion_length": 25.75,
      "epoch": 0.2672811059907834,
      "grad_norm": 5.814639568328857,
      "kl": 0.251953125,
      "learning_rate": 9.338461538461538e-07,
      "loss": 0.1945,
      "reward": 0.5500462651252747,
      "reward_std": 0.19469910115003586,
      "reward_std_all": 0.22208593785762787,
      "rewards-max/forensics_iou_reward": 0.8568697571754456,
      "rewards-max/forensics_strict_edge_reward": 0.14396477490663528,
      "rewards-max/forensics_tiou_reward": 0.7916713058948517,
      "rewards/forensics_iou_reward": 0.378400593996048,
      "rewards/forensics_strict_edge_reward": 0.008167692634742707,
      "rewards/forensics_tiou_reward": 0.31407977640628815,
      "step": 87
    },
    {
      "completion_length": 24.53125,
      "epoch": 0.27035330261136714,
      "grad_norm": 4.112408638000488,
      "kl": 0.22265625,
      "learning_rate": 9.33076923076923e-07,
      "loss": 0.1441,
      "reward": 0.5130638778209686,
      "reward_std": 0.25811371952295303,
      "reward_std_all": 0.2679404765367508,
      "rewards-max/forensics_iou_reward": 0.961439698934555,
      "rewards-max/forensics_strict_edge_reward": 0.14116808772087097,
      "rewards-max/forensics_tiou_reward": 0.95122230052948,
      "rewards/forensics_iou_reward": 0.3699004650115967,
      "rewards/forensics_strict_edge_reward": 0.006402947707101703,
      "rewards/forensics_tiou_reward": 0.32129186391830444,
      "step": 88
    },
    {
      "completion_length": 28.59375,
      "epoch": 0.27342549923195086,
      "grad_norm": 3.20141863822937,
      "kl": 0.2509765625,
      "learning_rate": 9.323076923076922e-07,
      "loss": 0.1164,
      "reward": 0.5905119180679321,
      "reward_std": 0.1819511577486992,
      "reward_std_all": 0.18911761790513992,
      "rewards-max/forensics_iou_reward": 0.859367161989212,
      "rewards-max/forensics_strict_edge_reward": 0.13787908107042313,
      "rewards-max/forensics_tiou_reward": 0.7955712974071503,
      "rewards/forensics_iou_reward": 0.3839772045612335,
      "rewards/forensics_strict_edge_reward": 0.007276904652826488,
      "rewards/forensics_tiou_reward": 0.3196520656347275,
      "step": 89
    },
    {
      "completion_length": 27.4375,
      "epoch": 0.2764976958525346,
      "grad_norm": 4.832082271575928,
      "kl": 0.2314453125,
      "learning_rate": 9.315384615384615e-07,
      "loss": 0.223,
      "reward": 0.5708345770835876,
      "reward_std": 0.18288780748844147,
      "reward_std_all": 0.19931749999523163,
      "rewards-max/forensics_iou_reward": 0.8715134859085083,
      "rewards-max/forensics_strict_edge_reward": 0.25590959936380386,
      "rewards-max/forensics_tiou_reward": 0.8301759958267212,
      "rewards/forensics_iou_reward": 0.3771332949399948,
      "rewards/forensics_strict_edge_reward": 0.01543415430933237,
      "rewards/forensics_tiou_reward": 0.3069350868463516,
      "step": 90
    },
    {
      "completion_length": 25.78125,
      "epoch": 0.27956989247311825,
      "grad_norm": 5.431726455688477,
      "kl": 0.22705078125,
      "learning_rate": 9.307692307692308e-07,
      "loss": 0.1143,
      "reward": 0.6079801321029663,
      "reward_std": 0.15236133337020874,
      "reward_std_all": 0.17376335710287094,
      "rewards-max/forensics_iou_reward": 0.8823750615119934,
      "rewards-max/forensics_strict_edge_reward": 0.20347648482129443,
      "rewards-max/forensics_tiou_reward": 0.8331768214702606,
      "rewards/forensics_iou_reward": 0.40619730949401855,
      "rewards/forensics_strict_edge_reward": 0.016301522891808418,
      "rewards/forensics_tiou_reward": 0.3318796306848526,
      "step": 91
    },
    {
      "completion_length": 27.234375,
      "epoch": 0.282642089093702,
      "grad_norm": 4.5101823806762695,
      "kl": 0.296875,
      "learning_rate": 9.3e-07,
      "loss": 0.1128,
      "reward": 0.5175267159938812,
      "reward_std": 0.18477009236812592,
      "reward_std_all": 0.22844815254211426,
      "rewards-max/forensics_iou_reward": 0.8556123077869415,
      "rewards-max/forensics_strict_edge_reward": 0.15670232335105538,
      "rewards-max/forensics_tiou_reward": 0.7850376069545746,
      "rewards/forensics_iou_reward": 0.3282439857721329,
      "rewards/forensics_strict_edge_reward": 0.006879277585539967,
      "rewards/forensics_tiou_reward": 0.25746526569128036,
      "step": 92
    },
    {
      "completion_length": 29.9375,
      "epoch": 0.2857142857142857,
      "grad_norm": 4.760862350463867,
      "kl": 0.3271484375,
      "learning_rate": 9.292307692307692e-07,
      "loss": 0.094,
      "reward": 0.5390646755695343,
      "reward_std": 0.20608854293823242,
      "reward_std_all": 0.2247534766793251,
      "rewards-max/forensics_iou_reward": 0.9019502699375153,
      "rewards-max/forensics_strict_edge_reward": 0.0275357267819345,
      "rewards-max/forensics_tiou_reward": 0.8588055670261383,
      "rewards/forensics_iou_reward": 0.3388601839542389,
      "rewards/forensics_strict_edge_reward": 0.0012838874536100775,
      "rewards/forensics_tiou_reward": 0.2709425836801529,
      "step": 93
    },
    {
      "completion_length": 25.609375,
      "epoch": 0.2887864823348694,
      "grad_norm": 6.271551132202148,
      "kl": 0.412109375,
      "learning_rate": 9.284615384615384e-07,
      "loss": 0.2024,
      "reward": 0.3925316035747528,
      "reward_std": 0.315187007188797,
      "reward_std_all": 0.3427477329969406,
      "rewards-max/forensics_iou_reward": 0.8210002481937408,
      "rewards-max/forensics_strict_edge_reward": 0.23881348222494125,
      "rewards-max/forensics_tiou_reward": 0.7947779595851898,
      "rewards/forensics_iou_reward": 0.2297389805316925,
      "rewards/forensics_strict_edge_reward": 0.013526901137083769,
      "rewards/forensics_tiou_reward": 0.191255584359169,
      "step": 94
    },
    {
      "completion_length": 28.421875,
      "epoch": 0.29185867895545314,
      "grad_norm": 5.500439167022705,
      "kl": 0.361328125,
      "learning_rate": 9.276923076923076e-07,
      "loss": 0.2315,
      "reward": 0.5015597343444824,
      "reward_std": 0.20880769193172455,
      "reward_std_all": 0.2628912329673767,
      "rewards-max/forensics_iou_reward": 0.9092299342155457,
      "rewards-max/forensics_strict_edge_reward": 0.15809179842472076,
      "rewards-max/forensics_tiou_reward": 0.8351675271987915,
      "rewards/forensics_iou_reward": 0.3297397047281265,
      "rewards/forensics_strict_edge_reward": 0.007745612878352404,
      "rewards/forensics_tiou_reward": 0.25372840464115143,
      "step": 95
    },
    {
      "completion_length": 28.515625,
      "epoch": 0.29493087557603687,
      "grad_norm": 3.761413812637329,
      "kl": 0.2724609375,
      "learning_rate": 9.26923076923077e-07,
      "loss": 0.1749,
      "reward": 0.5126199275255203,
      "reward_std": 0.1928786039352417,
      "reward_std_all": 0.23399817943572998,
      "rewards-max/forensics_iou_reward": 0.8843249082565308,
      "rewards-max/forensics_strict_edge_reward": 0.07614180818200111,
      "rewards-max/forensics_tiou_reward": 0.8398936688899994,
      "rewards/forensics_iou_reward": 0.33264797925949097,
      "rewards/forensics_strict_edge_reward": 0.007560073136119172,
      "rewards/forensics_tiou_reward": 0.2695017457008362,
      "step": 96
    },
    {
      "completion_length": 25.03125,
      "epoch": 0.2980030721966206,
      "grad_norm": 5.087399005889893,
      "kl": 0.248046875,
      "learning_rate": 9.261538461538461e-07,
      "loss": 0.1488,
      "reward": 0.5674364864826202,
      "reward_std": 0.11812364682555199,
      "reward_std_all": 0.17198893427848816,
      "rewards-max/forensics_iou_reward": 0.9666725099086761,
      "rewards-max/forensics_strict_edge_reward": 0.5352254435420036,
      "rewards-max/forensics_tiou_reward": 0.9542587995529175,
      "rewards/forensics_iou_reward": 0.38814133405685425,
      "rewards/forensics_strict_edge_reward": 0.0344904107041657,
      "rewards/forensics_tiou_reward": 0.32155296206474304,
      "step": 97
    },
    {
      "completion_length": 26.875,
      "epoch": 0.3010752688172043,
      "grad_norm": 3.9159023761749268,
      "kl": 0.2685546875,
      "learning_rate": 9.253846153846153e-07,
      "loss": 0.2047,
      "reward": 0.5584922432899475,
      "reward_std": 0.19089138507843018,
      "reward_std_all": 0.2115672528743744,
      "rewards-max/forensics_iou_reward": 0.8487966954708099,
      "rewards-max/forensics_strict_edge_reward": 0.2009410122409463,
      "rewards-max/forensics_tiou_reward": 0.7941075563430786,
      "rewards/forensics_iou_reward": 0.35445037484169006,
      "rewards/forensics_strict_edge_reward": 0.009403450996614993,
      "rewards/forensics_tiou_reward": 0.2976190075278282,
      "step": 98
    },
    {
      "completion_length": 27.90625,
      "epoch": 0.30414746543778803,
      "grad_norm": 5.769794464111328,
      "kl": 0.25,
      "learning_rate": 9.246153846153846e-07,
      "loss": 0.0495,
      "reward": 0.5296118557453156,
      "reward_std": 0.1699654757976532,
      "reward_std_all": 0.21215920895338058,
      "rewards-max/forensics_iou_reward": 0.8332633674144745,
      "rewards-max/forensics_strict_edge_reward": 0.4383902076515369,
      "rewards-max/forensics_tiou_reward": 0.7956685721874237,
      "rewards/forensics_iou_reward": 0.3392191529273987,
      "rewards/forensics_strict_edge_reward": 0.019036024208617164,
      "rewards/forensics_tiou_reward": 0.2798793241381645,
      "step": 99
    },
    {
      "completion_length": 29.4375,
      "epoch": 0.30721966205837176,
      "grad_norm": 4.479531288146973,
      "kl": 0.21337890625,
      "learning_rate": 9.238461538461538e-07,
      "loss": 0.095,
      "reward": 0.5658356100320816,
      "reward_std": 0.22718890011310577,
      "reward_std_all": 0.22391611337661743,
      "rewards-max/forensics_iou_reward": 0.9576822221279144,
      "rewards-max/forensics_strict_edge_reward": 0.29233212023973465,
      "rewards-max/forensics_tiou_reward": 0.9397736489772797,
      "rewards/forensics_iou_reward": 0.38603727519512177,
      "rewards/forensics_strict_edge_reward": 0.02469782531261444,
      "rewards/forensics_tiou_reward": 0.3194210082292557,
      "step": 100
    },
    {
      "completion_length": 24.765625,
      "epoch": 0.3102918586789555,
      "grad_norm": 5.522209644317627,
      "kl": 0.22412109375,
      "learning_rate": 9.230769230769231e-07,
      "loss": 0.157,
      "reward": 0.4672963172197342,
      "reward_std": 0.2866082265973091,
      "reward_std_all": 0.2980700880289078,
      "rewards-max/forensics_iou_reward": 0.8127003908157349,
      "rewards-max/forensics_strict_edge_reward": 0.16754169762134552,
      "rewards-max/forensics_tiou_reward": 0.7595911920070648,
      "rewards/forensics_iou_reward": 0.28936517238616943,
      "rewards/forensics_strict_edge_reward": 0.015187709592282772,
      "rewards/forensics_tiou_reward": 0.2554764747619629,
      "step": 101
    },
    {
      "completion_length": 26.625,
      "epoch": 0.31336405529953915,
      "grad_norm": 4.4775919914245605,
      "kl": 0.2294921875,
      "learning_rate": 9.223076923076922e-07,
      "loss": 0.1327,
      "reward": 0.4669466018676758,
      "reward_std": 0.23243586719036102,
      "reward_std_all": 0.26640674471855164,
      "rewards-max/forensics_iou_reward": 0.9503022134304047,
      "rewards-max/forensics_strict_edge_reward": 0.06728352466598153,
      "rewards-max/forensics_tiou_reward": 0.921110987663269,
      "rewards/forensics_iou_reward": 0.3125951290130615,
      "rewards/forensics_strict_edge_reward": 0.0034423606557538733,
      "rewards/forensics_tiou_reward": 0.27684134244918823,
      "step": 102
    },
    {
      "completion_length": 29.0,
      "epoch": 0.31643625192012287,
      "grad_norm": 4.121120929718018,
      "kl": 0.26513671875,
      "learning_rate": 9.215384615384614e-07,
      "loss": 0.1365,
      "reward": 0.4731803685426712,
      "reward_std": 0.2572288513183594,
      "reward_std_all": 0.27892565727233887,
      "rewards-max/forensics_iou_reward": 0.8065012395381927,
      "rewards-max/forensics_strict_edge_reward": 0.03551071509718895,
      "rewards-max/forensics_tiou_reward": 0.7651706635951996,
      "rewards/forensics_iou_reward": 0.28166620433330536,
      "rewards/forensics_strict_edge_reward": 0.0020412077428773046,
      "rewards/forensics_tiou_reward": 0.2241889387369156,
      "step": 103
    },
    {
      "completion_length": 24.734375,
      "epoch": 0.3195084485407066,
      "grad_norm": 4.57826566696167,
      "kl": 0.283203125,
      "learning_rate": 9.207692307692308e-07,
      "loss": 0.0774,
      "reward": 0.4659877419471741,
      "reward_std": 0.2373245656490326,
      "reward_std_all": 0.28088152408599854,
      "rewards-max/forensics_iou_reward": 0.8627562522888184,
      "rewards-max/forensics_strict_edge_reward": 0.006884799106046557,
      "rewards-max/forensics_tiou_reward": 0.7522123157978058,
      "rewards/forensics_iou_reward": 0.2984803766012192,
      "rewards/forensics_strict_edge_reward": 0.0003146558810840361,
      "rewards/forensics_tiou_reward": 0.2396874949336052,
      "step": 104
    },
    {
      "completion_length": 25.328125,
      "epoch": 0.3225806451612903,
      "grad_norm": 3.748857021331787,
      "kl": 0.2216796875,
      "learning_rate": 9.2e-07,
      "loss": 0.0148,
      "reward": 0.5839268565177917,
      "reward_std": 0.17850692570209503,
      "reward_std_all": 0.19675269722938538,
      "rewards-max/forensics_iou_reward": 0.9254024624824524,
      "rewards-max/forensics_strict_edge_reward": 0.2810596376657486,
      "rewards-max/forensics_tiou_reward": 0.9016701579093933,
      "rewards/forensics_iou_reward": 0.39126063883304596,
      "rewards/forensics_strict_edge_reward": 0.01626133732497692,
      "rewards/forensics_tiou_reward": 0.31557102501392365,
      "step": 105
    },
    {
      "completion_length": 28.046875,
      "epoch": 0.32565284178187404,
      "grad_norm": 3.728090286254883,
      "kl": 0.23974609375,
      "learning_rate": 9.192307692307692e-07,
      "loss": 0.0547,
      "reward": 0.5568867027759552,
      "reward_std": 0.20103032886981964,
      "reward_std_all": 0.21896136552095413,
      "rewards-max/forensics_iou_reward": 0.9161148071289062,
      "rewards-max/forensics_strict_edge_reward": 0.23994747549295425,
      "rewards-max/forensics_tiou_reward": 0.8879697024822235,
      "rewards/forensics_iou_reward": 0.3666999489068985,
      "rewards/forensics_strict_edge_reward": 0.013254744466394186,
      "rewards/forensics_tiou_reward": 0.30186179280281067,
      "step": 106
    },
    {
      "completion_length": 26.21875,
      "epoch": 0.32872503840245776,
      "grad_norm": 4.4950175285339355,
      "kl": 0.24560546875,
      "learning_rate": 9.184615384615384e-07,
      "loss": 0.0606,
      "reward": 0.541539877653122,
      "reward_std": 0.2207389920949936,
      "reward_std_all": 0.23703670501708984,
      "rewards-max/forensics_iou_reward": 0.9347496628761292,
      "rewards-max/forensics_strict_edge_reward": 0.3774634413421154,
      "rewards-max/forensics_tiou_reward": 0.9132851958274841,
      "rewards/forensics_iou_reward": 0.39182932674884796,
      "rewards/forensics_strict_edge_reward": 0.017170133301988244,
      "rewards/forensics_tiou_reward": 0.33384668827056885,
      "step": 107
    },
    {
      "completion_length": 24.734375,
      "epoch": 0.3317972350230415,
      "grad_norm": 4.850794315338135,
      "kl": 0.2158203125,
      "learning_rate": 9.176923076923076e-07,
      "loss": 0.2645,
      "reward": 0.5542496144771576,
      "reward_std": 0.19918350875377655,
      "reward_std_all": 0.2103368639945984,
      "rewards-max/forensics_iou_reward": 0.9380955994129181,
      "rewards-max/forensics_strict_edge_reward": 0.1751270443201065,
      "rewards-max/forensics_tiou_reward": 0.9154494106769562,
      "rewards/forensics_iou_reward": 0.3507597893476486,
      "rewards/forensics_strict_edge_reward": 0.008736473275348544,
      "rewards/forensics_tiou_reward": 0.28852352499961853,
      "step": 108
    },
    {
      "completion_length": 29.84375,
      "epoch": 0.3348694316436252,
      "grad_norm": 4.043277263641357,
      "kl": 0.2421875,
      "learning_rate": 9.169230769230769e-07,
      "loss": 0.3003,
      "reward": 0.5209079086780548,
      "reward_std": 0.23522069305181503,
      "reward_std_all": 0.2522331401705742,
      "rewards-max/forensics_iou_reward": 0.835798978805542,
      "rewards-max/forensics_strict_edge_reward": 0.2663553785532713,
      "rewards-max/forensics_tiou_reward": 0.7572996616363525,
      "rewards/forensics_iou_reward": 0.3216060698032379,
      "rewards/forensics_strict_edge_reward": 0.011603429331444204,
      "rewards/forensics_tiou_reward": 0.2634872943162918,
      "step": 109
    },
    {
      "completion_length": 26.328125,
      "epoch": 0.3379416282642089,
      "grad_norm": 6.187690258026123,
      "kl": 0.2265625,
      "learning_rate": 9.161538461538462e-07,
      "loss": 0.1477,
      "reward": 0.5909094512462616,
      "reward_std": 0.18214871734380722,
      "reward_std_all": 0.20042385160923004,
      "rewards-max/forensics_iou_reward": 0.8435258865356445,
      "rewards-max/forensics_strict_edge_reward": 0.09967953991144896,
      "rewards-max/forensics_tiou_reward": 0.7825811505317688,
      "rewards/forensics_iou_reward": 0.3909836560487747,
      "rewards/forensics_strict_edge_reward": 0.00481707789003849,
      "rewards/forensics_tiou_reward": 0.32330040633678436,
      "step": 110
    },
    {
      "completion_length": 25.328125,
      "epoch": 0.34101382488479265,
      "grad_norm": 5.810441970825195,
      "kl": 0.22412109375,
      "learning_rate": 9.153846153846153e-07,
      "loss": 0.1115,
      "reward": 0.5502722561359406,
      "reward_std": 0.2357710525393486,
      "reward_std_all": 0.23460101336240768,
      "rewards-max/forensics_iou_reward": 0.8766076564788818,
      "rewards-max/forensics_strict_edge_reward": 0.18569880165159702,
      "rewards-max/forensics_tiou_reward": 0.8228445649147034,
      "rewards/forensics_iou_reward": 0.3767555207014084,
      "rewards/forensics_strict_edge_reward": 0.011325266677886248,
      "rewards/forensics_tiou_reward": 0.32135844230651855,
      "step": 111
    },
    {
      "completion_length": 27.546875,
      "epoch": 0.34408602150537637,
      "grad_norm": 3.8910975456237793,
      "kl": 0.21044921875,
      "learning_rate": 9.146153846153846e-07,
      "loss": 0.1756,
      "reward": 0.5708464086055756,
      "reward_std": 0.12845101207494736,
      "reward_std_all": 0.17771758884191513,
      "rewards-max/forensics_iou_reward": 0.9123252928256989,
      "rewards-max/forensics_strict_edge_reward": 0.015071632340550423,
      "rewards-max/forensics_tiou_reward": 0.8693022727966309,
      "rewards/forensics_iou_reward": 0.38565686345100403,
      "rewards/forensics_strict_edge_reward": 0.0008283426868729293,
      "rewards/forensics_tiou_reward": 0.3189752697944641,
      "step": 112
    },
    {
      "completion_length": 26.296875,
      "epoch": 0.34715821812596004,
      "grad_norm": 4.051464080810547,
      "kl": 0.2236328125,
      "learning_rate": 9.138461538461538e-07,
      "loss": 0.1893,
      "reward": 0.5466814637184143,
      "reward_std": 0.2642985135316849,
      "reward_std_all": 0.25531117618083954,
      "rewards-max/forensics_iou_reward": 0.9516202509403229,
      "rewards-max/forensics_strict_edge_reward": 0.5788819789886475,
      "rewards-max/forensics_tiou_reward": 0.9296334087848663,
      "rewards/forensics_iou_reward": 0.38945648074150085,
      "rewards/forensics_strict_edge_reward": 0.03252029046416283,
      "rewards/forensics_tiou_reward": 0.3433162420988083,
      "step": 113
    },
    {
      "completion_length": 26.296875,
      "epoch": 0.35023041474654376,
      "grad_norm": 5.096777439117432,
      "kl": 0.21044921875,
      "learning_rate": 9.130769230769231e-07,
      "loss": 0.0286,
      "reward": 0.5293980538845062,
      "reward_std": 0.26216503232717514,
      "reward_std_all": 0.2622104734182358,
      "rewards-max/forensics_iou_reward": 0.7427642345428467,
      "rewards-max/forensics_strict_edge_reward": 0.16279505379498005,
      "rewards-max/forensics_tiou_reward": 0.7055136859416962,
      "rewards/forensics_iou_reward": 0.3233613818883896,
      "rewards/forensics_strict_edge_reward": 0.007104193093255162,
      "rewards/forensics_tiou_reward": 0.26931361854076385,
      "step": 114
    },
    {
      "completion_length": 26.5,
      "epoch": 0.3533026113671275,
      "grad_norm": 6.499116897583008,
      "kl": 0.19970703125,
      "learning_rate": 9.123076923076922e-07,
      "loss": 0.2032,
      "reward": 0.5213139057159424,
      "reward_std": 0.16671299934387207,
      "reward_std_all": 0.21117651462554932,
      "rewards-max/forensics_iou_reward": 0.7240506410598755,
      "rewards-max/forensics_strict_edge_reward": 0.0573459102306515,
      "rewards-max/forensics_tiou_reward": 0.6568599045276642,
      "rewards/forensics_iou_reward": 0.3112799748778343,
      "rewards/forensics_strict_edge_reward": 0.005610505642835051,
      "rewards/forensics_tiou_reward": 0.25140005350112915,
      "step": 115
    },
    {
      "completion_length": 26.859375,
      "epoch": 0.3563748079877112,
      "grad_norm": 3.917682647705078,
      "kl": 0.19384765625,
      "learning_rate": 9.115384615384614e-07,
      "loss": 0.0733,
      "reward": 0.5572815239429474,
      "reward_std": 0.19463800638914108,
      "reward_std_all": 0.2153046876192093,
      "rewards-max/forensics_iou_reward": 0.8971031010150909,
      "rewards-max/forensics_strict_edge_reward": 0.17274889221880585,
      "rewards-max/forensics_tiou_reward": 0.8413791656494141,
      "rewards/forensics_iou_reward": 0.3783129155635834,
      "rewards/forensics_strict_edge_reward": 0.00795028206994175,
      "rewards/forensics_tiou_reward": 0.3187480568885803,
      "step": 116
    },
    {
      "completion_length": 24.1875,
      "epoch": 0.35944700460829493,
      "grad_norm": 4.280894756317139,
      "kl": 0.23779296875,
      "learning_rate": 9.107692307692308e-07,
      "loss": 0.1419,
      "reward": 0.4512985646724701,
      "reward_std": 0.25947442650794983,
      "reward_std_all": 0.28916944563388824,
      "rewards-max/forensics_iou_reward": 0.9230381846427917,
      "rewards-max/forensics_strict_edge_reward": 0.241746187210083,
      "rewards-max/forensics_tiou_reward": 0.8997297286987305,
      "rewards/forensics_iou_reward": 0.2970297336578369,
      "rewards/forensics_strict_edge_reward": 0.013749956153333187,
      "rewards/forensics_tiou_reward": 0.25335440039634705,
      "step": 117
    },
    {
      "completion_length": 30.265625,
      "epoch": 0.36251920122887865,
      "grad_norm": 4.402066707611084,
      "kl": 0.22216796875,
      "learning_rate": 9.1e-07,
      "loss": 0.1167,
      "reward": 0.5355695188045502,
      "reward_std": 0.2039339691400528,
      "reward_std_all": 0.23352691531181335,
      "rewards-max/forensics_iou_reward": 0.9026848375797272,
      "rewards-max/forensics_strict_edge_reward": 0.14890070422552526,
      "rewards-max/forensics_tiou_reward": 0.87591752409935,
      "rewards/forensics_iou_reward": 0.3577735871076584,
      "rewards/forensics_strict_edge_reward": 0.006491008534794673,
      "rewards/forensics_tiou_reward": 0.30381014943122864,
      "step": 118
    },
    {
      "completion_length": 26.1875,
      "epoch": 0.3655913978494624,
      "grad_norm": 4.5856146812438965,
      "kl": 0.24072265625,
      "learning_rate": 9.092307692307692e-07,
      "loss": 0.3059,
      "reward": 0.5274766534566879,
      "reward_std": 0.16765379160642624,
      "reward_std_all": 0.21051771938800812,
      "rewards-max/forensics_iou_reward": 0.9074155986309052,
      "rewards-max/forensics_strict_edge_reward": 0.38241301476955414,
      "rewards-max/forensics_tiou_reward": 0.8867519199848175,
      "rewards/forensics_iou_reward": 0.3396589010953903,
      "rewards/forensics_strict_edge_reward": 0.0229605077765882,
      "rewards/forensics_tiou_reward": 0.28778311610221863,
      "step": 119
    },
    {
      "completion_length": 26.4375,
      "epoch": 0.3686635944700461,
      "grad_norm": 3.4189860820770264,
      "kl": 0.2783203125,
      "learning_rate": 9.084615384615384e-07,
      "loss": 0.1903,
      "reward": 0.4267066568136215,
      "reward_std": 0.26859094202518463,
      "reward_std_all": 0.3177901953458786,
      "rewards-max/forensics_iou_reward": 0.8586455583572388,
      "rewards-max/forensics_strict_edge_reward": 0.1414766723755747,
      "rewards-max/forensics_tiou_reward": 0.8121774196624756,
      "rewards/forensics_iou_reward": 0.2792898267507553,
      "rewards/forensics_strict_edge_reward": 0.00885942437525955,
      "rewards/forensics_tiou_reward": 0.22940316796302795,
      "step": 120
    },
    {
      "completion_length": 28.375,
      "epoch": 0.3717357910906298,
      "grad_norm": 5.087185382843018,
      "kl": 0.2041015625,
      "learning_rate": 9.076923076923076e-07,
      "loss": 0.0775,
      "reward": 0.5505377352237701,
      "reward_std": 0.211967334151268,
      "reward_std_all": 0.21908552944660187,
      "rewards-max/forensics_iou_reward": 0.6774488091468811,
      "rewards-max/forensics_strict_edge_reward": 0.028877034783363342,
      "rewards-max/forensics_tiou_reward": 0.6211104094982147,
      "rewards/forensics_iou_reward": 0.3053795397281647,
      "rewards/forensics_strict_edge_reward": 0.001291445893002674,
      "rewards/forensics_tiou_reward": 0.2481776475906372,
      "step": 121
    },
    {
      "completion_length": 26.8125,
      "epoch": 0.37480798771121354,
      "grad_norm": 4.4028239250183105,
      "kl": 0.234375,
      "learning_rate": 9.069230769230769e-07,
      "loss": 0.2091,
      "reward": 0.6297822594642639,
      "reward_std": 0.20845963805913925,
      "reward_std_all": 0.19451797008514404,
      "rewards-max/forensics_iou_reward": 0.9639306962490082,
      "rewards-max/forensics_strict_edge_reward": 0.42115044593811035,
      "rewards-max/forensics_tiou_reward": 0.9552880823612213,
      "rewards/forensics_iou_reward": 0.44591957330703735,
      "rewards/forensics_strict_edge_reward": 0.030654277419671416,
      "rewards/forensics_tiou_reward": 0.39194752275943756,
      "step": 122
    },
    {
      "completion_length": 28.15625,
      "epoch": 0.3778801843317972,
      "grad_norm": 3.735891342163086,
      "kl": 0.2333984375,
      "learning_rate": 9.061538461538461e-07,
      "loss": 0.144,
      "reward": 0.4945964515209198,
      "reward_std": 0.2066303864121437,
      "reward_std_all": 0.24834661185741425,
      "rewards-max/forensics_iou_reward": 0.7912752628326416,
      "rewards-max/forensics_strict_edge_reward": 0.13171655690530315,
      "rewards-max/forensics_tiou_reward": 0.7347529828548431,
      "rewards/forensics_iou_reward": 0.30725786089897156,
      "rewards/forensics_strict_edge_reward": 0.010446521060657687,
      "rewards/forensics_tiou_reward": 0.2545585036277771,
      "step": 123
    },
    {
      "completion_length": 25.109375,
      "epoch": 0.38095238095238093,
      "grad_norm": 3.912069082260132,
      "kl": 0.21875,
      "learning_rate": 9.053846153846153e-07,
      "loss": 0.3001,
      "reward": 0.508383572101593,
      "reward_std": 0.1472659856081009,
      "reward_std_all": 0.21475227177143097,
      "rewards-max/forensics_iou_reward": 0.9005177915096283,
      "rewards-max/forensics_strict_edge_reward": 0.016682819463312626,
      "rewards-max/forensics_tiou_reward": 0.8581817150115967,
      "rewards/forensics_iou_reward": 0.34944234788417816,
      "rewards/forensics_strict_edge_reward": 0.0008801744552329183,
      "rewards/forensics_tiou_reward": 0.28557127714157104,
      "step": 124
    },
    {
      "completion_length": 30.703125,
      "epoch": 0.38402457757296465,
      "grad_norm": 3.3633551597595215,
      "kl": 0.2744140625,
      "learning_rate": 9.046153846153846e-07,
      "loss": 0.1476,
      "reward": 0.5734530091285706,
      "reward_std": 0.21576064825057983,
      "reward_std_all": 0.21508053690195084,
      "rewards-max/forensics_iou_reward": 0.9002538919448853,
      "rewards-max/forensics_strict_edge_reward": 0.07887846417725086,
      "rewards-max/forensics_tiou_reward": 0.8666887879371643,
      "rewards/forensics_iou_reward": 0.3758891820907593,
      "rewards/forensics_strict_edge_reward": 0.005026907427236438,
      "rewards/forensics_tiou_reward": 0.31441669166088104,
      "step": 125
    },
    {
      "completion_length": 25.46875,
      "epoch": 0.3870967741935484,
      "grad_norm": 5.501532554626465,
      "kl": 0.25244140625,
      "learning_rate": 9.038461538461538e-07,
      "loss": -0.0289,
      "reward": 0.4763728082180023,
      "reward_std": 0.28925785422325134,
      "reward_std_all": 0.2980456203222275,
      "rewards-max/forensics_iou_reward": 0.8050174415111542,
      "rewards-max/forensics_strict_edge_reward": 0.05439786380156875,
      "rewards-max/forensics_tiou_reward": 0.7449977993965149,
      "rewards/forensics_iou_reward": 0.3016032725572586,
      "rewards/forensics_strict_edge_reward": 0.0037029579689260572,
      "rewards/forensics_tiou_reward": 0.2503037676215172,
      "step": 126
    },
    {
      "completion_length": 25.578125,
      "epoch": 0.3901689708141321,
      "grad_norm": 3.592749834060669,
      "kl": 0.21533203125,
      "learning_rate": 9.030769230769231e-07,
      "loss": 0.1895,
      "reward": 0.49004265666007996,
      "reward_std": 0.205914705991745,
      "reward_std_all": 0.24875520169734955,
      "rewards-max/forensics_iou_reward": 0.9463669061660767,
      "rewards-max/forensics_strict_edge_reward": 0.13856066763401031,
      "rewards-max/forensics_tiou_reward": 0.9233398735523224,
      "rewards/forensics_iou_reward": 0.33661511540412903,
      "rewards/forensics_strict_edge_reward": 0.0069838170893490314,
      "rewards/forensics_tiou_reward": 0.28265970945358276,
      "step": 127
    },
    {
      "completion_length": 26.46875,
      "epoch": 0.3932411674347158,
      "grad_norm": 3.8264973163604736,
      "kl": 0.24072265625,
      "learning_rate": 9.023076923076922e-07,
      "loss": 0.1457,
      "reward": 0.6028430461883545,
      "reward_std": 0.11063750833272934,
      "reward_std_all": 0.1615779846906662,
      "rewards-max/forensics_iou_reward": 0.9278228580951691,
      "rewards-max/forensics_strict_edge_reward": 0.06272164173424244,
      "rewards-max/forensics_tiou_reward": 0.9079649150371552,
      "rewards/forensics_iou_reward": 0.4399905353784561,
      "rewards/forensics_strict_edge_reward": 0.004586000461131334,
      "rewards/forensics_tiou_reward": 0.3934330642223358,
      "step": 128
    },
    {
      "completion_length": 22.40625,
      "epoch": 0.39631336405529954,
      "grad_norm": 4.457706928253174,
      "kl": 0.2734375,
      "learning_rate": 9.015384615384614e-07,
      "loss": 0.1152,
      "reward": 0.4995765686035156,
      "reward_std": 0.20044557750225067,
      "reward_std_all": 0.2559463828802109,
      "rewards-max/forensics_iou_reward": 0.7771499156951904,
      "rewards-max/forensics_strict_edge_reward": 0.22764848545193672,
      "rewards-max/forensics_tiou_reward": 0.7294720411300659,
      "rewards/forensics_iou_reward": 0.3292941451072693,
      "rewards/forensics_strict_edge_reward": 0.011999639216810465,
      "rewards/forensics_tiou_reward": 0.2805216312408447,
      "step": 129
    },
    {
      "completion_length": 23.15625,
      "epoch": 0.39938556067588327,
      "grad_norm": 4.83319616317749,
      "kl": 0.2314453125,
      "learning_rate": 9.007692307692307e-07,
      "loss": 0.2286,
      "reward": 0.37751011550426483,
      "reward_std": 0.1846921145915985,
      "reward_std_all": 0.29020603001117706,
      "rewards-max/forensics_iou_reward": 0.8300628364086151,
      "rewards-max/forensics_strict_edge_reward": 0.09146068058907986,
      "rewards-max/forensics_tiou_reward": 0.7868031561374664,
      "rewards/forensics_iou_reward": 0.22191333770751953,
      "rewards/forensics_strict_edge_reward": 0.005787848145700991,
      "rewards/forensics_tiou_reward": 0.18220416456460953,
      "step": 130
    },
    {
      "completion_length": 27.078125,
      "epoch": 0.402457757296467,
      "grad_norm": 4.183608531951904,
      "kl": 0.2109375,
      "learning_rate": 9e-07,
      "loss": 0.1288,
      "reward": 0.6206194162368774,
      "reward_std": 0.15917596966028214,
      "reward_std_all": 0.16841257363557816,
      "rewards-max/forensics_iou_reward": 0.9802321493625641,
      "rewards-max/forensics_strict_edge_reward": 0.3251090571284294,
      "rewards-max/forensics_tiou_reward": 0.9718814790248871,
      "rewards/forensics_iou_reward": 0.41855649650096893,
      "rewards/forensics_strict_edge_reward": 0.0147070053499192,
      "rewards/forensics_tiou_reward": 0.3486614525318146,
      "step": 131
    },
    {
      "completion_length": 24.09375,
      "epoch": 0.4055299539170507,
      "grad_norm": 4.9730119705200195,
      "kl": 0.23583984375,
      "learning_rate": 8.992307692307693e-07,
      "loss": 0.106,
      "reward": 0.4390377849340439,
      "reward_std": 0.27114337682724,
      "reward_std_all": 0.30626530945301056,
      "rewards-max/forensics_iou_reward": 0.8583946228027344,
      "rewards-max/forensics_strict_edge_reward": 0.07863380154594779,
      "rewards-max/forensics_tiou_reward": 0.8132591843605042,
      "rewards/forensics_iou_reward": 0.28774333000183105,
      "rewards/forensics_strict_edge_reward": 0.00673154869582504,
      "rewards/forensics_tiou_reward": 0.24840473383665085,
      "step": 132
    },
    {
      "completion_length": 26.765625,
      "epoch": 0.40860215053763443,
      "grad_norm": 4.51238489151001,
      "kl": 0.1826171875,
      "learning_rate": 8.984615384615384e-07,
      "loss": 0.1813,
      "reward": 0.5255513042211533,
      "reward_std": 0.2124636024236679,
      "reward_std_all": 0.23817423731088638,
      "rewards-max/forensics_iou_reward": 0.8141968250274658,
      "rewards-max/forensics_strict_edge_reward": 0.1001510787755251,
      "rewards-max/forensics_tiou_reward": 0.7529793679714203,
      "rewards/forensics_iou_reward": 0.3275206536054611,
      "rewards/forensics_strict_edge_reward": 0.005363024945836514,
      "rewards/forensics_tiou_reward": 0.266539990901947,
      "step": 133
    },
    {
      "completion_length": 26.1875,
      "epoch": 0.4116743471582181,
      "grad_norm": 4.5625152587890625,
      "kl": 0.2451171875,
      "learning_rate": 8.976923076923076e-07,
      "loss": 0.1855,
      "reward": 0.5050589740276337,
      "reward_std": 0.2139047458767891,
      "reward_std_all": 0.24293550103902817,
      "rewards-max/forensics_iou_reward": 0.9246704876422882,
      "rewards-max/forensics_strict_edge_reward": 0.047573022544384,
      "rewards-max/forensics_tiou_reward": 0.9050748646259308,
      "rewards/forensics_iou_reward": 0.3383651375770569,
      "rewards/forensics_strict_edge_reward": 0.002804807387292385,
      "rewards/forensics_tiou_reward": 0.2886497974395752,
      "step": 134
    },
    {
      "completion_length": 23.390625,
      "epoch": 0.4147465437788018,
      "grad_norm": 4.483931064605713,
      "kl": 0.2626953125,
      "learning_rate": 8.969230769230769e-07,
      "loss": 0.2339,
      "reward": 0.5382693409919739,
      "reward_std": 0.16660989075899124,
      "reward_std_all": 0.2183775007724762,
      "rewards-max/forensics_iou_reward": 0.8723226487636566,
      "rewards-max/forensics_strict_edge_reward": 0.08970520552247763,
      "rewards-max/forensics_tiou_reward": 0.8479194343090057,
      "rewards/forensics_iou_reward": 0.392449751496315,
      "rewards/forensics_strict_edge_reward": 0.006041931948857382,
      "rewards/forensics_tiou_reward": 0.3408152312040329,
      "step": 135
    },
    {
      "completion_length": 28.390625,
      "epoch": 0.41781874039938555,
      "grad_norm": 4.117213726043701,
      "kl": 0.24853515625,
      "learning_rate": 8.961538461538461e-07,
      "loss": 0.1933,
      "reward": 0.6371690034866333,
      "reward_std": 0.1805519238114357,
      "reward_std_all": 0.173760786652565,
      "rewards-max/forensics_iou_reward": 0.9102151393890381,
      "rewards-max/forensics_strict_edge_reward": 0.18124549556523561,
      "rewards-max/forensics_tiou_reward": 0.866637110710144,
      "rewards/forensics_iou_reward": 0.43991442024707794,
      "rewards/forensics_strict_edge_reward": 0.016916729393415153,
      "rewards/forensics_tiou_reward": 0.3892061859369278,
      "step": 136
    },
    {
      "completion_length": 23.9375,
      "epoch": 0.42089093701996927,
      "grad_norm": 5.296069145202637,
      "kl": 0.2841796875,
      "learning_rate": 8.953846153846152e-07,
      "loss": 0.1405,
      "reward": 0.5581614673137665,
      "reward_std": 0.2359139323234558,
      "reward_std_all": 0.23425276577472687,
      "rewards-max/forensics_iou_reward": 0.9518591463565826,
      "rewards-max/forensics_strict_edge_reward": 0.2592748908791691,
      "rewards-max/forensics_tiou_reward": 0.915619283914566,
      "rewards/forensics_iou_reward": 0.42976710200309753,
      "rewards/forensics_strict_edge_reward": 0.024411349644651636,
      "rewards/forensics_tiou_reward": 0.3630298376083374,
      "step": 137
    },
    {
      "completion_length": 26.765625,
      "epoch": 0.423963133640553,
      "grad_norm": 4.056783199310303,
      "kl": 0.2578125,
      "learning_rate": 8.946153846153846e-07,
      "loss": 0.0734,
      "reward": 0.5278457701206207,
      "reward_std": 0.2378951907157898,
      "reward_std_all": 0.24814803153276443,
      "rewards-max/forensics_iou_reward": 0.920308917760849,
      "rewards-max/forensics_strict_edge_reward": 0.3381679975427687,
      "rewards-max/forensics_tiou_reward": 0.8902405500411987,
      "rewards/forensics_iou_reward": 0.3421667367219925,
      "rewards/forensics_strict_edge_reward": 0.026277422643033788,
      "rewards/forensics_tiou_reward": 0.2941393703222275,
      "step": 138
    },
    {
      "completion_length": 27.875,
      "epoch": 0.4270353302611367,
      "grad_norm": 3.9878463745117188,
      "kl": 0.24755859375,
      "learning_rate": 8.938461538461538e-07,
      "loss": 0.2233,
      "reward": 0.6478437185287476,
      "reward_std": 0.12353164702653885,
      "reward_std_all": 0.14224950969219208,
      "rewards-max/forensics_iou_reward": 0.9832453727722168,
      "rewards-max/forensics_strict_edge_reward": 0.5541301667690277,
      "rewards-max/forensics_tiou_reward": 0.9681732356548309,
      "rewards/forensics_iou_reward": 0.4712558388710022,
      "rewards/forensics_strict_edge_reward": 0.02887143474072218,
      "rewards/forensics_tiou_reward": 0.4089406430721283,
      "step": 139
    },
    {
      "completion_length": 26.4375,
      "epoch": 0.43010752688172044,
      "grad_norm": 5.061419486999512,
      "kl": 0.26318359375,
      "learning_rate": 8.930769230769231e-07,
      "loss": 0.2394,
      "reward": 0.5921483039855957,
      "reward_std": 0.18535377085208893,
      "reward_std_all": 0.19882216304540634,
      "rewards-max/forensics_iou_reward": 0.8662628531455994,
      "rewards-max/forensics_strict_edge_reward": 0.14742095954716206,
      "rewards-max/forensics_tiou_reward": 0.8226812481880188,
      "rewards/forensics_iou_reward": 0.3982159197330475,
      "rewards/forensics_strict_edge_reward": 0.00834188365843147,
      "rewards/forensics_tiou_reward": 0.33781541883945465,
      "step": 140
    },
    {
      "completion_length": 25.125,
      "epoch": 0.43317972350230416,
      "grad_norm": 3.940487861633301,
      "kl": 0.349609375,
      "learning_rate": 8.923076923076923e-07,
      "loss": 0.2147,
      "reward": 0.5550680458545685,
      "reward_std": 0.1893555223941803,
      "reward_std_all": 0.21305706351995468,
      "rewards-max/forensics_iou_reward": 0.9841989874839783,
      "rewards-max/forensics_strict_edge_reward": 0.19442712143063545,
      "rewards-max/forensics_tiou_reward": 0.9716550409793854,
      "rewards/forensics_iou_reward": 0.40216101706027985,
      "rewards/forensics_strict_edge_reward": 0.010745072737336159,
      "rewards/forensics_tiou_reward": 0.32613831758499146,
      "step": 141
    },
    {
      "completion_length": 26.734375,
      "epoch": 0.4362519201228879,
      "grad_norm": 4.485960006713867,
      "kl": 0.20751953125,
      "learning_rate": 8.915384615384614e-07,
      "loss": 0.112,
      "reward": 0.6344891488552094,
      "reward_std": 0.13615184277296066,
      "reward_std_all": 0.14354440197348595,
      "rewards-max/forensics_iou_reward": 0.943946361541748,
      "rewards-max/forensics_strict_edge_reward": 0.28244586661458015,
      "rewards-max/forensics_tiou_reward": 0.9143102467060089,
      "rewards/forensics_iou_reward": 0.43607479333877563,
      "rewards/forensics_strict_edge_reward": 0.020124069415032864,
      "rewards/forensics_tiou_reward": 0.3648279756307602,
      "step": 142
    },
    {
      "completion_length": 26.875,
      "epoch": 0.4393241167434716,
      "grad_norm": 4.529175281524658,
      "kl": 0.31591796875,
      "learning_rate": 8.907692307692307e-07,
      "loss": 0.2719,
      "reward": 0.5786110460758209,
      "reward_std": 0.16648808121681213,
      "reward_std_all": 0.19260261952877045,
      "rewards-max/forensics_iou_reward": 0.8501929938793182,
      "rewards-max/forensics_strict_edge_reward": 0.03366016626750934,
      "rewards-max/forensics_tiou_reward": 0.8171542584896088,
      "rewards/forensics_iou_reward": 0.37118634581565857,
      "rewards/forensics_strict_edge_reward": 0.0038112443478439673,
      "rewards/forensics_tiou_reward": 0.31556185334920883,
      "step": 143
    },
    {
      "completion_length": 25.234375,
      "epoch": 0.4423963133640553,
      "grad_norm": 4.41459846496582,
      "kl": 0.2509765625,
      "learning_rate": 8.9e-07,
      "loss": 0.0439,
      "reward": 0.5727447867393494,
      "reward_std": 0.19561532139778137,
      "reward_std_all": 0.20620420575141907,
      "rewards-max/forensics_iou_reward": 0.8320413529872894,
      "rewards-max/forensics_strict_edge_reward": 0.006521089468151331,
      "rewards-max/forensics_tiou_reward": 0.7520349025726318,
      "rewards/forensics_iou_reward": 0.35783594846725464,
      "rewards/forensics_strict_edge_reward": 0.00047388307575602084,
      "rewards/forensics_tiou_reward": 0.28955772519111633,
      "step": 144
    },
    {
      "completion_length": 27.296875,
      "epoch": 0.445468509984639,
      "grad_norm": 4.777583599090576,
      "kl": 0.2646484375,
      "learning_rate": 8.892307692307693e-07,
      "loss": 0.1112,
      "reward": 0.6373766362667084,
      "reward_std": 0.12370003759860992,
      "reward_std_all": 0.1440705843269825,
      "rewards-max/forensics_iou_reward": 0.9567626118659973,
      "rewards-max/forensics_strict_edge_reward": 0.19200951606035233,
      "rewards-max/forensics_tiou_reward": 0.9256530106067657,
      "rewards/forensics_iou_reward": 0.47419123351573944,
      "rewards/forensics_strict_edge_reward": 0.012343506648903713,
      "rewards/forensics_tiou_reward": 0.40252505242824554,
      "step": 145
    },
    {
      "completion_length": 26.0,
      "epoch": 0.4485407066052227,
      "grad_norm": 4.955506324768066,
      "kl": 0.2099609375,
      "learning_rate": 8.884615384615384e-07,
      "loss": 0.0629,
      "reward": 0.5684885084629059,
      "reward_std": 0.16433849930763245,
      "reward_std_all": 0.1897851675748825,
      "rewards-max/forensics_iou_reward": 0.9138598740100861,
      "rewards-max/forensics_strict_edge_reward": 0.2010706141591072,
      "rewards-max/forensics_tiou_reward": 0.8788200914859772,
      "rewards/forensics_iou_reward": 0.38028231263160706,
      "rewards/forensics_strict_edge_reward": 0.010743033839389682,
      "rewards/forensics_tiou_reward": 0.32635942101478577,
      "step": 146
    },
    {
      "completion_length": 26.15625,
      "epoch": 0.45161290322580644,
      "grad_norm": 3.784916877746582,
      "kl": 0.2548828125,
      "learning_rate": 8.876923076923076e-07,
      "loss": 0.2313,
      "reward": 0.5976203382015228,
      "reward_std": 0.13249848783016205,
      "reward_std_all": 0.16534413397312164,
      "rewards-max/forensics_iou_reward": 0.9114814698696136,
      "rewards-max/forensics_strict_edge_reward": 0.17226452752947807,
      "rewards-max/forensics_tiou_reward": 0.8859074115753174,
      "rewards/forensics_iou_reward": 0.42884060740470886,
      "rewards/forensics_strict_edge_reward": 0.012101154890842736,
      "rewards/forensics_tiou_reward": 0.3565507084131241,
      "step": 147
    },
    {
      "completion_length": 25.578125,
      "epoch": 0.45468509984639016,
      "grad_norm": 5.874279975891113,
      "kl": 0.23583984375,
      "learning_rate": 8.869230769230769e-07,
      "loss": 0.1152,
      "reward": 0.5799611359834671,
      "reward_std": 0.15585340559482574,
      "reward_std_all": 0.17759962379932404,
      "rewards-max/forensics_iou_reward": 0.7218884527683258,
      "rewards-max/forensics_strict_edge_reward": 0.030405189652611853,
      "rewards-max/forensics_tiou_reward": 0.6638354361057281,
      "rewards/forensics_iou_reward": 0.362064391374588,
      "rewards/forensics_strict_edge_reward": 0.0022139707368413752,
      "rewards/forensics_tiou_reward": 0.28242165595293045,
      "step": 148
    },
    {
      "completion_length": 25.484375,
      "epoch": 0.4577572964669739,
      "grad_norm": 4.66207218170166,
      "kl": 0.24609375,
      "learning_rate": 8.861538461538461e-07,
      "loss": 0.1271,
      "reward": 0.6286194026470184,
      "reward_std": 0.18926595151424408,
      "reward_std_all": 0.18112237751483917,
      "rewards-max/forensics_iou_reward": 0.9495668113231659,
      "rewards-max/forensics_strict_edge_reward": 0.1665724664926529,
      "rewards-max/forensics_tiou_reward": 0.9285372793674469,
      "rewards/forensics_iou_reward": 0.44073939323425293,
      "rewards/forensics_strict_edge_reward": 0.011921316385269165,
      "rewards/forensics_tiou_reward": 0.3883602172136307,
      "step": 149
    },
    {
      "completion_length": 25.671875,
      "epoch": 0.4608294930875576,
      "grad_norm": 4.747864723205566,
      "kl": 0.22802734375,
      "learning_rate": 8.853846153846154e-07,
      "loss": 0.0969,
      "reward": 0.630802184343338,
      "reward_std": 0.1562199965119362,
      "reward_std_all": 0.16810940206050873,
      "rewards-max/forensics_iou_reward": 0.9297060072422028,
      "rewards-max/forensics_strict_edge_reward": 0.1280084252357483,
      "rewards-max/forensics_tiou_reward": 0.9140311479568481,
      "rewards/forensics_iou_reward": 0.4724843353033066,
      "rewards/forensics_strict_edge_reward": 0.010499550960958004,
      "rewards/forensics_tiou_reward": 0.41324830055236816,
      "step": 150
    },
    {
      "completion_length": 25.734375,
      "epoch": 0.46390168970814133,
      "grad_norm": 4.461247444152832,
      "kl": 0.24658203125,
      "learning_rate": 8.846153846153846e-07,
      "loss": 0.0926,
      "reward": 0.6317240595817566,
      "reward_std": 0.15825634449720383,
      "reward_std_all": 0.15910834074020386,
      "rewards-max/forensics_iou_reward": 0.9290103316307068,
      "rewards-max/forensics_strict_edge_reward": 0.3855445561930537,
      "rewards-max/forensics_tiou_reward": 0.9024952054023743,
      "rewards/forensics_iou_reward": 0.4589230716228485,
      "rewards/forensics_strict_edge_reward": 0.017744902448612265,
      "rewards/forensics_tiou_reward": 0.39433974027633667,
      "step": 151
    },
    {
      "completion_length": 22.53125,
      "epoch": 0.46697388632872505,
      "grad_norm": 5.751547336578369,
      "kl": 0.216796875,
      "learning_rate": 8.838461538461538e-07,
      "loss": 0.1664,
      "reward": 0.5616677403450012,
      "reward_std": 0.1830596700310707,
      "reward_std_all": 0.20636998862028122,
      "rewards-max/forensics_iou_reward": 0.9040781557559967,
      "rewards-max/forensics_strict_edge_reward": 0.4253032058477402,
      "rewards-max/forensics_tiou_reward": 0.8675554394721985,
      "rewards/forensics_iou_reward": 0.39657062292099,
      "rewards/forensics_strict_edge_reward": 0.029751085210591555,
      "rewards/forensics_tiou_reward": 0.33533884584903717,
      "step": 152
    },
    {
      "completion_length": 24.046875,
      "epoch": 0.4700460829493088,
      "grad_norm": 4.381823539733887,
      "kl": 0.2314453125,
      "learning_rate": 8.830769230769231e-07,
      "loss": 0.0786,
      "reward": 0.599806159734726,
      "reward_std": 0.19350264966487885,
      "reward_std_all": 0.1940365731716156,
      "rewards-max/forensics_iou_reward": 0.9651663303375244,
      "rewards-max/forensics_strict_edge_reward": 0.4855904970318079,
      "rewards-max/forensics_tiou_reward": 0.9513607919216156,
      "rewards/forensics_iou_reward": 0.45262420177459717,
      "rewards/forensics_strict_edge_reward": 0.023255302920006216,
      "rewards/forensics_tiou_reward": 0.40293456614017487,
      "step": 153
    },
    {
      "completion_length": 24.15625,
      "epoch": 0.4731182795698925,
      "grad_norm": 5.60158634185791,
      "kl": 0.2431640625,
      "learning_rate": 8.823076923076923e-07,
      "loss": 0.1566,
      "reward": 0.5467836856842041,
      "reward_std": 0.14460734277963638,
      "reward_std_all": 0.20185518264770508,
      "rewards-max/forensics_iou_reward": 0.8055598437786102,
      "rewards-max/forensics_strict_edge_reward": 0.01767953345552087,
      "rewards-max/forensics_tiou_reward": 0.760279655456543,
      "rewards/forensics_iou_reward": 0.3536805510520935,
      "rewards/forensics_strict_edge_reward": 0.0012924036709591746,
      "rewards/forensics_tiou_reward": 0.3023049235343933,
      "step": 154
    },
    {
      "completion_length": 24.890625,
      "epoch": 0.47619047619047616,
      "grad_norm": 5.676844596862793,
      "kl": 0.271484375,
      "learning_rate": 8.815384615384614e-07,
      "loss": 0.1813,
      "reward": 0.6884205341339111,
      "reward_std": 0.14995602518320084,
      "reward_std_all": 0.14075063169002533,
      "rewards-max/forensics_iou_reward": 0.9797961711883545,
      "rewards-max/forensics_strict_edge_reward": 0.5951731204986572,
      "rewards-max/forensics_tiou_reward": 0.9668153822422028,
      "rewards/forensics_iou_reward": 0.4928347319364548,
      "rewards/forensics_strict_edge_reward": 0.030273202806711197,
      "rewards/forensics_tiou_reward": 0.4026734530925751,
      "step": 155
    },
    {
      "completion_length": 22.953125,
      "epoch": 0.4792626728110599,
      "grad_norm": 4.3389387130737305,
      "kl": 0.234375,
      "learning_rate": 8.807692307692307e-07,
      "loss": 0.1429,
      "reward": 0.5455072373151779,
      "reward_std": 0.16776840761303902,
      "reward_std_all": 0.2092466577887535,
      "rewards-max/forensics_iou_reward": 0.8977731764316559,
      "rewards-max/forensics_strict_edge_reward": 0.12366575375199318,
      "rewards-max/forensics_tiou_reward": 0.833031564950943,
      "rewards/forensics_iou_reward": 0.3899182379245758,
      "rewards/forensics_strict_edge_reward": 0.0066808490082621574,
      "rewards/forensics_tiou_reward": 0.3222779333591461,
      "step": 156
    },
    {
      "completion_length": 24.203125,
      "epoch": 0.4823348694316436,
      "grad_norm": 4.207869529724121,
      "kl": 0.27978515625,
      "learning_rate": 8.799999999999999e-07,
      "loss": 0.272,
      "reward": 0.6399440169334412,
      "reward_std": 0.13383761793375015,
      "reward_std_all": 0.15304233878850937,
      "rewards-max/forensics_iou_reward": 0.9512975811958313,
      "rewards-max/forensics_strict_edge_reward": 0.24879217892885208,
      "rewards-max/forensics_tiou_reward": 0.9314219355583191,
      "rewards/forensics_iou_reward": 0.4903262257575989,
      "rewards/forensics_strict_edge_reward": 0.0198413897305727,
      "rewards/forensics_tiou_reward": 0.42804592847824097,
      "step": 157
    },
    {
      "completion_length": 24.59375,
      "epoch": 0.48540706605222733,
      "grad_norm": 4.664424419403076,
      "kl": 0.2275390625,
      "learning_rate": 8.792307692307693e-07,
      "loss": 0.1815,
      "reward": 0.5931989252567291,
      "reward_std": 0.14589983224868774,
      "reward_std_all": 0.17216261476278305,
      "rewards-max/forensics_iou_reward": 0.8844892978668213,
      "rewards-max/forensics_strict_edge_reward": 0.27192600443959236,
      "rewards-max/forensics_tiou_reward": 0.8520430624485016,
      "rewards/forensics_iou_reward": 0.409705713391304,
      "rewards/forensics_strict_edge_reward": 0.013074977090582252,
      "rewards/forensics_tiou_reward": 0.3553857207298279,
      "step": 158
    },
    {
      "completion_length": 24.046875,
      "epoch": 0.48847926267281105,
      "grad_norm": 6.376615047454834,
      "kl": 0.2685546875,
      "learning_rate": 8.784615384615385e-07,
      "loss": 0.0357,
      "reward": 0.5211546719074249,
      "reward_std": 0.22627487778663635,
      "reward_std_all": 0.2575038820505142,
      "rewards-max/forensics_iou_reward": 0.9275415539741516,
      "rewards-max/forensics_strict_edge_reward": 0.2007974460721016,
      "rewards-max/forensics_tiou_reward": 0.9060724973678589,
      "rewards/forensics_iou_reward": 0.36945849657058716,
      "rewards/forensics_strict_edge_reward": 0.008922813693061471,
      "rewards/forensics_tiou_reward": 0.32643719017505646,
      "step": 159
    },
    {
      "completion_length": 22.296875,
      "epoch": 0.4915514592933948,
      "grad_norm": 4.0595221519470215,
      "kl": 0.2802734375,
      "learning_rate": 8.776923076923076e-07,
      "loss": 0.1065,
      "reward": 0.6085573434829712,
      "reward_std": 0.16719705611467361,
      "reward_std_all": 0.18161307275295258,
      "rewards-max/forensics_iou_reward": 0.9660656452178955,
      "rewards-max/forensics_strict_edge_reward": 0.3392418399453163,
      "rewards-max/forensics_tiou_reward": 0.953160434961319,
      "rewards/forensics_iou_reward": 0.4510049819946289,
      "rewards/forensics_strict_edge_reward": 0.033492669463157654,
      "rewards/forensics_tiou_reward": 0.39990438520908356,
      "step": 160
    },
    {
      "completion_length": 21.953125,
      "epoch": 0.4946236559139785,
      "grad_norm": 5.140958786010742,
      "kl": 0.27587890625,
      "learning_rate": 8.769230769230769e-07,
      "loss": 0.1457,
      "reward": 0.5878470838069916,
      "reward_std": 0.17350473999977112,
      "reward_std_all": 0.19614507257938385,
      "rewards-max/forensics_iou_reward": 0.9311635196208954,
      "rewards-max/forensics_strict_edge_reward": 0.28727880120277405,
      "rewards-max/forensics_tiou_reward": 0.9057261645793915,
      "rewards/forensics_iou_reward": 0.4197988212108612,
      "rewards/forensics_strict_edge_reward": 0.023791968822479248,
      "rewards/forensics_tiou_reward": 0.3649199903011322,
      "step": 161
    },
    {
      "completion_length": 23.140625,
      "epoch": 0.4976958525345622,
      "grad_norm": 6.29982852935791,
      "kl": 0.25,
      "learning_rate": 8.761538461538461e-07,
      "loss": 0.2132,
      "reward": 0.5652699619531631,
      "reward_std": 0.1550530567765236,
      "reward_std_all": 0.20129839330911636,
      "rewards-max/forensics_iou_reward": 0.9464765787124634,
      "rewards-max/forensics_strict_edge_reward": 0.10031454637646675,
      "rewards-max/forensics_tiou_reward": 0.9328233599662781,
      "rewards/forensics_iou_reward": 0.4225059896707535,
      "rewards/forensics_strict_edge_reward": 0.012346496805548668,
      "rewards/forensics_tiou_reward": 0.3664074242115021,
      "step": 162
    },
    {
      "completion_length": 23.578125,
      "epoch": 0.500768049155146,
      "grad_norm": 4.550614356994629,
      "kl": 0.2001953125,
      "learning_rate": 8.753846153846154e-07,
      "loss": -0.0727,
      "reward": 0.5132776200771332,
      "reward_std": 0.16211777180433273,
      "reward_std_all": 0.21923501044511795,
      "rewards-max/forensics_iou_reward": 0.8417852222919464,
      "rewards-max/forensics_strict_edge_reward": 0.42007872462272644,
      "rewards-max/forensics_tiou_reward": 0.7961931228637695,
      "rewards/forensics_iou_reward": 0.351016029715538,
      "rewards/forensics_strict_edge_reward": 0.01819154853001237,
      "rewards/forensics_tiou_reward": 0.3066258132457733,
      "step": 163
    },
    {
      "completion_length": 24.15625,
      "epoch": 0.5038402457757296,
      "grad_norm": 7.493070125579834,
      "kl": 0.3134765625,
      "learning_rate": 8.746153846153845e-07,
      "loss": 0.2064,
      "reward": 0.5632559061050415,
      "reward_std": 0.22799241542816162,
      "reward_std_all": 0.22726089507341385,
      "rewards-max/forensics_iou_reward": 0.899983137845993,
      "rewards-max/forensics_strict_edge_reward": 0.07422233372926712,
      "rewards-max/forensics_tiou_reward": 0.847560465335846,
      "rewards/forensics_iou_reward": 0.36045004427433014,
      "rewards/forensics_strict_edge_reward": 0.00461723911575973,
      "rewards/forensics_tiou_reward": 0.3023092746734619,
      "step": 164
    },
    {
      "completion_length": 24.1875,
      "epoch": 0.5069124423963134,
      "grad_norm": 6.333958625793457,
      "kl": 0.1826171875,
      "learning_rate": 8.738461538461538e-07,
      "loss": 0.1615,
      "reward": 0.6131658852100372,
      "reward_std": 0.1337602585554123,
      "reward_std_all": 0.159057404845953,
      "rewards-max/forensics_iou_reward": 0.9536392986774445,
      "rewards-max/forensics_strict_edge_reward": 0.0769658237695694,
      "rewards-max/forensics_tiou_reward": 0.9401561915874481,
      "rewards/forensics_iou_reward": 0.43379220366477966,
      "rewards/forensics_strict_edge_reward": 0.004457362927496433,
      "rewards/forensics_tiou_reward": 0.3804488331079483,
      "step": 165
    },
    {
      "completion_length": 25.6875,
      "epoch": 0.5099846390168971,
      "grad_norm": 6.237347602844238,
      "kl": 0.2451171875,
      "learning_rate": 8.730769230769231e-07,
      "loss": 0.0425,
      "reward": 0.5751660466194153,
      "reward_std": 0.19765976071357727,
      "reward_std_all": 0.20621516555547714,
      "rewards-max/forensics_iou_reward": 0.8987660706043243,
      "rewards-max/forensics_strict_edge_reward": 0.027729765279218554,
      "rewards-max/forensics_tiou_reward": 0.8683838546276093,
      "rewards/forensics_iou_reward": 0.393051341176033,
      "rewards/forensics_strict_edge_reward": 0.0013797871361020952,
      "rewards/forensics_tiou_reward": 0.3338056802749634,
      "step": 166
    },
    {
      "completion_length": 23.859375,
      "epoch": 0.5130568356374808,
      "grad_norm": 11.122461318969727,
      "kl": 0.962890625,
      "learning_rate": 8.723076923076923e-07,
      "loss": 0.2493,
      "reward": 0.6075438261032104,
      "reward_std": 0.18132825940847397,
      "reward_std_all": 0.18196729570627213,
      "rewards-max/forensics_iou_reward": 0.9711242020130157,
      "rewards-max/forensics_strict_edge_reward": 0.474664568901062,
      "rewards-max/forensics_tiou_reward": 0.9618738293647766,
      "rewards/forensics_iou_reward": 0.4303985834121704,
      "rewards/forensics_strict_edge_reward": 0.02856874093413353,
      "rewards/forensics_tiou_reward": 0.37395811080932617,
      "step": 167
    },
    {
      "completion_length": 26.78125,
      "epoch": 0.5161290322580645,
      "grad_norm": 4.533309459686279,
      "kl": 0.2568359375,
      "learning_rate": 8.715384615384614e-07,
      "loss": 0.1027,
      "reward": 0.5902586281299591,
      "reward_std": 0.1486213579773903,
      "reward_std_all": 0.17585264146327972,
      "rewards-max/forensics_iou_reward": 0.9802786409854889,
      "rewards-max/forensics_strict_edge_reward": 0.4562618099153042,
      "rewards-max/forensics_tiou_reward": 0.9695737361907959,
      "rewards/forensics_iou_reward": 0.45726780593395233,
      "rewards/forensics_strict_edge_reward": 0.03270645812153816,
      "rewards/forensics_tiou_reward": 0.4097360074520111,
      "step": 168
    },
    {
      "completion_length": 28.25,
      "epoch": 0.5192012288786483,
      "grad_norm": 4.307915210723877,
      "kl": 0.2451171875,
      "learning_rate": 8.707692307692307e-07,
      "loss": 0.0466,
      "reward": 0.6227415800094604,
      "reward_std": 0.14674095064401627,
      "reward_std_all": 0.16469202190637589,
      "rewards-max/forensics_iou_reward": 0.8897083401679993,
      "rewards-max/forensics_strict_edge_reward": 0.4272540509700775,
      "rewards-max/forensics_tiou_reward": 0.8630139231681824,
      "rewards/forensics_iou_reward": 0.4231674373149872,
      "rewards/forensics_strict_edge_reward": 0.023033306002616882,
      "rewards/forensics_tiou_reward": 0.3597358167171478,
      "step": 169
    },
    {
      "completion_length": 25.625,
      "epoch": 0.522273425499232,
      "grad_norm": 5.4873528480529785,
      "kl": 0.2431640625,
      "learning_rate": 8.699999999999999e-07,
      "loss": 0.1707,
      "reward": 0.6249262392520905,
      "reward_std": 0.1668267622590065,
      "reward_std_all": 0.16797102987766266,
      "rewards-max/forensics_iou_reward": 0.9479196071624756,
      "rewards-max/forensics_strict_edge_reward": 0.12150871753692627,
      "rewards-max/forensics_tiou_reward": 0.9273711740970612,
      "rewards/forensics_iou_reward": 0.4505564868450165,
      "rewards/forensics_strict_edge_reward": 0.009444785071536899,
      "rewards/forensics_tiou_reward": 0.38723616302013397,
      "step": 170
    },
    {
      "completion_length": 29.640625,
      "epoch": 0.5253456221198156,
      "grad_norm": 4.751214504241943,
      "kl": 0.2626953125,
      "learning_rate": 8.692307692307692e-07,
      "loss": 0.0778,
      "reward": 0.5194190740585327,
      "reward_std": 0.12548451870679855,
      "reward_std_all": 0.19711577147245407,
      "rewards-max/forensics_iou_reward": 0.8888301849365234,
      "rewards-max/forensics_strict_edge_reward": 0.19523097621276975,
      "rewards-max/forensics_tiou_reward": 0.8511371612548828,
      "rewards/forensics_iou_reward": 0.3352101743221283,
      "rewards/forensics_strict_edge_reward": 0.008700145612237975,
      "rewards/forensics_tiou_reward": 0.28423040360212326,
      "step": 171
    },
    {
      "completion_length": 27.734375,
      "epoch": 0.5284178187403994,
      "grad_norm": 4.689931392669678,
      "kl": 0.23046875,
      "learning_rate": 8.684615384615385e-07,
      "loss": 0.0462,
      "reward": 0.62764772772789,
      "reward_std": 0.14677327871322632,
      "reward_std_all": 0.15342190861701965,
      "rewards-max/forensics_iou_reward": 0.8806005418300629,
      "rewards-max/forensics_strict_edge_reward": 0.03786717452749144,
      "rewards-max/forensics_tiou_reward": 0.8421096503734589,
      "rewards/forensics_iou_reward": 0.4150789827108383,
      "rewards/forensics_strict_edge_reward": 0.005347325470211217,
      "rewards/forensics_tiou_reward": 0.3430900424718857,
      "step": 172
    },
    {
      "completion_length": 24.40625,
      "epoch": 0.5314900153609831,
      "grad_norm": 3.9987576007843018,
      "kl": 0.3076171875,
      "learning_rate": 8.676923076923076e-07,
      "loss": 0.2243,
      "reward": 0.6673294603824615,
      "reward_std": 0.15662090480327606,
      "reward_std_all": 0.15131565928459167,
      "rewards-max/forensics_iou_reward": 0.9652666449546814,
      "rewards-max/forensics_strict_edge_reward": 0.45898057892918587,
      "rewards-max/forensics_tiou_reward": 0.9516382813453674,
      "rewards/forensics_iou_reward": 0.5174269080162048,
      "rewards/forensics_strict_edge_reward": 0.03310018393676728,
      "rewards/forensics_tiou_reward": 0.44653627276420593,
      "step": 173
    },
    {
      "completion_length": 28.671875,
      "epoch": 0.5345622119815668,
      "grad_norm": 4.076421737670898,
      "kl": 0.2353515625,
      "learning_rate": 8.669230769230769e-07,
      "loss": 0.2888,
      "reward": 0.65533846616745,
      "reward_std": 0.14730562269687653,
      "reward_std_all": 0.1462290808558464,
      "rewards-max/forensics_iou_reward": 0.9297003448009491,
      "rewards-max/forensics_strict_edge_reward": 0.3592977523803711,
      "rewards-max/forensics_tiou_reward": 0.9028410017490387,
      "rewards/forensics_iou_reward": 0.44575443863868713,
      "rewards/forensics_strict_edge_reward": 0.02477741427719593,
      "rewards/forensics_tiou_reward": 0.3759438991546631,
      "step": 174
    },
    {
      "completion_length": 27.140625,
      "epoch": 0.5376344086021505,
      "grad_norm": 4.3818039894104,
      "kl": 0.29296875,
      "learning_rate": 8.661538461538461e-07,
      "loss": 0.0199,
      "reward": 0.6167052984237671,
      "reward_std": 0.17017021775245667,
      "reward_std_all": 0.17097365111112595,
      "rewards-max/forensics_iou_reward": 0.918543666601181,
      "rewards-max/forensics_strict_edge_reward": 0.12827162072062492,
      "rewards-max/forensics_tiou_reward": 0.8870109617710114,
      "rewards/forensics_iou_reward": 0.43178442120552063,
      "rewards/forensics_strict_edge_reward": 0.01083945226855576,
      "rewards/forensics_tiou_reward": 0.3841570168733597,
      "step": 175
    },
    {
      "completion_length": 27.703125,
      "epoch": 0.5407066052227343,
      "grad_norm": 4.115788459777832,
      "kl": 0.24560546875,
      "learning_rate": 8.653846153846154e-07,
      "loss": 0.0166,
      "reward": 0.5726265907287598,
      "reward_std": 0.10424596071243286,
      "reward_std_all": 0.1687362715601921,
      "rewards-max/forensics_iou_reward": 0.8561015427112579,
      "rewards-max/forensics_strict_edge_reward": 0.2108158788178116,
      "rewards-max/forensics_tiou_reward": 0.8339466750621796,
      "rewards/forensics_iou_reward": 0.3694168031215668,
      "rewards/forensics_strict_edge_reward": 0.009167735603114124,
      "rewards/forensics_tiou_reward": 0.3088083416223526,
      "step": 176
    },
    {
      "completion_length": 27.21875,
      "epoch": 0.543778801843318,
      "grad_norm": 4.887269020080566,
      "kl": 0.24951171875,
      "learning_rate": 8.646153846153845e-07,
      "loss": 0.2092,
      "reward": 0.628240555524826,
      "reward_std": 0.09580620378255844,
      "reward_std_all": 0.1326424852013588,
      "rewards-max/forensics_iou_reward": 0.9311155676841736,
      "rewards-max/forensics_strict_edge_reward": 0.14021335542201996,
      "rewards-max/forensics_tiou_reward": 0.9075990915298462,
      "rewards/forensics_iou_reward": 0.4423634856939316,
      "rewards/forensics_strict_edge_reward": 0.006644371314905584,
      "rewards/forensics_tiou_reward": 0.38375788927078247,
      "step": 177
    },
    {
      "completion_length": 28.25,
      "epoch": 0.5468509984639017,
      "grad_norm": 5.225058555603027,
      "kl": 0.3193359375,
      "learning_rate": 8.638461538461537e-07,
      "loss": 0.1083,
      "reward": 0.5715783536434174,
      "reward_std": 0.15849781036376953,
      "reward_std_all": 0.18822942674160004,
      "rewards-max/forensics_iou_reward": 0.8534780442714691,
      "rewards-max/forensics_strict_edge_reward": 0.2513619896490127,
      "rewards-max/forensics_tiou_reward": 0.8126108944416046,
      "rewards/forensics_iou_reward": 0.37911301851272583,
      "rewards/forensics_strict_edge_reward": 0.01241071883123368,
      "rewards/forensics_tiou_reward": 0.3199513852596283,
      "step": 178
    },
    {
      "completion_length": 29.640625,
      "epoch": 0.5499231950844854,
      "grad_norm": 6.1768083572387695,
      "kl": 0.236328125,
      "learning_rate": 8.630769230769231e-07,
      "loss": 0.1339,
      "reward": 0.6654207110404968,
      "reward_std": 0.14805808663368225,
      "reward_std_all": 0.1441880278289318,
      "rewards-max/forensics_iou_reward": 0.9129913747310638,
      "rewards-max/forensics_strict_edge_reward": 0.31199593655765057,
      "rewards-max/forensics_tiou_reward": 0.8738873600959778,
      "rewards/forensics_iou_reward": 0.46868395805358887,
      "rewards/forensics_strict_edge_reward": 0.014545678161084652,
      "rewards/forensics_tiou_reward": 0.4071878343820572,
      "step": 179
    },
    {
      "completion_length": 27.828125,
      "epoch": 0.5529953917050692,
      "grad_norm": 4.333995819091797,
      "kl": 0.2626953125,
      "learning_rate": 8.623076923076923e-07,
      "loss": -0.0034,
      "reward": 0.5220185220241547,
      "reward_std": 0.2035994827747345,
      "reward_std_all": 0.2386251613497734,
      "rewards-max/forensics_iou_reward": 0.8610098958015442,
      "rewards-max/forensics_strict_edge_reward": 0.14548563584685326,
      "rewards-max/forensics_tiou_reward": 0.83152174949646,
      "rewards/forensics_iou_reward": 0.35180192440748215,
      "rewards/forensics_strict_edge_reward": 0.010127398069016635,
      "rewards/forensics_tiou_reward": 0.30845891684293747,
      "step": 180
    },
    {
      "completion_length": 26.296875,
      "epoch": 0.5560675883256528,
      "grad_norm": 4.493310451507568,
      "kl": 0.24609375,
      "learning_rate": 8.615384615384616e-07,
      "loss": 0.1398,
      "reward": 0.6536076366901398,
      "reward_std": 0.1298188716173172,
      "reward_std_all": 0.13846661150455475,
      "rewards-max/forensics_iou_reward": 0.9580757021903992,
      "rewards-max/forensics_strict_edge_reward": 0.3627530336380005,
      "rewards-max/forensics_tiou_reward": 0.9438543915748596,
      "rewards/forensics_iou_reward": 0.5304909199476242,
      "rewards/forensics_strict_edge_reward": 0.022453882731497288,
      "rewards/forensics_tiou_reward": 0.46897608041763306,
      "step": 181
    },
    {
      "completion_length": 25.921875,
      "epoch": 0.5591397849462365,
      "grad_norm": 3.5490620136260986,
      "kl": 0.25830078125,
      "learning_rate": 8.607692307692307e-07,
      "loss": 0.1758,
      "reward": 0.6595672070980072,
      "reward_std": 0.1559600681066513,
      "reward_std_all": 0.15255634486675262,
      "rewards-max/forensics_iou_reward": 0.9664455354213715,
      "rewards-max/forensics_strict_edge_reward": 0.17090171575546265,
      "rewards-max/forensics_tiou_reward": 0.9505072236061096,
      "rewards/forensics_iou_reward": 0.5131178498268127,
      "rewards/forensics_strict_edge_reward": 0.01719765830785036,
      "rewards/forensics_tiou_reward": 0.45620296895504,
      "step": 182
    },
    {
      "completion_length": 23.359375,
      "epoch": 0.5622119815668203,
      "grad_norm": 4.115595817565918,
      "kl": 0.25830078125,
      "learning_rate": 8.599999999999999e-07,
      "loss": 0.2268,
      "reward": 0.5362544655799866,
      "reward_std": 0.11702485010027885,
      "reward_std_all": 0.18548841029405594,
      "rewards-max/forensics_iou_reward": 0.8448732197284698,
      "rewards-max/forensics_strict_edge_reward": 0.09361124504357576,
      "rewards-max/forensics_tiou_reward": 0.8019120693206787,
      "rewards/forensics_iou_reward": 0.3835922181606293,
      "rewards/forensics_strict_edge_reward": 0.0043235140619799495,
      "rewards/forensics_tiou_reward": 0.32765811681747437,
      "step": 183
    },
    {
      "completion_length": 27.34375,
      "epoch": 0.565284178187404,
      "grad_norm": 4.26094388961792,
      "kl": 0.29296875,
      "learning_rate": 8.592307692307692e-07,
      "loss": 0.0721,
      "reward": 0.6189325153827667,
      "reward_std": 0.15778422355651855,
      "reward_std_all": 0.17303010076284409,
      "rewards-max/forensics_iou_reward": 0.9070600271224976,
      "rewards-max/forensics_strict_edge_reward": 0.31765156984329224,
      "rewards-max/forensics_tiou_reward": 0.8866591155529022,
      "rewards/forensics_iou_reward": 0.4504828155040741,
      "rewards/forensics_strict_edge_reward": 0.01840693224221468,
      "rewards/forensics_tiou_reward": 0.37479934096336365,
      "step": 184
    },
    {
      "completion_length": 25.703125,
      "epoch": 0.5683563748079877,
      "grad_norm": 5.746798038482666,
      "kl": 0.279296875,
      "learning_rate": 8.584615384615385e-07,
      "loss": 0.1588,
      "reward": 0.5906073451042175,
      "reward_std": 0.23278385400772095,
      "reward_std_all": 0.22242560237646103,
      "rewards-max/forensics_iou_reward": 0.8326314985752106,
      "rewards-max/forensics_strict_edge_reward": 0.2838740320876241,
      "rewards-max/forensics_tiou_reward": 0.8079499304294586,
      "rewards/forensics_iou_reward": 0.4026724398136139,
      "rewards/forensics_strict_edge_reward": 0.022887888422701508,
      "rewards/forensics_tiou_reward": 0.3473688066005707,
      "step": 185
    },
    {
      "completion_length": 25.46875,
      "epoch": 0.5714285714285714,
      "grad_norm": 6.254540920257568,
      "kl": 0.3388671875,
      "learning_rate": 8.576923076923076e-07,
      "loss": 0.1716,
      "reward": 0.5990366339683533,
      "reward_std": 0.20761693641543388,
      "reward_std_all": 0.20230204612016678,
      "rewards-max/forensics_iou_reward": 0.8127956390380859,
      "rewards-max/forensics_strict_edge_reward": 0.3637736812233925,
      "rewards-max/forensics_tiou_reward": 0.7822497189044952,
      "rewards/forensics_iou_reward": 0.4122840166091919,
      "rewards/forensics_strict_edge_reward": 0.021543595474213362,
      "rewards/forensics_tiou_reward": 0.3581804186105728,
      "step": 186
    },
    {
      "completion_length": 27.109375,
      "epoch": 0.5745007680491552,
      "grad_norm": 4.475863456726074,
      "kl": 0.259765625,
      "learning_rate": 8.569230769230769e-07,
      "loss": 0.1778,
      "reward": 0.5744885206222534,
      "reward_std": 0.22067642211914062,
      "reward_std_all": 0.22145260870456696,
      "rewards-max/forensics_iou_reward": 0.8452723622322083,
      "rewards-max/forensics_strict_edge_reward": 0.37459032610058784,
      "rewards-max/forensics_tiou_reward": 0.8159867227077484,
      "rewards/forensics_iou_reward": 0.3896876871585846,
      "rewards/forensics_strict_edge_reward": 0.02243044669739902,
      "rewards/forensics_tiou_reward": 0.3404766023159027,
      "step": 187
    },
    {
      "completion_length": 23.65625,
      "epoch": 0.5775729646697388,
      "grad_norm": 5.3239359855651855,
      "kl": 0.2939453125,
      "learning_rate": 8.561538461538461e-07,
      "loss": 0.0167,
      "reward": 0.6228339970111847,
      "reward_std": 0.1630309596657753,
      "reward_std_all": 0.17055152356624603,
      "rewards-max/forensics_iou_reward": 0.9386546313762665,
      "rewards-max/forensics_strict_edge_reward": 0.17124222964048386,
      "rewards-max/forensics_tiou_reward": 0.920200377702713,
      "rewards/forensics_iou_reward": 0.4213162213563919,
      "rewards/forensics_strict_edge_reward": 0.012993297539651394,
      "rewards/forensics_tiou_reward": 0.3717179596424103,
      "step": 188
    },
    {
      "completion_length": 22.5625,
      "epoch": 0.5806451612903226,
      "grad_norm": 5.428213119506836,
      "kl": 0.2939453125,
      "learning_rate": 8.553846153846154e-07,
      "loss": -0.0372,
      "reward": 0.6414920687675476,
      "reward_std": 0.13687554746866226,
      "reward_std_all": 0.15176746994256973,
      "rewards-max/forensics_iou_reward": 0.9593797624111176,
      "rewards-max/forensics_strict_edge_reward": 0.12054011225700378,
      "rewards-max/forensics_tiou_reward": 0.9412727653980255,
      "rewards/forensics_iou_reward": 0.47982752323150635,
      "rewards/forensics_strict_edge_reward": 0.011135640554130077,
      "rewards/forensics_tiou_reward": 0.41490045189857483,
      "step": 189
    },
    {
      "completion_length": 25.78125,
      "epoch": 0.5837173579109063,
      "grad_norm": 4.590858459472656,
      "kl": 0.294921875,
      "learning_rate": 8.546153846153846e-07,
      "loss": 0.207,
      "reward": 0.6126012802124023,
      "reward_std": 0.16383161768317223,
      "reward_std_all": 0.17088580131530762,
      "rewards-max/forensics_iou_reward": 0.9209688901901245,
      "rewards-max/forensics_strict_edge_reward": 0.18665628880262375,
      "rewards-max/forensics_tiou_reward": 0.8941348493099213,
      "rewards/forensics_iou_reward": 0.4320158064365387,
      "rewards/forensics_strict_edge_reward": 0.010669072158634663,
      "rewards/forensics_tiou_reward": 0.3678882122039795,
      "step": 190
    },
    {
      "completion_length": 24.0,
      "epoch": 0.5867895545314901,
      "grad_norm": 4.911631107330322,
      "kl": 0.3017578125,
      "learning_rate": 8.538461538461537e-07,
      "loss": 0.0977,
      "reward": 0.6123105883598328,
      "reward_std": 0.13286394998431206,
      "reward_std_all": 0.16253510117530823,
      "rewards-max/forensics_iou_reward": 0.840774267911911,
      "rewards-max/forensics_strict_edge_reward": 0.1988757811486721,
      "rewards-max/forensics_tiou_reward": 0.7976823151111603,
      "rewards/forensics_iou_reward": 0.41365747153759,
      "rewards/forensics_strict_edge_reward": 0.010308534721843898,
      "rewards/forensics_tiou_reward": 0.35752418637275696,
      "step": 191
    },
    {
      "completion_length": 25.0625,
      "epoch": 0.5898617511520737,
      "grad_norm": 4.62792444229126,
      "kl": 0.3193359375,
      "learning_rate": 8.530769230769231e-07,
      "loss": 0.1588,
      "reward": 0.6388689279556274,
      "reward_std": 0.11644092947244644,
      "reward_std_all": 0.14599108695983887,
      "rewards-max/forensics_iou_reward": 0.9573672711849213,
      "rewards-max/forensics_strict_edge_reward": 0.2510548457503319,
      "rewards-max/forensics_tiou_reward": 0.9411853551864624,
      "rewards/forensics_iou_reward": 0.5014620721340179,
      "rewards/forensics_strict_edge_reward": 0.017353836447000504,
      "rewards/forensics_tiou_reward": 0.4467619061470032,
      "step": 192
    },
    {
      "completion_length": 25.859375,
      "epoch": 0.5929339477726574,
      "grad_norm": 5.275009632110596,
      "kl": 0.28515625,
      "learning_rate": 8.523076923076923e-07,
      "loss": 0.2582,
      "reward": 0.6154012084007263,
      "reward_std": 0.14348585903644562,
      "reward_std_all": 0.16480319947004318,
      "rewards-max/forensics_iou_reward": 0.9163254499435425,
      "rewards-max/forensics_strict_edge_reward": 0.3040982857346535,
      "rewards-max/forensics_tiou_reward": 0.8956180810928345,
      "rewards/forensics_iou_reward": 0.4442593604326248,
      "rewards/forensics_strict_edge_reward": 0.02168034017086029,
      "rewards/forensics_tiou_reward": 0.3880544453859329,
      "step": 193
    },
    {
      "completion_length": 25.703125,
      "epoch": 0.5960061443932412,
      "grad_norm": 5.09995698928833,
      "kl": 0.2890625,
      "learning_rate": 8.515384615384616e-07,
      "loss": 0.1221,
      "reward": 0.6306104362010956,
      "reward_std": 0.11206883937120438,
      "reward_std_all": 0.13868827372789383,
      "rewards-max/forensics_iou_reward": 0.9901695251464844,
      "rewards-max/forensics_strict_edge_reward": 0.10974230850115418,
      "rewards-max/forensics_tiou_reward": 0.9826084077358246,
      "rewards/forensics_iou_reward": 0.4377991408109665,
      "rewards/forensics_strict_edge_reward": 0.005194180179387331,
      "rewards/forensics_tiou_reward": 0.37122824788093567,
      "step": 194
    },
    {
      "completion_length": 26.40625,
      "epoch": 0.5990783410138248,
      "grad_norm": 4.937005043029785,
      "kl": 0.2607421875,
      "learning_rate": 8.507692307692307e-07,
      "loss": -0.0262,
      "reward": 0.6332578957080841,
      "reward_std": 0.14776837080717087,
      "reward_std_all": 0.16160376369953156,
      "rewards-max/forensics_iou_reward": 0.8790937662124634,
      "rewards-max/forensics_strict_edge_reward": 0.126031544059515,
      "rewards-max/forensics_tiou_reward": 0.8313595950603485,
      "rewards/forensics_iou_reward": 0.473299503326416,
      "rewards/forensics_strict_edge_reward": 0.007196013815701008,
      "rewards/forensics_tiou_reward": 0.4090053290128708,
      "step": 195
    },
    {
      "completion_length": 29.90625,
      "epoch": 0.6021505376344086,
      "grad_norm": 4.340379238128662,
      "kl": 0.3125,
      "learning_rate": 8.499999999999999e-07,
      "loss": -0.0,
      "reward": 0.5167586803436279,
      "reward_std": 0.19748001545667648,
      "reward_std_all": 0.22953588515520096,
      "rewards-max/forensics_iou_reward": 0.844989687204361,
      "rewards-max/forensics_strict_edge_reward": 0.06020076549611986,
      "rewards-max/forensics_tiou_reward": 0.7521715462207794,
      "rewards/forensics_iou_reward": 0.35184288024902344,
      "rewards/forensics_strict_edge_reward": 0.0035812880341836717,
      "rewards/forensics_tiou_reward": 0.3016085773706436,
      "step": 196
    },
    {
      "completion_length": 28.578125,
      "epoch": 0.6052227342549923,
      "grad_norm": 5.227457046508789,
      "kl": 0.2783203125,
      "learning_rate": 8.492307692307692e-07,
      "loss": 0.1116,
      "reward": 0.5771773457527161,
      "reward_std": 0.16103430837392807,
      "reward_std_all": 0.1882934272289276,
      "rewards-max/forensics_iou_reward": 0.8806578814983368,
      "rewards-max/forensics_strict_edge_reward": 0.08238206431269646,
      "rewards-max/forensics_tiou_reward": 0.8516316115856171,
      "rewards/forensics_iou_reward": 0.3898821473121643,
      "rewards/forensics_strict_edge_reward": 0.004386749817058444,
      "rewards/forensics_tiou_reward": 0.3363287150859833,
      "step": 197
    },
    {
      "completion_length": 29.28125,
      "epoch": 0.6082949308755761,
      "grad_norm": 4.675616264343262,
      "kl": 0.3095703125,
      "learning_rate": 8.484615384615384e-07,
      "loss": 0.2233,
      "reward": 0.5362160354852676,
      "reward_std": 0.1828790232539177,
      "reward_std_all": 0.21901892870664597,
      "rewards-max/forensics_iou_reward": 0.878381073474884,
      "rewards-max/forensics_strict_edge_reward": 0.04353142762556672,
      "rewards-max/forensics_tiou_reward": 0.8434219658374786,
      "rewards/forensics_iou_reward": 0.34980328381061554,
      "rewards/forensics_strict_edge_reward": 0.002553820435423404,
      "rewards/forensics_tiou_reward": 0.30114123225212097,
      "step": 198
    },
    {
      "completion_length": 25.015625,
      "epoch": 0.6113671274961597,
      "grad_norm": 5.475357532501221,
      "kl": 0.298828125,
      "learning_rate": 8.476923076923077e-07,
      "loss": 0.0077,
      "reward": 0.6634179949760437,
      "reward_std": 0.12685857340693474,
      "reward_std_all": 0.13158506155014038,
      "rewards-max/forensics_iou_reward": 0.9553690850734711,
      "rewards-max/forensics_strict_edge_reward": 0.4716038852930069,
      "rewards-max/forensics_tiou_reward": 0.944955050945282,
      "rewards/forensics_iou_reward": 0.49851831793785095,
      "rewards/forensics_strict_edge_reward": 0.050233712419867516,
      "rewards/forensics_tiou_reward": 0.44043295085430145,
      "step": 199
    },
    {
      "completion_length": 26.515625,
      "epoch": 0.6144393241167435,
      "grad_norm": 5.225790500640869,
      "kl": 0.353515625,
      "learning_rate": 8.469230769230769e-07,
      "loss": 0.0145,
      "reward": 0.5760150998830795,
      "reward_std": 0.1598150134086609,
      "reward_std_all": 0.18703492730855942,
      "rewards-max/forensics_iou_reward": 0.9206743240356445,
      "rewards-max/forensics_strict_edge_reward": 0.11836010962724686,
      "rewards-max/forensics_tiou_reward": 0.8977431058883667,
      "rewards/forensics_iou_reward": 0.4045538455247879,
      "rewards/forensics_strict_edge_reward": 0.008297997061163187,
      "rewards/forensics_tiou_reward": 0.345546655356884,
      "step": 200
    },
    {
      "completion_length": 26.078125,
      "epoch": 0.6175115207373272,
      "grad_norm": 5.296114921569824,
      "kl": 0.3193359375,
      "learning_rate": 8.461538461538461e-07,
      "loss": 0.1626,
      "reward": 0.528567910194397,
      "reward_std": 0.1706678345799446,
      "reward_std_all": 0.2246677130460739,
      "rewards-max/forensics_iou_reward": 0.9695706963539124,
      "rewards-max/forensics_strict_edge_reward": 0.482279509305954,
      "rewards-max/forensics_tiou_reward": 0.9592782258987427,
      "rewards/forensics_iou_reward": 0.40807896852493286,
      "rewards/forensics_strict_edge_reward": 0.038541411980986595,
      "rewards/forensics_tiou_reward": 0.35465607047080994,
      "step": 201
    },
    {
      "completion_length": 28.625,
      "epoch": 0.620583717357911,
      "grad_norm": 5.453418254852295,
      "kl": 0.3330078125,
      "learning_rate": 8.453846153846154e-07,
      "loss": 0.0053,
      "reward": 0.5841101109981537,
      "reward_std": 0.1804114580154419,
      "reward_std_all": 0.19895734637975693,
      "rewards-max/forensics_iou_reward": 0.865174412727356,
      "rewards-max/forensics_strict_edge_reward": 0.18343859165906906,
      "rewards-max/forensics_tiou_reward": 0.8023044466972351,
      "rewards/forensics_iou_reward": 0.39482948184013367,
      "rewards/forensics_strict_edge_reward": 0.009288622066378593,
      "rewards/forensics_tiou_reward": 0.3251763582229614,
      "step": 202
    },
    {
      "completion_length": 26.109375,
      "epoch": 0.6236559139784946,
      "grad_norm": 5.795908451080322,
      "kl": 0.390625,
      "learning_rate": 8.446153846153846e-07,
      "loss": 0.1998,
      "reward": 0.6296715140342712,
      "reward_std": 0.14182131364941597,
      "reward_std_all": 0.1552141085267067,
      "rewards-max/forensics_iou_reward": 0.8133164048194885,
      "rewards-max/forensics_strict_edge_reward": 0.09525631135329604,
      "rewards-max/forensics_tiou_reward": 0.765648603439331,
      "rewards/forensics_iou_reward": 0.41195622086524963,
      "rewards/forensics_strict_edge_reward": 0.007102496558218263,
      "rewards/forensics_tiou_reward": 0.3363583981990814,
      "step": 203
    },
    {
      "completion_length": 25.953125,
      "epoch": 0.6267281105990783,
      "grad_norm": 5.132203102111816,
      "kl": 0.30859375,
      "learning_rate": 8.438461538461537e-07,
      "loss": 0.1657,
      "reward": 0.6763275861740112,
      "reward_std": 0.12152513861656189,
      "reward_std_all": 0.12411990016698837,
      "rewards-max/forensics_iou_reward": 0.969685435295105,
      "rewards-max/forensics_strict_edge_reward": 0.5711847841739655,
      "rewards-max/forensics_tiou_reward": 0.9632891416549683,
      "rewards/forensics_iou_reward": 0.49322132766246796,
      "rewards/forensics_strict_edge_reward": 0.03828962333500385,
      "rewards/forensics_tiou_reward": 0.43228013813495636,
      "step": 204
    },
    {
      "completion_length": 28.9375,
      "epoch": 0.6298003072196621,
      "grad_norm": 4.416938304901123,
      "kl": 0.236328125,
      "learning_rate": 8.43076923076923e-07,
      "loss": 0.1023,
      "reward": 0.6141711175441742,
      "reward_std": 0.1210714690387249,
      "reward_std_all": 0.14862313866615295,
      "rewards-max/forensics_iou_reward": 0.8211343884468079,
      "rewards-max/forensics_strict_edge_reward": 0.20490478077772423,
      "rewards-max/forensics_tiou_reward": 0.7996902763843536,
      "rewards/forensics_iou_reward": 0.4081299901008606,
      "rewards/forensics_strict_edge_reward": 0.010367513388246152,
      "rewards/forensics_tiou_reward": 0.3492664694786072,
      "step": 205
    },
    {
      "completion_length": 29.4375,
      "epoch": 0.6328725038402457,
      "grad_norm": 4.875456809997559,
      "kl": 0.326171875,
      "learning_rate": 8.423076923076923e-07,
      "loss": 0.1632,
      "reward": 0.6171256303787231,
      "reward_std": 0.17928820103406906,
      "reward_std_all": 0.17957528680562973,
      "rewards-max/forensics_iou_reward": 0.8724706470966339,
      "rewards-max/forensics_strict_edge_reward": 0.2506436218973249,
      "rewards-max/forensics_tiou_reward": 0.8353052139282227,
      "rewards/forensics_iou_reward": 0.40477415919303894,
      "rewards/forensics_strict_edge_reward": 0.03379550689714961,
      "rewards/forensics_tiou_reward": 0.3280218690633774,
      "step": 206
    },
    {
      "completion_length": 31.203125,
      "epoch": 0.6359447004608295,
      "grad_norm": 4.19541072845459,
      "kl": 0.318359375,
      "learning_rate": 8.415384615384616e-07,
      "loss": 0.2153,
      "reward": 0.6515748500823975,
      "reward_std": 0.10998119413852692,
      "reward_std_all": 0.1268802210688591,
      "rewards-max/forensics_iou_reward": 0.9384623169898987,
      "rewards-max/forensics_strict_edge_reward": 0.22466447949409485,
      "rewards-max/forensics_tiou_reward": 0.9273118674755096,
      "rewards/forensics_iou_reward": 0.46942926943302155,
      "rewards/forensics_strict_edge_reward": 0.013911866582930088,
      "rewards/forensics_tiou_reward": 0.4080813229084015,
      "step": 207
    },
    {
      "completion_length": 24.75,
      "epoch": 0.6390168970814132,
      "grad_norm": 4.569581985473633,
      "kl": 0.30078125,
      "learning_rate": 8.407692307692308e-07,
      "loss": 0.2135,
      "reward": 0.6624739766120911,
      "reward_std": 0.14937932789325714,
      "reward_std_all": 0.14401933923363686,
      "rewards-max/forensics_iou_reward": 0.9601360559463501,
      "rewards-max/forensics_strict_edge_reward": 0.3321225196123123,
      "rewards-max/forensics_tiou_reward": 0.9448045194149017,
      "rewards/forensics_iou_reward": 0.5039596408605576,
      "rewards/forensics_strict_edge_reward": 0.040633837692439556,
      "rewards/forensics_tiou_reward": 0.450106605887413,
      "step": 208
    },
    {
      "completion_length": 25.390625,
      "epoch": 0.642089093701997,
      "grad_norm": 5.1810407638549805,
      "kl": 0.3056640625,
      "learning_rate": 8.399999999999999e-07,
      "loss": 0.0289,
      "reward": 0.555902361869812,
      "reward_std": 0.15386545285582542,
      "reward_std_all": 0.19772309064865112,
      "rewards-max/forensics_iou_reward": 0.9322102963924408,
      "rewards-max/forensics_strict_edge_reward": 0.26147375255823135,
      "rewards-max/forensics_tiou_reward": 0.9118959009647369,
      "rewards/forensics_iou_reward": 0.38162675499916077,
      "rewards/forensics_strict_edge_reward": 0.011979278177022934,
      "rewards/forensics_tiou_reward": 0.3246801793575287,
      "step": 209
    },
    {
      "completion_length": 27.90625,
      "epoch": 0.6451612903225806,
      "grad_norm": 4.775360107421875,
      "kl": 0.27734375,
      "learning_rate": 8.392307692307692e-07,
      "loss": 0.1105,
      "reward": 0.6016300320625305,
      "reward_std": 0.16471631079912186,
      "reward_std_all": 0.1873091533780098,
      "rewards-max/forensics_iou_reward": 0.80391064286232,
      "rewards-max/forensics_strict_edge_reward": 0.06146889925003052,
      "rewards-max/forensics_tiou_reward": 0.7580412924289703,
      "rewards/forensics_iou_reward": 0.38602569699287415,
      "rewards/forensics_strict_edge_reward": 0.00441844342276454,
      "rewards/forensics_tiou_reward": 0.33391112089157104,
      "step": 210
    },
    {
      "completion_length": 25.125,
      "epoch": 0.6482334869431644,
      "grad_norm": 4.905134677886963,
      "kl": 0.29443359375,
      "learning_rate": 8.384615384615384e-07,
      "loss": 0.1904,
      "reward": 0.5675561130046844,
      "reward_std": 0.24552007764577866,
      "reward_std_all": 0.2374393343925476,
      "rewards-max/forensics_iou_reward": 0.9712248742580414,
      "rewards-max/forensics_strict_edge_reward": 0.017401418182998896,
      "rewards-max/forensics_tiou_reward": 0.9588602483272552,
      "rewards/forensics_iou_reward": 0.3943132609128952,
      "rewards/forensics_strict_edge_reward": 0.0010301090078428388,
      "rewards/forensics_tiou_reward": 0.3359861671924591,
      "step": 211
    },
    {
      "completion_length": 23.328125,
      "epoch": 0.6513056835637481,
      "grad_norm": 4.88679838180542,
      "kl": 0.26708984375,
      "learning_rate": 8.376923076923076e-07,
      "loss": 0.1745,
      "reward": 0.553420901298523,
      "reward_std": 0.14287611097097397,
      "reward_std_all": 0.18672136962413788,
      "rewards-max/forensics_iou_reward": 0.9440605342388153,
      "rewards-max/forensics_strict_edge_reward": 0.040003959147725254,
      "rewards-max/forensics_tiou_reward": 0.9053424596786499,
      "rewards/forensics_iou_reward": 0.38433937728405,
      "rewards/forensics_strict_edge_reward": 0.0025676605546323117,
      "rewards/forensics_tiou_reward": 0.3233056962490082,
      "step": 212
    },
    {
      "completion_length": 25.4375,
      "epoch": 0.6543778801843319,
      "grad_norm": 4.462191104888916,
      "kl": 0.35546875,
      "learning_rate": 8.369230769230769e-07,
      "loss": 0.025,
      "reward": 0.556279182434082,
      "reward_std": 0.25190597772598267,
      "reward_std_all": 0.2516551613807678,
      "rewards-max/forensics_iou_reward": 0.9411662220954895,
      "rewards-max/forensics_strict_edge_reward": 0.23179927794262767,
      "rewards-max/forensics_tiou_reward": 0.9198550283908844,
      "rewards/forensics_iou_reward": 0.3995877504348755,
      "rewards/forensics_strict_edge_reward": 0.012779257609508932,
      "rewards/forensics_tiou_reward": 0.34817075729370117,
      "step": 213
    },
    {
      "completion_length": 25.625,
      "epoch": 0.6574500768049155,
      "grad_norm": 5.067786693572998,
      "kl": 0.2861328125,
      "learning_rate": 8.361538461538461e-07,
      "loss": 0.0871,
      "reward": 0.5502604693174362,
      "reward_std": 0.16607612371444702,
      "reward_std_all": 0.2074940949678421,
      "rewards-max/forensics_iou_reward": 0.866918295621872,
      "rewards-max/forensics_strict_edge_reward": 0.25631625577807426,
      "rewards-max/forensics_tiou_reward": 0.8330397307872772,
      "rewards/forensics_iou_reward": 0.3907920569181442,
      "rewards/forensics_strict_edge_reward": 0.02910661231726408,
      "rewards/forensics_tiou_reward": 0.3446870446205139,
      "step": 214
    },
    {
      "completion_length": 24.671875,
      "epoch": 0.6605222734254992,
      "grad_norm": 5.01889181137085,
      "kl": 0.275390625,
      "learning_rate": 8.353846153846154e-07,
      "loss": 0.1387,
      "reward": 0.6305022537708282,
      "reward_std": 0.17612730711698532,
      "reward_std_all": 0.17135722190141678,
      "rewards-max/forensics_iou_reward": 0.9077041745185852,
      "rewards-max/forensics_strict_edge_reward": 0.5009789057075977,
      "rewards-max/forensics_tiou_reward": 0.8713439106941223,
      "rewards/forensics_iou_reward": 0.4496356099843979,
      "rewards/forensics_strict_edge_reward": 0.026578028686344624,
      "rewards/forensics_tiou_reward": 0.38469643890857697,
      "step": 215
    },
    {
      "completion_length": 24.546875,
      "epoch": 0.663594470046083,
      "grad_norm": 3.573815107345581,
      "kl": 0.2998046875,
      "learning_rate": 8.346153846153846e-07,
      "loss": 0.1893,
      "reward": 0.6041712164878845,
      "reward_std": 0.12818140536546707,
      "reward_std_all": 0.1601513773202896,
      "rewards-max/forensics_iou_reward": 0.9402507245540619,
      "rewards-max/forensics_strict_edge_reward": 0.08796381670981646,
      "rewards-max/forensics_tiou_reward": 0.905493289232254,
      "rewards/forensics_iou_reward": 0.43524688482284546,
      "rewards/forensics_strict_edge_reward": 0.004493406915571541,
      "rewards/forensics_tiou_reward": 0.3510928153991699,
      "step": 216
    },
    {
      "completion_length": 25.09375,
      "epoch": 0.6666666666666666,
      "grad_norm": 4.40750789642334,
      "kl": 0.24072265625,
      "learning_rate": 8.338461538461537e-07,
      "loss": 0.1726,
      "reward": 0.6571448147296906,
      "reward_std": 0.15206176415085793,
      "reward_std_all": 0.14995195716619492,
      "rewards-max/forensics_iou_reward": 0.9284950792789459,
      "rewards-max/forensics_strict_edge_reward": 0.353687047958374,
      "rewards-max/forensics_tiou_reward": 0.9121789038181305,
      "rewards/forensics_iou_reward": 0.5088297575712204,
      "rewards/forensics_strict_edge_reward": 0.041417865082621574,
      "rewards/forensics_tiou_reward": 0.4540805220603943,
      "step": 217
    },
    {
      "completion_length": 26.484375,
      "epoch": 0.6697388632872504,
      "grad_norm": 4.540378093719482,
      "kl": 0.3623046875,
      "learning_rate": 8.33076923076923e-07,
      "loss": 0.2255,
      "reward": 0.5981743335723877,
      "reward_std": 0.14584795385599136,
      "reward_std_all": 0.1777358278632164,
      "rewards-max/forensics_iou_reward": 0.9529333710670471,
      "rewards-max/forensics_strict_edge_reward": 0.17078380286693573,
      "rewards-max/forensics_tiou_reward": 0.9425852596759796,
      "rewards/forensics_iou_reward": 0.4492179751396179,
      "rewards/forensics_strict_edge_reward": 0.010292976163327694,
      "rewards/forensics_tiou_reward": 0.39932088553905487,
      "step": 218
    },
    {
      "completion_length": 23.96875,
      "epoch": 0.6728110599078341,
      "grad_norm": 4.636998653411865,
      "kl": 0.2607421875,
      "learning_rate": 8.323076923076922e-07,
      "loss": 0.1165,
      "reward": 0.6000527143478394,
      "reward_std": 0.15934810042381287,
      "reward_std_all": 0.17734525352716446,
      "rewards-max/forensics_iou_reward": 0.9301841855049133,
      "rewards-max/forensics_strict_edge_reward": 0.2748943828046322,
      "rewards-max/forensics_tiou_reward": 0.9125491678714752,
      "rewards/forensics_iou_reward": 0.43606290221214294,
      "rewards/forensics_strict_edge_reward": 0.013963275589048862,
      "rewards/forensics_tiou_reward": 0.3805103600025177,
      "step": 219
    },
    {
      "completion_length": 28.796875,
      "epoch": 0.6758832565284179,
      "grad_norm": 3.9705560207366943,
      "kl": 0.27734375,
      "learning_rate": 8.315384615384616e-07,
      "loss": 0.0565,
      "reward": 0.5372811257839203,
      "reward_std": 0.2029721811413765,
      "reward_std_all": 0.23270119726657867,
      "rewards-max/forensics_iou_reward": 0.8135357201099396,
      "rewards-max/forensics_strict_edge_reward": 0.22753343731164932,
      "rewards-max/forensics_tiou_reward": 0.8025053441524506,
      "rewards/forensics_iou_reward": 0.34312310814857483,
      "rewards/forensics_strict_edge_reward": 0.010308616561815143,
      "rewards/forensics_tiou_reward": 0.29095296561717987,
      "step": 220
    },
    {
      "completion_length": 22.75,
      "epoch": 0.6789554531490015,
      "grad_norm": 5.351088047027588,
      "kl": 0.24072265625,
      "learning_rate": 8.307692307692308e-07,
      "loss": 0.1303,
      "reward": 0.5336072444915771,
      "reward_std": 0.1651625856757164,
      "reward_std_all": 0.2078135833144188,
      "rewards-max/forensics_iou_reward": 0.908760130405426,
      "rewards-max/forensics_strict_edge_reward": 0.14943424239754677,
      "rewards-max/forensics_tiou_reward": 0.8820081651210785,
      "rewards/forensics_iou_reward": 0.37867869436740875,
      "rewards/forensics_strict_edge_reward": 0.012959879357367754,
      "rewards/forensics_tiou_reward": 0.328036829829216,
      "step": 221
    },
    {
      "completion_length": 25.15625,
      "epoch": 0.6820276497695853,
      "grad_norm": 4.605327129364014,
      "kl": 0.2763671875,
      "learning_rate": 8.299999999999999e-07,
      "loss": 0.0626,
      "reward": 0.635354220867157,
      "reward_std": 0.13151388615369797,
      "reward_std_all": 0.15075483918190002,
      "rewards-max/forensics_iou_reward": 0.9668154120445251,
      "rewards-max/forensics_strict_edge_reward": 0.38724619895219803,
      "rewards-max/forensics_tiou_reward": 0.9489284753799438,
      "rewards/forensics_iou_reward": 0.47158412635326385,
      "rewards/forensics_strict_edge_reward": 0.021273300051689148,
      "rewards/forensics_tiou_reward": 0.41944749653339386,
      "step": 222
    },
    {
      "completion_length": 22.90625,
      "epoch": 0.685099846390169,
      "grad_norm": 6.3089728355407715,
      "kl": 0.3662109375,
      "learning_rate": 8.292307692307692e-07,
      "loss": 0.2087,
      "reward": 0.6312633156776428,
      "reward_std": 0.1591639220714569,
      "reward_std_all": 0.16399677097797394,
      "rewards-max/forensics_iou_reward": 0.9345482885837555,
      "rewards-max/forensics_strict_edge_reward": 0.40399426221847534,
      "rewards-max/forensics_tiou_reward": 0.921597421169281,
      "rewards/forensics_iou_reward": 0.446233406662941,
      "rewards/forensics_strict_edge_reward": 0.01881756354123354,
      "rewards/forensics_tiou_reward": 0.3824605941772461,
      "step": 223
    },
    {
      "completion_length": 24.078125,
      "epoch": 0.6881720430107527,
      "grad_norm": 4.67911958694458,
      "kl": 0.3076171875,
      "learning_rate": 8.284615384615384e-07,
      "loss": 0.045,
      "reward": 0.568833738565445,
      "reward_std": 0.08653124794363976,
      "reward_std_all": 0.15742699056863785,
      "rewards-max/forensics_iou_reward": 0.9296231269836426,
      "rewards-max/forensics_strict_edge_reward": 0.27183256298303604,
      "rewards-max/forensics_tiou_reward": 0.9101759791374207,
      "rewards/forensics_iou_reward": 0.4060472100973129,
      "rewards/forensics_strict_edge_reward": 0.022190045565366745,
      "rewards/forensics_tiou_reward": 0.36140792071819305,
      "step": 224
    },
    {
      "completion_length": 24.484375,
      "epoch": 0.6912442396313364,
      "grad_norm": 4.510397911071777,
      "kl": 0.3037109375,
      "learning_rate": 8.276923076923077e-07,
      "loss": 0.147,
      "reward": 0.595058023929596,
      "reward_std": 0.173132486641407,
      "reward_std_all": 0.1886458396911621,
      "rewards-max/forensics_iou_reward": 0.9181032776832581,
      "rewards-max/forensics_strict_edge_reward": 0.17013695649802685,
      "rewards-max/forensics_tiou_reward": 0.8683034479618073,
      "rewards/forensics_iou_reward": 0.4373677968978882,
      "rewards/forensics_strict_edge_reward": 0.013796398416161537,
      "rewards/forensics_tiou_reward": 0.38746413588523865,
      "step": 225
    },
    {
      "completion_length": 24.78125,
      "epoch": 0.6943164362519201,
      "grad_norm": 5.050105571746826,
      "kl": 0.25048828125,
      "learning_rate": 8.269230769230768e-07,
      "loss": 0.0551,
      "reward": 0.5229488909244537,
      "reward_std": 0.10798756405711174,
      "reward_std_all": 0.18576648831367493,
      "rewards-max/forensics_iou_reward": 0.8432116210460663,
      "rewards-max/forensics_strict_edge_reward": 0.016716110754714464,
      "rewards-max/forensics_tiou_reward": 0.8136915862560272,
      "rewards/forensics_iou_reward": 0.3590339124202728,
      "rewards/forensics_strict_edge_reward": 0.0008493699299378932,
      "rewards/forensics_tiou_reward": 0.31490853428840637,
      "step": 226
    },
    {
      "completion_length": 22.359375,
      "epoch": 0.6973886328725039,
      "grad_norm": 4.963285446166992,
      "kl": 0.28076171875,
      "learning_rate": 8.261538461538461e-07,
      "loss": 0.2184,
      "reward": 0.6059739291667938,
      "reward_std": 0.12063891626894474,
      "reward_std_all": 0.15870409458875656,
      "rewards-max/forensics_iou_reward": 0.8470813930034637,
      "rewards-max/forensics_strict_edge_reward": 0.26200491935014725,
      "rewards-max/forensics_tiou_reward": 0.8002052009105682,
      "rewards/forensics_iou_reward": 0.43157562613487244,
      "rewards/forensics_strict_edge_reward": 0.023605678230524063,
      "rewards/forensics_tiou_reward": 0.38297079503536224,
      "step": 227
    },
    {
      "completion_length": 23.859375,
      "epoch": 0.7004608294930875,
      "grad_norm": 5.974629878997803,
      "kl": 0.25,
      "learning_rate": 8.253846153846154e-07,
      "loss": 0.0744,
      "reward": 0.5981205105781555,
      "reward_std": 0.19703390449285507,
      "reward_std_all": 0.1966269239783287,
      "rewards-max/forensics_iou_reward": 0.894390195608139,
      "rewards-max/forensics_strict_edge_reward": 0.19422993063926697,
      "rewards-max/forensics_tiou_reward": 0.8657428026199341,
      "rewards/forensics_iou_reward": 0.42005616426467896,
      "rewards/forensics_strict_edge_reward": 0.01623753528110683,
      "rewards/forensics_tiou_reward": 0.3591825067996979,
      "step": 228
    },
    {
      "completion_length": 22.46875,
      "epoch": 0.7035330261136713,
      "grad_norm": 6.24051570892334,
      "kl": 0.3154296875,
      "learning_rate": 8.246153846153846e-07,
      "loss": -0.0188,
      "reward": 0.6865294873714447,
      "reward_std": 0.111375592648983,
      "reward_std_all": 0.11229844763875008,
      "rewards-max/forensics_iou_reward": 0.859704315662384,
      "rewards-max/forensics_strict_edge_reward": 0.15622923895716667,
      "rewards-max/forensics_tiou_reward": 0.7916695177555084,
      "rewards/forensics_iou_reward": 0.4875213950872421,
      "rewards/forensics_strict_edge_reward": 0.009567293338477612,
      "rewards/forensics_tiou_reward": 0.42354816198349,
      "step": 229
    },
    {
      "completion_length": 23.5625,
      "epoch": 0.706605222734255,
      "grad_norm": 5.093001365661621,
      "kl": 0.29931640625,
      "learning_rate": 8.238461538461538e-07,
      "loss": 0.1847,
      "reward": 0.632160484790802,
      "reward_std": 0.219705231487751,
      "reward_std_all": 0.20214979350566864,
      "rewards-max/forensics_iou_reward": 0.9018082618713379,
      "rewards-max/forensics_strict_edge_reward": 0.11807510536164045,
      "rewards-max/forensics_tiou_reward": 0.8698149621486664,
      "rewards/forensics_iou_reward": 0.4821261018514633,
      "rewards/forensics_strict_edge_reward": 0.009955276153050363,
      "rewards/forensics_tiou_reward": 0.40498510003089905,
      "step": 230
    },
    {
      "completion_length": 24.1875,
      "epoch": 0.7096774193548387,
      "grad_norm": 4.8626322746276855,
      "kl": 0.369140625,
      "learning_rate": 8.23076923076923e-07,
      "loss": 0.1515,
      "reward": 0.5685059130191803,
      "reward_std": 0.15600376576185226,
      "reward_std_all": 0.18934866786003113,
      "rewards-max/forensics_iou_reward": 0.871656596660614,
      "rewards-max/forensics_strict_edge_reward": 0.018119526095688343,
      "rewards-max/forensics_tiou_reward": 0.8411922752857208,
      "rewards/forensics_iou_reward": 0.3877614885568619,
      "rewards/forensics_strict_edge_reward": 0.0013972108718007803,
      "rewards/forensics_tiou_reward": 0.33677026629447937,
      "step": 231
    },
    {
      "completion_length": 21.703125,
      "epoch": 0.7127496159754224,
      "grad_norm": 4.954957485198975,
      "kl": 0.359375,
      "learning_rate": 8.223076923076922e-07,
      "loss": 0.1911,
      "reward": 0.6143768727779388,
      "reward_std": 0.15013198927044868,
      "reward_std_all": 0.172356478869915,
      "rewards-max/forensics_iou_reward": 0.9823545217514038,
      "rewards-max/forensics_strict_edge_reward": 0.6009482592344284,
      "rewards-max/forensics_tiou_reward": 0.9769923090934753,
      "rewards/forensics_iou_reward": 0.4699818044900894,
      "rewards/forensics_strict_edge_reward": 0.03523845225572586,
      "rewards/forensics_tiou_reward": 0.4126202166080475,
      "step": 232
    },
    {
      "completion_length": 24.046875,
      "epoch": 0.7158218125960062,
      "grad_norm": 5.313076019287109,
      "kl": 0.2294921875,
      "learning_rate": 8.215384615384616e-07,
      "loss": 0.1312,
      "reward": 0.6518473327159882,
      "reward_std": 0.13803265616297722,
      "reward_std_all": 0.14081860333681107,
      "rewards-max/forensics_iou_reward": 0.9217159450054169,
      "rewards-max/forensics_strict_edge_reward": 0.3632843419909477,
      "rewards-max/forensics_tiou_reward": 0.8953217566013336,
      "rewards/forensics_iou_reward": 0.45637938380241394,
      "rewards/forensics_strict_edge_reward": 0.021006143651902676,
      "rewards/forensics_tiou_reward": 0.38929902017116547,
      "step": 233
    },
    {
      "completion_length": 24.03125,
      "epoch": 0.7188940092165899,
      "grad_norm": 6.228784561157227,
      "kl": 0.2939453125,
      "learning_rate": 8.207692307692308e-07,
      "loss": -0.0103,
      "reward": 0.5935058295726776,
      "reward_std": 0.17903979122638702,
      "reward_std_all": 0.19454525411128998,
      "rewards-max/forensics_iou_reward": 0.9567387998104095,
      "rewards-max/forensics_strict_edge_reward": 0.351408701390028,
      "rewards-max/forensics_tiou_reward": 0.9509362280368805,
      "rewards/forensics_iou_reward": 0.4362178444862366,
      "rewards/forensics_strict_edge_reward": 0.019698067801073194,
      "rewards/forensics_tiou_reward": 0.38290393352508545,
      "step": 234
    },
    {
      "completion_length": 27.125,
      "epoch": 0.7219662058371735,
      "grad_norm": 5.04624605178833,
      "kl": 0.32421875,
      "learning_rate": 8.199999999999999e-07,
      "loss": -0.0496,
      "reward": 0.5996688604354858,
      "reward_std": 0.11274429038167,
      "reward_std_all": 0.14943194389343262,
      "rewards-max/forensics_iou_reward": 0.7910332977771759,
      "rewards-max/forensics_strict_edge_reward": 0.00012237201372045092,
      "rewards-max/forensics_tiou_reward": 0.7376578748226166,
      "rewards/forensics_iou_reward": 0.40579378604888916,
      "rewards/forensics_strict_edge_reward": 6.0019017382728634e-06,
      "rewards/forensics_tiou_reward": 0.3368532806634903,
      "step": 235
    },
    {
      "completion_length": 22.1875,
      "epoch": 0.7250384024577573,
      "grad_norm": 5.36137580871582,
      "kl": 0.3125,
      "learning_rate": 8.192307692307692e-07,
      "loss": 0.0763,
      "reward": 0.5681068003177643,
      "reward_std": 0.18780629336833954,
      "reward_std_all": 0.20921996980905533,
      "rewards-max/forensics_iou_reward": 0.7939896285533905,
      "rewards-max/forensics_strict_edge_reward": 0.13437895756214857,
      "rewards-max/forensics_tiou_reward": 0.7661696076393127,
      "rewards/forensics_iou_reward": 0.3765195608139038,
      "rewards/forensics_strict_edge_reward": 0.007434326893417165,
      "rewards/forensics_tiou_reward": 0.31376999616622925,
      "step": 236
    },
    {
      "completion_length": 28.921875,
      "epoch": 0.728110599078341,
      "grad_norm": 5.240848541259766,
      "kl": 0.3857421875,
      "learning_rate": 8.184615384615384e-07,
      "loss": 0.1069,
      "reward": 0.6548833847045898,
      "reward_std": 0.17478805035352707,
      "reward_std_all": 0.1649678349494934,
      "rewards-max/forensics_iou_reward": 0.8487624228000641,
      "rewards-max/forensics_strict_edge_reward": 0.01912065502256155,
      "rewards-max/forensics_tiou_reward": 0.8038594722747803,
      "rewards/forensics_iou_reward": 0.4420791268348694,
      "rewards/forensics_strict_edge_reward": 0.0012895766412839293,
      "rewards/forensics_tiou_reward": 0.36802439391613007,
      "step": 237
    },
    {
      "completion_length": 23.703125,
      "epoch": 0.7311827956989247,
      "grad_norm": 5.60693883895874,
      "kl": 0.3271484375,
      "learning_rate": 8.176923076923077e-07,
      "loss": 0.1001,
      "reward": 0.6296290457248688,
      "reward_std": 0.16353868693113327,
      "reward_std_all": 0.16599342972040176,
      "rewards-max/forensics_iou_reward": 0.9552679359912872,
      "rewards-max/forensics_strict_edge_reward": 0.3851439654827118,
      "rewards-max/forensics_tiou_reward": 0.9442078769207001,
      "rewards/forensics_iou_reward": 0.4452676326036453,
      "rewards/forensics_strict_edge_reward": 0.02421808149665594,
      "rewards/forensics_tiou_reward": 0.3688637912273407,
      "step": 238
    },
    {
      "completion_length": 27.6875,
      "epoch": 0.7342549923195084,
      "grad_norm": 4.518022060394287,
      "kl": 0.28662109375,
      "learning_rate": 8.169230769230769e-07,
      "loss": 0.1991,
      "reward": 0.6851666569709778,
      "reward_std": 0.1340273767709732,
      "reward_std_all": 0.12747899815440178,
      "rewards-max/forensics_iou_reward": 0.9513373672962189,
      "rewards-max/forensics_strict_edge_reward": 0.40560540556907654,
      "rewards-max/forensics_tiou_reward": 0.938847005367279,
      "rewards/forensics_iou_reward": 0.530863344669342,
      "rewards/forensics_strict_edge_reward": 0.02751961722970009,
      "rewards/forensics_tiou_reward": 0.46293574571609497,
      "step": 239
    },
    {
      "completion_length": 29.046875,
      "epoch": 0.7373271889400922,
      "grad_norm": 4.449642181396484,
      "kl": 0.27734375,
      "learning_rate": 8.16153846153846e-07,
      "loss": 0.1569,
      "reward": 0.6373178064823151,
      "reward_std": 0.1551247462630272,
      "reward_std_all": 0.1558464914560318,
      "rewards-max/forensics_iou_reward": 0.9366898536682129,
      "rewards-max/forensics_strict_edge_reward": 0.11739117279648781,
      "rewards-max/forensics_tiou_reward": 0.9209886491298676,
      "rewards/forensics_iou_reward": 0.45833462476730347,
      "rewards/forensics_strict_edge_reward": 0.01268981653265655,
      "rewards/forensics_tiou_reward": 0.4098396301269531,
      "step": 240
    },
    {
      "completion_length": 30.96875,
      "epoch": 0.7403993855606759,
      "grad_norm": 4.259917736053467,
      "kl": 0.28271484375,
      "learning_rate": 8.153846153846154e-07,
      "loss": 0.1257,
      "reward": 0.7136097848415375,
      "reward_std": 0.11325427889823914,
      "reward_std_all": 0.10605510324239731,
      "rewards-max/forensics_iou_reward": 0.8266356885433197,
      "rewards-max/forensics_strict_edge_reward": 0.1691090576350689,
      "rewards-max/forensics_tiou_reward": 0.8005127310752869,
      "rewards/forensics_iou_reward": 0.5031890273094177,
      "rewards/forensics_strict_edge_reward": 0.011967082973569632,
      "rewards/forensics_tiou_reward": 0.4388902187347412,
      "step": 241
    },
    {
      "completion_length": 26.421875,
      "epoch": 0.7434715821812596,
      "grad_norm": 5.243143081665039,
      "kl": 0.26806640625,
      "learning_rate": 8.146153846153846e-07,
      "loss": 0.1454,
      "reward": 0.6188987195491791,
      "reward_std": 0.14233403652906418,
      "reward_std_all": 0.16346075385808945,
      "rewards-max/forensics_iou_reward": 0.7527129352092743,
      "rewards-max/forensics_strict_edge_reward": 0.19145724177360535,
      "rewards-max/forensics_tiou_reward": 0.7011316418647766,
      "rewards/forensics_iou_reward": 0.39222249388694763,
      "rewards/forensics_strict_edge_reward": 0.011196838226169348,
      "rewards/forensics_tiou_reward": 0.3269919455051422,
      "step": 242
    },
    {
      "completion_length": 25.1875,
      "epoch": 0.7465437788018433,
      "grad_norm": 6.362764358520508,
      "kl": 0.306640625,
      "learning_rate": 8.138461538461538e-07,
      "loss": 0.0963,
      "reward": 0.6940129101276398,
      "reward_std": 0.09370581060647964,
      "reward_std_all": 0.10023434832692146,
      "rewards-max/forensics_iou_reward": 0.8370077610015869,
      "rewards-max/forensics_strict_edge_reward": 0.19007771834731102,
      "rewards-max/forensics_tiou_reward": 0.8127366304397583,
      "rewards/forensics_iou_reward": 0.4956488013267517,
      "rewards/forensics_strict_edge_reward": 0.010325111215934157,
      "rewards/forensics_tiou_reward": 0.4308585375547409,
      "step": 243
    },
    {
      "completion_length": 29.78125,
      "epoch": 0.7496159754224271,
      "grad_norm": 4.936058521270752,
      "kl": 0.296875,
      "learning_rate": 8.13076923076923e-07,
      "loss": 0.1614,
      "reward": 0.5939138829708099,
      "reward_std": 0.16948941349983215,
      "reward_std_all": 0.1814384162425995,
      "rewards-max/forensics_iou_reward": 0.8688475489616394,
      "rewards-max/forensics_strict_edge_reward": 0.1688367947936058,
      "rewards-max/forensics_tiou_reward": 0.8304088115692139,
      "rewards/forensics_iou_reward": 0.3796490579843521,
      "rewards/forensics_strict_edge_reward": 0.011520545464009047,
      "rewards/forensics_tiou_reward": 0.32741886377334595,
      "step": 244
    },
    {
      "completion_length": 24.140625,
      "epoch": 0.7526881720430108,
      "grad_norm": 5.040585041046143,
      "kl": 0.3505859375,
      "learning_rate": 8.123076923076922e-07,
      "loss": 0.0873,
      "reward": 0.5527867376804352,
      "reward_std": 0.15057556331157684,
      "reward_std_all": 0.19578193873167038,
      "rewards-max/forensics_iou_reward": 0.910705029964447,
      "rewards-max/forensics_strict_edge_reward": 0.0832781046628952,
      "rewards-max/forensics_tiou_reward": 0.8790754973888397,
      "rewards/forensics_iou_reward": 0.3986360877752304,
      "rewards/forensics_strict_edge_reward": 0.006431512767449021,
      "rewards/forensics_tiou_reward": 0.348254919052124,
      "step": 245
    },
    {
      "completion_length": 23.5625,
      "epoch": 0.7557603686635944,
      "grad_norm": 4.767037868499756,
      "kl": 0.3564453125,
      "learning_rate": 8.115384615384615e-07,
      "loss": 0.0437,
      "reward": 0.6490560472011566,
      "reward_std": 0.09993396326899529,
      "reward_std_all": 0.12572015449404716,
      "rewards-max/forensics_iou_reward": 0.9533543884754181,
      "rewards-max/forensics_strict_edge_reward": 0.33018553256988525,
      "rewards-max/forensics_tiou_reward": 0.9404765367507935,
      "rewards/forensics_iou_reward": 0.47765618562698364,
      "rewards/forensics_strict_edge_reward": 0.02288992376998067,
      "rewards/forensics_tiou_reward": 0.40533629059791565,
      "step": 246
    },
    {
      "completion_length": 25.90625,
      "epoch": 0.7588325652841782,
      "grad_norm": 4.9993414878845215,
      "kl": 0.359375,
      "learning_rate": 8.107692307692308e-07,
      "loss": 0.175,
      "reward": 0.5642288327217102,
      "reward_std": 0.15818829461932182,
      "reward_std_all": 0.1976645663380623,
      "rewards-max/forensics_iou_reward": 0.9149064421653748,
      "rewards-max/forensics_strict_edge_reward": 0.3983701914548874,
      "rewards-max/forensics_tiou_reward": 0.88092240691185,
      "rewards/forensics_iou_reward": 0.4398011565208435,
      "rewards/forensics_strict_edge_reward": 0.02680296078324318,
      "rewards/forensics_tiou_reward": 0.3856393098831177,
      "step": 247
    },
    {
      "completion_length": 26.109375,
      "epoch": 0.7619047619047619,
      "grad_norm": 5.6372904777526855,
      "kl": 0.314453125,
      "learning_rate": 8.1e-07,
      "loss": 0.0707,
      "reward": 0.6602407097816467,
      "reward_std": 0.15734919160604477,
      "reward_std_all": 0.15125860273838043,
      "rewards-max/forensics_iou_reward": 0.8993397355079651,
      "rewards-max/forensics_strict_edge_reward": 0.20202163606882095,
      "rewards-max/forensics_tiou_reward": 0.8708082437515259,
      "rewards/forensics_iou_reward": 0.4772392362356186,
      "rewards/forensics_strict_edge_reward": 0.015299780759960413,
      "rewards/forensics_tiou_reward": 0.4210665822029114,
      "step": 248
    },
    {
      "completion_length": 27.515625,
      "epoch": 0.7649769585253456,
      "grad_norm": 4.269461154937744,
      "kl": 0.3564453125,
      "learning_rate": 8.092307692307692e-07,
      "loss": 0.1581,
      "reward": 0.643048495054245,
      "reward_std": 0.1649271547794342,
      "reward_std_all": 0.1647695153951645,
      "rewards-max/forensics_iou_reward": 0.8811081945896149,
      "rewards-max/forensics_strict_edge_reward": 0.23371074721217155,
      "rewards-max/forensics_tiou_reward": 0.8564851582050323,
      "rewards/forensics_iou_reward": 0.4697072058916092,
      "rewards/forensics_strict_edge_reward": 0.017044336069375277,
      "rewards/forensics_tiou_reward": 0.4147905856370926,
      "step": 249
    },
    {
      "completion_length": 25.265625,
      "epoch": 0.7680491551459293,
      "grad_norm": 4.165119647979736,
      "kl": 0.3173828125,
      "learning_rate": 8.084615384615384e-07,
      "loss": 0.0271,
      "reward": 0.5717836022377014,
      "reward_std": 0.13690905272960663,
      "reward_std_all": 0.18066194653511047,
      "rewards-max/forensics_iou_reward": 0.9203208982944489,
      "rewards-max/forensics_strict_edge_reward": 0.4368769954890013,
      "rewards-max/forensics_tiou_reward": 0.9024613201618195,
      "rewards/forensics_iou_reward": 0.4212707281112671,
      "rewards/forensics_strict_edge_reward": 0.039447033079341054,
      "rewards/forensics_tiou_reward": 0.3603920340538025,
      "step": 250
    },
    {
      "completion_length": 27.90625,
      "epoch": 0.7711213517665131,
      "grad_norm": 4.401180744171143,
      "kl": 0.3251953125,
      "learning_rate": 8.076923076923077e-07,
      "loss": 0.0582,
      "reward": 0.5453916192054749,
      "reward_std": 0.1602696105837822,
      "reward_std_all": 0.20284772664308548,
      "rewards-max/forensics_iou_reward": 0.9690601527690887,
      "rewards-max/forensics_strict_edge_reward": 0.6710383296012878,
      "rewards-max/forensics_tiou_reward": 0.9624332189559937,
      "rewards/forensics_iou_reward": 0.35851873457431793,
      "rewards/forensics_strict_edge_reward": 0.028669961728155613,
      "rewards/forensics_tiou_reward": 0.3137199133634567,
      "step": 251
    },
    {
      "completion_length": 25.34375,
      "epoch": 0.7741935483870968,
      "grad_norm": 4.642367839813232,
      "kl": 0.30078125,
      "learning_rate": 8.069230769230769e-07,
      "loss": 0.3243,
      "reward": 0.5078972280025482,
      "reward_std": 0.19103555381298065,
      "reward_std_all": 0.2411232441663742,
      "rewards-max/forensics_iou_reward": 0.9334015548229218,
      "rewards-max/forensics_strict_edge_reward": 0.2058773897588253,
      "rewards-max/forensics_tiou_reward": 0.9059178531169891,
      "rewards/forensics_iou_reward": 0.341066837310791,
      "rewards/forensics_strict_edge_reward": 0.011608931003138423,
      "rewards/forensics_tiou_reward": 0.28878558427095413,
      "step": 252
    },
    {
      "completion_length": 29.359375,
      "epoch": 0.7772657450076805,
      "grad_norm": 4.405031681060791,
      "kl": 0.2529296875,
      "learning_rate": 8.06153846153846e-07,
      "loss": 0.1595,
      "reward": 0.6364153623580933,
      "reward_std": 0.16499321162700653,
      "reward_std_all": 0.16545090824365616,
      "rewards-max/forensics_iou_reward": 0.8941890597343445,
      "rewards-max/forensics_strict_edge_reward": 0.05354675091803074,
      "rewards-max/forensics_tiou_reward": 0.866348534822464,
      "rewards/forensics_iou_reward": 0.44366858899593353,
      "rewards/forensics_strict_edge_reward": 0.004969606176018715,
      "rewards/forensics_tiou_reward": 0.3900666832923889,
      "step": 253
    },
    {
      "completion_length": 26.71875,
      "epoch": 0.7803379416282642,
      "grad_norm": 4.3035078048706055,
      "kl": 0.3193359375,
      "learning_rate": 8.053846153846154e-07,
      "loss": 0.1167,
      "reward": 0.4949420690536499,
      "reward_std": 0.16442079842090607,
      "reward_std_all": 0.23468293249607086,
      "rewards-max/forensics_iou_reward": 0.7207811772823334,
      "rewards-max/forensics_strict_edge_reward": 0.0894415593938902,
      "rewards-max/forensics_tiou_reward": 0.6708900928497314,
      "rewards/forensics_iou_reward": 0.29645131528377533,
      "rewards/forensics_strict_edge_reward": 0.004191774285573047,
      "rewards/forensics_tiou_reward": 0.2539491429924965,
      "step": 254
    },
    {
      "completion_length": 24.609375,
      "epoch": 0.783410138248848,
      "grad_norm": 5.621807098388672,
      "kl": 0.2705078125,
      "learning_rate": 8.046153846153846e-07,
      "loss": 0.2154,
      "reward": 0.6287672817707062,
      "reward_std": 0.13784048706293106,
      "reward_std_all": 0.15313978493213654,
      "rewards-max/forensics_iou_reward": 0.9215438663959503,
      "rewards-max/forensics_strict_edge_reward": 0.2749192491173744,
      "rewards-max/forensics_tiou_reward": 0.89615797996521,
      "rewards/forensics_iou_reward": 0.4568815529346466,
      "rewards/forensics_strict_edge_reward": 0.02720398223027587,
      "rewards/forensics_tiou_reward": 0.4084828644990921,
      "step": 255
    },
    {
      "completion_length": 27.140625,
      "epoch": 0.7864823348694316,
      "grad_norm": 4.7966694831848145,
      "kl": 0.2294921875,
      "learning_rate": 8.038461538461538e-07,
      "loss": 0.104,
      "reward": 0.647893637418747,
      "reward_std": 0.1665850132703781,
      "reward_std_all": 0.16022906452417374,
      "rewards-max/forensics_iou_reward": 0.8378232717514038,
      "rewards-max/forensics_strict_edge_reward": 0.2632426470518112,
      "rewards-max/forensics_tiou_reward": 0.7925612926483154,
      "rewards/forensics_iou_reward": 0.4415455460548401,
      "rewards/forensics_strict_edge_reward": 0.01974930427968502,
      "rewards/forensics_tiou_reward": 0.3906589448451996,
      "step": 256
    },
    {
      "completion_length": 27.015625,
      "epoch": 0.7895545314900153,
      "grad_norm": 6.163332462310791,
      "kl": 0.2958984375,
      "learning_rate": 8.03076923076923e-07,
      "loss": 0.1529,
      "reward": 0.6181724965572357,
      "reward_std": 0.1255202814936638,
      "reward_std_all": 0.15046679228544235,
      "rewards-max/forensics_iou_reward": 0.9315547943115234,
      "rewards-max/forensics_strict_edge_reward": 0.5167631208896637,
      "rewards-max/forensics_tiou_reward": 0.8995311558246613,
      "rewards/forensics_iou_reward": 0.44209912419319153,
      "rewards/forensics_strict_edge_reward": 0.027603798545897007,
      "rewards/forensics_tiou_reward": 0.39586664736270905,
      "step": 257
    },
    {
      "completion_length": 26.421875,
      "epoch": 0.7926267281105991,
      "grad_norm": 5.690183162689209,
      "kl": 0.361328125,
      "learning_rate": 8.023076923076922e-07,
      "loss": 0.2492,
      "reward": 0.5956485569477081,
      "reward_std": 0.22509460151195526,
      "reward_std_all": 0.21802770346403122,
      "rewards-max/forensics_iou_reward": 0.8773816227912903,
      "rewards-max/forensics_strict_edge_reward": 0.39113104343414307,
      "rewards-max/forensics_tiou_reward": 0.859332799911499,
      "rewards/forensics_iou_reward": 0.39960652589797974,
      "rewards/forensics_strict_edge_reward": 0.029282958712428808,
      "rewards/forensics_tiou_reward": 0.35819604992866516,
      "step": 258
    },
    {
      "completion_length": 27.640625,
      "epoch": 0.7956989247311828,
      "grad_norm": 4.555129051208496,
      "kl": 0.4619140625,
      "learning_rate": 8.015384615384615e-07,
      "loss": 0.1262,
      "reward": 0.5803945660591125,
      "reward_std": 0.1652454137802124,
      "reward_std_all": 0.1890750676393509,
      "rewards-max/forensics_iou_reward": 0.9491090178489685,
      "rewards-max/forensics_strict_edge_reward": 0.1765490584075451,
      "rewards-max/forensics_tiou_reward": 0.933198094367981,
      "rewards/forensics_iou_reward": 0.4267159253358841,
      "rewards/forensics_strict_edge_reward": 0.024623926728963852,
      "rewards/forensics_tiou_reward": 0.36991676688194275,
      "step": 259
    },
    {
      "completion_length": 26.609375,
      "epoch": 0.7987711213517665,
      "grad_norm": 5.114080905914307,
      "kl": 0.3876953125,
      "learning_rate": 8.007692307692308e-07,
      "loss": 0.1745,
      "reward": 0.667368471622467,
      "reward_std": 0.11908112466335297,
      "reward_std_all": 0.1314096823334694,
      "rewards-max/forensics_iou_reward": 0.9736400544643402,
      "rewards-max/forensics_strict_edge_reward": 0.536435529589653,
      "rewards-max/forensics_tiou_reward": 0.9606031775474548,
      "rewards/forensics_iou_reward": 0.48267944157123566,
      "rewards/forensics_strict_edge_reward": 0.05703994818031788,
      "rewards/forensics_tiou_reward": 0.42724891006946564,
      "step": 260
    },
    {
      "completion_length": 26.25,
      "epoch": 0.8018433179723502,
      "grad_norm": 4.028723239898682,
      "kl": 0.2587890625,
      "learning_rate": 8e-07,
      "loss": 0.0602,
      "reward": 0.6759015619754791,
      "reward_std": 0.09128256142139435,
      "reward_std_all": 0.1104610487818718,
      "rewards-max/forensics_iou_reward": 0.8992149531841278,
      "rewards-max/forensics_strict_edge_reward": 0.2654534410685301,
      "rewards-max/forensics_tiou_reward": 0.8756503760814667,
      "rewards/forensics_iou_reward": 0.4640226662158966,
      "rewards/forensics_strict_edge_reward": 0.022485866909846663,
      "rewards/forensics_tiou_reward": 0.39875245094299316,
      "step": 261
    },
    {
      "completion_length": 23.96875,
      "epoch": 0.804915514592934,
      "grad_norm": 5.632155418395996,
      "kl": 0.3955078125,
      "learning_rate": 7.992307692307692e-07,
      "loss": 0.0967,
      "reward": 0.6580916047096252,
      "reward_std": 0.14332620799541473,
      "reward_std_all": 0.1432834044098854,
      "rewards-max/forensics_iou_reward": 0.9075638055801392,
      "rewards-max/forensics_strict_edge_reward": 0.3415559455752373,
      "rewards-max/forensics_tiou_reward": 0.8787390291690826,
      "rewards/forensics_iou_reward": 0.46894365549087524,
      "rewards/forensics_strict_edge_reward": 0.026407694444060326,
      "rewards/forensics_tiou_reward": 0.4121723622083664,
      "step": 262
    },
    {
      "completion_length": 25.671875,
      "epoch": 0.8079877112135176,
      "grad_norm": 4.172552108764648,
      "kl": 0.251953125,
      "learning_rate": 7.984615384615384e-07,
      "loss": 0.1264,
      "reward": 0.621110737323761,
      "reward_std": 0.1620868369936943,
      "reward_std_all": 0.16588567942380905,
      "rewards-max/forensics_iou_reward": 0.8718145787715912,
      "rewards-max/forensics_strict_edge_reward": 0.15223029628396034,
      "rewards-max/forensics_tiou_reward": 0.8281777203083038,
      "rewards/forensics_iou_reward": 0.45961956679821014,
      "rewards/forensics_strict_edge_reward": 0.011929461499676108,
      "rewards/forensics_tiou_reward": 0.3927667737007141,
      "step": 263
    },
    {
      "completion_length": 26.953125,
      "epoch": 0.8110599078341014,
      "grad_norm": 3.5468738079071045,
      "kl": 0.2578125,
      "learning_rate": 7.976923076923077e-07,
      "loss": 0.1744,
      "reward": 0.6557634174823761,
      "reward_std": 0.1282593533396721,
      "reward_std_all": 0.1342688798904419,
      "rewards-max/forensics_iou_reward": 0.952987939119339,
      "rewards-max/forensics_strict_edge_reward": 0.547403559088707,
      "rewards-max/forensics_tiou_reward": 0.9295930564403534,
      "rewards/forensics_iou_reward": 0.49653396010398865,
      "rewards/forensics_strict_edge_reward": 0.05967644602060318,
      "rewards/forensics_tiou_reward": 0.4390723258256912,
      "step": 264
    },
    {
      "completion_length": 26.21875,
      "epoch": 0.8141321044546851,
      "grad_norm": 4.141299724578857,
      "kl": 0.28125,
      "learning_rate": 7.969230769230769e-07,
      "loss": 0.0162,
      "reward": 0.5571770519018173,
      "reward_std": 0.21384822577238083,
      "reward_std_all": 0.22453612089157104,
      "rewards-max/forensics_iou_reward": 0.7867443263530731,
      "rewards-max/forensics_strict_edge_reward": 0.11919254064559937,
      "rewards-max/forensics_tiou_reward": 0.7411581873893738,
      "rewards/forensics_iou_reward": 0.33866047859191895,
      "rewards/forensics_strict_edge_reward": 0.0058942255564033985,
      "rewards/forensics_tiou_reward": 0.2910624146461487,
      "step": 265
    },
    {
      "completion_length": 26.828125,
      "epoch": 0.8172043010752689,
      "grad_norm": 4.6851887702941895,
      "kl": 0.251953125,
      "learning_rate": 7.96153846153846e-07,
      "loss": 0.0147,
      "reward": 0.6730842888355255,
      "reward_std": 0.12540684640407562,
      "reward_std_all": 0.1262473538517952,
      "rewards-max/forensics_iou_reward": 0.9472531378269196,
      "rewards-max/forensics_strict_edge_reward": 0.39035897701978683,
      "rewards-max/forensics_tiou_reward": 0.9343628287315369,
      "rewards/forensics_iou_reward": 0.5331690907478333,
      "rewards/forensics_strict_edge_reward": 0.02957160584628582,
      "rewards/forensics_tiou_reward": 0.4650528281927109,
      "step": 266
    },
    {
      "completion_length": 35.40625,
      "epoch": 0.8202764976958525,
      "grad_norm": 3.7971253395080566,
      "kl": 0.28125,
      "learning_rate": 7.953846153846154e-07,
      "loss": 0.1406,
      "reward": 0.5958870947360992,
      "reward_std": 0.13507649675011635,
      "reward_std_all": 0.16592448949813843,
      "rewards-max/forensics_iou_reward": 0.9074924290180206,
      "rewards-max/forensics_strict_edge_reward": 0.39378954470157623,
      "rewards-max/forensics_tiou_reward": 0.8683167099952698,
      "rewards/forensics_iou_reward": 0.3747362196445465,
      "rewards/forensics_strict_edge_reward": 0.023480093106627464,
      "rewards/forensics_tiou_reward": 0.3199971914291382,
      "step": 267
    },
    {
      "completion_length": 26.796875,
      "epoch": 0.8233486943164362,
      "grad_norm": 5.495913982391357,
      "kl": 0.326171875,
      "learning_rate": 7.946153846153846e-07,
      "loss": 0.164,
      "reward": 0.5749382674694061,
      "reward_std": 0.18019504845142365,
      "reward_std_all": 0.1996030956506729,
      "rewards-max/forensics_iou_reward": 0.9214439392089844,
      "rewards-max/forensics_strict_edge_reward": 0.205037459731102,
      "rewards-max/forensics_tiou_reward": 0.8891902565956116,
      "rewards/forensics_iou_reward": 0.3909022808074951,
      "rewards/forensics_strict_edge_reward": 0.02070392156019807,
      "rewards/forensics_tiou_reward": 0.3406972885131836,
      "step": 268
    },
    {
      "completion_length": 25.4375,
      "epoch": 0.82642089093702,
      "grad_norm": 5.239014148712158,
      "kl": 0.3642578125,
      "learning_rate": 7.938461538461538e-07,
      "loss": 0.1205,
      "reward": 0.6306195259094238,
      "reward_std": 0.14830894023180008,
      "reward_std_all": 0.16494622826576233,
      "rewards-max/forensics_iou_reward": 0.957344263792038,
      "rewards-max/forensics_strict_edge_reward": 0.25506951101124287,
      "rewards-max/forensics_tiou_reward": 0.9392479360103607,
      "rewards/forensics_iou_reward": 0.46262364089488983,
      "rewards/forensics_strict_edge_reward": 0.018069099634885788,
      "rewards/forensics_tiou_reward": 0.4041335731744766,
      "step": 269
    },
    {
      "completion_length": 28.640625,
      "epoch": 0.8294930875576036,
      "grad_norm": 4.562281608581543,
      "kl": 0.248046875,
      "learning_rate": 7.930769230769231e-07,
      "loss": 0.245,
      "reward": 0.6871094703674316,
      "reward_std": 0.12378105521202087,
      "reward_std_all": 0.11994262784719467,
      "rewards-max/forensics_iou_reward": 0.8881869316101074,
      "rewards-max/forensics_strict_edge_reward": 0.26025206595659256,
      "rewards-max/forensics_tiou_reward": 0.8542794585227966,
      "rewards/forensics_iou_reward": 0.4616285115480423,
      "rewards/forensics_strict_edge_reward": 0.013880725484341383,
      "rewards/forensics_tiou_reward": 0.4084315448999405,
      "step": 270
    },
    {
      "completion_length": 28.9375,
      "epoch": 0.8325652841781874,
      "grad_norm": 4.7191481590271,
      "kl": 0.29736328125,
      "learning_rate": 7.923076923076922e-07,
      "loss": -0.0475,
      "reward": 0.5773130059242249,
      "reward_std": 0.0883794054389,
      "reward_std_all": 0.15648721903562546,
      "rewards-max/forensics_iou_reward": 0.9050700068473816,
      "rewards-max/forensics_strict_edge_reward": 0.2053181678056717,
      "rewards-max/forensics_tiou_reward": 0.8844956755638123,
      "rewards/forensics_iou_reward": 0.3993677347898483,
      "rewards/forensics_strict_edge_reward": 0.013096855022013187,
      "rewards/forensics_tiou_reward": 0.34371770918369293,
      "step": 271
    },
    {
      "completion_length": 26.3125,
      "epoch": 0.8356374807987711,
      "grad_norm": 5.146182060241699,
      "kl": 0.25634765625,
      "learning_rate": 7.915384615384615e-07,
      "loss": 0.1737,
      "reward": 0.7491104304790497,
      "reward_std": 0.08865664526820183,
      "reward_std_all": 0.07944067567586899,
      "rewards-max/forensics_iou_reward": 0.9722900092601776,
      "rewards-max/forensics_strict_edge_reward": 0.4021375998854637,
      "rewards-max/forensics_tiou_reward": 0.9585430324077606,
      "rewards/forensics_iou_reward": 0.6153632998466492,
      "rewards/forensics_strict_edge_reward": 0.020755993900820613,
      "rewards/forensics_tiou_reward": 0.5419601798057556,
      "step": 272
    },
    {
      "completion_length": 26.09375,
      "epoch": 0.8387096774193549,
      "grad_norm": 25.271039962768555,
      "kl": 1.4189453125,
      "learning_rate": 7.907692307692307e-07,
      "loss": 0.1305,
      "reward": 0.6587858498096466,
      "reward_std": 0.09987422823905945,
      "reward_std_all": 0.12158941850066185,
      "rewards-max/forensics_iou_reward": 0.9357240498065948,
      "rewards-max/forensics_strict_edge_reward": 0.34011197090148926,
      "rewards-max/forensics_tiou_reward": 0.9232673943042755,
      "rewards/forensics_iou_reward": 0.4593956768512726,
      "rewards/forensics_strict_edge_reward": 0.023379694670438766,
      "rewards/forensics_tiou_reward": 0.4031582921743393,
      "step": 273
    },
    {
      "completion_length": 26.6875,
      "epoch": 0.8417818740399385,
      "grad_norm": 5.52785062789917,
      "kl": 0.314453125,
      "learning_rate": 7.9e-07,
      "loss": 0.0943,
      "reward": 0.6024500727653503,
      "reward_std": 0.14205768704414368,
      "reward_std_all": 0.1679602861404419,
      "rewards-max/forensics_iou_reward": 0.8736460208892822,
      "rewards-max/forensics_strict_edge_reward": 0.030035708681680262,
      "rewards-max/forensics_tiou_reward": 0.8466441929340363,
      "rewards/forensics_iou_reward": 0.40605978667736053,
      "rewards/forensics_strict_edge_reward": 0.00228981624240987,
      "rewards/forensics_tiou_reward": 0.35246463119983673,
      "step": 274
    },
    {
      "completion_length": 27.65625,
      "epoch": 0.8448540706605223,
      "grad_norm": 4.024782180786133,
      "kl": 0.271484375,
      "learning_rate": 7.892307692307692e-07,
      "loss": 0.1682,
      "reward": 0.6602784097194672,
      "reward_std": 0.10193910449743271,
      "reward_std_all": 0.12265424057841301,
      "rewards-max/forensics_iou_reward": 0.9133521020412445,
      "rewards-max/forensics_strict_edge_reward": 0.2694724574685097,
      "rewards-max/forensics_tiou_reward": 0.8827593326568604,
      "rewards/forensics_iou_reward": 0.49117904901504517,
      "rewards/forensics_strict_edge_reward": 0.024778262712061405,
      "rewards/forensics_tiou_reward": 0.43115387856960297,
      "step": 275
    },
    {
      "completion_length": 24.359375,
      "epoch": 0.847926267281106,
      "grad_norm": 5.441979885101318,
      "kl": 0.314453125,
      "learning_rate": 7.884615384615384e-07,
      "loss": 0.1835,
      "reward": 0.6985437273979187,
      "reward_std": 0.1378973089158535,
      "reward_std_all": 0.12708551809191704,
      "rewards-max/forensics_iou_reward": 0.9361208379268646,
      "rewards-max/forensics_strict_edge_reward": 0.1632341779768467,
      "rewards-max/forensics_tiou_reward": 0.9172089099884033,
      "rewards/forensics_iou_reward": 0.5199223458766937,
      "rewards/forensics_strict_edge_reward": 0.017134284949861467,
      "rewards/forensics_tiou_reward": 0.45249180495738983,
      "step": 276
    },
    {
      "completion_length": 26.625,
      "epoch": 0.8509984639016898,
      "grad_norm": 4.90341854095459,
      "kl": 0.26171875,
      "learning_rate": 7.876923076923077e-07,
      "loss": 0.0484,
      "reward": 0.6800049245357513,
      "reward_std": 0.10525763034820557,
      "reward_std_all": 0.11341610550880432,
      "rewards-max/forensics_iou_reward": 0.9732137620449066,
      "rewards-max/forensics_strict_edge_reward": 0.5918685793876648,
      "rewards-max/forensics_tiou_reward": 0.9673368036746979,
      "rewards/forensics_iou_reward": 0.5033169090747833,
      "rewards/forensics_strict_edge_reward": 0.04706112667918205,
      "rewards/forensics_tiou_reward": 0.4439682960510254,
      "step": 277
    },
    {
      "completion_length": 25.625,
      "epoch": 0.8540706605222734,
      "grad_norm": 4.49712610244751,
      "kl": 0.25146484375,
      "learning_rate": 7.869230769230769e-07,
      "loss": 0.0677,
      "reward": 0.6437914371490479,
      "reward_std": 0.12120967358350754,
      "reward_std_all": 0.1406266652047634,
      "rewards-max/forensics_iou_reward": 0.8279853165149689,
      "rewards-max/forensics_strict_edge_reward": 0.13792927004396915,
      "rewards-max/forensics_tiou_reward": 0.766120582818985,
      "rewards/forensics_iou_reward": 0.43526023626327515,
      "rewards/forensics_strict_edge_reward": 0.008839771617203951,
      "rewards/forensics_tiou_reward": 0.38023553788661957,
      "step": 278
    },
    {
      "completion_length": 23.296875,
      "epoch": 0.8571428571428571,
      "grad_norm": 8.955095291137695,
      "kl": 0.271484375,
      "learning_rate": 7.861538461538461e-07,
      "loss": 0.089,
      "reward": 0.6218127310276031,
      "reward_std": 0.14907556399703026,
      "reward_std_all": 0.1665714681148529,
      "rewards-max/forensics_iou_reward": 0.93057981133461,
      "rewards-max/forensics_strict_edge_reward": 0.3432159423828125,
      "rewards-max/forensics_tiou_reward": 0.911533534526825,
      "rewards/forensics_iou_reward": 0.44555340707302094,
      "rewards/forensics_strict_edge_reward": 0.014517406467348337,
      "rewards/forensics_tiou_reward": 0.3952628821134567,
      "step": 279
    },
    {
      "completion_length": 25.203125,
      "epoch": 0.8602150537634409,
      "grad_norm": 5.419710159301758,
      "kl": 0.28515625,
      "learning_rate": 7.853846153846153e-07,
      "loss": 0.1114,
      "reward": 0.6791531145572662,
      "reward_std": 0.15793897956609726,
      "reward_std_all": 0.14829226583242416,
      "rewards-max/forensics_iou_reward": 0.8908074200153351,
      "rewards-max/forensics_strict_edge_reward": 0.05750522203743458,
      "rewards-max/forensics_tiou_reward": 0.8615337908267975,
      "rewards/forensics_iou_reward": 0.5014424622058868,
      "rewards/forensics_strict_edge_reward": 0.004533415660262108,
      "rewards/forensics_tiou_reward": 0.4345757067203522,
      "step": 280
    },
    {
      "completion_length": 25.0,
      "epoch": 0.8632872503840245,
      "grad_norm": 4.478894233703613,
      "kl": 0.3046875,
      "learning_rate": 7.846153846153846e-07,
      "loss": 0.1171,
      "reward": 0.6732680797576904,
      "reward_std": 0.10190117917954922,
      "reward_std_all": 0.11930075287818909,
      "rewards-max/forensics_iou_reward": 0.8750399351119995,
      "rewards-max/forensics_strict_edge_reward": 0.04089911747723818,
      "rewards-max/forensics_tiou_reward": 0.8518372774124146,
      "rewards/forensics_iou_reward": 0.49669089913368225,
      "rewards/forensics_strict_edge_reward": 0.0049891931703314185,
      "rewards/forensics_tiou_reward": 0.438277468085289,
      "step": 281
    },
    {
      "completion_length": 26.484375,
      "epoch": 0.8663594470046083,
      "grad_norm": 4.937830924987793,
      "kl": 0.25927734375,
      "learning_rate": 7.838461538461539e-07,
      "loss": 0.1648,
      "reward": 0.5957647860050201,
      "reward_std": 0.11744183488190174,
      "reward_std_all": 0.15921466797590256,
      "rewards-max/forensics_iou_reward": 0.8040477931499481,
      "rewards-max/forensics_strict_edge_reward": 0.14029270366881974,
      "rewards-max/forensics_tiou_reward": 0.7405947148799896,
      "rewards/forensics_iou_reward": 0.41607432067394257,
      "rewards/forensics_strict_edge_reward": 0.008403900585108204,
      "rewards/forensics_tiou_reward": 0.34209324419498444,
      "step": 282
    },
    {
      "completion_length": 24.984375,
      "epoch": 0.869431643625192,
      "grad_norm": 4.932636737823486,
      "kl": 0.3974609375,
      "learning_rate": 7.830769230769231e-07,
      "loss": 0.1514,
      "reward": 0.5930664539337158,
      "reward_std": 0.10338829457759857,
      "reward_std_all": 0.16417938470840454,
      "rewards-max/forensics_iou_reward": 0.9782480597496033,
      "rewards-max/forensics_strict_edge_reward": 0.3543456792831421,
      "rewards-max/forensics_tiou_reward": 0.9681394100189209,
      "rewards/forensics_iou_reward": 0.4272416681051254,
      "rewards/forensics_strict_edge_reward": 0.02067735744640231,
      "rewards/forensics_tiou_reward": 0.3805030435323715,
      "step": 283
    },
    {
      "completion_length": 24.390625,
      "epoch": 0.8725038402457758,
      "grad_norm": 4.962324142456055,
      "kl": 0.33203125,
      "learning_rate": 7.823076923076922e-07,
      "loss": 0.1498,
      "reward": 0.7288621664047241,
      "reward_std": 0.07788901031017303,
      "reward_std_all": 0.07667142152786255,
      "rewards-max/forensics_iou_reward": 0.9238847196102142,
      "rewards-max/forensics_strict_edge_reward": 0.3206252455711365,
      "rewards-max/forensics_tiou_reward": 0.908257246017456,
      "rewards/forensics_iou_reward": 0.5241449773311615,
      "rewards/forensics_strict_edge_reward": 0.019289907068014145,
      "rewards/forensics_tiou_reward": 0.4504087120294571,
      "step": 284
    },
    {
      "completion_length": 27.625,
      "epoch": 0.8755760368663594,
      "grad_norm": 4.681191444396973,
      "kl": 0.51171875,
      "learning_rate": 7.815384615384615e-07,
      "loss": 0.1139,
      "reward": 0.6096972227096558,
      "reward_std": 0.12124145030975342,
      "reward_std_all": 0.1548127345740795,
      "rewards-max/forensics_iou_reward": 0.7802633941173553,
      "rewards-max/forensics_strict_edge_reward": 0.25934258103370667,
      "rewards-max/forensics_tiou_reward": 0.7331116497516632,
      "rewards/forensics_iou_reward": 0.40505771338939667,
      "rewards/forensics_strict_edge_reward": 0.024764719186350703,
      "rewards/forensics_tiou_reward": 0.3431834876537323,
      "step": 285
    },
    {
      "completion_length": 26.203125,
      "epoch": 0.8786482334869432,
      "grad_norm": 4.946073055267334,
      "kl": 0.396484375,
      "learning_rate": 7.807692307692307e-07,
      "loss": 0.1267,
      "reward": 0.6054402887821198,
      "reward_std": 0.10845789685845375,
      "reward_std_all": 0.1464487835764885,
      "rewards-max/forensics_iou_reward": 0.9307358264923096,
      "rewards-max/forensics_strict_edge_reward": 0.30069421231746674,
      "rewards-max/forensics_tiou_reward": 0.9043792486190796,
      "rewards/forensics_iou_reward": 0.41890832781791687,
      "rewards/forensics_strict_edge_reward": 0.026400035247206688,
      "rewards/forensics_tiou_reward": 0.3677460104227066,
      "step": 286
    },
    {
      "completion_length": 23.796875,
      "epoch": 0.8817204301075269,
      "grad_norm": 5.278205394744873,
      "kl": 0.3330078125,
      "learning_rate": 7.799999999999999e-07,
      "loss": -0.0248,
      "reward": 0.6544205844402313,
      "reward_std": 0.12341462448239326,
      "reward_std_all": 0.1335332654416561,
      "rewards-max/forensics_iou_reward": 0.9166510701179504,
      "rewards-max/forensics_strict_edge_reward": 0.18874771147966385,
      "rewards-max/forensics_tiou_reward": 0.8761333227157593,
      "rewards/forensics_iou_reward": 0.4730149209499359,
      "rewards/forensics_strict_edge_reward": 0.015460473950952291,
      "rewards/forensics_tiou_reward": 0.4179508090019226,
      "step": 287
    },
    {
      "completion_length": 27.5,
      "epoch": 0.8847926267281107,
      "grad_norm": 5.384127140045166,
      "kl": 0.32421875,
      "learning_rate": 7.792307692307693e-07,
      "loss": 0.0003,
      "reward": 0.5631994009017944,
      "reward_std": 0.10443944856524467,
      "reward_std_all": 0.16728182137012482,
      "rewards-max/forensics_iou_reward": 0.8581432700157166,
      "rewards-max/forensics_strict_edge_reward": 0.1603640541434288,
      "rewards-max/forensics_tiou_reward": 0.8185721337795258,
      "rewards/forensics_iou_reward": 0.3680635541677475,
      "rewards/forensics_strict_edge_reward": 0.00782652385532856,
      "rewards/forensics_tiou_reward": 0.3145681619644165,
      "step": 288
    },
    {
      "completion_length": 24.234375,
      "epoch": 0.8878648233486943,
      "grad_norm": 4.701396942138672,
      "kl": 0.3291015625,
      "learning_rate": 7.784615384615384e-07,
      "loss": 0.1618,
      "reward": 0.6124502420425415,
      "reward_std": 0.2143089473247528,
      "reward_std_all": 0.20407527685165405,
      "rewards-max/forensics_iou_reward": 0.9544573128223419,
      "rewards-max/forensics_strict_edge_reward": 0.33508339524269104,
      "rewards-max/forensics_tiou_reward": 0.9399520754814148,
      "rewards/forensics_iou_reward": 0.4520060271024704,
      "rewards/forensics_strict_edge_reward": 0.02467009937390685,
      "rewards/forensics_tiou_reward": 0.3945499062538147,
      "step": 289
    },
    {
      "completion_length": 30.59375,
      "epoch": 0.890937019969278,
      "grad_norm": 4.409286022186279,
      "kl": 0.2568359375,
      "learning_rate": 7.776923076923077e-07,
      "loss": 0.1352,
      "reward": 0.6267493367195129,
      "reward_std": 0.14626670628786087,
      "reward_std_all": 0.1623149737715721,
      "rewards-max/forensics_iou_reward": 0.866131454706192,
      "rewards-max/forensics_strict_edge_reward": 0.10568396002054214,
      "rewards-max/forensics_tiou_reward": 0.834617555141449,
      "rewards/forensics_iou_reward": 0.46224333345890045,
      "rewards/forensics_strict_edge_reward": 0.006526428507640958,
      "rewards/forensics_tiou_reward": 0.41109447181224823,
      "step": 290
    },
    {
      "completion_length": 25.453125,
      "epoch": 0.8940092165898618,
      "grad_norm": 5.461556911468506,
      "kl": 0.33984375,
      "learning_rate": 7.769230769230769e-07,
      "loss": 0.0527,
      "reward": 0.6760388314723969,
      "reward_std": 0.14011052250862122,
      "reward_std_all": 0.13427652046084404,
      "rewards-max/forensics_iou_reward": 0.9903827607631683,
      "rewards-max/forensics_strict_edge_reward": 0.2440929301083088,
      "rewards-max/forensics_tiou_reward": 0.9848099648952484,
      "rewards/forensics_iou_reward": 0.5382115542888641,
      "rewards/forensics_strict_edge_reward": 0.023356223478913307,
      "rewards/forensics_tiou_reward": 0.4849073141813278,
      "step": 291
    },
    {
      "completion_length": 23.625,
      "epoch": 0.8970814132104454,
      "grad_norm": 6.15521240234375,
      "kl": 0.2783203125,
      "learning_rate": 7.761538461538461e-07,
      "loss": 0.1201,
      "reward": 0.6377768218517303,
      "reward_std": 0.17284902930259705,
      "reward_std_all": 0.16850164532661438,
      "rewards-max/forensics_iou_reward": 0.9434579312801361,
      "rewards-max/forensics_strict_edge_reward": 0.3934844136238098,
      "rewards-max/forensics_tiou_reward": 0.9307830631732941,
      "rewards/forensics_iou_reward": 0.4439314156770706,
      "rewards/forensics_strict_edge_reward": 0.03214411251246929,
      "rewards/forensics_tiou_reward": 0.3790383040904999,
      "step": 292
    },
    {
      "completion_length": 26.015625,
      "epoch": 0.9001536098310292,
      "grad_norm": 3.9645462036132812,
      "kl": 0.341796875,
      "learning_rate": 7.753846153846153e-07,
      "loss": 0.1094,
      "reward": 0.7121661305427551,
      "reward_std": 0.13251801580190659,
      "reward_std_all": 0.12014639750123024,
      "rewards-max/forensics_iou_reward": 0.9176804721355438,
      "rewards-max/forensics_strict_edge_reward": 0.286430224776268,
      "rewards-max/forensics_tiou_reward": 0.8787753283977509,
      "rewards/forensics_iou_reward": 0.529834657907486,
      "rewards/forensics_strict_edge_reward": 0.0181162569206208,
      "rewards/forensics_tiou_reward": 0.45648977160453796,
      "step": 293
    },
    {
      "completion_length": 31.234375,
      "epoch": 0.9032258064516129,
      "grad_norm": 4.844751834869385,
      "kl": 0.3203125,
      "learning_rate": 7.746153846153845e-07,
      "loss": 0.0356,
      "reward": 0.5684879124164581,
      "reward_std": 0.12035461887717247,
      "reward_std_all": 0.17922373861074448,
      "rewards-max/forensics_iou_reward": 0.854781299829483,
      "rewards-max/forensics_strict_edge_reward": 0.15763855539262295,
      "rewards-max/forensics_tiou_reward": 0.8158538639545441,
      "rewards/forensics_iou_reward": 0.408740296959877,
      "rewards/forensics_strict_edge_reward": 0.009133637533523142,
      "rewards/forensics_tiou_reward": 0.34114158153533936,
      "step": 294
    },
    {
      "completion_length": 26.171875,
      "epoch": 0.9062980030721967,
      "grad_norm": 5.276618480682373,
      "kl": 0.26904296875,
      "learning_rate": 7.738461538461539e-07,
      "loss": 0.1182,
      "reward": 0.6558846235275269,
      "reward_std": 0.09687726944684982,
      "reward_std_all": 0.1267629787325859,
      "rewards-max/forensics_iou_reward": 0.9525206089019775,
      "rewards-max/forensics_strict_edge_reward": 0.13723459839820862,
      "rewards-max/forensics_tiou_reward": 0.913235604763031,
      "rewards/forensics_iou_reward": 0.4679497331380844,
      "rewards/forensics_strict_edge_reward": 0.014308203011751175,
      "rewards/forensics_tiou_reward": 0.40660205483436584,
      "step": 295
    },
    {
      "completion_length": 27.234375,
      "epoch": 0.9093701996927803,
      "grad_norm": 4.60231351852417,
      "kl": 0.3330078125,
      "learning_rate": 7.730769230769231e-07,
      "loss": 0.1576,
      "reward": 0.5985935926437378,
      "reward_std": 0.13573186844587326,
      "reward_std_all": 0.17120149731636047,
      "rewards-max/forensics_iou_reward": 0.9195927381515503,
      "rewards-max/forensics_strict_edge_reward": 0.17785484343767166,
      "rewards-max/forensics_tiou_reward": 0.8950310349464417,
      "rewards/forensics_iou_reward": 0.42524999380111694,
      "rewards/forensics_strict_edge_reward": 0.01930570136755705,
      "rewards/forensics_tiou_reward": 0.36519622802734375,
      "step": 296
    },
    {
      "completion_length": 30.640625,
      "epoch": 0.9124423963133641,
      "grad_norm": 6.550318241119385,
      "kl": 0.3095703125,
      "learning_rate": 7.723076923076923e-07,
      "loss": 0.2046,
      "reward": 0.5578553527593613,
      "reward_std": 0.17205599322915077,
      "reward_std_all": 0.2061675861477852,
      "rewards-max/forensics_iou_reward": 0.9051546454429626,
      "rewards-max/forensics_strict_edge_reward": 0.2367031928151846,
      "rewards-max/forensics_tiou_reward": 0.8905333876609802,
      "rewards/forensics_iou_reward": 0.3629450798034668,
      "rewards/forensics_strict_edge_reward": 0.016678729560226202,
      "rewards/forensics_tiou_reward": 0.31278538703918457,
      "step": 297
    },
    {
      "completion_length": 28.609375,
      "epoch": 0.9155145929339478,
      "grad_norm": 5.829387187957764,
      "kl": 0.33984375,
      "learning_rate": 7.715384615384615e-07,
      "loss": 0.0929,
      "reward": 0.6315308511257172,
      "reward_std": 0.16788290441036224,
      "reward_std_all": 0.16644267365336418,
      "rewards-max/forensics_iou_reward": 0.9157074391841888,
      "rewards-max/forensics_strict_edge_reward": 0.15931903198361397,
      "rewards-max/forensics_tiou_reward": 0.8851836323738098,
      "rewards/forensics_iou_reward": 0.45351216197013855,
      "rewards/forensics_strict_edge_reward": 0.017248496296815574,
      "rewards/forensics_tiou_reward": 0.40055878460407257,
      "step": 298
    },
    {
      "completion_length": 27.53125,
      "epoch": 0.9185867895545314,
      "grad_norm": 4.757286548614502,
      "kl": 0.3564453125,
      "learning_rate": 7.707692307692307e-07,
      "loss": 0.0914,
      "reward": 0.6255050897598267,
      "reward_std": 0.07197491731494665,
      "reward_std_all": 0.12274879962205887,
      "rewards-max/forensics_iou_reward": 0.976161539554596,
      "rewards-max/forensics_strict_edge_reward": 0.4963888041675091,
      "rewards-max/forensics_tiou_reward": 0.958388090133667,
      "rewards/forensics_iou_reward": 0.47717761993408203,
      "rewards/forensics_strict_edge_reward": 0.021020642947405577,
      "rewards/forensics_tiou_reward": 0.4311273843050003,
      "step": 299
    },
    {
      "completion_length": 28.03125,
      "epoch": 0.9216589861751152,
      "grad_norm": 4.731748104095459,
      "kl": 0.2880859375,
      "learning_rate": 7.699999999999999e-07,
      "loss": 0.2487,
      "reward": 0.6447167098522186,
      "reward_std": 0.13527998328208923,
      "reward_std_all": 0.1496812254190445,
      "rewards-max/forensics_iou_reward": 0.9169410169124603,
      "rewards-max/forensics_strict_edge_reward": 0.2481778785586357,
      "rewards-max/forensics_tiou_reward": 0.8981022834777832,
      "rewards/forensics_iou_reward": 0.4553678631782532,
      "rewards/forensics_strict_edge_reward": 0.02521608304232359,
      "rewards/forensics_tiou_reward": 0.40534982085227966,
      "step": 300
    },
    {
      "completion_length": 24.484375,
      "epoch": 0.9247311827956989,
      "grad_norm": 4.799578666687012,
      "kl": 0.3125,
      "learning_rate": 7.692307692307693e-07,
      "loss": 0.1164,
      "reward": 0.6809831261634827,
      "reward_std": 0.0921798013150692,
      "reward_std_all": 0.11048144847154617,
      "rewards-max/forensics_iou_reward": 0.9449253976345062,
      "rewards-max/forensics_strict_edge_reward": 0.20527593046426773,
      "rewards-max/forensics_tiou_reward": 0.9081219434738159,
      "rewards/forensics_iou_reward": 0.5106658637523651,
      "rewards/forensics_strict_edge_reward": 0.019698362797498703,
      "rewards/forensics_tiou_reward": 0.4461038410663605,
      "step": 301
    },
    {
      "completion_length": 22.78125,
      "epoch": 0.9278033794162827,
      "grad_norm": 4.47173547744751,
      "kl": 0.3330078125,
      "learning_rate": 7.684615384615384e-07,
      "loss": 0.1097,
      "reward": 0.5990672409534454,
      "reward_std": 0.14403818920254707,
      "reward_std_all": 0.17359931021928787,
      "rewards-max/forensics_iou_reward": 0.9258159697055817,
      "rewards-max/forensics_strict_edge_reward": 0.2723190262913704,
      "rewards-max/forensics_tiou_reward": 0.9005509614944458,
      "rewards/forensics_iou_reward": 0.5007906556129456,
      "rewards/forensics_strict_edge_reward": 0.03128691914025694,
      "rewards/forensics_tiou_reward": 0.4407953917980194,
      "step": 302
    },
    {
      "completion_length": 25.8125,
      "epoch": 0.9308755760368663,
      "grad_norm": 6.49778938293457,
      "kl": 0.2802734375,
      "learning_rate": 7.676923076923077e-07,
      "loss": 0.1594,
      "reward": 0.6766138672828674,
      "reward_std": 0.14579251408576965,
      "reward_std_all": 0.13859029486775398,
      "rewards-max/forensics_iou_reward": 0.9472777545452118,
      "rewards-max/forensics_strict_edge_reward": 0.25162453576922417,
      "rewards-max/forensics_tiou_reward": 0.9141218364238739,
      "rewards/forensics_iou_reward": 0.5009176135063171,
      "rewards/forensics_strict_edge_reward": 0.02291624469216913,
      "rewards/forensics_tiou_reward": 0.446118101477623,
      "step": 303
    },
    {
      "completion_length": 24.546875,
      "epoch": 0.9339477726574501,
      "grad_norm": 5.018259048461914,
      "kl": 0.318359375,
      "learning_rate": 7.669230769230769e-07,
      "loss": 0.1035,
      "reward": 0.658437043428421,
      "reward_std": 0.16550658643245697,
      "reward_std_all": 0.157168447971344,
      "rewards-max/forensics_iou_reward": 0.9134979546070099,
      "rewards-max/forensics_strict_edge_reward": 0.4369185008108616,
      "rewards-max/forensics_tiou_reward": 0.8877617418766022,
      "rewards/forensics_iou_reward": 0.49796512722969055,
      "rewards/forensics_strict_edge_reward": 0.028912223875522614,
      "rewards/forensics_tiou_reward": 0.44799911975860596,
      "step": 304
    },
    {
      "completion_length": 24.140625,
      "epoch": 0.9370199692780338,
      "grad_norm": 5.372569561004639,
      "kl": 0.32421875,
      "learning_rate": 7.661538461538461e-07,
      "loss": 0.1138,
      "reward": 0.5703432857990265,
      "reward_std": 0.16943491995334625,
      "reward_std_all": 0.20239891111850739,
      "rewards-max/forensics_iou_reward": 0.9602669179439545,
      "rewards-max/forensics_strict_edge_reward": 0.21196052804589272,
      "rewards-max/forensics_tiou_reward": 0.9386211335659027,
      "rewards/forensics_iou_reward": 0.4488829970359802,
      "rewards/forensics_strict_edge_reward": 0.01420343853533268,
      "rewards/forensics_tiou_reward": 0.4049952030181885,
      "step": 305
    },
    {
      "completion_length": 25.96875,
      "epoch": 0.9400921658986175,
      "grad_norm": 5.2107439041137695,
      "kl": 0.2744140625,
      "learning_rate": 7.653846153846153e-07,
      "loss": 0.0132,
      "reward": 0.5829049348831177,
      "reward_std": 0.163441251963377,
      "reward_std_all": 0.18785995990037918,
      "rewards-max/forensics_iou_reward": 0.9000371396541595,
      "rewards-max/forensics_strict_edge_reward": 0.12345781177282333,
      "rewards-max/forensics_tiou_reward": 0.8750698268413544,
      "rewards/forensics_iou_reward": 0.4181412607431412,
      "rewards/forensics_strict_edge_reward": 0.007836558390408754,
      "rewards/forensics_tiou_reward": 0.36424393951892853,
      "step": 306
    },
    {
      "completion_length": 25.140625,
      "epoch": 0.9431643625192012,
      "grad_norm": 5.20530366897583,
      "kl": 0.291015625,
      "learning_rate": 7.646153846153845e-07,
      "loss": 0.1091,
      "reward": 0.7029237747192383,
      "reward_std": 0.1046278290450573,
      "reward_std_all": 0.10326610878109932,
      "rewards-max/forensics_iou_reward": 0.8502512574195862,
      "rewards-max/forensics_strict_edge_reward": 0.267203958472237,
      "rewards-max/forensics_tiou_reward": 0.8086514174938202,
      "rewards/forensics_iou_reward": 0.48717200756073,
      "rewards/forensics_strict_edge_reward": 0.020867863961029798,
      "rewards/forensics_tiou_reward": 0.4239548593759537,
      "step": 307
    },
    {
      "completion_length": 27.265625,
      "epoch": 0.946236559139785,
      "grad_norm": 6.050049781799316,
      "kl": 0.4130859375,
      "learning_rate": 7.638461538461539e-07,
      "loss": 0.159,
      "reward": 0.5706203579902649,
      "reward_std": 0.10625646263360977,
      "reward_std_all": 0.16615941375494003,
      "rewards-max/forensics_iou_reward": 0.9556576311588287,
      "rewards-max/forensics_strict_edge_reward": 0.2972389757633209,
      "rewards-max/forensics_tiou_reward": 0.9465419352054596,
      "rewards/forensics_iou_reward": 0.43747514486312866,
      "rewards/forensics_strict_edge_reward": 0.015646771993488073,
      "rewards/forensics_tiou_reward": 0.38764502108097076,
      "step": 308
    },
    {
      "completion_length": 27.25,
      "epoch": 0.9493087557603687,
      "grad_norm": 5.323083877563477,
      "kl": 0.24267578125,
      "learning_rate": 7.630769230769231e-07,
      "loss": 0.0528,
      "reward": 0.6560548543930054,
      "reward_std": 0.1339862272143364,
      "reward_std_all": 0.14306408166885376,
      "rewards-max/forensics_iou_reward": 0.9302127361297607,
      "rewards-max/forensics_strict_edge_reward": 0.11621283460408449,
      "rewards-max/forensics_tiou_reward": 0.9056243002414703,
      "rewards/forensics_iou_reward": 0.48659947514533997,
      "rewards/forensics_strict_edge_reward": 0.013079004012979567,
      "rewards/forensics_tiou_reward": 0.4271360784769058,
      "step": 309
    },
    {
      "completion_length": 27.734375,
      "epoch": 0.9523809523809523,
      "grad_norm": 5.096236228942871,
      "kl": 0.2646484375,
      "learning_rate": 7.623076923076923e-07,
      "loss": 0.164,
      "reward": 0.7049401998519897,
      "reward_std": 0.1089000292122364,
      "reward_std_all": 0.10482081212103367,
      "rewards-max/forensics_iou_reward": 0.9233466386795044,
      "rewards-max/forensics_strict_edge_reward": 0.16466598212718964,
      "rewards-max/forensics_tiou_reward": 0.8936831057071686,
      "rewards/forensics_iou_reward": 0.5102517008781433,
      "rewards/forensics_strict_edge_reward": 0.022284044651314616,
      "rewards/forensics_tiou_reward": 0.4374263733625412,
      "step": 310
    },
    {
      "completion_length": 26.859375,
      "epoch": 0.9554531490015361,
      "grad_norm": 4.450344085693359,
      "kl": 0.28125,
      "learning_rate": 7.615384615384615e-07,
      "loss": -0.0338,
      "reward": 0.5919316411018372,
      "reward_std": 0.1676870957016945,
      "reward_std_all": 0.18327542394399643,
      "rewards-max/forensics_iou_reward": 0.9364270865917206,
      "rewards-max/forensics_strict_edge_reward": 0.29388464987277985,
      "rewards-max/forensics_tiou_reward": 0.9225465655326843,
      "rewards/forensics_iou_reward": 0.4165644645690918,
      "rewards/forensics_strict_edge_reward": 0.024369585793465376,
      "rewards/forensics_tiou_reward": 0.3599425107240677,
      "step": 311
    },
    {
      "completion_length": 25.78125,
      "epoch": 0.9585253456221198,
      "grad_norm": 4.847181797027588,
      "kl": 0.3779296875,
      "learning_rate": 7.607692307692307e-07,
      "loss": 0.1338,
      "reward": 0.7036131322383881,
      "reward_std": 0.07599114440381527,
      "reward_std_all": 0.08837134391069412,
      "rewards-max/forensics_iou_reward": 0.9375963807106018,
      "rewards-max/forensics_strict_edge_reward": 0.38765060901641846,
      "rewards-max/forensics_tiou_reward": 0.9196044206619263,
      "rewards/forensics_iou_reward": 0.5242145657539368,
      "rewards/forensics_strict_edge_reward": 0.030524183064699173,
      "rewards/forensics_tiou_reward": 0.4637179672718048,
      "step": 312
    },
    {
      "completion_length": 25.3125,
      "epoch": 0.9615975422427036,
      "grad_norm": 5.70274543762207,
      "kl": 0.4697265625,
      "learning_rate": 7.599999999999999e-07,
      "loss": 0.155,
      "reward": 0.566803514957428,
      "reward_std": 0.1536557674407959,
      "reward_std_all": 0.18784942477941513,
      "rewards-max/forensics_iou_reward": 0.9141510725021362,
      "rewards-max/forensics_strict_edge_reward": 0.2873297519981861,
      "rewards-max/forensics_tiou_reward": 0.8821091651916504,
      "rewards/forensics_iou_reward": 0.4019906222820282,
      "rewards/forensics_strict_edge_reward": 0.0238357987254858,
      "rewards/forensics_tiou_reward": 0.35246293246746063,
      "step": 313
    },
    {
      "completion_length": 22.84375,
      "epoch": 0.9646697388632872,
      "grad_norm": 6.785489559173584,
      "kl": 0.27685546875,
      "learning_rate": 7.592307692307692e-07,
      "loss": 0.2552,
      "reward": 0.6018435955047607,
      "reward_std": 0.15866340696811676,
      "reward_std_all": 0.1803078055381775,
      "rewards-max/forensics_iou_reward": 0.8880398273468018,
      "rewards-max/forensics_strict_edge_reward": 0.20525825396180153,
      "rewards-max/forensics_tiou_reward": 0.8609002232551575,
      "rewards/forensics_iou_reward": 0.42049503326416016,
      "rewards/forensics_strict_edge_reward": 0.014016882283613086,
      "rewards/forensics_tiou_reward": 0.36036644876003265,
      "step": 314
    },
    {
      "completion_length": 25.984375,
      "epoch": 0.967741935483871,
      "grad_norm": 6.588428974151611,
      "kl": 0.2998046875,
      "learning_rate": 7.584615384615384e-07,
      "loss": 0.168,
      "reward": 0.6248716413974762,
      "reward_std": 0.18339256197214127,
      "reward_std_all": 0.18367186188697815,
      "rewards-max/forensics_iou_reward": 0.9158351719379425,
      "rewards-max/forensics_strict_edge_reward": 0.21376504004001617,
      "rewards-max/forensics_tiou_reward": 0.9005416035652161,
      "rewards/forensics_iou_reward": 0.4713594913482666,
      "rewards/forensics_strict_edge_reward": 0.02086418028920889,
      "rewards/forensics_tiou_reward": 0.4122525304555893,
      "step": 315
    },
    {
      "completion_length": 29.390625,
      "epoch": 0.9708141321044547,
      "grad_norm": 5.127074241638184,
      "kl": 0.2880859375,
      "learning_rate": 7.576923076923077e-07,
      "loss": 0.1127,
      "reward": 0.6547266840934753,
      "reward_std": 0.14034434035420418,
      "reward_std_all": 0.13978241384029388,
      "rewards-max/forensics_iou_reward": 0.9156233072280884,
      "rewards-max/forensics_strict_edge_reward": 0.435134157538414,
      "rewards-max/forensics_tiou_reward": 0.8816012442111969,
      "rewards/forensics_iou_reward": 0.4574274867773056,
      "rewards/forensics_strict_edge_reward": 0.02099964488297701,
      "rewards/forensics_tiou_reward": 0.4028649479150772,
      "step": 316
    },
    {
      "completion_length": 24.984375,
      "epoch": 0.9738863287250384,
      "grad_norm": 5.677720069885254,
      "kl": 0.302734375,
      "learning_rate": 7.569230769230769e-07,
      "loss": 0.1938,
      "reward": 0.6820032596588135,
      "reward_std": 0.13277825713157654,
      "reward_std_all": 0.12740132957696915,
      "rewards-max/forensics_iou_reward": 0.917346179485321,
      "rewards-max/forensics_strict_edge_reward": 0.3412692993879318,
      "rewards-max/forensics_tiou_reward": 0.8991129398345947,
      "rewards/forensics_iou_reward": 0.5175939202308655,
      "rewards/forensics_strict_edge_reward": 0.024643155746161938,
      "rewards/forensics_tiou_reward": 0.4611753821372986,
      "step": 317
    },
    {
      "completion_length": 25.984375,
      "epoch": 0.9769585253456221,
      "grad_norm": 5.369248390197754,
      "kl": 0.376953125,
      "learning_rate": 7.561538461538461e-07,
      "loss": 0.243,
      "reward": 0.6137372255325317,
      "reward_std": 0.17906038463115692,
      "reward_std_all": 0.18926743417978287,
      "rewards-max/forensics_iou_reward": 0.8999571800231934,
      "rewards-max/forensics_strict_edge_reward": 0.18334465473890305,
      "rewards-max/forensics_tiou_reward": 0.8724161982536316,
      "rewards/forensics_iou_reward": 0.4541601836681366,
      "rewards/forensics_strict_edge_reward": 0.017436412628740072,
      "rewards/forensics_tiou_reward": 0.4086136668920517,
      "step": 318
    },
    {
      "completion_length": 26.125,
      "epoch": 0.9800307219662059,
      "grad_norm": 4.121177673339844,
      "kl": 0.29345703125,
      "learning_rate": 7.553846153846154e-07,
      "loss": 0.1159,
      "reward": 0.6272352337837219,
      "reward_std": 0.1459747478365898,
      "reward_std_all": 0.15767960250377655,
      "rewards-max/forensics_iou_reward": 0.9130028188228607,
      "rewards-max/forensics_strict_edge_reward": 0.2266138717532158,
      "rewards-max/forensics_tiou_reward": 0.8871938288211823,
      "rewards/forensics_iou_reward": 0.4465431720018387,
      "rewards/forensics_strict_edge_reward": 0.018869644030928612,
      "rewards/forensics_tiou_reward": 0.397488608956337,
      "step": 319
    },
    {
      "completion_length": 25.703125,
      "epoch": 0.9831029185867896,
      "grad_norm": 4.141445636749268,
      "kl": 0.3095703125,
      "learning_rate": 7.546153846153845e-07,
      "loss": 0.1458,
      "reward": 0.7480201125144958,
      "reward_std": 0.05479382164776325,
      "reward_std_all": 0.05628768168389797,
      "rewards-max/forensics_iou_reward": 0.9277571737766266,
      "rewards-max/forensics_strict_edge_reward": 0.2431606389582157,
      "rewards-max/forensics_tiou_reward": 0.9041071832180023,
      "rewards/forensics_iou_reward": 0.5710091590881348,
      "rewards/forensics_strict_edge_reward": 0.017054743133485317,
      "rewards/forensics_tiou_reward": 0.5036123842000961,
      "step": 320
    },
    {
      "completion_length": 26.484375,
      "epoch": 0.9861751152073732,
      "grad_norm": 4.380834579467773,
      "kl": 0.3359375,
      "learning_rate": 7.538461538461538e-07,
      "loss": 0.1024,
      "reward": 0.6695278584957123,
      "reward_std": 0.14208004623651505,
      "reward_std_all": 0.13765359669923782,
      "rewards-max/forensics_iou_reward": 0.9372658431529999,
      "rewards-max/forensics_strict_edge_reward": 0.3432159423828125,
      "rewards-max/forensics_tiou_reward": 0.9207672774791718,
      "rewards/forensics_iou_reward": 0.4771345257759094,
      "rewards/forensics_strict_edge_reward": 0.0241501796990633,
      "rewards/forensics_tiou_reward": 0.4134266972541809,
      "step": 321
    },
    {
      "completion_length": 27.09375,
      "epoch": 0.989247311827957,
      "grad_norm": 5.460525035858154,
      "kl": 0.3017578125,
      "learning_rate": 7.530769230769231e-07,
      "loss": 0.1818,
      "reward": 0.5828295052051544,
      "reward_std": 0.1501273736357689,
      "reward_std_all": 0.1827269345521927,
      "rewards-max/forensics_iou_reward": 0.9203079342842102,
      "rewards-max/forensics_strict_edge_reward": 0.2911505401134491,
      "rewards-max/forensics_tiou_reward": 0.8971579074859619,
      "rewards/forensics_iou_reward": 0.407784566283226,
      "rewards/forensics_strict_edge_reward": 0.02363352570682764,
      "rewards/forensics_tiou_reward": 0.3480824828147888,
      "step": 322
    },
    {
      "completion_length": 25.640625,
      "epoch": 0.9923195084485407,
      "grad_norm": 4.720059394836426,
      "kl": 0.4404296875,
      "learning_rate": 7.523076923076923e-07,
      "loss": 0.1054,
      "reward": 0.5935072898864746,
      "reward_std": 0.11924709752202034,
      "reward_std_all": 0.16229983419179916,
      "rewards-max/forensics_iou_reward": 0.927510529756546,
      "rewards-max/forensics_strict_edge_reward": 0.20898250862956047,
      "rewards-max/forensics_tiou_reward": 0.9077260792255402,
      "rewards/forensics_iou_reward": 0.41692790389060974,
      "rewards/forensics_strict_edge_reward": 0.015567688271403313,
      "rewards/forensics_tiou_reward": 0.36391448974609375,
      "step": 323
    },
    {
      "completion_length": 23.984375,
      "epoch": 0.9953917050691244,
      "grad_norm": 6.067975997924805,
      "kl": 0.36328125,
      "learning_rate": 7.515384615384615e-07,
      "loss": 0.1806,
      "reward": 0.7006658017635345,
      "reward_std": 0.11800489202141762,
      "reward_std_all": 0.11601320654153824,
      "rewards-max/forensics_iou_reward": 0.9773939847946167,
      "rewards-max/forensics_strict_edge_reward": 0.5470698475837708,
      "rewards-max/forensics_tiou_reward": 0.9706441462039948,
      "rewards/forensics_iou_reward": 0.580429345369339,
      "rewards/forensics_strict_edge_reward": 0.059791794046759605,
      "rewards/forensics_tiou_reward": 0.5311107635498047,
      "step": 324
    },
    {
      "completion_length": 27.234375,
      "epoch": 0.9984639016897081,
      "grad_norm": 5.368013858795166,
      "kl": 0.349609375,
      "learning_rate": 7.507692307692307e-07,
      "loss": 0.1252,
      "reward": 0.6976352334022522,
      "reward_std": 0.08800298348069191,
      "reward_std_all": 0.09690163657069206,
      "rewards-max/forensics_iou_reward": 0.8320682048797607,
      "rewards-max/forensics_strict_edge_reward": 0.10560822300612926,
      "rewards-max/forensics_tiou_reward": 0.8027742803096771,
      "rewards/forensics_iou_reward": 0.4875895828008652,
      "rewards/forensics_strict_edge_reward": 0.005787632428109646,
      "rewards/forensics_tiou_reward": 0.43739959597587585,
      "step": 325
    },
    {
      "completion_length": 18.80000114440918,
      "epoch": 1.0,
      "grad_norm": 3.026134967803955,
      "kl": 0.29296875,
      "learning_rate": 7.5e-07,
      "loss": 0.0925,
      "reward": 0.517526388168335,
      "reward_std": 0.09962587803602219,
      "reward_std_all": 0.18489493429660797,
      "rewards-max/forensics_iou_reward": 0.9016133546829224,
      "rewards-max/forensics_strict_edge_reward": 1.1272469237155747e-05,
      "rewards-max/forensics_tiou_reward": 0.8577871918678284,
      "rewards/forensics_iou_reward": 0.35843682289123535,
      "rewards/forensics_strict_edge_reward": 3.0782543944951613e-06,
      "rewards/forensics_tiou_reward": 0.3045940399169922,
      "step": 326
    },
    {
      "completion_length": 28.734375,
      "epoch": 1.0030721966205838,
      "grad_norm": 5.181419849395752,
      "kl": 0.3154296875,
      "learning_rate": 7.492307692307692e-07,
      "loss": 0.266,
      "reward": 0.6182974576950073,
      "reward_std": 0.11721571534872055,
      "reward_std_all": 0.14766644686460495,
      "rewards-max/forensics_iou_reward": 0.9201797842979431,
      "rewards-max/forensics_strict_edge_reward": 0.20459160208702087,
      "rewards-max/forensics_tiou_reward": 0.8999118208885193,
      "rewards/forensics_iou_reward": 0.4472507834434509,
      "rewards/forensics_strict_edge_reward": 0.013208567281253636,
      "rewards/forensics_tiou_reward": 0.38882264494895935,
      "step": 327
    },
    {
      "completion_length": 25.59375,
      "epoch": 1.0061443932411673,
      "grad_norm": 6.5638203620910645,
      "kl": 0.3349609375,
      "learning_rate": 7.484615384615384e-07,
      "loss": 0.1603,
      "reward": 0.587489053606987,
      "reward_std": 0.13864705711603165,
      "reward_std_all": 0.18439077585935593,
      "rewards-max/forensics_iou_reward": 0.9815655648708344,
      "rewards-max/forensics_strict_edge_reward": 0.5925651341676712,
      "rewards-max/forensics_tiou_reward": 0.9730634987354279,
      "rewards/forensics_iou_reward": 0.46620604395866394,
      "rewards/forensics_strict_edge_reward": 0.04931095987558365,
      "rewards/forensics_tiou_reward": 0.42720502614974976,
      "step": 328
    },
    {
      "completion_length": 28.9375,
      "epoch": 1.0092165898617511,
      "grad_norm": 4.781493663787842,
      "kl": 0.357421875,
      "learning_rate": 7.476923076923077e-07,
      "loss": 0.2108,
      "reward": 0.6885753870010376,
      "reward_std": 0.1004723384976387,
      "reward_std_all": 0.10334033146500587,
      "rewards-max/forensics_iou_reward": 0.9850185811519623,
      "rewards-max/forensics_strict_edge_reward": 0.3018669933080673,
      "rewards-max/forensics_tiou_reward": 0.9810612499713898,
      "rewards/forensics_iou_reward": 0.5433582663536072,
      "rewards/forensics_strict_edge_reward": 0.022225359454751015,
      "rewards/forensics_tiou_reward": 0.4754832088947296,
      "step": 329
    },
    {
      "completion_length": 26.828125,
      "epoch": 1.012288786482335,
      "grad_norm": 4.513436317443848,
      "kl": 0.3447265625,
      "learning_rate": 7.469230769230769e-07,
      "loss": 0.1479,
      "reward": 0.6670734584331512,
      "reward_std": 0.11104244366288185,
      "reward_std_all": 0.12211953103542328,
      "rewards-max/forensics_iou_reward": 0.9254308342933655,
      "rewards-max/forensics_strict_edge_reward": 0.11999992281198502,
      "rewards-max/forensics_tiou_reward": 0.906541109085083,
      "rewards/forensics_iou_reward": 0.5053602904081345,
      "rewards/forensics_strict_edge_reward": 0.011992265935987234,
      "rewards/forensics_tiou_reward": 0.44404152035713196,
      "step": 330
    },
    {
      "completion_length": 28.625,
      "epoch": 1.0153609831029187,
      "grad_norm": 5.38625431060791,
      "kl": 0.341796875,
      "learning_rate": 7.461538461538461e-07,
      "loss": 0.1142,
      "reward": 0.6596730947494507,
      "reward_std": 0.07360683754086494,
      "reward_std_all": 0.11175478622317314,
      "rewards-max/forensics_iou_reward": 0.9730328023433685,
      "rewards-max/forensics_strict_edge_reward": 0.29335225746035576,
      "rewards-max/forensics_tiou_reward": 0.9557672142982483,
      "rewards/forensics_iou_reward": 0.5075802803039551,
      "rewards/forensics_strict_edge_reward": 0.024195270147174597,
      "rewards/forensics_tiou_reward": 0.45327696204185486,
      "step": 331
    },
    {
      "completion_length": 27.640625,
      "epoch": 1.0184331797235022,
      "grad_norm": 4.554852485656738,
      "kl": 0.296875,
      "learning_rate": 7.453846153846154e-07,
      "loss": 0.0707,
      "reward": 0.5962306261062622,
      "reward_std": 0.12683333456516266,
      "reward_std_all": 0.16793136671185493,
      "rewards-max/forensics_iou_reward": 0.9456632733345032,
      "rewards-max/forensics_strict_edge_reward": 0.229200785048306,
      "rewards-max/forensics_tiou_reward": 0.9141506850719452,
      "rewards/forensics_iou_reward": 0.4248644709587097,
      "rewards/forensics_strict_edge_reward": 0.01759334868984297,
      "rewards/forensics_tiou_reward": 0.36829350888729095,
      "step": 332
    },
    {
      "completion_length": 24.65625,
      "epoch": 1.021505376344086,
      "grad_norm": 4.871731758117676,
      "kl": 0.3212890625,
      "learning_rate": 7.446153846153845e-07,
      "loss": 0.1007,
      "reward": 0.6986538171768188,
      "reward_std": 0.10235099121928215,
      "reward_std_all": 0.10264533385634422,
      "rewards-max/forensics_iou_reward": 0.9443061053752899,
      "rewards-max/forensics_strict_edge_reward": 0.17262601107358932,
      "rewards-max/forensics_tiou_reward": 0.9226212203502655,
      "rewards/forensics_iou_reward": 0.5211266875267029,
      "rewards/forensics_strict_edge_reward": 0.01461168727837503,
      "rewards/forensics_tiou_reward": 0.4615246057510376,
      "step": 333
    },
    {
      "completion_length": 29.296875,
      "epoch": 1.0245775729646698,
      "grad_norm": 4.8856706619262695,
      "kl": 0.3115234375,
      "learning_rate": 7.438461538461538e-07,
      "loss": 0.1529,
      "reward": 0.6019526124000549,
      "reward_std": 0.1479923203587532,
      "reward_std_all": 0.1709250956773758,
      "rewards-max/forensics_iou_reward": 0.9932957887649536,
      "rewards-max/forensics_strict_edge_reward": 0.9211947023868561,
      "rewards-max/forensics_tiou_reward": 0.9920476377010345,
      "rewards/forensics_iou_reward": 0.4588107168674469,
      "rewards/forensics_strict_edge_reward": 0.07968547940254211,
      "rewards/forensics_tiou_reward": 0.4151465594768524,
      "step": 334
    },
    {
      "completion_length": 26.984375,
      "epoch": 1.0276497695852536,
      "grad_norm": 4.67686128616333,
      "kl": 0.3251953125,
      "learning_rate": 7.43076923076923e-07,
      "loss": -0.0276,
      "reward": 0.6268685162067413,
      "reward_std": 0.07870391011238098,
      "reward_std_all": 0.12664270401000977,
      "rewards-max/forensics_iou_reward": 0.9602825939655304,
      "rewards-max/forensics_strict_edge_reward": 0.4427416920661926,
      "rewards-max/forensics_tiou_reward": 0.9443954527378082,
      "rewards/forensics_iou_reward": 0.47609128057956696,
      "rewards/forensics_strict_edge_reward": 0.031903814524412155,
      "rewards/forensics_tiou_reward": 0.4275786280632019,
      "step": 335
    },
    {
      "completion_length": 30.203125,
      "epoch": 1.0307219662058371,
      "grad_norm": 5.0078654289245605,
      "kl": 0.3291015625,
      "learning_rate": 7.423076923076923e-07,
      "loss": 0.0693,
      "reward": 0.7216479182243347,
      "reward_std": 0.07662538439035416,
      "reward_std_all": 0.0849379450082779,
      "rewards-max/forensics_iou_reward": 0.946815013885498,
      "rewards-max/forensics_strict_edge_reward": 0.30885880440473557,
      "rewards-max/forensics_tiou_reward": 0.9252052903175354,
      "rewards/forensics_iou_reward": 0.5534360706806183,
      "rewards/forensics_strict_edge_reward": 0.0237793056294322,
      "rewards/forensics_tiou_reward": 0.49311594665050507,
      "step": 336
    },
    {
      "completion_length": 23.546875,
      "epoch": 1.033794162826421,
      "grad_norm": 5.719845294952393,
      "kl": 0.44921875,
      "learning_rate": 7.415384615384616e-07,
      "loss": 0.2194,
      "reward": 0.6851388514041901,
      "reward_std": 0.09449143335223198,
      "reward_std_all": 0.10748769342899323,
      "rewards-max/forensics_iou_reward": 0.9454935491085052,
      "rewards-max/forensics_strict_edge_reward": 0.27228695154190063,
      "rewards-max/forensics_tiou_reward": 0.9319276213645935,
      "rewards/forensics_iou_reward": 0.4908487945795059,
      "rewards/forensics_strict_edge_reward": 0.0232488876208663,
      "rewards/forensics_tiou_reward": 0.41730719804763794,
      "step": 337
    },
    {
      "completion_length": 25.328125,
      "epoch": 1.0368663594470047,
      "grad_norm": 4.837698459625244,
      "kl": 0.3408203125,
      "learning_rate": 7.407692307692307e-07,
      "loss": 0.1637,
      "reward": 0.6504988372325897,
      "reward_std": 0.1188659593462944,
      "reward_std_all": 0.13805614411830902,
      "rewards-max/forensics_iou_reward": 0.9412408471107483,
      "rewards-max/forensics_strict_edge_reward": 0.030305764521472156,
      "rewards-max/forensics_tiou_reward": 0.909209281206131,
      "rewards/forensics_iou_reward": 0.4661286175251007,
      "rewards/forensics_strict_edge_reward": 0.0016079373308457434,
      "rewards/forensics_tiou_reward": 0.38284240663051605,
      "step": 338
    },
    {
      "completion_length": 26.9375,
      "epoch": 1.0399385560675882,
      "grad_norm": 5.481497764587402,
      "kl": 0.27734375,
      "learning_rate": 7.4e-07,
      "loss": 0.1612,
      "reward": 0.6777791678905487,
      "reward_std": 0.1441504880785942,
      "reward_std_all": 0.13783183693885803,
      "rewards-max/forensics_iou_reward": 0.9677584171295166,
      "rewards-max/forensics_strict_edge_reward": 0.4964817464351654,
      "rewards-max/forensics_tiou_reward": 0.9580804407596588,
      "rewards/forensics_iou_reward": 0.5592396110296249,
      "rewards/forensics_strict_edge_reward": 0.042500246316194534,
      "rewards/forensics_tiou_reward": 0.49802011251449585,
      "step": 339
    },
    {
      "completion_length": 27.328125,
      "epoch": 1.043010752688172,
      "grad_norm": 4.976227283477783,
      "kl": 0.2919921875,
      "learning_rate": 7.392307692307692e-07,
      "loss": 0.0631,
      "reward": 0.7257958650588989,
      "reward_std": 0.09745461121201515,
      "reward_std_all": 0.09168065711855888,
      "rewards-max/forensics_iou_reward": 0.9408477544784546,
      "rewards-max/forensics_strict_edge_reward": 0.4385186433792114,
      "rewards-max/forensics_tiou_reward": 0.9258729219436646,
      "rewards/forensics_iou_reward": 0.5228666961193085,
      "rewards/forensics_strict_edge_reward": 0.03115787636488676,
      "rewards/forensics_tiou_reward": 0.4468979388475418,
      "step": 340
    },
    {
      "completion_length": 23.6875,
      "epoch": 1.0460829493087558,
      "grad_norm": 4.285808563232422,
      "kl": 0.29296875,
      "learning_rate": 7.384615384615384e-07,
      "loss": 0.0939,
      "reward": 0.7016782760620117,
      "reward_std": 0.07654663175344467,
      "reward_std_all": 0.09395278617739677,
      "rewards-max/forensics_iou_reward": 0.9683758318424225,
      "rewards-max/forensics_strict_edge_reward": 0.4086042046546936,
      "rewards-max/forensics_tiou_reward": 0.9552508890628815,
      "rewards/forensics_iou_reward": 0.5554543733596802,
      "rewards/forensics_strict_edge_reward": 0.020583425648510456,
      "rewards/forensics_tiou_reward": 0.49355366826057434,
      "step": 341
    },
    {
      "completion_length": 27.5625,
      "epoch": 1.0491551459293396,
      "grad_norm": 4.653296947479248,
      "kl": 0.3232421875,
      "learning_rate": 7.376923076923076e-07,
      "loss": 0.0889,
      "reward": 0.6506225168704987,
      "reward_std": 0.15740066766738892,
      "reward_std_all": 0.15416159108281136,
      "rewards-max/forensics_iou_reward": 0.9387403428554535,
      "rewards-max/forensics_strict_edge_reward": 0.4085611943155527,
      "rewards-max/forensics_tiou_reward": 0.9097488820552826,
      "rewards/forensics_iou_reward": 0.4758751392364502,
      "rewards/forensics_strict_edge_reward": 0.019467639736831188,
      "rewards/forensics_tiou_reward": 0.4239942133426666,
      "step": 342
    },
    {
      "completion_length": 28.5,
      "epoch": 1.0522273425499231,
      "grad_norm": 4.99662971496582,
      "kl": 0.3466796875,
      "learning_rate": 7.369230769230769e-07,
      "loss": 0.1646,
      "reward": 0.6163907647132874,
      "reward_std": 0.12026020139455795,
      "reward_std_all": 0.15759557485580444,
      "rewards-max/forensics_iou_reward": 0.9978794157505035,
      "rewards-max/forensics_strict_edge_reward": 0.8333333432674408,
      "rewards-max/forensics_tiou_reward": 0.9973620772361755,
      "rewards/forensics_iou_reward": 0.48101815581321716,
      "rewards/forensics_strict_edge_reward": 0.07975563034415245,
      "rewards/forensics_tiou_reward": 0.4547930359840393,
      "step": 343
    },
    {
      "completion_length": 26.71875,
      "epoch": 1.055299539170507,
      "grad_norm": 5.045009136199951,
      "kl": 0.3076171875,
      "learning_rate": 7.361538461538461e-07,
      "loss": 0.1503,
      "reward": 0.6598545014858246,
      "reward_std": 0.07984042540192604,
      "reward_std_all": 0.11247415468096733,
      "rewards-max/forensics_iou_reward": 0.9651764929294586,
      "rewards-max/forensics_strict_edge_reward": 0.6355590522289276,
      "rewards-max/forensics_tiou_reward": 0.9502190947532654,
      "rewards/forensics_iou_reward": 0.4970192164182663,
      "rewards/forensics_strict_edge_reward": 0.04016128648072481,
      "rewards/forensics_tiou_reward": 0.44811540842056274,
      "step": 344
    },
    {
      "completion_length": 25.09375,
      "epoch": 1.0583717357910907,
      "grad_norm": 7.936818599700928,
      "kl": 0.3291015625,
      "learning_rate": 7.353846153846154e-07,
      "loss": -0.0623,
      "reward": 0.737260639667511,
      "reward_std": 0.059243980795145035,
      "reward_std_all": 0.06188058294355869,
      "rewards-max/forensics_iou_reward": 0.9513199925422668,
      "rewards-max/forensics_strict_edge_reward": 0.401060089468956,
      "rewards-max/forensics_tiou_reward": 0.9300880134105682,
      "rewards/forensics_iou_reward": 0.626918613910675,
      "rewards/forensics_strict_edge_reward": 0.042897085659205914,
      "rewards/forensics_tiou_reward": 0.5695928335189819,
      "step": 345
    },
    {
      "completion_length": 26.6875,
      "epoch": 1.0614439324116745,
      "grad_norm": 5.286695957183838,
      "kl": 0.3505859375,
      "learning_rate": 7.346153846153846e-07,
      "loss": 0.0818,
      "reward": 0.7466453313827515,
      "reward_std": 0.050025250762701035,
      "reward_std_all": 0.05308623984456062,
      "rewards-max/forensics_iou_reward": 0.9482509195804596,
      "rewards-max/forensics_strict_edge_reward": 0.2927777022123337,
      "rewards-max/forensics_tiou_reward": 0.9345672130584717,
      "rewards/forensics_iou_reward": 0.600050538778305,
      "rewards/forensics_strict_edge_reward": 0.026101259514689445,
      "rewards/forensics_tiou_reward": 0.5161907970905304,
      "step": 346
    },
    {
      "completion_length": 25.484375,
      "epoch": 1.064516129032258,
      "grad_norm": 5.494155406951904,
      "kl": 0.3046875,
      "learning_rate": 7.338461538461538e-07,
      "loss": 0.09,
      "reward": 0.7385982275009155,
      "reward_std": 0.08799631893634796,
      "reward_std_all": 0.0817649345844984,
      "rewards-max/forensics_iou_reward": 0.9514105319976807,
      "rewards-max/forensics_strict_edge_reward": 0.3467385023832321,
      "rewards-max/forensics_tiou_reward": 0.9241865277290344,
      "rewards/forensics_iou_reward": 0.5923957228660583,
      "rewards/forensics_strict_edge_reward": 0.03153565339744091,
      "rewards/forensics_tiou_reward": 0.526244193315506,
      "step": 347
    },
    {
      "completion_length": 24.734375,
      "epoch": 1.0675883256528418,
      "grad_norm": 5.242945671081543,
      "kl": 0.3779296875,
      "learning_rate": 7.33076923076923e-07,
      "loss": 0.0416,
      "reward": 0.6555297076702118,
      "reward_std": 0.05604681745171547,
      "reward_std_all": 0.10080708377063274,
      "rewards-max/forensics_iou_reward": 0.9569478332996368,
      "rewards-max/forensics_strict_edge_reward": 0.3560166209936142,
      "rewards-max/forensics_tiou_reward": 0.9429133236408234,
      "rewards/forensics_iou_reward": 0.5003628730773926,
      "rewards/forensics_strict_edge_reward": 0.02344309352338314,
      "rewards/forensics_tiou_reward": 0.45194390416145325,
      "step": 348
    },
    {
      "completion_length": 25.296875,
      "epoch": 1.0706605222734256,
      "grad_norm": 5.809732913970947,
      "kl": 0.3251953125,
      "learning_rate": 7.323076923076923e-07,
      "loss": 0.0419,
      "reward": 0.700203001499176,
      "reward_std": 0.07859614491462708,
      "reward_std_all": 0.09015700407326221,
      "rewards-max/forensics_iou_reward": 0.9367758333683014,
      "rewards-max/forensics_strict_edge_reward": 0.24964904505759478,
      "rewards-max/forensics_tiou_reward": 0.9192878901958466,
      "rewards/forensics_iou_reward": 0.5755559802055359,
      "rewards/forensics_strict_edge_reward": 0.015580260893329978,
      "rewards/forensics_tiou_reward": 0.5233669877052307,
      "step": 349
    },
    {
      "completion_length": 25.921875,
      "epoch": 1.0737327188940091,
      "grad_norm": 7.405223846435547,
      "kl": 0.3203125,
      "learning_rate": 7.315384615384616e-07,
      "loss": 0.1055,
      "reward": 0.6827880442142487,
      "reward_std": 0.08321498520672321,
      "reward_std_all": 0.10573352500796318,
      "rewards-max/forensics_iou_reward": 0.9760775566101074,
      "rewards-max/forensics_strict_edge_reward": 0.5854001492261887,
      "rewards-max/forensics_tiou_reward": 0.9686932265758514,
      "rewards/forensics_iou_reward": 0.6054860949516296,
      "rewards/forensics_strict_edge_reward": 0.05278575513511896,
      "rewards/forensics_tiou_reward": 0.5546073168516159,
      "step": 350
    },
    {
      "completion_length": 29.71875,
      "epoch": 1.076804915514593,
      "grad_norm": 6.552340984344482,
      "kl": 0.4580078125,
      "learning_rate": 7.307692307692307e-07,
      "loss": 0.0762,
      "reward": 0.6489191055297852,
      "reward_std": 0.12954793870449066,
      "reward_std_all": 0.14241661131381989,
      "rewards-max/forensics_iou_reward": 0.95792156457901,
      "rewards-max/forensics_strict_edge_reward": 0.46168768405914307,
      "rewards-max/forensics_tiou_reward": 0.9349387288093567,
      "rewards/forensics_iou_reward": 0.503116711974144,
      "rewards/forensics_strict_edge_reward": 0.04718577302992344,
      "rewards/forensics_tiou_reward": 0.4646502435207367,
      "step": 351
    },
    {
      "completion_length": 26.0625,
      "epoch": 1.0798771121351767,
      "grad_norm": 6.972104549407959,
      "kl": 0.3671875,
      "learning_rate": 7.3e-07,
      "loss": 0.0649,
      "reward": 0.6216030418872833,
      "reward_std": 0.06931589357554913,
      "reward_std_all": 0.12325618416070938,
      "rewards-max/forensics_iou_reward": 0.9639157950878143,
      "rewards-max/forensics_strict_edge_reward": 0.2864609733223915,
      "rewards-max/forensics_tiou_reward": 0.9434744417667389,
      "rewards/forensics_iou_reward": 0.4657323807477951,
      "rewards/forensics_strict_edge_reward": 0.013639328768476844,
      "rewards/forensics_tiou_reward": 0.4106491804122925,
      "step": 352
    },
    {
      "completion_length": 25.21875,
      "epoch": 1.0829493087557605,
      "grad_norm": 6.257562160491943,
      "kl": 0.3828125,
      "learning_rate": 7.292307692307692e-07,
      "loss": 0.1907,
      "reward": 0.6664153933525085,
      "reward_std": 0.14388053119182587,
      "reward_std_all": 0.14233437925577164,
      "rewards-max/forensics_iou_reward": 0.9372038543224335,
      "rewards-max/forensics_strict_edge_reward": 0.06155669689178467,
      "rewards-max/forensics_tiou_reward": 0.9168443977832794,
      "rewards/forensics_iou_reward": 0.49897706508636475,
      "rewards/forensics_strict_edge_reward": 0.007594612892717123,
      "rewards/forensics_tiou_reward": 0.442454069852829,
      "step": 353
    },
    {
      "completion_length": 26.453125,
      "epoch": 1.086021505376344,
      "grad_norm": 5.20625114440918,
      "kl": 0.361328125,
      "learning_rate": 7.284615384615384e-07,
      "loss": 0.1267,
      "reward": 0.6328364908695221,
      "reward_std": 0.13448325544595718,
      "reward_std_all": 0.15209773555397987,
      "rewards-max/forensics_iou_reward": 0.8642052412033081,
      "rewards-max/forensics_strict_edge_reward": 0.13114038482308388,
      "rewards-max/forensics_tiou_reward": 0.833762526512146,
      "rewards/forensics_iou_reward": 0.41727274656295776,
      "rewards/forensics_strict_edge_reward": 0.006640736246481538,
      "rewards/forensics_tiou_reward": 0.35947853326797485,
      "step": 354
    },
    {
      "completion_length": 26.28125,
      "epoch": 1.0890937019969278,
      "grad_norm": 4.6722917556762695,
      "kl": 0.435546875,
      "learning_rate": 7.276923076923076e-07,
      "loss": 0.1129,
      "reward": 0.6284451484680176,
      "reward_std": 0.09167583100497723,
      "reward_std_all": 0.1322268806397915,
      "rewards-max/forensics_iou_reward": 0.9433096349239349,
      "rewards-max/forensics_strict_edge_reward": 0.11273747030645609,
      "rewards-max/forensics_tiou_reward": 0.9250865876674652,
      "rewards/forensics_iou_reward": 0.5432338118553162,
      "rewards/forensics_strict_edge_reward": 0.015948643907904625,
      "rewards/forensics_tiou_reward": 0.4796673655509949,
      "step": 355
    },
    {
      "completion_length": 26.828125,
      "epoch": 1.0921658986175116,
      "grad_norm": 8.087933540344238,
      "kl": 0.30859375,
      "learning_rate": 7.269230769230769e-07,
      "loss": 0.1667,
      "reward": 0.6375557780265808,
      "reward_std": 0.12811143323779106,
      "reward_std_all": 0.14756474643945694,
      "rewards-max/forensics_iou_reward": 0.8321272134780884,
      "rewards-max/forensics_strict_edge_reward": 0.0863406864227727,
      "rewards-max/forensics_tiou_reward": 0.7744722068309784,
      "rewards/forensics_iou_reward": 0.4311215579509735,
      "rewards/forensics_strict_edge_reward": 0.004845793908316409,
      "rewards/forensics_tiou_reward": 0.373974934220314,
      "step": 356
    },
    {
      "completion_length": 28.734375,
      "epoch": 1.0952380952380953,
      "grad_norm": 5.53927755355835,
      "kl": 0.3115234375,
      "learning_rate": 7.261538461538461e-07,
      "loss": 0.0746,
      "reward": 0.7408724427223206,
      "reward_std": 0.07670662179589272,
      "reward_std_all": 0.07299032807350159,
      "rewards-max/forensics_iou_reward": 0.925111711025238,
      "rewards-max/forensics_strict_edge_reward": 0.380371056497097,
      "rewards-max/forensics_tiou_reward": 0.8996535539627075,
      "rewards/forensics_iou_reward": 0.61080601811409,
      "rewards/forensics_strict_edge_reward": 0.028257804922759533,
      "rewards/forensics_tiou_reward": 0.5532075762748718,
      "step": 357
    },
    {
      "completion_length": 26.703125,
      "epoch": 1.098310291858679,
      "grad_norm": 5.392632484436035,
      "kl": 0.328125,
      "learning_rate": 7.253846153846154e-07,
      "loss": 0.1118,
      "reward": 0.6691882610321045,
      "reward_std": 0.10987690836191177,
      "reward_std_all": 0.1260334551334381,
      "rewards-max/forensics_iou_reward": 0.9893224537372589,
      "rewards-max/forensics_strict_edge_reward": 0.7580138444900513,
      "rewards-max/forensics_tiou_reward": 0.9847047626972198,
      "rewards/forensics_iou_reward": 0.49044401943683624,
      "rewards/forensics_strict_edge_reward": 0.058150889351964,
      "rewards/forensics_tiou_reward": 0.4317081719636917,
      "step": 358
    },
    {
      "completion_length": 26.90625,
      "epoch": 1.1013824884792627,
      "grad_norm": 4.6182146072387695,
      "kl": 0.3203125,
      "learning_rate": 7.246153846153846e-07,
      "loss": 0.0071,
      "reward": 0.7012999951839447,
      "reward_std": 0.11531546525657177,
      "reward_std_all": 0.11489555146545172,
      "rewards-max/forensics_iou_reward": 0.9230867326259613,
      "rewards-max/forensics_strict_edge_reward": 0.34409867227077484,
      "rewards-max/forensics_tiou_reward": 0.9042111039161682,
      "rewards/forensics_iou_reward": 0.5209736824035645,
      "rewards/forensics_strict_edge_reward": 0.0215502199716866,
      "rewards/forensics_tiou_reward": 0.45476675033569336,
      "step": 359
    },
    {
      "completion_length": 24.796875,
      "epoch": 1.1044546850998465,
      "grad_norm": 5.944077014923096,
      "kl": 0.4375,
      "learning_rate": 7.238461538461538e-07,
      "loss": 0.1592,
      "reward": 0.7275583744049072,
      "reward_std": 0.07014560140669346,
      "reward_std_all": 0.0757329910993576,
      "rewards-max/forensics_iou_reward": 0.9887732565402985,
      "rewards-max/forensics_strict_edge_reward": 0.33736710250377655,
      "rewards-max/forensics_tiou_reward": 0.9814418852329254,
      "rewards/forensics_iou_reward": 0.5630441606044769,
      "rewards/forensics_strict_edge_reward": 0.020248742774128914,
      "rewards/forensics_tiou_reward": 0.4981839060783386,
      "step": 360
    },
    {
      "completion_length": 25.09375,
      "epoch": 1.10752688172043,
      "grad_norm": 5.411959648132324,
      "kl": 0.3564453125,
      "learning_rate": 7.23076923076923e-07,
      "loss": 0.1557,
      "reward": 0.6980282962322235,
      "reward_std": 0.10551998764276505,
      "reward_std_all": 0.106391541659832,
      "rewards-max/forensics_iou_reward": 0.9328840970993042,
      "rewards-max/forensics_strict_edge_reward": 0.39796263352036476,
      "rewards-max/forensics_tiou_reward": 0.90975221991539,
      "rewards/forensics_iou_reward": 0.5208893120288849,
      "rewards/forensics_strict_edge_reward": 0.030378964729607105,
      "rewards/forensics_tiou_reward": 0.457116037607193,
      "step": 361
    },
    {
      "completion_length": 30.296875,
      "epoch": 1.1105990783410138,
      "grad_norm": 4.673335552215576,
      "kl": 0.3408203125,
      "learning_rate": 7.223076923076922e-07,
      "loss": 0.1735,
      "reward": 0.6703773736953735,
      "reward_std": 0.1086144708096981,
      "reward_std_all": 0.11693403869867325,
      "rewards-max/forensics_iou_reward": 0.9807926118373871,
      "rewards-max/forensics_strict_edge_reward": 0.5205644369125366,
      "rewards-max/forensics_tiou_reward": 0.9765179455280304,
      "rewards/forensics_iou_reward": 0.47853976488113403,
      "rewards/forensics_strict_edge_reward": 0.03400464542210102,
      "rewards/forensics_tiou_reward": 0.4280690848827362,
      "step": 362
    },
    {
      "completion_length": 27.0625,
      "epoch": 1.1136712749615976,
      "grad_norm": 5.480749130249023,
      "kl": 0.4345703125,
      "learning_rate": 7.215384615384616e-07,
      "loss": 0.1793,
      "reward": 0.6673795878887177,
      "reward_std": 0.11067326739430428,
      "reward_std_all": 0.13098105788230896,
      "rewards-max/forensics_iou_reward": 0.9235333204269409,
      "rewards-max/forensics_strict_edge_reward": 0.22973909229040146,
      "rewards-max/forensics_tiou_reward": 0.899935781955719,
      "rewards/forensics_iou_reward": 0.5510047078132629,
      "rewards/forensics_strict_edge_reward": 0.025188494473695755,
      "rewards/forensics_tiou_reward": 0.507035955786705,
      "step": 363
    },
    {
      "completion_length": 27.53125,
      "epoch": 1.1167434715821813,
      "grad_norm": 4.382523536682129,
      "kl": 0.4716796875,
      "learning_rate": 7.207692307692307e-07,
      "loss": -0.0186,
      "reward": 0.6993107497692108,
      "reward_std": 0.04110816493630409,
      "reward_std_all": 0.07178555987775326,
      "rewards-max/forensics_iou_reward": 0.9679413735866547,
      "rewards-max/forensics_strict_edge_reward": 0.5953177660703659,
      "rewards-max/forensics_tiou_reward": 0.9591753780841827,
      "rewards/forensics_iou_reward": 0.5463508665561676,
      "rewards/forensics_strict_edge_reward": 0.048158248886466026,
      "rewards/forensics_tiou_reward": 0.5006009042263031,
      "step": 364
    },
    {
      "completion_length": 23.53125,
      "epoch": 1.119815668202765,
      "grad_norm": 6.317041873931885,
      "kl": 0.3251953125,
      "learning_rate": 7.2e-07,
      "loss": 0.1017,
      "reward": 0.6746923923492432,
      "reward_std": 0.1428893655538559,
      "reward_std_all": 0.13699080795049667,
      "rewards-max/forensics_iou_reward": 0.9508595764636993,
      "rewards-max/forensics_strict_edge_reward": 0.17747382819652557,
      "rewards-max/forensics_tiou_reward": 0.9383438229560852,
      "rewards/forensics_iou_reward": 0.5151604413986206,
      "rewards/forensics_strict_edge_reward": 0.01909230975434184,
      "rewards/forensics_tiou_reward": 0.44194208085536957,
      "step": 365
    },
    {
      "completion_length": 23.8125,
      "epoch": 1.1228878648233487,
      "grad_norm": 7.436250686645508,
      "kl": 0.3935546875,
      "learning_rate": 7.192307692307692e-07,
      "loss": 0.0878,
      "reward": 0.6774152219295502,
      "reward_std": 0.0909146573394537,
      "reward_std_all": 0.10906577855348587,
      "rewards-max/forensics_iou_reward": 0.8921104967594147,
      "rewards-max/forensics_strict_edge_reward": 0.1253891233354807,
      "rewards-max/forensics_tiou_reward": 0.8567657172679901,
      "rewards/forensics_iou_reward": 0.534772202372551,
      "rewards/forensics_strict_edge_reward": 0.01312240357947303,
      "rewards/forensics_tiou_reward": 0.47618332505226135,
      "step": 366
    },
    {
      "completion_length": 27.109375,
      "epoch": 1.1259600614439325,
      "grad_norm": 5.7571210861206055,
      "kl": 0.3505859375,
      "learning_rate": 7.184615384615384e-07,
      "loss": 0.0853,
      "reward": 0.71578049659729,
      "reward_std": 0.09812110289931297,
      "reward_std_all": 0.10015853121876717,
      "rewards-max/forensics_iou_reward": 0.8832264244556427,
      "rewards-max/forensics_strict_edge_reward": 0.36647093296051025,
      "rewards-max/forensics_tiou_reward": 0.8581278026103973,
      "rewards/forensics_iou_reward": 0.5469577014446259,
      "rewards/forensics_strict_edge_reward": 0.03238959889858961,
      "rewards/forensics_tiou_reward": 0.4946732819080353,
      "step": 367
    },
    {
      "completion_length": 27.828125,
      "epoch": 1.129032258064516,
      "grad_norm": 5.745141506195068,
      "kl": 0.31005859375,
      "learning_rate": 7.176923076923077e-07,
      "loss": 0.1252,
      "reward": 0.6538509130477905,
      "reward_std": 0.14279965311288834,
      "reward_std_all": 0.1421680897474289,
      "rewards-max/forensics_iou_reward": 0.9384523332118988,
      "rewards-max/forensics_strict_edge_reward": 0.21167714521288872,
      "rewards-max/forensics_tiou_reward": 0.9188855886459351,
      "rewards/forensics_iou_reward": 0.45373208820819855,
      "rewards/forensics_strict_edge_reward": 0.01701620640233159,
      "rewards/forensics_tiou_reward": 0.3947412222623825,
      "step": 368
    },
    {
      "completion_length": 28.625,
      "epoch": 1.1321044546850998,
      "grad_norm": 4.908281326293945,
      "kl": 0.3330078125,
      "learning_rate": 7.169230769230768e-07,
      "loss": 0.1611,
      "reward": 0.6531899571418762,
      "reward_std": 0.10060055553913116,
      "reward_std_all": 0.12992813810706139,
      "rewards-max/forensics_iou_reward": 0.9634775221347809,
      "rewards-max/forensics_strict_edge_reward": 0.4054059237241745,
      "rewards-max/forensics_tiou_reward": 0.9354294836521149,
      "rewards/forensics_iou_reward": 0.5359242260456085,
      "rewards/forensics_strict_edge_reward": 0.02938995324075222,
      "rewards/forensics_tiou_reward": 0.49096736311912537,
      "step": 369
    },
    {
      "completion_length": 28.328125,
      "epoch": 1.1351766513056836,
      "grad_norm": 7.162904739379883,
      "kl": 0.4765625,
      "learning_rate": 7.161538461538461e-07,
      "loss": 0.1656,
      "reward": 0.6855986416339874,
      "reward_std": 0.09210175462067127,
      "reward_std_all": 0.10442740097641945,
      "rewards-max/forensics_iou_reward": 0.8640615344047546,
      "rewards-max/forensics_strict_edge_reward": 0.3399255871772766,
      "rewards-max/forensics_tiou_reward": 0.8345925211906433,
      "rewards/forensics_iou_reward": 0.5005155354738235,
      "rewards/forensics_strict_edge_reward": 0.029848560690879822,
      "rewards/forensics_tiou_reward": 0.44304700195789337,
      "step": 370
    },
    {
      "completion_length": 23.875,
      "epoch": 1.1382488479262673,
      "grad_norm": 6.210855007171631,
      "kl": 0.4560546875,
      "learning_rate": 7.153846153846154e-07,
      "loss": 0.0565,
      "reward": 0.6416936814785004,
      "reward_std": 0.13386185839772224,
      "reward_std_all": 0.15173491835594177,
      "rewards-max/forensics_iou_reward": 0.9216943085193634,
      "rewards-max/forensics_strict_edge_reward": 0.31253018975257874,
      "rewards-max/forensics_tiou_reward": 0.9025176465511322,
      "rewards/forensics_iou_reward": 0.48630809783935547,
      "rewards/forensics_strict_edge_reward": 0.01875001471489668,
      "rewards/forensics_tiou_reward": 0.42662279307842255,
      "step": 371
    },
    {
      "completion_length": 28.75,
      "epoch": 1.141321044546851,
      "grad_norm": 5.857396602630615,
      "kl": 0.333984375,
      "learning_rate": 7.146153846153846e-07,
      "loss": 0.169,
      "reward": 0.6586109697818756,
      "reward_std": 0.17778772115707397,
      "reward_std_all": 0.16597609966993332,
      "rewards-max/forensics_iou_reward": 0.9242377579212189,
      "rewards-max/forensics_strict_edge_reward": 0.21804559230804443,
      "rewards-max/forensics_tiou_reward": 0.9013288617134094,
      "rewards/forensics_iou_reward": 0.5233640670776367,
      "rewards/forensics_strict_edge_reward": 0.014550544321537018,
      "rewards/forensics_tiou_reward": 0.47467567026615143,
      "step": 372
    },
    {
      "completion_length": 26.109375,
      "epoch": 1.1443932411674347,
      "grad_norm": 4.881885051727295,
      "kl": 0.400390625,
      "learning_rate": 7.138461538461538e-07,
      "loss": 0.0089,
      "reward": 0.7129403948783875,
      "reward_std": 0.12243838980793953,
      "reward_std_all": 0.11301771551370621,
      "rewards-max/forensics_iou_reward": 0.9176328182220459,
      "rewards-max/forensics_strict_edge_reward": 0.2619753032922745,
      "rewards-max/forensics_tiou_reward": 0.9020480513572693,
      "rewards/forensics_iou_reward": 0.5505021214485168,
      "rewards/forensics_strict_edge_reward": 0.021014251746237278,
      "rewards/forensics_tiou_reward": 0.49511362612247467,
      "step": 373
    },
    {
      "completion_length": 25.578125,
      "epoch": 1.1474654377880185,
      "grad_norm": 5.414888858795166,
      "kl": 0.45703125,
      "learning_rate": 7.13076923076923e-07,
      "loss": 0.0777,
      "reward": 0.614096611738205,
      "reward_std": 0.13829674012959003,
      "reward_std_all": 0.16569653525948524,
      "rewards-max/forensics_iou_reward": 0.942716658115387,
      "rewards-max/forensics_strict_edge_reward": 0.5495419353246689,
      "rewards-max/forensics_tiou_reward": 0.926045298576355,
      "rewards/forensics_iou_reward": 0.4781370162963867,
      "rewards/forensics_strict_edge_reward": 0.05361126735806465,
      "rewards/forensics_tiou_reward": 0.42762626707553864,
      "step": 374
    },
    {
      "completion_length": 29.296875,
      "epoch": 1.1505376344086022,
      "grad_norm": 4.439238548278809,
      "kl": 0.271484375,
      "learning_rate": 7.123076923076922e-07,
      "loss": 0.1229,
      "reward": 0.672029584646225,
      "reward_std": 0.1507391408085823,
      "reward_std_all": 0.14433559775352478,
      "rewards-max/forensics_iou_reward": 0.8515720367431641,
      "rewards-max/forensics_strict_edge_reward": 0.18133402988314629,
      "rewards-max/forensics_tiou_reward": 0.8051484227180481,
      "rewards/forensics_iou_reward": 0.4906162917613983,
      "rewards/forensics_strict_edge_reward": 0.017487567383795977,
      "rewards/forensics_tiou_reward": 0.4439588040113449,
      "step": 375
    },
    {
      "completion_length": 29.4375,
      "epoch": 1.1536098310291858,
      "grad_norm": 5.116947650909424,
      "kl": 0.3193359375,
      "learning_rate": 7.115384615384616e-07,
      "loss": 0.03,
      "reward": 0.6072134375572205,
      "reward_std": 0.12522033601999283,
      "reward_std_all": 0.15734677016735077,
      "rewards-max/forensics_iou_reward": 0.929254025220871,
      "rewards-max/forensics_strict_edge_reward": 0.24121301621198654,
      "rewards-max/forensics_tiou_reward": 0.9104019403457642,
      "rewards/forensics_iou_reward": 0.4380485266447067,
      "rewards/forensics_strict_edge_reward": 0.012934351805597544,
      "rewards/forensics_tiou_reward": 0.3810907006263733,
      "step": 376
    },
    {
      "completion_length": 28.3125,
      "epoch": 1.1566820276497696,
      "grad_norm": 5.790285587310791,
      "kl": 0.3232421875,
      "learning_rate": 7.107692307692308e-07,
      "loss": 0.0029,
      "reward": 0.6974495053291321,
      "reward_std": 0.10223827324807644,
      "reward_std_all": 0.10612921789288521,
      "rewards-max/forensics_iou_reward": 0.9215195775032043,
      "rewards-max/forensics_strict_edge_reward": 0.23682181164622307,
      "rewards-max/forensics_tiou_reward": 0.9065398871898651,
      "rewards/forensics_iou_reward": 0.4987029731273651,
      "rewards/forensics_strict_edge_reward": 0.027614694787189364,
      "rewards/forensics_tiou_reward": 0.4370187073945999,
      "step": 377
    },
    {
      "completion_length": 26.28125,
      "epoch": 1.1597542242703534,
      "grad_norm": 4.962922096252441,
      "kl": 0.3623046875,
      "learning_rate": 7.1e-07,
      "loss": 0.2299,
      "reward": 0.7236630022525787,
      "reward_std": 0.0745040774345398,
      "reward_std_all": 0.08080527186393738,
      "rewards-max/forensics_iou_reward": 0.9737837612628937,
      "rewards-max/forensics_strict_edge_reward": 0.5889939665794373,
      "rewards-max/forensics_tiou_reward": 0.965134859085083,
      "rewards/forensics_iou_reward": 0.5721753537654877,
      "rewards/forensics_strict_edge_reward": 0.055990759283304214,
      "rewards/forensics_tiou_reward": 0.49659618735313416,
      "step": 378
    },
    {
      "completion_length": 24.328125,
      "epoch": 1.1628264208909371,
      "grad_norm": 5.643325328826904,
      "kl": 0.421875,
      "learning_rate": 7.092307692307692e-07,
      "loss": 0.049,
      "reward": 0.6399545073509216,
      "reward_std": 0.1757364198565483,
      "reward_std_all": 0.17063534259796143,
      "rewards-max/forensics_iou_reward": 0.9184992909431458,
      "rewards-max/forensics_strict_edge_reward": 0.4021375998854637,
      "rewards-max/forensics_tiou_reward": 0.8998909890651703,
      "rewards/forensics_iou_reward": 0.45801201462745667,
      "rewards/forensics_strict_edge_reward": 0.051205269526690245,
      "rewards/forensics_tiou_reward": 0.4035579562187195,
      "step": 379
    },
    {
      "completion_length": 22.28125,
      "epoch": 1.1658986175115207,
      "grad_norm": 5.9761247634887695,
      "kl": 0.3583984375,
      "learning_rate": 7.084615384615384e-07,
      "loss": 0.0486,
      "reward": 0.6378279030323029,
      "reward_std": 0.126654714345932,
      "reward_std_all": 0.14661503583192825,
      "rewards-max/forensics_iou_reward": 0.979450911283493,
      "rewards-max/forensics_strict_edge_reward": 0.30387546122074127,
      "rewards-max/forensics_tiou_reward": 0.9673358201980591,
      "rewards/forensics_iou_reward": 0.5262531638145447,
      "rewards/forensics_strict_edge_reward": 0.014573969878256321,
      "rewards/forensics_tiou_reward": 0.47581154108047485,
      "step": 380
    },
    {
      "completion_length": 23.34375,
      "epoch": 1.1689708141321045,
      "grad_norm": 6.1159281730651855,
      "kl": 0.47265625,
      "learning_rate": 7.076923076923077e-07,
      "loss": 0.1091,
      "reward": 0.6383824646472931,
      "reward_std": 0.16435659676790237,
      "reward_std_all": 0.16107749566435814,
      "rewards-max/forensics_iou_reward": 0.906339555978775,
      "rewards-max/forensics_strict_edge_reward": 0.38022976741194725,
      "rewards-max/forensics_tiou_reward": 0.8876066505908966,
      "rewards/forensics_iou_reward": 0.48727743327617645,
      "rewards/forensics_strict_edge_reward": 0.02141713723540306,
      "rewards/forensics_tiou_reward": 0.43717220425605774,
      "step": 381
    },
    {
      "completion_length": 23.703125,
      "epoch": 1.1720430107526882,
      "grad_norm": 5.154820919036865,
      "kl": 0.4111328125,
      "learning_rate": 7.069230769230768e-07,
      "loss": 0.1025,
      "reward": 0.6199551224708557,
      "reward_std": 0.13513625413179398,
      "reward_std_all": 0.15401875972747803,
      "rewards-max/forensics_iou_reward": 0.9430191218852997,
      "rewards-max/forensics_strict_edge_reward": 0.2868630141019821,
      "rewards-max/forensics_tiou_reward": 0.917182445526123,
      "rewards/forensics_iou_reward": 0.45508207380771637,
      "rewards/forensics_strict_edge_reward": 0.019201485440135002,
      "rewards/forensics_tiou_reward": 0.381790429353714,
      "step": 382
    },
    {
      "completion_length": 27.359375,
      "epoch": 1.1751152073732718,
      "grad_norm": 4.988589286804199,
      "kl": 0.443359375,
      "learning_rate": 7.061538461538462e-07,
      "loss": 0.0198,
      "reward": 0.6912389397621155,
      "reward_std": 0.13774242997169495,
      "reward_std_all": 0.13243526220321655,
      "rewards-max/forensics_iou_reward": 0.9835793972015381,
      "rewards-max/forensics_strict_edge_reward": 0.5634074360132217,
      "rewards-max/forensics_tiou_reward": 0.9792746901512146,
      "rewards/forensics_iou_reward": 0.5422775745391846,
      "rewards/forensics_strict_edge_reward": 0.046923963353037834,
      "rewards/forensics_tiou_reward": 0.48431625962257385,
      "step": 383
    },
    {
      "completion_length": 29.0625,
      "epoch": 1.1781874039938556,
      "grad_norm": 5.813256740570068,
      "kl": 0.572265625,
      "learning_rate": 7.053846153846154e-07,
      "loss": 0.2136,
      "reward": 0.6384368240833282,
      "reward_std": 0.1297735534608364,
      "reward_std_all": 0.14792650192975998,
      "rewards-max/forensics_iou_reward": 0.989695131778717,
      "rewards-max/forensics_strict_edge_reward": 0.7268245220184326,
      "rewards-max/forensics_tiou_reward": 0.9864796996116638,
      "rewards/forensics_iou_reward": 0.48445266485214233,
      "rewards/forensics_strict_edge_reward": 0.0652141161262989,
      "rewards/forensics_tiou_reward": 0.43448078632354736,
      "step": 384
    },
    {
      "completion_length": 24.1875,
      "epoch": 1.1812596006144394,
      "grad_norm": 7.134328365325928,
      "kl": 0.3701171875,
      "learning_rate": 7.046153846153846e-07,
      "loss": 0.1433,
      "reward": 0.577454686164856,
      "reward_std": 0.14279060438275337,
      "reward_std_all": 0.18367557227611542,
      "rewards-max/forensics_iou_reward": 0.9787418246269226,
      "rewards-max/forensics_strict_edge_reward": 0.47726282477378845,
      "rewards-max/forensics_tiou_reward": 0.9701572358608246,
      "rewards/forensics_iou_reward": 0.4284605085849762,
      "rewards/forensics_strict_edge_reward": 0.026846216060221195,
      "rewards/forensics_tiou_reward": 0.3752191811800003,
      "step": 385
    },
    {
      "completion_length": 22.59375,
      "epoch": 1.1843317972350231,
      "grad_norm": 5.77748966217041,
      "kl": 0.296875,
      "learning_rate": 7.038461538461539e-07,
      "loss": -0.0164,
      "reward": 0.6474794447422028,
      "reward_std": 0.10490826144814491,
      "reward_std_all": 0.12756092101335526,
      "rewards-max/forensics_iou_reward": 0.940514475107193,
      "rewards-max/forensics_strict_edge_reward": 0.645659938454628,
      "rewards-max/forensics_tiou_reward": 0.9323807954788208,
      "rewards/forensics_iou_reward": 0.5056403130292892,
      "rewards/forensics_strict_edge_reward": 0.04927685018628836,
      "rewards/forensics_tiou_reward": 0.4575905501842499,
      "step": 386
    },
    {
      "completion_length": 26.34375,
      "epoch": 1.1874039938556067,
      "grad_norm": 4.796816825866699,
      "kl": 0.302734375,
      "learning_rate": 7.03076923076923e-07,
      "loss": 0.188,
      "reward": 0.6894265711307526,
      "reward_std": 0.11021260544657707,
      "reward_std_all": 0.11588968336582184,
      "rewards-max/forensics_iou_reward": 0.9431110322475433,
      "rewards-max/forensics_strict_edge_reward": 0.4690026119351387,
      "rewards-max/forensics_tiou_reward": 0.9207848310470581,
      "rewards/forensics_iou_reward": 0.5210225880146027,
      "rewards/forensics_strict_edge_reward": 0.03394236508756876,
      "rewards/forensics_tiou_reward": 0.435106560587883,
      "step": 387
    },
    {
      "completion_length": 22.421875,
      "epoch": 1.1904761904761905,
      "grad_norm": 4.663717746734619,
      "kl": 0.2734375,
      "learning_rate": 7.023076923076922e-07,
      "loss": 0.0544,
      "reward": 0.712236762046814,
      "reward_std": 0.10794687271118164,
      "reward_std_all": 0.10393980145454407,
      "rewards-max/forensics_iou_reward": 0.9606973230838776,
      "rewards-max/forensics_strict_edge_reward": 0.13044166658073664,
      "rewards-max/forensics_tiou_reward": 0.9392737448215485,
      "rewards/forensics_iou_reward": 0.5857908725738525,
      "rewards/forensics_strict_edge_reward": 0.012780507444404066,
      "rewards/forensics_tiou_reward": 0.5312025845050812,
      "step": 388
    },
    {
      "completion_length": 28.078125,
      "epoch": 1.1935483870967742,
      "grad_norm": 4.5224409103393555,
      "kl": 0.37109375,
      "learning_rate": 7.015384615384615e-07,
      "loss": 0.1736,
      "reward": 0.6334557235240936,
      "reward_std": 0.1425420418381691,
      "reward_std_all": 0.15824173390865326,
      "rewards-max/forensics_iou_reward": 0.9334706366062164,
      "rewards-max/forensics_strict_edge_reward": 0.4307064116001129,
      "rewards-max/forensics_tiou_reward": 0.9149753749370575,
      "rewards/forensics_iou_reward": 0.4902528375387192,
      "rewards/forensics_strict_edge_reward": 0.03225766867399216,
      "rewards/forensics_tiou_reward": 0.4473242908716202,
      "step": 389
    },
    {
      "completion_length": 27.40625,
      "epoch": 1.1966205837173578,
      "grad_norm": 4.379509449005127,
      "kl": 0.3173828125,
      "learning_rate": 7.007692307692308e-07,
      "loss": 0.1867,
      "reward": 0.6608555018901825,
      "reward_std": 0.08270366303622723,
      "reward_std_all": 0.11067749932408333,
      "rewards-max/forensics_iou_reward": 0.9288243055343628,
      "rewards-max/forensics_strict_edge_reward": 0.112779401242733,
      "rewards-max/forensics_tiou_reward": 0.9025797843933105,
      "rewards/forensics_iou_reward": 0.49332788586616516,
      "rewards/forensics_strict_edge_reward": 0.011434062151238322,
      "rewards/forensics_tiou_reward": 0.438642755150795,
      "step": 390
    },
    {
      "completion_length": 28.28125,
      "epoch": 1.1996927803379416,
      "grad_norm": 4.1106343269348145,
      "kl": 0.2587890625,
      "learning_rate": 7e-07,
      "loss": 0.0437,
      "reward": 0.6808880269527435,
      "reward_std": 0.11636628955602646,
      "reward_std_all": 0.11938884854316711,
      "rewards-max/forensics_iou_reward": 0.9391139149665833,
      "rewards-max/forensics_strict_edge_reward": 0.42574505507946014,
      "rewards-max/forensics_tiou_reward": 0.9185143113136292,
      "rewards/forensics_iou_reward": 0.5233786702156067,
      "rewards/forensics_strict_edge_reward": 0.03141276352107525,
      "rewards/forensics_tiou_reward": 0.47641292214393616,
      "step": 391
    },
    {
      "completion_length": 28.34375,
      "epoch": 1.2027649769585254,
      "grad_norm": 5.198550701141357,
      "kl": 0.4609375,
      "learning_rate": 6.992307692307692e-07,
      "loss": 0.0096,
      "reward": 0.7154836654663086,
      "reward_std": 0.10817627236247063,
      "reward_std_all": 0.10183480754494667,
      "rewards-max/forensics_iou_reward": 0.9524325728416443,
      "rewards-max/forensics_strict_edge_reward": 0.2868630141019821,
      "rewards-max/forensics_tiou_reward": 0.9398005604743958,
      "rewards/forensics_iou_reward": 0.583118736743927,
      "rewards/forensics_strict_edge_reward": 0.028456601779907942,
      "rewards/forensics_tiou_reward": 0.5086081624031067,
      "step": 392
    },
    {
      "completion_length": 30.640625,
      "epoch": 1.2058371735791091,
      "grad_norm": 6.121063232421875,
      "kl": 0.306640625,
      "learning_rate": 6.984615384615384e-07,
      "loss": 0.0128,
      "reward": 0.6142295598983765,
      "reward_std": 0.08795073814690113,
      "reward_std_all": 0.14249156415462494,
      "rewards-max/forensics_iou_reward": 0.9750609695911407,
      "rewards-max/forensics_strict_edge_reward": 0.348579540848732,
      "rewards-max/forensics_tiou_reward": 0.9636037349700928,
      "rewards/forensics_iou_reward": 0.46987128257751465,
      "rewards/forensics_strict_edge_reward": 0.043643354438245296,
      "rewards/forensics_tiou_reward": 0.42378565669059753,
      "step": 393
    },
    {
      "completion_length": 22.25,
      "epoch": 1.2089093701996927,
      "grad_norm": 7.359542369842529,
      "kl": 0.3984375,
      "learning_rate": 6.976923076923077e-07,
      "loss": 0.1417,
      "reward": 0.6966115236282349,
      "reward_std": 0.13556235656142235,
      "reward_std_all": 0.12632966041564941,
      "rewards-max/forensics_iou_reward": 0.9289531707763672,
      "rewards-max/forensics_strict_edge_reward": 0.406333327293396,
      "rewards-max/forensics_tiou_reward": 0.9121406376361847,
      "rewards/forensics_iou_reward": 0.5380692929029465,
      "rewards/forensics_strict_edge_reward": 0.022767159156501293,
      "rewards/forensics_tiou_reward": 0.486469030380249,
      "step": 394
    },
    {
      "completion_length": 24.1875,
      "epoch": 1.2119815668202765,
      "grad_norm": 5.399829387664795,
      "kl": 0.3740234375,
      "learning_rate": 6.969230769230769e-07,
      "loss": 0.0799,
      "reward": 0.6793746650218964,
      "reward_std": 0.07857543602585793,
      "reward_std_all": 0.09858028590679169,
      "rewards-max/forensics_iou_reward": 0.9725877046585083,
      "rewards-max/forensics_strict_edge_reward": 0.44064828753471375,
      "rewards-max/forensics_tiou_reward": 0.9642997980117798,
      "rewards/forensics_iou_reward": 0.5356078594923019,
      "rewards/forensics_strict_edge_reward": 0.027541235089302063,
      "rewards/forensics_tiou_reward": 0.47532200813293457,
      "step": 395
    },
    {
      "completion_length": 27.078125,
      "epoch": 1.2150537634408602,
      "grad_norm": 4.67580509185791,
      "kl": 0.3671875,
      "learning_rate": 6.961538461538461e-07,
      "loss": 0.1928,
      "reward": 0.6922877132892609,
      "reward_std": 0.1046410147100687,
      "reward_std_all": 0.10811719298362732,
      "rewards-max/forensics_iou_reward": 0.9118386805057526,
      "rewards-max/forensics_strict_edge_reward": 0.28660978376865387,
      "rewards-max/forensics_tiou_reward": 0.896657258272171,
      "rewards/forensics_iou_reward": 0.5269803106784821,
      "rewards/forensics_strict_edge_reward": 0.027976377867162228,
      "rewards/forensics_tiou_reward": 0.4712718427181244,
      "step": 396
    },
    {
      "completion_length": 27.53125,
      "epoch": 1.218125960061444,
      "grad_norm": 5.089437961578369,
      "kl": 0.43359375,
      "learning_rate": 6.953846153846154e-07,
      "loss": 0.1104,
      "reward": 0.6663090288639069,
      "reward_std": 0.11372436210513115,
      "reward_std_all": 0.12958228960633278,
      "rewards-max/forensics_iou_reward": 0.8718827664852142,
      "rewards-max/forensics_strict_edge_reward": 0.24912343337200582,
      "rewards-max/forensics_tiou_reward": 0.830550491809845,
      "rewards/forensics_iou_reward": 0.45898081362247467,
      "rewards/forensics_strict_edge_reward": 0.011318008924718015,
      "rewards/forensics_tiou_reward": 0.4004589021205902,
      "step": 397
    },
    {
      "completion_length": 24.25,
      "epoch": 1.2211981566820276,
      "grad_norm": 7.5883989334106445,
      "kl": 0.3525390625,
      "learning_rate": 6.946153846153846e-07,
      "loss": 0.0661,
      "reward": 0.7094787955284119,
      "reward_std": 0.0959471333771944,
      "reward_std_all": 0.0980714000761509,
      "rewards-max/forensics_iou_reward": 0.9854397475719452,
      "rewards-max/forensics_strict_edge_reward": 0.5523883104324341,
      "rewards-max/forensics_tiou_reward": 0.9797331690788269,
      "rewards/forensics_iou_reward": 0.5716182291507721,
      "rewards/forensics_strict_edge_reward": 0.061540067195892334,
      "rewards/forensics_tiou_reward": 0.5106492042541504,
      "step": 398
    },
    {
      "completion_length": 25.171875,
      "epoch": 1.2242703533026114,
      "grad_norm": 4.709245681762695,
      "kl": 0.353515625,
      "learning_rate": 6.938461538461539e-07,
      "loss": 0.1838,
      "reward": 0.6799820959568024,
      "reward_std": 0.085597675293684,
      "reward_std_all": 0.1052485778927803,
      "rewards-max/forensics_iou_reward": 0.9111609160900116,
      "rewards-max/forensics_strict_edge_reward": 0.24126016907393932,
      "rewards-max/forensics_tiou_reward": 0.8865351974964142,
      "rewards/forensics_iou_reward": 0.523368239402771,
      "rewards/forensics_strict_edge_reward": 0.018924534204415977,
      "rewards/forensics_tiou_reward": 0.4743986278772354,
      "step": 399
    },
    {
      "completion_length": 26.421875,
      "epoch": 1.2273425499231951,
      "grad_norm": 3.985250234603882,
      "kl": 0.376953125,
      "learning_rate": 6.93076923076923e-07,
      "loss": 0.0662,
      "reward": 0.7430840134620667,
      "reward_std": 0.0589281152933836,
      "reward_std_all": 0.06046874448657036,
      "rewards-max/forensics_iou_reward": 0.9539972245693207,
      "rewards-max/forensics_strict_edge_reward": 0.47245824337005615,
      "rewards-max/forensics_tiou_reward": 0.9419643580913544,
      "rewards/forensics_iou_reward": 0.6276074349880219,
      "rewards/forensics_strict_edge_reward": 0.03362833522260189,
      "rewards/forensics_tiou_reward": 0.5808787941932678,
      "step": 400
    },
    {
      "completion_length": 27.28125,
      "epoch": 1.230414746543779,
      "grad_norm": 4.491318702697754,
      "kl": 0.4150390625,
      "learning_rate": 6.923076923076922e-07,
      "loss": 0.0576,
      "reward": 0.7295409142971039,
      "reward_std": 0.09302358329296112,
      "reward_std_all": 0.08733504265546799,
      "rewards-max/forensics_iou_reward": 0.9605381190776825,
      "rewards-max/forensics_strict_edge_reward": 0.45717276632785797,
      "rewards-max/forensics_tiou_reward": 0.9497023522853851,
      "rewards/forensics_iou_reward": 0.5934478938579559,
      "rewards/forensics_strict_edge_reward": 0.029497671872377396,
      "rewards/forensics_tiou_reward": 0.5413393676280975,
      "step": 401
    },
    {
      "completion_length": 28.21875,
      "epoch": 1.2334869431643625,
      "grad_norm": 4.295539855957031,
      "kl": 0.3515625,
      "learning_rate": 6.915384615384615e-07,
      "loss": 0.124,
      "reward": 0.6085595488548279,
      "reward_std": 0.1496400572359562,
      "reward_std_all": 0.1716046929359436,
      "rewards-max/forensics_iou_reward": 0.9162530899047852,
      "rewards-max/forensics_strict_edge_reward": 0.25912253744900227,
      "rewards-max/forensics_tiou_reward": 0.8874388337135315,
      "rewards/forensics_iou_reward": 0.4622986614704132,
      "rewards/forensics_strict_edge_reward": 0.02263943722937256,
      "rewards/forensics_tiou_reward": 0.4050554186105728,
      "step": 402
    },
    {
      "completion_length": 24.796875,
      "epoch": 1.2365591397849462,
      "grad_norm": 6.213072776794434,
      "kl": 0.404296875,
      "learning_rate": 6.907692307692307e-07,
      "loss": 0.182,
      "reward": 0.7215113639831543,
      "reward_std": 0.13343599066138268,
      "reward_std_all": 0.11863743513822556,
      "rewards-max/forensics_iou_reward": 0.975121796131134,
      "rewards-max/forensics_strict_edge_reward": 0.3813142329454422,
      "rewards-max/forensics_tiou_reward": 0.9684800803661346,
      "rewards/forensics_iou_reward": 0.5692819952964783,
      "rewards/forensics_strict_edge_reward": 0.04710538499057293,
      "rewards/forensics_tiou_reward": 0.5218226611614227,
      "step": 403
    },
    {
      "completion_length": 26.046875,
      "epoch": 1.23963133640553,
      "grad_norm": 4.292879581451416,
      "kl": 0.3740234375,
      "learning_rate": 6.9e-07,
      "loss": 0.0554,
      "reward": 0.7084376513957977,
      "reward_std": 0.05327938683331013,
      "reward_std_all": 0.07588911056518555,
      "rewards-max/forensics_iou_reward": 0.9450841546058655,
      "rewards-max/forensics_strict_edge_reward": 0.21567560732364655,
      "rewards-max/forensics_tiou_reward": 0.9254224896430969,
      "rewards/forensics_iou_reward": 0.5419138669967651,
      "rewards/forensics_strict_edge_reward": 0.01916087418794632,
      "rewards/forensics_tiou_reward": 0.4868512451648712,
      "step": 404
    },
    {
      "completion_length": 24.625,
      "epoch": 1.2427035330261136,
      "grad_norm": 7.5151824951171875,
      "kl": 0.3369140625,
      "learning_rate": 6.892307692307692e-07,
      "loss": 0.0423,
      "reward": 0.5876176655292511,
      "reward_std": 0.09087680280208588,
      "reward_std_all": 0.14690714329481125,
      "rewards-max/forensics_iou_reward": 0.9858080744743347,
      "rewards-max/forensics_strict_edge_reward": 0.3726309004705399,
      "rewards-max/forensics_tiou_reward": 0.9810121357440948,
      "rewards/forensics_iou_reward": 0.4779444485902786,
      "rewards/forensics_strict_edge_reward": 0.02527461331919767,
      "rewards/forensics_tiou_reward": 0.4283474087715149,
      "step": 405
    },
    {
      "completion_length": 23.6875,
      "epoch": 1.2457757296466974,
      "grad_norm": 4.695733547210693,
      "kl": 0.4873046875,
      "learning_rate": 6.884615384615384e-07,
      "loss": 0.0667,
      "reward": 0.6987440884113312,
      "reward_std": 0.09418067708611488,
      "reward_std_all": 0.10067925602197647,
      "rewards-max/forensics_iou_reward": 0.9751295149326324,
      "rewards-max/forensics_strict_edge_reward": 0.3931644558906555,
      "rewards-max/forensics_tiou_reward": 0.9651648998260498,
      "rewards/forensics_iou_reward": 0.5455645322799683,
      "rewards/forensics_strict_edge_reward": 0.038990442641079426,
      "rewards/forensics_tiou_reward": 0.46999962627887726,
      "step": 406
    },
    {
      "completion_length": 26.40625,
      "epoch": 1.2488479262672811,
      "grad_norm": 4.971114635467529,
      "kl": 0.435546875,
      "learning_rate": 6.876923076923077e-07,
      "loss": 0.1357,
      "reward": 0.6261473894119263,
      "reward_std": 0.11892925575375557,
      "reward_std_all": 0.1476937234401703,
      "rewards-max/forensics_iou_reward": 0.9491349458694458,
      "rewards-max/forensics_strict_edge_reward": 0.4114123433828354,
      "rewards-max/forensics_tiou_reward": 0.935449093580246,
      "rewards/forensics_iou_reward": 0.483304426074028,
      "rewards/forensics_strict_edge_reward": 0.03764178417623043,
      "rewards/forensics_tiou_reward": 0.4248673617839813,
      "step": 407
    },
    {
      "completion_length": 25.5,
      "epoch": 1.251920122887865,
      "grad_norm": 5.563323497772217,
      "kl": 0.310546875,
      "learning_rate": 6.869230769230769e-07,
      "loss": 0.0967,
      "reward": 0.7165287733078003,
      "reward_std": 0.1148027703166008,
      "reward_std_all": 0.1065291054546833,
      "rewards-max/forensics_iou_reward": 0.9458849728107452,
      "rewards-max/forensics_strict_edge_reward": 0.43538494408130646,
      "rewards-max/forensics_tiou_reward": 0.9366888999938965,
      "rewards/forensics_iou_reward": 0.5356719493865967,
      "rewards/forensics_strict_edge_reward": 0.02573973871767521,
      "rewards/forensics_tiou_reward": 0.47743406891822815,
      "step": 408
    },
    {
      "completion_length": 25.625,
      "epoch": 1.2549923195084485,
      "grad_norm": 4.65983247756958,
      "kl": 0.3701171875,
      "learning_rate": 6.861538461538461e-07,
      "loss": 0.0737,
      "reward": 0.7076778709888458,
      "reward_std": 0.096199631690979,
      "reward_std_all": 0.09578394517302513,
      "rewards-max/forensics_iou_reward": 0.8856456577777863,
      "rewards-max/forensics_strict_edge_reward": 0.2547980658710003,
      "rewards-max/forensics_tiou_reward": 0.8678461909294128,
      "rewards/forensics_iou_reward": 0.5359740853309631,
      "rewards/forensics_strict_edge_reward": 0.012388092814944685,
      "rewards/forensics_tiou_reward": 0.46812307834625244,
      "step": 409
    },
    {
      "completion_length": 23.765625,
      "epoch": 1.2580645161290323,
      "grad_norm": 5.330956935882568,
      "kl": 0.2939453125,
      "learning_rate": 6.853846153846153e-07,
      "loss": 0.1278,
      "reward": 0.666419267654419,
      "reward_std": 0.13055618107318878,
      "reward_std_all": 0.14174681901931763,
      "rewards-max/forensics_iou_reward": 0.9379788041114807,
      "rewards-max/forensics_strict_edge_reward": 0.17923272401094437,
      "rewards-max/forensics_tiou_reward": 0.9047726690769196,
      "rewards/forensics_iou_reward": 0.5168602168560028,
      "rewards/forensics_strict_edge_reward": 0.015972835943102837,
      "rewards/forensics_tiou_reward": 0.45809750258922577,
      "step": 410
    },
    {
      "completion_length": 26.828125,
      "epoch": 1.261136712749616,
      "grad_norm": 5.395296096801758,
      "kl": 0.3359375,
      "learning_rate": 6.846153846153846e-07,
      "loss": 0.0585,
      "reward": 0.7081268131732941,
      "reward_std": 0.07322489097714424,
      "reward_std_all": 0.08488348498940468,
      "rewards-max/forensics_iou_reward": 0.9804423451423645,
      "rewards-max/forensics_strict_edge_reward": 0.3955056071281433,
      "rewards-max/forensics_tiou_reward": 0.9731764495372772,
      "rewards/forensics_iou_reward": 0.585449606180191,
      "rewards/forensics_strict_edge_reward": 0.03484600968658924,
      "rewards/forensics_tiou_reward": 0.5247723758220673,
      "step": 411
    },
    {
      "completion_length": 25.796875,
      "epoch": 1.2642089093701996,
      "grad_norm": 6.631100654602051,
      "kl": 0.31640625,
      "learning_rate": 6.838461538461539e-07,
      "loss": 0.1164,
      "reward": 0.7071212530136108,
      "reward_std": 0.1324734166264534,
      "reward_std_all": 0.12158282101154327,
      "rewards-max/forensics_iou_reward": 0.9049012064933777,
      "rewards-max/forensics_strict_edge_reward": 0.158625528216362,
      "rewards-max/forensics_tiou_reward": 0.8614924550056458,
      "rewards/forensics_iou_reward": 0.5853302776813507,
      "rewards/forensics_strict_edge_reward": 0.01777790393680334,
      "rewards/forensics_tiou_reward": 0.5262127220630646,
      "step": 412
    },
    {
      "completion_length": 24.890625,
      "epoch": 1.2672811059907834,
      "grad_norm": 4.83495569229126,
      "kl": 0.3046875,
      "learning_rate": 6.83076923076923e-07,
      "loss": 0.0755,
      "reward": 0.6943685412406921,
      "reward_std": 0.10070270672440529,
      "reward_std_all": 0.10191483423113823,
      "rewards-max/forensics_iou_reward": 0.9675419926643372,
      "rewards-max/forensics_strict_edge_reward": 0.29475584626197815,
      "rewards-max/forensics_tiou_reward": 0.9516000151634216,
      "rewards/forensics_iou_reward": 0.5419074892997742,
      "rewards/forensics_strict_edge_reward": 0.028906051069498062,
      "rewards/forensics_tiou_reward": 0.4815121442079544,
      "step": 413
    },
    {
      "completion_length": 25.15625,
      "epoch": 1.2703533026113671,
      "grad_norm": 5.428030014038086,
      "kl": 0.298828125,
      "learning_rate": 6.823076923076922e-07,
      "loss": -0.006,
      "reward": 0.6226261854171753,
      "reward_std": 0.1264333575963974,
      "reward_std_all": 0.15035201609134674,
      "rewards-max/forensics_iou_reward": 0.8761227428913116,
      "rewards-max/forensics_strict_edge_reward": 0.1836656853556633,
      "rewards-max/forensics_tiou_reward": 0.8558592200279236,
      "rewards/forensics_iou_reward": 0.4360823333263397,
      "rewards/forensics_strict_edge_reward": 0.009081199998036027,
      "rewards/forensics_tiou_reward": 0.38950181007385254,
      "step": 414
    },
    {
      "completion_length": 30.8125,
      "epoch": 1.273425499231951,
      "grad_norm": 4.45082950592041,
      "kl": 0.376953125,
      "learning_rate": 6.815384615384615e-07,
      "loss": 0.062,
      "reward": 0.7147040367126465,
      "reward_std": 0.1173991709947586,
      "reward_std_all": 0.10896581783890724,
      "rewards-max/forensics_iou_reward": 0.9607562124729156,
      "rewards-max/forensics_strict_edge_reward": 0.27988626807928085,
      "rewards-max/forensics_tiou_reward": 0.9450211822986603,
      "rewards/forensics_iou_reward": 0.5581183731555939,
      "rewards/forensics_strict_edge_reward": 0.01911128731444478,
      "rewards/forensics_tiou_reward": 0.5114318579435349,
      "step": 415
    },
    {
      "completion_length": 27.9375,
      "epoch": 1.2764976958525347,
      "grad_norm": 4.6011834144592285,
      "kl": 0.3134765625,
      "learning_rate": 6.807692307692307e-07,
      "loss": 0.1801,
      "reward": 0.7169470191001892,
      "reward_std": 0.09334215149283409,
      "reward_std_all": 0.09132837876677513,
      "rewards-max/forensics_iou_reward": 0.906707376241684,
      "rewards-max/forensics_strict_edge_reward": 0.4128633141517639,
      "rewards-max/forensics_tiou_reward": 0.8821638226509094,
      "rewards/forensics_iou_reward": 0.5305453836917877,
      "rewards/forensics_strict_edge_reward": 0.0377718354575336,
      "rewards/forensics_tiou_reward": 0.4732276648283005,
      "step": 416
    },
    {
      "completion_length": 26.3125,
      "epoch": 1.2795698924731183,
      "grad_norm": 4.9572343826293945,
      "kl": 0.29296875,
      "learning_rate": 6.800000000000001e-07,
      "loss": 0.0518,
      "reward": 0.6101362705230713,
      "reward_std": 0.15200001746416092,
      "reward_std_all": 0.16989104449748993,
      "rewards-max/forensics_iou_reward": 0.9166151583194733,
      "rewards-max/forensics_strict_edge_reward": 0.17715754359960556,
      "rewards-max/forensics_tiou_reward": 0.8945786654949188,
      "rewards/forensics_iou_reward": 0.46204379200935364,
      "rewards/forensics_strict_edge_reward": 0.016664195572957397,
      "rewards/forensics_tiou_reward": 0.40765948593616486,
      "step": 417
    },
    {
      "completion_length": 26.671875,
      "epoch": 1.282642089093702,
      "grad_norm": 4.8412580490112305,
      "kl": 0.3212890625,
      "learning_rate": 6.792307692307692e-07,
      "loss": 0.1798,
      "reward": 0.6767293214797974,
      "reward_std": 0.08020274713635445,
      "reward_std_all": 0.10284964926540852,
      "rewards-max/forensics_iou_reward": 0.9136669337749481,
      "rewards-max/forensics_strict_edge_reward": 0.09754503518342972,
      "rewards-max/forensics_tiou_reward": 0.8785578608512878,
      "rewards/forensics_iou_reward": 0.5128375440835953,
      "rewards/forensics_strict_edge_reward": 0.008962791878730059,
      "rewards/forensics_tiou_reward": 0.4591089338064194,
      "step": 418
    },
    {
      "completion_length": 22.375,
      "epoch": 1.2857142857142856,
      "grad_norm": 5.645130634307861,
      "kl": 0.27490234375,
      "learning_rate": 6.784615384615384e-07,
      "loss": 0.1495,
      "reward": 0.6984778940677643,
      "reward_std": 0.1151353009045124,
      "reward_std_all": 0.11094411835074425,
      "rewards-max/forensics_iou_reward": 0.9355656802654266,
      "rewards-max/forensics_strict_edge_reward": 0.4542374759912491,
      "rewards-max/forensics_tiou_reward": 0.9174922704696655,
      "rewards/forensics_iou_reward": 0.5391016602516174,
      "rewards/forensics_strict_edge_reward": 0.04473350569605827,
      "rewards/forensics_tiou_reward": 0.4862753450870514,
      "step": 419
    },
    {
      "completion_length": 30.109375,
      "epoch": 1.2887864823348694,
      "grad_norm": 4.799912452697754,
      "kl": 0.32421875,
      "learning_rate": 6.776923076923077e-07,
      "loss": 0.062,
      "reward": 0.6301959156990051,
      "reward_std": 0.10808531194925308,
      "reward_std_all": 0.14164762943983078,
      "rewards-max/forensics_iou_reward": 0.9275577664375305,
      "rewards-max/forensics_strict_edge_reward": 0.23273001611232758,
      "rewards-max/forensics_tiou_reward": 0.8977026045322418,
      "rewards/forensics_iou_reward": 0.4717865288257599,
      "rewards/forensics_strict_edge_reward": 0.015475575346499681,
      "rewards/forensics_tiou_reward": 0.4312225729227066,
      "step": 420
    },
    {
      "completion_length": 28.734375,
      "epoch": 1.2918586789554531,
      "grad_norm": 6.848349571228027,
      "kl": 0.439453125,
      "learning_rate": 6.769230769230769e-07,
      "loss": 0.2245,
      "reward": 0.6947938799858093,
      "reward_std": 0.08391905203461647,
      "reward_std_all": 0.1010599285364151,
      "rewards-max/forensics_iou_reward": 0.9659665524959564,
      "rewards-max/forensics_strict_edge_reward": 0.2711085230112076,
      "rewards-max/forensics_tiou_reward": 0.9526141583919525,
      "rewards/forensics_iou_reward": 0.5289985537528992,
      "rewards/forensics_strict_edge_reward": 0.02281992044299841,
      "rewards/forensics_tiou_reward": 0.46823257207870483,
      "step": 421
    },
    {
      "completion_length": 26.859375,
      "epoch": 1.294930875576037,
      "grad_norm": 4.983539581298828,
      "kl": 0.3671875,
      "learning_rate": 6.761538461538461e-07,
      "loss": 0.1796,
      "reward": 0.6233045160770416,
      "reward_std": 0.17346014827489853,
      "reward_std_all": 0.17816881835460663,
      "rewards-max/forensics_iou_reward": 0.9233961701393127,
      "rewards-max/forensics_strict_edge_reward": 0.24479330703616142,
      "rewards-max/forensics_tiou_reward": 0.8971448540687561,
      "rewards/forensics_iou_reward": 0.462787002325058,
      "rewards/forensics_strict_edge_reward": 0.012983550084754825,
      "rewards/forensics_tiou_reward": 0.4187743067741394,
      "step": 422
    },
    {
      "completion_length": 26.484375,
      "epoch": 1.2980030721966207,
      "grad_norm": 5.049590587615967,
      "kl": 0.5400390625,
      "learning_rate": 6.753846153846153e-07,
      "loss": 0.1755,
      "reward": 0.7099997103214264,
      "reward_std": 0.12489666417241096,
      "reward_std_all": 0.11503598466515541,
      "rewards-max/forensics_iou_reward": 0.9638391435146332,
      "rewards-max/forensics_strict_edge_reward": 0.5119633823633194,
      "rewards-max/forensics_tiou_reward": 0.9523839354515076,
      "rewards/forensics_iou_reward": 0.5611762702465057,
      "rewards/forensics_strict_edge_reward": 0.06435856595635414,
      "rewards/forensics_tiou_reward": 0.5136266350746155,
      "step": 423
    },
    {
      "completion_length": 25.96875,
      "epoch": 1.3010752688172043,
      "grad_norm": 4.597318172454834,
      "kl": 0.357421875,
      "learning_rate": 6.746153846153845e-07,
      "loss": 0.0952,
      "reward": 0.7030077576637268,
      "reward_std": 0.10947634279727936,
      "reward_std_all": 0.10625866428017616,
      "rewards-max/forensics_iou_reward": 0.9806720614433289,
      "rewards-max/forensics_strict_edge_reward": 0.5877082049846649,
      "rewards-max/forensics_tiou_reward": 0.9710822701454163,
      "rewards/forensics_iou_reward": 0.5592564642429352,
      "rewards/forensics_strict_edge_reward": 0.03841729275882244,
      "rewards/forensics_tiou_reward": 0.5041415989398956,
      "step": 424
    },
    {
      "completion_length": 27.203125,
      "epoch": 1.304147465437788,
      "grad_norm": 4.534994125366211,
      "kl": 0.3544921875,
      "learning_rate": 6.738461538461539e-07,
      "loss": 0.2277,
      "reward": 0.7215213775634766,
      "reward_std": 0.12076478451490402,
      "reward_std_all": 0.10982729122042656,
      "rewards-max/forensics_iou_reward": 0.8613872826099396,
      "rewards-max/forensics_strict_edge_reward": 0.21966427564620972,
      "rewards-max/forensics_tiou_reward": 0.8369512557983398,
      "rewards/forensics_iou_reward": 0.5675634741783142,
      "rewards/forensics_strict_edge_reward": 0.021953104063868523,
      "rewards/forensics_tiou_reward": 0.517267718911171,
      "step": 425
    },
    {
      "completion_length": 24.078125,
      "epoch": 1.3072196620583718,
      "grad_norm": 5.392589569091797,
      "kl": 0.3076171875,
      "learning_rate": 6.730769230769231e-07,
      "loss": 0.1449,
      "reward": 0.7303985953330994,
      "reward_std": 0.09794117137789726,
      "reward_std_all": 0.09321304224431515,
      "rewards-max/forensics_iou_reward": 0.9325352609157562,
      "rewards-max/forensics_strict_edge_reward": 0.19139372557401657,
      "rewards-max/forensics_tiou_reward": 0.9151075184345245,
      "rewards/forensics_iou_reward": 0.5862432420253754,
      "rewards/forensics_strict_edge_reward": 0.016663051676005125,
      "rewards/forensics_tiou_reward": 0.5170280039310455,
      "step": 426
    },
    {
      "completion_length": 26.1875,
      "epoch": 1.3102918586789554,
      "grad_norm": 4.91376256942749,
      "kl": 0.3740234375,
      "learning_rate": 6.723076923076922e-07,
      "loss": 0.1642,
      "reward": 0.7557329535484314,
      "reward_std": 0.08635840192437172,
      "reward_std_all": 0.07655566185712814,
      "rewards-max/forensics_iou_reward": 0.9514542818069458,
      "rewards-max/forensics_strict_edge_reward": 0.2404046133160591,
      "rewards-max/forensics_tiou_reward": 0.9341234564781189,
      "rewards/forensics_iou_reward": 0.6855731308460236,
      "rewards/forensics_strict_edge_reward": 0.017301530577242374,
      "rewards/forensics_tiou_reward": 0.6146720349788666,
      "step": 427
    },
    {
      "completion_length": 24.234375,
      "epoch": 1.3133640552995391,
      "grad_norm": 5.808231830596924,
      "kl": 0.458984375,
      "learning_rate": 6.715384615384615e-07,
      "loss": 0.1214,
      "reward": 0.7044443488121033,
      "reward_std": 0.11291200667619705,
      "reward_std_all": 0.10825047641992569,
      "rewards-max/forensics_iou_reward": 0.9720000922679901,
      "rewards-max/forensics_strict_edge_reward": 0.46017518639564514,
      "rewards-max/forensics_tiou_reward": 0.9593481421470642,
      "rewards/forensics_iou_reward": 0.6202768981456757,
      "rewards/forensics_strict_edge_reward": 0.06578964367508888,
      "rewards/forensics_tiou_reward": 0.5600546300411224,
      "step": 428
    },
    {
      "completion_length": 26.53125,
      "epoch": 1.316436251920123,
      "grad_norm": 7.5450615882873535,
      "kl": 0.5537109375,
      "learning_rate": 6.707692307692307e-07,
      "loss": -0.0259,
      "reward": 0.7044095396995544,
      "reward_std": 0.07760858908295631,
      "reward_std_all": 0.08481718972325325,
      "rewards-max/forensics_iou_reward": 0.9702602028846741,
      "rewards-max/forensics_strict_edge_reward": 0.49658530950546265,
      "rewards-max/forensics_tiou_reward": 0.9620523750782013,
      "rewards/forensics_iou_reward": 0.5898522734642029,
      "rewards/forensics_strict_edge_reward": 0.05477621965110302,
      "rewards/forensics_tiou_reward": 0.5425275266170502,
      "step": 429
    },
    {
      "completion_length": 29.28125,
      "epoch": 1.3195084485407067,
      "grad_norm": 4.5378642082214355,
      "kl": 0.3408203125,
      "learning_rate": 6.7e-07,
      "loss": 0.02,
      "reward": 0.763820469379425,
      "reward_std": 0.045320115983486176,
      "reward_std_all": 0.04492449760437012,
      "rewards-max/forensics_iou_reward": 0.982151061296463,
      "rewards-max/forensics_strict_edge_reward": 0.6247820556163788,
      "rewards-max/forensics_tiou_reward": 0.9770424067974091,
      "rewards/forensics_iou_reward": 0.5840093195438385,
      "rewards/forensics_strict_edge_reward": 0.03878673166036606,
      "rewards/forensics_tiou_reward": 0.5176779925823212,
      "step": 430
    },
    {
      "completion_length": 28.578125,
      "epoch": 1.3225806451612903,
      "grad_norm": 5.407444477081299,
      "kl": 0.4111328125,
      "learning_rate": 6.692307692307692e-07,
      "loss": 0.1065,
      "reward": 0.6947424411773682,
      "reward_std": 0.07963033020496368,
      "reward_std_all": 0.0931342002004385,
      "rewards-max/forensics_iou_reward": 0.8800907731056213,
      "rewards-max/forensics_strict_edge_reward": 0.021174901630729437,
      "rewards-max/forensics_tiou_reward": 0.8586150109767914,
      "rewards/forensics_iou_reward": 0.502872958779335,
      "rewards/forensics_strict_edge_reward": 0.0016033617430366576,
      "rewards/forensics_tiou_reward": 0.4496801197528839,
      "step": 431
    },
    {
      "completion_length": 26.15625,
      "epoch": 1.325652841781874,
      "grad_norm": 6.951718807220459,
      "kl": 0.34375,
      "learning_rate": 6.684615384615384e-07,
      "loss": 0.1631,
      "reward": 0.6815795600414276,
      "reward_std": 0.14602766931056976,
      "reward_std_all": 0.13768059015274048,
      "rewards-max/forensics_iou_reward": 0.9872523844242096,
      "rewards-max/forensics_strict_edge_reward": 0.5905451774597168,
      "rewards-max/forensics_tiou_reward": 0.9833265244960785,
      "rewards/forensics_iou_reward": 0.4983731806278229,
      "rewards/forensics_strict_edge_reward": 0.035561319440603256,
      "rewards/forensics_tiou_reward": 0.4430391937494278,
      "step": 432
    },
    {
      "completion_length": 27.109375,
      "epoch": 1.3287250384024578,
      "grad_norm": 5.3442559242248535,
      "kl": 0.28564453125,
      "learning_rate": 6.676923076923077e-07,
      "loss": 0.0163,
      "reward": 0.6677992641925812,
      "reward_std": 0.07306753098964691,
      "reward_std_all": 0.10350489616394043,
      "rewards-max/forensics_iou_reward": 0.9318181872367859,
      "rewards-max/forensics_strict_edge_reward": 0.5036259908229113,
      "rewards-max/forensics_tiou_reward": 0.9004856646060944,
      "rewards/forensics_iou_reward": 0.5330692529678345,
      "rewards/forensics_strict_edge_reward": 0.02356150926789269,
      "rewards/forensics_tiou_reward": 0.47222961485385895,
      "step": 433
    },
    {
      "completion_length": 31.203125,
      "epoch": 1.3317972350230414,
      "grad_norm": 5.028726577758789,
      "kl": 0.37890625,
      "learning_rate": 6.669230769230769e-07,
      "loss": 0.0995,
      "reward": 0.7111109793186188,
      "reward_std": 0.09952152706682682,
      "reward_std_all": 0.10040091350674629,
      "rewards-max/forensics_iou_reward": 0.9544236660003662,
      "rewards-max/forensics_strict_edge_reward": 0.28589269518852234,
      "rewards-max/forensics_tiou_reward": 0.9422581195831299,
      "rewards/forensics_iou_reward": 0.5321940034627914,
      "rewards/forensics_strict_edge_reward": 0.015056682284921408,
      "rewards/forensics_tiou_reward": 0.48079095780849457,
      "step": 434
    },
    {
      "completion_length": 28.921875,
      "epoch": 1.3348694316436251,
      "grad_norm": 5.290640354156494,
      "kl": 0.3232421875,
      "learning_rate": 6.661538461538462e-07,
      "loss": 0.1488,
      "reward": 0.6848060190677643,
      "reward_std": 0.13820572197437286,
      "reward_std_all": 0.1323230341076851,
      "rewards-max/forensics_iou_reward": 0.9459359347820282,
      "rewards-max/forensics_strict_edge_reward": 0.2952784150838852,
      "rewards-max/forensics_tiou_reward": 0.9169932007789612,
      "rewards/forensics_iou_reward": 0.524754136800766,
      "rewards/forensics_strict_edge_reward": 0.026692481711506844,
      "rewards/forensics_tiou_reward": 0.466025173664093,
      "step": 435
    },
    {
      "completion_length": 25.578125,
      "epoch": 1.337941628264209,
      "grad_norm": 6.272456169128418,
      "kl": 0.314453125,
      "learning_rate": 6.653846153846153e-07,
      "loss": -0.0056,
      "reward": 0.6684680283069611,
      "reward_std": 0.08070736937224865,
      "reward_std_all": 0.10631930828094482,
      "rewards-max/forensics_iou_reward": 0.9188208878040314,
      "rewards-max/forensics_strict_edge_reward": 0.19336214661598206,
      "rewards-max/forensics_tiou_reward": 0.8911210000514984,
      "rewards/forensics_iou_reward": 0.500039279460907,
      "rewards/forensics_strict_edge_reward": 0.013550067786127329,
      "rewards/forensics_tiou_reward": 0.4481806606054306,
      "step": 436
    },
    {
      "completion_length": 27.1875,
      "epoch": 1.3410138248847927,
      "grad_norm": 5.283590316772461,
      "kl": 0.3916015625,
      "learning_rate": 6.646153846153845e-07,
      "loss": 0.1022,
      "reward": 0.7219942510128021,
      "reward_std": 0.03885000664740801,
      "reward_std_all": 0.060541353188455105,
      "rewards-max/forensics_iou_reward": 0.9660661816596985,
      "rewards-max/forensics_strict_edge_reward": 0.402368500828743,
      "rewards-max/forensics_tiou_reward": 0.9541006982326508,
      "rewards/forensics_iou_reward": 0.585720956325531,
      "rewards/forensics_strict_edge_reward": 0.03858433663845062,
      "rewards/forensics_tiou_reward": 0.5313186198472977,
      "step": 437
    },
    {
      "completion_length": 26.734375,
      "epoch": 1.3440860215053765,
      "grad_norm": 5.6161627769470215,
      "kl": 0.3271484375,
      "learning_rate": 6.638461538461539e-07,
      "loss": 0.0763,
      "reward": 0.5805649757385254,
      "reward_std": 0.13848429173231125,
      "reward_std_all": 0.176822692155838,
      "rewards-max/forensics_iou_reward": 0.8341059684753418,
      "rewards-max/forensics_strict_edge_reward": 0.22229879349470139,
      "rewards-max/forensics_tiou_reward": 0.8028726875782013,
      "rewards/forensics_iou_reward": 0.36675402522087097,
      "rewards/forensics_strict_edge_reward": 0.017620235215872526,
      "rewards/forensics_tiou_reward": 0.30078624188899994,
      "step": 438
    },
    {
      "completion_length": 24.0,
      "epoch": 1.34715821812596,
      "grad_norm": 5.5821027755737305,
      "kl": 0.38671875,
      "learning_rate": 6.630769230769231e-07,
      "loss": 0.1057,
      "reward": 0.565624326467514,
      "reward_std": 0.1287822499871254,
      "reward_std_all": 0.1726781576871872,
      "rewards-max/forensics_iou_reward": 0.9428586363792419,
      "rewards-max/forensics_strict_edge_reward": 0.412410169839859,
      "rewards-max/forensics_tiou_reward": 0.9258729517459869,
      "rewards/forensics_iou_reward": 0.40657053887844086,
      "rewards/forensics_strict_edge_reward": 0.030106274411082268,
      "rewards/forensics_tiou_reward": 0.3577086478471756,
      "step": 439
    },
    {
      "completion_length": 30.453125,
      "epoch": 1.3502304147465438,
      "grad_norm": 4.5520734786987305,
      "kl": 0.361328125,
      "learning_rate": 6.623076923076923e-07,
      "loss": 0.2009,
      "reward": 0.6926144361495972,
      "reward_std": 0.1387789472937584,
      "reward_std_all": 0.1291813738644123,
      "rewards-max/forensics_iou_reward": 0.9472526907920837,
      "rewards-max/forensics_strict_edge_reward": 0.18758921325206757,
      "rewards-max/forensics_tiou_reward": 0.934260904788971,
      "rewards/forensics_iou_reward": 0.5294598788022995,
      "rewards/forensics_strict_edge_reward": 0.01969800889492035,
      "rewards/forensics_tiou_reward": 0.4674179404973984,
      "step": 440
    },
    {
      "completion_length": 27.578125,
      "epoch": 1.3533026113671274,
      "grad_norm": 6.129795074462891,
      "kl": 0.4384765625,
      "learning_rate": 6.615384615384615e-07,
      "loss": 0.0833,
      "reward": 0.620326817035675,
      "reward_std": 0.06380476616322994,
      "reward_std_all": 0.12236585840582848,
      "rewards-max/forensics_iou_reward": 0.9414556920528412,
      "rewards-max/forensics_strict_edge_reward": 0.2955048084259033,
      "rewards-max/forensics_tiou_reward": 0.9264464378356934,
      "rewards/forensics_iou_reward": 0.45043011009693146,
      "rewards/forensics_strict_edge_reward": 0.014338667038828135,
      "rewards/forensics_tiou_reward": 0.38930508494377136,
      "step": 441
    },
    {
      "completion_length": 22.015625,
      "epoch": 1.3563748079877112,
      "grad_norm": 5.31343412399292,
      "kl": 0.4189453125,
      "learning_rate": 6.607692307692307e-07,
      "loss": 0.1438,
      "reward": 0.560270294547081,
      "reward_std": 0.13068528473377228,
      "reward_std_all": 0.18061736971139908,
      "rewards-max/forensics_iou_reward": 0.9547534883022308,
      "rewards-max/forensics_strict_edge_reward": 0.4102710783481598,
      "rewards-max/forensics_tiou_reward": 0.9383343458175659,
      "rewards/forensics_iou_reward": 0.4488929808139801,
      "rewards/forensics_strict_edge_reward": 0.047425375785678625,
      "rewards/forensics_tiou_reward": 0.40075427293777466,
      "step": 442
    },
    {
      "completion_length": 28.09375,
      "epoch": 1.359447004608295,
      "grad_norm": 5.56125020980835,
      "kl": 0.4140625,
      "learning_rate": 6.6e-07,
      "loss": -0.0047,
      "reward": 0.634241908788681,
      "reward_std": 0.07377712428569794,
      "reward_std_all": 0.11888246983289719,
      "rewards-max/forensics_iou_reward": 0.9423218369483948,
      "rewards-max/forensics_strict_edge_reward": 0.17935939133167267,
      "rewards-max/forensics_tiou_reward": 0.9224111139774323,
      "rewards/forensics_iou_reward": 0.4992073178291321,
      "rewards/forensics_strict_edge_reward": 0.011965625919401646,
      "rewards/forensics_tiou_reward": 0.4444965422153473,
      "step": 443
    },
    {
      "completion_length": 23.9375,
      "epoch": 1.3625192012288787,
      "grad_norm": 5.193134307861328,
      "kl": 0.341796875,
      "learning_rate": 6.592307692307691e-07,
      "loss": 0.0696,
      "reward": 0.7427846491336823,
      "reward_std": 0.07034000661224127,
      "reward_std_all": 0.06897543650120497,
      "rewards-max/forensics_iou_reward": 0.9736842215061188,
      "rewards-max/forensics_strict_edge_reward": 0.7246644794940948,
      "rewards-max/forensics_tiou_reward": 0.9685441851615906,
      "rewards/forensics_iou_reward": 0.6271576285362244,
      "rewards/forensics_strict_edge_reward": 0.06033835932612419,
      "rewards/forensics_tiou_reward": 0.5704471170902252,
      "step": 444
    },
    {
      "completion_length": 24.640625,
      "epoch": 1.3655913978494625,
      "grad_norm": 14.57470703125,
      "kl": 1.28515625,
      "learning_rate": 6.584615384615384e-07,
      "loss": 0.1695,
      "reward": 0.6831235289573669,
      "reward_std": 0.09852023795247078,
      "reward_std_all": 0.11417122557759285,
      "rewards-max/forensics_iou_reward": 0.9835051596164703,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.9796443581581116,
      "rewards/forensics_iou_reward": 0.5751062035560608,
      "rewards/forensics_strict_edge_reward": 0.049557456746697426,
      "rewards/forensics_tiou_reward": 0.5216486155986786,
      "step": 445
    },
    {
      "completion_length": 25.984375,
      "epoch": 1.368663594470046,
      "grad_norm": 5.033536434173584,
      "kl": 0.310546875,
      "learning_rate": 6.576923076923077e-07,
      "loss": 0.009,
      "reward": 0.7532825171947479,
      "reward_std": 0.07398543879389763,
      "reward_std_all": 0.06807068176567554,
      "rewards-max/forensics_iou_reward": 0.9940607249736786,
      "rewards-max/forensics_strict_edge_reward": 0.7815065383911133,
      "rewards-max/forensics_tiou_reward": 0.9918533563613892,
      "rewards/forensics_iou_reward": 0.6418806612491608,
      "rewards/forensics_strict_edge_reward": 0.08447512239217758,
      "rewards/forensics_tiou_reward": 0.5922163426876068,
      "step": 446
    },
    {
      "completion_length": 26.4375,
      "epoch": 1.3717357910906298,
      "grad_norm": 5.313849449157715,
      "kl": 0.275390625,
      "learning_rate": 6.569230769230769e-07,
      "loss": 0.1775,
      "reward": 0.6228043437004089,
      "reward_std": 0.1664419323205948,
      "reward_std_all": 0.17780111730098724,
      "rewards-max/forensics_iou_reward": 0.9876222014427185,
      "rewards-max/forensics_strict_edge_reward": 0.297258660197258,
      "rewards-max/forensics_tiou_reward": 0.9809819757938385,
      "rewards/forensics_iou_reward": 0.48232781887054443,
      "rewards/forensics_strict_edge_reward": 0.03045880701392889,
      "rewards/forensics_tiou_reward": 0.4381704479455948,
      "step": 447
    },
    {
      "completion_length": 22.625,
      "epoch": 1.3748079877112136,
      "grad_norm": 4.873108863830566,
      "kl": 0.484375,
      "learning_rate": 6.561538461538462e-07,
      "loss": 0.0541,
      "reward": 0.5690045356750488,
      "reward_std": 0.1763528436422348,
      "reward_std_all": 0.2016080990433693,
      "rewards-max/forensics_iou_reward": 0.881112664937973,
      "rewards-max/forensics_strict_edge_reward": 0.23870950192213058,
      "rewards-max/forensics_tiou_reward": 0.8606652021408081,
      "rewards/forensics_iou_reward": 0.3997015655040741,
      "rewards/forensics_strict_edge_reward": 0.015703186159953475,
      "rewards/forensics_tiou_reward": 0.35190731287002563,
      "step": 448
    },
    {
      "completion_length": 22.546875,
      "epoch": 1.3778801843317972,
      "grad_norm": 5.533677101135254,
      "kl": 0.298828125,
      "learning_rate": 6.553846153846153e-07,
      "loss": 0.0024,
      "reward": 0.6700518727302551,
      "reward_std": 0.07927838154137135,
      "reward_std_all": 0.10618618130683899,
      "rewards-max/forensics_iou_reward": 0.9983866214752197,
      "rewards-max/forensics_strict_edge_reward": 0.6419725120067596,
      "rewards-max/forensics_tiou_reward": 0.9974254071712494,
      "rewards/forensics_iou_reward": 0.593960165977478,
      "rewards/forensics_strict_edge_reward": 0.0443322379142046,
      "rewards/forensics_tiou_reward": 0.5412552058696747,
      "step": 449
    },
    {
      "completion_length": 29.0,
      "epoch": 1.380952380952381,
      "grad_norm": 7.94912052154541,
      "kl": 0.66796875,
      "learning_rate": 6.546153846153845e-07,
      "loss": 0.1725,
      "reward": 0.6294633746147156,
      "reward_std": 0.13696391135454178,
      "reward_std_all": 0.15824128314852715,
      "rewards-max/forensics_iou_reward": 0.9134518504142761,
      "rewards-max/forensics_strict_edge_reward": 0.3265264481306076,
      "rewards-max/forensics_tiou_reward": 0.8982100784778595,
      "rewards/forensics_iou_reward": 0.4704541265964508,
      "rewards/forensics_strict_edge_reward": 0.029135121032595634,
      "rewards/forensics_tiou_reward": 0.42772528529167175,
      "step": 450
    },
    {
      "completion_length": 27.21875,
      "epoch": 1.3840245775729647,
      "grad_norm": 6.9923319816589355,
      "kl": 0.3779296875,
      "learning_rate": 6.538461538461538e-07,
      "loss": 0.1163,
      "reward": 0.652923047542572,
      "reward_std": 0.08958507515490055,
      "reward_std_all": 0.11909551173448563,
      "rewards-max/forensics_iou_reward": 0.9227674007415771,
      "rewards-max/forensics_strict_edge_reward": 0.22803419828414917,
      "rewards-max/forensics_tiou_reward": 0.9015920758247375,
      "rewards/forensics_iou_reward": 0.49112868309020996,
      "rewards/forensics_strict_edge_reward": 0.010781664401292801,
      "rewards/forensics_tiou_reward": 0.4240158498287201,
      "step": 451
    },
    {
      "completion_length": 23.265625,
      "epoch": 1.3870967741935485,
      "grad_norm": 7.158519744873047,
      "kl": 0.25341796875,
      "learning_rate": 6.530769230769231e-07,
      "loss": 0.1379,
      "reward": 0.63118577003479,
      "reward_std": 0.1434144675731659,
      "reward_std_all": 0.15944281220436096,
      "rewards-max/forensics_iou_reward": 0.9441290497779846,
      "rewards-max/forensics_strict_edge_reward": 0.1498393788933754,
      "rewards-max/forensics_tiou_reward": 0.9263838529586792,
      "rewards/forensics_iou_reward": 0.49709028005599976,
      "rewards/forensics_strict_edge_reward": 0.0135544347576797,
      "rewards/forensics_tiou_reward": 0.45116403698921204,
      "step": 452
    },
    {
      "completion_length": 22.609375,
      "epoch": 1.390168970814132,
      "grad_norm": 5.0814595222473145,
      "kl": 0.326171875,
      "learning_rate": 6.523076923076923e-07,
      "loss": -0.0565,
      "reward": 0.7233020663261414,
      "reward_std": 0.0832149013876915,
      "reward_std_all": 0.08259089104831219,
      "rewards-max/forensics_iou_reward": 0.9574726819992065,
      "rewards-max/forensics_strict_edge_reward": 0.41128233075141907,
      "rewards-max/forensics_tiou_reward": 0.9440834820270538,
      "rewards/forensics_iou_reward": 0.586801290512085,
      "rewards/forensics_strict_edge_reward": 0.04887716472148895,
      "rewards/forensics_tiou_reward": 0.5341518223285675,
      "step": 453
    },
    {
      "completion_length": 22.828125,
      "epoch": 1.3932411674347158,
      "grad_norm": 5.239348411560059,
      "kl": 0.396484375,
      "learning_rate": 6.515384615384615e-07,
      "loss": 0.0744,
      "reward": 0.7642963528633118,
      "reward_std": 0.0711839497089386,
      "reward_std_all": 0.06356754712760448,
      "rewards-max/forensics_iou_reward": 0.9591346979141235,
      "rewards-max/forensics_strict_edge_reward": 0.46161630749702454,
      "rewards-max/forensics_tiou_reward": 0.9502312242984772,
      "rewards/forensics_iou_reward": 0.6461427807807922,
      "rewards/forensics_strict_edge_reward": 0.055915359407663345,
      "rewards/forensics_tiou_reward": 0.5887811481952667,
      "step": 454
    },
    {
      "completion_length": 23.953125,
      "epoch": 1.3963133640552996,
      "grad_norm": 6.680677890777588,
      "kl": 0.3720703125,
      "learning_rate": 6.507692307692307e-07,
      "loss": 0.0088,
      "reward": 0.6879346072673798,
      "reward_std": 0.0858320789411664,
      "reward_std_all": 0.09559804573655128,
      "rewards-max/forensics_iou_reward": 0.9700866937637329,
      "rewards-max/forensics_strict_edge_reward": 0.5103136450052261,
      "rewards-max/forensics_tiou_reward": 0.95859295129776,
      "rewards/forensics_iou_reward": 0.5657570660114288,
      "rewards/forensics_strict_edge_reward": 0.03092574095353484,
      "rewards/forensics_tiou_reward": 0.5090218484401703,
      "step": 455
    },
    {
      "completion_length": 25.546875,
      "epoch": 1.3993855606758832,
      "grad_norm": 5.748649597167969,
      "kl": 0.33203125,
      "learning_rate": 6.5e-07,
      "loss": 0.1628,
      "reward": 0.6666893362998962,
      "reward_std": 0.16212105378508568,
      "reward_std_all": 0.1534728966653347,
      "rewards-max/forensics_iou_reward": 0.974022626876831,
      "rewards-max/forensics_strict_edge_reward": 0.5488366484642029,
      "rewards-max/forensics_tiou_reward": 0.9665674567222595,
      "rewards/forensics_iou_reward": 0.5553788989782333,
      "rewards/forensics_strict_edge_reward": 0.038200367242097855,
      "rewards/forensics_tiou_reward": 0.4962223172187805,
      "step": 456
    },
    {
      "completion_length": 22.390625,
      "epoch": 1.402457757296467,
      "grad_norm": 5.072146892547607,
      "kl": 0.4521484375,
      "learning_rate": 6.492307692307692e-07,
      "loss": 0.1507,
      "reward": 0.6521851718425751,
      "reward_std": 0.07423754781484604,
      "reward_std_all": 0.11607379466295242,
      "rewards-max/forensics_iou_reward": 0.9724538922309875,
      "rewards-max/forensics_strict_edge_reward": 0.6924179792404175,
      "rewards-max/forensics_tiou_reward": 0.9633187353610992,
      "rewards/forensics_iou_reward": 0.5202349424362183,
      "rewards/forensics_strict_edge_reward": 0.10609496757388115,
      "rewards/forensics_tiou_reward": 0.47820915281772614,
      "step": 457
    },
    {
      "completion_length": 25.890625,
      "epoch": 1.4055299539170507,
      "grad_norm": 6.252277851104736,
      "kl": 0.380859375,
      "learning_rate": 6.484615384615383e-07,
      "loss": 0.0098,
      "reward": 0.6375003159046173,
      "reward_std": 0.14581776410341263,
      "reward_std_all": 0.16114775836467743,
      "rewards-max/forensics_iou_reward": 0.913091391324997,
      "rewards-max/forensics_strict_edge_reward": 0.14407383650541306,
      "rewards-max/forensics_tiou_reward": 0.8899465203285217,
      "rewards/forensics_iou_reward": 0.48439520597457886,
      "rewards/forensics_strict_edge_reward": 0.011287650791928172,
      "rewards/forensics_tiou_reward": 0.4474344104528427,
      "step": 458
    },
    {
      "completion_length": 23.671875,
      "epoch": 1.4086021505376345,
      "grad_norm": 4.142276287078857,
      "kl": 0.345703125,
      "learning_rate": 6.476923076923077e-07,
      "loss": -0.0091,
      "reward": 0.7298089563846588,
      "reward_std": 0.04329187236726284,
      "reward_std_all": 0.06324986647814512,
      "rewards-max/forensics_iou_reward": 0.9694709181785583,
      "rewards-max/forensics_strict_edge_reward": 0.6483443379402161,
      "rewards-max/forensics_tiou_reward": 0.9628623425960541,
      "rewards/forensics_iou_reward": 0.6223074197769165,
      "rewards/forensics_strict_edge_reward": 0.04295025393366814,
      "rewards/forensics_tiou_reward": 0.5673373639583588,
      "step": 459
    },
    {
      "completion_length": 25.390625,
      "epoch": 1.411674347158218,
      "grad_norm": 5.407529830932617,
      "kl": 0.3779296875,
      "learning_rate": 6.469230769230769e-07,
      "loss": 0.0962,
      "reward": 0.6213875114917755,
      "reward_std": 0.1068835873156786,
      "reward_std_all": 0.14999115094542503,
      "rewards-max/forensics_iou_reward": 0.9833064079284668,
      "rewards-max/forensics_strict_edge_reward": 0.3499405235052109,
      "rewards-max/forensics_tiou_reward": 0.9747227132320404,
      "rewards/forensics_iou_reward": 0.4800952076911926,
      "rewards/forensics_strict_edge_reward": 0.027741973754018545,
      "rewards/forensics_tiou_reward": 0.4332116097211838,
      "step": 460
    },
    {
      "completion_length": 22.859375,
      "epoch": 1.4147465437788018,
      "grad_norm": 6.112362861633301,
      "kl": 0.3359375,
      "learning_rate": 6.461538461538462e-07,
      "loss": 0.2112,
      "reward": 0.6409595012664795,
      "reward_std": 0.14235924184322357,
      "reward_std_all": 0.1520768254995346,
      "rewards-max/forensics_iou_reward": 0.9502758383750916,
      "rewards-max/forensics_strict_edge_reward": 0.219696044921875,
      "rewards-max/forensics_tiou_reward": 0.9238390922546387,
      "rewards/forensics_iou_reward": 0.5049586147069931,
      "rewards/forensics_strict_edge_reward": 0.02305492665618658,
      "rewards/forensics_tiou_reward": 0.45087260007858276,
      "step": 461
    },
    {
      "completion_length": 24.875,
      "epoch": 1.4178187403993856,
      "grad_norm": 5.173044681549072,
      "kl": 0.3251953125,
      "learning_rate": 6.453846153846153e-07,
      "loss": 0.0346,
      "reward": 0.7384224832057953,
      "reward_std": 0.06632105261087418,
      "reward_std_all": 0.06609673425555229,
      "rewards-max/forensics_iou_reward": 0.9332659244537354,
      "rewards-max/forensics_strict_edge_reward": 0.29944878816604614,
      "rewards-max/forensics_tiou_reward": 0.9157986640930176,
      "rewards/forensics_iou_reward": 0.6030387580394745,
      "rewards/forensics_strict_edge_reward": 0.023665947373956442,
      "rewards/forensics_tiou_reward": 0.5417693555355072,
      "step": 462
    },
    {
      "completion_length": 27.09375,
      "epoch": 1.4208909370199692,
      "grad_norm": 4.891208171844482,
      "kl": 0.3251953125,
      "learning_rate": 6.446153846153845e-07,
      "loss": 0.17,
      "reward": 0.6532774567604065,
      "reward_std": 0.09594962000846863,
      "reward_std_all": 0.12466928362846375,
      "rewards-max/forensics_iou_reward": 0.9479199349880219,
      "rewards-max/forensics_strict_edge_reward": 0.17485969187691808,
      "rewards-max/forensics_tiou_reward": 0.9366315305233002,
      "rewards/forensics_iou_reward": 0.4960761219263077,
      "rewards/forensics_strict_edge_reward": 0.0081591626003501,
      "rewards/forensics_tiou_reward": 0.43432237207889557,
      "step": 463
    },
    {
      "completion_length": 25.40625,
      "epoch": 1.423963133640553,
      "grad_norm": 4.996828556060791,
      "kl": 0.3720703125,
      "learning_rate": 6.438461538461538e-07,
      "loss": -0.1191,
      "reward": 0.6122208833694458,
      "reward_std": 0.11945977807044983,
      "reward_std_all": 0.15277986228466034,
      "rewards-max/forensics_iou_reward": 0.9521198272705078,
      "rewards-max/forensics_strict_edge_reward": 0.5147627294063568,
      "rewards-max/forensics_tiou_reward": 0.9410130083560944,
      "rewards/forensics_iou_reward": 0.4714304059743881,
      "rewards/forensics_strict_edge_reward": 0.02773395087569952,
      "rewards/forensics_tiou_reward": 0.42651090025901794,
      "step": 464
    },
    {
      "completion_length": 27.4375,
      "epoch": 1.4270353302611367,
      "grad_norm": 6.867246627807617,
      "kl": 0.466796875,
      "learning_rate": 6.43076923076923e-07,
      "loss": 0.0453,
      "reward": 0.6977837681770325,
      "reward_std": 0.02093851286917925,
      "reward_std_all": 0.06662820279598236,
      "rewards-max/forensics_iou_reward": 0.9598673284053802,
      "rewards-max/forensics_strict_edge_reward": 0.32198866456747055,
      "rewards-max/forensics_tiou_reward": 0.9477216601371765,
      "rewards/forensics_iou_reward": 0.5862761437892914,
      "rewards/forensics_strict_edge_reward": 0.017159256618469954,
      "rewards/forensics_tiou_reward": 0.5287370681762695,
      "step": 465
    },
    {
      "completion_length": 29.03125,
      "epoch": 1.4301075268817205,
      "grad_norm": 3.6565229892730713,
      "kl": 0.330078125,
      "learning_rate": 6.423076923076924e-07,
      "loss": 0.0872,
      "reward": 0.7121166884899139,
      "reward_std": 0.08547373488545418,
      "reward_std_all": 0.0894966870546341,
      "rewards-max/forensics_iou_reward": 0.968329906463623,
      "rewards-max/forensics_strict_edge_reward": 0.3882795572280884,
      "rewards-max/forensics_tiou_reward": 0.9538579881191254,
      "rewards/forensics_iou_reward": 0.5184357762336731,
      "rewards/forensics_strict_edge_reward": 0.043173227459192276,
      "rewards/forensics_tiou_reward": 0.47334274649620056,
      "step": 466
    },
    {
      "completion_length": 25.890625,
      "epoch": 1.4331797235023043,
      "grad_norm": 5.4338836669921875,
      "kl": 0.5029296875,
      "learning_rate": 6.415384615384615e-07,
      "loss": 0.1081,
      "reward": 0.7018802464008331,
      "reward_std": 0.1220010332763195,
      "reward_std_all": 0.11521472036838531,
      "rewards-max/forensics_iou_reward": 0.9433930218219757,
      "rewards-max/forensics_strict_edge_reward": 0.16367435455322266,
      "rewards-max/forensics_tiou_reward": 0.9178730845451355,
      "rewards/forensics_iou_reward": 0.5608360171318054,
      "rewards/forensics_strict_edge_reward": 0.015891857910901308,
      "rewards/forensics_tiou_reward": 0.4793364703655243,
      "step": 467
    },
    {
      "completion_length": 24.546875,
      "epoch": 1.4362519201228878,
      "grad_norm": 6.503259658813477,
      "kl": 0.365234375,
      "learning_rate": 6.407692307692307e-07,
      "loss": 0.108,
      "reward": 0.5961237549781799,
      "reward_std": 0.11578844860196114,
      "reward_std_all": 0.15871620178222656,
      "rewards-max/forensics_iou_reward": 0.959668904542923,
      "rewards-max/forensics_strict_edge_reward": 0.41361792385578156,
      "rewards-max/forensics_tiou_reward": 0.9317748248577118,
      "rewards/forensics_iou_reward": 0.4115607440471649,
      "rewards/forensics_strict_edge_reward": 0.028699157759547234,
      "rewards/forensics_tiou_reward": 0.3626396358013153,
      "step": 468
    },
    {
      "completion_length": 28.609375,
      "epoch": 1.4393241167434716,
      "grad_norm": 5.042266368865967,
      "kl": 0.408203125,
      "learning_rate": 6.4e-07,
      "loss": 0.0746,
      "reward": 0.683025598526001,
      "reward_std": 0.08404945954680443,
      "reward_std_all": 0.10131489485502243,
      "rewards-max/forensics_iou_reward": 0.9619409441947937,
      "rewards-max/forensics_strict_edge_reward": 0.5498600900173187,
      "rewards-max/forensics_tiou_reward": 0.9516618847846985,
      "rewards/forensics_iou_reward": 0.4969891905784607,
      "rewards/forensics_strict_edge_reward": 0.02971789985895157,
      "rewards/forensics_tiou_reward": 0.44470421969890594,
      "step": 469
    },
    {
      "completion_length": 23.09375,
      "epoch": 1.4423963133640554,
      "grad_norm": 4.49998664855957,
      "kl": 0.4443359375,
      "learning_rate": 6.392307692307692e-07,
      "loss": 0.0956,
      "reward": 0.668908029794693,
      "reward_std": 0.13639061897993088,
      "reward_std_all": 0.14223026484251022,
      "rewards-max/forensics_iou_reward": 0.9475840330123901,
      "rewards-max/forensics_strict_edge_reward": 0.254275880753994,
      "rewards-max/forensics_tiou_reward": 0.9287840723991394,
      "rewards/forensics_iou_reward": 0.5485864877700806,
      "rewards/forensics_strict_edge_reward": 0.024652190040796995,
      "rewards/forensics_tiou_reward": 0.5111265927553177,
      "step": 470
    },
    {
      "completion_length": 25.15625,
      "epoch": 1.445468509984639,
      "grad_norm": 4.54750394821167,
      "kl": 0.3291015625,
      "learning_rate": 6.384615384615383e-07,
      "loss": 0.1028,
      "reward": 0.6643122136592865,
      "reward_std": 0.10670556128025055,
      "reward_std_all": 0.12561754509806633,
      "rewards-max/forensics_iou_reward": 0.9502435624599457,
      "rewards-max/forensics_strict_edge_reward": 0.48720505833625793,
      "rewards-max/forensics_tiou_reward": 0.933752179145813,
      "rewards/forensics_iou_reward": 0.5446836948394775,
      "rewards/forensics_strict_edge_reward": 0.042369596660137177,
      "rewards/forensics_tiou_reward": 0.5045469999313354,
      "step": 471
    },
    {
      "completion_length": 25.921875,
      "epoch": 1.4485407066052227,
      "grad_norm": 4.876191139221191,
      "kl": 0.3759765625,
      "learning_rate": 6.376923076923077e-07,
      "loss": 0.0315,
      "reward": 0.7820148468017578,
      "reward_std": 0.03230788931250572,
      "reward_std_all": 0.031228145584464073,
      "rewards-max/forensics_iou_reward": 0.9795485138893127,
      "rewards-max/forensics_strict_edge_reward": 0.5547693520784378,
      "rewards-max/forensics_tiou_reward": 0.9712738990783691,
      "rewards/forensics_iou_reward": 0.6673344373703003,
      "rewards/forensics_strict_edge_reward": 0.03595694340765476,
      "rewards/forensics_tiou_reward": 0.6127685904502869,
      "step": 472
    },
    {
      "completion_length": 27.0625,
      "epoch": 1.4516129032258065,
      "grad_norm": 5.5333428382873535,
      "kl": 0.345703125,
      "learning_rate": 6.369230769230769e-07,
      "loss": 0.0391,
      "reward": 0.7099296450614929,
      "reward_std": 0.08638626802712679,
      "reward_std_all": 0.087684640660882,
      "rewards-max/forensics_iou_reward": 0.9521216750144958,
      "rewards-max/forensics_strict_edge_reward": 0.2893613278865814,
      "rewards-max/forensics_tiou_reward": 0.9374245703220367,
      "rewards/forensics_iou_reward": 0.5594172030687332,
      "rewards/forensics_strict_edge_reward": 0.020254437811672688,
      "rewards/forensics_tiou_reward": 0.5108682662248611,
      "step": 473
    },
    {
      "completion_length": 27.75,
      "epoch": 1.4546850998463903,
      "grad_norm": 5.78364372253418,
      "kl": 0.298828125,
      "learning_rate": 6.361538461538462e-07,
      "loss": 0.0184,
      "reward": 0.6639237403869629,
      "reward_std": 0.06803126633167267,
      "reward_std_all": 0.10122213512659073,
      "rewards-max/forensics_iou_reward": 0.9771517217159271,
      "rewards-max/forensics_strict_edge_reward": 0.5038549304008484,
      "rewards-max/forensics_tiou_reward": 0.9702689051628113,
      "rewards/forensics_iou_reward": 0.564005583524704,
      "rewards/forensics_strict_edge_reward": 0.039224009960889816,
      "rewards/forensics_tiou_reward": 0.5182730853557587,
      "step": 474
    },
    {
      "completion_length": 26.328125,
      "epoch": 1.4577572964669738,
      "grad_norm": 6.138429164886475,
      "kl": 0.328125,
      "learning_rate": 6.353846153846154e-07,
      "loss": 0.2347,
      "reward": 0.6364185214042664,
      "reward_std": 0.17742565274238586,
      "reward_std_all": 0.17068522423505783,
      "rewards-max/forensics_iou_reward": 0.9773926436901093,
      "rewards-max/forensics_strict_edge_reward": 0.48442691564559937,
      "rewards-max/forensics_tiou_reward": 0.9682197570800781,
      "rewards/forensics_iou_reward": 0.4949900954961777,
      "rewards/forensics_strict_edge_reward": 0.02911949809640646,
      "rewards/forensics_tiou_reward": 0.44842541217803955,
      "step": 475
    },
    {
      "completion_length": 22.96875,
      "epoch": 1.4608294930875576,
      "grad_norm": 5.5980916023254395,
      "kl": 0.3134765625,
      "learning_rate": 6.346153846153845e-07,
      "loss": 0.1082,
      "reward": 0.6685767769813538,
      "reward_std": 0.06004274729639292,
      "reward_std_all": 0.1018824353814125,
      "rewards-max/forensics_iou_reward": 0.996028482913971,
      "rewards-max/forensics_strict_edge_reward": 0.6416448950767517,
      "rewards-max/forensics_tiou_reward": 0.9923886954784393,
      "rewards/forensics_iou_reward": 0.5203126221895218,
      "rewards/forensics_strict_edge_reward": 0.042392488569021225,
      "rewards/forensics_tiou_reward": 0.47809460759162903,
      "step": 476
    },
    {
      "completion_length": 24.125,
      "epoch": 1.4639016897081414,
      "grad_norm": 6.316850185394287,
      "kl": 0.341796875,
      "learning_rate": 6.338461538461538e-07,
      "loss": 0.0162,
      "reward": 0.7527750730514526,
      "reward_std": 0.04591474309563637,
      "reward_std_all": 0.04912247881293297,
      "rewards-max/forensics_iou_reward": 0.9947492182254791,
      "rewards-max/forensics_strict_edge_reward": 0.6934256851673126,
      "rewards-max/forensics_tiou_reward": 0.9925904870033264,
      "rewards/forensics_iou_reward": 0.6167150437831879,
      "rewards/forensics_strict_edge_reward": 0.053121183067560196,
      "rewards/forensics_tiou_reward": 0.558030515909195,
      "step": 477
    },
    {
      "completion_length": 24.4375,
      "epoch": 1.466973886328725,
      "grad_norm": 5.973873615264893,
      "kl": 0.33984375,
      "learning_rate": 6.33076923076923e-07,
      "loss": 0.175,
      "reward": 0.6922146081924438,
      "reward_std": 0.1308194138109684,
      "reward_std_all": 0.12610311061143875,
      "rewards-max/forensics_iou_reward": 0.9309617877006531,
      "rewards-max/forensics_strict_edge_reward": 0.3020675405859947,
      "rewards-max/forensics_tiou_reward": 0.9121718406677246,
      "rewards/forensics_iou_reward": 0.5452599227428436,
      "rewards/forensics_strict_edge_reward": 0.03018319606781006,
      "rewards/forensics_tiou_reward": 0.4949573874473572,
      "step": 478
    },
    {
      "completion_length": 22.875,
      "epoch": 1.4700460829493087,
      "grad_norm": 7.6434197425842285,
      "kl": 0.40625,
      "learning_rate": 6.323076923076924e-07,
      "loss": 0.1921,
      "reward": 0.6279900670051575,
      "reward_std": 0.14759209752082825,
      "reward_std_all": 0.16489923745393753,
      "rewards-max/forensics_iou_reward": 0.9050567150115967,
      "rewards-max/forensics_strict_edge_reward": 0.22299528494477272,
      "rewards-max/forensics_tiou_reward": 0.8793206810951233,
      "rewards/forensics_iou_reward": 0.4535537660121918,
      "rewards/forensics_strict_edge_reward": 0.026823554886505008,
      "rewards/forensics_tiou_reward": 0.392890065908432,
      "step": 479
    },
    {
      "completion_length": 24.0625,
      "epoch": 1.4731182795698925,
      "grad_norm": 6.06693696975708,
      "kl": 0.349609375,
      "learning_rate": 6.315384615384615e-07,
      "loss": 0.1386,
      "reward": 0.7076850831508636,
      "reward_std": 0.12297254800796509,
      "reward_std_all": 0.11773275956511497,
      "rewards-max/forensics_iou_reward": 0.9706349074840546,
      "rewards-max/forensics_strict_edge_reward": 0.47731224820017815,
      "rewards-max/forensics_tiou_reward": 0.9620364010334015,
      "rewards/forensics_iou_reward": 0.5595030784606934,
      "rewards/forensics_strict_edge_reward": 0.0440658344887197,
      "rewards/forensics_tiou_reward": 0.5142242163419724,
      "step": 480
    },
    {
      "completion_length": 28.546875,
      "epoch": 1.4761904761904763,
      "grad_norm": 4.748407363891602,
      "kl": 0.3544921875,
      "learning_rate": 6.307692307692307e-07,
      "loss": 0.0936,
      "reward": 0.7318531572818756,
      "reward_std": 0.10247278213500977,
      "reward_std_all": 0.09402942657470703,
      "rewards-max/forensics_iou_reward": 0.9663960039615631,
      "rewards-max/forensics_strict_edge_reward": 0.3907497674226761,
      "rewards-max/forensics_tiou_reward": 0.95533287525177,
      "rewards/forensics_iou_reward": 0.5609209537506104,
      "rewards/forensics_strict_edge_reward": 0.01883348962292075,
      "rewards/forensics_tiou_reward": 0.5176016092300415,
      "step": 481
    },
    {
      "completion_length": 23.5,
      "epoch": 1.4792626728110598,
      "grad_norm": 5.986233234405518,
      "kl": 0.37890625,
      "learning_rate": 6.3e-07,
      "loss": 0.1509,
      "reward": 0.6349869668483734,
      "reward_std": 0.11118028312921524,
      "reward_std_all": 0.13752485811710358,
      "rewards-max/forensics_iou_reward": 0.8509619832038879,
      "rewards-max/forensics_strict_edge_reward": 0.17803993867710233,
      "rewards-max/forensics_tiou_reward": 0.8137508928775787,
      "rewards/forensics_iou_reward": 0.4778190851211548,
      "rewards/forensics_strict_edge_reward": 0.01606503626680933,
      "rewards/forensics_tiou_reward": 0.4224878251552582,
      "step": 482
    },
    {
      "completion_length": 24.515625,
      "epoch": 1.4823348694316436,
      "grad_norm": 4.983444690704346,
      "kl": 0.40234375,
      "learning_rate": 6.292307692307692e-07,
      "loss": 0.0439,
      "reward": 0.7007707953453064,
      "reward_std": 0.12685536593198776,
      "reward_std_all": 0.11939101293683052,
      "rewards-max/forensics_iou_reward": 0.971688836812973,
      "rewards-max/forensics_strict_edge_reward": 0.25927625596523285,
      "rewards-max/forensics_tiou_reward": 0.9564028382301331,
      "rewards/forensics_iou_reward": 0.5700190365314484,
      "rewards/forensics_strict_edge_reward": 0.023974135518074036,
      "rewards/forensics_tiou_reward": 0.5177991092205048,
      "step": 483
    },
    {
      "completion_length": 24.953125,
      "epoch": 1.4854070660522274,
      "grad_norm": 5.228448867797852,
      "kl": 0.6015625,
      "learning_rate": 6.284615384615385e-07,
      "loss": 0.0944,
      "reward": 0.6995936036109924,
      "reward_std": 0.1236780434846878,
      "reward_std_all": 0.11646902188658714,
      "rewards-max/forensics_iou_reward": 0.9666478335857391,
      "rewards-max/forensics_strict_edge_reward": 0.273137629032135,
      "rewards-max/forensics_tiou_reward": 0.9525490999221802,
      "rewards/forensics_iou_reward": 0.5367884039878845,
      "rewards/forensics_strict_edge_reward": 0.028089101426303387,
      "rewards/forensics_tiou_reward": 0.484500527381897,
      "step": 484
    },
    {
      "completion_length": 29.375,
      "epoch": 1.488479262672811,
      "grad_norm": 5.131180763244629,
      "kl": 0.357421875,
      "learning_rate": 6.276923076923076e-07,
      "loss": 0.0925,
      "reward": 0.7271488606929779,
      "reward_std": 0.09759427327662706,
      "reward_std_all": 0.09149017091840506,
      "rewards-max/forensics_iou_reward": 0.9466017484664917,
      "rewards-max/forensics_strict_edge_reward": 0.3144293576478958,
      "rewards-max/forensics_tiou_reward": 0.9319519102573395,
      "rewards/forensics_iou_reward": 0.5753459930419922,
      "rewards/forensics_strict_edge_reward": 0.030580685939639807,
      "rewards/forensics_tiou_reward": 0.5252577662467957,
      "step": 485
    },
    {
      "completion_length": 26.03125,
      "epoch": 1.4915514592933947,
      "grad_norm": 5.798637390136719,
      "kl": 0.326171875,
      "learning_rate": 6.269230769230769e-07,
      "loss": 0.0112,
      "reward": 0.6102528870105743,
      "reward_std": 0.1014399565756321,
      "reward_std_all": 0.13774704933166504,
      "rewards-max/forensics_iou_reward": 0.822021096944809,
      "rewards-max/forensics_strict_edge_reward": 0.3160267360508442,
      "rewards-max/forensics_tiou_reward": 0.8189474940299988,
      "rewards/forensics_iou_reward": 0.4522120952606201,
      "rewards/forensics_strict_edge_reward": 0.017111361368733924,
      "rewards/forensics_tiou_reward": 0.40867385268211365,
      "step": 486
    },
    {
      "completion_length": 26.671875,
      "epoch": 1.4946236559139785,
      "grad_norm": 6.495605945587158,
      "kl": 0.349609375,
      "learning_rate": 6.261538461538462e-07,
      "loss": -0.0289,
      "reward": 0.7149217128753662,
      "reward_std": 0.07513854838907719,
      "reward_std_all": 0.0799688883125782,
      "rewards-max/forensics_iou_reward": 0.9763972461223602,
      "rewards-max/forensics_strict_edge_reward": 0.43545667827129364,
      "rewards-max/forensics_tiou_reward": 0.9694432616233826,
      "rewards/forensics_iou_reward": 0.5627192854881287,
      "rewards/forensics_strict_edge_reward": 0.04528965428471565,
      "rewards/forensics_tiou_reward": 0.5053860247135162,
      "step": 487
    },
    {
      "completion_length": 26.96875,
      "epoch": 1.4976958525345623,
      "grad_norm": 5.848689079284668,
      "kl": 0.318359375,
      "learning_rate": 6.253846153846154e-07,
      "loss": 0.1908,
      "reward": 0.6891620755195618,
      "reward_std": 0.09715444222092628,
      "reward_std_all": 0.11356155201792717,
      "rewards-max/forensics_iou_reward": 0.9723919928073883,
      "rewards-max/forensics_strict_edge_reward": 0.6934256851673126,
      "rewards-max/forensics_tiou_reward": 0.9678677916526794,
      "rewards/forensics_iou_reward": 0.5610040426254272,
      "rewards/forensics_strict_edge_reward": 0.050892289727926254,
      "rewards/forensics_tiou_reward": 0.511266827583313,
      "step": 488
    },
    {
      "completion_length": 30.953125,
      "epoch": 1.500768049155146,
      "grad_norm": 5.059872150421143,
      "kl": 0.3134765625,
      "learning_rate": 6.246153846153845e-07,
      "loss": 0.0126,
      "reward": 0.6535323858261108,
      "reward_std": 0.11670014262199402,
      "reward_std_all": 0.13630925118923187,
      "rewards-max/forensics_iou_reward": 0.895530492067337,
      "rewards-max/forensics_strict_edge_reward": 0.17418761923909187,
      "rewards-max/forensics_tiou_reward": 0.8740650117397308,
      "rewards/forensics_iou_reward": 0.51125168800354,
      "rewards/forensics_strict_edge_reward": 0.02905715099768713,
      "rewards/forensics_tiou_reward": 0.462589293718338,
      "step": 489
    },
    {
      "completion_length": 26.390625,
      "epoch": 1.5038402457757296,
      "grad_norm": 6.775400161743164,
      "kl": 0.3408203125,
      "learning_rate": 6.238461538461538e-07,
      "loss": 0.1341,
      "reward": 0.6764044761657715,
      "reward_std": 0.1155969500541687,
      "reward_std_all": 0.12869831919670105,
      "rewards-max/forensics_iou_reward": 0.9943052530288696,
      "rewards-max/forensics_strict_edge_reward": 0.557279422879219,
      "rewards-max/forensics_tiou_reward": 0.9918281733989716,
      "rewards/forensics_iou_reward": 0.52678182721138,
      "rewards/forensics_strict_edge_reward": 0.03500758996233344,
      "rewards/forensics_tiou_reward": 0.46893975138664246,
      "step": 490
    },
    {
      "completion_length": 22.140625,
      "epoch": 1.5069124423963134,
      "grad_norm": 6.35023307800293,
      "kl": 0.314453125,
      "learning_rate": 6.23076923076923e-07,
      "loss": -0.0748,
      "reward": 0.6732905507087708,
      "reward_std": 0.11436184495687485,
      "reward_std_all": 0.12519168853759766,
      "rewards-max/forensics_iou_reward": 0.9863455295562744,
      "rewards-max/forensics_strict_edge_reward": 0.6236754059791565,
      "rewards-max/forensics_tiou_reward": 0.9816732406616211,
      "rewards/forensics_iou_reward": 0.5683950781822205,
      "rewards/forensics_strict_edge_reward": 0.055614128708839417,
      "rewards/forensics_tiou_reward": 0.5241069495677948,
      "step": 491
    },
    {
      "completion_length": 24.65625,
      "epoch": 1.509984639016897,
      "grad_norm": 6.259958744049072,
      "kl": 0.4306640625,
      "learning_rate": 6.223076923076924e-07,
      "loss": 0.0149,
      "reward": 0.7138673663139343,
      "reward_std": 0.030766712501645088,
      "reward_std_all": 0.061590700410306454,
      "rewards-max/forensics_iou_reward": 0.9460338652133942,
      "rewards-max/forensics_strict_edge_reward": 0.26217253506183624,
      "rewards-max/forensics_tiou_reward": 0.9229298830032349,
      "rewards/forensics_iou_reward": 0.582701787352562,
      "rewards/forensics_strict_edge_reward": 0.022015956230461597,
      "rewards/forensics_tiou_reward": 0.5206962078809738,
      "step": 492
    },
    {
      "completion_length": 26.1875,
      "epoch": 1.5130568356374807,
      "grad_norm": 5.921154022216797,
      "kl": 0.337890625,
      "learning_rate": 6.215384615384615e-07,
      "loss": 0.0988,
      "reward": 0.7660057544708252,
      "reward_std": 0.027062295004725456,
      "reward_std_all": 0.03374883811920881,
      "rewards-max/forensics_iou_reward": 0.9604771733283997,
      "rewards-max/forensics_strict_edge_reward": 0.27708966471254826,
      "rewards-max/forensics_tiou_reward": 0.9401997625827789,
      "rewards/forensics_iou_reward": 0.655440628528595,
      "rewards/forensics_strict_edge_reward": 0.020505023188889027,
      "rewards/forensics_tiou_reward": 0.5884033739566803,
      "step": 493
    },
    {
      "completion_length": 25.9375,
      "epoch": 1.5161290322580645,
      "grad_norm": 5.098851680755615,
      "kl": 0.330078125,
      "learning_rate": 6.207692307692307e-07,
      "loss": 0.1313,
      "reward": 0.7372184693813324,
      "reward_std": 0.05771014094352722,
      "reward_std_all": 0.06391842570155859,
      "rewards-max/forensics_iou_reward": 0.9919030070304871,
      "rewards-max/forensics_strict_edge_reward": 0.661531001329422,
      "rewards-max/forensics_tiou_reward": 0.9886443912982941,
      "rewards/forensics_iou_reward": 0.617232084274292,
      "rewards/forensics_strict_edge_reward": 0.0534660778939724,
      "rewards/forensics_tiou_reward": 0.548339307308197,
      "step": 494
    },
    {
      "completion_length": 23.234375,
      "epoch": 1.5192012288786483,
      "grad_norm": 6.492910385131836,
      "kl": 0.3349609375,
      "learning_rate": 6.2e-07,
      "loss": 0.1414,
      "reward": 0.7042137980461121,
      "reward_std": 0.04248142056167126,
      "reward_std_all": 0.07167481631040573,
      "rewards-max/forensics_iou_reward": 0.9739266932010651,
      "rewards-max/forensics_strict_edge_reward": 0.4982926547527313,
      "rewards-max/forensics_tiou_reward": 0.9593704342842102,
      "rewards/forensics_iou_reward": 0.6231836080551147,
      "rewards/forensics_strict_edge_reward": 0.040813472121953964,
      "rewards/forensics_tiou_reward": 0.5635472983121872,
      "step": 495
    },
    {
      "completion_length": 21.8125,
      "epoch": 1.522273425499232,
      "grad_norm": 4.930464267730713,
      "kl": 0.517578125,
      "learning_rate": 6.192307692307692e-07,
      "loss": 0.0739,
      "reward": 0.7171686291694641,
      "reward_std": 0.07775493338704109,
      "reward_std_all": 0.07990790903568268,
      "rewards-max/forensics_iou_reward": 0.8753772974014282,
      "rewards-max/forensics_strict_edge_reward": 0.0768254604190588,
      "rewards-max/forensics_tiou_reward": 0.8349394798278809,
      "rewards/forensics_iou_reward": 0.538688600063324,
      "rewards/forensics_strict_edge_reward": 0.0064443935407325625,
      "rewards/forensics_tiou_reward": 0.4661414474248886,
      "step": 496
    },
    {
      "completion_length": 24.40625,
      "epoch": 1.5253456221198156,
      "grad_norm": 5.78355598449707,
      "kl": 0.4013671875,
      "learning_rate": 6.184615384615385e-07,
      "loss": 0.152,
      "reward": 0.6461977958679199,
      "reward_std": 0.19446170330047607,
      "reward_std_all": 0.18154309689998627,
      "rewards-max/forensics_iou_reward": 0.8433448076248169,
      "rewards-max/forensics_strict_edge_reward": 0.21479890868067741,
      "rewards-max/forensics_tiou_reward": 0.7933209836483002,
      "rewards/forensics_iou_reward": 0.42959365248680115,
      "rewards/forensics_strict_edge_reward": 0.010608062613755465,
      "rewards/forensics_tiou_reward": 0.3709120899438858,
      "step": 497
    },
    {
      "completion_length": 24.546875,
      "epoch": 1.5284178187403994,
      "grad_norm": 6.810635089874268,
      "kl": 0.3486328125,
      "learning_rate": 6.176923076923076e-07,
      "loss": 0.1254,
      "reward": 0.6507909893989563,
      "reward_std": 0.062195638194680214,
      "reward_std_all": 0.10611281543970108,
      "rewards-max/forensics_iou_reward": 0.9604738056659698,
      "rewards-max/forensics_strict_edge_reward": 0.31587664037942886,
      "rewards-max/forensics_tiou_reward": 0.9478084146976471,
      "rewards/forensics_iou_reward": 0.5245119035243988,
      "rewards/forensics_strict_edge_reward": 0.025479106232523918,
      "rewards/forensics_tiou_reward": 0.4724101722240448,
      "step": 498
    },
    {
      "completion_length": 24.28125,
      "epoch": 1.531490015360983,
      "grad_norm": 8.104079246520996,
      "kl": 0.3349609375,
      "learning_rate": 6.169230769230768e-07,
      "loss": 0.0935,
      "reward": 0.6913678646087646,
      "reward_std": 0.10781843960285187,
      "reward_std_all": 0.10743297822773457,
      "rewards-max/forensics_iou_reward": 0.7386850416660309,
      "rewards-max/forensics_strict_edge_reward": 0.1467638574540615,
      "rewards-max/forensics_tiou_reward": 0.6707067787647247,
      "rewards/forensics_iou_reward": 0.47751379013061523,
      "rewards/forensics_strict_edge_reward": 0.013081966899335384,
      "rewards/forensics_tiou_reward": 0.4197104871273041,
      "step": 499
    },
    {
      "completion_length": 25.46875,
      "epoch": 1.5345622119815667,
      "grad_norm": 6.078488349914551,
      "kl": 0.375,
      "learning_rate": 6.161538461538462e-07,
      "loss": 0.0906,
      "reward": 0.6388368010520935,
      "reward_std": 0.11640679091215134,
      "reward_std_all": 0.13981958106160164,
      "rewards-max/forensics_iou_reward": 0.958302229642868,
      "rewards-max/forensics_strict_edge_reward": 0.39243943989276886,
      "rewards-max/forensics_tiou_reward": 0.9402360618114471,
      "rewards/forensics_iou_reward": 0.48629169166088104,
      "rewards/forensics_strict_edge_reward": 0.04356814827769995,
      "rewards/forensics_tiou_reward": 0.4341449439525604,
      "step": 500
    },
    {
      "completion_length": 25.453125,
      "epoch": 1.5376344086021505,
      "grad_norm": 5.434355735778809,
      "kl": 0.361328125,
      "learning_rate": 6.153846153846154e-07,
      "loss": 0.1799,
      "reward": 0.5867560505867004,
      "reward_std": 0.08144303783774376,
      "reward_std_all": 0.145419679582119,
      "rewards-max/forensics_iou_reward": 0.9771880209445953,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9726913273334503,
      "rewards/forensics_iou_reward": 0.4160909950733185,
      "rewards/forensics_strict_edge_reward": 0.038799822330474854,
      "rewards/forensics_tiou_reward": 0.36504198610782623,
      "step": 501
    },
    {
      "completion_length": 27.765625,
      "epoch": 1.5407066052227343,
      "grad_norm": 6.481152057647705,
      "kl": 0.3330078125,
      "learning_rate": 6.146153846153845e-07,
      "loss": 0.2183,
      "reward": 0.6925890147686005,
      "reward_std": 0.13536009564995766,
      "reward_std_all": 0.1286926530301571,
      "rewards-max/forensics_iou_reward": 0.9222125113010406,
      "rewards-max/forensics_strict_edge_reward": 0.2046772465109825,
      "rewards-max/forensics_tiou_reward": 0.9100228548049927,
      "rewards/forensics_iou_reward": 0.546537384390831,
      "rewards/forensics_strict_edge_reward": 0.012944599613547325,
      "rewards/forensics_tiou_reward": 0.4942961186170578,
      "step": 502
    },
    {
      "completion_length": 29.921875,
      "epoch": 1.543778801843318,
      "grad_norm": 4.977537631988525,
      "kl": 0.3447265625,
      "learning_rate": 6.138461538461538e-07,
      "loss": 0.0603,
      "reward": 0.7709044516086578,
      "reward_std": 0.03688027337193489,
      "reward_std_all": 0.03769378364086151,
      "rewards-max/forensics_iou_reward": 0.9565724432468414,
      "rewards-max/forensics_strict_edge_reward": 0.4426772743463516,
      "rewards-max/forensics_tiou_reward": 0.9387744069099426,
      "rewards/forensics_iou_reward": 0.6730614900588989,
      "rewards/forensics_strict_edge_reward": 0.04722192697227001,
      "rewards/forensics_tiou_reward": 0.628656268119812,
      "step": 503
    },
    {
      "completion_length": 28.3125,
      "epoch": 1.5468509984639018,
      "grad_norm": 5.3720011711120605,
      "kl": 0.390625,
      "learning_rate": 6.13076923076923e-07,
      "loss": 0.2133,
      "reward": 0.7005058228969574,
      "reward_std": 0.12488564103841782,
      "reward_std_all": 0.11773902177810669,
      "rewards-max/forensics_iou_reward": 0.9208754897117615,
      "rewards-max/forensics_strict_edge_reward": 0.43896108865737915,
      "rewards-max/forensics_tiou_reward": 0.903470903635025,
      "rewards/forensics_iou_reward": 0.5967674851417542,
      "rewards/forensics_strict_edge_reward": 0.023649179376661777,
      "rewards/forensics_tiou_reward": 0.5422756671905518,
      "step": 504
    },
    {
      "completion_length": 22.9375,
      "epoch": 1.5499231950844854,
      "grad_norm": 5.87509822845459,
      "kl": 0.3193359375,
      "learning_rate": 6.123076923076923e-07,
      "loss": 0.1038,
      "reward": 0.6992156505584717,
      "reward_std": 0.07240438275039196,
      "reward_std_all": 0.09037252888083458,
      "rewards-max/forensics_iou_reward": 0.9943144023418427,
      "rewards-max/forensics_strict_edge_reward": 0.9360267519950867,
      "rewards-max/forensics_tiou_reward": 0.9925777912139893,
      "rewards/forensics_iou_reward": 0.5575670599937439,
      "rewards/forensics_strict_edge_reward": 0.06468822062015533,
      "rewards/forensics_tiou_reward": 0.5060132145881653,
      "step": 505
    },
    {
      "completion_length": 27.953125,
      "epoch": 1.5529953917050692,
      "grad_norm": 7.164451599121094,
      "kl": 0.38671875,
      "learning_rate": 6.115384615384616e-07,
      "loss": 0.1152,
      "reward": 0.7203928530216217,
      "reward_std": 0.09441771730780602,
      "reward_std_all": 0.0928325206041336,
      "rewards-max/forensics_iou_reward": 0.9121475517749786,
      "rewards-max/forensics_strict_edge_reward": 0.5567733645439148,
      "rewards-max/forensics_tiou_reward": 0.8903803825378418,
      "rewards/forensics_iou_reward": 0.5743277370929718,
      "rewards/forensics_strict_edge_reward": 0.040998950600624084,
      "rewards/forensics_tiou_reward": 0.5297406315803528,
      "step": 506
    },
    {
      "completion_length": 28.140625,
      "epoch": 1.5560675883256527,
      "grad_norm": 6.079713821411133,
      "kl": 0.5166015625,
      "learning_rate": 6.107692307692307e-07,
      "loss": 0.1646,
      "reward": 0.6878632009029388,
      "reward_std": 0.14133137464523315,
      "reward_std_all": 0.13254162669181824,
      "rewards-max/forensics_iou_reward": 0.9700339436531067,
      "rewards-max/forensics_strict_edge_reward": 0.2596410159021616,
      "rewards-max/forensics_tiou_reward": 0.9557884335517883,
      "rewards/forensics_iou_reward": 0.5702784657478333,
      "rewards/forensics_strict_edge_reward": 0.02929805777966976,
      "rewards/forensics_tiou_reward": 0.5179702639579773,
      "step": 507
    },
    {
      "completion_length": 31.515625,
      "epoch": 1.5591397849462365,
      "grad_norm": 5.733565807342529,
      "kl": 0.314453125,
      "learning_rate": 6.1e-07,
      "loss": 0.0741,
      "reward": 0.7260502874851227,
      "reward_std": 0.09532849863171577,
      "reward_std_all": 0.09028683975338936,
      "rewards-max/forensics_iou_reward": 0.9432649612426758,
      "rewards-max/forensics_strict_edge_reward": 0.2560736611485481,
      "rewards-max/forensics_tiou_reward": 0.932841420173645,
      "rewards/forensics_iou_reward": 0.5666466653347015,
      "rewards/forensics_strict_edge_reward": 0.0302846971899271,
      "rewards/forensics_tiou_reward": 0.5224935710430145,
      "step": 508
    },
    {
      "completion_length": 24.4375,
      "epoch": 1.5622119815668203,
      "grad_norm": 7.273051738739014,
      "kl": 0.4013671875,
      "learning_rate": 6.092307692307692e-07,
      "loss": -0.0179,
      "reward": 0.7387804687023163,
      "reward_std": 0.06158989667892456,
      "reward_std_all": 0.06393428519368172,
      "rewards-max/forensics_iou_reward": 0.9683313965797424,
      "rewards-max/forensics_strict_edge_reward": 0.308086059987545,
      "rewards-max/forensics_tiou_reward": 0.949010968208313,
      "rewards/forensics_iou_reward": 0.6358799338340759,
      "rewards/forensics_strict_edge_reward": 0.028435428626835346,
      "rewards/forensics_tiou_reward": 0.5850743651390076,
      "step": 509
    },
    {
      "completion_length": 23.5,
      "epoch": 1.565284178187404,
      "grad_norm": 5.694919586181641,
      "kl": 0.3212890625,
      "learning_rate": 6.084615384615385e-07,
      "loss": 0.0825,
      "reward": 0.6872869729995728,
      "reward_std": 0.07698857970535755,
      "reward_std_all": 0.10487331822514534,
      "rewards-max/forensics_iou_reward": 0.9192808866500854,
      "rewards-max/forensics_strict_edge_reward": 0.46472586691379547,
      "rewards-max/forensics_tiou_reward": 0.904169350862503,
      "rewards/forensics_iou_reward": 0.5461429953575134,
      "rewards/forensics_strict_edge_reward": 0.04080696776509285,
      "rewards/forensics_tiou_reward": 0.49637527763843536,
      "step": 510
    },
    {
      "completion_length": 24.84375,
      "epoch": 1.5683563748079878,
      "grad_norm": 6.252741813659668,
      "kl": 0.3759765625,
      "learning_rate": 6.076923076923076e-07,
      "loss": 0.0217,
      "reward": 0.7696179449558258,
      "reward_std": 0.04418101720511913,
      "reward_std_all": 0.042739156633615494,
      "rewards-max/forensics_iou_reward": 0.9656719863414764,
      "rewards-max/forensics_strict_edge_reward": 0.5114051550626755,
      "rewards-max/forensics_tiou_reward": 0.9528596699237823,
      "rewards/forensics_iou_reward": 0.6269819140434265,
      "rewards/forensics_strict_edge_reward": 0.04724462702870369,
      "rewards/forensics_tiou_reward": 0.5675744116306305,
      "step": 511
    },
    {
      "completion_length": 23.390625,
      "epoch": 1.5714285714285714,
      "grad_norm": 5.074002265930176,
      "kl": 0.40234375,
      "learning_rate": 6.069230769230768e-07,
      "loss": 0.0179,
      "reward": 0.6768105030059814,
      "reward_std": 0.09923660475760698,
      "reward_std_all": 0.11664412170648575,
      "rewards-max/forensics_iou_reward": 0.9671550095081329,
      "rewards-max/forensics_strict_edge_reward": 0.36635348200798035,
      "rewards-max/forensics_tiou_reward": 0.9550677835941315,
      "rewards/forensics_iou_reward": 0.5555225014686584,
      "rewards/forensics_strict_edge_reward": 0.025777935050427914,
      "rewards/forensics_tiou_reward": 0.5082696825265884,
      "step": 512
    },
    {
      "completion_length": 26.4375,
      "epoch": 1.5745007680491552,
      "grad_norm": 6.831339359283447,
      "kl": 0.2890625,
      "learning_rate": 6.061538461538462e-07,
      "loss": 0.1588,
      "reward": 0.7050434052944183,
      "reward_std": 0.05957622453570366,
      "reward_std_all": 0.08081146329641342,
      "rewards-max/forensics_iou_reward": 0.8988061547279358,
      "rewards-max/forensics_strict_edge_reward": 0.24600800126791,
      "rewards-max/forensics_tiou_reward": 0.8786234855651855,
      "rewards/forensics_iou_reward": 0.5069581866264343,
      "rewards/forensics_strict_edge_reward": 0.02284041978418827,
      "rewards/forensics_tiou_reward": 0.4663676917552948,
      "step": 513
    },
    {
      "completion_length": 26.703125,
      "epoch": 1.5775729646697387,
      "grad_norm": 5.600632190704346,
      "kl": 0.3173828125,
      "learning_rate": 6.053846153846154e-07,
      "loss": -0.0348,
      "reward": 0.6816267967224121,
      "reward_std": 0.11595869436860085,
      "reward_std_all": 0.11701234430074692,
      "rewards-max/forensics_iou_reward": 0.9708406031131744,
      "rewards-max/forensics_strict_edge_reward": 0.48137304186820984,
      "rewards-max/forensics_tiou_reward": 0.9618867933750153,
      "rewards/forensics_iou_reward": 0.551824614405632,
      "rewards/forensics_strict_edge_reward": 0.03344344347715378,
      "rewards/forensics_tiou_reward": 0.5019850432872772,
      "step": 514
    },
    {
      "completion_length": 26.84375,
      "epoch": 1.5806451612903225,
      "grad_norm": 7.3572797775268555,
      "kl": 0.46484375,
      "learning_rate": 6.046153846153847e-07,
      "loss": 0.2196,
      "reward": 0.5864600539207458,
      "reward_std": 0.12136460095643997,
      "reward_std_all": 0.16624469310045242,
      "rewards-max/forensics_iou_reward": 0.9304259717464447,
      "rewards-max/forensics_strict_edge_reward": 0.149413836421445,
      "rewards-max/forensics_tiou_reward": 0.9179606437683105,
      "rewards/forensics_iou_reward": 0.44580136239528656,
      "rewards/forensics_strict_edge_reward": 0.009944707999238744,
      "rewards/forensics_tiou_reward": 0.3880927860736847,
      "step": 515
    },
    {
      "completion_length": 26.703125,
      "epoch": 1.5837173579109063,
      "grad_norm": 5.889714241027832,
      "kl": 0.3798828125,
      "learning_rate": 6.038461538461538e-07,
      "loss": 0.1479,
      "reward": 0.6607548594474792,
      "reward_std": 0.07465434074401855,
      "reward_std_all": 0.10567852109670639,
      "rewards-max/forensics_iou_reward": 0.9681463837623596,
      "rewards-max/forensics_strict_edge_reward": 0.33065591752529144,
      "rewards-max/forensics_tiou_reward": 0.9566411077976227,
      "rewards/forensics_iou_reward": 0.5390790402889252,
      "rewards/forensics_strict_edge_reward": 0.02678382396697998,
      "rewards/forensics_tiou_reward": 0.49143363535404205,
      "step": 516
    },
    {
      "completion_length": 24.546875,
      "epoch": 1.58678955453149,
      "grad_norm": 5.548431873321533,
      "kl": 0.3291015625,
      "learning_rate": 6.03076923076923e-07,
      "loss": 0.1086,
      "reward": 0.7074256241321564,
      "reward_std": 0.09202251210808754,
      "reward_std_all": 0.09320541843771935,
      "rewards-max/forensics_iou_reward": 0.9360355138778687,
      "rewards-max/forensics_strict_edge_reward": 0.2946265935897827,
      "rewards-max/forensics_tiou_reward": 0.9242870211601257,
      "rewards/forensics_iou_reward": 0.5186434388160706,
      "rewards/forensics_strict_edge_reward": 0.026510052382946014,
      "rewards/forensics_tiou_reward": 0.4662500470876694,
      "step": 517
    },
    {
      "completion_length": 26.96875,
      "epoch": 1.5898617511520738,
      "grad_norm": 6.942874908447266,
      "kl": 0.3310546875,
      "learning_rate": 6.023076923076923e-07,
      "loss": 0.1295,
      "reward": 0.5929481089115143,
      "reward_std": 0.15046587586402893,
      "reward_std_all": 0.17505471408367157,
      "rewards-max/forensics_iou_reward": 0.9807362258434296,
      "rewards-max/forensics_strict_edge_reward": 0.2412191480398178,
      "rewards-max/forensics_tiou_reward": 0.964973509311676,
      "rewards/forensics_iou_reward": 0.42505383491516113,
      "rewards/forensics_strict_edge_reward": 0.01789275323972106,
      "rewards/forensics_tiou_reward": 0.37892909348011017,
      "step": 518
    },
    {
      "completion_length": 23.53125,
      "epoch": 1.5929339477726574,
      "grad_norm": 6.346518516540527,
      "kl": 0.478515625,
      "learning_rate": 6.015384615384615e-07,
      "loss": 0.0921,
      "reward": 0.7394133806228638,
      "reward_std": 0.07453072443604469,
      "reward_std_all": 0.07199034281075001,
      "rewards-max/forensics_iou_reward": 0.960657924413681,
      "rewards-max/forensics_strict_edge_reward": 0.5931393504142761,
      "rewards-max/forensics_tiou_reward": 0.9541113078594208,
      "rewards/forensics_iou_reward": 0.5926640182733536,
      "rewards/forensics_strict_edge_reward": 0.0652324203401804,
      "rewards/forensics_tiou_reward": 0.5522612035274506,
      "step": 519
    },
    {
      "completion_length": 26.875,
      "epoch": 1.5960061443932412,
      "grad_norm": 6.301579475402832,
      "kl": 0.49609375,
      "learning_rate": 6.007692307692307e-07,
      "loss": -0.055,
      "reward": 0.6641726791858673,
      "reward_std": 0.08702556602656841,
      "reward_std_all": 0.11313091777265072,
      "rewards-max/forensics_iou_reward": 0.9357650279998779,
      "rewards-max/forensics_strict_edge_reward": 0.4230326861143112,
      "rewards-max/forensics_tiou_reward": 0.899374783039093,
      "rewards/forensics_iou_reward": 0.5133432745933533,
      "rewards/forensics_strict_edge_reward": 0.03159322403371334,
      "rewards/forensics_tiou_reward": 0.4589826315641403,
      "step": 520
    },
    {
      "completion_length": 28.0625,
      "epoch": 1.5990783410138247,
      "grad_norm": 4.759422779083252,
      "kl": 0.3955078125,
      "learning_rate": 6e-07,
      "loss": 0.2721,
      "reward": 0.6559609770774841,
      "reward_std": 0.13485419750213623,
      "reward_std_all": 0.1477423831820488,
      "rewards-max/forensics_iou_reward": 0.985516756772995,
      "rewards-max/forensics_strict_edge_reward": 0.4789964258670807,
      "rewards-max/forensics_tiou_reward": 0.9775703251361847,
      "rewards/forensics_iou_reward": 0.514865905046463,
      "rewards/forensics_strict_edge_reward": 0.04367386735975742,
      "rewards/forensics_tiou_reward": 0.45974327623844147,
      "step": 521
    },
    {
      "completion_length": 25.5,
      "epoch": 1.6021505376344085,
      "grad_norm": 6.362585544586182,
      "kl": 0.3525390625,
      "learning_rate": 5.992307692307692e-07,
      "loss": 0.0104,
      "reward": 0.6520935297012329,
      "reward_std": 0.07298726588487625,
      "reward_std_all": 0.10989371314644814,
      "rewards-max/forensics_iou_reward": 0.9036238491535187,
      "rewards-max/forensics_strict_edge_reward": 0.21644389256834984,
      "rewards-max/forensics_tiou_reward": 0.8644867837429047,
      "rewards/forensics_iou_reward": 0.49715785682201385,
      "rewards/forensics_strict_edge_reward": 0.020076482091099024,
      "rewards/forensics_tiou_reward": 0.43764403462409973,
      "step": 522
    },
    {
      "completion_length": 29.9375,
      "epoch": 1.6052227342549923,
      "grad_norm": 5.131672382354736,
      "kl": 0.3837890625,
      "learning_rate": 5.984615384615385e-07,
      "loss": 0.0445,
      "reward": 0.7218395471572876,
      "reward_std": 0.0709923803806305,
      "reward_std_all": 0.07695522718131542,
      "rewards-max/forensics_iou_reward": 0.963248610496521,
      "rewards-max/forensics_strict_edge_reward": 0.23372641578316689,
      "rewards-max/forensics_tiou_reward": 0.9492840766906738,
      "rewards/forensics_iou_reward": 0.5731652826070786,
      "rewards/forensics_strict_edge_reward": 0.016888048965483904,
      "rewards/forensics_tiou_reward": 0.5076210796833038,
      "step": 523
    },
    {
      "completion_length": 23.8125,
      "epoch": 1.608294930875576,
      "grad_norm": 5.263246059417725,
      "kl": 0.3349609375,
      "learning_rate": 5.976923076923077e-07,
      "loss": 0.1322,
      "reward": 0.766115665435791,
      "reward_std": 0.055428996216505766,
      "reward_std_all": 0.05177097860723734,
      "rewards-max/forensics_iou_reward": 0.9415404796600342,
      "rewards-max/forensics_strict_edge_reward": 0.5591417104005814,
      "rewards-max/forensics_tiou_reward": 0.9286306798458099,
      "rewards/forensics_iou_reward": 0.6369601488113403,
      "rewards/forensics_strict_edge_reward": 0.048313165083527565,
      "rewards/forensics_tiou_reward": 0.5746022462844849,
      "step": 524
    },
    {
      "completion_length": 25.34375,
      "epoch": 1.6113671274961598,
      "grad_norm": 6.682897090911865,
      "kl": 0.35546875,
      "learning_rate": 5.969230769230768e-07,
      "loss": 0.1455,
      "reward": 0.5592459738254547,
      "reward_std": 0.10644872672855854,
      "reward_std_all": 0.16804521530866623,
      "rewards-max/forensics_iou_reward": 0.8565465807914734,
      "rewards-max/forensics_strict_edge_reward": 0.07576400972902775,
      "rewards-max/forensics_tiou_reward": 0.823161393404007,
      "rewards/forensics_iou_reward": 0.3730347454547882,
      "rewards/forensics_strict_edge_reward": 0.004598553525283933,
      "rewards/forensics_tiou_reward": 0.3086520656943321,
      "step": 525
    },
    {
      "completion_length": 30.875,
      "epoch": 1.6144393241167436,
      "grad_norm": 4.671828269958496,
      "kl": 0.3759765625,
      "learning_rate": 5.961538461538461e-07,
      "loss": 0.1968,
      "reward": 0.7670530080795288,
      "reward_std": 0.04557513818144798,
      "reward_std_all": 0.043998527340590954,
      "rewards-max/forensics_iou_reward": 0.9493606984615326,
      "rewards-max/forensics_strict_edge_reward": 0.4302518963813782,
      "rewards-max/forensics_tiou_reward": 0.9342910051345825,
      "rewards/forensics_iou_reward": 0.6079569458961487,
      "rewards/forensics_strict_edge_reward": 0.05041656829416752,
      "rewards/forensics_tiou_reward": 0.5439977943897247,
      "step": 526
    },
    {
      "completion_length": 24.984375,
      "epoch": 1.6175115207373272,
      "grad_norm": 4.977200984954834,
      "kl": 0.37109375,
      "learning_rate": 5.953846153846154e-07,
      "loss": 0.0497,
      "reward": 0.7539638578891754,
      "reward_std": 0.0712474100291729,
      "reward_std_all": 0.06592851132154465,
      "rewards-max/forensics_iou_reward": 0.9868428707122803,
      "rewards-max/forensics_strict_edge_reward": 0.7952725887298584,
      "rewards-max/forensics_tiou_reward": 0.9844822585582733,
      "rewards/forensics_iou_reward": 0.6271571218967438,
      "rewards/forensics_strict_edge_reward": 0.07228368148207664,
      "rewards/forensics_tiou_reward": 0.5509681105613708,
      "step": 527
    },
    {
      "completion_length": 23.15625,
      "epoch": 1.620583717357911,
      "grad_norm": 5.343601703643799,
      "kl": 0.2998046875,
      "learning_rate": 5.946153846153847e-07,
      "loss": 0.085,
      "reward": 0.6397322416305542,
      "reward_std": 0.10407789796590805,
      "reward_std_all": 0.13677244633436203,
      "rewards-max/forensics_iou_reward": 0.9553165435791016,
      "rewards-max/forensics_strict_edge_reward": 0.5066853165626526,
      "rewards-max/forensics_tiou_reward": 0.9455433487892151,
      "rewards/forensics_iou_reward": 0.48854196071624756,
      "rewards/forensics_strict_edge_reward": 0.0314276535063982,
      "rewards/forensics_tiou_reward": 0.43285658955574036,
      "step": 528
    },
    {
      "completion_length": 25.296875,
      "epoch": 1.6236559139784945,
      "grad_norm": 5.536044597625732,
      "kl": 0.3740234375,
      "learning_rate": 5.938461538461538e-07,
      "loss": 0.105,
      "reward": 0.6992334723472595,
      "reward_std": 0.11713453754782677,
      "reward_std_all": 0.11676027625799179,
      "rewards-max/forensics_iou_reward": 0.9771181344985962,
      "rewards-max/forensics_strict_edge_reward": 0.5484400391578674,
      "rewards-max/forensics_tiou_reward": 0.969064474105835,
      "rewards/forensics_iou_reward": 0.5625126212835312,
      "rewards/forensics_strict_edge_reward": 0.06098152883350849,
      "rewards/forensics_tiou_reward": 0.5199841260910034,
      "step": 529
    },
    {
      "completion_length": 23.609375,
      "epoch": 1.6267281105990783,
      "grad_norm": 5.0781378746032715,
      "kl": 0.4765625,
      "learning_rate": 5.93076923076923e-07,
      "loss": 0.1647,
      "reward": 0.7348948121070862,
      "reward_std": 0.09835804253816605,
      "reward_std_all": 0.0903867855668068,
      "rewards-max/forensics_iou_reward": 0.9433927237987518,
      "rewards-max/forensics_strict_edge_reward": 0.3219582736492157,
      "rewards-max/forensics_tiou_reward": 0.927389532327652,
      "rewards/forensics_iou_reward": 0.5739831924438477,
      "rewards/forensics_strict_edge_reward": 0.02740615140646696,
      "rewards/forensics_tiou_reward": 0.4955136775970459,
      "step": 530
    },
    {
      "completion_length": 28.0,
      "epoch": 1.629800307219662,
      "grad_norm": 6.020941257476807,
      "kl": 0.4072265625,
      "learning_rate": 5.923076923076923e-07,
      "loss": 0.0515,
      "reward": 0.660510927438736,
      "reward_std": 0.07059885933995247,
      "reward_std_all": 0.10600362345576286,
      "rewards-max/forensics_iou_reward": 0.9249706268310547,
      "rewards-max/forensics_strict_edge_reward": 0.14964693412184715,
      "rewards-max/forensics_tiou_reward": 0.900126725435257,
      "rewards/forensics_iou_reward": 0.4993487596511841,
      "rewards/forensics_strict_edge_reward": 0.014900965616106987,
      "rewards/forensics_tiou_reward": 0.4389919191598892,
      "step": 531
    },
    {
      "completion_length": 29.859375,
      "epoch": 1.6328725038402458,
      "grad_norm": 5.347428798675537,
      "kl": 0.537109375,
      "learning_rate": 5.915384615384615e-07,
      "loss": 0.0831,
      "reward": 0.6852619349956512,
      "reward_std": 0.11366268247365952,
      "reward_std_all": 0.12396754324436188,
      "rewards-max/forensics_iou_reward": 0.92333984375,
      "rewards-max/forensics_strict_edge_reward": 0.41449280083179474,
      "rewards-max/forensics_tiou_reward": 0.9090839326381683,
      "rewards/forensics_iou_reward": 0.5465753078460693,
      "rewards/forensics_strict_edge_reward": 0.04360131546854973,
      "rewards/forensics_tiou_reward": 0.5069753229618073,
      "step": 532
    },
    {
      "completion_length": 26.9375,
      "epoch": 1.6359447004608296,
      "grad_norm": 7.156416416168213,
      "kl": 0.4072265625,
      "learning_rate": 5.907692307692307e-07,
      "loss": 0.1626,
      "reward": 0.7055999338626862,
      "reward_std": 0.08750888705253601,
      "reward_std_all": 0.09818154945969582,
      "rewards-max/forensics_iou_reward": 0.9747059047222137,
      "rewards-max/forensics_strict_edge_reward": 0.41893986612558365,
      "rewards-max/forensics_tiou_reward": 0.9644879102706909,
      "rewards/forensics_iou_reward": 0.5626793503761292,
      "rewards/forensics_strict_edge_reward": 0.04358363896608353,
      "rewards/forensics_tiou_reward": 0.5132338404655457,
      "step": 533
    },
    {
      "completion_length": 22.796875,
      "epoch": 1.6390168970814132,
      "grad_norm": 6.116817474365234,
      "kl": 0.3671875,
      "learning_rate": 5.9e-07,
      "loss": 0.0242,
      "reward": 0.7295878231525421,
      "reward_std": 0.06666452810168266,
      "reward_std_all": 0.06916601210832596,
      "rewards-max/forensics_iou_reward": 0.8618371784687042,
      "rewards-max/forensics_strict_edge_reward": 0.1672469601035118,
      "rewards-max/forensics_tiou_reward": 0.7771495580673218,
      "rewards/forensics_iou_reward": 0.5232333540916443,
      "rewards/forensics_strict_edge_reward": 0.015332126524299383,
      "rewards/forensics_tiou_reward": 0.4552913159132004,
      "step": 534
    },
    {
      "completion_length": 26.03125,
      "epoch": 1.642089093701997,
      "grad_norm": 6.571786880493164,
      "kl": 0.349609375,
      "learning_rate": 5.892307692307692e-07,
      "loss": 0.0449,
      "reward": 0.6981253325939178,
      "reward_std": 0.13830798491835594,
      "reward_std_all": 0.1278842445462942,
      "rewards-max/forensics_iou_reward": 0.9507855772972107,
      "rewards-max/forensics_strict_edge_reward": 0.20677004009485245,
      "rewards-max/forensics_tiou_reward": 0.9197841882705688,
      "rewards/forensics_iou_reward": 0.5149306654930115,
      "rewards/forensics_strict_edge_reward": 0.01618689065799117,
      "rewards/forensics_tiou_reward": 0.4598912447690964,
      "step": 535
    },
    {
      "completion_length": 26.671875,
      "epoch": 1.6451612903225805,
      "grad_norm": 4.423155307769775,
      "kl": 0.3759765625,
      "learning_rate": 5.884615384615385e-07,
      "loss": -0.0219,
      "reward": 0.7076018154621124,
      "reward_std": 0.08747665584087372,
      "reward_std_all": 0.09146945178508759,
      "rewards-max/forensics_iou_reward": 0.9476108551025391,
      "rewards-max/forensics_strict_edge_reward": 0.39651863276958466,
      "rewards-max/forensics_tiou_reward": 0.930021196603775,
      "rewards/forensics_iou_reward": 0.5343376100063324,
      "rewards/forensics_strict_edge_reward": 0.04072747752070427,
      "rewards/forensics_tiou_reward": 0.4920772612094879,
      "step": 536
    },
    {
      "completion_length": 24.625,
      "epoch": 1.6482334869431643,
      "grad_norm": 4.625014781951904,
      "kl": 0.42578125,
      "learning_rate": 5.876923076923077e-07,
      "loss": 0.1102,
      "reward": 0.6967154741287231,
      "reward_std": 0.08121757954359055,
      "reward_std_all": 0.09923043847084045,
      "rewards-max/forensics_iou_reward": 0.8504667282104492,
      "rewards-max/forensics_strict_edge_reward": 0.31069888174533844,
      "rewards-max/forensics_tiou_reward": 0.806421160697937,
      "rewards/forensics_iou_reward": 0.5685208737850189,
      "rewards/forensics_strict_edge_reward": 0.026365169323980808,
      "rewards/forensics_tiou_reward": 0.5209028422832489,
      "step": 537
    },
    {
      "completion_length": 27.75,
      "epoch": 1.651305683563748,
      "grad_norm": 5.47194766998291,
      "kl": 0.3828125,
      "learning_rate": 5.869230769230768e-07,
      "loss": 0.1175,
      "reward": 0.7028439044952393,
      "reward_std": 0.05941906198859215,
      "reward_std_all": 0.07961013168096542,
      "rewards-max/forensics_iou_reward": 0.9856953620910645,
      "rewards-max/forensics_strict_edge_reward": 0.30350157618522644,
      "rewards-max/forensics_tiou_reward": 0.9776774942874908,
      "rewards/forensics_iou_reward": 0.6059349179267883,
      "rewards/forensics_strict_edge_reward": 0.029971297830343246,
      "rewards/forensics_tiou_reward": 0.5518141388893127,
      "step": 538
    },
    {
      "completion_length": 31.15625,
      "epoch": 1.6543778801843319,
      "grad_norm": 4.214058876037598,
      "kl": 0.3837890625,
      "learning_rate": 5.861538461538461e-07,
      "loss": 0.039,
      "reward": 0.6582809388637543,
      "reward_std": 0.10495511442422867,
      "reward_std_all": 0.12183548510074615,
      "rewards-max/forensics_iou_reward": 0.9816461503505707,
      "rewards-max/forensics_strict_edge_reward": 0.5896764099597931,
      "rewards-max/forensics_tiou_reward": 0.9754330813884735,
      "rewards/forensics_iou_reward": 0.5090757310390472,
      "rewards/forensics_strict_edge_reward": 0.05696943774819374,
      "rewards/forensics_tiou_reward": 0.4617954343557358,
      "step": 539
    },
    {
      "completion_length": 28.953125,
      "epoch": 1.6574500768049156,
      "grad_norm": 5.678452491760254,
      "kl": 0.3671875,
      "learning_rate": 5.853846153846154e-07,
      "loss": 0.1673,
      "reward": 0.7222826480865479,
      "reward_std": 0.07992222905158997,
      "reward_std_all": 0.08164644613862038,
      "rewards-max/forensics_iou_reward": 0.9821810126304626,
      "rewards-max/forensics_strict_edge_reward": 0.2660123109817505,
      "rewards-max/forensics_tiou_reward": 0.9750307500362396,
      "rewards/forensics_iou_reward": 0.6180272996425629,
      "rewards/forensics_strict_edge_reward": 0.02156386524438858,
      "rewards/forensics_tiou_reward": 0.5543503761291504,
      "step": 540
    },
    {
      "completion_length": 24.109375,
      "epoch": 1.6605222734254992,
      "grad_norm": 4.773214817047119,
      "kl": 0.38671875,
      "learning_rate": 5.846153846153847e-07,
      "loss": 0.0491,
      "reward": 0.7110738754272461,
      "reward_std": 0.10395535081624985,
      "reward_std_all": 0.09977266564965248,
      "rewards-max/forensics_iou_reward": 0.9892315864562988,
      "rewards-max/forensics_strict_edge_reward": 0.6770831942558289,
      "rewards-max/forensics_tiou_reward": 0.9844617247581482,
      "rewards/forensics_iou_reward": 0.6269782781600952,
      "rewards/forensics_strict_edge_reward": 0.05862126871943474,
      "rewards/forensics_tiou_reward": 0.5767623782157898,
      "step": 541
    },
    {
      "completion_length": 26.0625,
      "epoch": 1.663594470046083,
      "grad_norm": 7.549363613128662,
      "kl": 0.65625,
      "learning_rate": 5.838461538461538e-07,
      "loss": 0.0733,
      "reward": 0.7231553196907043,
      "reward_std": 0.08813530392944813,
      "reward_std_all": 0.08544601872563362,
      "rewards-max/forensics_iou_reward": 0.9949820935726166,
      "rewards-max/forensics_strict_edge_reward": 0.661531001329422,
      "rewards-max/forensics_tiou_reward": 0.9934089779853821,
      "rewards/forensics_iou_reward": 0.6454022228717804,
      "rewards/forensics_strict_edge_reward": 0.08310303092002869,
      "rewards/forensics_tiou_reward": 0.5914998948574066,
      "step": 542
    },
    {
      "completion_length": 27.921875,
      "epoch": 1.6666666666666665,
      "grad_norm": 6.501559257507324,
      "kl": 0.333984375,
      "learning_rate": 5.83076923076923e-07,
      "loss": 0.1739,
      "reward": 0.6159343719482422,
      "reward_std": 0.09582941979169846,
      "reward_std_all": 0.1415456347167492,
      "rewards-max/forensics_iou_reward": 0.8865624964237213,
      "rewards-max/forensics_strict_edge_reward": 0.36777692288160324,
      "rewards-max/forensics_tiou_reward": 0.8455349206924438,
      "rewards/forensics_iou_reward": 0.45625558495521545,
      "rewards/forensics_strict_edge_reward": 0.026248806156218052,
      "rewards/forensics_tiou_reward": 0.40627536177635193,
      "step": 543
    },
    {
      "completion_length": 28.859375,
      "epoch": 1.6697388632872503,
      "grad_norm": 5.026043891906738,
      "kl": 0.591796875,
      "learning_rate": 5.823076923076923e-07,
      "loss": 0.149,
      "reward": 0.6656212210655212,
      "reward_std": 0.10189307853579521,
      "reward_std_all": 0.12167903035879135,
      "rewards-max/forensics_iou_reward": 0.9688229262828827,
      "rewards-max/forensics_strict_edge_reward": 0.5459358394145966,
      "rewards-max/forensics_tiou_reward": 0.9620601534843445,
      "rewards/forensics_iou_reward": 0.5156023949384689,
      "rewards/forensics_strict_edge_reward": 0.047630425542593,
      "rewards/forensics_tiou_reward": 0.4744754880666733,
      "step": 544
    },
    {
      "completion_length": 25.296875,
      "epoch": 1.672811059907834,
      "grad_norm": 6.221490859985352,
      "kl": 0.2978515625,
      "learning_rate": 5.815384615384615e-07,
      "loss": 0.0942,
      "reward": 0.7411509454250336,
      "reward_std": 0.07368669658899307,
      "reward_std_all": 0.0709633119404316,
      "rewards-max/forensics_iou_reward": 0.8417071402072906,
      "rewards-max/forensics_strict_edge_reward": 0.3084772452712059,
      "rewards-max/forensics_tiou_reward": 0.8058986067771912,
      "rewards/forensics_iou_reward": 0.5526996552944183,
      "rewards/forensics_strict_edge_reward": 0.017344163730740547,
      "rewards/forensics_tiou_reward": 0.4993338882923126,
      "step": 545
    },
    {
      "completion_length": 26.28125,
      "epoch": 1.6758832565284179,
      "grad_norm": 5.147511959075928,
      "kl": 0.43359375,
      "learning_rate": 5.807692307692307e-07,
      "loss": 0.1769,
      "reward": 0.7157659232616425,
      "reward_std": 0.08500936627388,
      "reward_std_all": 0.08955264836549759,
      "rewards-max/forensics_iou_reward": 0.9803657531738281,
      "rewards-max/forensics_strict_edge_reward": 0.5555379986763,
      "rewards-max/forensics_tiou_reward": 0.9741642773151398,
      "rewards/forensics_iou_reward": 0.5502553582191467,
      "rewards/forensics_strict_edge_reward": 0.0529110599309206,
      "rewards/forensics_tiou_reward": 0.4985974580049515,
      "step": 546
    },
    {
      "completion_length": 25.15625,
      "epoch": 1.6789554531490016,
      "grad_norm": 5.249520301818848,
      "kl": 0.359375,
      "learning_rate": 5.8e-07,
      "loss": 0.1167,
      "reward": 0.7720809876918793,
      "reward_std": 0.0401818361133337,
      "reward_std_all": 0.0389731265604496,
      "rewards-max/forensics_iou_reward": 0.957581639289856,
      "rewards-max/forensics_strict_edge_reward": 0.362475261092186,
      "rewards-max/forensics_tiou_reward": 0.9336786568164825,
      "rewards/forensics_iou_reward": 0.5895887613296509,
      "rewards/forensics_strict_edge_reward": 0.04581666179001331,
      "rewards/forensics_tiou_reward": 0.5242884755134583,
      "step": 547
    },
    {
      "completion_length": 24.359375,
      "epoch": 1.6820276497695854,
      "grad_norm": 4.985254287719727,
      "kl": 0.2802734375,
      "learning_rate": 5.792307692307692e-07,
      "loss": 0.1331,
      "reward": 0.6713274717330933,
      "reward_std": 0.13926564902067184,
      "reward_std_all": 0.13495244830846786,
      "rewards-max/forensics_iou_reward": 0.9172368943691254,
      "rewards-max/forensics_strict_edge_reward": 0.2662566900253296,
      "rewards-max/forensics_tiou_reward": 0.8975791037082672,
      "rewards/forensics_iou_reward": 0.48374874889850616,
      "rewards/forensics_strict_edge_reward": 0.021620734594762325,
      "rewards/forensics_tiou_reward": 0.42048977315425873,
      "step": 548
    },
    {
      "completion_length": 24.921875,
      "epoch": 1.685099846390169,
      "grad_norm": 4.618947982788086,
      "kl": 0.32421875,
      "learning_rate": 5.784615384615385e-07,
      "loss": -0.0198,
      "reward": 0.6940315067768097,
      "reward_std": 0.04214156046509743,
      "reward_std_all": 0.07833217829465866,
      "rewards-max/forensics_iou_reward": 0.9791587591171265,
      "rewards-max/forensics_strict_edge_reward": 0.37049978971481323,
      "rewards-max/forensics_tiou_reward": 0.9689604938030243,
      "rewards/forensics_iou_reward": 0.5836219191551208,
      "rewards/forensics_strict_edge_reward": 0.03844573348760605,
      "rewards/forensics_tiou_reward": 0.5273071527481079,
      "step": 549
    },
    {
      "completion_length": 26.453125,
      "epoch": 1.6881720430107527,
      "grad_norm": 5.611331462860107,
      "kl": 0.296875,
      "learning_rate": 5.776923076923077e-07,
      "loss": 0.0256,
      "reward": 0.753007173538208,
      "reward_std": 0.053962450474500656,
      "reward_std_all": 0.05484665185213089,
      "rewards-max/forensics_iou_reward": 0.9777877628803253,
      "rewards-max/forensics_strict_edge_reward": 0.6443193107843399,
      "rewards-max/forensics_tiou_reward": 0.97041255235672,
      "rewards/forensics_iou_reward": 0.5811669230461121,
      "rewards/forensics_strict_edge_reward": 0.06519387196749449,
      "rewards/forensics_tiou_reward": 0.5079202651977539,
      "step": 550
    },
    {
      "completion_length": 24.453125,
      "epoch": 1.6912442396313363,
      "grad_norm": 5.766304016113281,
      "kl": 0.2861328125,
      "learning_rate": 5.769230769230768e-07,
      "loss": 0.088,
      "reward": 0.6794623732566833,
      "reward_std": 0.10339191975072026,
      "reward_std_all": 0.11862220801413059,
      "rewards-max/forensics_iou_reward": 0.9650121331214905,
      "rewards-max/forensics_strict_edge_reward": 0.39879319816827774,
      "rewards-max/forensics_tiou_reward": 0.9506050050258636,
      "rewards/forensics_iou_reward": 0.5602190345525742,
      "rewards/forensics_strict_edge_reward": 0.05204861983656883,
      "rewards/forensics_tiou_reward": 0.5021954774856567,
      "step": 551
    },
    {
      "completion_length": 25.265625,
      "epoch": 1.69431643625192,
      "grad_norm": 5.842685222625732,
      "kl": 0.328125,
      "learning_rate": 5.761538461538461e-07,
      "loss": 0.088,
      "reward": 0.7514115273952484,
      "reward_std": 0.07594096660614014,
      "reward_std_all": 0.07015008106827736,
      "rewards-max/forensics_iou_reward": 0.9727029204368591,
      "rewards-max/forensics_strict_edge_reward": 0.40373896062374115,
      "rewards-max/forensics_tiou_reward": 0.9616412222385406,
      "rewards/forensics_iou_reward": 0.6344605982303619,
      "rewards/forensics_strict_edge_reward": 0.025127751752734184,
      "rewards/forensics_tiou_reward": 0.5765671730041504,
      "step": 552
    },
    {
      "completion_length": 24.65625,
      "epoch": 1.6973886328725039,
      "grad_norm": 5.497591972351074,
      "kl": 0.2880859375,
      "learning_rate": 5.753846153846153e-07,
      "loss": 0.0109,
      "reward": 0.6655690371990204,
      "reward_std": 0.10411987453699112,
      "reward_std_all": 0.11949627101421356,
      "rewards-max/forensics_iou_reward": 0.9306897819042206,
      "rewards-max/forensics_strict_edge_reward": 0.37141383811831474,
      "rewards-max/forensics_tiou_reward": 0.9152293503284454,
      "rewards/forensics_iou_reward": 0.48105764389038086,
      "rewards/forensics_strict_edge_reward": 0.02195964171551168,
      "rewards/forensics_tiou_reward": 0.4131942689418793,
      "step": 553
    },
    {
      "completion_length": 28.375,
      "epoch": 1.7004608294930876,
      "grad_norm": 6.493183135986328,
      "kl": 0.25244140625,
      "learning_rate": 5.746153846153847e-07,
      "loss": 0.0756,
      "reward": 0.7306863963603973,
      "reward_std": 0.07068098708987236,
      "reward_std_all": 0.07207122817635536,
      "rewards-max/forensics_iou_reward": 0.9210218489170074,
      "rewards-max/forensics_strict_edge_reward": 0.31817932426929474,
      "rewards-max/forensics_tiou_reward": 0.9017215669155121,
      "rewards/forensics_iou_reward": 0.5743555128574371,
      "rewards/forensics_strict_edge_reward": 0.03393779322504997,
      "rewards/forensics_tiou_reward": 0.5185331851243973,
      "step": 554
    },
    {
      "completion_length": 27.625,
      "epoch": 1.7035330261136714,
      "grad_norm": 5.102501392364502,
      "kl": 0.26123046875,
      "learning_rate": 5.738461538461539e-07,
      "loss": 0.1292,
      "reward": 0.7200193107128143,
      "reward_std": 0.10276253521442413,
      "reward_std_all": 0.09772966429591179,
      "rewards-max/forensics_iou_reward": 0.8771211206912994,
      "rewards-max/forensics_strict_edge_reward": 0.23115862533450127,
      "rewards-max/forensics_tiou_reward": 0.8582819998264313,
      "rewards/forensics_iou_reward": 0.5527332723140717,
      "rewards/forensics_strict_edge_reward": 0.03462729882448912,
      "rewards/forensics_tiou_reward": 0.48961421847343445,
      "step": 555
    },
    {
      "completion_length": 24.34375,
      "epoch": 1.706605222734255,
      "grad_norm": 6.44612979888916,
      "kl": 0.357421875,
      "learning_rate": 5.73076923076923e-07,
      "loss": 0.0075,
      "reward": 0.672197699546814,
      "reward_std": 0.13709725812077522,
      "reward_std_all": 0.13391730561852455,
      "rewards-max/forensics_iou_reward": 0.9821318686008453,
      "rewards-max/forensics_strict_edge_reward": 0.5067439526319504,
      "rewards-max/forensics_tiou_reward": 0.976553350687027,
      "rewards/forensics_iou_reward": 0.5264998376369476,
      "rewards/forensics_strict_edge_reward": 0.03240285627543926,
      "rewards/forensics_tiou_reward": 0.47781604528427124,
      "step": 556
    },
    {
      "completion_length": 24.75,
      "epoch": 1.7096774193548387,
      "grad_norm": 6.1855645179748535,
      "kl": 0.349609375,
      "learning_rate": 5.723076923076923e-07,
      "loss": 0.0311,
      "reward": 0.6774688065052032,
      "reward_std": 0.10651480406522751,
      "reward_std_all": 0.11128789559006691,
      "rewards-max/forensics_iou_reward": 0.9589131474494934,
      "rewards-max/forensics_strict_edge_reward": 0.37661072611808777,
      "rewards-max/forensics_tiou_reward": 0.9424150288105011,
      "rewards/forensics_iou_reward": 0.5402716994285583,
      "rewards/forensics_strict_edge_reward": 0.04119240678846836,
      "rewards/forensics_tiou_reward": 0.48304447531700134,
      "step": 557
    },
    {
      "completion_length": 22.875,
      "epoch": 1.7127496159754223,
      "grad_norm": 5.252291679382324,
      "kl": 0.3310546875,
      "learning_rate": 5.715384615384615e-07,
      "loss": 0.08,
      "reward": 0.7130309343338013,
      "reward_std": 0.09501217305660248,
      "reward_std_all": 0.0948457345366478,
      "rewards-max/forensics_iou_reward": 0.9526783227920532,
      "rewards-max/forensics_strict_edge_reward": 0.41257937997579575,
      "rewards-max/forensics_tiou_reward": 0.942396879196167,
      "rewards/forensics_iou_reward": 0.5723592638969421,
      "rewards/forensics_strict_edge_reward": 0.03392605250701308,
      "rewards/forensics_tiou_reward": 0.5172817260026932,
      "step": 558
    },
    {
      "completion_length": 26.515625,
      "epoch": 1.715821812596006,
      "grad_norm": 5.214343070983887,
      "kl": 0.630859375,
      "learning_rate": 5.707692307692307e-07,
      "loss": 0.0413,
      "reward": 0.7317701280117035,
      "reward_std": 0.09129847958683968,
      "reward_std_all": 0.08584766834974289,
      "rewards-max/forensics_iou_reward": 0.9778204560279846,
      "rewards-max/forensics_strict_edge_reward": 0.19211883191019297,
      "rewards-max/forensics_tiou_reward": 0.9623851180076599,
      "rewards/forensics_iou_reward": 0.5993037223815918,
      "rewards/forensics_strict_edge_reward": 0.01644785888493061,
      "rewards/forensics_tiou_reward": 0.5465463399887085,
      "step": 559
    },
    {
      "completion_length": 30.59375,
      "epoch": 1.7188940092165899,
      "grad_norm": 4.563684940338135,
      "kl": 0.3271484375,
      "learning_rate": 5.699999999999999e-07,
      "loss": -0.0406,
      "reward": 0.6971209943294525,
      "reward_std": 0.06347452802583575,
      "reward_std_all": 0.08680659905076027,
      "rewards-max/forensics_iou_reward": 0.8403326869010925,
      "rewards-max/forensics_strict_edge_reward": 0.3361583650112152,
      "rewards-max/forensics_tiou_reward": 0.8072880506515503,
      "rewards/forensics_iou_reward": 0.5139483958482742,
      "rewards/forensics_strict_edge_reward": 0.02773583075031638,
      "rewards/forensics_tiou_reward": 0.4704973101615906,
      "step": 560
    },
    {
      "completion_length": 24.96875,
      "epoch": 1.7219662058371736,
      "grad_norm": 6.275681018829346,
      "kl": 0.3193359375,
      "learning_rate": 5.692307692307692e-07,
      "loss": 0.1604,
      "reward": 0.6756048798561096,
      "reward_std": 0.14422497153282166,
      "reward_std_all": 0.13897130638360977,
      "rewards-max/forensics_iou_reward": 0.9305911064147949,
      "rewards-max/forensics_strict_edge_reward": 0.16685901209712029,
      "rewards-max/forensics_tiou_reward": 0.910474568605423,
      "rewards/forensics_iou_reward": 0.5408107340335846,
      "rewards/forensics_strict_edge_reward": 0.021868595387786627,
      "rewards/forensics_tiou_reward": 0.47871866822242737,
      "step": 561
    },
    {
      "completion_length": 26.109375,
      "epoch": 1.7250384024577574,
      "grad_norm": 5.1843461990356445,
      "kl": 0.3671875,
      "learning_rate": 5.684615384615385e-07,
      "loss": 0.1018,
      "reward": 0.6561224460601807,
      "reward_std": 0.08953625708818436,
      "reward_std_all": 0.12578636966645718,
      "rewards-max/forensics_iou_reward": 0.9739913642406464,
      "rewards-max/forensics_strict_edge_reward": 0.3721938282251358,
      "rewards-max/forensics_tiou_reward": 0.9664513170719147,
      "rewards/forensics_iou_reward": 0.5398204773664474,
      "rewards/forensics_strict_edge_reward": 0.04663436487317085,
      "rewards/forensics_tiou_reward": 0.4988595098257065,
      "step": 562
    },
    {
      "completion_length": 27.953125,
      "epoch": 1.728110599078341,
      "grad_norm": 6.032698631286621,
      "kl": 0.29052734375,
      "learning_rate": 5.676923076923077e-07,
      "loss": 0.1543,
      "reward": 0.656226396560669,
      "reward_std": 0.08855139464139938,
      "reward_std_all": 0.12132674735039473,
      "rewards-max/forensics_iou_reward": 0.941101461648941,
      "rewards-max/forensics_strict_edge_reward": 0.2714177258312702,
      "rewards-max/forensics_tiou_reward": 0.9220395684242249,
      "rewards/forensics_iou_reward": 0.4928453266620636,
      "rewards/forensics_strict_edge_reward": 0.027999382466077805,
      "rewards/forensics_tiou_reward": 0.4423779547214508,
      "step": 563
    },
    {
      "completion_length": 27.75,
      "epoch": 1.7311827956989247,
      "grad_norm": 7.2532148361206055,
      "kl": 0.3603515625,
      "learning_rate": 5.669230769230769e-07,
      "loss": 0.1609,
      "reward": 0.6401008665561676,
      "reward_std": 0.11823871731758118,
      "reward_std_all": 0.14010706543922424,
      "rewards-max/forensics_iou_reward": 0.9155379235744476,
      "rewards-max/forensics_strict_edge_reward": 0.21561894053593278,
      "rewards-max/forensics_tiou_reward": 0.8976519107818604,
      "rewards/forensics_iou_reward": 0.4865521788597107,
      "rewards/forensics_strict_edge_reward": 0.01610192278167233,
      "rewards/forensics_tiou_reward": 0.4367073178291321,
      "step": 564
    },
    {
      "completion_length": 27.40625,
      "epoch": 1.7342549923195083,
      "grad_norm": 5.5415496826171875,
      "kl": 0.4638671875,
      "learning_rate": 5.661538461538461e-07,
      "loss": 0.1005,
      "reward": 0.7356618344783783,
      "reward_std": 0.07202977687120438,
      "reward_std_all": 0.0707891397178173,
      "rewards-max/forensics_iou_reward": 0.9991191625595093,
      "rewards-max/forensics_strict_edge_reward": 0.9211947023868561,
      "rewards-max/forensics_tiou_reward": 0.9988203644752502,
      "rewards/forensics_iou_reward": 0.6207605600357056,
      "rewards/forensics_strict_edge_reward": 0.06610672548413277,
      "rewards/forensics_tiou_reward": 0.5510446429252625,
      "step": 565
    },
    {
      "completion_length": 24.875,
      "epoch": 1.737327188940092,
      "grad_norm": 7.123325347900391,
      "kl": 0.3349609375,
      "learning_rate": 5.653846153846153e-07,
      "loss": 0.1427,
      "reward": 0.644225537776947,
      "reward_std": 0.09797625988721848,
      "reward_std_all": 0.1348518505692482,
      "rewards-max/forensics_iou_reward": 0.9808917343616486,
      "rewards-max/forensics_strict_edge_reward": 0.7744058072566986,
      "rewards-max/forensics_tiou_reward": 0.975136399269104,
      "rewards/forensics_iou_reward": 0.5249466598033905,
      "rewards/forensics_strict_edge_reward": 0.046523382887244225,
      "rewards/forensics_tiou_reward": 0.48453810811042786,
      "step": 566
    },
    {
      "completion_length": 24.875,
      "epoch": 1.7403993855606759,
      "grad_norm": 7.0809431076049805,
      "kl": 0.30224609375,
      "learning_rate": 5.646153846153847e-07,
      "loss": 0.0329,
      "reward": 0.7282427549362183,
      "reward_std": 0.03639743383973837,
      "reward_std_all": 0.055082542821764946,
      "rewards-max/forensics_iou_reward": 0.9662162065505981,
      "rewards-max/forensics_strict_edge_reward": 0.5284441187977791,
      "rewards-max/forensics_tiou_reward": 0.9518356919288635,
      "rewards/forensics_iou_reward": 0.6007470190525055,
      "rewards/forensics_strict_edge_reward": 0.029348970390856266,
      "rewards/forensics_tiou_reward": 0.5365546345710754,
      "step": 567
    },
    {
      "completion_length": 27.9375,
      "epoch": 1.7434715821812596,
      "grad_norm": 5.326152324676514,
      "kl": 0.3369140625,
      "learning_rate": 5.638461538461539e-07,
      "loss": 0.1468,
      "reward": 0.7603586316108704,
      "reward_std": 0.06803673133254051,
      "reward_std_all": 0.06237940676510334,
      "rewards-max/forensics_iou_reward": 0.8505973517894745,
      "rewards-max/forensics_strict_edge_reward": 0.04350937623530626,
      "rewards-max/forensics_tiou_reward": 0.7807875275611877,
      "rewards/forensics_iou_reward": 0.5905404388904572,
      "rewards/forensics_strict_edge_reward": 0.006075908429920673,
      "rewards/forensics_tiou_reward": 0.5298052728176117,
      "step": 568
    },
    {
      "completion_length": 29.09375,
      "epoch": 1.7465437788018434,
      "grad_norm": 5.931325912475586,
      "kl": 0.341796875,
      "learning_rate": 5.63076923076923e-07,
      "loss": 0.0326,
      "reward": 0.6877670288085938,
      "reward_std": 0.06750488094985485,
      "reward_std_all": 0.08788532298058271,
      "rewards-max/forensics_iou_reward": 0.9122877418994904,
      "rewards-max/forensics_strict_edge_reward": 0.030095928348600864,
      "rewards-max/forensics_tiou_reward": 0.869563490152359,
      "rewards/forensics_iou_reward": 0.5364712476730347,
      "rewards/forensics_strict_edge_reward": 0.0032958497758954763,
      "rewards/forensics_tiou_reward": 0.477867066860199,
      "step": 569
    },
    {
      "completion_length": 28.59375,
      "epoch": 1.7496159754224272,
      "grad_norm": 4.908440589904785,
      "kl": 0.3828125,
      "learning_rate": 5.623076923076923e-07,
      "loss": 0.0161,
      "reward": 0.7382054626941681,
      "reward_std": 0.07602469623088837,
      "reward_std_all": 0.07339576631784439,
      "rewards-max/forensics_iou_reward": 0.986589640378952,
      "rewards-max/forensics_strict_edge_reward": 0.6886506080627441,
      "rewards-max/forensics_tiou_reward": 0.9817180037498474,
      "rewards/forensics_iou_reward": 0.6572973728179932,
      "rewards/forensics_strict_edge_reward": 0.09431571513414383,
      "rewards/forensics_tiou_reward": 0.6047417521476746,
      "step": 570
    },
    {
      "completion_length": 28.8125,
      "epoch": 1.7526881720430108,
      "grad_norm": 5.43999719619751,
      "kl": 0.4609375,
      "learning_rate": 5.615384615384615e-07,
      "loss": -0.0048,
      "reward": 0.7659429013729095,
      "reward_std": 0.050572301930515096,
      "reward_std_all": 0.048579185269773006,
      "rewards-max/forensics_iou_reward": 0.9375995993614197,
      "rewards-max/forensics_strict_edge_reward": 0.26729515567421913,
      "rewards-max/forensics_tiou_reward": 0.9128367304801941,
      "rewards/forensics_iou_reward": 0.6192480623722076,
      "rewards/forensics_strict_edge_reward": 0.03034716099500656,
      "rewards/forensics_tiou_reward": 0.5615475177764893,
      "step": 571
    },
    {
      "completion_length": 28.578125,
      "epoch": 1.7557603686635943,
      "grad_norm": 6.198525905609131,
      "kl": 0.3759765625,
      "learning_rate": 5.607692307692308e-07,
      "loss": 0.0719,
      "reward": 0.6041449904441833,
      "reward_std": 0.07993212714791298,
      "reward_std_all": 0.14155128970742226,
      "rewards-max/forensics_iou_reward": 0.8813965022563934,
      "rewards-max/forensics_strict_edge_reward": 0.20604118704795837,
      "rewards-max/forensics_tiou_reward": 0.8584607839584351,
      "rewards/forensics_iou_reward": 0.47601318359375,
      "rewards/forensics_strict_edge_reward": 0.015051278285682201,
      "rewards/forensics_tiou_reward": 0.4286522716283798,
      "step": 572
    },
    {
      "completion_length": 29.09375,
      "epoch": 1.758832565284178,
      "grad_norm": 4.586952209472656,
      "kl": 0.392578125,
      "learning_rate": 5.6e-07,
      "loss": 0.0476,
      "reward": 0.7079582512378693,
      "reward_std": 0.09051384404301643,
      "reward_std_all": 0.09324020892381668,
      "rewards-max/forensics_iou_reward": 0.9722631275653839,
      "rewards-max/forensics_strict_edge_reward": 0.4102914482355118,
      "rewards-max/forensics_tiou_reward": 0.9641562402248383,
      "rewards/forensics_iou_reward": 0.6153424978256226,
      "rewards/forensics_strict_edge_reward": 0.036282723769545555,
      "rewards/forensics_tiou_reward": 0.5679584741592407,
      "step": 573
    },
    {
      "completion_length": 24.953125,
      "epoch": 1.7619047619047619,
      "grad_norm": 5.039916038513184,
      "kl": 0.4580078125,
      "learning_rate": 5.592307692307691e-07,
      "loss": 0.1168,
      "reward": 0.6848735213279724,
      "reward_std": 0.09278926253318787,
      "reward_std_all": 0.10512648895382881,
      "rewards-max/forensics_iou_reward": 0.929560124874115,
      "rewards-max/forensics_strict_edge_reward": 0.6459447741508484,
      "rewards-max/forensics_tiou_reward": 0.916029155254364,
      "rewards/forensics_iou_reward": 0.5360775291919708,
      "rewards/forensics_strict_edge_reward": 0.04524325765669346,
      "rewards/forensics_tiou_reward": 0.48146240413188934,
      "step": 574
    },
    {
      "completion_length": 26.984375,
      "epoch": 1.7649769585253456,
      "grad_norm": 4.837973117828369,
      "kl": 0.34375,
      "learning_rate": 5.584615384615385e-07,
      "loss": 0.0329,
      "reward": 0.7234191596508026,
      "reward_std": 0.05731804668903351,
      "reward_std_all": 0.06708662584424019,
      "rewards-max/forensics_iou_reward": 0.9868998527526855,
      "rewards-max/forensics_strict_edge_reward": 0.4188033491373062,
      "rewards-max/forensics_tiou_reward": 0.9803498387336731,
      "rewards/forensics_iou_reward": 0.571753591299057,
      "rewards/forensics_strict_edge_reward": 0.042606028728187084,
      "rewards/forensics_tiou_reward": 0.5183884799480438,
      "step": 575
    },
    {
      "completion_length": 25.84375,
      "epoch": 1.7680491551459294,
      "grad_norm": 4.862114429473877,
      "kl": 0.58203125,
      "learning_rate": 5.576923076923077e-07,
      "loss": 0.1822,
      "reward": 0.6626302897930145,
      "reward_std": 0.12227138131856918,
      "reward_std_all": 0.13641687855124474,
      "rewards-max/forensics_iou_reward": 0.9633654356002808,
      "rewards-max/forensics_strict_edge_reward": 0.2694481462240219,
      "rewards-max/forensics_tiou_reward": 0.9526698589324951,
      "rewards/forensics_iou_reward": 0.5324433445930481,
      "rewards/forensics_strict_edge_reward": 0.033739715814590454,
      "rewards/forensics_tiou_reward": 0.48907333612442017,
      "step": 576
    },
    {
      "completion_length": 24.96875,
      "epoch": 1.7711213517665132,
      "grad_norm": 6.1510539054870605,
      "kl": 0.3466796875,
      "learning_rate": 5.569230769230769e-07,
      "loss": 0.1007,
      "reward": 0.6529504656791687,
      "reward_std": 0.11043452844023705,
      "reward_std_all": 0.13388557359576225,
      "rewards-max/forensics_iou_reward": 0.979598194360733,
      "rewards-max/forensics_strict_edge_reward": 0.3732265755534172,
      "rewards-max/forensics_tiou_reward": 0.9727855026721954,
      "rewards/forensics_iou_reward": 0.5347002893686295,
      "rewards/forensics_strict_edge_reward": 0.028782925568521023,
      "rewards/forensics_tiou_reward": 0.485905259847641,
      "step": 577
    },
    {
      "completion_length": 26.046875,
      "epoch": 1.7741935483870968,
      "grad_norm": 5.856914520263672,
      "kl": 0.40625,
      "learning_rate": 5.561538461538461e-07,
      "loss": 0.0483,
      "reward": 0.6959594786167145,
      "reward_std": 0.09614831209182739,
      "reward_std_all": 0.10974935069680214,
      "rewards-max/forensics_iou_reward": 0.9680876731872559,
      "rewards-max/forensics_strict_edge_reward": 0.4231441020965576,
      "rewards-max/forensics_tiou_reward": 0.9567058980464935,
      "rewards/forensics_iou_reward": 0.5469853281974792,
      "rewards/forensics_strict_edge_reward": 0.05150532349944115,
      "rewards/forensics_tiou_reward": 0.4951717406511307,
      "step": 578
    },
    {
      "completion_length": 22.453125,
      "epoch": 1.7772657450076805,
      "grad_norm": 7.008564472198486,
      "kl": 0.3427734375,
      "learning_rate": 5.553846153846153e-07,
      "loss": 0.0937,
      "reward": 0.7108333110809326,
      "reward_std": 0.11776075512170792,
      "reward_std_all": 0.11187265813350677,
      "rewards-max/forensics_iou_reward": 0.9013758897781372,
      "rewards-max/forensics_strict_edge_reward": 0.11401275405660272,
      "rewards-max/forensics_tiou_reward": 0.8543151915073395,
      "rewards/forensics_iou_reward": 0.51541967689991,
      "rewards/forensics_strict_edge_reward": 0.011691737221553922,
      "rewards/forensics_tiou_reward": 0.4496423602104187,
      "step": 579
    },
    {
      "completion_length": 29.90625,
      "epoch": 1.780337941628264,
      "grad_norm": 4.493292808532715,
      "kl": 0.3369140625,
      "learning_rate": 5.546153846153846e-07,
      "loss": 0.3046,
      "reward": 0.6685271859169006,
      "reward_std": 0.14278292655944824,
      "reward_std_all": 0.1404452994465828,
      "rewards-max/forensics_iou_reward": 0.9959930777549744,
      "rewards-max/forensics_strict_edge_reward": 0.7528755068778992,
      "rewards-max/forensics_tiou_reward": 0.9946257770061493,
      "rewards/forensics_iou_reward": 0.4863555580377579,
      "rewards/forensics_strict_edge_reward": 0.06320931948721409,
      "rewards/forensics_tiou_reward": 0.44277316331863403,
      "step": 580
    },
    {
      "completion_length": 22.46875,
      "epoch": 1.7834101382488479,
      "grad_norm": 8.422765731811523,
      "kl": 0.4560546875,
      "learning_rate": 5.538461538461539e-07,
      "loss": 0.0479,
      "reward": 0.7410516738891602,
      "reward_std": 0.050849923864006996,
      "reward_std_all": 0.055352842435240746,
      "rewards-max/forensics_iou_reward": 0.9900985360145569,
      "rewards-max/forensics_strict_edge_reward": 0.3720445856451988,
      "rewards-max/forensics_tiou_reward": 0.984283059835434,
      "rewards/forensics_iou_reward": 0.6072263419628143,
      "rewards/forensics_strict_edge_reward": 0.03593978937715292,
      "rewards/forensics_tiou_reward": 0.541249543428421,
      "step": 581
    },
    {
      "completion_length": 26.25,
      "epoch": 1.7864823348694316,
      "grad_norm": 6.503252029418945,
      "kl": 0.490234375,
      "learning_rate": 5.53076923076923e-07,
      "loss": 0.1422,
      "reward": 0.659517765045166,
      "reward_std": 0.12628880888223648,
      "reward_std_all": 0.13543836399912834,
      "rewards-max/forensics_iou_reward": 0.9499658048152924,
      "rewards-max/forensics_strict_edge_reward": 0.38654085993766785,
      "rewards-max/forensics_tiou_reward": 0.9428023099899292,
      "rewards/forensics_iou_reward": 0.520160436630249,
      "rewards/forensics_strict_edge_reward": 0.02579207858070731,
      "rewards/forensics_tiou_reward": 0.4747369885444641,
      "step": 582
    },
    {
      "completion_length": 24.703125,
      "epoch": 1.7895545314900154,
      "grad_norm": 5.622392654418945,
      "kl": 0.388671875,
      "learning_rate": 5.523076923076923e-07,
      "loss": 0.1902,
      "reward": 0.6651831865310669,
      "reward_std": 0.10927269235253334,
      "reward_std_all": 0.12276102602481842,
      "rewards-max/forensics_iou_reward": 0.915248841047287,
      "rewards-max/forensics_strict_edge_reward": 0.2905754968523979,
      "rewards-max/forensics_tiou_reward": 0.8996697664260864,
      "rewards/forensics_iou_reward": 0.48706233501434326,
      "rewards/forensics_strict_edge_reward": 0.02059646788984537,
      "rewards/forensics_tiou_reward": 0.439334511756897,
      "step": 583
    },
    {
      "completion_length": 27.515625,
      "epoch": 1.7926267281105992,
      "grad_norm": 5.626106262207031,
      "kl": 0.384765625,
      "learning_rate": 5.515384615384615e-07,
      "loss": 0.1458,
      "reward": 0.7070770859718323,
      "reward_std": 0.09617743641138077,
      "reward_std_all": 0.09547964110970497,
      "rewards-max/forensics_iou_reward": 0.967610627412796,
      "rewards-max/forensics_strict_edge_reward": 0.17399121075868607,
      "rewards-max/forensics_tiou_reward": 0.9529969096183777,
      "rewards/forensics_iou_reward": 0.5661157071590424,
      "rewards/forensics_strict_edge_reward": 0.01658596657216549,
      "rewards/forensics_tiou_reward": 0.506911426782608,
      "step": 584
    },
    {
      "completion_length": 24.09375,
      "epoch": 1.7956989247311828,
      "grad_norm": 5.678775787353516,
      "kl": 0.30078125,
      "learning_rate": 5.507692307692308e-07,
      "loss": -0.0166,
      "reward": 0.6438083350658417,
      "reward_std": 0.07673193514347076,
      "reward_std_all": 0.11464791372418404,
      "rewards-max/forensics_iou_reward": 0.9465516209602356,
      "rewards-max/forensics_strict_edge_reward": 0.39750565588474274,
      "rewards-max/forensics_tiou_reward": 0.9289960861206055,
      "rewards/forensics_iou_reward": 0.4823557883501053,
      "rewards/forensics_strict_edge_reward": 0.024858348071575165,
      "rewards/forensics_tiou_reward": 0.4219762235879898,
      "step": 585
    },
    {
      "completion_length": 25.71875,
      "epoch": 1.7987711213517665,
      "grad_norm": 5.780706405639648,
      "kl": 0.3857421875,
      "learning_rate": 5.5e-07,
      "loss": 0.0728,
      "reward": 0.6869577169418335,
      "reward_std": 0.08277747221291065,
      "reward_std_all": 0.10416175425052643,
      "rewards-max/forensics_iou_reward": 0.9806890189647675,
      "rewards-max/forensics_strict_edge_reward": 0.3361583650112152,
      "rewards-max/forensics_tiou_reward": 0.9741076529026031,
      "rewards/forensics_iou_reward": 0.5541692078113556,
      "rewards/forensics_strict_edge_reward": 0.03969608061015606,
      "rewards/forensics_tiou_reward": 0.5076305270195007,
      "step": 586
    },
    {
      "completion_length": 23.671875,
      "epoch": 1.80184331797235,
      "grad_norm": 4.811446666717529,
      "kl": 0.40234375,
      "learning_rate": 5.492307692307691e-07,
      "loss": 0.0967,
      "reward": 0.6645070910453796,
      "reward_std": 0.08006803691387177,
      "reward_std_all": 0.11301304399967194,
      "rewards-max/forensics_iou_reward": 0.9893176555633545,
      "rewards-max/forensics_strict_edge_reward": 0.6965087056159973,
      "rewards-max/forensics_tiou_reward": 0.9853450059890747,
      "rewards/forensics_iou_reward": 0.5453099012374878,
      "rewards/forensics_strict_edge_reward": 0.05222063697874546,
      "rewards/forensics_tiou_reward": 0.49929994344711304,
      "step": 587
    },
    {
      "completion_length": 27.71875,
      "epoch": 1.8049155145929339,
      "grad_norm": 4.9157490730285645,
      "kl": 0.3798828125,
      "learning_rate": 5.484615384615385e-07,
      "loss": 0.1096,
      "reward": 0.7294518649578094,
      "reward_std": 0.08964584395289421,
      "reward_std_all": 0.08551367744803429,
      "rewards-max/forensics_iou_reward": 0.9787384867668152,
      "rewards-max/forensics_strict_edge_reward": 0.4389830380678177,
      "rewards-max/forensics_tiou_reward": 0.9716561138629913,
      "rewards/forensics_iou_reward": 0.5940988063812256,
      "rewards/forensics_strict_edge_reward": 0.026352081447839737,
      "rewards/forensics_tiou_reward": 0.5464979708194733,
      "step": 588
    },
    {
      "completion_length": 24.75,
      "epoch": 1.8079877112135176,
      "grad_norm": 5.352863311767578,
      "kl": 0.4033203125,
      "learning_rate": 5.476923076923077e-07,
      "loss": 0.1415,
      "reward": 0.6740870773792267,
      "reward_std": 0.08464943617582321,
      "reward_std_all": 0.10868275165557861,
      "rewards-max/forensics_iou_reward": 0.96827632188797,
      "rewards-max/forensics_strict_edge_reward": 0.51341712474823,
      "rewards-max/forensics_tiou_reward": 0.9594029784202576,
      "rewards/forensics_iou_reward": 0.5353632867336273,
      "rewards/forensics_strict_edge_reward": 0.034499332308769226,
      "rewards/forensics_tiou_reward": 0.4864101856946945,
      "step": 589
    },
    {
      "completion_length": 25.09375,
      "epoch": 1.8110599078341014,
      "grad_norm": 4.924567222595215,
      "kl": 0.4853515625,
      "learning_rate": 5.469230769230769e-07,
      "loss": 0.1348,
      "reward": 0.7110334634780884,
      "reward_std": 0.10295740514993668,
      "reward_std_all": 0.09883055090904236,
      "rewards-max/forensics_iou_reward": 0.9898719489574432,
      "rewards-max/forensics_strict_edge_reward": 0.7517415881156921,
      "rewards-max/forensics_tiou_reward": 0.9860741496086121,
      "rewards/forensics_iou_reward": 0.6063417196273804,
      "rewards/forensics_strict_edge_reward": 0.07991832681000233,
      "rewards/forensics_tiou_reward": 0.5578413307666779,
      "step": 590
    },
    {
      "completion_length": 24.921875,
      "epoch": 1.8141321044546852,
      "grad_norm": 5.0763258934021,
      "kl": 0.3701171875,
      "learning_rate": 5.461538461538461e-07,
      "loss": 0.0087,
      "reward": 0.7082967162132263,
      "reward_std": 0.051334040239453316,
      "reward_std_all": 0.07401853799819946,
      "rewards-max/forensics_iou_reward": 0.9627797901630402,
      "rewards-max/forensics_strict_edge_reward": 0.43393993377685547,
      "rewards-max/forensics_tiou_reward": 0.9517515897750854,
      "rewards/forensics_iou_reward": 0.6082926094532013,
      "rewards/forensics_strict_edge_reward": 0.054952844977378845,
      "rewards/forensics_tiou_reward": 0.5463666617870331,
      "step": 591
    },
    {
      "completion_length": 29.625,
      "epoch": 1.817204301075269,
      "grad_norm": 4.419260501861572,
      "kl": 0.451171875,
      "learning_rate": 5.453846153846153e-07,
      "loss": 0.1193,
      "reward": 0.7179202437400818,
      "reward_std": 0.09687063843011856,
      "reward_std_all": 0.09287535771727562,
      "rewards-max/forensics_iou_reward": 0.9759891629219055,
      "rewards-max/forensics_strict_edge_reward": 0.5756610408425331,
      "rewards-max/forensics_tiou_reward": 0.9684669077396393,
      "rewards/forensics_iou_reward": 0.5742126107215881,
      "rewards/forensics_strict_edge_reward": 0.05649377405643463,
      "rewards/forensics_tiou_reward": 0.513947069644928,
      "step": 592
    },
    {
      "completion_length": 29.328125,
      "epoch": 1.8202764976958525,
      "grad_norm": 5.82667350769043,
      "kl": 0.41796875,
      "learning_rate": 5.446153846153846e-07,
      "loss": 0.0848,
      "reward": 0.652537614107132,
      "reward_std": 0.11745032295584679,
      "reward_std_all": 0.12768059596419334,
      "rewards-max/forensics_iou_reward": 0.9277240037918091,
      "rewards-max/forensics_strict_edge_reward": 0.36162828654050827,
      "rewards-max/forensics_tiou_reward": 0.9095823466777802,
      "rewards/forensics_iou_reward": 0.5249245762825012,
      "rewards/forensics_strict_edge_reward": 0.036969877779483795,
      "rewards/forensics_tiou_reward": 0.47436027228832245,
      "step": 593
    },
    {
      "completion_length": 24.71875,
      "epoch": 1.823348694316436,
      "grad_norm": 6.586769104003906,
      "kl": 0.4736328125,
      "learning_rate": 5.438461538461538e-07,
      "loss": 0.1067,
      "reward": 0.6793156862258911,
      "reward_std": 0.07543215341866016,
      "reward_std_all": 0.09842178598046303,
      "rewards-max/forensics_iou_reward": 0.9708185493946075,
      "rewards-max/forensics_strict_edge_reward": 0.4624108225107193,
      "rewards-max/forensics_tiou_reward": 0.9597480893135071,
      "rewards/forensics_iou_reward": 0.528855174779892,
      "rewards/forensics_strict_edge_reward": 0.03545496053993702,
      "rewards/forensics_tiou_reward": 0.4737379848957062,
      "step": 594
    },
    {
      "completion_length": 27.78125,
      "epoch": 1.8264208909370199,
      "grad_norm": 5.683557987213135,
      "kl": 0.328125,
      "learning_rate": 5.430769230769231e-07,
      "loss": 0.1771,
      "reward": 0.6939166188240051,
      "reward_std": 0.06544658914208412,
      "reward_std_all": 0.08820916712284088,
      "rewards-max/forensics_iou_reward": 0.8660362660884857,
      "rewards-max/forensics_strict_edge_reward": 0.23069029673933983,
      "rewards-max/forensics_tiou_reward": 0.8231179416179657,
      "rewards/forensics_iou_reward": 0.4954662024974823,
      "rewards/forensics_strict_edge_reward": 0.015030541806481779,
      "rewards/forensics_tiou_reward": 0.4385591149330139,
      "step": 595
    },
    {
      "completion_length": 23.828125,
      "epoch": 1.8294930875576036,
      "grad_norm": 5.779327392578125,
      "kl": 0.3544921875,
      "learning_rate": 5.423076923076923e-07,
      "loss": 0.052,
      "reward": 0.6275364756584167,
      "reward_std": 0.09000028856098652,
      "reward_std_all": 0.12803667783737183,
      "rewards-max/forensics_iou_reward": 0.9306722581386566,
      "rewards-max/forensics_strict_edge_reward": 0.35114636272192,
      "rewards-max/forensics_tiou_reward": 0.9205948412418365,
      "rewards/forensics_iou_reward": 0.4717932194471359,
      "rewards/forensics_strict_edge_reward": 0.023237102199345827,
      "rewards/forensics_tiou_reward": 0.410065233707428,
      "step": 596
    },
    {
      "completion_length": 23.859375,
      "epoch": 1.8325652841781874,
      "grad_norm": 5.923611640930176,
      "kl": 0.29296875,
      "learning_rate": 5.415384615384615e-07,
      "loss": 0.0863,
      "reward": 0.7388899922370911,
      "reward_std": 0.07022954523563385,
      "reward_std_all": 0.0692854281514883,
      "rewards-max/forensics_iou_reward": 0.9735500514507294,
      "rewards-max/forensics_strict_edge_reward": 0.3564630448818207,
      "rewards-max/forensics_tiou_reward": 0.9659806191921234,
      "rewards/forensics_iou_reward": 0.6048950850963593,
      "rewards/forensics_strict_edge_reward": 0.048792192712426186,
      "rewards/forensics_tiou_reward": 0.5496989786624908,
      "step": 597
    },
    {
      "completion_length": 24.453125,
      "epoch": 1.8356374807987712,
      "grad_norm": 5.192111492156982,
      "kl": 0.3671875,
      "learning_rate": 5.407692307692308e-07,
      "loss": 0.1249,
      "reward": 0.6645417809486389,
      "reward_std": 0.13238553702831268,
      "reward_std_all": 0.13510657474398613,
      "rewards-max/forensics_iou_reward": 0.9430474936962128,
      "rewards-max/forensics_strict_edge_reward": 0.23274682462215424,
      "rewards-max/forensics_tiou_reward": 0.9230546355247498,
      "rewards/forensics_iou_reward": 0.5575107038021088,
      "rewards/forensics_strict_edge_reward": 0.0176444579847157,
      "rewards/forensics_tiou_reward": 0.5084756016731262,
      "step": 598
    },
    {
      "completion_length": 27.859375,
      "epoch": 1.838709677419355,
      "grad_norm": 3.772642135620117,
      "kl": 0.2978515625,
      "learning_rate": 5.4e-07,
      "loss": 0.0274,
      "reward": 0.6573884189128876,
      "reward_std": 0.09488915279507637,
      "reward_std_all": 0.11950847133994102,
      "rewards-max/forensics_iou_reward": 0.9648409187793732,
      "rewards-max/forensics_strict_edge_reward": 0.3299398720264435,
      "rewards-max/forensics_tiou_reward": 0.9456553161144257,
      "rewards/forensics_iou_reward": 0.5037258416414261,
      "rewards/forensics_strict_edge_reward": 0.03542467299848795,
      "rewards/forensics_tiou_reward": 0.4431760460138321,
      "step": 599
    },
    {
      "completion_length": 26.15625,
      "epoch": 1.8417818740399385,
      "grad_norm": 4.709334373474121,
      "kl": 0.3203125,
      "learning_rate": 5.392307692307691e-07,
      "loss": 0.0543,
      "reward": 0.7336516976356506,
      "reward_std": 0.07734828302636743,
      "reward_std_all": 0.07548095658421516,
      "rewards-max/forensics_iou_reward": 0.9400433599948883,
      "rewards-max/forensics_strict_edge_reward": 0.2030430231243372,
      "rewards-max/forensics_tiou_reward": 0.925230473279953,
      "rewards/forensics_iou_reward": 0.5990494191646576,
      "rewards/forensics_strict_edge_reward": 0.02138356352224946,
      "rewards/forensics_tiou_reward": 0.5410927832126617,
      "step": 600
    },
    {
      "completion_length": 23.234375,
      "epoch": 1.8448540706605223,
      "grad_norm": 6.899382591247559,
      "kl": 0.373046875,
      "learning_rate": 5.384615384615384e-07,
      "loss": 0.1303,
      "reward": 0.6512640416622162,
      "reward_std": 0.16543250530958176,
      "reward_std_all": 0.16013360768556595,
      "rewards-max/forensics_iou_reward": 0.9908740818500519,
      "rewards-max/forensics_strict_edge_reward": 0.620985820889473,
      "rewards-max/forensics_tiou_reward": 0.9859908223152161,
      "rewards/forensics_iou_reward": 0.5620950609445572,
      "rewards/forensics_strict_edge_reward": 0.06988580524921417,
      "rewards/forensics_tiou_reward": 0.516973227262497,
      "step": 601
    },
    {
      "completion_length": 25.171875,
      "epoch": 1.8479262672811059,
      "grad_norm": 5.118136882781982,
      "kl": 0.4296875,
      "learning_rate": 5.376923076923077e-07,
      "loss": 0.1825,
      "reward": 0.6811306774616241,
      "reward_std": 0.12945066764950752,
      "reward_std_all": 0.13461555168032646,
      "rewards-max/forensics_iou_reward": 0.9763483703136444,
      "rewards-max/forensics_strict_edge_reward": 0.493121400475502,
      "rewards-max/forensics_tiou_reward": 0.9656452834606171,
      "rewards/forensics_iou_reward": 0.587218850851059,
      "rewards/forensics_strict_edge_reward": 0.04624152462929487,
      "rewards/forensics_tiou_reward": 0.5440165102481842,
      "step": 602
    },
    {
      "completion_length": 25.453125,
      "epoch": 1.8509984639016897,
      "grad_norm": 4.705447673797607,
      "kl": 0.3603515625,
      "learning_rate": 5.369230769230769e-07,
      "loss": 0.1876,
      "reward": 0.7293674051761627,
      "reward_std": 0.07423270493745804,
      "reward_std_all": 0.07749275304377079,
      "rewards-max/forensics_iou_reward": 0.9713014960289001,
      "rewards-max/forensics_strict_edge_reward": 0.41878940537571907,
      "rewards-max/forensics_tiou_reward": 0.9615910947322845,
      "rewards/forensics_iou_reward": 0.5457961559295654,
      "rewards/forensics_strict_edge_reward": 0.03231999650597572,
      "rewards/forensics_tiou_reward": 0.501152366399765,
      "step": 603
    },
    {
      "completion_length": 26.140625,
      "epoch": 1.8540706605222734,
      "grad_norm": 5.1669440269470215,
      "kl": 0.330078125,
      "learning_rate": 5.361538461538462e-07,
      "loss": 0.0577,
      "reward": 0.7622332572937012,
      "reward_std": 0.07092739176005125,
      "reward_std_all": 0.06409966154024005,
      "rewards-max/forensics_iou_reward": 0.9713448882102966,
      "rewards-max/forensics_strict_edge_reward": 0.15090805664658546,
      "rewards-max/forensics_tiou_reward": 0.9553632438182831,
      "rewards/forensics_iou_reward": 0.6641240119934082,
      "rewards/forensics_strict_edge_reward": 0.014935638522729278,
      "rewards/forensics_tiou_reward": 0.6168197691440582,
      "step": 604
    },
    {
      "completion_length": 23.703125,
      "epoch": 1.8571428571428572,
      "grad_norm": 5.578022480010986,
      "kl": 0.3330078125,
      "learning_rate": 5.353846153846153e-07,
      "loss": 0.0533,
      "reward": 0.7937824130058289,
      "reward_std": 0.011510349344462156,
      "reward_std_all": 0.013517549959942698,
      "rewards-max/forensics_iou_reward": 0.933941513299942,
      "rewards-max/forensics_strict_edge_reward": 0.34701039642095566,
      "rewards-max/forensics_tiou_reward": 0.9160723686218262,
      "rewards/forensics_iou_reward": 0.6501255333423615,
      "rewards/forensics_strict_edge_reward": 0.01814843504689634,
      "rewards/forensics_tiou_reward": 0.5831384658813477,
      "step": 605
    },
    {
      "completion_length": 25.578125,
      "epoch": 1.860215053763441,
      "grad_norm": 8.648148536682129,
      "kl": 1.0,
      "learning_rate": 5.346153846153846e-07,
      "loss": -0.0194,
      "reward": 0.602517157793045,
      "reward_std": 0.1322338730096817,
      "reward_std_all": 0.16302119940519333,
      "rewards-max/forensics_iou_reward": 0.9777754545211792,
      "rewards-max/forensics_strict_edge_reward": 0.38586731255054474,
      "rewards-max/forensics_tiou_reward": 0.9694041609764099,
      "rewards/forensics_iou_reward": 0.4468119144439697,
      "rewards/forensics_strict_edge_reward": 0.03149146493524313,
      "rewards/forensics_tiou_reward": 0.39987270534038544,
      "step": 606
    },
    {
      "completion_length": 26.984375,
      "epoch": 1.8632872503840245,
      "grad_norm": 7.838016033172607,
      "kl": 0.6591796875,
      "learning_rate": 5.338461538461538e-07,
      "loss": 0.0959,
      "reward": 0.717077374458313,
      "reward_std": 0.0958139793947339,
      "reward_std_all": 0.09483647346496582,
      "rewards-max/forensics_iou_reward": 0.9316627979278564,
      "rewards-max/forensics_strict_edge_reward": 0.1836952492594719,
      "rewards-max/forensics_tiou_reward": 0.9091658890247345,
      "rewards/forensics_iou_reward": 0.5960373878479004,
      "rewards/forensics_strict_edge_reward": 0.020373203326016665,
      "rewards/forensics_tiou_reward": 0.5403841435909271,
      "step": 607
    },
    {
      "completion_length": 28.0625,
      "epoch": 1.8663594470046083,
      "grad_norm": 4.27046537399292,
      "kl": 0.3046875,
      "learning_rate": 5.33076923076923e-07,
      "loss": 0.0719,
      "reward": 0.6977997124195099,
      "reward_std": 0.05344676971435547,
      "reward_std_all": 0.08211325854063034,
      "rewards-max/forensics_iou_reward": 0.9820723831653595,
      "rewards-max/forensics_strict_edge_reward": 0.6153696477413177,
      "rewards-max/forensics_tiou_reward": 0.9752947986125946,
      "rewards/forensics_iou_reward": 0.6111133694648743,
      "rewards/forensics_strict_edge_reward": 0.053386472165584564,
      "rewards/forensics_tiou_reward": 0.5654116272926331,
      "step": 608
    },
    {
      "completion_length": 28.1875,
      "epoch": 1.8694316436251919,
      "grad_norm": 5.6060590744018555,
      "kl": 0.400390625,
      "learning_rate": 5.323076923076923e-07,
      "loss": 0.1047,
      "reward": 0.697593092918396,
      "reward_std": 0.1007302775979042,
      "reward_std_all": 0.1018412597477436,
      "rewards-max/forensics_iou_reward": 0.9704392850399017,
      "rewards-max/forensics_strict_edge_reward": 0.4154136925935745,
      "rewards-max/forensics_tiou_reward": 0.953972727060318,
      "rewards/forensics_iou_reward": 0.5419498682022095,
      "rewards/forensics_strict_edge_reward": 0.03791728988289833,
      "rewards/forensics_tiou_reward": 0.5019550323486328,
      "step": 609
    },
    {
      "completion_length": 26.578125,
      "epoch": 1.8725038402457757,
      "grad_norm": 5.483158111572266,
      "kl": 0.349609375,
      "learning_rate": 5.315384615384615e-07,
      "loss": 0.0557,
      "reward": 0.6734490692615509,
      "reward_std": 0.09666464291512966,
      "reward_std_all": 0.11250683665275574,
      "rewards-max/forensics_iou_reward": 0.9418854415416718,
      "rewards-max/forensics_strict_edge_reward": 0.3336062878370285,
      "rewards-max/forensics_tiou_reward": 0.9286927282810211,
      "rewards/forensics_iou_reward": 0.5075167864561081,
      "rewards/forensics_strict_edge_reward": 0.019898767583072186,
      "rewards/forensics_tiou_reward": 0.44691747426986694,
      "step": 610
    },
    {
      "completion_length": 21.453125,
      "epoch": 1.8755760368663594,
      "grad_norm": 5.481334686279297,
      "kl": 0.3232421875,
      "learning_rate": 5.307692307692308e-07,
      "loss": 0.0163,
      "reward": 0.7851961255073547,
      "reward_std": 0.023966305889189243,
      "reward_std_all": 0.024591099470853806,
      "rewards-max/forensics_iou_reward": 0.9725620150566101,
      "rewards-max/forensics_strict_edge_reward": 0.6271176636219025,
      "rewards-max/forensics_tiou_reward": 0.9661875367164612,
      "rewards/forensics_iou_reward": 0.6751643121242523,
      "rewards/forensics_strict_edge_reward": 0.0803824570029974,
      "rewards/forensics_tiou_reward": 0.6151793897151947,
      "step": 611
    },
    {
      "completion_length": 22.390625,
      "epoch": 1.8786482334869432,
      "grad_norm": 7.380529880523682,
      "kl": 0.4912109375,
      "learning_rate": 5.3e-07,
      "loss": 0.086,
      "reward": 0.6710951328277588,
      "reward_std": 0.12237808294594288,
      "reward_std_all": 0.12373360991477966,
      "rewards-max/forensics_iou_reward": 0.956986665725708,
      "rewards-max/forensics_strict_edge_reward": 0.33851243555545807,
      "rewards-max/forensics_tiou_reward": 0.9469067752361298,
      "rewards/forensics_iou_reward": 0.5190462172031403,
      "rewards/forensics_strict_edge_reward": 0.025850875303149223,
      "rewards/forensics_tiou_reward": 0.45079652965068817,
      "step": 612
    },
    {
      "completion_length": 24.046875,
      "epoch": 1.881720430107527,
      "grad_norm": 5.630440711975098,
      "kl": 0.3837890625,
      "learning_rate": 5.292307692307692e-07,
      "loss": 0.0267,
      "reward": 0.7551712989807129,
      "reward_std": 0.07948329672217369,
      "reward_std_all": 0.07161122933030128,
      "rewards-max/forensics_iou_reward": 0.97581946849823,
      "rewards-max/forensics_strict_edge_reward": 0.36554911732673645,
      "rewards-max/forensics_tiou_reward": 0.963831216096878,
      "rewards/forensics_iou_reward": 0.6654097735881805,
      "rewards/forensics_strict_edge_reward": 0.048626627773046494,
      "rewards/forensics_tiou_reward": 0.6147285997867584,
      "step": 613
    },
    {
      "completion_length": 24.109375,
      "epoch": 1.8847926267281108,
      "grad_norm": 5.885781764984131,
      "kl": 0.458984375,
      "learning_rate": 5.284615384615384e-07,
      "loss": 0.0536,
      "reward": 0.6180300712585449,
      "reward_std": 0.09955986775457859,
      "reward_std_all": 0.14098582416772842,
      "rewards-max/forensics_iou_reward": 0.9731330573558807,
      "rewards-max/forensics_strict_edge_reward": 0.4466163069009781,
      "rewards-max/forensics_tiou_reward": 0.963207334280014,
      "rewards/forensics_iou_reward": 0.5300302356481552,
      "rewards/forensics_strict_edge_reward": 0.034242562018334866,
      "rewards/forensics_tiou_reward": 0.48098696768283844,
      "step": 614
    },
    {
      "completion_length": 27.640625,
      "epoch": 1.8878648233486943,
      "grad_norm": 5.536015510559082,
      "kl": 0.4599609375,
      "learning_rate": 5.276923076923076e-07,
      "loss": 0.2016,
      "reward": 0.6997630000114441,
      "reward_std": 0.10097743943333626,
      "reward_std_all": 0.1058880090713501,
      "rewards-max/forensics_iou_reward": 0.9551691710948944,
      "rewards-max/forensics_strict_edge_reward": 0.3856436163187027,
      "rewards-max/forensics_tiou_reward": 0.941381424665451,
      "rewards/forensics_iou_reward": 0.5002451837062836,
      "rewards/forensics_strict_edge_reward": 0.02362764533609152,
      "rewards/forensics_tiou_reward": 0.43909111618995667,
      "step": 615
    },
    {
      "completion_length": 26.625,
      "epoch": 1.8909370199692779,
      "grad_norm": 5.491552829742432,
      "kl": 0.36328125,
      "learning_rate": 5.269230769230769e-07,
      "loss": -0.0205,
      "reward": 0.762693464756012,
      "reward_std": 0.054039787501096725,
      "reward_std_all": 0.05175945907831192,
      "rewards-max/forensics_iou_reward": 0.9327864050865173,
      "rewards-max/forensics_strict_edge_reward": 0.3141128122806549,
      "rewards-max/forensics_tiou_reward": 0.8976015746593475,
      "rewards/forensics_iou_reward": 0.6120665669441223,
      "rewards/forensics_strict_edge_reward": 0.019631543196737766,
      "rewards/forensics_tiou_reward": 0.5498892664909363,
      "step": 616
    },
    {
      "completion_length": 22.90625,
      "epoch": 1.8940092165898617,
      "grad_norm": 5.682499408721924,
      "kl": 0.45703125,
      "learning_rate": 5.261538461538462e-07,
      "loss": 0.0619,
      "reward": 0.6784323155879974,
      "reward_std": 0.10982400923967361,
      "reward_std_all": 0.11933556012809277,
      "rewards-max/forensics_iou_reward": 0.9578515291213989,
      "rewards-max/forensics_strict_edge_reward": 0.3421594202518463,
      "rewards-max/forensics_tiou_reward": 0.9461172521114349,
      "rewards/forensics_iou_reward": 0.5724507123231888,
      "rewards/forensics_strict_edge_reward": 0.039767845533788204,
      "rewards/forensics_tiou_reward": 0.5240561813116074,
      "step": 617
    },
    {
      "completion_length": 26.515625,
      "epoch": 1.8970814132104454,
      "grad_norm": 5.2772955894470215,
      "kl": 0.3955078125,
      "learning_rate": 5.253846153846153e-07,
      "loss": 0.1421,
      "reward": 0.7307818830013275,
      "reward_std": 0.03884080797433853,
      "reward_std_all": 0.059729790315032005,
      "rewards-max/forensics_iou_reward": 0.9627373516559601,
      "rewards-max/forensics_strict_edge_reward": 0.3787757158279419,
      "rewards-max/forensics_tiou_reward": 0.9416983425617218,
      "rewards/forensics_iou_reward": 0.6056979298591614,
      "rewards/forensics_strict_edge_reward": 0.05192145146429539,
      "rewards/forensics_tiou_reward": 0.5479510128498077,
      "step": 618
    },
    {
      "completion_length": 25.453125,
      "epoch": 1.9001536098310292,
      "grad_norm": 6.447849273681641,
      "kl": 0.359375,
      "learning_rate": 5.246153846153846e-07,
      "loss": 0.0556,
      "reward": 0.6680534183979034,
      "reward_std": 0.11821247264742851,
      "reward_std_all": 0.12237796187400818,
      "rewards-max/forensics_iou_reward": 0.9438939094543457,
      "rewards-max/forensics_strict_edge_reward": 0.4406117331236601,
      "rewards-max/forensics_tiou_reward": 0.9368945360183716,
      "rewards/forensics_iou_reward": 0.5271411240100861,
      "rewards/forensics_strict_edge_reward": 0.03079179860651493,
      "rewards/forensics_tiou_reward": 0.4670659154653549,
      "step": 619
    },
    {
      "completion_length": 25.296875,
      "epoch": 1.903225806451613,
      "grad_norm": 4.963466167449951,
      "kl": 0.3828125,
      "learning_rate": 5.238461538461538e-07,
      "loss": 0.0911,
      "reward": 0.7431831955909729,
      "reward_std": 0.09355117753148079,
      "reward_std_all": 0.08481410518288612,
      "rewards-max/forensics_iou_reward": 0.998527467250824,
      "rewards-max/forensics_strict_edge_reward": 0.6779094636440277,
      "rewards-max/forensics_tiou_reward": 0.997432291507721,
      "rewards/forensics_iou_reward": 0.5682777464389801,
      "rewards/forensics_strict_edge_reward": 0.038983505219221115,
      "rewards/forensics_tiou_reward": 0.5108459442853928,
      "step": 620
    },
    {
      "completion_length": 24.875,
      "epoch": 1.9062980030721968,
      "grad_norm": 5.085384845733643,
      "kl": 0.3759765625,
      "learning_rate": 5.23076923076923e-07,
      "loss": 0.1133,
      "reward": 0.7748657464981079,
      "reward_std": 0.040479617193341255,
      "reward_std_all": 0.038739001378417015,
      "rewards-max/forensics_iou_reward": 0.9518000483512878,
      "rewards-max/forensics_strict_edge_reward": 0.31755489110946655,
      "rewards-max/forensics_tiou_reward": 0.9396051466464996,
      "rewards/forensics_iou_reward": 0.6890048086643219,
      "rewards/forensics_strict_edge_reward": 0.02442004345357418,
      "rewards/forensics_tiou_reward": 0.6331255435943604,
      "step": 621
    },
    {
      "completion_length": 24.8125,
      "epoch": 1.9093701996927803,
      "grad_norm": 5.562001705169678,
      "kl": 0.34375,
      "learning_rate": 5.223076923076924e-07,
      "loss": -0.0299,
      "reward": 0.6880466938018799,
      "reward_std": 0.08286594599485397,
      "reward_std_all": 0.10259514302015305,
      "rewards-max/forensics_iou_reward": 0.9464754462242126,
      "rewards-max/forensics_strict_edge_reward": 0.19684057123959064,
      "rewards-max/forensics_tiou_reward": 0.9326080679893494,
      "rewards/forensics_iou_reward": 0.5759584307670593,
      "rewards/forensics_strict_edge_reward": 0.02335759811103344,
      "rewards/forensics_tiou_reward": 0.5273776799440384,
      "step": 622
    },
    {
      "completion_length": 27.078125,
      "epoch": 1.912442396313364,
      "grad_norm": 5.040302276611328,
      "kl": 0.4287109375,
      "learning_rate": 5.215384615384615e-07,
      "loss": -0.0136,
      "reward": 0.7323206961154938,
      "reward_std": 0.03770469035953283,
      "reward_std_all": 0.05886602960526943,
      "rewards-max/forensics_iou_reward": 0.9436053335666656,
      "rewards-max/forensics_strict_edge_reward": 0.11189350299537182,
      "rewards-max/forensics_tiou_reward": 0.9279710352420807,
      "rewards/forensics_iou_reward": 0.6274089813232422,
      "rewards/forensics_strict_edge_reward": 0.006532947416417301,
      "rewards/forensics_tiou_reward": 0.5625239610671997,
      "step": 623
    },
    {
      "completion_length": 28.125,
      "epoch": 1.9155145929339477,
      "grad_norm": 5.152557849884033,
      "kl": 0.3193359375,
      "learning_rate": 5.207692307692308e-07,
      "loss": 0.0409,
      "reward": 0.6893633306026459,
      "reward_std": 0.11737917736172676,
      "reward_std_all": 0.12219009920954704,
      "rewards-max/forensics_iou_reward": 0.9653390049934387,
      "rewards-max/forensics_strict_edge_reward": 0.2332761362195015,
      "rewards-max/forensics_tiou_reward": 0.9515413343906403,
      "rewards/forensics_iou_reward": 0.5599703788757324,
      "rewards/forensics_strict_edge_reward": 0.018223816994577646,
      "rewards/forensics_tiou_reward": 0.5112634599208832,
      "step": 624
    },
    {
      "completion_length": 28.453125,
      "epoch": 1.9185867895545314,
      "grad_norm": 5.411959648132324,
      "kl": 0.328125,
      "learning_rate": 5.2e-07,
      "loss": 0.0638,
      "reward": 0.6755846440792084,
      "reward_std": 0.09972844272851944,
      "reward_std_all": 0.11700716614723206,
      "rewards-max/forensics_iou_reward": 0.981249988079071,
      "rewards-max/forensics_strict_edge_reward": 0.8032653331756592,
      "rewards-max/forensics_tiou_reward": 0.9776459336280823,
      "rewards/forensics_iou_reward": 0.5450021326541901,
      "rewards/forensics_strict_edge_reward": 0.07178851030766964,
      "rewards/forensics_tiou_reward": 0.5050869286060333,
      "step": 625
    },
    {
      "completion_length": 25.609375,
      "epoch": 1.9216589861751152,
      "grad_norm": 5.029561996459961,
      "kl": 0.3271484375,
      "learning_rate": 5.192307692307692e-07,
      "loss": 0.0546,
      "reward": 0.7368032932281494,
      "reward_std": 0.03923179302364588,
      "reward_std_all": 0.05900863837450743,
      "rewards-max/forensics_iou_reward": 0.9599410593509674,
      "rewards-max/forensics_strict_edge_reward": 0.4226568788290024,
      "rewards-max/forensics_tiou_reward": 0.9499863088130951,
      "rewards/forensics_iou_reward": 0.622316837310791,
      "rewards/forensics_strict_edge_reward": 0.028671376407146454,
      "rewards/forensics_tiou_reward": 0.5600205361843109,
      "step": 626
    },
    {
      "completion_length": 24.625,
      "epoch": 1.924731182795699,
      "grad_norm": 5.760429859161377,
      "kl": 0.3466796875,
      "learning_rate": 5.184615384615384e-07,
      "loss": 0.0328,
      "reward": 0.7239162027835846,
      "reward_std": 0.07868343032896519,
      "reward_std_all": 0.0788128636777401,
      "rewards-max/forensics_iou_reward": 0.9698265492916107,
      "rewards-max/forensics_strict_edge_reward": 0.1095364922657609,
      "rewards-max/forensics_tiou_reward": 0.9571642279624939,
      "rewards/forensics_iou_reward": 0.631047934293747,
      "rewards/forensics_strict_edge_reward": 0.00861406757030636,
      "rewards/forensics_tiou_reward": 0.5875123739242554,
      "step": 627
    },
    {
      "completion_length": 25.484375,
      "epoch": 1.9278033794162828,
      "grad_norm": 4.695008277893066,
      "kl": 0.4130859375,
      "learning_rate": 5.176923076923076e-07,
      "loss": 0.2147,
      "reward": 0.6582622230052948,
      "reward_std": 0.087090153247118,
      "reward_std_all": 0.11856008693575859,
      "rewards-max/forensics_iou_reward": 0.9389077723026276,
      "rewards-max/forensics_strict_edge_reward": 0.2911340296268463,
      "rewards-max/forensics_tiou_reward": 0.9206303358078003,
      "rewards/forensics_iou_reward": 0.5485706329345703,
      "rewards/forensics_strict_edge_reward": 0.022790037095546722,
      "rewards/forensics_tiou_reward": 0.48552653193473816,
      "step": 628
    },
    {
      "completion_length": 25.5,
      "epoch": 1.9308755760368663,
      "grad_norm": 5.255928993225098,
      "kl": 0.2734375,
      "learning_rate": 5.16923076923077e-07,
      "loss": 0.0603,
      "reward": 0.6939323842525482,
      "reward_std": 0.09900976344943047,
      "reward_std_all": 0.10166433826088905,
      "rewards-max/forensics_iou_reward": 0.9299714267253876,
      "rewards-max/forensics_strict_edge_reward": 0.43661171942949295,
      "rewards-max/forensics_tiou_reward": 0.9200832843780518,
      "rewards/forensics_iou_reward": 0.5487945973873138,
      "rewards/forensics_strict_edge_reward": 0.048778366297483444,
      "rewards/forensics_tiou_reward": 0.4991714656352997,
      "step": 629
    },
    {
      "completion_length": 30.125,
      "epoch": 1.93394777265745,
      "grad_norm": 4.961763858795166,
      "kl": 0.494140625,
      "learning_rate": 5.161538461538462e-07,
      "loss": 0.0805,
      "reward": 0.6825579404830933,
      "reward_std": 0.08441886305809021,
      "reward_std_all": 0.10496199317276478,
      "rewards-max/forensics_iou_reward": 0.9471911787986755,
      "rewards-max/forensics_strict_edge_reward": 0.3308589681982994,
      "rewards-max/forensics_tiou_reward": 0.9380380511283875,
      "rewards/forensics_iou_reward": 0.526629775762558,
      "rewards/forensics_strict_edge_reward": 0.03591251699253917,
      "rewards/forensics_tiou_reward": 0.48409122228622437,
      "step": 630
    },
    {
      "completion_length": 26.328125,
      "epoch": 1.9370199692780337,
      "grad_norm": 5.032156467437744,
      "kl": 0.4765625,
      "learning_rate": 5.153846153846153e-07,
      "loss": 0.1315,
      "reward": 0.6526670753955841,
      "reward_std": 0.11235947906970978,
      "reward_std_all": 0.1371065154671669,
      "rewards-max/forensics_iou_reward": 0.9731404483318329,
      "rewards-max/forensics_strict_edge_reward": 0.4113115221261978,
      "rewards-max/forensics_tiou_reward": 0.9621555805206299,
      "rewards/forensics_iou_reward": 0.5535082519054413,
      "rewards/forensics_strict_edge_reward": 0.03875126503407955,
      "rewards/forensics_tiou_reward": 0.5079823434352875,
      "step": 631
    },
    {
      "completion_length": 23.75,
      "epoch": 1.9400921658986174,
      "grad_norm": 5.7803754806518555,
      "kl": 0.431640625,
      "learning_rate": 5.146153846153846e-07,
      "loss": 0.0956,
      "reward": 0.5975916981697083,
      "reward_std": 0.059606995433568954,
      "reward_std_all": 0.13478617370128632,
      "rewards-max/forensics_iou_reward": 0.9370532929897308,
      "rewards-max/forensics_strict_edge_reward": 0.36448411643505096,
      "rewards-max/forensics_tiou_reward": 0.9183236956596375,
      "rewards/forensics_iou_reward": 0.46606943011283875,
      "rewards/forensics_strict_edge_reward": 0.0340759065002203,
      "rewards/forensics_tiou_reward": 0.41932517290115356,
      "step": 632
    },
    {
      "completion_length": 27.53125,
      "epoch": 1.9431643625192012,
      "grad_norm": 6.1813764572143555,
      "kl": 0.390625,
      "learning_rate": 5.138461538461538e-07,
      "loss": 0.1492,
      "reward": 0.5978701710700989,
      "reward_std": 0.14547103270888329,
      "reward_std_all": 0.17300143837928772,
      "rewards-max/forensics_iou_reward": 0.8928853571414948,
      "rewards-max/forensics_strict_edge_reward": 0.28160209581255913,
      "rewards-max/forensics_tiou_reward": 0.8705379366874695,
      "rewards/forensics_iou_reward": 0.4229910969734192,
      "rewards/forensics_strict_edge_reward": 0.03725527483038604,
      "rewards/forensics_tiou_reward": 0.37299442291259766,
      "step": 633
    },
    {
      "completion_length": 24.265625,
      "epoch": 1.946236559139785,
      "grad_norm": 5.687857151031494,
      "kl": 0.375,
      "learning_rate": 5.13076923076923e-07,
      "loss": 0.1006,
      "reward": 0.7072724997997284,
      "reward_std": 0.10389645025134087,
      "reward_std_all": 0.10188686102628708,
      "rewards-max/forensics_iou_reward": 0.9160856604576111,
      "rewards-max/forensics_strict_edge_reward": 0.3269258551299572,
      "rewards-max/forensics_tiou_reward": 0.891514390707016,
      "rewards/forensics_iou_reward": 0.5487186014652252,
      "rewards/forensics_strict_edge_reward": 0.016949155367910862,
      "rewards/forensics_tiou_reward": 0.4805307388305664,
      "step": 634
    },
    {
      "completion_length": 30.25,
      "epoch": 1.9493087557603688,
      "grad_norm": 5.325872421264648,
      "kl": 0.359375,
      "learning_rate": 5.123076923076923e-07,
      "loss": 0.0231,
      "reward": 0.6095866858959198,
      "reward_std": 0.057514776359312236,
      "reward_std_all": 0.1262219063937664,
      "rewards-max/forensics_iou_reward": 0.9359737038612366,
      "rewards-max/forensics_strict_edge_reward": 0.29245471209287643,
      "rewards-max/forensics_tiou_reward": 0.9250588715076447,
      "rewards/forensics_iou_reward": 0.4572453945875168,
      "rewards/forensics_strict_edge_reward": 0.02683845814317465,
      "rewards/forensics_tiou_reward": 0.397964671254158,
      "step": 635
    },
    {
      "completion_length": 26.296875,
      "epoch": 1.9523809523809523,
      "grad_norm": 6.566474437713623,
      "kl": 0.55078125,
      "learning_rate": 5.115384615384615e-07,
      "loss": 0.135,
      "reward": 0.7278622686862946,
      "reward_std": 0.06300000473856926,
      "reward_std_all": 0.07198002561926842,
      "rewards-max/forensics_iou_reward": 0.9296244382858276,
      "rewards-max/forensics_strict_edge_reward": 0.29667849838733673,
      "rewards-max/forensics_tiou_reward": 0.907367080450058,
      "rewards/forensics_iou_reward": 0.5563235878944397,
      "rewards/forensics_strict_edge_reward": 0.020757691003382206,
      "rewards/forensics_tiou_reward": 0.4980601519346237,
      "step": 636
    },
    {
      "completion_length": 27.390625,
      "epoch": 1.955453149001536,
      "grad_norm": 6.0231499671936035,
      "kl": 0.349609375,
      "learning_rate": 5.107692307692308e-07,
      "loss": 0.065,
      "reward": 0.6971627473831177,
      "reward_std": 0.10986370965838432,
      "reward_std_all": 0.10807382315397263,
      "rewards-max/forensics_iou_reward": 0.9127014875411987,
      "rewards-max/forensics_strict_edge_reward": 0.2618055194616318,
      "rewards-max/forensics_tiou_reward": 0.8905892968177795,
      "rewards/forensics_iou_reward": 0.5298943519592285,
      "rewards/forensics_strict_edge_reward": 0.024917734786868095,
      "rewards/forensics_tiou_reward": 0.47508566081523895,
      "step": 637
    },
    {
      "completion_length": 32.15625,
      "epoch": 1.9585253456221197,
      "grad_norm": 5.277888774871826,
      "kl": 0.4638671875,
      "learning_rate": 5.1e-07,
      "loss": 0.243,
      "reward": 0.7316622138023376,
      "reward_std": 0.09944676980376244,
      "reward_std_all": 0.09160896390676498,
      "rewards-max/forensics_iou_reward": 0.9169279038906097,
      "rewards-max/forensics_strict_edge_reward": 0.2730702608823776,
      "rewards-max/forensics_tiou_reward": 0.9007195830345154,
      "rewards/forensics_iou_reward": 0.5609210133552551,
      "rewards/forensics_strict_edge_reward": 0.037440731190145016,
      "rewards/forensics_tiou_reward": 0.5195767879486084,
      "step": 638
    },
    {
      "completion_length": 26.84375,
      "epoch": 1.9615975422427034,
      "grad_norm": 6.283299446105957,
      "kl": 0.34765625,
      "learning_rate": 5.092307692307692e-07,
      "loss": 0.0749,
      "reward": 0.7002196907997131,
      "reward_std": 0.06523387413471937,
      "reward_std_all": 0.08307107724249363,
      "rewards-max/forensics_iou_reward": 0.9511854350566864,
      "rewards-max/forensics_strict_edge_reward": 0.3436185121536255,
      "rewards-max/forensics_tiou_reward": 0.9420082867145538,
      "rewards/forensics_iou_reward": 0.5749596357345581,
      "rewards/forensics_strict_edge_reward": 0.0468729492276907,
      "rewards/forensics_tiou_reward": 0.5150737017393112,
      "step": 639
    },
    {
      "completion_length": 21.6875,
      "epoch": 1.9646697388632872,
      "grad_norm": 5.784713268280029,
      "kl": 0.388671875,
      "learning_rate": 5.084615384615384e-07,
      "loss": -0.01,
      "reward": 0.7045331597328186,
      "reward_std": 0.0855504460632801,
      "reward_std_all": 0.0884130522608757,
      "rewards-max/forensics_iou_reward": 0.9745581746101379,
      "rewards-max/forensics_strict_edge_reward": 0.5179875791072845,
      "rewards-max/forensics_tiou_reward": 0.9644378125667572,
      "rewards/forensics_iou_reward": 0.6134348213672638,
      "rewards/forensics_strict_edge_reward": 0.06603764742612839,
      "rewards/forensics_tiou_reward": 0.5718625038862228,
      "step": 640
    },
    {
      "completion_length": 24.9375,
      "epoch": 1.967741935483871,
      "grad_norm": 4.657483100891113,
      "kl": 0.3564453125,
      "learning_rate": 5.076923076923076e-07,
      "loss": 0.0426,
      "reward": 0.7327158451080322,
      "reward_std": 0.07225774973630905,
      "reward_std_all": 0.07421972416341305,
      "rewards-max/forensics_iou_reward": 0.9684931039810181,
      "rewards-max/forensics_strict_edge_reward": 0.5515579879283905,
      "rewards-max/forensics_tiou_reward": 0.9607957899570465,
      "rewards/forensics_iou_reward": 0.5914792120456696,
      "rewards/forensics_strict_edge_reward": 0.05904591269791126,
      "rewards/forensics_tiou_reward": 0.5468182116746902,
      "step": 641
    },
    {
      "completion_length": 24.078125,
      "epoch": 1.9708141321044548,
      "grad_norm": 5.3423004150390625,
      "kl": 0.38671875,
      "learning_rate": 5.06923076923077e-07,
      "loss": 0.0556,
      "reward": 0.7640455067157745,
      "reward_std": 0.06727716140449047,
      "reward_std_all": 0.060803381726145744,
      "rewards-max/forensics_iou_reward": 0.9802631437778473,
      "rewards-max/forensics_strict_edge_reward": 0.6505971103906631,
      "rewards-max/forensics_tiou_reward": 0.9749499261379242,
      "rewards/forensics_iou_reward": 0.6560899019241333,
      "rewards/forensics_strict_edge_reward": 0.07463331148028374,
      "rewards/forensics_tiou_reward": 0.6041378378868103,
      "step": 642
    },
    {
      "completion_length": 23.015625,
      "epoch": 1.9738863287250386,
      "grad_norm": 7.21133279800415,
      "kl": 0.4150390625,
      "learning_rate": 5.061538461538462e-07,
      "loss": -0.0254,
      "reward": 0.6088319718837738,
      "reward_std": 0.14922895282506943,
      "reward_std_all": 0.1690821498632431,
      "rewards-max/forensics_iou_reward": 0.9089052677154541,
      "rewards-max/forensics_strict_edge_reward": 0.11957495100796223,
      "rewards-max/forensics_tiou_reward": 0.8785176277160645,
      "rewards/forensics_iou_reward": 0.45253537595272064,
      "rewards/forensics_strict_edge_reward": 0.014767147600650787,
      "rewards/forensics_tiou_reward": 0.3926197290420532,
      "step": 643
    },
    {
      "completion_length": 24.265625,
      "epoch": 1.976958525345622,
      "grad_norm": 6.8910112380981445,
      "kl": 0.4384765625,
      "learning_rate": 5.053846153846154e-07,
      "loss": 0.1851,
      "reward": 0.6759990453720093,
      "reward_std": 0.105633195489645,
      "reward_std_all": 0.11825602874159813,
      "rewards-max/forensics_iou_reward": 0.958215206861496,
      "rewards-max/forensics_strict_edge_reward": 0.3482530266046524,
      "rewards-max/forensics_tiou_reward": 0.9372414350509644,
      "rewards/forensics_iou_reward": 0.5673912167549133,
      "rewards/forensics_strict_edge_reward": 0.031521046068519354,
      "rewards/forensics_tiou_reward": 0.5081933438777924,
      "step": 644
    },
    {
      "completion_length": 26.21875,
      "epoch": 1.9800307219662059,
      "grad_norm": 5.107605457305908,
      "kl": 0.2822265625,
      "learning_rate": 5.046153846153846e-07,
      "loss": 0.0352,
      "reward": 0.6842684149742126,
      "reward_std": 0.09926614165306091,
      "reward_std_all": 0.11012380570173264,
      "rewards-max/forensics_iou_reward": 0.9608312249183655,
      "rewards-max/forensics_strict_edge_reward": 0.4468829333782196,
      "rewards-max/forensics_tiou_reward": 0.9508665502071381,
      "rewards/forensics_iou_reward": 0.562358945608139,
      "rewards/forensics_strict_edge_reward": 0.03674107976257801,
      "rewards/forensics_tiou_reward": 0.523996889591217,
      "step": 645
    },
    {
      "completion_length": 26.453125,
      "epoch": 1.9831029185867894,
      "grad_norm": 4.9402852058410645,
      "kl": 0.3564453125,
      "learning_rate": 5.038461538461538e-07,
      "loss": 0.0686,
      "reward": 0.7504750788211823,
      "reward_std": 0.06995655968785286,
      "reward_std_all": 0.06580747291445732,
      "rewards-max/forensics_iou_reward": 0.9930070042610168,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.989068865776062,
      "rewards/forensics_iou_reward": 0.628340482711792,
      "rewards/forensics_strict_edge_reward": 0.05862164497375488,
      "rewards/forensics_tiou_reward": 0.5767245590686798,
      "step": 646
    },
    {
      "completion_length": 25.625,
      "epoch": 1.9861751152073732,
      "grad_norm": 4.847559452056885,
      "kl": 0.3232421875,
      "learning_rate": 5.03076923076923e-07,
      "loss": 0.0181,
      "reward": 0.7611930072307587,
      "reward_std": 0.04213671013712883,
      "reward_std_all": 0.04484911821782589,
      "rewards-max/forensics_iou_reward": 0.9596500098705292,
      "rewards-max/forensics_strict_edge_reward": 0.43997953832149506,
      "rewards-max/forensics_tiou_reward": 0.946975976228714,
      "rewards/forensics_iou_reward": 0.6104921102523804,
      "rewards/forensics_strict_edge_reward": 0.035402562469244,
      "rewards/forensics_tiou_reward": 0.5580801367759705,
      "step": 647
    },
    {
      "completion_length": 28.015625,
      "epoch": 1.989247311827957,
      "grad_norm": 5.417877197265625,
      "kl": 0.345703125,
      "learning_rate": 5.023076923076923e-07,
      "loss": 0.1673,
      "reward": 0.6966165602207184,
      "reward_std": 0.07189582660794258,
      "reward_std_all": 0.09228907711803913,
      "rewards-max/forensics_iou_reward": 0.9425334334373474,
      "rewards-max/forensics_strict_edge_reward": 0.36699535697698593,
      "rewards-max/forensics_tiou_reward": 0.9262197315692902,
      "rewards/forensics_iou_reward": 0.5427925437688828,
      "rewards/forensics_strict_edge_reward": 0.0440313471481204,
      "rewards/forensics_tiou_reward": 0.5055981278419495,
      "step": 648
    },
    {
      "completion_length": 27.046875,
      "epoch": 1.9923195084485408,
      "grad_norm": 5.13401985168457,
      "kl": 0.3759765625,
      "learning_rate": 5.015384615384614e-07,
      "loss": -0.0451,
      "reward": 0.713324785232544,
      "reward_std": 0.047750428318977356,
      "reward_std_all": 0.07060030288994312,
      "rewards-max/forensics_iou_reward": 0.965132474899292,
      "rewards-max/forensics_strict_edge_reward": 0.44934946298599243,
      "rewards-max/forensics_tiou_reward": 0.9524069428443909,
      "rewards/forensics_iou_reward": 0.6112656593322754,
      "rewards/forensics_strict_edge_reward": 0.03497235011309385,
      "rewards/forensics_tiou_reward": 0.5635899305343628,
      "step": 649
    },
    {
      "completion_length": 21.5,
      "epoch": 1.9953917050691246,
      "grad_norm": 5.047667503356934,
      "kl": 0.3740234375,
      "learning_rate": 5.007692307692308e-07,
      "loss": 0.0192,
      "reward": 0.6741236746311188,
      "reward_std": 0.04348496533930302,
      "reward_std_all": 0.0891432985663414,
      "rewards-max/forensics_iou_reward": 0.9947149753570557,
      "rewards-max/forensics_strict_edge_reward": 0.4968930333852768,
      "rewards-max/forensics_tiou_reward": 0.9898125231266022,
      "rewards/forensics_iou_reward": 0.5538812279701233,
      "rewards/forensics_strict_edge_reward": 0.03820755099877715,
      "rewards/forensics_tiou_reward": 0.5084552466869354,
      "step": 650
    },
    {
      "completion_length": 24.734375,
      "epoch": 1.998463901689708,
      "grad_norm": 5.845298767089844,
      "kl": 0.423828125,
      "learning_rate": 5e-07,
      "loss": 0.0089,
      "reward": 0.7326377332210541,
      "reward_std": 0.04154553264379501,
      "reward_std_all": 0.06132984720170498,
      "rewards-max/forensics_iou_reward": 0.9987642765045166,
      "rewards-max/forensics_strict_edge_reward": 0.6779286861419678,
      "rewards-max/forensics_tiou_reward": 0.9979070723056793,
      "rewards/forensics_iou_reward": 0.6433477997779846,
      "rewards/forensics_strict_edge_reward": 0.06094170734286308,
      "rewards/forensics_tiou_reward": 0.5945318639278412,
      "step": 651
    },
    {
      "completion_length": 24.80000114440918,
      "epoch": 2.0,
      "grad_norm": 4.546806335449219,
      "kl": 1.1328125,
      "learning_rate": 4.992307692307692e-07,
      "loss": 0.0111,
      "reward": 0.6610226035118103,
      "reward_std": 0.1669221818447113,
      "reward_std_all": 0.21795181930065155,
      "rewards-max/forensics_iou_reward": 0.786057710647583,
      "rewards-max/forensics_strict_edge_reward": 0.29186201095581055,
      "rewards-max/forensics_tiou_reward": 0.7514904141426086,
      "rewards/forensics_iou_reward": 0.5427140593528748,
      "rewards/forensics_strict_edge_reward": 0.06973861902952194,
      "rewards/forensics_tiou_reward": 0.5120722055435181,
      "step": 652
    },
    {
      "completion_length": 22.8125,
      "epoch": 2.003072196620584,
      "grad_norm": 5.922970771789551,
      "kl": 0.3369140625,
      "learning_rate": 4.984615384615384e-07,
      "loss": 0.026,
      "reward": 0.7259584665298462,
      "reward_std": 0.055031053721904755,
      "reward_std_all": 0.07286545261740685,
      "rewards-max/forensics_iou_reward": 0.9566395580768585,
      "rewards-max/forensics_strict_edge_reward": 0.6720768958330154,
      "rewards-max/forensics_tiou_reward": 0.9493441581726074,
      "rewards/forensics_iou_reward": 0.5780272483825684,
      "rewards/forensics_strict_edge_reward": 0.06876778230071068,
      "rewards/forensics_tiou_reward": 0.5414396822452545,
      "step": 653
    },
    {
      "completion_length": 25.3125,
      "epoch": 2.0061443932411676,
      "grad_norm": 4.874382972717285,
      "kl": 0.3876953125,
      "learning_rate": 4.976923076923077e-07,
      "loss": 0.1114,
      "reward": 0.7038946449756622,
      "reward_std": 0.09787208214402199,
      "reward_std_all": 0.0970931425690651,
      "rewards-max/forensics_iou_reward": 0.9797516167163849,
      "rewards-max/forensics_strict_edge_reward": 0.527929812669754,
      "rewards-max/forensics_tiou_reward": 0.9614807367324829,
      "rewards/forensics_iou_reward": 0.6164079904556274,
      "rewards/forensics_strict_edge_reward": 0.0510999970138073,
      "rewards/forensics_tiou_reward": 0.5678020417690277,
      "step": 654
    },
    {
      "completion_length": 28.640625,
      "epoch": 2.0092165898617513,
      "grad_norm": 4.518969535827637,
      "kl": 0.4091796875,
      "learning_rate": 4.969230769230769e-07,
      "loss": 0.0244,
      "reward": 0.7348554134368896,
      "reward_std": 0.08288195729255676,
      "reward_std_all": 0.07871592417359352,
      "rewards-max/forensics_iou_reward": 0.9724629521369934,
      "rewards-max/forensics_strict_edge_reward": 0.32475871220231056,
      "rewards-max/forensics_tiou_reward": 0.958762526512146,
      "rewards/forensics_iou_reward": 0.5712502896785736,
      "rewards/forensics_strict_edge_reward": 0.031646995805203915,
      "rewards/forensics_tiou_reward": 0.5126788169145584,
      "step": 655
    },
    {
      "completion_length": 26.3125,
      "epoch": 2.0122887864823347,
      "grad_norm": 5.789950847625732,
      "kl": 0.341796875,
      "learning_rate": 4.961538461538462e-07,
      "loss": 0.1292,
      "reward": 0.7352333068847656,
      "reward_std": 0.085776811465621,
      "reward_std_all": 0.08074658177793026,
      "rewards-max/forensics_iou_reward": 0.970067024230957,
      "rewards-max/forensics_strict_edge_reward": 0.5221183001995087,
      "rewards-max/forensics_tiou_reward": 0.9559964239597321,
      "rewards/forensics_iou_reward": 0.6172741949558258,
      "rewards/forensics_strict_edge_reward": 0.05524051561951637,
      "rewards/forensics_tiou_reward": 0.5809731483459473,
      "step": 656
    },
    {
      "completion_length": 26.28125,
      "epoch": 2.0153609831029184,
      "grad_norm": 5.641947269439697,
      "kl": 0.4873046875,
      "learning_rate": 4.953846153846154e-07,
      "loss": 0.0548,
      "reward": 0.7112511992454529,
      "reward_std": 0.09679313376545906,
      "reward_std_all": 0.0942798312753439,
      "rewards-max/forensics_iou_reward": 0.9922807514667511,
      "rewards-max/forensics_strict_edge_reward": 0.8053101599216461,
      "rewards-max/forensics_tiou_reward": 0.9909501075744629,
      "rewards/forensics_iou_reward": 0.6164866089820862,
      "rewards/forensics_strict_edge_reward": 0.09267798066139221,
      "rewards/forensics_tiou_reward": 0.5776968598365784,
      "step": 657
    },
    {
      "completion_length": 27.203125,
      "epoch": 2.0184331797235022,
      "grad_norm": 5.317221164703369,
      "kl": 0.408203125,
      "learning_rate": 4.946153846153846e-07,
      "loss": -0.0208,
      "reward": 0.7067424952983856,
      "reward_std": 0.05322413519024849,
      "reward_std_all": 0.07867134734988213,
      "rewards-max/forensics_iou_reward": 0.9652146995067596,
      "rewards-max/forensics_strict_edge_reward": 0.47400766611099243,
      "rewards-max/forensics_tiou_reward": 0.9543946087360382,
      "rewards/forensics_iou_reward": 0.5638845413923264,
      "rewards/forensics_strict_edge_reward": 0.049796149134635925,
      "rewards/forensics_tiou_reward": 0.5131117403507233,
      "step": 658
    },
    {
      "completion_length": 25.3125,
      "epoch": 2.021505376344086,
      "grad_norm": 4.890255928039551,
      "kl": 0.390625,
      "learning_rate": 4.938461538461538e-07,
      "loss": 0.1382,
      "reward": 0.6914847791194916,
      "reward_std": 0.11769947782158852,
      "reward_std_all": 0.11934912204742432,
      "rewards-max/forensics_iou_reward": 0.9797301292419434,
      "rewards-max/forensics_strict_edge_reward": 0.33019154518842697,
      "rewards-max/forensics_tiou_reward": 0.9736238121986389,
      "rewards/forensics_iou_reward": 0.6067361235618591,
      "rewards/forensics_strict_edge_reward": 0.04333534464240074,
      "rewards/forensics_tiou_reward": 0.5608957260847092,
      "step": 659
    },
    {
      "completion_length": 23.671875,
      "epoch": 2.02457757296467,
      "grad_norm": 5.136505603790283,
      "kl": 0.3671875,
      "learning_rate": 4.93076923076923e-07,
      "loss": -0.017,
      "reward": 0.6941166818141937,
      "reward_std": 0.07985812984406948,
      "reward_std_all": 0.09891800116747618,
      "rewards-max/forensics_iou_reward": 0.977259635925293,
      "rewards-max/forensics_strict_edge_reward": 0.4245914816856384,
      "rewards-max/forensics_tiou_reward": 0.9721031188964844,
      "rewards/forensics_iou_reward": 0.5661566704511642,
      "rewards/forensics_strict_edge_reward": 0.056963070295751095,
      "rewards/forensics_tiou_reward": 0.5252889394760132,
      "step": 660
    },
    {
      "completion_length": 29.53125,
      "epoch": 2.0276497695852536,
      "grad_norm": 5.07091760635376,
      "kl": 0.4423828125,
      "learning_rate": 4.923076923076923e-07,
      "loss": -0.0425,
      "reward": 0.7112119495868683,
      "reward_std": 0.08833043277263641,
      "reward_std_all": 0.08931674435734749,
      "rewards-max/forensics_iou_reward": 0.9287717044353485,
      "rewards-max/forensics_strict_edge_reward": 0.2905498594045639,
      "rewards-max/forensics_tiou_reward": 0.9058716595172882,
      "rewards/forensics_iou_reward": 0.6128187775611877,
      "rewards/forensics_strict_edge_reward": 0.027468949556350708,
      "rewards/forensics_tiou_reward": 0.5665885508060455,
      "step": 661
    },
    {
      "completion_length": 24.59375,
      "epoch": 2.0307219662058373,
      "grad_norm": 4.374538421630859,
      "kl": 0.4169921875,
      "learning_rate": 4.915384615384615e-07,
      "loss": 0.0975,
      "reward": 0.7629688680171967,
      "reward_std": 0.03598199598491192,
      "reward_std_all": 0.041095539927482605,
      "rewards-max/forensics_iou_reward": 0.9510561227798462,
      "rewards-max/forensics_strict_edge_reward": 0.488550528883934,
      "rewards-max/forensics_tiou_reward": 0.9347515404224396,
      "rewards/forensics_iou_reward": 0.6511210799217224,
      "rewards/forensics_strict_edge_reward": 0.042628299444913864,
      "rewards/forensics_tiou_reward": 0.5795740783214569,
      "step": 662
    },
    {
      "completion_length": 27.328125,
      "epoch": 2.033794162826421,
      "grad_norm": 5.855571746826172,
      "kl": 0.42578125,
      "learning_rate": 4.907692307692308e-07,
      "loss": 0.0401,
      "reward": 0.7749729156494141,
      "reward_std": 0.03996859723702073,
      "reward_std_all": 0.038315887562930584,
      "rewards-max/forensics_iou_reward": 0.9785079658031464,
      "rewards-max/forensics_strict_edge_reward": 0.6886506080627441,
      "rewards-max/forensics_tiou_reward": 0.970740020275116,
      "rewards/forensics_iou_reward": 0.6852515637874603,
      "rewards/forensics_strict_edge_reward": 0.05418017506599426,
      "rewards/forensics_tiou_reward": 0.6361493468284607,
      "step": 663
    },
    {
      "completion_length": 29.46875,
      "epoch": 2.0368663594470044,
      "grad_norm": 5.1849541664123535,
      "kl": 0.36328125,
      "learning_rate": 4.9e-07,
      "loss": -0.02,
      "reward": 0.7450951039791107,
      "reward_std": 0.05249364650808275,
      "reward_std_all": 0.05681507755070925,
      "rewards-max/forensics_iou_reward": 0.8645444512367249,
      "rewards-max/forensics_strict_edge_reward": 0.48549328185617924,
      "rewards-max/forensics_tiou_reward": 0.8491839468479156,
      "rewards/forensics_iou_reward": 0.6088326871395111,
      "rewards/forensics_strict_edge_reward": 0.07792975299526006,
      "rewards/forensics_tiou_reward": 0.55349962413311,
      "step": 664
    },
    {
      "completion_length": 28.21875,
      "epoch": 2.0399385560675882,
      "grad_norm": 4.570839881896973,
      "kl": 0.37890625,
      "learning_rate": 4.892307692307692e-07,
      "loss": 0.0227,
      "reward": 0.792013555765152,
      "reward_std": 0.014377731597051024,
      "reward_std_all": 0.015939688310027122,
      "rewards-max/forensics_iou_reward": 0.9728070199489594,
      "rewards-max/forensics_strict_edge_reward": 0.36029765009880066,
      "rewards-max/forensics_tiou_reward": 0.9622257947921753,
      "rewards/forensics_iou_reward": 0.6786108016967773,
      "rewards/forensics_strict_edge_reward": 0.0348155852407217,
      "rewards/forensics_tiou_reward": 0.6223200857639313,
      "step": 665
    },
    {
      "completion_length": 23.875,
      "epoch": 2.043010752688172,
      "grad_norm": 5.994181156158447,
      "kl": 0.3623046875,
      "learning_rate": 4.884615384615384e-07,
      "loss": 0.0662,
      "reward": 0.7557077705860138,
      "reward_std": 0.07199829071760178,
      "reward_std_all": 0.06631030328571796,
      "rewards-max/forensics_iou_reward": 0.9725677967071533,
      "rewards-max/forensics_strict_edge_reward": 0.6526533365249634,
      "rewards-max/forensics_tiou_reward": 0.9677166938781738,
      "rewards/forensics_iou_reward": 0.620369553565979,
      "rewards/forensics_strict_edge_reward": 0.08289481699466705,
      "rewards/forensics_tiou_reward": 0.5546817779541016,
      "step": 666
    },
    {
      "completion_length": 26.828125,
      "epoch": 2.046082949308756,
      "grad_norm": 5.29351806640625,
      "kl": 0.41796875,
      "learning_rate": 4.876923076923077e-07,
      "loss": 0.133,
      "reward": 0.7370971143245697,
      "reward_std": 0.06276397779583931,
      "reward_std_all": 0.06723970174789429,
      "rewards-max/forensics_iou_reward": 0.9875222742557526,
      "rewards-max/forensics_strict_edge_reward": 0.6966203600168228,
      "rewards-max/forensics_tiou_reward": 0.9807189106941223,
      "rewards/forensics_iou_reward": 0.6355805397033691,
      "rewards/forensics_strict_edge_reward": 0.07554742321372032,
      "rewards/forensics_tiou_reward": 0.5793663859367371,
      "step": 667
    },
    {
      "completion_length": 23.515625,
      "epoch": 2.0491551459293396,
      "grad_norm": 17.95240020751953,
      "kl": 1.0966796875,
      "learning_rate": 4.869230769230769e-07,
      "loss": 0.0165,
      "reward": 0.7038888335227966,
      "reward_std": 0.07436582073569298,
      "reward_std_all": 0.09195934236049652,
      "rewards-max/forensics_iou_reward": 0.9966885149478912,
      "rewards-max/forensics_strict_edge_reward": 0.5435412377119064,
      "rewards-max/forensics_tiou_reward": 0.9938434958457947,
      "rewards/forensics_iou_reward": 0.6649656891822815,
      "rewards/forensics_strict_edge_reward": 0.09476523101329803,
      "rewards/forensics_tiou_reward": 0.6294729113578796,
      "step": 668
    },
    {
      "completion_length": 24.5625,
      "epoch": 2.0522273425499233,
      "grad_norm": 4.628787517547607,
      "kl": 0.4951171875,
      "learning_rate": 4.861538461538461e-07,
      "loss": 0.0772,
      "reward": 0.7234466969966888,
      "reward_std": 0.10344668477773666,
      "reward_std_all": 0.09643999487161636,
      "rewards-max/forensics_iou_reward": 0.9522912502288818,
      "rewards-max/forensics_strict_edge_reward": 0.37195122241973877,
      "rewards-max/forensics_tiou_reward": 0.9353817999362946,
      "rewards/forensics_iou_reward": 0.6278667449951172,
      "rewards/forensics_strict_edge_reward": 0.05107389576733112,
      "rewards/forensics_tiou_reward": 0.5801106095314026,
      "step": 669
    },
    {
      "completion_length": 27.953125,
      "epoch": 2.055299539170507,
      "grad_norm": 5.600971698760986,
      "kl": 0.404296875,
      "learning_rate": 4.853846153846154e-07,
      "loss": 0.0524,
      "reward": 0.7170052230358124,
      "reward_std": 0.0885305255651474,
      "reward_std_all": 0.08793758600950241,
      "rewards-max/forensics_iou_reward": 0.9644587934017181,
      "rewards-max/forensics_strict_edge_reward": 0.7090296894311905,
      "rewards-max/forensics_tiou_reward": 0.9543166160583496,
      "rewards/forensics_iou_reward": 0.5958745181560516,
      "rewards/forensics_strict_edge_reward": 0.05117661692202091,
      "rewards/forensics_tiou_reward": 0.5539168119430542,
      "step": 670
    },
    {
      "completion_length": 23.28125,
      "epoch": 2.0583717357910905,
      "grad_norm": 6.171682357788086,
      "kl": 0.302734375,
      "learning_rate": 4.846153846153846e-07,
      "loss": -0.039,
      "reward": 0.7131829261779785,
      "reward_std": 0.08406972885131836,
      "reward_std_all": 0.0859428271651268,
      "rewards-max/forensics_iou_reward": 0.9781509339809418,
      "rewards-max/forensics_strict_edge_reward": 0.580880731344223,
      "rewards-max/forensics_tiou_reward": 0.9720512330532074,
      "rewards/forensics_iou_reward": 0.5786090195178986,
      "rewards/forensics_strict_edge_reward": 0.054884377866983414,
      "rewards/forensics_tiou_reward": 0.5274840891361237,
      "step": 671
    },
    {
      "completion_length": 34.171875,
      "epoch": 2.0614439324116742,
      "grad_norm": 4.671652793884277,
      "kl": 0.3349609375,
      "learning_rate": 4.838461538461539e-07,
      "loss": 0.0614,
      "reward": 0.7340250015258789,
      "reward_std": 0.03953965101391077,
      "reward_std_all": 0.057444196194410324,
      "rewards-max/forensics_iou_reward": 0.984332799911499,
      "rewards-max/forensics_strict_edge_reward": 0.580880731344223,
      "rewards-max/forensics_tiou_reward": 0.9787564277648926,
      "rewards/forensics_iou_reward": 0.6323755383491516,
      "rewards/forensics_strict_edge_reward": 0.06183707155287266,
      "rewards/forensics_tiou_reward": 0.5832285284996033,
      "step": 672
    },
    {
      "completion_length": 29.828125,
      "epoch": 2.064516129032258,
      "grad_norm": 6.086270809173584,
      "kl": 0.5693359375,
      "learning_rate": 4.83076923076923e-07,
      "loss": 0.2155,
      "reward": 0.7375962138175964,
      "reward_std": 0.057742414996027946,
      "reward_std_all": 0.06401808373630047,
      "rewards-max/forensics_iou_reward": 0.9700463116168976,
      "rewards-max/forensics_strict_edge_reward": 0.7029964029788971,
      "rewards-max/forensics_tiou_reward": 0.9647244811058044,
      "rewards/forensics_iou_reward": 0.6296302676200867,
      "rewards/forensics_strict_edge_reward": 0.05450606904923916,
      "rewards/forensics_tiou_reward": 0.5631434321403503,
      "step": 673
    },
    {
      "completion_length": 22.65625,
      "epoch": 2.067588325652842,
      "grad_norm": 5.8413004875183105,
      "kl": 0.34375,
      "learning_rate": 4.823076923076923e-07,
      "loss": 0.0429,
      "reward": 0.692342072725296,
      "reward_std": 0.02936001680791378,
      "reward_std_all": 0.07402899861335754,
      "rewards-max/forensics_iou_reward": 0.9735187590122223,
      "rewards-max/forensics_strict_edge_reward": 0.5431406646966934,
      "rewards-max/forensics_tiou_reward": 0.9626544415950775,
      "rewards/forensics_iou_reward": 0.5830996185541153,
      "rewards/forensics_strict_edge_reward": 0.038189755752682686,
      "rewards/forensics_tiou_reward": 0.5367859601974487,
      "step": 674
    },
    {
      "completion_length": 27.0,
      "epoch": 2.0706605222734256,
      "grad_norm": 5.524745941162109,
      "kl": 0.3349609375,
      "learning_rate": 4.815384615384615e-07,
      "loss": -0.0107,
      "reward": 0.7087573409080505,
      "reward_std": 0.06312945578247309,
      "reward_std_all": 0.08269160985946655,
      "rewards-max/forensics_iou_reward": 0.9966942071914673,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9957011938095093,
      "rewards/forensics_iou_reward": 0.5915577709674835,
      "rewards/forensics_strict_edge_reward": 0.07354575395584106,
      "rewards/forensics_tiou_reward": 0.5409446358680725,
      "step": 675
    },
    {
      "completion_length": 24.421875,
      "epoch": 2.0737327188940093,
      "grad_norm": 6.354891300201416,
      "kl": 0.3701171875,
      "learning_rate": 4.807692307692307e-07,
      "loss": 0.0271,
      "reward": 0.7854792475700378,
      "reward_std": 0.029864483512938023,
      "reward_std_all": 0.028654368594288826,
      "rewards-max/forensics_iou_reward": 0.9957264959812164,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9943761229515076,
      "rewards/forensics_iou_reward": 0.6744315028190613,
      "rewards/forensics_strict_edge_reward": 0.06891575083136559,
      "rewards/forensics_tiou_reward": 0.6332001686096191,
      "step": 676
    },
    {
      "completion_length": 25.625,
      "epoch": 2.076804915514593,
      "grad_norm": 5.970175266265869,
      "kl": 0.3193359375,
      "learning_rate": 4.8e-07,
      "loss": -0.0336,
      "reward": 0.7669289708137512,
      "reward_std": 0.04991414211690426,
      "reward_std_all": 0.04754951596260071,
      "rewards-max/forensics_iou_reward": 0.9847545623779297,
      "rewards-max/forensics_strict_edge_reward": 0.5726078450679779,
      "rewards-max/forensics_tiou_reward": 0.9799601137638092,
      "rewards/forensics_iou_reward": 0.6419031918048859,
      "rewards/forensics_strict_edge_reward": 0.05702671594917774,
      "rewards/forensics_tiou_reward": 0.5895706117153168,
      "step": 677
    },
    {
      "completion_length": 25.625,
      "epoch": 2.0798771121351765,
      "grad_norm": 6.597629070281982,
      "kl": 0.369140625,
      "learning_rate": 4.792307692307692e-07,
      "loss": 0.0289,
      "reward": 0.7545068562030792,
      "reward_std": 0.06017875671386719,
      "reward_std_all": 0.06199384108185768,
      "rewards-max/forensics_iou_reward": 0.9662674069404602,
      "rewards-max/forensics_strict_edge_reward": 0.5829301476478577,
      "rewards-max/forensics_tiou_reward": 0.9533158838748932,
      "rewards/forensics_iou_reward": 0.6470352411270142,
      "rewards/forensics_strict_edge_reward": 0.08135846443474293,
      "rewards/forensics_tiou_reward": 0.6025397181510925,
      "step": 678
    },
    {
      "completion_length": 24.453125,
      "epoch": 2.0829493087557602,
      "grad_norm": 6.384382247924805,
      "kl": 0.384765625,
      "learning_rate": 4.784615384615384e-07,
      "loss": 0.0625,
      "reward": 0.6515230536460876,
      "reward_std": 0.041268179193139076,
      "reward_std_all": 0.09576687216758728,
      "rewards-max/forensics_iou_reward": 0.9442134499549866,
      "rewards-max/forensics_strict_edge_reward": 0.45854125916957855,
      "rewards-max/forensics_tiou_reward": 0.9337225556373596,
      "rewards/forensics_iou_reward": 0.4828142821788788,
      "rewards/forensics_strict_edge_reward": 0.023533841595053673,
      "rewards/forensics_tiou_reward": 0.42812494933605194,
      "step": 679
    },
    {
      "completion_length": 28.546875,
      "epoch": 2.086021505376344,
      "grad_norm": 5.3894147872924805,
      "kl": 0.3798828125,
      "learning_rate": 4.776923076923077e-07,
      "loss": 0.0385,
      "reward": 0.6851821839809418,
      "reward_std": 0.03882589511340484,
      "reward_std_all": 0.08254082128405571,
      "rewards-max/forensics_iou_reward": 0.9594303071498871,
      "rewards-max/forensics_strict_edge_reward": 0.5513441562652588,
      "rewards-max/forensics_tiou_reward": 0.9422908127307892,
      "rewards/forensics_iou_reward": 0.5888898968696594,
      "rewards/forensics_strict_edge_reward": 0.0720023475587368,
      "rewards/forensics_tiou_reward": 0.5458546429872513,
      "step": 680
    },
    {
      "completion_length": 25.4375,
      "epoch": 2.089093701996928,
      "grad_norm": 8.825240135192871,
      "kl": 0.3505859375,
      "learning_rate": 4.769230769230769e-07,
      "loss": -0.0489,
      "reward": 0.7748197019100189,
      "reward_std": 0.032418728806078434,
      "reward_std_all": 0.03309089317917824,
      "rewards-max/forensics_iou_reward": 0.9858682751655579,
      "rewards-max/forensics_strict_edge_reward": 0.5790818333625793,
      "rewards-max/forensics_tiou_reward": 0.9777128994464874,
      "rewards/forensics_iou_reward": 0.6831821501255035,
      "rewards/forensics_strict_edge_reward": 0.09263929352164268,
      "rewards/forensics_tiou_reward": 0.6425755023956299,
      "step": 681
    },
    {
      "completion_length": 25.21875,
      "epoch": 2.0921658986175116,
      "grad_norm": 6.291936874389648,
      "kl": 0.3818359375,
      "learning_rate": 4.7615384615384613e-07,
      "loss": 0.1105,
      "reward": 0.7377504706382751,
      "reward_std": 0.09772241115570068,
      "reward_std_all": 0.08907048963010311,
      "rewards-max/forensics_iou_reward": 0.9810129106044769,
      "rewards-max/forensics_strict_edge_reward": 0.48195670545101166,
      "rewards-max/forensics_tiou_reward": 0.971927285194397,
      "rewards/forensics_iou_reward": 0.6035144925117493,
      "rewards/forensics_strict_edge_reward": 0.04570045694708824,
      "rewards/forensics_tiou_reward": 0.5464145839214325,
      "step": 682
    },
    {
      "completion_length": 22.796875,
      "epoch": 2.0952380952380953,
      "grad_norm": 6.168593883514404,
      "kl": 0.3515625,
      "learning_rate": 4.753846153846154e-07,
      "loss": 0.0614,
      "reward": 0.7745780944824219,
      "reward_std": 0.053336337208747864,
      "reward_std_all": 0.04809685982763767,
      "rewards-max/forensics_iou_reward": 0.9579192996025085,
      "rewards-max/forensics_strict_edge_reward": 0.45157748460769653,
      "rewards-max/forensics_tiou_reward": 0.9464910328388214,
      "rewards/forensics_iou_reward": 0.6500299870967865,
      "rewards/forensics_strict_edge_reward": 0.06333708018064499,
      "rewards/forensics_tiou_reward": 0.6094386279582977,
      "step": 683
    },
    {
      "completion_length": 24.4375,
      "epoch": 2.098310291858679,
      "grad_norm": 5.371388912200928,
      "kl": 0.3251953125,
      "learning_rate": 4.7461538461538456e-07,
      "loss": 0.0992,
      "reward": 0.742817759513855,
      "reward_std": 0.0691755898296833,
      "reward_std_all": 0.06858455389738083,
      "rewards-max/forensics_iou_reward": 0.9531035423278809,
      "rewards-max/forensics_strict_edge_reward": 0.4294941872358322,
      "rewards-max/forensics_tiou_reward": 0.9328278601169586,
      "rewards/forensics_iou_reward": 0.6722034215927124,
      "rewards/forensics_strict_edge_reward": 0.046729592606425285,
      "rewards/forensics_tiou_reward": 0.6212604939937592,
      "step": 684
    },
    {
      "completion_length": 25.8125,
      "epoch": 2.1013824884792625,
      "grad_norm": 63.2021484375,
      "kl": 3.3427734375,
      "learning_rate": 4.738461538461538e-07,
      "loss": 0.1155,
      "reward": 0.6624036431312561,
      "reward_std": 0.11054741963744164,
      "reward_std_all": 0.13157648593187332,
      "rewards-max/forensics_iou_reward": 0.982950896024704,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9772939383983612,
      "rewards/forensics_iou_reward": 0.5819404125213623,
      "rewards/forensics_strict_edge_reward": 0.0696889441460371,
      "rewards/forensics_tiou_reward": 0.5428206324577332,
      "step": 685
    },
    {
      "completion_length": 25.03125,
      "epoch": 2.1044546850998462,
      "grad_norm": 5.038314342498779,
      "kl": 0.4912109375,
      "learning_rate": 4.7307692307692304e-07,
      "loss": 0.0195,
      "reward": 0.7667058408260345,
      "reward_std": 0.053501298651099205,
      "reward_std_all": 0.0501901600509882,
      "rewards-max/forensics_iou_reward": 0.9725581109523773,
      "rewards-max/forensics_strict_edge_reward": 0.48137304186820984,
      "rewards-max/forensics_tiou_reward": 0.9602009057998657,
      "rewards/forensics_iou_reward": 0.6484197080135345,
      "rewards/forensics_strict_edge_reward": 0.04792158119380474,
      "rewards/forensics_tiou_reward": 0.6036548018455505,
      "step": 686
    },
    {
      "completion_length": 25.484375,
      "epoch": 2.10752688172043,
      "grad_norm": 5.825042724609375,
      "kl": 0.3359375,
      "learning_rate": 4.723076923076923e-07,
      "loss": -0.0279,
      "reward": 0.7646511197090149,
      "reward_std": 0.05170526448637247,
      "reward_std_all": 0.04941197484731674,
      "rewards-max/forensics_iou_reward": 0.9457094073295593,
      "rewards-max/forensics_strict_edge_reward": 0.3614702522754669,
      "rewards-max/forensics_tiou_reward": 0.9279983639717102,
      "rewards/forensics_iou_reward": 0.6386978924274445,
      "rewards/forensics_strict_edge_reward": 0.022751525975763798,
      "rewards/forensics_tiou_reward": 0.5789570212364197,
      "step": 687
    },
    {
      "completion_length": 29.125,
      "epoch": 2.110599078341014,
      "grad_norm": 6.3350605964660645,
      "kl": 0.34765625,
      "learning_rate": 4.715384615384615e-07,
      "loss": 0.1452,
      "reward": 0.7069782912731171,
      "reward_std": 0.0688210055232048,
      "reward_std_all": 0.0870291106402874,
      "rewards-max/forensics_iou_reward": 0.9641998112201691,
      "rewards-max/forensics_strict_edge_reward": 0.2796490117907524,
      "rewards-max/forensics_tiou_reward": 0.9506990909576416,
      "rewards/forensics_iou_reward": 0.5867511332035065,
      "rewards/forensics_strict_edge_reward": 0.04581313859671354,
      "rewards/forensics_tiou_reward": 0.5510700941085815,
      "step": 688
    },
    {
      "completion_length": 29.875,
      "epoch": 2.1136712749615976,
      "grad_norm": 5.648200988769531,
      "kl": 0.3466796875,
      "learning_rate": 4.7076923076923073e-07,
      "loss": -0.0767,
      "reward": 0.6549176573753357,
      "reward_std": 0.07391712069511414,
      "reward_std_all": 0.11336362361907959,
      "rewards-max/forensics_iou_reward": 0.9991360604763031,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9987602829933167,
      "rewards/forensics_iou_reward": 0.5897671580314636,
      "rewards/forensics_strict_edge_reward": 0.07308326661586761,
      "rewards/forensics_tiou_reward": 0.5490166395902634,
      "step": 689
    },
    {
      "completion_length": 25.6875,
      "epoch": 2.1167434715821813,
      "grad_norm": 5.744733810424805,
      "kl": 0.41015625,
      "learning_rate": 4.6999999999999995e-07,
      "loss": 0.0702,
      "reward": 0.7070682644844055,
      "reward_std": 0.08811966329813004,
      "reward_std_all": 0.09165192395448685,
      "rewards-max/forensics_iou_reward": 0.9714589416980743,
      "rewards-max/forensics_strict_edge_reward": 0.4626377075910568,
      "rewards-max/forensics_tiou_reward": 0.9596373438835144,
      "rewards/forensics_iou_reward": 0.577940970659256,
      "rewards/forensics_strict_edge_reward": 0.02342122932896018,
      "rewards/forensics_tiou_reward": 0.5245059430599213,
      "step": 690
    },
    {
      "completion_length": 22.078125,
      "epoch": 2.119815668202765,
      "grad_norm": 8.779399871826172,
      "kl": 0.3798828125,
      "learning_rate": 4.692307692307692e-07,
      "loss": 0.0222,
      "reward": 0.7872713506221771,
      "reward_std": 0.0299015260534361,
      "reward_std_all": 0.028083251556381583,
      "rewards-max/forensics_iou_reward": 0.973440557718277,
      "rewards-max/forensics_strict_edge_reward": 0.444107111543417,
      "rewards-max/forensics_tiou_reward": 0.9627246558666229,
      "rewards/forensics_iou_reward": 0.7017103433609009,
      "rewards/forensics_strict_edge_reward": 0.08069798536598682,
      "rewards/forensics_tiou_reward": 0.6486102640628815,
      "step": 691
    },
    {
      "completion_length": 26.578125,
      "epoch": 2.122887864823349,
      "grad_norm": 4.973483562469482,
      "kl": 0.3671875,
      "learning_rate": 4.684615384615385e-07,
      "loss": -0.0273,
      "reward": 0.6973575353622437,
      "reward_std": 0.0528530552983284,
      "reward_std_all": 0.08289765752851963,
      "rewards-max/forensics_iou_reward": 0.9954369366168976,
      "rewards-max/forensics_strict_edge_reward": 0.5739265382289886,
      "rewards-max/forensics_tiou_reward": 0.9922980666160583,
      "rewards/forensics_iou_reward": 0.6095913648605347,
      "rewards/forensics_strict_edge_reward": 0.059364414773881435,
      "rewards/forensics_tiou_reward": 0.5606018006801605,
      "step": 692
    },
    {
      "completion_length": 24.0625,
      "epoch": 2.1259600614439322,
      "grad_norm": 6.198493957519531,
      "kl": 0.2666015625,
      "learning_rate": 4.6769230769230764e-07,
      "loss": -0.0317,
      "reward": 0.7738404273986816,
      "reward_std": 0.033070665784180164,
      "reward_std_all": 0.03401302732527256,
      "rewards-max/forensics_iou_reward": 0.9711267352104187,
      "rewards-max/forensics_strict_edge_reward": 0.6317985653877258,
      "rewards-max/forensics_tiou_reward": 0.9643933475017548,
      "rewards/forensics_iou_reward": 0.6418478786945343,
      "rewards/forensics_strict_edge_reward": 0.043068982660770416,
      "rewards/forensics_tiou_reward": 0.593434602022171,
      "step": 693
    },
    {
      "completion_length": 27.359375,
      "epoch": 2.129032258064516,
      "grad_norm": 5.6686320304870605,
      "kl": 0.580078125,
      "learning_rate": 4.669230769230769e-07,
      "loss": 0.0216,
      "reward": 0.7129563391208649,
      "reward_std": 0.09062027186155319,
      "reward_std_all": 0.09332232177257538,
      "rewards-max/forensics_iou_reward": 0.951862096786499,
      "rewards-max/forensics_strict_edge_reward": 0.3924011290073395,
      "rewards-max/forensics_tiou_reward": 0.931995302438736,
      "rewards/forensics_iou_reward": 0.5310648381710052,
      "rewards/forensics_strict_edge_reward": 0.034087641164660454,
      "rewards/forensics_tiou_reward": 0.4861298054456711,
      "step": 694
    },
    {
      "completion_length": 22.796875,
      "epoch": 2.1321044546851,
      "grad_norm": 6.0869293212890625,
      "kl": 0.3720703125,
      "learning_rate": 4.661538461538461e-07,
      "loss": 0.0202,
      "reward": 0.753446102142334,
      "reward_std": 0.06408285908401012,
      "reward_std_all": 0.06092364341020584,
      "rewards-max/forensics_iou_reward": 0.94954514503479,
      "rewards-max/forensics_strict_edge_reward": 0.5491165816783905,
      "rewards-max/forensics_tiou_reward": 0.9273973703384399,
      "rewards/forensics_iou_reward": 0.6172256469726562,
      "rewards/forensics_strict_edge_reward": 0.053126681596040726,
      "rewards/forensics_tiou_reward": 0.5671513080596924,
      "step": 695
    },
    {
      "completion_length": 25.359375,
      "epoch": 2.1351766513056836,
      "grad_norm": 5.327826023101807,
      "kl": 0.4560546875,
      "learning_rate": 4.653846153846154e-07,
      "loss": 0.0768,
      "reward": 0.7142784893512726,
      "reward_std": 0.0789017453789711,
      "reward_std_all": 0.08748099580407143,
      "rewards-max/forensics_iou_reward": 0.973073273897171,
      "rewards-max/forensics_strict_edge_reward": 0.5599738955497742,
      "rewards-max/forensics_tiou_reward": 0.9680635631084442,
      "rewards/forensics_iou_reward": 0.5792425870895386,
      "rewards/forensics_strict_edge_reward": 0.048304131254553795,
      "rewards/forensics_tiou_reward": 0.5314480364322662,
      "step": 696
    },
    {
      "completion_length": 25.15625,
      "epoch": 2.1382488479262673,
      "grad_norm": 6.4038920402526855,
      "kl": 0.3515625,
      "learning_rate": 4.646153846153846e-07,
      "loss": 0.002,
      "reward": 0.7143445312976837,
      "reward_std": 0.08993944898247719,
      "reward_std_all": 0.08948870375752449,
      "rewards-max/forensics_iou_reward": 0.9811513423919678,
      "rewards-max/forensics_strict_edge_reward": 0.7065972685813904,
      "rewards-max/forensics_tiou_reward": 0.9742233455181122,
      "rewards/forensics_iou_reward": 0.6432749629020691,
      "rewards/forensics_strict_edge_reward": 0.08800134062767029,
      "rewards/forensics_tiou_reward": 0.5987033843994141,
      "step": 697
    },
    {
      "completion_length": 23.5,
      "epoch": 2.141321044546851,
      "grad_norm": 5.8507280349731445,
      "kl": 0.4716796875,
      "learning_rate": 4.638461538461538e-07,
      "loss": 0.1878,
      "reward": 0.6533415615558624,
      "reward_std": 0.07589292153716087,
      "reward_std_all": 0.11320817843079567,
      "rewards-max/forensics_iou_reward": 0.9839887619018555,
      "rewards-max/forensics_strict_edge_reward": 0.6555203795433044,
      "rewards-max/forensics_tiou_reward": 0.9773053824901581,
      "rewards/forensics_iou_reward": 0.5566313415765762,
      "rewards/forensics_strict_edge_reward": 0.063164621591568,
      "rewards/forensics_tiou_reward": 0.5037341713905334,
      "step": 698
    },
    {
      "completion_length": 27.765625,
      "epoch": 2.144393241167435,
      "grad_norm": 5.297074317932129,
      "kl": 0.3916015625,
      "learning_rate": 4.6307692307692303e-07,
      "loss": 0.0857,
      "reward": 0.7618860602378845,
      "reward_std": 0.03712731692939997,
      "reward_std_all": 0.04423963278532028,
      "rewards-max/forensics_iou_reward": 0.983187198638916,
      "rewards-max/forensics_strict_edge_reward": 0.7419291436672211,
      "rewards-max/forensics_tiou_reward": 0.9789750576019287,
      "rewards/forensics_iou_reward": 0.6654049158096313,
      "rewards/forensics_strict_edge_reward": 0.049468742683529854,
      "rewards/forensics_tiou_reward": 0.6117617785930634,
      "step": 699
    },
    {
      "completion_length": 23.734375,
      "epoch": 2.1474654377880182,
      "grad_norm": 5.203183174133301,
      "kl": 0.3984375,
      "learning_rate": 4.623076923076923e-07,
      "loss": 0.0664,
      "reward": 0.7213658094406128,
      "reward_std": 0.08960385993123055,
      "reward_std_all": 0.08686240389943123,
      "rewards-max/forensics_iou_reward": 0.9912305474281311,
      "rewards-max/forensics_strict_edge_reward": 0.6718102097511292,
      "rewards-max/forensics_tiou_reward": 0.9885626137256622,
      "rewards/forensics_iou_reward": 0.642585813999176,
      "rewards/forensics_strict_edge_reward": 0.0710509605705738,
      "rewards/forensics_tiou_reward": 0.5962831377983093,
      "step": 700
    },
    {
      "completion_length": 22.140625,
      "epoch": 2.150537634408602,
      "grad_norm": 5.518761157989502,
      "kl": 0.40966796875,
      "learning_rate": 4.6153846153846156e-07,
      "loss": 0.0403,
      "reward": 0.7568574547767639,
      "reward_std": 0.04123270884156227,
      "reward_std_all": 0.0451742522418499,
      "rewards-max/forensics_iou_reward": 0.9816384315490723,
      "rewards-max/forensics_strict_edge_reward": 0.7101751863956451,
      "rewards-max/forensics_tiou_reward": 0.9753877222537994,
      "rewards/forensics_iou_reward": 0.6356269419193268,
      "rewards/forensics_strict_edge_reward": 0.06293412297964096,
      "rewards/forensics_tiou_reward": 0.5773345232009888,
      "step": 701
    },
    {
      "completion_length": 27.765625,
      "epoch": 2.153609831029186,
      "grad_norm": 5.6400322914123535,
      "kl": 0.5048828125,
      "learning_rate": 4.607692307692307e-07,
      "loss": 0.1162,
      "reward": 0.6886001229286194,
      "reward_std": 0.1424654982984066,
      "reward_std_all": 0.13416315242648125,
      "rewards-max/forensics_iou_reward": 0.9924414753913879,
      "rewards-max/forensics_strict_edge_reward": 0.5320693850517273,
      "rewards-max/forensics_tiou_reward": 0.9901250302791595,
      "rewards/forensics_iou_reward": 0.6182834208011627,
      "rewards/forensics_strict_edge_reward": 0.07050463184714317,
      "rewards/forensics_tiou_reward": 0.5803187489509583,
      "step": 702
    },
    {
      "completion_length": 30.34375,
      "epoch": 2.1566820276497696,
      "grad_norm": 6.2682342529296875,
      "kl": 0.330078125,
      "learning_rate": 4.6e-07,
      "loss": 0.0125,
      "reward": 0.6756006181240082,
      "reward_std": 0.09163710405118763,
      "reward_std_all": 0.1121553834527731,
      "rewards-max/forensics_iou_reward": 0.8776373863220215,
      "rewards-max/forensics_strict_edge_reward": 0.023422075901180506,
      "rewards-max/forensics_tiou_reward": 0.8262804746627808,
      "rewards/forensics_iou_reward": 0.5625511407852173,
      "rewards/forensics_strict_edge_reward": 0.0034041407488984987,
      "rewards/forensics_tiou_reward": 0.5088230073451996,
      "step": 703
    },
    {
      "completion_length": 24.671875,
      "epoch": 2.1597542242703534,
      "grad_norm": 6.660946846008301,
      "kl": 0.388671875,
      "learning_rate": 4.592307692307692e-07,
      "loss": 0.0263,
      "reward": 0.7602174282073975,
      "reward_std": 0.039466314017772675,
      "reward_std_all": 0.0424779299646616,
      "rewards-max/forensics_iou_reward": 0.9818494617938995,
      "rewards-max/forensics_strict_edge_reward": 0.5394537225365639,
      "rewards-max/forensics_tiou_reward": 0.9694373607635498,
      "rewards/forensics_iou_reward": 0.628024160861969,
      "rewards/forensics_strict_edge_reward": 0.03821983560919762,
      "rewards/forensics_tiou_reward": 0.5742130875587463,
      "step": 704
    },
    {
      "completion_length": 25.90625,
      "epoch": 2.162826420890937,
      "grad_norm": 6.880037307739258,
      "kl": 0.392578125,
      "learning_rate": 4.5846153846153847e-07,
      "loss": 0.0877,
      "reward": 0.7333402931690216,
      "reward_std": 0.08405398204922676,
      "reward_std_all": 0.08005104586482048,
      "rewards-max/forensics_iou_reward": 0.9820517599582672,
      "rewards-max/forensics_strict_edge_reward": 0.5221183001995087,
      "rewards-max/forensics_tiou_reward": 0.974564403295517,
      "rewards/forensics_iou_reward": 0.6395117938518524,
      "rewards/forensics_strict_edge_reward": 0.06887045502662659,
      "rewards/forensics_tiou_reward": 0.5844515860080719,
      "step": 705
    },
    {
      "completion_length": 24.0,
      "epoch": 2.165898617511521,
      "grad_norm": 5.853061199188232,
      "kl": 0.3955078125,
      "learning_rate": 4.5769230769230763e-07,
      "loss": 0.0685,
      "reward": 0.729112833738327,
      "reward_std": 0.08474216982722282,
      "reward_std_all": 0.08278154954314232,
      "rewards-max/forensics_iou_reward": 0.9847889244556427,
      "rewards-max/forensics_strict_edge_reward": 0.49006423354148865,
      "rewards-max/forensics_tiou_reward": 0.9767722487449646,
      "rewards/forensics_iou_reward": 0.6078842282295227,
      "rewards/forensics_strict_edge_reward": 0.03691352903842926,
      "rewards/forensics_tiou_reward": 0.5450984835624695,
      "step": 706
    },
    {
      "completion_length": 25.390625,
      "epoch": 2.1689708141321047,
      "grad_norm": 5.700376987457275,
      "kl": 0.3330078125,
      "learning_rate": 4.569230769230769e-07,
      "loss": 0.108,
      "reward": 0.7260954976081848,
      "reward_std": 0.08292336761951447,
      "reward_std_all": 0.08268126100301743,
      "rewards-max/forensics_iou_reward": 0.9843177199363708,
      "rewards-max/forensics_strict_edge_reward": 0.48699167370796204,
      "rewards-max/forensics_tiou_reward": 0.9764152765274048,
      "rewards/forensics_iou_reward": 0.613813042640686,
      "rewards/forensics_strict_edge_reward": 0.05331077426671982,
      "rewards/forensics_tiou_reward": 0.5683003962039948,
      "step": 707
    },
    {
      "completion_length": 26.75,
      "epoch": 2.172043010752688,
      "grad_norm": 4.992285251617432,
      "kl": 0.359375,
      "learning_rate": 4.561538461538461e-07,
      "loss": -0.0206,
      "reward": 0.7478355765342712,
      "reward_std": 0.07744137570261955,
      "reward_std_all": 0.07178647071123123,
      "rewards-max/forensics_iou_reward": 0.9560102224349976,
      "rewards-max/forensics_strict_edge_reward": 0.5185712575912476,
      "rewards-max/forensics_tiou_reward": 0.9385807812213898,
      "rewards/forensics_iou_reward": 0.6558705270290375,
      "rewards/forensics_strict_edge_reward": 0.03943127207458019,
      "rewards/forensics_tiou_reward": 0.5999763906002045,
      "step": 708
    },
    {
      "completion_length": 25.90625,
      "epoch": 2.175115207373272,
      "grad_norm": 6.170876979827881,
      "kl": 0.474609375,
      "learning_rate": 4.553846153846154e-07,
      "loss": -0.0105,
      "reward": 0.751001387834549,
      "reward_std": 0.0723661445081234,
      "reward_std_all": 0.06758714094758034,
      "rewards-max/forensics_iou_reward": 0.9311220049858093,
      "rewards-max/forensics_strict_edge_reward": 0.41929639875888824,
      "rewards-max/forensics_tiou_reward": 0.9097822606563568,
      "rewards/forensics_iou_reward": 0.6038799583911896,
      "rewards/forensics_strict_edge_reward": 0.04234742559492588,
      "rewards/forensics_tiou_reward": 0.5401411354541779,
      "step": 709
    },
    {
      "completion_length": 24.421875,
      "epoch": 2.1781874039938556,
      "grad_norm": 6.181998252868652,
      "kl": 0.3408203125,
      "learning_rate": 4.546153846153846e-07,
      "loss": 0.073,
      "reward": 0.7519322633743286,
      "reward_std": 0.08152194693684578,
      "reward_std_all": 0.07373857870697975,
      "rewards-max/forensics_iou_reward": 0.987386554479599,
      "rewards-max/forensics_strict_edge_reward": 0.709127277135849,
      "rewards-max/forensics_tiou_reward": 0.9831078350543976,
      "rewards/forensics_iou_reward": 0.6349080502986908,
      "rewards/forensics_strict_edge_reward": 0.07520831376314163,
      "rewards/forensics_tiou_reward": 0.5922009646892548,
      "step": 710
    },
    {
      "completion_length": 24.484375,
      "epoch": 2.1812596006144394,
      "grad_norm": 6.103133201599121,
      "kl": 0.3173828125,
      "learning_rate": 4.538461538461538e-07,
      "loss": 0.0107,
      "reward": 0.7492916882038116,
      "reward_std": 0.045135414926335216,
      "reward_std_all": 0.049471001606434584,
      "rewards-max/forensics_iou_reward": 0.9811645448207855,
      "rewards-max/forensics_strict_edge_reward": 0.5243366360664368,
      "rewards-max/forensics_tiou_reward": 0.9766192138195038,
      "rewards/forensics_iou_reward": 0.6940382719039917,
      "rewards/forensics_strict_edge_reward": 0.06469651125371456,
      "rewards/forensics_tiou_reward": 0.6543197631835938,
      "step": 711
    },
    {
      "completion_length": 26.890625,
      "epoch": 2.184331797235023,
      "grad_norm": 7.105763912200928,
      "kl": 0.3876953125,
      "learning_rate": 4.5307692307692307e-07,
      "loss": 0.0542,
      "reward": 0.6689321398735046,
      "reward_std": 0.053721094503998756,
      "reward_std_all": 0.09815587475895882,
      "rewards-max/forensics_iou_reward": 0.9753756821155548,
      "rewards-max/forensics_strict_edge_reward": 0.3619372546672821,
      "rewards-max/forensics_tiou_reward": 0.9664216935634613,
      "rewards/forensics_iou_reward": 0.5694465339183807,
      "rewards/forensics_strict_edge_reward": 0.03943450190126896,
      "rewards/forensics_tiou_reward": 0.5295359790325165,
      "step": 712
    },
    {
      "completion_length": 21.59375,
      "epoch": 2.187403993855607,
      "grad_norm": 7.3588690757751465,
      "kl": 0.408203125,
      "learning_rate": 4.523076923076923e-07,
      "loss": -0.0624,
      "reward": 0.7524601817131042,
      "reward_std": 0.055721936747431755,
      "reward_std_all": 0.055992305278778076,
      "rewards-max/forensics_iou_reward": 0.9745141267776489,
      "rewards-max/forensics_strict_edge_reward": 0.6716593652963638,
      "rewards-max/forensics_tiou_reward": 0.9658465385437012,
      "rewards/forensics_iou_reward": 0.6220953464508057,
      "rewards/forensics_strict_edge_reward": 0.08516399934887886,
      "rewards/forensics_tiou_reward": 0.5671118795871735,
      "step": 713
    },
    {
      "completion_length": 25.046875,
      "epoch": 2.1904761904761907,
      "grad_norm": 5.458496570587158,
      "kl": 0.421875,
      "learning_rate": 4.5153846153846155e-07,
      "loss": -0.034,
      "reward": 0.7912976145744324,
      "reward_std": 0.023844701005145907,
      "reward_std_all": 0.022577370749786496,
      "rewards-max/forensics_iou_reward": 0.9850897789001465,
      "rewards-max/forensics_strict_edge_reward": 0.6820428818464279,
      "rewards-max/forensics_tiou_reward": 0.9796872138977051,
      "rewards/forensics_iou_reward": 0.6784518361091614,
      "rewards/forensics_strict_edge_reward": 0.08487404510378838,
      "rewards/forensics_tiou_reward": 0.6269824504852295,
      "step": 714
    },
    {
      "completion_length": 25.109375,
      "epoch": 2.193548387096774,
      "grad_norm": 7.184045791625977,
      "kl": 0.38671875,
      "learning_rate": 4.507692307692307e-07,
      "loss": 0.0429,
      "reward": 0.7106894850730896,
      "reward_std": 0.09229669719934464,
      "reward_std_all": 0.09142928570508957,
      "rewards-max/forensics_iou_reward": 0.9703191816806793,
      "rewards-max/forensics_strict_edge_reward": 0.6469956934452057,
      "rewards-max/forensics_tiou_reward": 0.959702342748642,
      "rewards/forensics_iou_reward": 0.5442575514316559,
      "rewards/forensics_strict_edge_reward": 0.06586900539696217,
      "rewards/forensics_tiou_reward": 0.492782786488533,
      "step": 715
    },
    {
      "completion_length": 28.171875,
      "epoch": 2.196620583717358,
      "grad_norm": 5.145171642303467,
      "kl": 0.4072265625,
      "learning_rate": 4.5e-07,
      "loss": 0.0994,
      "reward": 0.7725936770439148,
      "reward_std": 0.039881469681859016,
      "reward_std_all": 0.03885582461953163,
      "rewards-max/forensics_iou_reward": 0.9626824259757996,
      "rewards-max/forensics_strict_edge_reward": 0.2211335375905037,
      "rewards-max/forensics_tiou_reward": 0.9462741911411285,
      "rewards/forensics_iou_reward": 0.6201644539833069,
      "rewards/forensics_strict_edge_reward": 0.027874240186065435,
      "rewards/forensics_tiou_reward": 0.5665372312068939,
      "step": 716
    },
    {
      "completion_length": 27.953125,
      "epoch": 2.1996927803379416,
      "grad_norm": 5.561338424682617,
      "kl": 0.39453125,
      "learning_rate": 4.492307692307692e-07,
      "loss": -0.0109,
      "reward": 0.7527249455451965,
      "reward_std": 0.041678338311612606,
      "reward_std_all": 0.04637328814715147,
      "rewards-max/forensics_iou_reward": 0.9906839728355408,
      "rewards-max/forensics_strict_edge_reward": 0.5072436034679413,
      "rewards-max/forensics_tiou_reward": 0.9867891669273376,
      "rewards/forensics_iou_reward": 0.7207341492176056,
      "rewards/forensics_strict_edge_reward": 0.06114416755735874,
      "rewards/forensics_tiou_reward": 0.6729492545127869,
      "step": 717
    },
    {
      "completion_length": 25.59375,
      "epoch": 2.2027649769585254,
      "grad_norm": 6.619394779205322,
      "kl": 0.3173828125,
      "learning_rate": 4.4846153846153846e-07,
      "loss": 0.023,
      "reward": 0.7287341952323914,
      "reward_std": 0.09194807708263397,
      "reward_std_all": 0.09128198027610779,
      "rewards-max/forensics_iou_reward": 0.9553361237049103,
      "rewards-max/forensics_strict_edge_reward": 0.47400766611099243,
      "rewards-max/forensics_tiou_reward": 0.9438152313232422,
      "rewards/forensics_iou_reward": 0.6130984723567963,
      "rewards/forensics_strict_edge_reward": 0.04802101477980614,
      "rewards/forensics_tiou_reward": 0.5490699410438538,
      "step": 718
    },
    {
      "completion_length": 23.9375,
      "epoch": 2.205837173579109,
      "grad_norm": 5.539004802703857,
      "kl": 0.3798828125,
      "learning_rate": 4.476923076923076e-07,
      "loss": 0.1416,
      "reward": 0.7570146024227142,
      "reward_std": 0.08111289143562317,
      "reward_std_all": 0.07262303680181503,
      "rewards-max/forensics_iou_reward": 0.9457232356071472,
      "rewards-max/forensics_strict_edge_reward": 0.4862638860940933,
      "rewards-max/forensics_tiou_reward": 0.9335415065288544,
      "rewards/forensics_iou_reward": 0.6679219603538513,
      "rewards/forensics_strict_edge_reward": 0.03790730517357588,
      "rewards/forensics_tiou_reward": 0.6061467230319977,
      "step": 719
    },
    {
      "completion_length": 26.71875,
      "epoch": 2.208909370199693,
      "grad_norm": 6.559072494506836,
      "kl": 0.3515625,
      "learning_rate": 4.469230769230769e-07,
      "loss": -0.1016,
      "reward": 0.8021098375320435,
      "reward_std": 0.0076288990676403046,
      "reward_std_all": 0.008568446850404143,
      "rewards-max/forensics_iou_reward": 0.9768005013465881,
      "rewards-max/forensics_strict_edge_reward": 0.4063575565814972,
      "rewards-max/forensics_tiou_reward": 0.9599569141864777,
      "rewards/forensics_iou_reward": 0.7837197780609131,
      "rewards/forensics_strict_edge_reward": 0.05574073549360037,
      "rewards/forensics_tiou_reward": 0.7426222562789917,
      "step": 720
    },
    {
      "completion_length": 23.046875,
      "epoch": 2.2119815668202767,
      "grad_norm": 7.9215216636657715,
      "kl": 0.36328125,
      "learning_rate": 4.4615384615384615e-07,
      "loss": 0.1292,
      "reward": 0.7753323018550873,
      "reward_std": 0.04536166414618492,
      "reward_std_all": 0.04232491087168455,
      "rewards-max/forensics_iou_reward": 0.963224470615387,
      "rewards-max/forensics_strict_edge_reward": 0.466240257024765,
      "rewards-max/forensics_tiou_reward": 0.9523400664329529,
      "rewards/forensics_iou_reward": 0.7034306824207306,
      "rewards/forensics_strict_edge_reward": 0.05162061657756567,
      "rewards/forensics_tiou_reward": 0.6448315382003784,
      "step": 721
    },
    {
      "completion_length": 24.96875,
      "epoch": 2.21505376344086,
      "grad_norm": 7.663481712341309,
      "kl": 0.443359375,
      "learning_rate": 4.4538461538461537e-07,
      "loss": 0.0676,
      "reward": 0.7642995715141296,
      "reward_std": 0.052967846393585205,
      "reward_std_all": 0.05026434734463692,
      "rewards-max/forensics_iou_reward": 0.9926552176475525,
      "rewards-max/forensics_strict_edge_reward": 0.4599933922290802,
      "rewards-max/forensics_tiou_reward": 0.989637941122055,
      "rewards/forensics_iou_reward": 0.6770839989185333,
      "rewards/forensics_strict_edge_reward": 0.08370490744709969,
      "rewards/forensics_tiou_reward": 0.6246164739131927,
      "step": 722
    },
    {
      "completion_length": 27.765625,
      "epoch": 2.218125960061444,
      "grad_norm": 6.552161693572998,
      "kl": 0.392578125,
      "learning_rate": 4.4461538461538463e-07,
      "loss": 0.0926,
      "reward": 0.692083865404129,
      "reward_std": 0.13154777884483337,
      "reward_std_all": 0.1248064935207367,
      "rewards-max/forensics_iou_reward": 0.9885260760784149,
      "rewards-max/forensics_strict_edge_reward": 0.4461440145969391,
      "rewards-max/forensics_tiou_reward": 0.9859954416751862,
      "rewards/forensics_iou_reward": 0.6014841198921204,
      "rewards/forensics_strict_edge_reward": 0.03363912831991911,
      "rewards/forensics_tiou_reward": 0.5561446845531464,
      "step": 723
    },
    {
      "completion_length": 23.96875,
      "epoch": 2.2211981566820276,
      "grad_norm": 6.715676307678223,
      "kl": 0.48388671875,
      "learning_rate": 4.438461538461538e-07,
      "loss": 0.0783,
      "reward": 0.7826258540153503,
      "reward_std": 0.03871992928907275,
      "reward_std_all": 0.03598335664719343,
      "rewards-max/forensics_iou_reward": 0.9717262089252472,
      "rewards-max/forensics_strict_edge_reward": 0.6515887677669525,
      "rewards-max/forensics_tiou_reward": 0.9663626253604889,
      "rewards/forensics_iou_reward": 0.6715875566005707,
      "rewards/forensics_strict_edge_reward": 0.05861877091228962,
      "rewards/forensics_tiou_reward": 0.6221953630447388,
      "step": 724
    },
    {
      "completion_length": 28.71875,
      "epoch": 2.2242703533026114,
      "grad_norm": 7.187520980834961,
      "kl": 0.3544921875,
      "learning_rate": 4.4307692307692306e-07,
      "loss": 0.0559,
      "reward": 0.7295376658439636,
      "reward_std": 0.050346434116363525,
      "reward_std_all": 0.06574265658855438,
      "rewards-max/forensics_iou_reward": 0.9654845297336578,
      "rewards-max/forensics_strict_edge_reward": 0.3610784709453583,
      "rewards-max/forensics_tiou_reward": 0.951076090335846,
      "rewards/forensics_iou_reward": 0.6333847939968109,
      "rewards/forensics_strict_edge_reward": 0.0470631904900074,
      "rewards/forensics_tiou_reward": 0.5850043594837189,
      "step": 725
    },
    {
      "completion_length": 24.3125,
      "epoch": 2.227342549923195,
      "grad_norm": 6.3125834465026855,
      "kl": 0.396484375,
      "learning_rate": 4.423076923076923e-07,
      "loss": 0.0222,
      "reward": 0.7752902209758759,
      "reward_std": 0.05375943426042795,
      "reward_std_all": 0.048395607620477676,
      "rewards-max/forensics_iou_reward": 0.9781668484210968,
      "rewards-max/forensics_strict_edge_reward": 0.5770299434661865,
      "rewards-max/forensics_tiou_reward": 0.9691416323184967,
      "rewards/forensics_iou_reward": 0.7260697185993195,
      "rewards/forensics_strict_edge_reward": 0.07304645515978336,
      "rewards/forensics_tiou_reward": 0.6848708391189575,
      "step": 726
    },
    {
      "completion_length": 25.609375,
      "epoch": 2.230414746543779,
      "grad_norm": 5.686318397521973,
      "kl": 0.373046875,
      "learning_rate": 4.4153846153846154e-07,
      "loss": 0.0634,
      "reward": 0.7535757720470428,
      "reward_std": 0.057451240718364716,
      "reward_std_all": 0.055980670265853405,
      "rewards-max/forensics_iou_reward": 0.9891090095043182,
      "rewards-max/forensics_strict_edge_reward": 0.3527763783931732,
      "rewards-max/forensics_tiou_reward": 0.9820838570594788,
      "rewards/forensics_iou_reward": 0.6841461658477783,
      "rewards/forensics_strict_edge_reward": 0.04171494580805302,
      "rewards/forensics_tiou_reward": 0.6327199041843414,
      "step": 727
    },
    {
      "completion_length": 25.0625,
      "epoch": 2.2334869431643627,
      "grad_norm": 7.658685684204102,
      "kl": 0.4130859375,
      "learning_rate": 4.407692307692307e-07,
      "loss": 0.031,
      "reward": 0.6420654058456421,
      "reward_std": 0.07070502080023289,
      "reward_std_all": 0.1200304888188839,
      "rewards-max/forensics_iou_reward": 0.9752777218818665,
      "rewards-max/forensics_strict_edge_reward": 0.2015906125307083,
      "rewards-max/forensics_tiou_reward": 0.9571651816368103,
      "rewards/forensics_iou_reward": 0.49327221512794495,
      "rewards/forensics_strict_edge_reward": 0.020807284861803055,
      "rewards/forensics_tiou_reward": 0.45079895853996277,
      "step": 728
    },
    {
      "completion_length": 26.125,
      "epoch": 2.236559139784946,
      "grad_norm": 11.91535758972168,
      "kl": 0.4052734375,
      "learning_rate": 4.3999999999999997e-07,
      "loss": 0.0775,
      "reward": 0.6891575455665588,
      "reward_std": 0.1301991418004036,
      "reward_std_all": 0.12401849031448364,
      "rewards-max/forensics_iou_reward": 0.945720374584198,
      "rewards-max/forensics_strict_edge_reward": 0.5435085296630859,
      "rewards-max/forensics_tiou_reward": 0.930403083562851,
      "rewards/forensics_iou_reward": 0.544548511505127,
      "rewards/forensics_strict_edge_reward": 0.07404318824410439,
      "rewards/forensics_tiou_reward": 0.5068837851285934,
      "step": 729
    },
    {
      "completion_length": 28.625,
      "epoch": 2.23963133640553,
      "grad_norm": 6.782750606536865,
      "kl": 0.34765625,
      "learning_rate": 4.3923076923076924e-07,
      "loss": 0.0998,
      "reward": 0.7455032169818878,
      "reward_std": 0.06838432652875781,
      "reward_std_all": 0.0664125308394432,
      "rewards-max/forensics_iou_reward": 0.8946205079555511,
      "rewards-max/forensics_strict_edge_reward": 0.1543818861246109,
      "rewards-max/forensics_tiou_reward": 0.8723460733890533,
      "rewards/forensics_iou_reward": 0.5869577527046204,
      "rewards/forensics_strict_edge_reward": 0.01997220516204834,
      "rewards/forensics_tiou_reward": 0.5305507779121399,
      "step": 730
    },
    {
      "completion_length": 23.765625,
      "epoch": 2.2427035330261136,
      "grad_norm": 8.508056640625,
      "kl": 0.564453125,
      "learning_rate": 4.3846153846153845e-07,
      "loss": 0.0529,
      "reward": 0.7766202390193939,
      "reward_std": 0.03472255356609821,
      "reward_std_all": 0.03432763274759054,
      "rewards-max/forensics_iou_reward": 0.9100103974342346,
      "rewards-max/forensics_strict_edge_reward": 0.2802576348185539,
      "rewards-max/forensics_tiou_reward": 0.8937669098377228,
      "rewards/forensics_iou_reward": 0.6135519444942474,
      "rewards/forensics_strict_edge_reward": 0.020796322263777256,
      "rewards/forensics_tiou_reward": 0.5569565296173096,
      "step": 731
    },
    {
      "completion_length": 24.984375,
      "epoch": 2.2457757296466974,
      "grad_norm": 6.227051258087158,
      "kl": 0.42578125,
      "learning_rate": 4.376923076923077e-07,
      "loss": 0.0554,
      "reward": 0.750638335943222,
      "reward_std": 0.07003879360854626,
      "reward_std_all": 0.06635030684992671,
      "rewards-max/forensics_iou_reward": 0.9309732913970947,
      "rewards-max/forensics_strict_edge_reward": 0.3139119893312454,
      "rewards-max/forensics_tiou_reward": 0.9160839319229126,
      "rewards/forensics_iou_reward": 0.6518891453742981,
      "rewards/forensics_strict_edge_reward": 0.027643423527479172,
      "rewards/forensics_tiou_reward": 0.6021182537078857,
      "step": 732
    },
    {
      "completion_length": 27.25,
      "epoch": 2.248847926267281,
      "grad_norm": 6.554233074188232,
      "kl": 0.4091796875,
      "learning_rate": 4.369230769230769e-07,
      "loss": -0.0284,
      "reward": 0.6894166171550751,
      "reward_std": 0.06698403880000114,
      "reward_std_all": 0.09188620187342167,
      "rewards-max/forensics_iou_reward": 0.9819356203079224,
      "rewards-max/forensics_strict_edge_reward": 0.4905853122472763,
      "rewards-max/forensics_tiou_reward": 0.9745037257671356,
      "rewards/forensics_iou_reward": 0.5960884392261505,
      "rewards/forensics_strict_edge_reward": 0.05748065002262592,
      "rewards/forensics_tiou_reward": 0.5552616119384766,
      "step": 733
    },
    {
      "completion_length": 25.375,
      "epoch": 2.251920122887865,
      "grad_norm": 6.5185956954956055,
      "kl": 0.4755859375,
      "learning_rate": 4.3615384615384614e-07,
      "loss": 0.0555,
      "reward": 0.7240894734859467,
      "reward_std": 0.06434610299766064,
      "reward_std_all": 0.07244749739766121,
      "rewards-max/forensics_iou_reward": 0.9940574169158936,
      "rewards-max/forensics_strict_edge_reward": 0.6273068785667419,
      "rewards-max/forensics_tiou_reward": 0.9919559359550476,
      "rewards/forensics_iou_reward": 0.6236771941184998,
      "rewards/forensics_strict_edge_reward": 0.07702705264091492,
      "rewards/forensics_tiou_reward": 0.582769364118576,
      "step": 734
    },
    {
      "completion_length": 27.109375,
      "epoch": 2.2549923195084487,
      "grad_norm": 9.18233585357666,
      "kl": 0.4453125,
      "learning_rate": 4.3538461538461536e-07,
      "loss": -0.1049,
      "reward": 0.6991877555847168,
      "reward_std": 0.03737770183943212,
      "reward_std_all": 0.06982531026005745,
      "rewards-max/forensics_iou_reward": 0.9817768931388855,
      "rewards-max/forensics_strict_edge_reward": 0.36234288662672043,
      "rewards-max/forensics_tiou_reward": 0.9765099287033081,
      "rewards/forensics_iou_reward": 0.6140537858009338,
      "rewards/forensics_strict_edge_reward": 0.03257198119536042,
      "rewards/forensics_tiou_reward": 0.57322958111763,
      "step": 735
    },
    {
      "completion_length": 31.1875,
      "epoch": 2.258064516129032,
      "grad_norm": 7.468700408935547,
      "kl": 0.400390625,
      "learning_rate": 4.346153846153846e-07,
      "loss": 0.0953,
      "reward": 0.7438145577907562,
      "reward_std": 0.027946973219513893,
      "reward_std_all": 0.046113042160868645,
      "rewards-max/forensics_iou_reward": 0.9695252180099487,
      "rewards-max/forensics_strict_edge_reward": 0.4899376779794693,
      "rewards-max/forensics_tiou_reward": 0.9597388207912445,
      "rewards/forensics_iou_reward": 0.6904430389404297,
      "rewards/forensics_strict_edge_reward": 0.045814476907253265,
      "rewards/forensics_tiou_reward": 0.6479495167732239,
      "step": 736
    },
    {
      "completion_length": 27.421875,
      "epoch": 2.261136712749616,
      "grad_norm": 5.714119911193848,
      "kl": 0.556640625,
      "learning_rate": 4.338461538461538e-07,
      "loss": -0.0795,
      "reward": 0.7859080135822296,
      "reward_std": 0.035793447867035866,
      "reward_std_all": 0.03285432327538729,
      "rewards-max/forensics_iou_reward": 0.9954627454280853,
      "rewards-max/forensics_strict_edge_reward": 0.6017574965953827,
      "rewards-max/forensics_tiou_reward": 0.992580235004425,
      "rewards/forensics_iou_reward": 0.7626115381717682,
      "rewards/forensics_strict_edge_reward": 0.08737242594361305,
      "rewards/forensics_tiou_reward": 0.7160248458385468,
      "step": 737
    },
    {
      "completion_length": 24.203125,
      "epoch": 2.2642089093701996,
      "grad_norm": 7.5723114013671875,
      "kl": 0.4462890625,
      "learning_rate": 4.3307692307692305e-07,
      "loss": 0.1144,
      "reward": 0.7001951336860657,
      "reward_std": 0.1114581860601902,
      "reward_std_all": 0.10754287242889404,
      "rewards-max/forensics_iou_reward": 0.9595238268375397,
      "rewards-max/forensics_strict_edge_reward": 0.5621400475502014,
      "rewards-max/forensics_tiou_reward": 0.9471771121025085,
      "rewards/forensics_iou_reward": 0.6003568768501282,
      "rewards/forensics_strict_edge_reward": 0.06771476566791534,
      "rewards/forensics_tiou_reward": 0.5559262037277222,
      "step": 738
    },
    {
      "completion_length": 23.640625,
      "epoch": 2.2672811059907834,
      "grad_norm": 6.90202522277832,
      "kl": 0.3271484375,
      "learning_rate": 4.3230769230769227e-07,
      "loss": 0.0821,
      "reward": 0.7470619678497314,
      "reward_std": 0.08208617195487022,
      "reward_std_all": 0.07563608884811401,
      "rewards-max/forensics_iou_reward": 0.9934137165546417,
      "rewards-max/forensics_strict_edge_reward": 0.7770941257476807,
      "rewards-max/forensics_tiou_reward": 0.9900646507740021,
      "rewards/forensics_iou_reward": 0.6296913921833038,
      "rewards/forensics_strict_edge_reward": 0.08426877856254578,
      "rewards/forensics_tiou_reward": 0.5784001350402832,
      "step": 739
    },
    {
      "completion_length": 27.109375,
      "epoch": 2.270353302611367,
      "grad_norm": 5.424894332885742,
      "kl": 0.427734375,
      "learning_rate": 4.3153846153846153e-07,
      "loss": 0.1045,
      "reward": 0.7239365875720978,
      "reward_std": 0.08841346576809883,
      "reward_std_all": 0.08689495734870434,
      "rewards-max/forensics_iou_reward": 0.8432471752166748,
      "rewards-max/forensics_strict_edge_reward": 0.19012529961764812,
      "rewards-max/forensics_tiou_reward": 0.8206171691417694,
      "rewards/forensics_iou_reward": 0.4763312339782715,
      "rewards/forensics_strict_edge_reward": 0.010194429545663297,
      "rewards/forensics_tiou_reward": 0.4078325778245926,
      "step": 740
    },
    {
      "completion_length": 27.34375,
      "epoch": 2.273425499231951,
      "grad_norm": 6.295006275177002,
      "kl": 0.4921875,
      "learning_rate": 4.307692307692308e-07,
      "loss": 0.0985,
      "reward": 0.7643300890922546,
      "reward_std": 0.05035174824297428,
      "reward_std_all": 0.048448966816067696,
      "rewards-max/forensics_iou_reward": 0.9765227735042572,
      "rewards-max/forensics_strict_edge_reward": 0.6892678737640381,
      "rewards-max/forensics_tiou_reward": 0.9709195792675018,
      "rewards/forensics_iou_reward": 0.6323167085647583,
      "rewards/forensics_strict_edge_reward": 0.06044092774391174,
      "rewards/forensics_tiou_reward": 0.5815565884113312,
      "step": 741
    },
    {
      "completion_length": 21.734375,
      "epoch": 2.2764976958525347,
      "grad_norm": 5.7434916496276855,
      "kl": 0.611328125,
      "learning_rate": 4.2999999999999996e-07,
      "loss": 0.1074,
      "reward": 0.6871061325073242,
      "reward_std": 0.0812664907425642,
      "reward_std_all": 0.10013925656676292,
      "rewards-max/forensics_iou_reward": 0.9959862232208252,
      "rewards-max/forensics_strict_edge_reward": 0.5986756533384323,
      "rewards-max/forensics_tiou_reward": 0.9950019121170044,
      "rewards/forensics_iou_reward": 0.555371031165123,
      "rewards/forensics_strict_edge_reward": 0.05284034460783005,
      "rewards/forensics_tiou_reward": 0.4942670911550522,
      "step": 742
    },
    {
      "completion_length": 27.25,
      "epoch": 2.279569892473118,
      "grad_norm": 6.209193706512451,
      "kl": 0.427734375,
      "learning_rate": 4.292307692307692e-07,
      "loss": 0.0628,
      "reward": 0.7298926413059235,
      "reward_std": 0.06898093037307262,
      "reward_std_all": 0.0738045834004879,
      "rewards-max/forensics_iou_reward": 0.9644641876220703,
      "rewards-max/forensics_strict_edge_reward": 0.364642009139061,
      "rewards-max/forensics_tiou_reward": 0.9393566846847534,
      "rewards/forensics_iou_reward": 0.6265202462673187,
      "rewards/forensics_strict_edge_reward": 0.040628667920827866,
      "rewards/forensics_tiou_reward": 0.5891439318656921,
      "step": 743
    },
    {
      "completion_length": 22.9375,
      "epoch": 2.282642089093702,
      "grad_norm": 9.076930046081543,
      "kl": 0.482421875,
      "learning_rate": 4.2846153846153844e-07,
      "loss": 0.0738,
      "reward": 0.6671183407306671,
      "reward_std": 0.10855013132095337,
      "reward_std_all": 0.12181559950113297,
      "rewards-max/forensics_iou_reward": 0.9821159243583679,
      "rewards-max/forensics_strict_edge_reward": 0.6487045586109161,
      "rewards-max/forensics_tiou_reward": 0.9757405519485474,
      "rewards/forensics_iou_reward": 0.5647064447402954,
      "rewards/forensics_strict_edge_reward": 0.07292553409934044,
      "rewards/forensics_tiou_reward": 0.5214741080999374,
      "step": 744
    },
    {
      "completion_length": 24.4375,
      "epoch": 2.2857142857142856,
      "grad_norm": 8.215060234069824,
      "kl": 0.498046875,
      "learning_rate": 4.276923076923077e-07,
      "loss": 0.1319,
      "reward": 0.7171645164489746,
      "reward_std": 0.05260521359741688,
      "reward_std_all": 0.06965003348886967,
      "rewards-max/forensics_iou_reward": 0.9714528322219849,
      "rewards-max/forensics_strict_edge_reward": 0.27133943140506744,
      "rewards-max/forensics_tiou_reward": 0.9682852029800415,
      "rewards/forensics_iou_reward": 0.6074650585651398,
      "rewards/forensics_strict_edge_reward": 0.041995180770754814,
      "rewards/forensics_tiou_reward": 0.5623068809509277,
      "step": 745
    },
    {
      "completion_length": 25.4375,
      "epoch": 2.2887864823348694,
      "grad_norm": 6.405143737792969,
      "kl": 0.375,
      "learning_rate": 4.2692307692307687e-07,
      "loss": -0.0855,
      "reward": 0.7668105959892273,
      "reward_std": 0.02561577595770359,
      "reward_std_all": 0.03155944682657719,
      "rewards-max/forensics_iou_reward": 0.9862269759178162,
      "rewards-max/forensics_strict_edge_reward": 0.6994845867156982,
      "rewards-max/forensics_tiou_reward": 0.9819486737251282,
      "rewards/forensics_iou_reward": 0.7503357827663422,
      "rewards/forensics_strict_edge_reward": 0.10016510635614395,
      "rewards/forensics_tiou_reward": 0.7132892608642578,
      "step": 746
    },
    {
      "completion_length": 24.6875,
      "epoch": 2.291858678955453,
      "grad_norm": 8.540313720703125,
      "kl": 0.3916015625,
      "learning_rate": 4.2615384615384613e-07,
      "loss": -0.0684,
      "reward": 0.7266754508018494,
      "reward_std": 0.07220531068742275,
      "reward_std_all": 0.07467521168291569,
      "rewards-max/forensics_iou_reward": 0.9898967146873474,
      "rewards-max/forensics_strict_edge_reward": 0.4749826490879059,
      "rewards-max/forensics_tiou_reward": 0.9857870638370514,
      "rewards/forensics_iou_reward": 0.6310926079750061,
      "rewards/forensics_strict_edge_reward": 0.06840064376592636,
      "rewards/forensics_tiou_reward": 0.5859372913837433,
      "step": 747
    },
    {
      "completion_length": 25.734375,
      "epoch": 2.294930875576037,
      "grad_norm": 6.131796836853027,
      "kl": 0.39453125,
      "learning_rate": 4.2538461538461535e-07,
      "loss": 0.0667,
      "reward": 0.7584161758422852,
      "reward_std": 0.08059170097112656,
      "reward_std_all": 0.07201160117983818,
      "rewards-max/forensics_iou_reward": 0.9303112328052521,
      "rewards-max/forensics_strict_edge_reward": 0.39178532361984253,
      "rewards-max/forensics_tiou_reward": 0.9102092087268829,
      "rewards/forensics_iou_reward": 0.6339536011219025,
      "rewards/forensics_strict_edge_reward": 0.055066476576030254,
      "rewards/forensics_tiou_reward": 0.5935841500759125,
      "step": 748
    },
    {
      "completion_length": 22.921875,
      "epoch": 2.2980030721966207,
      "grad_norm": 6.753705024719238,
      "kl": 0.4482421875,
      "learning_rate": 4.246153846153846e-07,
      "loss": 0.067,
      "reward": 0.7545226216316223,
      "reward_std": 0.05680788308382034,
      "reward_std_all": 0.0555108617991209,
      "rewards-max/forensics_iou_reward": 0.9739060401916504,
      "rewards-max/forensics_strict_edge_reward": 0.3274146765470505,
      "rewards-max/forensics_tiou_reward": 0.9644089639186859,
      "rewards/forensics_iou_reward": 0.6658117175102234,
      "rewards/forensics_strict_edge_reward": 0.027349725365638733,
      "rewards/forensics_tiou_reward": 0.5913524031639099,
      "step": 749
    },
    {
      "completion_length": 28.6875,
      "epoch": 2.3010752688172045,
      "grad_norm": 7.149511337280273,
      "kl": 0.576171875,
      "learning_rate": 4.2384615384615383e-07,
      "loss": -0.0204,
      "reward": 0.7236492931842804,
      "reward_std": 0.051494841231033206,
      "reward_std_all": 0.07105577830225229,
      "rewards-max/forensics_iou_reward": 0.9644717276096344,
      "rewards-max/forensics_strict_edge_reward": 0.3939525932073593,
      "rewards-max/forensics_tiou_reward": 0.9452650845050812,
      "rewards/forensics_iou_reward": 0.613616555929184,
      "rewards/forensics_strict_edge_reward": 0.048095833510160446,
      "rewards/forensics_tiou_reward": 0.5633358955383301,
      "step": 750
    },
    {
      "completion_length": 26.828125,
      "epoch": 2.3041474654377883,
      "grad_norm": 5.574870586395264,
      "kl": 0.421875,
      "learning_rate": 4.2307692307692304e-07,
      "loss": -0.0862,
      "reward": 0.7355061769485474,
      "reward_std": 0.032293134834617376,
      "reward_std_all": 0.053873246535658836,
      "rewards-max/forensics_iou_reward": 0.9735812842845917,
      "rewards-max/forensics_strict_edge_reward": 0.5221183001995087,
      "rewards-max/forensics_tiou_reward": 0.966507613658905,
      "rewards/forensics_iou_reward": 0.6340981721878052,
      "rewards/forensics_strict_edge_reward": 0.05180243030190468,
      "rewards/forensics_tiou_reward": 0.5787186324596405,
      "step": 751
    },
    {
      "completion_length": 25.1875,
      "epoch": 2.3072196620583716,
      "grad_norm": 6.463418960571289,
      "kl": 0.3427734375,
      "learning_rate": 4.223076923076923e-07,
      "loss": 0.1088,
      "reward": 0.7651747167110443,
      "reward_std": 0.04514170251786709,
      "reward_std_all": 0.044000507332384586,
      "rewards-max/forensics_iou_reward": 0.9884056746959686,
      "rewards-max/forensics_strict_edge_reward": 0.707136332988739,
      "rewards-max/forensics_tiou_reward": 0.985177218914032,
      "rewards/forensics_iou_reward": 0.6908928453922272,
      "rewards/forensics_strict_edge_reward": 0.09625421836972237,
      "rewards/forensics_tiou_reward": 0.6410420536994934,
      "step": 752
    },
    {
      "completion_length": 25.21875,
      "epoch": 2.3102918586789554,
      "grad_norm": 6.044689178466797,
      "kl": 0.4677734375,
      "learning_rate": 4.215384615384615e-07,
      "loss": -0.0464,
      "reward": 0.7340041399002075,
      "reward_std": 0.05773387255612761,
      "reward_std_all": 0.06425075931474566,
      "rewards-max/forensics_iou_reward": 0.9785398840904236,
      "rewards-max/forensics_strict_edge_reward": 0.4848052114248276,
      "rewards-max/forensics_tiou_reward": 0.9720690548419952,
      "rewards/forensics_iou_reward": 0.6467465460300446,
      "rewards/forensics_strict_edge_reward": 0.030968744307756424,
      "rewards/forensics_tiou_reward": 0.595038503408432,
      "step": 753
    },
    {
      "completion_length": 28.9375,
      "epoch": 2.313364055299539,
      "grad_norm": 4.97150993347168,
      "kl": 0.5029296875,
      "learning_rate": 4.207692307692308e-07,
      "loss": 0.0889,
      "reward": 0.7624214589595795,
      "reward_std": 0.059823885560035706,
      "reward_std_all": 0.055837979540228844,
      "rewards-max/forensics_iou_reward": 0.9779459536075592,
      "rewards-max/forensics_strict_edge_reward": 0.5547898411750793,
      "rewards-max/forensics_tiou_reward": 0.9721894264221191,
      "rewards/forensics_iou_reward": 0.6762973964214325,
      "rewards/forensics_strict_edge_reward": 0.06235189363360405,
      "rewards/forensics_tiou_reward": 0.633137047290802,
      "step": 754
    },
    {
      "completion_length": 24.21875,
      "epoch": 2.316436251920123,
      "grad_norm": 8.60204792022705,
      "kl": 0.451171875,
      "learning_rate": 4.1999999999999995e-07,
      "loss": -0.0288,
      "reward": 0.7209655940532684,
      "reward_std": 0.04081322066485882,
      "reward_std_all": 0.06369729246944189,
      "rewards-max/forensics_iou_reward": 0.9800215363502502,
      "rewards-max/forensics_strict_edge_reward": 0.6360974311828613,
      "rewards-max/forensics_tiou_reward": 0.9747499525547028,
      "rewards/forensics_iou_reward": 0.5967169404029846,
      "rewards/forensics_strict_edge_reward": 0.04642779193818569,
      "rewards/forensics_tiou_reward": 0.5337581932544708,
      "step": 755
    },
    {
      "completion_length": 26.421875,
      "epoch": 2.3195084485407067,
      "grad_norm": 5.242404937744141,
      "kl": 0.3916015625,
      "learning_rate": 4.192307692307692e-07,
      "loss": 0.012,
      "reward": 0.7785816788673401,
      "reward_std": 0.03136684279888868,
      "reward_std_all": 0.031048528850078583,
      "rewards-max/forensics_iou_reward": 0.9784263670444489,
      "rewards-max/forensics_strict_edge_reward": 0.659181222319603,
      "rewards-max/forensics_tiou_reward": 0.9719595313072205,
      "rewards/forensics_iou_reward": 0.7260168790817261,
      "rewards/forensics_strict_edge_reward": 0.0892750509083271,
      "rewards/forensics_tiou_reward": 0.6724215447902679,
      "step": 756
    },
    {
      "completion_length": 24.65625,
      "epoch": 2.3225806451612905,
      "grad_norm": 6.8992719650268555,
      "kl": 0.404296875,
      "learning_rate": 4.1846153846153843e-07,
      "loss": 0.1441,
      "reward": 0.7029363512992859,
      "reward_std": 0.10613638907670975,
      "reward_std_all": 0.10840645805001259,
      "rewards-max/forensics_iou_reward": 0.9590438902378082,
      "rewards-max/forensics_strict_edge_reward": 0.41781647503376007,
      "rewards-max/forensics_tiou_reward": 0.9439679384231567,
      "rewards/forensics_iou_reward": 0.5560435652732849,
      "rewards/forensics_strict_edge_reward": 0.0474514476954937,
      "rewards/forensics_tiou_reward": 0.508812353014946,
      "step": 757
    },
    {
      "completion_length": 22.359375,
      "epoch": 2.3256528417818743,
      "grad_norm": 6.525073528289795,
      "kl": 0.3486328125,
      "learning_rate": 4.176923076923077e-07,
      "loss": 0.125,
      "reward": 0.7254318594932556,
      "reward_std": 0.06562348082661629,
      "reward_std_all": 0.07221709936857224,
      "rewards-max/forensics_iou_reward": 0.9626623392105103,
      "rewards-max/forensics_strict_edge_reward": 0.5550253242254257,
      "rewards-max/forensics_tiou_reward": 0.9499645829200745,
      "rewards/forensics_iou_reward": 0.5922697186470032,
      "rewards/forensics_strict_edge_reward": 0.048511928878724575,
      "rewards/forensics_tiou_reward": 0.5374014377593994,
      "step": 758
    },
    {
      "completion_length": 25.703125,
      "epoch": 2.3287250384024576,
      "grad_norm": 6.886958599090576,
      "kl": 0.4296875,
      "learning_rate": 4.1692307692307686e-07,
      "loss": 0.047,
      "reward": 0.7403656542301178,
      "reward_std": 0.026703225448727608,
      "reward_std_all": 0.04817925859242678,
      "rewards-max/forensics_iou_reward": 0.9411719441413879,
      "rewards-max/forensics_strict_edge_reward": 0.37321021035313606,
      "rewards-max/forensics_tiou_reward": 0.9263417720794678,
      "rewards/forensics_iou_reward": 0.6443027853965759,
      "rewards/forensics_strict_edge_reward": 0.02758494671434164,
      "rewards/forensics_tiou_reward": 0.5856006741523743,
      "step": 759
    },
    {
      "completion_length": 24.421875,
      "epoch": 2.3317972350230414,
      "grad_norm": 6.941173553466797,
      "kl": 0.3623046875,
      "learning_rate": 4.161538461538461e-07,
      "loss": 0.0777,
      "reward": 0.7262603640556335,
      "reward_std": 0.06358984112739563,
      "reward_std_all": 0.06865935400128365,
      "rewards-max/forensics_iou_reward": 0.9499113261699677,
      "rewards-max/forensics_strict_edge_reward": 0.31121845357120037,
      "rewards-max/forensics_tiou_reward": 0.935140997171402,
      "rewards/forensics_iou_reward": 0.5617846548557281,
      "rewards/forensics_strict_edge_reward": 0.026339310919865966,
      "rewards/forensics_tiou_reward": 0.5125731229782104,
      "step": 760
    },
    {
      "completion_length": 26.34375,
      "epoch": 2.334869431643625,
      "grad_norm": 6.799370288848877,
      "kl": 0.3818359375,
      "learning_rate": 4.153846153846154e-07,
      "loss": -0.0268,
      "reward": 0.7625369429588318,
      "reward_std": 0.055376202799379826,
      "reward_std_all": 0.0525334682315588,
      "rewards-max/forensics_iou_reward": 0.9810763895511627,
      "rewards-max/forensics_strict_edge_reward": 0.41435663402080536,
      "rewards-max/forensics_tiou_reward": 0.9674349427223206,
      "rewards/forensics_iou_reward": 0.6633405983448029,
      "rewards/forensics_strict_edge_reward": 0.04284394346177578,
      "rewards/forensics_tiou_reward": 0.6148476600646973,
      "step": 761
    },
    {
      "completion_length": 29.21875,
      "epoch": 2.337941628264209,
      "grad_norm": 6.31936502456665,
      "kl": 0.537109375,
      "learning_rate": 4.146153846153846e-07,
      "loss": 0.0808,
      "reward": 0.7776410281658173,
      "reward_std": 0.048606357304379344,
      "reward_std_all": 0.044136310927569866,
      "rewards-max/forensics_iou_reward": 0.9882038533687592,
      "rewards-max/forensics_strict_edge_reward": 0.4710261672735214,
      "rewards-max/forensics_tiou_reward": 0.9845537841320038,
      "rewards/forensics_iou_reward": 0.6802886426448822,
      "rewards/forensics_strict_edge_reward": 0.05634220503270626,
      "rewards/forensics_tiou_reward": 0.6361604630947113,
      "step": 762
    },
    {
      "completion_length": 26.28125,
      "epoch": 2.3410138248847927,
      "grad_norm": 7.2669830322265625,
      "kl": 0.3974609375,
      "learning_rate": 4.1384615384615387e-07,
      "loss": 0.0326,
      "reward": 0.7845320701599121,
      "reward_std": 0.02536489348858595,
      "reward_std_all": 0.0254331910982728,
      "rewards-max/forensics_iou_reward": 0.9909860789775848,
      "rewards-max/forensics_strict_edge_reward": 0.6824378371238708,
      "rewards-max/forensics_tiou_reward": 0.9881566166877747,
      "rewards/forensics_iou_reward": 0.7382664680480957,
      "rewards/forensics_strict_edge_reward": 0.07622338086366653,
      "rewards/forensics_tiou_reward": 0.6883118450641632,
      "step": 763
    },
    {
      "completion_length": 23.78125,
      "epoch": 2.3440860215053765,
      "grad_norm": 5.74207067489624,
      "kl": 0.55859375,
      "learning_rate": 4.1307692307692303e-07,
      "loss": 0.1182,
      "reward": 0.7233707308769226,
      "reward_std": 0.07534490525722504,
      "reward_std_all": 0.08582168817520142,
      "rewards-max/forensics_iou_reward": 0.9902590215206146,
      "rewards-max/forensics_strict_edge_reward": 0.4438335746526718,
      "rewards-max/forensics_tiou_reward": 0.9850662648677826,
      "rewards/forensics_iou_reward": 0.6633629500865936,
      "rewards/forensics_strict_edge_reward": 0.044648267328739166,
      "rewards/forensics_tiou_reward": 0.616829127073288,
      "step": 764
    },
    {
      "completion_length": 23.046875,
      "epoch": 2.3471582181259603,
      "grad_norm": 32.968360900878906,
      "kl": 1.5771484375,
      "learning_rate": 4.123076923076923e-07,
      "loss": 0.1506,
      "reward": 0.7421791851520538,
      "reward_std": 0.10372928716242313,
      "reward_std_all": 0.0927351713180542,
      "rewards-max/forensics_iou_reward": 0.9132606387138367,
      "rewards-max/forensics_strict_edge_reward": 0.10209526866674423,
      "rewards-max/forensics_tiou_reward": 0.8752623200416565,
      "rewards/forensics_iou_reward": 0.6560977399349213,
      "rewards/forensics_strict_edge_reward": 0.017421803437173367,
      "rewards/forensics_tiou_reward": 0.6021546125411987,
      "step": 765
    },
    {
      "completion_length": 27.03125,
      "epoch": 2.3502304147465436,
      "grad_norm": 6.715386390686035,
      "kl": 0.359375,
      "learning_rate": 4.115384615384615e-07,
      "loss": -0.0852,
      "reward": 0.7789673209190369,
      "reward_std": 0.03645239397883415,
      "reward_std_all": 0.03484475426375866,
      "rewards-max/forensics_iou_reward": 0.9575513005256653,
      "rewards-max/forensics_strict_edge_reward": 0.6115466058254242,
      "rewards-max/forensics_tiou_reward": 0.949251264333725,
      "rewards/forensics_iou_reward": 0.6499024033546448,
      "rewards/forensics_strict_edge_reward": 0.05043062008917332,
      "rewards/forensics_tiou_reward": 0.6005806624889374,
      "step": 766
    },
    {
      "completion_length": 23.453125,
      "epoch": 2.3533026113671274,
      "grad_norm": 6.371396541595459,
      "kl": 0.361328125,
      "learning_rate": 4.107692307692308e-07,
      "loss": 0.0602,
      "reward": 0.7048138380050659,
      "reward_std": 0.09565339982509613,
      "reward_std_all": 0.10358546674251556,
      "rewards-max/forensics_iou_reward": 0.9772883653640747,
      "rewards-max/forensics_strict_edge_reward": 0.7418040186166763,
      "rewards-max/forensics_tiou_reward": 0.9712241888046265,
      "rewards/forensics_iou_reward": 0.6117483973503113,
      "rewards/forensics_strict_edge_reward": 0.0855184905230999,
      "rewards/forensics_tiou_reward": 0.5772754549980164,
      "step": 767
    },
    {
      "completion_length": 24.359375,
      "epoch": 2.356374807987711,
      "grad_norm": 6.075585842132568,
      "kl": 0.4580078125,
      "learning_rate": 4.0999999999999994e-07,
      "loss": 0.0205,
      "reward": 0.7546735107898712,
      "reward_std": 0.06893622688949108,
      "reward_std_all": 0.06410754472017288,
      "rewards-max/forensics_iou_reward": 0.9939276278018951,
      "rewards-max/forensics_strict_edge_reward": 0.45591895282268524,
      "rewards-max/forensics_tiou_reward": 0.991555780172348,
      "rewards/forensics_iou_reward": 0.640811026096344,
      "rewards/forensics_strict_edge_reward": 0.06098024547100067,
      "rewards/forensics_tiou_reward": 0.5825367867946625,
      "step": 768
    },
    {
      "completion_length": 27.109375,
      "epoch": 2.359447004608295,
      "grad_norm": 6.71072244644165,
      "kl": 0.4716796875,
      "learning_rate": 4.092307692307692e-07,
      "loss": 0.1566,
      "reward": 0.7415719032287598,
      "reward_std": 0.08286832459270954,
      "reward_std_all": 0.07768153958022594,
      "rewards-max/forensics_iou_reward": 0.9613623023033142,
      "rewards-max/forensics_strict_edge_reward": 0.40730567276477814,
      "rewards-max/forensics_tiou_reward": 0.948017418384552,
      "rewards/forensics_iou_reward": 0.647182047367096,
      "rewards/forensics_strict_edge_reward": 0.04265128634870052,
      "rewards/forensics_tiou_reward": 0.6006799340248108,
      "step": 769
    },
    {
      "completion_length": 27.78125,
      "epoch": 2.3625192012288787,
      "grad_norm": 5.1425089836120605,
      "kl": 0.412109375,
      "learning_rate": 4.0846153846153847e-07,
      "loss": 0.093,
      "reward": 0.6995870471000671,
      "reward_std": 0.09236789867281914,
      "reward_std_all": 0.09488388150930405,
      "rewards-max/forensics_iou_reward": 0.9676861464977264,
      "rewards-max/forensics_strict_edge_reward": 0.4322323799133301,
      "rewards-max/forensics_tiou_reward": 0.9587017595767975,
      "rewards/forensics_iou_reward": 0.5718411803245544,
      "rewards/forensics_strict_edge_reward": 0.03943165764212608,
      "rewards/forensics_tiou_reward": 0.5277426242828369,
      "step": 770
    },
    {
      "completion_length": 28.609375,
      "epoch": 2.3655913978494625,
      "grad_norm": 5.600930690765381,
      "kl": 0.4052734375,
      "learning_rate": 4.076923076923077e-07,
      "loss": 0.067,
      "reward": 0.7777245938777924,
      "reward_std": 0.03931836038827896,
      "reward_std_all": 0.03715123049914837,
      "rewards-max/forensics_iou_reward": 0.9898347556591034,
      "rewards-max/forensics_strict_edge_reward": 0.5500316619873047,
      "rewards-max/forensics_tiou_reward": 0.9862752854824066,
      "rewards/forensics_iou_reward": 0.6790240108966827,
      "rewards/forensics_strict_edge_reward": 0.10861701518297195,
      "rewards/forensics_tiou_reward": 0.6219380497932434,
      "step": 771
    },
    {
      "completion_length": 24.703125,
      "epoch": 2.3686635944700463,
      "grad_norm": 7.071643829345703,
      "kl": 0.4501953125,
      "learning_rate": 4.069230769230769e-07,
      "loss": 0.0752,
      "reward": 0.7426106631755829,
      "reward_std": 0.06680511310696602,
      "reward_std_all": 0.06571150943636894,
      "rewards-max/forensics_iou_reward": 0.9565202593803406,
      "rewards-max/forensics_strict_edge_reward": 0.37205515056848526,
      "rewards-max/forensics_tiou_reward": 0.9320786595344543,
      "rewards/forensics_iou_reward": 0.6369972825050354,
      "rewards/forensics_strict_edge_reward": 0.04907946474850178,
      "rewards/forensics_tiou_reward": 0.5868052244186401,
      "step": 772
    },
    {
      "completion_length": 23.75,
      "epoch": 2.3717357910906296,
      "grad_norm": 7.859052658081055,
      "kl": 0.28662109375,
      "learning_rate": 4.061538461538461e-07,
      "loss": 0.0924,
      "reward": 0.7235138118267059,
      "reward_std": 0.0970061905682087,
      "reward_std_all": 0.0919203869998455,
      "rewards-max/forensics_iou_reward": 0.9551825523376465,
      "rewards-max/forensics_strict_edge_reward": 0.3503752648830414,
      "rewards-max/forensics_tiou_reward": 0.943269282579422,
      "rewards/forensics_iou_reward": 0.5660370290279388,
      "rewards/forensics_strict_edge_reward": 0.027245139703154564,
      "rewards/forensics_tiou_reward": 0.5089370012283325,
      "step": 773
    },
    {
      "completion_length": 26.0,
      "epoch": 2.3748079877112134,
      "grad_norm": 6.1542181968688965,
      "kl": 0.369140625,
      "learning_rate": 4.053846153846154e-07,
      "loss": 0.0827,
      "reward": 0.7746516466140747,
      "reward_std": 0.0477842828258872,
      "reward_std_all": 0.04448211193084717,
      "rewards-max/forensics_iou_reward": 0.9385806620121002,
      "rewards-max/forensics_strict_edge_reward": 0.2677118256688118,
      "rewards-max/forensics_tiou_reward": 0.9271343946456909,
      "rewards/forensics_iou_reward": 0.6097593903541565,
      "rewards/forensics_strict_edge_reward": 0.021434281021356583,
      "rewards/forensics_tiou_reward": 0.5536507964134216,
      "step": 774
    },
    {
      "completion_length": 27.21875,
      "epoch": 2.377880184331797,
      "grad_norm": 5.878857135772705,
      "kl": 0.482421875,
      "learning_rate": 4.046153846153846e-07,
      "loss": 0.0223,
      "reward": 0.7716777324676514,
      "reward_std": 0.048450666945427656,
      "reward_std_all": 0.045498230028897524,
      "rewards-max/forensics_iou_reward": 0.990080714225769,
      "rewards-max/forensics_strict_edge_reward": 0.49370086938142776,
      "rewards-max/forensics_tiou_reward": 0.9843878149986267,
      "rewards/forensics_iou_reward": 0.6642528176307678,
      "rewards/forensics_strict_edge_reward": 0.046682024374604225,
      "rewards/forensics_tiou_reward": 0.6116377115249634,
      "step": 775
    },
    {
      "completion_length": 25.234375,
      "epoch": 2.380952380952381,
      "grad_norm": 6.534191608428955,
      "kl": 0.458984375,
      "learning_rate": 4.0384615384615386e-07,
      "loss": -0.0128,
      "reward": 0.7618821561336517,
      "reward_std": 0.051735895685851574,
      "reward_std_all": 0.05053982697427273,
      "rewards-max/forensics_iou_reward": 0.979885458946228,
      "rewards-max/forensics_strict_edge_reward": 0.6200447529554367,
      "rewards-max/forensics_tiou_reward": 0.9738857746124268,
      "rewards/forensics_iou_reward": 0.5847795307636261,
      "rewards/forensics_strict_edge_reward": 0.047794269397854805,
      "rewards/forensics_tiou_reward": 0.5443894267082214,
      "step": 776
    },
    {
      "completion_length": 22.953125,
      "epoch": 2.3840245775729647,
      "grad_norm": 7.028572082519531,
      "kl": 0.4716796875,
      "learning_rate": 4.03076923076923e-07,
      "loss": 0.1332,
      "reward": 0.7283078134059906,
      "reward_std": 0.09917701780796051,
      "reward_std_all": 0.09234148263931274,
      "rewards-max/forensics_iou_reward": 0.9954156577587128,
      "rewards-max/forensics_strict_edge_reward": 0.586972177028656,
      "rewards-max/forensics_tiou_reward": 0.9938764274120331,
      "rewards/forensics_iou_reward": 0.6836768090724945,
      "rewards/forensics_strict_edge_reward": 0.04793109372258186,
      "rewards/forensics_tiou_reward": 0.6360251903533936,
      "step": 777
    },
    {
      "completion_length": 25.125,
      "epoch": 2.3870967741935485,
      "grad_norm": 6.936661243438721,
      "kl": 0.591796875,
      "learning_rate": 4.023076923076923e-07,
      "loss": 0.0002,
      "reward": 0.7738454937934875,
      "reward_std": 0.03936841245740652,
      "reward_std_all": 0.03802401479333639,
      "rewards-max/forensics_iou_reward": 0.9702497124671936,
      "rewards-max/forensics_strict_edge_reward": 0.5337068438529968,
      "rewards-max/forensics_tiou_reward": 0.9602318406105042,
      "rewards/forensics_iou_reward": 0.6909664273262024,
      "rewards/forensics_strict_edge_reward": 0.05543817952275276,
      "rewards/forensics_tiou_reward": 0.6195439696311951,
      "step": 778
    },
    {
      "completion_length": 29.765625,
      "epoch": 2.3901689708141323,
      "grad_norm": 6.0256805419921875,
      "kl": 0.4228515625,
      "learning_rate": 4.015384615384615e-07,
      "loss": 0.1055,
      "reward": 0.7625948190689087,
      "reward_std": 0.05891532450914383,
      "reward_std_all": 0.055017873644828796,
      "rewards-max/forensics_iou_reward": 0.9626044034957886,
      "rewards-max/forensics_strict_edge_reward": 0.43277521431446075,
      "rewards-max/forensics_tiou_reward": 0.9534371197223663,
      "rewards/forensics_iou_reward": 0.6338987052440643,
      "rewards/forensics_strict_edge_reward": 0.04700641520321369,
      "rewards/forensics_tiou_reward": 0.5853508412837982,
      "step": 779
    },
    {
      "completion_length": 25.796875,
      "epoch": 2.3932411674347156,
      "grad_norm": 5.411549091339111,
      "kl": 0.5126953125,
      "learning_rate": 4.0076923076923077e-07,
      "loss": 0.0207,
      "reward": 0.7541415095329285,
      "reward_std": 0.05905374506255612,
      "reward_std_all": 0.05721388501115143,
      "rewards-max/forensics_iou_reward": 0.9745328426361084,
      "rewards-max/forensics_strict_edge_reward": 0.4686945676803589,
      "rewards-max/forensics_tiou_reward": 0.9680841863155365,
      "rewards/forensics_iou_reward": 0.6788561940193176,
      "rewards/forensics_strict_edge_reward": 0.06312614865601063,
      "rewards/forensics_tiou_reward": 0.6225503981113434,
      "step": 780
    },
    {
      "completion_length": 24.828125,
      "epoch": 2.3963133640552994,
      "grad_norm": 6.296042442321777,
      "kl": 0.4599609375,
      "learning_rate": 4e-07,
      "loss": 0.0823,
      "reward": 0.7559781670570374,
      "reward_std": 0.06584534049034119,
      "reward_std_all": 0.06380113959312439,
      "rewards-max/forensics_iou_reward": 0.988518238067627,
      "rewards-max/forensics_strict_edge_reward": 0.4899434596300125,
      "rewards-max/forensics_tiou_reward": 0.9825673699378967,
      "rewards/forensics_iou_reward": 0.651542603969574,
      "rewards/forensics_strict_edge_reward": 0.07592137530446053,
      "rewards/forensics_tiou_reward": 0.6073615252971649,
      "step": 781
    },
    {
      "completion_length": 26.109375,
      "epoch": 2.399385560675883,
      "grad_norm": 5.800296783447266,
      "kl": 0.423828125,
      "learning_rate": 3.992307692307692e-07,
      "loss": -0.0206,
      "reward": 0.7530806958675385,
      "reward_std": 0.056766179390251637,
      "reward_std_all": 0.055805766955018044,
      "rewards-max/forensics_iou_reward": 0.9713009297847748,
      "rewards-max/forensics_strict_edge_reward": 0.34750107303261757,
      "rewards-max/forensics_tiou_reward": 0.960954874753952,
      "rewards/forensics_iou_reward": 0.6535865664482117,
      "rewards/forensics_strict_edge_reward": 0.0349590468686074,
      "rewards/forensics_tiou_reward": 0.5968562662601471,
      "step": 782
    },
    {
      "completion_length": 25.53125,
      "epoch": 2.402457757296467,
      "grad_norm": 7.631347179412842,
      "kl": 0.478515625,
      "learning_rate": 3.9846153846153846e-07,
      "loss": 0.0171,
      "reward": 0.7438930571079254,
      "reward_std": 0.06524569541215897,
      "reward_std_all": 0.0652774553745985,
      "rewards-max/forensics_iou_reward": 0.9682061076164246,
      "rewards-max/forensics_strict_edge_reward": 0.6483443379402161,
      "rewards-max/forensics_tiou_reward": 0.9619168937206268,
      "rewards/forensics_iou_reward": 0.6380727887153625,
      "rewards/forensics_strict_edge_reward": 0.05072860233485699,
      "rewards/forensics_tiou_reward": 0.5966845154762268,
      "step": 783
    },
    {
      "completion_length": 25.03125,
      "epoch": 2.4055299539170507,
      "grad_norm": 7.2600274085998535,
      "kl": 0.498046875,
      "learning_rate": 3.976923076923077e-07,
      "loss": 0.0803,
      "reward": 0.7913739383220673,
      "reward_std": 0.014180436730384827,
      "reward_std_all": 0.01572805969044566,
      "rewards-max/forensics_iou_reward": 0.9871641993522644,
      "rewards-max/forensics_strict_edge_reward": 0.37977246940135956,
      "rewards-max/forensics_tiou_reward": 0.9827105402946472,
      "rewards/forensics_iou_reward": 0.6728078424930573,
      "rewards/forensics_strict_edge_reward": 0.05372529290616512,
      "rewards/forensics_tiou_reward": 0.6122748851776123,
      "step": 784
    },
    {
      "completion_length": 26.359375,
      "epoch": 2.4086021505376345,
      "grad_norm": 6.112404823303223,
      "kl": 0.4248046875,
      "learning_rate": 3.969230769230769e-07,
      "loss": 0.1116,
      "reward": 0.6564125120639801,
      "reward_std": 0.0782078467309475,
      "reward_std_all": 0.11533528938889503,
      "rewards-max/forensics_iou_reward": 0.9853812754154205,
      "rewards-max/forensics_strict_edge_reward": 0.450267031788826,
      "rewards-max/forensics_tiou_reward": 0.9753808975219727,
      "rewards/forensics_iou_reward": 0.5910342633724213,
      "rewards/forensics_strict_edge_reward": 0.057852206751704216,
      "rewards/forensics_tiou_reward": 0.5385519117116928,
      "step": 785
    },
    {
      "completion_length": 24.5,
      "epoch": 2.4116743471582183,
      "grad_norm": 6.673244476318359,
      "kl": 0.4111328125,
      "learning_rate": 3.961538461538461e-07,
      "loss": 0.0033,
      "reward": 0.7619183659553528,
      "reward_std": 0.043334368616342545,
      "reward_std_all": 0.04492894280701876,
      "rewards-max/forensics_iou_reward": 0.9345820546150208,
      "rewards-max/forensics_strict_edge_reward": 0.26107342168688774,
      "rewards-max/forensics_tiou_reward": 0.9135480225086212,
      "rewards/forensics_iou_reward": 0.6461972296237946,
      "rewards/forensics_strict_edge_reward": 0.0308974701911211,
      "rewards/forensics_tiou_reward": 0.5866618156433105,
      "step": 786
    },
    {
      "completion_length": 21.203125,
      "epoch": 2.4147465437788016,
      "grad_norm": 6.479431629180908,
      "kl": 0.3896484375,
      "learning_rate": 3.9538461538461537e-07,
      "loss": -0.0456,
      "reward": 0.7871895432472229,
      "reward_std": 0.018116185907274485,
      "reward_std_all": 0.019856294384226203,
      "rewards-max/forensics_iou_reward": 0.9721414744853973,
      "rewards-max/forensics_strict_edge_reward": 0.46688374876976013,
      "rewards-max/forensics_tiou_reward": 0.9597237706184387,
      "rewards/forensics_iou_reward": 0.7083703875541687,
      "rewards/forensics_strict_edge_reward": 0.046871867030858994,
      "rewards/forensics_tiou_reward": 0.653045117855072,
      "step": 787
    },
    {
      "completion_length": 24.75,
      "epoch": 2.4178187403993854,
      "grad_norm": 7.607930660247803,
      "kl": 0.359375,
      "learning_rate": 3.946153846153846e-07,
      "loss": 0.0352,
      "reward": 0.7997158169746399,
      "reward_std": 0.009667434613220394,
      "reward_std_all": 0.010833746753633022,
      "rewards-max/forensics_iou_reward": 0.9778623282909393,
      "rewards-max/forensics_strict_edge_reward": 0.4419209808111191,
      "rewards-max/forensics_tiou_reward": 0.9648492634296417,
      "rewards/forensics_iou_reward": 0.7051713466644287,
      "rewards/forensics_strict_edge_reward": 0.042172444984316826,
      "rewards/forensics_tiou_reward": 0.6543338894844055,
      "step": 788
    },
    {
      "completion_length": 24.5625,
      "epoch": 2.420890937019969,
      "grad_norm": 6.999288082122803,
      "kl": 0.4169921875,
      "learning_rate": 3.9384615384615385e-07,
      "loss": -0.0064,
      "reward": 0.7054951786994934,
      "reward_std": 0.01704871142283082,
      "reward_std_all": 0.061839308589696884,
      "rewards-max/forensics_iou_reward": 0.9776119291782379,
      "rewards-max/forensics_strict_edge_reward": 0.7744058072566986,
      "rewards-max/forensics_tiou_reward": 0.9733188152313232,
      "rewards/forensics_iou_reward": 0.5666385889053345,
      "rewards/forensics_strict_edge_reward": 0.05619368888437748,
      "rewards/forensics_tiou_reward": 0.5260712206363678,
      "step": 789
    },
    {
      "completion_length": 24.609375,
      "epoch": 2.423963133640553,
      "grad_norm": 6.006939888000488,
      "kl": 0.40234375,
      "learning_rate": 3.9307692307692306e-07,
      "loss": 0.0256,
      "reward": 0.7052322626113892,
      "reward_std": 0.058634910732507706,
      "reward_std_all": 0.08147936128079891,
      "rewards-max/forensics_iou_reward": 0.9693236351013184,
      "rewards-max/forensics_strict_edge_reward": 0.4789094477891922,
      "rewards-max/forensics_tiou_reward": 0.9602360725402832,
      "rewards/forensics_iou_reward": 0.6272103786468506,
      "rewards/forensics_strict_edge_reward": 0.056217487901449203,
      "rewards/forensics_tiou_reward": 0.5829910337924957,
      "step": 790
    },
    {
      "completion_length": 24.671875,
      "epoch": 2.4270353302611367,
      "grad_norm": 6.953779697418213,
      "kl": 0.494140625,
      "learning_rate": 3.923076923076923e-07,
      "loss": 0.1674,
      "reward": 0.7326588928699493,
      "reward_std": 0.0569177120923996,
      "reward_std_all": 0.06497040763497353,
      "rewards-max/forensics_iou_reward": 0.9912489652633667,
      "rewards-max/forensics_strict_edge_reward": 0.7837068438529968,
      "rewards-max/forensics_tiou_reward": 0.9877610802650452,
      "rewards/forensics_iou_reward": 0.6053502261638641,
      "rewards/forensics_strict_edge_reward": 0.08653360605239868,
      "rewards/forensics_tiou_reward": 0.5529071688652039,
      "step": 791
    },
    {
      "completion_length": 27.59375,
      "epoch": 2.4301075268817205,
      "grad_norm": 6.505720138549805,
      "kl": 0.4609375,
      "learning_rate": 3.9153846153846154e-07,
      "loss": 0.0403,
      "reward": 0.7639647722244263,
      "reward_std": 0.046157920733094215,
      "reward_std_all": 0.045643653720617294,
      "rewards-max/forensics_iou_reward": 0.9936675727367401,
      "rewards-max/forensics_strict_edge_reward": 0.5675660967826843,
      "rewards-max/forensics_tiou_reward": 0.9922943711280823,
      "rewards/forensics_iou_reward": 0.6592788696289062,
      "rewards/forensics_strict_edge_reward": 0.08304383605718613,
      "rewards/forensics_tiou_reward": 0.6229474544525146,
      "step": 792
    },
    {
      "completion_length": 24.5,
      "epoch": 2.4331797235023043,
      "grad_norm": 7.342344284057617,
      "kl": 0.626953125,
      "learning_rate": 3.9076923076923076e-07,
      "loss": 0.156,
      "reward": 0.7693173885345459,
      "reward_std": 0.04349474608898163,
      "reward_std_all": 0.043308766558766365,
      "rewards-max/forensics_iou_reward": 0.9963893294334412,
      "rewards-max/forensics_strict_edge_reward": 0.8381626009941101,
      "rewards-max/forensics_tiou_reward": 0.9952377676963806,
      "rewards/forensics_iou_reward": 0.6511808931827545,
      "rewards/forensics_strict_edge_reward": 0.09049167484045029,
      "rewards/forensics_tiou_reward": 0.5952334403991699,
      "step": 793
    },
    {
      "completion_length": 26.75,
      "epoch": 2.436251920122888,
      "grad_norm": 7.013576507568359,
      "kl": 0.404296875,
      "learning_rate": 3.8999999999999997e-07,
      "loss": -0.0074,
      "reward": 0.7255874574184418,
      "reward_std": 0.020870725624263287,
      "reward_std_all": 0.05061627924442291,
      "rewards-max/forensics_iou_reward": 0.9727486073970795,
      "rewards-max/forensics_strict_edge_reward": 0.33740925043821335,
      "rewards-max/forensics_tiou_reward": 0.9641396403312683,
      "rewards/forensics_iou_reward": 0.6383228898048401,
      "rewards/forensics_strict_edge_reward": 0.03336372599005699,
      "rewards/forensics_tiou_reward": 0.5791196823120117,
      "step": 794
    },
    {
      "completion_length": 32.234375,
      "epoch": 2.439324116743472,
      "grad_norm": 4.9831223487854,
      "kl": 0.451171875,
      "learning_rate": 3.892307692307692e-07,
      "loss": 0.1504,
      "reward": 0.71775883436203,
      "reward_std": 0.08758858032524586,
      "reward_std_all": 0.08753248676657677,
      "rewards-max/forensics_iou_reward": 0.9482014775276184,
      "rewards-max/forensics_strict_edge_reward": 0.29926537722349167,
      "rewards-max/forensics_tiou_reward": 0.9359093010425568,
      "rewards/forensics_iou_reward": 0.5835924744606018,
      "rewards/forensics_strict_edge_reward": 0.02333475463092327,
      "rewards/forensics_tiou_reward": 0.5450865477323532,
      "step": 795
    },
    {
      "completion_length": 27.6875,
      "epoch": 2.442396313364055,
      "grad_norm": 5.935353755950928,
      "kl": 0.3291015625,
      "learning_rate": 3.8846153846153845e-07,
      "loss": -0.0466,
      "reward": 0.7342129647731781,
      "reward_std": 0.07129317708313465,
      "reward_std_all": 0.07100637257099152,
      "rewards-max/forensics_iou_reward": 0.9818786680698395,
      "rewards-max/forensics_strict_edge_reward": 0.5192448049783707,
      "rewards-max/forensics_tiou_reward": 0.9707050919532776,
      "rewards/forensics_iou_reward": 0.7001903057098389,
      "rewards/forensics_strict_edge_reward": 0.06616567075252533,
      "rewards/forensics_tiou_reward": 0.6617204248905182,
      "step": 796
    },
    {
      "completion_length": 23.265625,
      "epoch": 2.445468509984639,
      "grad_norm": 7.470708847045898,
      "kl": 0.3740234375,
      "learning_rate": 3.8769230769230767e-07,
      "loss": 0.0314,
      "reward": 0.7931258082389832,
      "reward_std": 0.02172830607742071,
      "reward_std_all": 0.02061760611832142,
      "rewards-max/forensics_iou_reward": 0.9822483956813812,
      "rewards-max/forensics_strict_edge_reward": 0.6763102114200592,
      "rewards-max/forensics_tiou_reward": 0.9757565557956696,
      "rewards/forensics_iou_reward": 0.6884133219718933,
      "rewards/forensics_strict_edge_reward": 0.0678216777741909,
      "rewards/forensics_tiou_reward": 0.6411395072937012,
      "step": 797
    },
    {
      "completion_length": 23.890625,
      "epoch": 2.4485407066052227,
      "grad_norm": 8.8650541305542,
      "kl": 0.822265625,
      "learning_rate": 3.8692307692307693e-07,
      "loss": 0.052,
      "reward": 0.704494297504425,
      "reward_std": 0.07104014232754707,
      "reward_std_all": 0.08516932092607021,
      "rewards-max/forensics_iou_reward": 0.9882000982761383,
      "rewards-max/forensics_strict_edge_reward": 0.3517128527164459,
      "rewards-max/forensics_tiou_reward": 0.9844017922878265,
      "rewards/forensics_iou_reward": 0.6004227846860886,
      "rewards/forensics_strict_edge_reward": 0.05806344375014305,
      "rewards/forensics_tiou_reward": 0.547542542219162,
      "step": 798
    },
    {
      "completion_length": 30.890625,
      "epoch": 2.4516129032258065,
      "grad_norm": 5.8805131912231445,
      "kl": 0.509765625,
      "learning_rate": 3.8615384615384615e-07,
      "loss": 0.0864,
      "reward": 0.7456673085689545,
      "reward_std": 0.05501653626561165,
      "reward_std_all": 0.05904535576701164,
      "rewards-max/forensics_iou_reward": 0.9841746985912323,
      "rewards-max/forensics_strict_edge_reward": 0.5214770957827568,
      "rewards-max/forensics_tiou_reward": 0.9735396504402161,
      "rewards/forensics_iou_reward": 0.6441813111305237,
      "rewards/forensics_strict_edge_reward": 0.06113812141120434,
      "rewards/forensics_tiou_reward": 0.5943801999092102,
      "step": 799
    },
    {
      "completion_length": 24.9375,
      "epoch": 2.4546850998463903,
      "grad_norm": 6.300044059753418,
      "kl": 0.390625,
      "learning_rate": 3.8538461538461536e-07,
      "loss": 0.0466,
      "reward": 0.760625034570694,
      "reward_std": 0.07677732780575752,
      "reward_std_all": 0.06828097067773342,
      "rewards-max/forensics_iou_reward": 0.9771498739719391,
      "rewards-max/forensics_strict_edge_reward": 0.5422053933143616,
      "rewards-max/forensics_tiou_reward": 0.9705112278461456,
      "rewards/forensics_iou_reward": 0.6706370711326599,
      "rewards/forensics_strict_edge_reward": 0.07598315365612507,
      "rewards/forensics_tiou_reward": 0.6291817724704742,
      "step": 800
    },
    {
      "completion_length": 27.015625,
      "epoch": 2.457757296466974,
      "grad_norm": 7.4545392990112305,
      "kl": 0.388671875,
      "learning_rate": 3.8461538461538463e-07,
      "loss": 0.0491,
      "reward": 0.7237531244754791,
      "reward_std": 0.09117197804152966,
      "reward_std_all": 0.08822820894420147,
      "rewards-max/forensics_iou_reward": 0.9822516441345215,
      "rewards-max/forensics_strict_edge_reward": 0.3420734480023384,
      "rewards-max/forensics_tiou_reward": 0.9716016054153442,
      "rewards/forensics_iou_reward": 0.6238819658756256,
      "rewards/forensics_strict_edge_reward": 0.031907163094729185,
      "rewards/forensics_tiou_reward": 0.5785589516162872,
      "step": 801
    },
    {
      "completion_length": 24.546875,
      "epoch": 2.460829493087558,
      "grad_norm": 8.622176170349121,
      "kl": 0.962890625,
      "learning_rate": 3.8384615384615384e-07,
      "loss": 0.0721,
      "reward": 0.6980981230735779,
      "reward_std": 0.08860900066792965,
      "reward_std_all": 0.10360308736562729,
      "rewards-max/forensics_iou_reward": 0.9744715988636017,
      "rewards-max/forensics_strict_edge_reward": 0.6698661744594574,
      "rewards-max/forensics_tiou_reward": 0.967985063791275,
      "rewards/forensics_iou_reward": 0.5634116232395172,
      "rewards/forensics_strict_edge_reward": 0.05945851653814316,
      "rewards/forensics_tiou_reward": 0.5090046375989914,
      "step": 802
    },
    {
      "completion_length": 27.546875,
      "epoch": 2.463901689708141,
      "grad_norm": 6.812963485717773,
      "kl": 0.421875,
      "learning_rate": 3.8307692307692305e-07,
      "loss": 0.0429,
      "reward": 0.7477059960365295,
      "reward_std": 0.05287003889679909,
      "reward_std_all": 0.056189537048339844,
      "rewards-max/forensics_iou_reward": 0.9730096459388733,
      "rewards-max/forensics_strict_edge_reward": 0.3218531608581543,
      "rewards-max/forensics_tiou_reward": 0.9642508924007416,
      "rewards/forensics_iou_reward": 0.6389262080192566,
      "rewards/forensics_strict_edge_reward": 0.03224107436835766,
      "rewards/forensics_tiou_reward": 0.5932841002941132,
      "step": 803
    },
    {
      "completion_length": 24.75,
      "epoch": 2.466973886328725,
      "grad_norm": 7.434890270233154,
      "kl": 0.48046875,
      "learning_rate": 3.8230769230769227e-07,
      "loss": 0.1389,
      "reward": 0.6591319441795349,
      "reward_std": 0.13762284070253372,
      "reward_std_all": 0.14091132581233978,
      "rewards-max/forensics_iou_reward": 0.9754834473133087,
      "rewards-max/forensics_strict_edge_reward": 0.4289195388555527,
      "rewards-max/forensics_tiou_reward": 0.9723018407821655,
      "rewards/forensics_iou_reward": 0.5538240969181061,
      "rewards/forensics_strict_edge_reward": 0.03486552648246288,
      "rewards/forensics_tiou_reward": 0.5046744793653488,
      "step": 804
    },
    {
      "completion_length": 27.046875,
      "epoch": 2.4700460829493087,
      "grad_norm": 5.996354579925537,
      "kl": 0.3662109375,
      "learning_rate": 3.8153846153846153e-07,
      "loss": 0.0992,
      "reward": 0.7502055764198303,
      "reward_std": 0.08174098841845989,
      "reward_std_all": 0.07442802004516125,
      "rewards-max/forensics_iou_reward": 0.9812059700489044,
      "rewards-max/forensics_strict_edge_reward": 0.3766644448041916,
      "rewards-max/forensics_tiou_reward": 0.9755539894104004,
      "rewards/forensics_iou_reward": 0.6099531948566437,
      "rewards/forensics_strict_edge_reward": 0.039738661609590054,
      "rewards/forensics_tiou_reward": 0.5564155578613281,
      "step": 805
    },
    {
      "completion_length": 24.8125,
      "epoch": 2.4731182795698925,
      "grad_norm": 7.579362392425537,
      "kl": 0.3291015625,
      "learning_rate": 3.8076923076923075e-07,
      "loss": 0.2046,
      "reward": 0.7258580029010773,
      "reward_std": 0.10255761072039604,
      "reward_std_all": 0.09541954845190048,
      "rewards-max/forensics_iou_reward": 0.9366804361343384,
      "rewards-max/forensics_strict_edge_reward": 0.22165070101618767,
      "rewards-max/forensics_tiou_reward": 0.8970104455947876,
      "rewards/forensics_iou_reward": 0.5638522803783417,
      "rewards/forensics_strict_edge_reward": 0.019890389405190945,
      "rewards/forensics_tiou_reward": 0.5071914345026016,
      "step": 806
    },
    {
      "completion_length": 25.53125,
      "epoch": 2.4761904761904763,
      "grad_norm": 7.482072830200195,
      "kl": 0.544921875,
      "learning_rate": 3.7999999999999996e-07,
      "loss": 0.1067,
      "reward": 0.7463106215000153,
      "reward_std": 0.07848163414746523,
      "reward_std_all": 0.07520774379372597,
      "rewards-max/forensics_iou_reward": 0.9502873718738556,
      "rewards-max/forensics_strict_edge_reward": 0.4263888746500015,
      "rewards-max/forensics_tiou_reward": 0.9410389065742493,
      "rewards/forensics_iou_reward": 0.5647304058074951,
      "rewards/forensics_strict_edge_reward": 0.03062162920832634,
      "rewards/forensics_tiou_reward": 0.5098792612552643,
      "step": 807
    },
    {
      "completion_length": 25.75,
      "epoch": 2.47926267281106,
      "grad_norm": 6.186276912689209,
      "kl": 0.3369140625,
      "learning_rate": 3.792307692307692e-07,
      "loss": 0.1042,
      "reward": 0.7382218539714813,
      "reward_std": 0.08850893005728722,
      "reward_std_all": 0.08200463466346264,
      "rewards-max/forensics_iou_reward": 0.9597321450710297,
      "rewards-max/forensics_strict_edge_reward": 0.25739095360040665,
      "rewards-max/forensics_tiou_reward": 0.9490264356136322,
      "rewards/forensics_iou_reward": 0.6518747806549072,
      "rewards/forensics_strict_edge_reward": 0.026854638010263443,
      "rewards/forensics_tiou_reward": 0.6048410534858704,
      "step": 808
    },
    {
      "completion_length": 26.9375,
      "epoch": 2.482334869431644,
      "grad_norm": 6.557861804962158,
      "kl": 0.455078125,
      "learning_rate": 3.7846153846153844e-07,
      "loss": 0.0404,
      "reward": 0.7475854456424713,
      "reward_std": 0.04444231139495969,
      "reward_std_all": 0.04931094637140632,
      "rewards-max/forensics_iou_reward": 0.9888723790645599,
      "rewards-max/forensics_strict_edge_reward": 0.6265945732593536,
      "rewards-max/forensics_tiou_reward": 0.9853388667106628,
      "rewards/forensics_iou_reward": 0.6315411031246185,
      "rewards/forensics_strict_edge_reward": 0.05987736955285072,
      "rewards/forensics_tiou_reward": 0.5850198864936829,
      "step": 809
    },
    {
      "completion_length": 27.65625,
      "epoch": 2.485407066052227,
      "grad_norm": 7.515693187713623,
      "kl": 0.427734375,
      "learning_rate": 3.776923076923077e-07,
      "loss": 0.0368,
      "reward": 0.7199711203575134,
      "reward_std": 0.07716306671500206,
      "reward_std_all": 0.08342155441641808,
      "rewards-max/forensics_iou_reward": 0.9781218469142914,
      "rewards-max/forensics_strict_edge_reward": 0.7069477140903473,
      "rewards-max/forensics_tiou_reward": 0.9732823371887207,
      "rewards/forensics_iou_reward": 0.5998386144638062,
      "rewards/forensics_strict_edge_reward": 0.07553007453680038,
      "rewards/forensics_tiou_reward": 0.5509154796600342,
      "step": 810
    },
    {
      "completion_length": 26.859375,
      "epoch": 2.488479262672811,
      "grad_norm": 5.997760772705078,
      "kl": 0.3935546875,
      "learning_rate": 3.769230769230769e-07,
      "loss": 0.0361,
      "reward": 0.7900407314300537,
      "reward_std": 0.018689191434532404,
      "reward_std_all": 0.01941005839034915,
      "rewards-max/forensics_iou_reward": 0.9680080413818359,
      "rewards-max/forensics_strict_edge_reward": 0.4493289589881897,
      "rewards-max/forensics_tiou_reward": 0.956413060426712,
      "rewards/forensics_iou_reward": 0.6910163462162018,
      "rewards/forensics_strict_edge_reward": 0.03203254472464323,
      "rewards/forensics_tiou_reward": 0.6286502182483673,
      "step": 811
    },
    {
      "completion_length": 27.734375,
      "epoch": 2.4915514592933947,
      "grad_norm": 5.5963921546936035,
      "kl": 0.3369140625,
      "learning_rate": 3.7615384615384614e-07,
      "loss": 0.1947,
      "reward": 0.6953662037849426,
      "reward_std": 0.12817614153027534,
      "reward_std_all": 0.12127123773097992,
      "rewards-max/forensics_iou_reward": 0.9448826611042023,
      "rewards-max/forensics_strict_edge_reward": 0.5885358899831772,
      "rewards-max/forensics_tiou_reward": 0.9270835518836975,
      "rewards/forensics_iou_reward": 0.5257848501205444,
      "rewards/forensics_strict_edge_reward": 0.0499050822108984,
      "rewards/forensics_tiou_reward": 0.48513978719711304,
      "step": 812
    },
    {
      "completion_length": 28.3125,
      "epoch": 2.4946236559139785,
      "grad_norm": 7.435177326202393,
      "kl": 0.453125,
      "learning_rate": 3.7538461538461535e-07,
      "loss": 0.0067,
      "reward": 0.7849756479263306,
      "reward_std": 0.02288325224071741,
      "reward_std_all": 0.02380344830453396,
      "rewards-max/forensics_iou_reward": 0.978353351354599,
      "rewards-max/forensics_strict_edge_reward": 0.46338221430778503,
      "rewards-max/forensics_tiou_reward": 0.969030350446701,
      "rewards/forensics_iou_reward": 0.7304853200912476,
      "rewards/forensics_strict_edge_reward": 0.05903041921555996,
      "rewards/forensics_tiou_reward": 0.67128124833107,
      "step": 813
    },
    {
      "completion_length": 30.203125,
      "epoch": 2.4976958525345623,
      "grad_norm": 6.509426116943359,
      "kl": 0.5,
      "learning_rate": 3.746153846153846e-07,
      "loss": -0.0198,
      "reward": 0.7865855097770691,
      "reward_std": 0.03327127720694989,
      "reward_std_all": 0.030864700442180037,
      "rewards-max/forensics_iou_reward": 0.9841609597206116,
      "rewards-max/forensics_strict_edge_reward": 0.5088357329368591,
      "rewards-max/forensics_tiou_reward": 0.9779722392559052,
      "rewards/forensics_iou_reward": 0.7123958170413971,
      "rewards/forensics_strict_edge_reward": 0.06043485924601555,
      "rewards/forensics_tiou_reward": 0.6747485399246216,
      "step": 814
    },
    {
      "completion_length": 24.171875,
      "epoch": 2.500768049155146,
      "grad_norm": 7.149110317230225,
      "kl": 0.392578125,
      "learning_rate": 3.7384615384615383e-07,
      "loss": 0.0637,
      "reward": 0.771695464849472,
      "reward_std": 0.04955677501857281,
      "reward_std_all": 0.046012984588742256,
      "rewards-max/forensics_iou_reward": 0.9780662357807159,
      "rewards-max/forensics_strict_edge_reward": 0.6153465881943703,
      "rewards-max/forensics_tiou_reward": 0.9751094579696655,
      "rewards/forensics_iou_reward": 0.6744533181190491,
      "rewards/forensics_strict_edge_reward": 0.09118517488241196,
      "rewards/forensics_tiou_reward": 0.6249558329582214,
      "step": 815
    },
    {
      "completion_length": 24.453125,
      "epoch": 2.50384024577573,
      "grad_norm": 6.96450662612915,
      "kl": 0.4580078125,
      "learning_rate": 3.7307692307692304e-07,
      "loss": 0.0253,
      "reward": 0.7705929279327393,
      "reward_std": 0.05252561601810157,
      "reward_std_all": 0.04820676217786968,
      "rewards-max/forensics_iou_reward": 0.9921368658542633,
      "rewards-max/forensics_strict_edge_reward": 0.8667194843292236,
      "rewards-max/forensics_tiou_reward": 0.9903486669063568,
      "rewards/forensics_iou_reward": 0.727972686290741,
      "rewards/forensics_strict_edge_reward": 0.10377291589975357,
      "rewards/forensics_tiou_reward": 0.6842313408851624,
      "step": 816
    },
    {
      "completion_length": 23.234375,
      "epoch": 2.506912442396313,
      "grad_norm": 6.151392936706543,
      "kl": 0.458984375,
      "learning_rate": 3.7230769230769226e-07,
      "loss": 0.0185,
      "reward": 0.7180259823799133,
      "reward_std": 0.10831825062632561,
      "reward_std_all": 0.10146840289235115,
      "rewards-max/forensics_iou_reward": 0.9519686102867126,
      "rewards-max/forensics_strict_edge_reward": 0.37537892162799835,
      "rewards-max/forensics_tiou_reward": 0.9395642876625061,
      "rewards/forensics_iou_reward": 0.5955116450786591,
      "rewards/forensics_strict_edge_reward": 0.04246027022600174,
      "rewards/forensics_tiou_reward": 0.5494821667671204,
      "step": 817
    },
    {
      "completion_length": 25.203125,
      "epoch": 2.509984639016897,
      "grad_norm": 7.119509220123291,
      "kl": 0.462890625,
      "learning_rate": 3.715384615384615e-07,
      "loss": 0.1335,
      "reward": 0.7191360294818878,
      "reward_std": 0.09270882606506348,
      "reward_std_all": 0.08954321965575218,
      "rewards-max/forensics_iou_reward": 0.9715957641601562,
      "rewards-max/forensics_strict_edge_reward": 0.4352063685655594,
      "rewards-max/forensics_tiou_reward": 0.9571668803691864,
      "rewards/forensics_iou_reward": 0.5665056705474854,
      "rewards/forensics_strict_edge_reward": 0.03548811003565788,
      "rewards/forensics_tiou_reward": 0.507142573595047,
      "step": 818
    },
    {
      "completion_length": 26.828125,
      "epoch": 2.5130568356374807,
      "grad_norm": 7.01637601852417,
      "kl": 0.515625,
      "learning_rate": 3.707692307692308e-07,
      "loss": 0.0998,
      "reward": 0.7813579738140106,
      "reward_std": 0.03835479589179158,
      "reward_std_all": 0.03553358791396022,
      "rewards-max/forensics_iou_reward": 0.9856206774711609,
      "rewards-max/forensics_strict_edge_reward": 0.5686750710010529,
      "rewards-max/forensics_tiou_reward": 0.9805781245231628,
      "rewards/forensics_iou_reward": 0.7167094051837921,
      "rewards/forensics_strict_edge_reward": 0.08265215903520584,
      "rewards/forensics_tiou_reward": 0.6654514074325562,
      "step": 819
    },
    {
      "completion_length": 26.234375,
      "epoch": 2.5161290322580645,
      "grad_norm": 6.449347019195557,
      "kl": 0.341796875,
      "learning_rate": 3.7e-07,
      "loss": -0.0002,
      "reward": 0.7934025824069977,
      "reward_std": 0.01375977008137852,
      "reward_std_all": 0.015386674087494612,
      "rewards-max/forensics_iou_reward": 0.9736022055149078,
      "rewards-max/forensics_strict_edge_reward": 0.4592139273881912,
      "rewards-max/forensics_tiou_reward": 0.967212051153183,
      "rewards/forensics_iou_reward": 0.6740031540393829,
      "rewards/forensics_strict_edge_reward": 0.023940503131598234,
      "rewards/forensics_tiou_reward": 0.6182790994644165,
      "step": 820
    },
    {
      "completion_length": 29.546875,
      "epoch": 2.5192012288786483,
      "grad_norm": 5.473827838897705,
      "kl": 0.4228515625,
      "learning_rate": 3.692307692307692e-07,
      "loss": 0.1483,
      "reward": 0.7451261579990387,
      "reward_std": 0.06060793809592724,
      "reward_std_all": 0.06030910834670067,
      "rewards-max/forensics_iou_reward": 0.9834830462932587,
      "rewards-max/forensics_strict_edge_reward": 0.580880731344223,
      "rewards-max/forensics_tiou_reward": 0.9775108098983765,
      "rewards/forensics_iou_reward": 0.6500125527381897,
      "rewards/forensics_strict_edge_reward": 0.08950764685869217,
      "rewards/forensics_tiou_reward": 0.6104640662670135,
      "step": 821
    },
    {
      "completion_length": 26.125,
      "epoch": 2.522273425499232,
      "grad_norm": 5.2376708984375,
      "kl": 0.3662109375,
      "learning_rate": 3.6846153846153843e-07,
      "loss": -0.0357,
      "reward": 0.7646281719207764,
      "reward_std": 0.06961522251367569,
      "reward_std_all": 0.06227155029773712,
      "rewards-max/forensics_iou_reward": 0.9765872061252594,
      "rewards-max/forensics_strict_edge_reward": 0.6421464383602142,
      "rewards-max/forensics_tiou_reward": 0.9714799523353577,
      "rewards/forensics_iou_reward": 0.7267504632472992,
      "rewards/forensics_strict_edge_reward": 0.09296566620469093,
      "rewards/forensics_tiou_reward": 0.6794247031211853,
      "step": 822
    },
    {
      "completion_length": 25.828125,
      "epoch": 2.525345622119816,
      "grad_norm": 10.391417503356934,
      "kl": 0.4404296875,
      "learning_rate": 3.676923076923077e-07,
      "loss": 0.1048,
      "reward": 0.7571222186088562,
      "reward_std": 0.060798101127147675,
      "reward_std_all": 0.058073271065950394,
      "rewards-max/forensics_iou_reward": 0.9443430602550507,
      "rewards-max/forensics_strict_edge_reward": 0.5382633358240128,
      "rewards-max/forensics_tiou_reward": 0.9313846528530121,
      "rewards/forensics_iou_reward": 0.6173295974731445,
      "rewards/forensics_strict_edge_reward": 0.06898034736514091,
      "rewards/forensics_tiou_reward": 0.5666837990283966,
      "step": 823
    },
    {
      "completion_length": 27.734375,
      "epoch": 2.528417818740399,
      "grad_norm": 6.286891460418701,
      "kl": 0.470703125,
      "learning_rate": 3.669230769230769e-07,
      "loss": 0.0506,
      "reward": 0.7522702217102051,
      "reward_std": 0.053369857370853424,
      "reward_std_all": 0.05361863225698471,
      "rewards-max/forensics_iou_reward": 0.9603420794010162,
      "rewards-max/forensics_strict_edge_reward": 0.42128483951091766,
      "rewards-max/forensics_tiou_reward": 0.9417926371097565,
      "rewards/forensics_iou_reward": 0.6247033476829529,
      "rewards/forensics_strict_edge_reward": 0.06908626854419708,
      "rewards/forensics_tiou_reward": 0.5775221884250641,
      "step": 824
    },
    {
      "completion_length": 25.734375,
      "epoch": 2.531490015360983,
      "grad_norm": 8.213274002075195,
      "kl": 0.4599609375,
      "learning_rate": 3.661538461538461e-07,
      "loss": -0.0179,
      "reward": 0.763606607913971,
      "reward_std": 0.06746174581348896,
      "reward_std_all": 0.061334168538451195,
      "rewards-max/forensics_iou_reward": 0.9069435298442841,
      "rewards-max/forensics_strict_edge_reward": 0.24896756373345852,
      "rewards-max/forensics_tiou_reward": 0.8799168467521667,
      "rewards/forensics_iou_reward": 0.6619933247566223,
      "rewards/forensics_strict_edge_reward": 0.018755112774670124,
      "rewards/forensics_tiou_reward": 0.6119988858699799,
      "step": 825
    },
    {
      "completion_length": 27.4375,
      "epoch": 2.5345622119815667,
      "grad_norm": 5.816349983215332,
      "kl": 0.5029296875,
      "learning_rate": 3.6538461538461534e-07,
      "loss": 0.0914,
      "reward": 0.703826367855072,
      "reward_std": 0.07994630187749863,
      "reward_std_all": 0.09576386958360672,
      "rewards-max/forensics_iou_reward": 0.992689847946167,
      "rewards-max/forensics_strict_edge_reward": 0.6405399739742279,
      "rewards-max/forensics_tiou_reward": 0.9901228249073029,
      "rewards/forensics_iou_reward": 0.6461336612701416,
      "rewards/forensics_strict_edge_reward": 0.08746140822768211,
      "rewards/forensics_tiou_reward": 0.6014814376831055,
      "step": 826
    },
    {
      "completion_length": 23.390625,
      "epoch": 2.5376344086021505,
      "grad_norm": 8.723883628845215,
      "kl": 0.865234375,
      "learning_rate": 3.646153846153846e-07,
      "loss": 0.0103,
      "reward": 0.743281215429306,
      "reward_std": 0.05714372918009758,
      "reward_std_all": 0.05840518698096275,
      "rewards-max/forensics_iou_reward": 0.9754111766815186,
      "rewards-max/forensics_strict_edge_reward": 0.6271176636219025,
      "rewards-max/forensics_tiou_reward": 0.9689755737781525,
      "rewards/forensics_iou_reward": 0.6503970324993134,
      "rewards/forensics_strict_edge_reward": 0.06496772542595863,
      "rewards/forensics_tiou_reward": 0.5986534357070923,
      "step": 827
    },
    {
      "completion_length": 24.9375,
      "epoch": 2.5407066052227343,
      "grad_norm": 6.809017658233643,
      "kl": 0.376953125,
      "learning_rate": 3.638461538461538e-07,
      "loss": -0.049,
      "reward": 0.6609519720077515,
      "reward_std": 0.10911311022937298,
      "reward_std_all": 0.13129431754350662,
      "rewards-max/forensics_iou_reward": 0.9785151779651642,
      "rewards-max/forensics_strict_edge_reward": 0.418794646859169,
      "rewards-max/forensics_tiou_reward": 0.9661760926246643,
      "rewards/forensics_iou_reward": 0.5747271776199341,
      "rewards/forensics_strict_edge_reward": 0.05348120629787445,
      "rewards/forensics_tiou_reward": 0.5320282578468323,
      "step": 828
    },
    {
      "completion_length": 24.453125,
      "epoch": 2.543778801843318,
      "grad_norm": 7.839186668395996,
      "kl": 0.435546875,
      "learning_rate": 3.6307692307692303e-07,
      "loss": -0.0052,
      "reward": 0.7144120037555695,
      "reward_std": 0.04090686049312353,
      "reward_std_all": 0.06648565270006657,
      "rewards-max/forensics_iou_reward": 0.9390702247619629,
      "rewards-max/forensics_strict_edge_reward": 0.32959966361522675,
      "rewards-max/forensics_tiou_reward": 0.9234041273593903,
      "rewards/forensics_iou_reward": 0.633338063955307,
      "rewards/forensics_strict_edge_reward": 0.05444205924868584,
      "rewards/forensics_tiou_reward": 0.5842376947402954,
      "step": 829
    },
    {
      "completion_length": 27.3125,
      "epoch": 2.546850998463902,
      "grad_norm": 8.148487091064453,
      "kl": 0.48828125,
      "learning_rate": 3.623076923076923e-07,
      "loss": 0.0919,
      "reward": 0.7450321912765503,
      "reward_std": 0.06637008115649223,
      "reward_std_all": 0.06503280811011791,
      "rewards-max/forensics_iou_reward": 0.978110671043396,
      "rewards-max/forensics_strict_edge_reward": 0.40861257910728455,
      "rewards-max/forensics_tiou_reward": 0.9716630280017853,
      "rewards/forensics_iou_reward": 0.583246260881424,
      "rewards/forensics_strict_edge_reward": 0.0492205573245883,
      "rewards/forensics_tiou_reward": 0.5381107926368713,
      "step": 830
    },
    {
      "completion_length": 27.09375,
      "epoch": 2.549923195084485,
      "grad_norm": 7.73259973526001,
      "kl": 0.3837890625,
      "learning_rate": 3.615384615384615e-07,
      "loss": -0.0171,
      "reward": 0.6840311884880066,
      "reward_std": 0.0836135521531105,
      "reward_std_all": 0.10484351217746735,
      "rewards-max/forensics_iou_reward": 0.9384601712226868,
      "rewards-max/forensics_strict_edge_reward": 0.6347885429859161,
      "rewards-max/forensics_tiou_reward": 0.9251707494258881,
      "rewards/forensics_iou_reward": 0.5672737658023834,
      "rewards/forensics_strict_edge_reward": 0.07141515053808689,
      "rewards/forensics_tiou_reward": 0.5303204655647278,
      "step": 831
    },
    {
      "completion_length": 25.9375,
      "epoch": 2.5529953917050694,
      "grad_norm": 7.5906267166137695,
      "kl": 0.4560546875,
      "learning_rate": 3.607692307692308e-07,
      "loss": 0.0305,
      "reward": 0.7866950333118439,
      "reward_std": 0.03125400934368372,
      "reward_std_all": 0.029081476852297783,
      "rewards-max/forensics_iou_reward": 0.9777564406394958,
      "rewards-max/forensics_strict_edge_reward": 0.4906764030456543,
      "rewards-max/forensics_tiou_reward": 0.9705145955085754,
      "rewards/forensics_iou_reward": 0.7452190220355988,
      "rewards/forensics_strict_edge_reward": 0.0498039610683918,
      "rewards/forensics_tiou_reward": 0.6893862187862396,
      "step": 832
    },
    {
      "completion_length": 26.421875,
      "epoch": 2.5560675883256527,
      "grad_norm": 7.379814624786377,
      "kl": 0.4384765625,
      "learning_rate": 3.6e-07,
      "loss": -0.0278,
      "reward": 0.7102840840816498,
      "reward_std": 0.07189846783876419,
      "reward_std_all": 0.08902943879365921,
      "rewards-max/forensics_iou_reward": 0.9824002683162689,
      "rewards-max/forensics_strict_edge_reward": 0.5275620073080063,
      "rewards-max/forensics_tiou_reward": 0.9763094186782837,
      "rewards/forensics_iou_reward": 0.6355639398097992,
      "rewards/forensics_strict_edge_reward": 0.061623286455869675,
      "rewards/forensics_tiou_reward": 0.5897193253040314,
      "step": 833
    },
    {
      "completion_length": 23.078125,
      "epoch": 2.5591397849462365,
      "grad_norm": 6.425806045532227,
      "kl": 0.390625,
      "learning_rate": 3.592307692307692e-07,
      "loss": 0.0521,
      "reward": 0.7324657142162323,
      "reward_std": 0.031361622735857964,
      "reward_std_all": 0.054238567128777504,
      "rewards-max/forensics_iou_reward": 0.992272675037384,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9894757866859436,
      "rewards/forensics_iou_reward": 0.6559462547302246,
      "rewards/forensics_strict_edge_reward": 0.07547461986541748,
      "rewards/forensics_tiou_reward": 0.5979273021221161,
      "step": 834
    },
    {
      "completion_length": 24.875,
      "epoch": 2.5622119815668203,
      "grad_norm": 7.23925256729126,
      "kl": 0.580078125,
      "learning_rate": 3.584615384615384e-07,
      "loss": 0.1252,
      "reward": 0.7361398339271545,
      "reward_std": 0.09008648619055748,
      "reward_std_all": 0.08371456153690815,
      "rewards-max/forensics_iou_reward": 0.9752411842346191,
      "rewards-max/forensics_strict_edge_reward": 0.5570182800292969,
      "rewards-max/forensics_tiou_reward": 0.9665703177452087,
      "rewards/forensics_iou_reward": 0.6190418004989624,
      "rewards/forensics_strict_edge_reward": 0.0611252561211586,
      "rewards/forensics_tiou_reward": 0.5647920072078705,
      "step": 835
    },
    {
      "completion_length": 27.28125,
      "epoch": 2.565284178187404,
      "grad_norm": 7.234572887420654,
      "kl": 0.44140625,
      "learning_rate": 3.576923076923077e-07,
      "loss": 0.1178,
      "reward": 0.6759903728961945,
      "reward_std": 0.15807469189167023,
      "reward_std_all": 0.14782720059156418,
      "rewards-max/forensics_iou_reward": 0.9705069959163666,
      "rewards-max/forensics_strict_edge_reward": 0.43758223950862885,
      "rewards-max/forensics_tiou_reward": 0.962394118309021,
      "rewards/forensics_iou_reward": 0.5143847167491913,
      "rewards/forensics_strict_edge_reward": 0.025095317512750626,
      "rewards/forensics_tiou_reward": 0.45414409041404724,
      "step": 836
    },
    {
      "completion_length": 23.03125,
      "epoch": 2.568356374807988,
      "grad_norm": 6.772005081176758,
      "kl": 0.365234375,
      "learning_rate": 3.569230769230769e-07,
      "loss": 0.1002,
      "reward": 0.7565299272537231,
      "reward_std": 0.08356946334242821,
      "reward_std_all": 0.0742633119225502,
      "rewards-max/forensics_iou_reward": 0.9773494005203247,
      "rewards-max/forensics_strict_edge_reward": 0.3085371255874634,
      "rewards-max/forensics_tiou_reward": 0.9696835279464722,
      "rewards/forensics_iou_reward": 0.6584239602088928,
      "rewards/forensics_strict_edge_reward": 0.05189056135714054,
      "rewards/forensics_tiou_reward": 0.6069239675998688,
      "step": 837
    },
    {
      "completion_length": 24.875,
      "epoch": 2.571428571428571,
      "grad_norm": 6.978824138641357,
      "kl": 0.3681640625,
      "learning_rate": 3.561538461538461e-07,
      "loss": -0.0255,
      "reward": 0.7831944823265076,
      "reward_std": 0.01971989928279072,
      "reward_std_all": 0.02219242916908115,
      "rewards-max/forensics_iou_reward": 0.9285806715488434,
      "rewards-max/forensics_strict_edge_reward": 0.34556837379932404,
      "rewards-max/forensics_tiou_reward": 0.9066461026668549,
      "rewards/forensics_iou_reward": 0.7141275703907013,
      "rewards/forensics_strict_edge_reward": 0.031086193397641182,
      "rewards/forensics_tiou_reward": 0.6628246605396271,
      "step": 838
    },
    {
      "completion_length": 33.09375,
      "epoch": 2.5745007680491554,
      "grad_norm": 6.8765668869018555,
      "kl": 0.404296875,
      "learning_rate": 3.553846153846154e-07,
      "loss": 0.0197,
      "reward": 0.6756727397441864,
      "reward_std": 0.1168789304792881,
      "reward_std_all": 0.12161093950271606,
      "rewards-max/forensics_iou_reward": 0.9450868666172028,
      "rewards-max/forensics_strict_edge_reward": 0.26604741252958775,
      "rewards-max/forensics_tiou_reward": 0.9301161468029022,
      "rewards/forensics_iou_reward": 0.5476749539375305,
      "rewards/forensics_strict_edge_reward": 0.015317022567614913,
      "rewards/forensics_tiou_reward": 0.505865141749382,
      "step": 839
    },
    {
      "completion_length": 23.71875,
      "epoch": 2.5775729646697387,
      "grad_norm": 6.05040979385376,
      "kl": 0.392578125,
      "learning_rate": 3.546153846153846e-07,
      "loss": 0.0951,
      "reward": 0.7757143676280975,
      "reward_std": 0.041744086891412735,
      "reward_std_all": 0.03969058580696583,
      "rewards-max/forensics_iou_reward": 0.9359093010425568,
      "rewards-max/forensics_strict_edge_reward": 0.4860214702785015,
      "rewards-max/forensics_tiou_reward": 0.9173970222473145,
      "rewards/forensics_iou_reward": 0.6565067172050476,
      "rewards/forensics_strict_edge_reward": 0.07058892399072647,
      "rewards/forensics_tiou_reward": 0.606913149356842,
      "step": 840
    },
    {
      "completion_length": 26.328125,
      "epoch": 2.5806451612903225,
      "grad_norm": 6.090779781341553,
      "kl": 0.517578125,
      "learning_rate": 3.5384615384615386e-07,
      "loss": 0.031,
      "reward": 0.7760331332683563,
      "reward_std": 0.040425101295113564,
      "reward_std_all": 0.038838597014546394,
      "rewards-max/forensics_iou_reward": 0.9831667244434357,
      "rewards-max/forensics_strict_edge_reward": 0.6095658242702484,
      "rewards-max/forensics_tiou_reward": 0.9758034944534302,
      "rewards/forensics_iou_reward": 0.6546481847763062,
      "rewards/forensics_strict_edge_reward": 0.04453008063137531,
      "rewards/forensics_tiou_reward": 0.6058976948261261,
      "step": 841
    },
    {
      "completion_length": 22.546875,
      "epoch": 2.5837173579109063,
      "grad_norm": 7.397331714630127,
      "kl": 0.4130859375,
      "learning_rate": 3.530769230769231e-07,
      "loss": 0.0857,
      "reward": 0.7496857643127441,
      "reward_std": 0.03946629632264376,
      "reward_std_all": 0.04690632736310363,
      "rewards-max/forensics_iou_reward": 0.9746900498867035,
      "rewards-max/forensics_strict_edge_reward": 0.4199727475643158,
      "rewards-max/forensics_tiou_reward": 0.9693639278411865,
      "rewards/forensics_iou_reward": 0.6032889783382416,
      "rewards/forensics_strict_edge_reward": 0.0515426192432642,
      "rewards/forensics_tiou_reward": 0.5406626760959625,
      "step": 842
    },
    {
      "completion_length": 31.5625,
      "epoch": 2.58678955453149,
      "grad_norm": 11.412700653076172,
      "kl": 0.4267578125,
      "learning_rate": 3.523076923076923e-07,
      "loss": 0.0805,
      "reward": 0.7692177295684814,
      "reward_std": 0.04249801766127348,
      "reward_std_all": 0.0422714501619339,
      "rewards-max/forensics_iou_reward": 0.966619461774826,
      "rewards-max/forensics_strict_edge_reward": 0.4858183264732361,
      "rewards-max/forensics_tiou_reward": 0.9574716687202454,
      "rewards/forensics_iou_reward": 0.6304206550121307,
      "rewards/forensics_strict_edge_reward": 0.05895613506436348,
      "rewards/forensics_tiou_reward": 0.585921585559845,
      "step": 843
    },
    {
      "completion_length": 27.4375,
      "epoch": 2.589861751152074,
      "grad_norm": 8.137136459350586,
      "kl": 0.4072265625,
      "learning_rate": 3.515384615384615e-07,
      "loss": 0.0436,
      "reward": 0.7663967609405518,
      "reward_std": 0.04270612448453903,
      "reward_std_all": 0.042513396590948105,
      "rewards-max/forensics_iou_reward": 0.9913491308689117,
      "rewards-max/forensics_strict_edge_reward": 0.5003146827220917,
      "rewards-max/forensics_tiou_reward": 0.9865045249462128,
      "rewards/forensics_iou_reward": 0.7050576508045197,
      "rewards/forensics_strict_edge_reward": 0.049270983785390854,
      "rewards/forensics_tiou_reward": 0.6488156914710999,
      "step": 844
    },
    {
      "completion_length": 31.03125,
      "epoch": 2.592933947772657,
      "grad_norm": 6.512624740600586,
      "kl": 0.458984375,
      "learning_rate": 3.5076923076923077e-07,
      "loss": 0.0707,
      "reward": 0.6862281262874603,
      "reward_std": 0.14451555162668228,
      "reward_std_all": 0.13689086586236954,
      "rewards-max/forensics_iou_reward": 0.9808022677898407,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.975093424320221,
      "rewards/forensics_iou_reward": 0.557392418384552,
      "rewards/forensics_strict_edge_reward": 0.06562844291329384,
      "rewards/forensics_tiou_reward": 0.5098351538181305,
      "step": 845
    },
    {
      "completion_length": 29.40625,
      "epoch": 2.5960061443932414,
      "grad_norm": 6.389960765838623,
      "kl": 0.5068359375,
      "learning_rate": 3.5e-07,
      "loss": 0.1059,
      "reward": 0.766464114189148,
      "reward_std": 0.03762726637069136,
      "reward_std_all": 0.04117250721901655,
      "rewards-max/forensics_iou_reward": 0.9860042631626129,
      "rewards-max/forensics_strict_edge_reward": 0.6079092025756836,
      "rewards-max/forensics_tiou_reward": 0.9826246201992035,
      "rewards/forensics_iou_reward": 0.68864706158638,
      "rewards/forensics_strict_edge_reward": 0.06485243141651154,
      "rewards/forensics_tiou_reward": 0.642108291387558,
      "step": 846
    },
    {
      "completion_length": 26.59375,
      "epoch": 2.5990783410138247,
      "grad_norm": 8.578204154968262,
      "kl": 0.32763671875,
      "learning_rate": 3.492307692307692e-07,
      "loss": 0.083,
      "reward": 0.7520252764225006,
      "reward_std": 0.03154192119836807,
      "reward_std_all": 0.0447094663977623,
      "rewards-max/forensics_iou_reward": 0.9586184322834015,
      "rewards-max/forensics_strict_edge_reward": 0.3845105320215225,
      "rewards-max/forensics_tiou_reward": 0.9430593848228455,
      "rewards/forensics_iou_reward": 0.6376864612102509,
      "rewards/forensics_strict_edge_reward": 0.06139921769499779,
      "rewards/forensics_tiou_reward": 0.5908141136169434,
      "step": 847
    },
    {
      "completion_length": 28.0,
      "epoch": 2.6021505376344085,
      "grad_norm": 7.539186000823975,
      "kl": 0.3994140625,
      "learning_rate": 3.4846153846153846e-07,
      "loss": 0.0196,
      "reward": 0.6766782104969025,
      "reward_std": 0.0511786974966526,
      "reward_std_all": 0.09225179255008698,
      "rewards-max/forensics_iou_reward": 0.9421469271183014,
      "rewards-max/forensics_strict_edge_reward": 0.4827862083911896,
      "rewards-max/forensics_tiou_reward": 0.9270274937152863,
      "rewards/forensics_iou_reward": 0.5428409576416016,
      "rewards/forensics_strict_edge_reward": 0.047611622139811516,
      "rewards/forensics_tiou_reward": 0.49552440643310547,
      "step": 848
    },
    {
      "completion_length": 24.40625,
      "epoch": 2.6052227342549923,
      "grad_norm": 6.586092472076416,
      "kl": 0.4267578125,
      "learning_rate": 3.476923076923077e-07,
      "loss": 0.0263,
      "reward": 0.7782400250434875,
      "reward_std": 0.049036160577088594,
      "reward_std_all": 0.0440469766035676,
      "rewards-max/forensics_iou_reward": 0.988698273897171,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.985660195350647,
      "rewards/forensics_iou_reward": 0.7391799092292786,
      "rewards/forensics_strict_edge_reward": 0.07302867621183395,
      "rewards/forensics_tiou_reward": 0.7026739716529846,
      "step": 849
    },
    {
      "completion_length": 21.171875,
      "epoch": 2.608294930875576,
      "grad_norm": 7.289525508880615,
      "kl": 0.384765625,
      "learning_rate": 3.4692307692307694e-07,
      "loss": 0.0673,
      "reward": 0.7883538603782654,
      "reward_std": 0.023919397965073586,
      "reward_std_all": 0.023447119630873203,
      "rewards-max/forensics_iou_reward": 0.9837725758552551,
      "rewards-max/forensics_strict_edge_reward": 0.5931393504142761,
      "rewards-max/forensics_tiou_reward": 0.9777780175209045,
      "rewards/forensics_iou_reward": 0.7179355323314667,
      "rewards/forensics_strict_edge_reward": 0.07801688648760319,
      "rewards/forensics_tiou_reward": 0.6529169678688049,
      "step": 850
    },
    {
      "completion_length": 22.796875,
      "epoch": 2.61136712749616,
      "grad_norm": 6.261707305908203,
      "kl": 0.4169921875,
      "learning_rate": 3.461538461538461e-07,
      "loss": 0.0575,
      "reward": 0.7538490295410156,
      "reward_std": 0.0657455325126648,
      "reward_std_all": 0.06283775344491005,
      "rewards-max/forensics_iou_reward": 0.9629983305931091,
      "rewards-max/forensics_strict_edge_reward": 0.5031317323446274,
      "rewards-max/forensics_tiou_reward": 0.946690559387207,
      "rewards/forensics_iou_reward": 0.6319113075733185,
      "rewards/forensics_strict_edge_reward": 0.056456558406353,
      "rewards/forensics_tiou_reward": 0.5874131917953491,
      "step": 851
    },
    {
      "completion_length": 27.265625,
      "epoch": 2.6144393241167436,
      "grad_norm": 7.342206001281738,
      "kl": 0.5166015625,
      "learning_rate": 3.4538461538461537e-07,
      "loss": 0.1111,
      "reward": 0.7574250102043152,
      "reward_std": 0.06233329093083739,
      "reward_std_all": 0.05979485251009464,
      "rewards-max/forensics_iou_reward": 0.9733401834964752,
      "rewards-max/forensics_strict_edge_reward": 0.3157382905483246,
      "rewards-max/forensics_tiou_reward": 0.9628272652626038,
      "rewards/forensics_iou_reward": 0.5952125191688538,
      "rewards/forensics_strict_edge_reward": 0.022153940983116627,
      "rewards/forensics_tiou_reward": 0.5462013781070709,
      "step": 852
    },
    {
      "completion_length": 27.65625,
      "epoch": 2.6175115207373274,
      "grad_norm": 6.411808967590332,
      "kl": 0.52734375,
      "learning_rate": 3.446153846153846e-07,
      "loss": 0.2042,
      "reward": 0.7294881641864777,
      "reward_std": 0.10413189604878426,
      "reward_std_all": 0.09550974518060684,
      "rewards-max/forensics_iou_reward": 0.9885560572147369,
      "rewards-max/forensics_strict_edge_reward": 0.7055691182613373,
      "rewards-max/forensics_tiou_reward": 0.984439879655838,
      "rewards/forensics_iou_reward": 0.6119088232517242,
      "rewards/forensics_strict_edge_reward": 0.06268180720508099,
      "rewards/forensics_tiou_reward": 0.565582275390625,
      "step": 853
    },
    {
      "completion_length": 27.625,
      "epoch": 2.6205837173579107,
      "grad_norm": 6.814645290374756,
      "kl": 0.4482421875,
      "learning_rate": 3.4384615384615385e-07,
      "loss": -0.1026,
      "reward": 0.7619608640670776,
      "reward_std": 0.05000345781445503,
      "reward_std_all": 0.04861089028418064,
      "rewards-max/forensics_iou_reward": 0.9875693321228027,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9838093221187592,
      "rewards/forensics_iou_reward": 0.6944141983985901,
      "rewards/forensics_strict_edge_reward": 0.08078432455658913,
      "rewards/forensics_tiou_reward": 0.648807018995285,
      "step": 854
    },
    {
      "completion_length": 23.4375,
      "epoch": 2.6236559139784945,
      "grad_norm": 9.839612007141113,
      "kl": 0.421875,
      "learning_rate": 3.4307692307692307e-07,
      "loss": 0.0765,
      "reward": 0.686882734298706,
      "reward_std": 0.05889146588742733,
      "reward_std_all": 0.08569186553359032,
      "rewards-max/forensics_iou_reward": 0.9779822826385498,
      "rewards-max/forensics_strict_edge_reward": 0.6965087056159973,
      "rewards-max/forensics_tiou_reward": 0.9726079404354095,
      "rewards/forensics_iou_reward": 0.6026560962200165,
      "rewards/forensics_strict_edge_reward": 0.04746874421834946,
      "rewards/forensics_tiou_reward": 0.556708574295044,
      "step": 855
    },
    {
      "completion_length": 26.0625,
      "epoch": 2.6267281105990783,
      "grad_norm": 8.503473281860352,
      "kl": 0.435546875,
      "learning_rate": 3.423076923076923e-07,
      "loss": 0.0024,
      "reward": 0.7676400244235992,
      "reward_std": 0.04247916303575039,
      "reward_std_all": 0.04258854500949383,
      "rewards-max/forensics_iou_reward": 0.946672797203064,
      "rewards-max/forensics_strict_edge_reward": 0.34257786720991135,
      "rewards-max/forensics_tiou_reward": 0.9266908168792725,
      "rewards/forensics_iou_reward": 0.6068524420261383,
      "rewards/forensics_strict_edge_reward": 0.031218924093991518,
      "rewards/forensics_tiou_reward": 0.5462542474269867,
      "step": 856
    },
    {
      "completion_length": 24.046875,
      "epoch": 2.629800307219662,
      "grad_norm": 10.823972702026367,
      "kl": 0.4619140625,
      "learning_rate": 3.415384615384615e-07,
      "loss": 0.0993,
      "reward": 0.7796507775783539,
      "reward_std": 0.03428862616419792,
      "reward_std_all": 0.03290488012135029,
      "rewards-max/forensics_iou_reward": 0.994140625,
      "rewards-max/forensics_strict_edge_reward": 0.8704091012477875,
      "rewards-max/forensics_tiou_reward": 0.9930572509765625,
      "rewards/forensics_iou_reward": 0.6378234624862671,
      "rewards/forensics_strict_edge_reward": 0.09690752252936363,
      "rewards/forensics_tiou_reward": 0.588411808013916,
      "step": 857
    },
    {
      "completion_length": 27.46875,
      "epoch": 2.632872503840246,
      "grad_norm": 7.453158378601074,
      "kl": 0.4951171875,
      "learning_rate": 3.4076923076923076e-07,
      "loss": 0.0784,
      "reward": 0.6656877398490906,
      "reward_std": 0.056991470977663994,
      "reward_std_all": 0.10023250430822372,
      "rewards-max/forensics_iou_reward": 0.9277781546115875,
      "rewards-max/forensics_strict_edge_reward": 0.4618668928742409,
      "rewards-max/forensics_tiou_reward": 0.9104092121124268,
      "rewards/forensics_iou_reward": 0.5110113620758057,
      "rewards/forensics_strict_edge_reward": 0.039875549264252186,
      "rewards/forensics_tiou_reward": 0.47254735231399536,
      "step": 858
    },
    {
      "completion_length": 25.109375,
      "epoch": 2.6359447004608296,
      "grad_norm": 6.998828411102295,
      "kl": 0.4404296875,
      "learning_rate": 3.4000000000000003e-07,
      "loss": 0.0785,
      "reward": 0.7151376605033875,
      "reward_std": 0.07507813349366188,
      "reward_std_all": 0.0898880660533905,
      "rewards-max/forensics_iou_reward": 0.9617562294006348,
      "rewards-max/forensics_strict_edge_reward": 0.49686121940612793,
      "rewards-max/forensics_tiou_reward": 0.9519914090633392,
      "rewards/forensics_iou_reward": 0.6474761366844177,
      "rewards/forensics_strict_edge_reward": 0.03906339965760708,
      "rewards/forensics_tiou_reward": 0.6089489161968231,
      "step": 859
    },
    {
      "completion_length": 29.171875,
      "epoch": 2.6390168970814134,
      "grad_norm": 9.436544418334961,
      "kl": 0.3837890625,
      "learning_rate": 3.392307692307692e-07,
      "loss": 0.0319,
      "reward": 0.6924847960472107,
      "reward_std": 0.04025392699986696,
      "reward_std_all": 0.07503937184810638,
      "rewards-max/forensics_iou_reward": 0.9414839148521423,
      "rewards-max/forensics_strict_edge_reward": 0.19446904212236404,
      "rewards-max/forensics_tiou_reward": 0.9316657185554504,
      "rewards/forensics_iou_reward": 0.591220498085022,
      "rewards/forensics_strict_edge_reward": 0.019342048559337854,
      "rewards/forensics_tiou_reward": 0.552214503288269,
      "step": 860
    },
    {
      "completion_length": 24.203125,
      "epoch": 2.6420890937019967,
      "grad_norm": 8.578768730163574,
      "kl": 0.4306640625,
      "learning_rate": 3.3846153846153845e-07,
      "loss": 0.0402,
      "reward": 0.7733163237571716,
      "reward_std": 0.044767238199710846,
      "reward_std_all": 0.04240783676505089,
      "rewards-max/forensics_iou_reward": 0.9583431780338287,
      "rewards-max/forensics_strict_edge_reward": 0.2612588554620743,
      "rewards-max/forensics_tiou_reward": 0.9486380815505981,
      "rewards/forensics_iou_reward": 0.6772263050079346,
      "rewards/forensics_strict_edge_reward": 0.032501320354640484,
      "rewards/forensics_tiou_reward": 0.6294361650943756,
      "step": 861
    },
    {
      "completion_length": 24.140625,
      "epoch": 2.6451612903225805,
      "grad_norm": 8.220610618591309,
      "kl": 0.5029296875,
      "learning_rate": 3.3769230769230767e-07,
      "loss": 0.132,
      "reward": 0.6685587167739868,
      "reward_std": 0.08489101007580757,
      "reward_std_all": 0.11694605648517609,
      "rewards-max/forensics_iou_reward": 0.9827239513397217,
      "rewards-max/forensics_strict_edge_reward": 0.5521765947341919,
      "rewards-max/forensics_tiou_reward": 0.9798007905483246,
      "rewards/forensics_iou_reward": 0.5721638798713684,
      "rewards/forensics_strict_edge_reward": 0.07015050202608109,
      "rewards/forensics_tiou_reward": 0.5369892418384552,
      "step": 862
    },
    {
      "completion_length": 23.6875,
      "epoch": 2.6482334869431643,
      "grad_norm": 5.801358222961426,
      "kl": 0.3984375,
      "learning_rate": 3.3692307692307693e-07,
      "loss": -0.018,
      "reward": 0.761396735906601,
      "reward_std": 0.061813101172447205,
      "reward_std_all": 0.058140337467193604,
      "rewards-max/forensics_iou_reward": 0.990622878074646,
      "rewards-max/forensics_strict_edge_reward": 0.31668807566165924,
      "rewards-max/forensics_tiou_reward": 0.9870409071445465,
      "rewards/forensics_iou_reward": 0.7235338389873505,
      "rewards/forensics_strict_edge_reward": 0.03319538105279207,
      "rewards/forensics_tiou_reward": 0.6779819428920746,
      "step": 863
    },
    {
      "completion_length": 27.109375,
      "epoch": 2.651305683563748,
      "grad_norm": 7.351811408996582,
      "kl": 0.529296875,
      "learning_rate": 3.361538461538461e-07,
      "loss": -0.0127,
      "reward": 0.7838671803474426,
      "reward_std": 0.03229132352862507,
      "reward_std_all": 0.030560373794287443,
      "rewards-max/forensics_iou_reward": 0.9950562119483948,
      "rewards-max/forensics_strict_edge_reward": 0.6151081919670105,
      "rewards-max/forensics_tiou_reward": 0.9941422045230865,
      "rewards/forensics_iou_reward": 0.7387442886829376,
      "rewards/forensics_strict_edge_reward": 0.08453135192394257,
      "rewards/forensics_tiou_reward": 0.6954982876777649,
      "step": 864
    },
    {
      "completion_length": 23.140625,
      "epoch": 2.654377880184332,
      "grad_norm": 9.664400100708008,
      "kl": 0.302734375,
      "learning_rate": 3.3538461538461536e-07,
      "loss": 0.1176,
      "reward": 0.7328985333442688,
      "reward_std": 0.07307539880275726,
      "reward_std_all": 0.07272317633032799,
      "rewards-max/forensics_iou_reward": 0.9882466495037079,
      "rewards-max/forensics_strict_edge_reward": 0.6952291429042816,
      "rewards-max/forensics_tiou_reward": 0.9839659035205841,
      "rewards/forensics_iou_reward": 0.6085675060749054,
      "rewards/forensics_strict_edge_reward": 0.06165000610053539,
      "rewards/forensics_tiou_reward": 0.5424724221229553,
      "step": 865
    },
    {
      "completion_length": 25.703125,
      "epoch": 2.6574500768049156,
      "grad_norm": 7.300561428070068,
      "kl": 0.3115234375,
      "learning_rate": 3.346153846153846e-07,
      "loss": -0.0039,
      "reward": 0.751451849937439,
      "reward_std": 0.037664078176021576,
      "reward_std_all": 0.045712387189269066,
      "rewards-max/forensics_iou_reward": 0.9798847734928131,
      "rewards-max/forensics_strict_edge_reward": 0.51341712474823,
      "rewards-max/forensics_tiou_reward": 0.9728696644306183,
      "rewards/forensics_iou_reward": 0.694745272397995,
      "rewards/forensics_strict_edge_reward": 0.06587615422904491,
      "rewards/forensics_tiou_reward": 0.6608645915985107,
      "step": 866
    },
    {
      "completion_length": 25.453125,
      "epoch": 2.6605222734254994,
      "grad_norm": 6.465716361999512,
      "kl": 0.4189453125,
      "learning_rate": 3.3384615384615384e-07,
      "loss": 0.0518,
      "reward": 0.6970037817955017,
      "reward_std": 0.08604399859905243,
      "reward_std_all": 0.10219182074069977,
      "rewards-max/forensics_iou_reward": 0.9667131900787354,
      "rewards-max/forensics_strict_edge_reward": 0.5201957523822784,
      "rewards-max/forensics_tiou_reward": 0.9623494744300842,
      "rewards/forensics_iou_reward": 0.5826675891876221,
      "rewards/forensics_strict_edge_reward": 0.04014075547456741,
      "rewards/forensics_tiou_reward": 0.537729024887085,
      "step": 867
    },
    {
      "completion_length": 27.34375,
      "epoch": 2.6635944700460827,
      "grad_norm": 7.138583660125732,
      "kl": 0.3525390625,
      "learning_rate": 3.330769230769231e-07,
      "loss": -0.0024,
      "reward": 0.7436958849430084,
      "reward_std": 0.028255432844161987,
      "reward_std_all": 0.04942585714161396,
      "rewards-max/forensics_iou_reward": 0.992348849773407,
      "rewards-max/forensics_strict_edge_reward": 0.37146827578544617,
      "rewards-max/forensics_tiou_reward": 0.9875521659851074,
      "rewards/forensics_iou_reward": 0.6565606594085693,
      "rewards/forensics_strict_edge_reward": 0.0365336649119854,
      "rewards/forensics_tiou_reward": 0.6102955043315887,
      "step": 868
    },
    {
      "completion_length": 26.015625,
      "epoch": 2.6666666666666665,
      "grad_norm": 7.585115432739258,
      "kl": 0.470703125,
      "learning_rate": 3.3230769230769227e-07,
      "loss": 0.007,
      "reward": 0.6997724771499634,
      "reward_std": 0.06080034375190735,
      "reward_std_all": 0.08264121413230896,
      "rewards-max/forensics_iou_reward": 0.9585681557655334,
      "rewards-max/forensics_strict_edge_reward": 0.38056008517742157,
      "rewards-max/forensics_tiou_reward": 0.9522231817245483,
      "rewards/forensics_iou_reward": 0.6598618030548096,
      "rewards/forensics_strict_edge_reward": 0.03582155145704746,
      "rewards/forensics_tiou_reward": 0.6241903603076935,
      "step": 869
    },
    {
      "completion_length": 24.71875,
      "epoch": 2.6697388632872503,
      "grad_norm": 6.088283538818359,
      "kl": 0.548828125,
      "learning_rate": 3.3153846153846154e-07,
      "loss": 0.0136,
      "reward": 0.7605848908424377,
      "reward_std": 0.07873491011559963,
      "reward_std_all": 0.06995199806988239,
      "rewards-max/forensics_iou_reward": 0.9938760995864868,
      "rewards-max/forensics_strict_edge_reward": 0.4346180260181427,
      "rewards-max/forensics_tiou_reward": 0.9914463758468628,
      "rewards/forensics_iou_reward": 0.6969775855541229,
      "rewards/forensics_strict_edge_reward": 0.08120676688849926,
      "rewards/forensics_tiou_reward": 0.6514353454113007,
      "step": 870
    },
    {
      "completion_length": 28.59375,
      "epoch": 2.672811059907834,
      "grad_norm": 44.607452392578125,
      "kl": 2.5771484375,
      "learning_rate": 3.3076923076923075e-07,
      "loss": 0.1308,
      "reward": 0.6850443482398987,
      "reward_std": 0.06802203319966793,
      "reward_std_all": 0.09871730208396912,
      "rewards-max/forensics_iou_reward": 0.9697004854679108,
      "rewards-max/forensics_strict_edge_reward": 0.3946065902709961,
      "rewards-max/forensics_tiou_reward": 0.9583593606948853,
      "rewards/forensics_iou_reward": 0.5703112483024597,
      "rewards/forensics_strict_edge_reward": 0.04657500609755516,
      "rewards/forensics_tiou_reward": 0.5288245379924774,
      "step": 871
    },
    {
      "completion_length": 22.59375,
      "epoch": 2.675883256528418,
      "grad_norm": 9.624167442321777,
      "kl": 0.349609375,
      "learning_rate": 3.3e-07,
      "loss": 0.0152,
      "reward": 0.6340622305870056,
      "reward_std": 0.12575366348028183,
      "reward_std_all": 0.14588724821805954,
      "rewards-max/forensics_iou_reward": 0.9565747976303101,
      "rewards-max/forensics_strict_edge_reward": 0.575356051325798,
      "rewards-max/forensics_tiou_reward": 0.936583399772644,
      "rewards/forensics_iou_reward": 0.5667822062969208,
      "rewards/forensics_strict_edge_reward": 0.07968987803906202,
      "rewards/forensics_tiou_reward": 0.5323996543884277,
      "step": 872
    },
    {
      "completion_length": 23.65625,
      "epoch": 2.6789554531490016,
      "grad_norm": 8.221134185791016,
      "kl": 0.48828125,
      "learning_rate": 3.292307692307692e-07,
      "loss": 0.0066,
      "reward": 0.7892253398895264,
      "reward_std": 0.02911765454337001,
      "reward_std_all": 0.027166794752702117,
      "rewards-max/forensics_iou_reward": 0.9887798130512238,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.9848812818527222,
      "rewards/forensics_iou_reward": 0.7104445397853851,
      "rewards/forensics_strict_edge_reward": 0.08368382975459099,
      "rewards/forensics_tiou_reward": 0.6680516600608826,
      "step": 873
    },
    {
      "completion_length": 25.09375,
      "epoch": 2.6820276497695854,
      "grad_norm": 7.429564476013184,
      "kl": 0.4140625,
      "learning_rate": 3.2846153846153844e-07,
      "loss": -0.0204,
      "reward": 0.7302161157131195,
      "reward_std": 0.08893406018614769,
      "reward_std_all": 0.08447111956775188,
      "rewards-max/forensics_iou_reward": 0.9468004107475281,
      "rewards-max/forensics_strict_edge_reward": 0.3940344303846359,
      "rewards-max/forensics_tiou_reward": 0.9364750385284424,
      "rewards/forensics_iou_reward": 0.6266844272613525,
      "rewards/forensics_strict_edge_reward": 0.028434330597519875,
      "rewards/forensics_tiou_reward": 0.5779748857021332,
      "step": 874
    },
    {
      "completion_length": 26.046875,
      "epoch": 2.6850998463901687,
      "grad_norm": 9.405591011047363,
      "kl": 0.4072265625,
      "learning_rate": 3.2769230769230766e-07,
      "loss": 0.0985,
      "reward": 0.7389892637729645,
      "reward_std": 0.023970041424036026,
      "reward_std_all": 0.04808466508984566,
      "rewards-max/forensics_iou_reward": 0.9647475183010101,
      "rewards-max/forensics_strict_edge_reward": 0.32594865560531616,
      "rewards-max/forensics_tiou_reward": 0.9531941711902618,
      "rewards/forensics_iou_reward": 0.5934782922267914,
      "rewards/forensics_strict_edge_reward": 0.03493691608309746,
      "rewards/forensics_tiou_reward": 0.5407915413379669,
      "step": 875
    },
    {
      "completion_length": 28.265625,
      "epoch": 2.688172043010753,
      "grad_norm": 6.022268295288086,
      "kl": 0.5390625,
      "learning_rate": 3.269230769230769e-07,
      "loss": 0.0731,
      "reward": 0.7031719386577606,
      "reward_std": 0.10043404251337051,
      "reward_std_all": 0.10076761990785599,
      "rewards-max/forensics_iou_reward": 0.9688895046710968,
      "rewards-max/forensics_strict_edge_reward": 0.34546563774347305,
      "rewards-max/forensics_tiou_reward": 0.9447320699691772,
      "rewards/forensics_iou_reward": 0.5900837779045105,
      "rewards/forensics_strict_edge_reward": 0.022001863457262516,
      "rewards/forensics_tiou_reward": 0.537966325879097,
      "step": 876
    },
    {
      "completion_length": 22.671875,
      "epoch": 2.6912442396313363,
      "grad_norm": 7.679985046386719,
      "kl": 0.5,
      "learning_rate": 3.2615384615384614e-07,
      "loss": -0.0061,
      "reward": 0.7557604312896729,
      "reward_std": 0.006223017117008567,
      "reward_std_all": 0.030503838788717985,
      "rewards-max/forensics_iou_reward": 0.9764139652252197,
      "rewards-max/forensics_strict_edge_reward": 0.5702531039714813,
      "rewards-max/forensics_tiou_reward": 0.9697936177253723,
      "rewards/forensics_iou_reward": 0.704531341791153,
      "rewards/forensics_strict_edge_reward": 0.040729052387177944,
      "rewards/forensics_tiou_reward": 0.6583465933799744,
      "step": 877
    },
    {
      "completion_length": 22.96875,
      "epoch": 2.69431643625192,
      "grad_norm": 6.8623175621032715,
      "kl": 0.4658203125,
      "learning_rate": 3.2538461538461535e-07,
      "loss": 0.118,
      "reward": 0.7671217918395996,
      "reward_std": 0.06276625581085682,
      "reward_std_all": 0.05684700235724449,
      "rewards-max/forensics_iou_reward": 0.9768031239509583,
      "rewards-max/forensics_strict_edge_reward": 0.43940845131874084,
      "rewards-max/forensics_tiou_reward": 0.9657179713249207,
      "rewards/forensics_iou_reward": 0.6812449097633362,
      "rewards/forensics_strict_edge_reward": 0.049943313002586365,
      "rewards/forensics_tiou_reward": 0.6326550841331482,
      "step": 878
    },
    {
      "completion_length": 25.671875,
      "epoch": 2.697388632872504,
      "grad_norm": 6.627684116363525,
      "kl": 0.3447265625,
      "learning_rate": 3.246153846153846e-07,
      "loss": 0.0643,
      "reward": 0.7955822944641113,
      "reward_std": 0.016956686216872185,
      "reward_std_all": 0.017011765157803893,
      "rewards-max/forensics_iou_reward": 0.9335800111293793,
      "rewards-max/forensics_strict_edge_reward": 0.6470369100570679,
      "rewards-max/forensics_tiou_reward": 0.9224434792995453,
      "rewards/forensics_iou_reward": 0.6832915842533112,
      "rewards/forensics_strict_edge_reward": 0.0627684686332941,
      "rewards/forensics_tiou_reward": 0.629639744758606,
      "step": 879
    },
    {
      "completion_length": 30.640625,
      "epoch": 2.7004608294930876,
      "grad_norm": 8.33450984954834,
      "kl": 0.41796875,
      "learning_rate": 3.2384615384615383e-07,
      "loss": -0.035,
      "reward": 0.6998040974140167,
      "reward_std": 0.06798087432980537,
      "reward_std_all": 0.08874000608921051,
      "rewards-max/forensics_iou_reward": 0.9548526108264923,
      "rewards-max/forensics_strict_edge_reward": 0.32213714718818665,
      "rewards-max/forensics_tiou_reward": 0.9375849664211273,
      "rewards/forensics_iou_reward": 0.6260171830654144,
      "rewards/forensics_strict_edge_reward": 0.043912144377827644,
      "rewards/forensics_tiou_reward": 0.584968775510788,
      "step": 880
    },
    {
      "completion_length": 24.5625,
      "epoch": 2.7035330261136714,
      "grad_norm": 5.820157051086426,
      "kl": 0.4228515625,
      "learning_rate": 3.230769230769231e-07,
      "loss": 0.127,
      "reward": 0.6958608329296112,
      "reward_std": 0.15503714233636856,
      "reward_std_all": 0.14128657802939415,
      "rewards-max/forensics_iou_reward": 0.97279092669487,
      "rewards-max/forensics_strict_edge_reward": 0.41257381439208984,
      "rewards-max/forensics_tiou_reward": 0.9647817313671112,
      "rewards/forensics_iou_reward": 0.6128769814968109,
      "rewards/forensics_strict_edge_reward": 0.07112188264727592,
      "rewards/forensics_tiou_reward": 0.5692967772483826,
      "step": 881
    },
    {
      "completion_length": 27.328125,
      "epoch": 2.7066052227342547,
      "grad_norm": 6.765076160430908,
      "kl": 0.482421875,
      "learning_rate": 3.2230769230769226e-07,
      "loss": -0.0436,
      "reward": 0.7250678539276123,
      "reward_std": 0.050921766087412834,
      "reward_std_all": 0.0631510354578495,
      "rewards-max/forensics_iou_reward": 0.9833134710788727,
      "rewards-max/forensics_strict_edge_reward": 0.6526533365249634,
      "rewards-max/forensics_tiou_reward": 0.9783625304698944,
      "rewards/forensics_iou_reward": 0.6560807228088379,
      "rewards/forensics_strict_edge_reward": 0.07487649470567703,
      "rewards/forensics_tiou_reward": 0.6147456765174866,
      "step": 882
    },
    {
      "completion_length": 27.421875,
      "epoch": 2.709677419354839,
      "grad_norm": 5.760947227478027,
      "kl": 0.3505859375,
      "learning_rate": 3.215384615384615e-07,
      "loss": 0.097,
      "reward": 0.7547538876533508,
      "reward_std": 0.05941379815340042,
      "reward_std_all": 0.05724918469786644,
      "rewards-max/forensics_iou_reward": 0.9834101796150208,
      "rewards-max/forensics_strict_edge_reward": 0.5304600298404694,
      "rewards-max/forensics_tiou_reward": 0.9775664806365967,
      "rewards/forensics_iou_reward": 0.6947034597396851,
      "rewards/forensics_strict_edge_reward": 0.07412349805235863,
      "rewards/forensics_tiou_reward": 0.6567357182502747,
      "step": 883
    },
    {
      "completion_length": 29.671875,
      "epoch": 2.7127496159754223,
      "grad_norm": 5.2492475509643555,
      "kl": 0.435546875,
      "learning_rate": 3.2076923076923074e-07,
      "loss": 0.1288,
      "reward": 0.7327496111392975,
      "reward_std": 0.04013501945883036,
      "reward_std_all": 0.0565862562507391,
      "rewards-max/forensics_iou_reward": 0.9650344550609589,
      "rewards-max/forensics_strict_edge_reward": 0.3778143525123596,
      "rewards-max/forensics_tiou_reward": 0.9508911669254303,
      "rewards/forensics_iou_reward": 0.6346573233604431,
      "rewards/forensics_strict_edge_reward": 0.0526515394449234,
      "rewards/forensics_tiou_reward": 0.5919855535030365,
      "step": 884
    },
    {
      "completion_length": 28.859375,
      "epoch": 2.715821812596006,
      "grad_norm": 7.132052421569824,
      "kl": 0.533203125,
      "learning_rate": 3.2e-07,
      "loss": 0.1552,
      "reward": 0.6925901770591736,
      "reward_std": 0.06275083497166634,
      "reward_std_all": 0.08701624348759651,
      "rewards-max/forensics_iou_reward": 0.9514959156513214,
      "rewards-max/forensics_strict_edge_reward": 0.3397350013256073,
      "rewards-max/forensics_tiou_reward": 0.9417948126792908,
      "rewards/forensics_iou_reward": 0.5648084431886673,
      "rewards/forensics_strict_edge_reward": 0.04634745605289936,
      "rewards/forensics_tiou_reward": 0.5182072967290878,
      "step": 885
    },
    {
      "completion_length": 27.84375,
      "epoch": 2.71889400921659,
      "grad_norm": 5.922994613647461,
      "kl": 0.3271484375,
      "learning_rate": 3.1923076923076917e-07,
      "loss": 0.0804,
      "reward": 0.7518614232540131,
      "reward_std": 0.07364814728498459,
      "reward_std_all": 0.06848889589309692,
      "rewards-max/forensics_iou_reward": 0.9917326867580414,
      "rewards-max/forensics_strict_edge_reward": 0.7418042421340942,
      "rewards-max/forensics_tiou_reward": 0.987036794424057,
      "rewards/forensics_iou_reward": 0.6526628136634827,
      "rewards/forensics_strict_edge_reward": 0.0689922533929348,
      "rewards/forensics_tiou_reward": 0.6062372326850891,
      "step": 886
    },
    {
      "completion_length": 24.5625,
      "epoch": 2.7219662058371736,
      "grad_norm": 9.564075469970703,
      "kl": 0.4716796875,
      "learning_rate": 3.1846153846153843e-07,
      "loss": -0.0251,
      "reward": 0.7159976065158844,
      "reward_std": 0.05529842618852854,
      "reward_std_all": 0.07447627559304237,
      "rewards-max/forensics_iou_reward": 0.9833242893218994,
      "rewards-max/forensics_strict_edge_reward": 0.6326714754104614,
      "rewards-max/forensics_tiou_reward": 0.979050487279892,
      "rewards/forensics_iou_reward": 0.6180408298969269,
      "rewards/forensics_strict_edge_reward": 0.04944186843931675,
      "rewards/forensics_tiou_reward": 0.5544497221708298,
      "step": 887
    },
    {
      "completion_length": 22.609375,
      "epoch": 2.7250384024577574,
      "grad_norm": 7.372291088104248,
      "kl": 0.4443359375,
      "learning_rate": 3.176923076923077e-07,
      "loss": -0.0229,
      "reward": 0.7360760569572449,
      "reward_std": 0.07342511415481567,
      "reward_std_all": 0.07238973304629326,
      "rewards-max/forensics_iou_reward": 0.9900648593902588,
      "rewards-max/forensics_strict_edge_reward": 0.8260191679000854,
      "rewards-max/forensics_tiou_reward": 0.9871044456958771,
      "rewards/forensics_iou_reward": 0.6890160441398621,
      "rewards/forensics_strict_edge_reward": 0.08843842893838882,
      "rewards/forensics_tiou_reward": 0.6475245654582977,
      "step": 888
    },
    {
      "completion_length": 24.734375,
      "epoch": 2.7281105990783407,
      "grad_norm": 7.577263355255127,
      "kl": 0.6875,
      "learning_rate": 3.169230769230769e-07,
      "loss": 0.1223,
      "reward": 0.747745543718338,
      "reward_std": 0.07687727361917496,
      "reward_std_all": 0.07158002257347107,
      "rewards-max/forensics_iou_reward": 0.9592983424663544,
      "rewards-max/forensics_strict_edge_reward": 0.6083278059959412,
      "rewards-max/forensics_tiou_reward": 0.947455108165741,
      "rewards/forensics_iou_reward": 0.6361099779605865,
      "rewards/forensics_strict_edge_reward": 0.08449195697903633,
      "rewards/forensics_tiou_reward": 0.5788044929504395,
      "step": 889
    },
    {
      "completion_length": 25.890625,
      "epoch": 2.731182795698925,
      "grad_norm": 6.1780686378479,
      "kl": 0.42578125,
      "learning_rate": 3.161538461538462e-07,
      "loss": 0.0836,
      "reward": 0.7388617992401123,
      "reward_std": 0.06941646710038185,
      "reward_std_all": 0.06806071102619171,
      "rewards-max/forensics_iou_reward": 0.9472042620182037,
      "rewards-max/forensics_strict_edge_reward": 0.3523929864168167,
      "rewards-max/forensics_tiou_reward": 0.9334251582622528,
      "rewards/forensics_iou_reward": 0.5978904962539673,
      "rewards/forensics_strict_edge_reward": 0.03134967852383852,
      "rewards/forensics_tiou_reward": 0.5437687635421753,
      "step": 890
    },
    {
      "completion_length": 28.9375,
      "epoch": 2.7342549923195083,
      "grad_norm": 5.991843223571777,
      "kl": 0.5302734375,
      "learning_rate": 3.1538461538461534e-07,
      "loss": 0.0677,
      "reward": 0.7144355773925781,
      "reward_std": 0.06955966865643859,
      "reward_std_all": 0.08604225143790245,
      "rewards-max/forensics_iou_reward": 0.9808050692081451,
      "rewards-max/forensics_strict_edge_reward": 0.588684618473053,
      "rewards-max/forensics_tiou_reward": 0.9736176133155823,
      "rewards/forensics_iou_reward": 0.6441166996955872,
      "rewards/forensics_strict_edge_reward": 0.08564302511513233,
      "rewards/forensics_tiou_reward": 0.6047617495059967,
      "step": 891
    },
    {
      "completion_length": 25.90625,
      "epoch": 2.737327188940092,
      "grad_norm": 8.94775104522705,
      "kl": 0.423828125,
      "learning_rate": 3.146153846153846e-07,
      "loss": 0.1417,
      "reward": 0.7276463210582733,
      "reward_std": 0.09061996638774872,
      "reward_std_all": 0.08668739348649979,
      "rewards-max/forensics_iou_reward": 0.9880987703800201,
      "rewards-max/forensics_strict_edge_reward": 0.7610765993595123,
      "rewards-max/forensics_tiou_reward": 0.9822493493556976,
      "rewards/forensics_iou_reward": 0.6363347470760345,
      "rewards/forensics_strict_edge_reward": 0.05784839391708374,
      "rewards/forensics_tiou_reward": 0.5854899883270264,
      "step": 892
    },
    {
      "completion_length": 22.6875,
      "epoch": 2.740399385560676,
      "grad_norm": 6.691854476928711,
      "kl": 0.4091796875,
      "learning_rate": 3.138461538461538e-07,
      "loss": 0.1115,
      "reward": 0.7687439620494843,
      "reward_std": 0.06479676440358162,
      "reward_std_all": 0.057861159555613995,
      "rewards-max/forensics_iou_reward": 0.9626097083091736,
      "rewards-max/forensics_strict_edge_reward": 0.5319994986057281,
      "rewards-max/forensics_tiou_reward": 0.9517742097377777,
      "rewards/forensics_iou_reward": 0.7187656760215759,
      "rewards/forensics_strict_edge_reward": 0.0507214218378067,
      "rewards/forensics_tiou_reward": 0.6628172397613525,
      "step": 893
    },
    {
      "completion_length": 28.046875,
      "epoch": 2.7434715821812596,
      "grad_norm": 7.669386386871338,
      "kl": 0.8125,
      "learning_rate": 3.130769230769231e-07,
      "loss": 0.1073,
      "reward": 0.6912490725517273,
      "reward_std": 0.11997237801551819,
      "reward_std_all": 0.11637048795819283,
      "rewards-max/forensics_iou_reward": 0.9710240960121155,
      "rewards-max/forensics_strict_edge_reward": 0.42036955058574677,
      "rewards-max/forensics_tiou_reward": 0.9621224105358124,
      "rewards/forensics_iou_reward": 0.5708962678909302,
      "rewards/forensics_strict_edge_reward": 0.039273710921406746,
      "rewards/forensics_tiou_reward": 0.5311077982187271,
      "step": 894
    },
    {
      "completion_length": 27.3125,
      "epoch": 2.7465437788018434,
      "grad_norm": 6.062201976776123,
      "kl": 0.521484375,
      "learning_rate": 3.1230769230769225e-07,
      "loss": 0.0496,
      "reward": 0.7873242795467377,
      "reward_std": 0.022480787243694067,
      "reward_std_all": 0.023136934731155634,
      "rewards-max/forensics_iou_reward": 0.9625928401947021,
      "rewards-max/forensics_strict_edge_reward": 0.21064262092113495,
      "rewards-max/forensics_tiou_reward": 0.9504390060901642,
      "rewards/forensics_iou_reward": 0.6690514087677002,
      "rewards/forensics_strict_edge_reward": 0.02720455452799797,
      "rewards/forensics_tiou_reward": 0.6209248006343842,
      "step": 895
    },
    {
      "completion_length": 26.96875,
      "epoch": 2.749615975422427,
      "grad_norm": 8.780461311340332,
      "kl": 0.4296875,
      "learning_rate": 3.115384615384615e-07,
      "loss": 0.0462,
      "reward": 0.7779543995857239,
      "reward_std": 0.0490210372954607,
      "reward_std_all": 0.04436399135738611,
      "rewards-max/forensics_iou_reward": 0.9924559891223907,
      "rewards-max/forensics_strict_edge_reward": 0.7046436965465546,
      "rewards-max/forensics_tiou_reward": 0.9894231259822845,
      "rewards/forensics_iou_reward": 0.7099418044090271,
      "rewards/forensics_strict_edge_reward": 0.09460839256644249,
      "rewards/forensics_tiou_reward": 0.6711284518241882,
      "step": 896
    },
    {
      "completion_length": 25.921875,
      "epoch": 2.752688172043011,
      "grad_norm": 9.638466835021973,
      "kl": 0.8671875,
      "learning_rate": 3.1076923076923073e-07,
      "loss": 0.0694,
      "reward": 0.7437321841716766,
      "reward_std": 0.0968032912351191,
      "reward_std_all": 0.08682613633573055,
      "rewards-max/forensics_iou_reward": 0.9835275709629059,
      "rewards-max/forensics_strict_edge_reward": 0.6419725120067596,
      "rewards-max/forensics_tiou_reward": 0.9769184589385986,
      "rewards/forensics_iou_reward": 0.72834512591362,
      "rewards/forensics_strict_edge_reward": 0.08417877554893494,
      "rewards/forensics_tiou_reward": 0.6953580379486084,
      "step": 897
    },
    {
      "completion_length": 24.328125,
      "epoch": 2.7557603686635943,
      "grad_norm": 8.024072647094727,
      "kl": 0.357421875,
      "learning_rate": 3.1e-07,
      "loss": 0.0672,
      "reward": 0.7238291501998901,
      "reward_std": 0.08471459150314331,
      "reward_std_all": 0.08428288623690605,
      "rewards-max/forensics_iou_reward": 0.9842554330825806,
      "rewards-max/forensics_strict_edge_reward": 0.6154739558696747,
      "rewards-max/forensics_tiou_reward": 0.9755086302757263,
      "rewards/forensics_iou_reward": 0.62088543176651,
      "rewards/forensics_strict_edge_reward": 0.06950123980641365,
      "rewards/forensics_tiou_reward": 0.5414051413536072,
      "step": 898
    },
    {
      "completion_length": 25.96875,
      "epoch": 2.758832565284178,
      "grad_norm": 6.568661689758301,
      "kl": 0.3974609375,
      "learning_rate": 3.0923076923076926e-07,
      "loss": 0.0369,
      "reward": 0.7268308699131012,
      "reward_std": 0.042982444167137146,
      "reward_std_all": 0.06360415183007717,
      "rewards-max/forensics_iou_reward": 0.9734523594379425,
      "rewards-max/forensics_strict_edge_reward": 0.5190997421741486,
      "rewards-max/forensics_tiou_reward": 0.9657281637191772,
      "rewards/forensics_iou_reward": 0.6614890992641449,
      "rewards/forensics_strict_edge_reward": 0.03614790178835392,
      "rewards/forensics_tiou_reward": 0.6176185309886932,
      "step": 899
    },
    {
      "completion_length": 27.09375,
      "epoch": 2.761904761904762,
      "grad_norm": 7.189581394195557,
      "kl": 0.451171875,
      "learning_rate": 3.084615384615384e-07,
      "loss": 0.0683,
      "reward": 0.760861724615097,
      "reward_std": 0.03068901877850294,
      "reward_std_all": 0.03907717578113079,
      "rewards-max/forensics_iou_reward": 0.9787425100803375,
      "rewards-max/forensics_strict_edge_reward": 0.6302271485328674,
      "rewards-max/forensics_tiou_reward": 0.973198264837265,
      "rewards/forensics_iou_reward": 0.664892315864563,
      "rewards/forensics_strict_edge_reward": 0.07210896164178848,
      "rewards/forensics_tiou_reward": 0.6082796454429626,
      "step": 900
    },
    {
      "completion_length": 24.796875,
      "epoch": 2.7649769585253456,
      "grad_norm": 6.146605968475342,
      "kl": 0.501953125,
      "learning_rate": 3.076923076923077e-07,
      "loss": 0.0375,
      "reward": 0.7409258484840393,
      "reward_std": 0.07514450699090958,
      "reward_std_all": 0.07236065901815891,
      "rewards-max/forensics_iou_reward": 0.9586817920207977,
      "rewards-max/forensics_strict_edge_reward": 0.43546412885189056,
      "rewards-max/forensics_tiou_reward": 0.9430316090583801,
      "rewards/forensics_iou_reward": 0.6466091275215149,
      "rewards/forensics_strict_edge_reward": 0.04591597616672516,
      "rewards/forensics_tiou_reward": 0.6002173125743866,
      "step": 901
    },
    {
      "completion_length": 26.34375,
      "epoch": 2.7680491551459294,
      "grad_norm": 6.494825839996338,
      "kl": 0.48046875,
      "learning_rate": 3.069230769230769e-07,
      "loss": 0.0336,
      "reward": 0.7586143314838409,
      "reward_std": 0.04962364211678505,
      "reward_std_all": 0.049898626282811165,
      "rewards-max/forensics_iou_reward": 0.9750633537769318,
      "rewards-max/forensics_strict_edge_reward": 0.5985779762268066,
      "rewards-max/forensics_tiou_reward": 0.9633528292179108,
      "rewards/forensics_iou_reward": 0.6520346999168396,
      "rewards/forensics_strict_edge_reward": 0.045962851494550705,
      "rewards/forensics_tiou_reward": 0.608405590057373,
      "step": 902
    },
    {
      "completion_length": 27.71875,
      "epoch": 2.771121351766513,
      "grad_norm": 6.668495178222656,
      "kl": 0.52734375,
      "learning_rate": 3.0615384615384617e-07,
      "loss": 0.1307,
      "reward": 0.7701463997364044,
      "reward_std": 0.047746967524290085,
      "reward_std_all": 0.04493769444525242,
      "rewards-max/forensics_iou_reward": 0.9663276970386505,
      "rewards-max/forensics_strict_edge_reward": 0.326495997607708,
      "rewards-max/forensics_tiou_reward": 0.9561835527420044,
      "rewards/forensics_iou_reward": 0.6822397112846375,
      "rewards/forensics_strict_edge_reward": 0.04195536207407713,
      "rewards/forensics_tiou_reward": 0.6268961727619171,
      "step": 903
    },
    {
      "completion_length": 30.25,
      "epoch": 2.774193548387097,
      "grad_norm": 5.912859916687012,
      "kl": 0.3251953125,
      "learning_rate": 3.0538461538461533e-07,
      "loss": 0.0746,
      "reward": 0.7667776942253113,
      "reward_std": 0.06350979208946228,
      "reward_std_all": 0.05752037838101387,
      "rewards-max/forensics_iou_reward": 0.9852729737758636,
      "rewards-max/forensics_strict_edge_reward": 0.7711026966571808,
      "rewards-max/forensics_tiou_reward": 0.9809050559997559,
      "rewards/forensics_iou_reward": 0.6558200716972351,
      "rewards/forensics_strict_edge_reward": 0.07303644344210625,
      "rewards/forensics_tiou_reward": 0.6079963147640228,
      "step": 904
    },
    {
      "completion_length": 27.921875,
      "epoch": 2.7772657450076803,
      "grad_norm": 7.28029727935791,
      "kl": 0.3681640625,
      "learning_rate": 3.046153846153846e-07,
      "loss": 0.1045,
      "reward": 0.7617859840393066,
      "reward_std": 0.05020125024020672,
      "reward_std_all": 0.04902426898479462,
      "rewards-max/forensics_iou_reward": 0.9427184462547302,
      "rewards-max/forensics_strict_edge_reward": 0.37697096541523933,
      "rewards-max/forensics_tiou_reward": 0.9212045073509216,
      "rewards/forensics_iou_reward": 0.6144179701805115,
      "rewards/forensics_strict_edge_reward": 0.022613544948399067,
      "rewards/forensics_tiou_reward": 0.5443194508552551,
      "step": 905
    },
    {
      "completion_length": 24.96875,
      "epoch": 2.780337941628264,
      "grad_norm": 5.8201518058776855,
      "kl": 0.4873046875,
      "learning_rate": 3.038461538461538e-07,
      "loss": -0.0884,
      "reward": 0.7569148242473602,
      "reward_std": 0.057421013712882996,
      "reward_std_all": 0.05520947836339474,
      "rewards-max/forensics_iou_reward": 0.9785237908363342,
      "rewards-max/forensics_strict_edge_reward": 0.3486747592687607,
      "rewards-max/forensics_tiou_reward": 0.9676265716552734,
      "rewards/forensics_iou_reward": 0.680716335773468,
      "rewards/forensics_strict_edge_reward": 0.06623777840286493,
      "rewards/forensics_tiou_reward": 0.6329208016395569,
      "step": 906
    },
    {
      "completion_length": 25.546875,
      "epoch": 2.783410138248848,
      "grad_norm": 7.011776924133301,
      "kl": 0.4140625,
      "learning_rate": 3.030769230769231e-07,
      "loss": 0.111,
      "reward": 0.7618140280246735,
      "reward_std": 0.04712178558111191,
      "reward_std_all": 0.046834275126457214,
      "rewards-max/forensics_iou_reward": 0.945544421672821,
      "rewards-max/forensics_strict_edge_reward": 0.4589935690164566,
      "rewards-max/forensics_tiou_reward": 0.9360890090465546,
      "rewards/forensics_iou_reward": 0.6186388731002808,
      "rewards/forensics_strict_edge_reward": 0.0378294475376606,
      "rewards/forensics_tiou_reward": 0.5718570649623871,
      "step": 907
    },
    {
      "completion_length": 24.828125,
      "epoch": 2.7864823348694316,
      "grad_norm": 7.80635404586792,
      "kl": 0.68359375,
      "learning_rate": 3.0230769230769235e-07,
      "loss": -0.0101,
      "reward": 0.6997381448745728,
      "reward_std": 0.07558324560523033,
      "reward_std_all": 0.09372961893677711,
      "rewards-max/forensics_iou_reward": 0.9918166399002075,
      "rewards-max/forensics_strict_edge_reward": 0.638425350189209,
      "rewards-max/forensics_tiou_reward": 0.9879861772060394,
      "rewards/forensics_iou_reward": 0.6220198571681976,
      "rewards/forensics_strict_edge_reward": 0.05906057171523571,
      "rewards/forensics_tiou_reward": 0.5754473805427551,
      "step": 908
    },
    {
      "completion_length": 27.40625,
      "epoch": 2.7895545314900154,
      "grad_norm": 12.080392837524414,
      "kl": 0.740234375,
      "learning_rate": 3.015384615384615e-07,
      "loss": 0.0208,
      "reward": 0.7408975064754486,
      "reward_std": 0.04025100916624069,
      "reward_std_all": 0.05517561919987202,
      "rewards-max/forensics_iou_reward": 0.969673752784729,
      "rewards-max/forensics_strict_edge_reward": 0.7575493454933167,
      "rewards-max/forensics_tiou_reward": 0.9658639430999756,
      "rewards/forensics_iou_reward": 0.7167079150676727,
      "rewards/forensics_strict_edge_reward": 0.09080665558576584,
      "rewards/forensics_tiou_reward": 0.6804671287536621,
      "step": 909
    },
    {
      "completion_length": 26.0625,
      "epoch": 2.792626728110599,
      "grad_norm": 5.196066379547119,
      "kl": 0.5234375,
      "learning_rate": 3.0076923076923077e-07,
      "loss": 0.0527,
      "reward": 0.7346873879432678,
      "reward_std": 0.028113340958952904,
      "reward_std_all": 0.05105533730238676,
      "rewards-max/forensics_iou_reward": 0.984681248664856,
      "rewards-max/forensics_strict_edge_reward": 0.39007145166397095,
      "rewards-max/forensics_tiou_reward": 0.9792925715446472,
      "rewards/forensics_iou_reward": 0.6665700972080231,
      "rewards/forensics_strict_edge_reward": 0.05463628610596061,
      "rewards/forensics_tiou_reward": 0.6163574159145355,
      "step": 910
    },
    {
      "completion_length": 27.171875,
      "epoch": 2.795698924731183,
      "grad_norm": 6.1541337966918945,
      "kl": 0.3193359375,
      "learning_rate": 3e-07,
      "loss": 0.1678,
      "reward": 0.7369529604911804,
      "reward_std": 0.09171712771058083,
      "reward_std_all": 0.08643971383571625,
      "rewards-max/forensics_iou_reward": 0.9881576299667358,
      "rewards-max/forensics_strict_edge_reward": 0.8418737053871155,
      "rewards-max/forensics_tiou_reward": 0.9851719737052917,
      "rewards/forensics_iou_reward": 0.7075441479682922,
      "rewards/forensics_strict_edge_reward": 0.08699135109782219,
      "rewards/forensics_tiou_reward": 0.6669214367866516,
      "step": 911
    },
    {
      "completion_length": 28.078125,
      "epoch": 2.7987711213517663,
      "grad_norm": 5.525783061981201,
      "kl": 0.3759765625,
      "learning_rate": 2.9923076923076925e-07,
      "loss": 0.0589,
      "reward": 0.6754509508609772,
      "reward_std": 0.07182486355304718,
      "reward_std_all": 0.09906203672289848,
      "rewards-max/forensics_iou_reward": 0.9719151258468628,
      "rewards-max/forensics_strict_edge_reward": 0.549151748418808,
      "rewards-max/forensics_tiou_reward": 0.9639406502246857,
      "rewards/forensics_iou_reward": 0.5638337135314941,
      "rewards/forensics_strict_edge_reward": 0.06003329157829285,
      "rewards/forensics_tiou_reward": 0.5194248706102371,
      "step": 912
    },
    {
      "completion_length": 28.484375,
      "epoch": 2.80184331797235,
      "grad_norm": 6.962789058685303,
      "kl": 0.4150390625,
      "learning_rate": 2.984615384615384e-07,
      "loss": 0.0623,
      "reward": 0.7628476023674011,
      "reward_std": 0.05106547102332115,
      "reward_std_all": 0.04951692000031471,
      "rewards-max/forensics_iou_reward": 0.9614070951938629,
      "rewards-max/forensics_strict_edge_reward": 0.48441818356513977,
      "rewards-max/forensics_tiou_reward": 0.9501977562904358,
      "rewards/forensics_iou_reward": 0.6679906249046326,
      "rewards/forensics_strict_edge_reward": 0.06328438222408295,
      "rewards/forensics_tiou_reward": 0.6103461980819702,
      "step": 913
    },
    {
      "completion_length": 27.28125,
      "epoch": 2.804915514592934,
      "grad_norm": 7.215058326721191,
      "kl": 0.357421875,
      "learning_rate": 2.976923076923077e-07,
      "loss": -0.0367,
      "reward": 0.7841930985450745,
      "reward_std": 0.01362541550770402,
      "reward_std_all": 0.01815005368553102,
      "rewards-max/forensics_iou_reward": 0.9680178463459015,
      "rewards-max/forensics_strict_edge_reward": 0.5274569988250732,
      "rewards-max/forensics_tiou_reward": 0.9603745937347412,
      "rewards/forensics_iou_reward": 0.6683396995067596,
      "rewards/forensics_strict_edge_reward": 0.049685368314385414,
      "rewards/forensics_tiou_reward": 0.6168304085731506,
      "step": 914
    },
    {
      "completion_length": 25.234375,
      "epoch": 2.8079877112135176,
      "grad_norm": 9.057650566101074,
      "kl": 0.5078125,
      "learning_rate": 2.969230769230769e-07,
      "loss": 0.1019,
      "reward": 0.7000953853130341,
      "reward_std": 0.06114185135811567,
      "reward_std_all": 0.08348546177148819,
      "rewards-max/forensics_iou_reward": 0.9659357070922852,
      "rewards-max/forensics_strict_edge_reward": 0.5971094369888306,
      "rewards-max/forensics_tiou_reward": 0.9609659314155579,
      "rewards/forensics_iou_reward": 0.5646691918373108,
      "rewards/forensics_strict_edge_reward": 0.07416452094912529,
      "rewards/forensics_tiou_reward": 0.5259219408035278,
      "step": 915
    },
    {
      "completion_length": 23.921875,
      "epoch": 2.8110599078341014,
      "grad_norm": 6.508062839508057,
      "kl": 0.55859375,
      "learning_rate": 2.9615384615384616e-07,
      "loss": -0.0364,
      "reward": 0.6486698389053345,
      "reward_std": 0.09509018436074257,
      "reward_std_all": 0.13264600187540054,
      "rewards-max/forensics_iou_reward": 0.9848821461200714,
      "rewards-max/forensics_strict_edge_reward": 0.4136439710855484,
      "rewards-max/forensics_tiou_reward": 0.9780077040195465,
      "rewards/forensics_iou_reward": 0.5676804482936859,
      "rewards/forensics_strict_edge_reward": 0.04618922434747219,
      "rewards/forensics_tiou_reward": 0.5313675105571747,
      "step": 916
    },
    {
      "completion_length": 30.265625,
      "epoch": 2.814132104454685,
      "grad_norm": 7.769415855407715,
      "kl": 0.5107421875,
      "learning_rate": 2.953846153846154e-07,
      "loss": -0.0063,
      "reward": 0.7602744996547699,
      "reward_std": 0.03963900217786431,
      "reward_std_all": 0.042101224418729544,
      "rewards-max/forensics_iou_reward": 0.9847451150417328,
      "rewards-max/forensics_strict_edge_reward": 0.4434666633605957,
      "rewards-max/forensics_tiou_reward": 0.97941455245018,
      "rewards/forensics_iou_reward": 0.6560500264167786,
      "rewards/forensics_strict_edge_reward": 0.04336793161928654,
      "rewards/forensics_tiou_reward": 0.6054152846336365,
      "step": 917
    },
    {
      "completion_length": 23.375,
      "epoch": 2.817204301075269,
      "grad_norm": 8.08956527709961,
      "kl": 0.3447265625,
      "learning_rate": 2.946153846153846e-07,
      "loss": 0.0068,
      "reward": 0.7660403251647949,
      "reward_std": 0.04626301862299442,
      "reward_std_all": 0.04578592162579298,
      "rewards-max/forensics_iou_reward": 0.982229083776474,
      "rewards-max/forensics_strict_edge_reward": 0.5065501630306244,
      "rewards-max/forensics_tiou_reward": 0.9795056283473969,
      "rewards/forensics_iou_reward": 0.6814336478710175,
      "rewards/forensics_strict_edge_reward": 0.0491644162684679,
      "rewards/forensics_tiou_reward": 0.6419600546360016,
      "step": 918
    },
    {
      "completion_length": 26.5,
      "epoch": 2.8202764976958523,
      "grad_norm": 7.884687423706055,
      "kl": 0.67578125,
      "learning_rate": 2.9384615384615385e-07,
      "loss": 0.0422,
      "reward": 0.7258503437042236,
      "reward_std": 0.07556670904159546,
      "reward_std_all": 0.07588346302509308,
      "rewards-max/forensics_iou_reward": 0.9714156091213226,
      "rewards-max/forensics_strict_edge_reward": 0.33045870810747147,
      "rewards-max/forensics_tiou_reward": 0.9629822075366974,
      "rewards/forensics_iou_reward": 0.6248294711112976,
      "rewards/forensics_strict_edge_reward": 0.02776904171332717,
      "rewards/forensics_tiou_reward": 0.566921591758728,
      "step": 919
    },
    {
      "completion_length": 29.296875,
      "epoch": 2.823348694316436,
      "grad_norm": 8.686197280883789,
      "kl": 0.3525390625,
      "learning_rate": 2.9307692307692307e-07,
      "loss": -0.0696,
      "reward": 0.7570480704307556,
      "reward_std": 0.049585304223001,
      "reward_std_all": 0.05141147971153259,
      "rewards-max/forensics_iou_reward": 0.9650638103485107,
      "rewards-max/forensics_strict_edge_reward": 0.32409805059432983,
      "rewards-max/forensics_tiou_reward": 0.9520482420921326,
      "rewards/forensics_iou_reward": 0.660101443529129,
      "rewards/forensics_strict_edge_reward": 0.035391341894865036,
      "rewards/forensics_tiou_reward": 0.6015453636646271,
      "step": 920
    },
    {
      "completion_length": 28.109375,
      "epoch": 2.82642089093702,
      "grad_norm": 5.436910629272461,
      "kl": 0.486328125,
      "learning_rate": 2.9230769230769234e-07,
      "loss": -0.0185,
      "reward": 0.7527697682380676,
      "reward_std": 0.037815398536622524,
      "reward_std_all": 0.044494278728961945,
      "rewards-max/forensics_iou_reward": 0.9692219495773315,
      "rewards-max/forensics_strict_edge_reward": 0.3110484331846237,
      "rewards-max/forensics_tiou_reward": 0.9600802361965179,
      "rewards/forensics_iou_reward": 0.684612512588501,
      "rewards/forensics_strict_edge_reward": 0.04631154052913189,
      "rewards/forensics_tiou_reward": 0.6499828398227692,
      "step": 921
    },
    {
      "completion_length": 25.8125,
      "epoch": 2.8294930875576036,
      "grad_norm": 6.96835470199585,
      "kl": 0.41796875,
      "learning_rate": 2.915384615384615e-07,
      "loss": 0.0175,
      "reward": 0.7525922954082489,
      "reward_std": 0.02439900953322649,
      "reward_std_all": 0.03929363843053579,
      "rewards-max/forensics_iou_reward": 0.9939002990722656,
      "rewards-max/forensics_strict_edge_reward": 0.6377167105674744,
      "rewards-max/forensics_tiou_reward": 0.9897537231445312,
      "rewards/forensics_iou_reward": 0.656758576631546,
      "rewards/forensics_strict_edge_reward": 0.07808931171894073,
      "rewards/forensics_tiou_reward": 0.5964985489845276,
      "step": 922
    },
    {
      "completion_length": 27.1875,
      "epoch": 2.8325652841781874,
      "grad_norm": 5.68394660949707,
      "kl": 0.369140625,
      "learning_rate": 2.9076923076923076e-07,
      "loss": 0.031,
      "reward": 0.7544147372245789,
      "reward_std": 0.008791365660727024,
      "reward_std_all": 0.032628222135826945,
      "rewards-max/forensics_iou_reward": 0.958081990480423,
      "rewards-max/forensics_strict_edge_reward": 0.34842032939195633,
      "rewards-max/forensics_tiou_reward": 0.9504440724849701,
      "rewards/forensics_iou_reward": 0.6559630036354065,
      "rewards/forensics_strict_edge_reward": 0.04926226567476988,
      "rewards/forensics_tiou_reward": 0.6096771061420441,
      "step": 923
    },
    {
      "completion_length": 28.171875,
      "epoch": 2.835637480798771,
      "grad_norm": 6.63532829284668,
      "kl": 0.416015625,
      "learning_rate": 2.9e-07,
      "loss": -0.0014,
      "reward": 0.7209538519382477,
      "reward_std": 0.0671672411262989,
      "reward_std_all": 0.07655870541930199,
      "rewards-max/forensics_iou_reward": 0.9632900953292847,
      "rewards-max/forensics_strict_edge_reward": 0.49006423354148865,
      "rewards-max/forensics_tiou_reward": 0.9534343481063843,
      "rewards/forensics_iou_reward": 0.6062563359737396,
      "rewards/forensics_strict_edge_reward": 0.036049097776412964,
      "rewards/forensics_tiou_reward": 0.5690058171749115,
      "step": 924
    },
    {
      "completion_length": 24.390625,
      "epoch": 2.838709677419355,
      "grad_norm": 7.431890964508057,
      "kl": 0.380859375,
      "learning_rate": 2.8923076923076924e-07,
      "loss": 0.0822,
      "reward": 0.7544631361961365,
      "reward_std": 0.07725219428539276,
      "reward_std_all": 0.07073693163692951,
      "rewards-max/forensics_iou_reward": 0.947778970003128,
      "rewards-max/forensics_strict_edge_reward": 0.29184726625680923,
      "rewards-max/forensics_tiou_reward": 0.9304586350917816,
      "rewards/forensics_iou_reward": 0.6188063025474548,
      "rewards/forensics_strict_edge_reward": 0.040849081706255674,
      "rewards/forensics_tiou_reward": 0.5560160577297211,
      "step": 925
    },
    {
      "completion_length": 26.5625,
      "epoch": 2.8417818740399383,
      "grad_norm": 7.285758972167969,
      "kl": 0.390625,
      "learning_rate": 2.884615384615384e-07,
      "loss": 0.0857,
      "reward": 0.7420162558555603,
      "reward_std": 0.07849411014467478,
      "reward_std_all": 0.07415544800460339,
      "rewards-max/forensics_iou_reward": 0.9412618279457092,
      "rewards-max/forensics_strict_edge_reward": 0.31600624322891235,
      "rewards-max/forensics_tiou_reward": 0.9248436987400055,
      "rewards/forensics_iou_reward": 0.6145810484886169,
      "rewards/forensics_strict_edge_reward": 0.038379453122615814,
      "rewards/forensics_tiou_reward": 0.5669043064117432,
      "step": 926
    },
    {
      "completion_length": 29.9375,
      "epoch": 2.8448540706605225,
      "grad_norm": 7.8378005027771,
      "kl": 0.419921875,
      "learning_rate": 2.8769230769230767e-07,
      "loss": 0.0259,
      "reward": 0.7560508847236633,
      "reward_std": 0.04561254195868969,
      "reward_std_all": 0.04856240935623646,
      "rewards-max/forensics_iou_reward": 0.9885927736759186,
      "rewards-max/forensics_strict_edge_reward": 0.7203783690929413,
      "rewards-max/forensics_tiou_reward": 0.9850362539291382,
      "rewards/forensics_iou_reward": 0.6528312563896179,
      "rewards/forensics_strict_edge_reward": 0.07672746665775776,
      "rewards/forensics_tiou_reward": 0.6090271472930908,
      "step": 927
    },
    {
      "completion_length": 24.484375,
      "epoch": 2.847926267281106,
      "grad_norm": 5.632308006286621,
      "kl": 0.453125,
      "learning_rate": 2.8692307692307694e-07,
      "loss": -0.0381,
      "reward": 0.7551351189613342,
      "reward_std": 0.04416192136704922,
      "reward_std_all": 0.04793219827115536,
      "rewards-max/forensics_iou_reward": 0.9566931426525116,
      "rewards-max/forensics_strict_edge_reward": 0.5227276980876923,
      "rewards-max/forensics_tiou_reward": 0.9427596926689148,
      "rewards/forensics_iou_reward": 0.6672747731208801,
      "rewards/forensics_strict_edge_reward": 0.06321667786687613,
      "rewards/forensics_tiou_reward": 0.6124555468559265,
      "step": 928
    },
    {
      "completion_length": 26.25,
      "epoch": 2.8509984639016897,
      "grad_norm": 7.0113420486450195,
      "kl": 0.3310546875,
      "learning_rate": 2.8615384615384615e-07,
      "loss": 0.1107,
      "reward": 0.7341091334819794,
      "reward_std": 0.10192567482590675,
      "reward_std_all": 0.09318384528160095,
      "rewards-max/forensics_iou_reward": 0.9792720973491669,
      "rewards-max/forensics_strict_edge_reward": 0.6816803216934204,
      "rewards-max/forensics_tiou_reward": 0.9753828942775726,
      "rewards/forensics_iou_reward": 0.6792469024658203,
      "rewards/forensics_strict_edge_reward": 0.07661503180861473,
      "rewards/forensics_tiou_reward": 0.6306360065937042,
      "step": 929
    },
    {
      "completion_length": 25.578125,
      "epoch": 2.8540706605222734,
      "grad_norm": 7.655855178833008,
      "kl": 0.353515625,
      "learning_rate": 2.8538461538461536e-07,
      "loss": 0.0139,
      "reward": 0.731088399887085,
      "reward_std": 0.0635988914873451,
      "reward_std_all": 0.06696855463087559,
      "rewards-max/forensics_iou_reward": 0.9826149642467499,
      "rewards-max/forensics_strict_edge_reward": 0.4753422290086746,
      "rewards-max/forensics_tiou_reward": 0.9764703214168549,
      "rewards/forensics_iou_reward": 0.6363109052181244,
      "rewards/forensics_strict_edge_reward": 0.07623487524688244,
      "rewards/forensics_tiou_reward": 0.5899683833122253,
      "step": 930
    },
    {
      "completion_length": 23.015625,
      "epoch": 2.857142857142857,
      "grad_norm": 8.589834213256836,
      "kl": 0.486328125,
      "learning_rate": 2.846153846153846e-07,
      "loss": 0.0418,
      "reward": 0.7360235452651978,
      "reward_std": 0.08255447167903185,
      "reward_std_all": 0.0784481717273593,
      "rewards-max/forensics_iou_reward": 0.9726698696613312,
      "rewards-max/forensics_strict_edge_reward": 0.3456393852829933,
      "rewards-max/forensics_tiou_reward": 0.9629294574260712,
      "rewards/forensics_iou_reward": 0.6169187724590302,
      "rewards/forensics_strict_edge_reward": 0.03913251683115959,
      "rewards/forensics_tiou_reward": 0.5632770657539368,
      "step": 931
    },
    {
      "completion_length": 29.515625,
      "epoch": 2.860215053763441,
      "grad_norm": 5.993714332580566,
      "kl": 0.5400390625,
      "learning_rate": 2.8384615384615384e-07,
      "loss": 0.0134,
      "reward": 0.7786427140235901,
      "reward_std": 0.03906979039311409,
      "reward_std_all": 0.0369354672729969,
      "rewards-max/forensics_iou_reward": 0.9819121360778809,
      "rewards-max/forensics_strict_edge_reward": 0.8135445415973663,
      "rewards-max/forensics_tiou_reward": 0.9785469472408295,
      "rewards/forensics_iou_reward": 0.5997903943061829,
      "rewards/forensics_strict_edge_reward": 0.06532508134841919,
      "rewards/forensics_tiou_reward": 0.5524360835552216,
      "step": 932
    },
    {
      "completion_length": 24.3125,
      "epoch": 2.8632872503840243,
      "grad_norm": 7.343483924865723,
      "kl": 0.3662109375,
      "learning_rate": 2.8307692307692306e-07,
      "loss": 0.0264,
      "reward": 0.7475863397121429,
      "reward_std": 0.025018130894750357,
      "reward_std_all": 0.044816067442297935,
      "rewards-max/forensics_iou_reward": 0.9898989796638489,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9867885112762451,
      "rewards/forensics_iou_reward": 0.6210797727108002,
      "rewards/forensics_strict_edge_reward": 0.07955779507756233,
      "rewards/forensics_tiou_reward": 0.5789403021335602,
      "step": 933
    },
    {
      "completion_length": 22.8125,
      "epoch": 2.8663594470046085,
      "grad_norm": 5.729289531707764,
      "kl": 0.4052734375,
      "learning_rate": 2.823076923076923e-07,
      "loss": 0.1263,
      "reward": 0.785767138004303,
      "reward_std": 0.026143668685108423,
      "reward_std_all": 0.02694118395447731,
      "rewards-max/forensics_iou_reward": 0.986175537109375,
      "rewards-max/forensics_strict_edge_reward": 0.4374866336584091,
      "rewards-max/forensics_tiou_reward": 0.980118602514267,
      "rewards/forensics_iou_reward": 0.7482132911682129,
      "rewards/forensics_strict_edge_reward": 0.06696443632245064,
      "rewards/forensics_tiou_reward": 0.6857967972755432,
      "step": 934
    },
    {
      "completion_length": 28.765625,
      "epoch": 2.869431643625192,
      "grad_norm": 6.213136196136475,
      "kl": 0.45703125,
      "learning_rate": 2.815384615384615e-07,
      "loss": 0.0102,
      "reward": 0.7364554107189178,
      "reward_std": 0.028812095522880554,
      "reward_std_all": 0.051146382465958595,
      "rewards-max/forensics_iou_reward": 0.9748701751232147,
      "rewards-max/forensics_strict_edge_reward": 0.2629244551062584,
      "rewards-max/forensics_tiou_reward": 0.9535419940948486,
      "rewards/forensics_iou_reward": 0.5898257493972778,
      "rewards/forensics_strict_edge_reward": 0.022887525148689747,
      "rewards/forensics_tiou_reward": 0.5356161594390869,
      "step": 935
    },
    {
      "completion_length": 28.890625,
      "epoch": 2.8725038402457757,
      "grad_norm": 8.281655311584473,
      "kl": 0.3466796875,
      "learning_rate": 2.8076923076923075e-07,
      "loss": 0.1431,
      "reward": 0.762902557849884,
      "reward_std": 0.03781286487355828,
      "reward_std_all": 0.04037144593894482,
      "rewards-max/forensics_iou_reward": 0.940433919429779,
      "rewards-max/forensics_strict_edge_reward": 0.3004312515258789,
      "rewards-max/forensics_tiou_reward": 0.9277157485485077,
      "rewards/forensics_iou_reward": 0.6273378729820251,
      "rewards/forensics_strict_edge_reward": 0.03219273593276739,
      "rewards/forensics_tiou_reward": 0.5840350389480591,
      "step": 936
    },
    {
      "completion_length": 26.296875,
      "epoch": 2.8755760368663594,
      "grad_norm": 6.881389617919922,
      "kl": 0.3798828125,
      "learning_rate": 2.8e-07,
      "loss": 0.0369,
      "reward": 0.7756307125091553,
      "reward_std": 0.030412628315389156,
      "reward_std_all": 0.031221451237797737,
      "rewards-max/forensics_iou_reward": 0.9804486036300659,
      "rewards-max/forensics_strict_edge_reward": 0.5658674240112305,
      "rewards-max/forensics_tiou_reward": 0.9706381559371948,
      "rewards/forensics_iou_reward": 0.7229205965995789,
      "rewards/forensics_strict_edge_reward": 0.04825112037360668,
      "rewards/forensics_tiou_reward": 0.6700567901134491,
      "step": 937
    },
    {
      "completion_length": 22.75,
      "epoch": 2.878648233486943,
      "grad_norm": 8.400850296020508,
      "kl": 0.5068359375,
      "learning_rate": 2.7923076923076923e-07,
      "loss": 0.0353,
      "reward": 0.7678440809249878,
      "reward_std": 0.04850351973436773,
      "reward_std_all": 0.046189614571630955,
      "rewards-max/forensics_iou_reward": 0.984375,
      "rewards-max/forensics_strict_edge_reward": 0.7388437688350677,
      "rewards-max/forensics_tiou_reward": 0.9737536311149597,
      "rewards/forensics_iou_reward": 0.6784816086292267,
      "rewards/forensics_strict_edge_reward": 0.09825970977544785,
      "rewards/forensics_tiou_reward": 0.6336122453212738,
      "step": 938
    },
    {
      "completion_length": 25.671875,
      "epoch": 2.881720430107527,
      "grad_norm": 7.236908912658691,
      "kl": 0.4462890625,
      "learning_rate": 2.7846153846153845e-07,
      "loss": 0.1064,
      "reward": 0.7062440514564514,
      "reward_std": 0.041483656503260136,
      "reward_std_all": 0.07161579467356205,
      "rewards-max/forensics_iou_reward": 0.9841949343681335,
      "rewards-max/forensics_strict_edge_reward": 0.4566798061132431,
      "rewards-max/forensics_tiou_reward": 0.9784855246543884,
      "rewards/forensics_iou_reward": 0.6052396893501282,
      "rewards/forensics_strict_edge_reward": 0.036100570345297456,
      "rewards/forensics_tiou_reward": 0.552818775177002,
      "step": 939
    },
    {
      "completion_length": 28.046875,
      "epoch": 2.8847926267281108,
      "grad_norm": 7.769891738891602,
      "kl": 0.349609375,
      "learning_rate": 2.7769230769230766e-07,
      "loss": 0.0584,
      "reward": 0.7668539583683014,
      "reward_std": 0.052230662666261196,
      "reward_std_all": 0.04919137340039015,
      "rewards-max/forensics_iou_reward": 0.978012204170227,
      "rewards-max/forensics_strict_edge_reward": 0.5221183001995087,
      "rewards-max/forensics_tiou_reward": 0.972574919462204,
      "rewards/forensics_iou_reward": 0.6559784710407257,
      "rewards/forensics_strict_edge_reward": 0.05127084255218506,
      "rewards/forensics_tiou_reward": 0.6224625706672668,
      "step": 940
    },
    {
      "completion_length": 23.765625,
      "epoch": 2.8878648233486945,
      "grad_norm": 7.748489856719971,
      "kl": 0.333984375,
      "learning_rate": 2.7692307692307693e-07,
      "loss": 0.0602,
      "reward": 0.7007001042366028,
      "reward_std": 0.046182578429579735,
      "reward_std_all": 0.0747667457908392,
      "rewards-max/forensics_iou_reward": 0.9214204549789429,
      "rewards-max/forensics_strict_edge_reward": 0.26740886457264423,
      "rewards-max/forensics_tiou_reward": 0.8898572623729706,
      "rewards/forensics_iou_reward": 0.6032579392194748,
      "rewards/forensics_strict_edge_reward": 0.03538160375319421,
      "rewards/forensics_tiou_reward": 0.5549110770225525,
      "step": 941
    },
    {
      "completion_length": 23.15625,
      "epoch": 2.890937019969278,
      "grad_norm": 6.723114490509033,
      "kl": 0.5390625,
      "learning_rate": 2.7615384615384614e-07,
      "loss": 0.1244,
      "reward": 0.7313859164714813,
      "reward_std": 0.053345488384366035,
      "reward_std_all": 0.06316082924604416,
      "rewards-max/forensics_iou_reward": 0.9883005023002625,
      "rewards-max/forensics_strict_edge_reward": 0.7188692688941956,
      "rewards-max/forensics_tiou_reward": 0.9842797219753265,
      "rewards/forensics_iou_reward": 0.6566545069217682,
      "rewards/forensics_strict_edge_reward": 0.08202754706144333,
      "rewards/forensics_tiou_reward": 0.6052831411361694,
      "step": 942
    },
    {
      "completion_length": 26.390625,
      "epoch": 2.8940092165898617,
      "grad_norm": 7.049102306365967,
      "kl": 0.4248046875,
      "learning_rate": 2.753846153846154e-07,
      "loss": 0.0643,
      "reward": 0.7647331655025482,
      "reward_std": 0.05275685340166092,
      "reward_std_all": 0.049993555061519146,
      "rewards-max/forensics_iou_reward": 0.9880642294883728,
      "rewards-max/forensics_strict_edge_reward": 0.6836083829402924,
      "rewards-max/forensics_tiou_reward": 0.9868179261684418,
      "rewards/forensics_iou_reward": 0.6589964628219604,
      "rewards/forensics_strict_edge_reward": 0.07815692946314812,
      "rewards/forensics_tiou_reward": 0.6075297594070435,
      "step": 943
    },
    {
      "completion_length": 23.75,
      "epoch": 2.8970814132104454,
      "grad_norm": 8.434550285339355,
      "kl": 0.3603515625,
      "learning_rate": 2.7461538461538457e-07,
      "loss": 0.0029,
      "reward": 0.7269346714019775,
      "reward_std": 0.06969672068953514,
      "reward_std_all": 0.07347723469138145,
      "rewards-max/forensics_iou_reward": 0.9676796197891235,
      "rewards-max/forensics_strict_edge_reward": 0.5404154062271118,
      "rewards-max/forensics_tiou_reward": 0.9597780108451843,
      "rewards/forensics_iou_reward": 0.6229605674743652,
      "rewards/forensics_strict_edge_reward": 0.091713797301054,
      "rewards/forensics_tiou_reward": 0.5826152563095093,
      "step": 944
    },
    {
      "completion_length": 23.015625,
      "epoch": 2.900153609831029,
      "grad_norm": 7.172518253326416,
      "kl": 0.408203125,
      "learning_rate": 2.7384615384615383e-07,
      "loss": 0.1252,
      "reward": 0.6478870511054993,
      "reward_std": 0.051473382860422134,
      "reward_std_all": 0.10510344430804253,
      "rewards-max/forensics_iou_reward": 0.908730149269104,
      "rewards-max/forensics_strict_edge_reward": 0.34755249321460724,
      "rewards-max/forensics_tiou_reward": 0.8852099180221558,
      "rewards/forensics_iou_reward": 0.5058070123195648,
      "rewards/forensics_strict_edge_reward": 0.03161030588671565,
      "rewards/forensics_tiou_reward": 0.4646373838186264,
      "step": 945
    },
    {
      "completion_length": 27.203125,
      "epoch": 2.903225806451613,
      "grad_norm": 6.370201110839844,
      "kl": 0.451171875,
      "learning_rate": 2.7307692307692305e-07,
      "loss": 0.1653,
      "reward": 0.7269655168056488,
      "reward_std": 0.06473633274435997,
      "reward_std_all": 0.07442372292280197,
      "rewards-max/forensics_iou_reward": 0.925214409828186,
      "rewards-max/forensics_strict_edge_reward": 0.47082987427711487,
      "rewards-max/forensics_tiou_reward": 0.8999879658222198,
      "rewards/forensics_iou_reward": 0.6321187019348145,
      "rewards/forensics_strict_edge_reward": 0.05390596482902765,
      "rewards/forensics_tiou_reward": 0.580823540687561,
      "step": 946
    },
    {
      "completion_length": 26.28125,
      "epoch": 2.9062980030721968,
      "grad_norm": 8.162545204162598,
      "kl": 0.3681640625,
      "learning_rate": 2.723076923076923e-07,
      "loss": -0.0909,
      "reward": 0.7474346160888672,
      "reward_std": 0.04939187690615654,
      "reward_std_all": 0.053545963019132614,
      "rewards-max/forensics_iou_reward": 0.9982206523418427,
      "rewards-max/forensics_strict_edge_reward": 0.9836080372333527,
      "rewards-max/forensics_tiou_reward": 0.9972281754016876,
      "rewards/forensics_iou_reward": 0.6761017441749573,
      "rewards/forensics_strict_edge_reward": 0.11355603486299515,
      "rewards/forensics_tiou_reward": 0.6151900887489319,
      "step": 947
    },
    {
      "completion_length": 23.640625,
      "epoch": 2.9093701996927805,
      "grad_norm": 7.356562614440918,
      "kl": 0.6533203125,
      "learning_rate": 2.7153846153846153e-07,
      "loss": 0.0538,
      "reward": 0.7624189853668213,
      "reward_std": 0.0763792097568512,
      "reward_std_all": 0.06772049516439438,
      "rewards-max/forensics_iou_reward": 0.980949193239212,
      "rewards-max/forensics_strict_edge_reward": 0.6326714754104614,
      "rewards-max/forensics_tiou_reward": 0.9739843904972076,
      "rewards/forensics_iou_reward": 0.7244712114334106,
      "rewards/forensics_strict_edge_reward": 0.09386342391371727,
      "rewards/forensics_tiou_reward": 0.6686836183071136,
      "step": 948
    },
    {
      "completion_length": 25.578125,
      "epoch": 2.912442396313364,
      "grad_norm": 7.770138263702393,
      "kl": 0.4619140625,
      "learning_rate": 2.7076923076923074e-07,
      "loss": 0.0641,
      "reward": 0.7452874183654785,
      "reward_std": 0.028678476810455322,
      "reward_std_all": 0.04694796120747924,
      "rewards-max/forensics_iou_reward": 0.9620685577392578,
      "rewards-max/forensics_strict_edge_reward": 0.5863399803638458,
      "rewards-max/forensics_tiou_reward": 0.9529821574687958,
      "rewards/forensics_iou_reward": 0.6494860053062439,
      "rewards/forensics_strict_edge_reward": 0.08248855452984571,
      "rewards/forensics_tiou_reward": 0.6100147366523743,
      "step": 949
    },
    {
      "completion_length": 25.15625,
      "epoch": 2.9155145929339477,
      "grad_norm": 9.539434432983398,
      "kl": 0.431640625,
      "learning_rate": 2.7e-07,
      "loss": 0.0092,
      "reward": 0.7540200650691986,
      "reward_std": 0.04826098773628473,
      "reward_std_all": 0.05014963215216994,
      "rewards-max/forensics_iou_reward": 0.9587767124176025,
      "rewards-max/forensics_strict_edge_reward": 0.506792426109314,
      "rewards-max/forensics_tiou_reward": 0.9415032267570496,
      "rewards/forensics_iou_reward": 0.6328709721565247,
      "rewards/forensics_strict_edge_reward": 0.058294810354709625,
      "rewards/forensics_tiou_reward": 0.5831506550312042,
      "step": 950
    },
    {
      "completion_length": 26.3125,
      "epoch": 2.9185867895545314,
      "grad_norm": 5.7470879554748535,
      "kl": 0.384765625,
      "learning_rate": 2.692307692307692e-07,
      "loss": 0.0343,
      "reward": 0.7446380257606506,
      "reward_std": 0.07194763515144587,
      "reward_std_all": 0.06862628180533648,
      "rewards-max/forensics_iou_reward": 0.9691233038902283,
      "rewards-max/forensics_strict_edge_reward": 0.4482892006635666,
      "rewards-max/forensics_tiou_reward": 0.9604366719722748,
      "rewards/forensics_iou_reward": 0.6654715240001678,
      "rewards/forensics_strict_edge_reward": 0.06558824609965086,
      "rewards/forensics_tiou_reward": 0.6210125386714935,
      "step": 951
    },
    {
      "completion_length": 22.296875,
      "epoch": 2.921658986175115,
      "grad_norm": 7.673227310180664,
      "kl": 0.380859375,
      "learning_rate": 2.6846153846153844e-07,
      "loss": 0.0598,
      "reward": 0.7266337275505066,
      "reward_std": 0.02878954354673624,
      "reward_std_all": 0.05097840540111065,
      "rewards-max/forensics_iou_reward": 0.9295426905155182,
      "rewards-max/forensics_strict_edge_reward": 0.26341865956783295,
      "rewards-max/forensics_tiou_reward": 0.9138940572738647,
      "rewards/forensics_iou_reward": 0.6321516931056976,
      "rewards/forensics_strict_edge_reward": 0.03428985737264156,
      "rewards/forensics_tiou_reward": 0.5836754143238068,
      "step": 952
    },
    {
      "completion_length": 28.5,
      "epoch": 2.924731182795699,
      "grad_norm": 6.91480827331543,
      "kl": 0.388671875,
      "learning_rate": 2.6769230769230765e-07,
      "loss": 0.0112,
      "reward": 0.7794619202613831,
      "reward_std": 0.04697979614138603,
      "reward_std_all": 0.04241950064897537,
      "rewards-max/forensics_iou_reward": 0.984154224395752,
      "rewards-max/forensics_strict_edge_reward": 0.6201116293668747,
      "rewards-max/forensics_tiou_reward": 0.9787929356098175,
      "rewards/forensics_iou_reward": 0.7385838329792023,
      "rewards/forensics_strict_edge_reward": 0.08638369850814342,
      "rewards/forensics_tiou_reward": 0.6973586678504944,
      "step": 953
    },
    {
      "completion_length": 23.3125,
      "epoch": 2.9278033794162828,
      "grad_norm": 7.158670425415039,
      "kl": 0.3447265625,
      "learning_rate": 2.669230769230769e-07,
      "loss": -0.1435,
      "reward": 0.7131638526916504,
      "reward_std": 0.06608288362622261,
      "reward_std_all": 0.08392278663814068,
      "rewards-max/forensics_iou_reward": 0.9904059171676636,
      "rewards-max/forensics_strict_edge_reward": 0.7169294059276581,
      "rewards-max/forensics_tiou_reward": 0.988974004983902,
      "rewards/forensics_iou_reward": 0.630697637796402,
      "rewards/forensics_strict_edge_reward": 0.07716729491949081,
      "rewards/forensics_tiou_reward": 0.5930259525775909,
      "step": 954
    },
    {
      "completion_length": 25.65625,
      "epoch": 2.9308755760368665,
      "grad_norm": 7.498308181762695,
      "kl": 0.32421875,
      "learning_rate": 2.6615384615384613e-07,
      "loss": 0.093,
      "reward": 0.764225959777832,
      "reward_std": 0.04699087701737881,
      "reward_std_all": 0.04726244788616896,
      "rewards-max/forensics_iou_reward": 0.9685919284820557,
      "rewards-max/forensics_strict_edge_reward": 0.5154386982321739,
      "rewards-max/forensics_tiou_reward": 0.9466205835342407,
      "rewards/forensics_iou_reward": 0.7031350135803223,
      "rewards/forensics_strict_edge_reward": 0.07357539236545563,
      "rewards/forensics_tiou_reward": 0.6449195146560669,
      "step": 955
    },
    {
      "completion_length": 27.515625,
      "epoch": 2.93394777265745,
      "grad_norm": 5.92336368560791,
      "kl": 0.4267578125,
      "learning_rate": 2.653846153846154e-07,
      "loss": 0.0467,
      "reward": 0.7466865181922913,
      "reward_std": 0.04354195203632116,
      "reward_std_all": 0.053559962660074234,
      "rewards-max/forensics_iou_reward": 0.9935513734817505,
      "rewards-max/forensics_strict_edge_reward": 0.6666916608810425,
      "rewards-max/forensics_tiou_reward": 0.9912818372249603,
      "rewards/forensics_iou_reward": 0.6743379831314087,
      "rewards/forensics_strict_edge_reward": 0.08754650317132473,
      "rewards/forensics_tiou_reward": 0.638333797454834,
      "step": 956
    },
    {
      "completion_length": 24.5,
      "epoch": 2.9370199692780337,
      "grad_norm": 7.757746696472168,
      "kl": 0.34765625,
      "learning_rate": 2.646153846153846e-07,
      "loss": -0.0823,
      "reward": 0.7095827758312225,
      "reward_std": 0.05034831538796425,
      "reward_std_all": 0.07851642929017544,
      "rewards-max/forensics_iou_reward": 0.9708688855171204,
      "rewards-max/forensics_strict_edge_reward": 0.45645682513713837,
      "rewards-max/forensics_tiou_reward": 0.9635503888130188,
      "rewards/forensics_iou_reward": 0.6037594974040985,
      "rewards/forensics_strict_edge_reward": 0.04033895954489708,
      "rewards/forensics_tiou_reward": 0.5518999099731445,
      "step": 957
    },
    {
      "completion_length": 24.625,
      "epoch": 2.9400921658986174,
      "grad_norm": 8.328022956848145,
      "kl": 0.388671875,
      "learning_rate": 2.638461538461538e-07,
      "loss": 0.0292,
      "reward": 0.7492140233516693,
      "reward_std": 0.07352142035961151,
      "reward_std_all": 0.06846298463642597,
      "rewards-max/forensics_iou_reward": 0.9657131433486938,
      "rewards-max/forensics_strict_edge_reward": 0.44064828753471375,
      "rewards-max/forensics_tiou_reward": 0.9535112380981445,
      "rewards/forensics_iou_reward": 0.6788819432258606,
      "rewards/forensics_strict_edge_reward": 0.058085158467292786,
      "rewards/forensics_tiou_reward": 0.6234221160411835,
      "step": 958
    },
    {
      "completion_length": 24.21875,
      "epoch": 2.943164362519201,
      "grad_norm": 7.445954322814941,
      "kl": 0.505859375,
      "learning_rate": 2.630769230769231e-07,
      "loss": 0.1089,
      "reward": 0.7598387897014618,
      "reward_std": 0.040073781507089734,
      "reward_std_all": 0.04401697497814894,
      "rewards-max/forensics_iou_reward": 0.9461032748222351,
      "rewards-max/forensics_strict_edge_reward": 0.40399426221847534,
      "rewards-max/forensics_tiou_reward": 0.9308564066886902,
      "rewards/forensics_iou_reward": 0.6461308002471924,
      "rewards/forensics_strict_edge_reward": 0.04472973197698593,
      "rewards/forensics_tiou_reward": 0.5774803757667542,
      "step": 959
    },
    {
      "completion_length": 28.1875,
      "epoch": 2.946236559139785,
      "grad_norm": 6.712944030761719,
      "kl": 0.3955078125,
      "learning_rate": 2.623076923076923e-07,
      "loss": -0.063,
      "reward": 0.6755678951740265,
      "reward_std": 0.014131520874798298,
      "reward_std_all": 0.0732591561973095,
      "rewards-max/forensics_iou_reward": 0.9977269470691681,
      "rewards-max/forensics_strict_edge_reward": 0.6490193009376526,
      "rewards-max/forensics_tiou_reward": 0.9961977899074554,
      "rewards/forensics_iou_reward": 0.5905832350254059,
      "rewards/forensics_strict_edge_reward": 0.06273963302373886,
      "rewards/forensics_tiou_reward": 0.5535238683223724,
      "step": 960
    },
    {
      "completion_length": 26.8125,
      "epoch": 2.9493087557603688,
      "grad_norm": 6.987604141235352,
      "kl": 0.4140625,
      "learning_rate": 2.615384615384615e-07,
      "loss": 0.1072,
      "reward": 0.6949133276939392,
      "reward_std": 0.08328864350914955,
      "reward_std_all": 0.09283927641808987,
      "rewards-max/forensics_iou_reward": 0.8992874026298523,
      "rewards-max/forensics_strict_edge_reward": 0.11615362390875816,
      "rewards-max/forensics_tiou_reward": 0.8775607943534851,
      "rewards/forensics_iou_reward": 0.5245668143033981,
      "rewards/forensics_strict_edge_reward": 0.01135462149977684,
      "rewards/forensics_tiou_reward": 0.46947894990444183,
      "step": 961
    },
    {
      "completion_length": 22.6875,
      "epoch": 2.9523809523809526,
      "grad_norm": 6.536511421203613,
      "kl": 0.30859375,
      "learning_rate": 2.6076923076923073e-07,
      "loss": -0.0732,
      "reward": 0.7557247281074524,
      "reward_std": 0.051476532593369484,
      "reward_std_all": 0.052698105573654175,
      "rewards-max/forensics_iou_reward": 0.9592070579528809,
      "rewards-max/forensics_strict_edge_reward": 0.469832107424736,
      "rewards-max/forensics_tiou_reward": 0.9486063718795776,
      "rewards/forensics_iou_reward": 0.7080176472663879,
      "rewards/forensics_strict_edge_reward": 0.05859971605241299,
      "rewards/forensics_tiou_reward": 0.6610442698001862,
      "step": 962
    },
    {
      "completion_length": 27.6875,
      "epoch": 2.955453149001536,
      "grad_norm": 6.573411464691162,
      "kl": 0.4736328125,
      "learning_rate": 2.6e-07,
      "loss": 0.005,
      "reward": 0.7811138927936554,
      "reward_std": 0.028259528684429824,
      "reward_std_all": 0.028384355828166008,
      "rewards-max/forensics_iou_reward": 0.9796665608882904,
      "rewards-max/forensics_strict_edge_reward": 0.26625459641218185,
      "rewards-max/forensics_tiou_reward": 0.9695073664188385,
      "rewards/forensics_iou_reward": 0.680613249540329,
      "rewards/forensics_strict_edge_reward": 0.0254118621814996,
      "rewards/forensics_tiou_reward": 0.6215685904026031,
      "step": 963
    },
    {
      "completion_length": 27.40625,
      "epoch": 2.9585253456221197,
      "grad_norm": 6.690407752990723,
      "kl": 0.3876953125,
      "learning_rate": 2.592307692307692e-07,
      "loss": 0.0469,
      "reward": 0.7771070599555969,
      "reward_std": 0.029701558873057365,
      "reward_std_all": 0.030810724943876266,
      "rewards-max/forensics_iou_reward": 0.9756960272789001,
      "rewards-max/forensics_strict_edge_reward": 0.5262769460678101,
      "rewards-max/forensics_tiou_reward": 0.9659606516361237,
      "rewards/forensics_iou_reward": 0.6831961870193481,
      "rewards/forensics_strict_edge_reward": 0.05892286077141762,
      "rewards/forensics_tiou_reward": 0.6380254030227661,
      "step": 964
    },
    {
      "completion_length": 28.28125,
      "epoch": 2.9615975422427034,
      "grad_norm": 8.415180206298828,
      "kl": 0.31640625,
      "learning_rate": 2.584615384615385e-07,
      "loss": -0.0111,
      "reward": 0.7757983505725861,
      "reward_std": 0.043784622102975845,
      "reward_std_all": 0.04085927829146385,
      "rewards-max/forensics_iou_reward": 0.9893743991851807,
      "rewards-max/forensics_strict_edge_reward": 0.5926170498132706,
      "rewards-max/forensics_tiou_reward": 0.985621303319931,
      "rewards/forensics_iou_reward": 0.7289409041404724,
      "rewards/forensics_strict_edge_reward": 0.04852496460080147,
      "rewards/forensics_tiou_reward": 0.6850577890872955,
      "step": 965
    },
    {
      "completion_length": 29.234375,
      "epoch": 2.964669738863287,
      "grad_norm": 5.873526573181152,
      "kl": 0.3173828125,
      "learning_rate": 2.5769230769230764e-07,
      "loss": 0.0081,
      "reward": 0.7536133527755737,
      "reward_std": 0.04421073757112026,
      "reward_std_all": 0.04826832376420498,
      "rewards-max/forensics_iou_reward": 0.967774361371994,
      "rewards-max/forensics_strict_edge_reward": 0.39585426449775696,
      "rewards-max/forensics_tiou_reward": 0.9545159339904785,
      "rewards/forensics_iou_reward": 0.6570964157581329,
      "rewards/forensics_strict_edge_reward": 0.045539941638708115,
      "rewards/forensics_tiou_reward": 0.6200266778469086,
      "step": 966
    },
    {
      "completion_length": 26.265625,
      "epoch": 2.967741935483871,
      "grad_norm": 9.559520721435547,
      "kl": 0.3984375,
      "learning_rate": 2.569230769230769e-07,
      "loss": 0.1305,
      "reward": 0.7189819514751434,
      "reward_std": 0.11140014603734016,
      "reward_std_all": 0.1034056767821312,
      "rewards-max/forensics_iou_reward": 0.9892336428165436,
      "rewards-max/forensics_strict_edge_reward": 0.5538144111633301,
      "rewards-max/forensics_tiou_reward": 0.9862777292728424,
      "rewards/forensics_iou_reward": 0.5850256681442261,
      "rewards/forensics_strict_edge_reward": 0.041130177676677704,
      "rewards/forensics_tiou_reward": 0.5406114459037781,
      "step": 967
    },
    {
      "completion_length": 22.09375,
      "epoch": 2.9708141321044548,
      "grad_norm": 10.6248140335083,
      "kl": 0.451171875,
      "learning_rate": 2.5615384615384617e-07,
      "loss": 0.1133,
      "reward": 0.6947821974754333,
      "reward_std": 0.0890408456325531,
      "reward_std_all": 0.10480611771345139,
      "rewards-max/forensics_iou_reward": 0.9457831382751465,
      "rewards-max/forensics_strict_edge_reward": 0.6505971103906631,
      "rewards-max/forensics_tiou_reward": 0.9377911388874054,
      "rewards/forensics_iou_reward": 0.6264837086200714,
      "rewards/forensics_strict_edge_reward": 0.07266221009194851,
      "rewards/forensics_tiou_reward": 0.5856305658817291,
      "step": 968
    },
    {
      "completion_length": 27.65625,
      "epoch": 2.9738863287250386,
      "grad_norm": 6.720517158508301,
      "kl": 0.611328125,
      "learning_rate": 2.553846153846154e-07,
      "loss": 0.0994,
      "reward": 0.7842558324337006,
      "reward_std": 0.035053880885243416,
      "reward_std_all": 0.03241295134648681,
      "rewards-max/forensics_iou_reward": 0.9617498517036438,
      "rewards-max/forensics_strict_edge_reward": 0.6122131198644638,
      "rewards-max/forensics_tiou_reward": 0.9547967314720154,
      "rewards/forensics_iou_reward": 0.611198753118515,
      "rewards/forensics_strict_edge_reward": 0.06453688815236092,
      "rewards/forensics_tiou_reward": 0.5582593977451324,
      "step": 969
    },
    {
      "completion_length": 27.140625,
      "epoch": 2.976958525345622,
      "grad_norm": 8.363274574279785,
      "kl": 0.462890625,
      "learning_rate": 2.546153846153846e-07,
      "loss": 0.0881,
      "reward": 0.7465644776821136,
      "reward_std": 0.055249689146876335,
      "reward_std_all": 0.05634554103016853,
      "rewards-max/forensics_iou_reward": 0.9935228228569031,
      "rewards-max/forensics_strict_edge_reward": 0.7542104423046112,
      "rewards-max/forensics_tiou_reward": 0.9916230738162994,
      "rewards/forensics_iou_reward": 0.6330455839633942,
      "rewards/forensics_strict_edge_reward": 0.07078736647963524,
      "rewards/forensics_tiou_reward": 0.587529331445694,
      "step": 970
    },
    {
      "completion_length": 29.234375,
      "epoch": 2.980030721966206,
      "grad_norm": 7.519860744476318,
      "kl": 0.4609375,
      "learning_rate": 2.538461538461538e-07,
      "loss": -0.0127,
      "reward": 0.6920731961727142,
      "reward_std": 0.06542436499148607,
      "reward_std_all": 0.0960082896053791,
      "rewards-max/forensics_iou_reward": 0.9726074934005737,
      "rewards-max/forensics_strict_edge_reward": 0.6453147530555725,
      "rewards-max/forensics_tiou_reward": 0.9638018608093262,
      "rewards/forensics_iou_reward": 0.6475532054901123,
      "rewards/forensics_strict_edge_reward": 0.06672246940433979,
      "rewards/forensics_tiou_reward": 0.6184661388397217,
      "step": 971
    },
    {
      "completion_length": 27.90625,
      "epoch": 2.9831029185867894,
      "grad_norm": 6.426028251647949,
      "kl": 0.41796875,
      "learning_rate": 2.530769230769231e-07,
      "loss": 0.0296,
      "reward": 0.7460977137088776,
      "reward_std": 0.06918261898681521,
      "reward_std_all": 0.0662659655790776,
      "rewards-max/forensics_iou_reward": 0.9605334997177124,
      "rewards-max/forensics_strict_edge_reward": 0.44554485380649567,
      "rewards-max/forensics_tiou_reward": 0.9384075403213501,
      "rewards/forensics_iou_reward": 0.6582667827606201,
      "rewards/forensics_strict_edge_reward": 0.06904922798275948,
      "rewards/forensics_tiou_reward": 0.6192340552806854,
      "step": 972
    },
    {
      "completion_length": 25.75,
      "epoch": 2.986175115207373,
      "grad_norm": 6.72426700592041,
      "kl": 0.4111328125,
      "learning_rate": 2.523076923076923e-07,
      "loss": 0.1007,
      "reward": 0.7712558209896088,
      "reward_std": 0.0384134016931057,
      "reward_std_all": 0.038249324075877666,
      "rewards-max/forensics_iou_reward": 0.9572345316410065,
      "rewards-max/forensics_strict_edge_reward": 0.21046458184719086,
      "rewards-max/forensics_tiou_reward": 0.9499178528785706,
      "rewards/forensics_iou_reward": 0.6888976991176605,
      "rewards/forensics_strict_edge_reward": 0.026326528750360012,
      "rewards/forensics_tiou_reward": 0.6539295315742493,
      "step": 973
    },
    {
      "completion_length": 22.015625,
      "epoch": 2.989247311827957,
      "grad_norm": 8.67856216430664,
      "kl": 0.390625,
      "learning_rate": 2.515384615384615e-07,
      "loss": -0.026,
      "reward": 0.7103263437747955,
      "reward_std": 0.05654008314013481,
      "reward_std_all": 0.07739812135696411,
      "rewards-max/forensics_iou_reward": 0.9858190715312958,
      "rewards-max/forensics_strict_edge_reward": 0.6302271485328674,
      "rewards-max/forensics_tiou_reward": 0.978338211774826,
      "rewards/forensics_iou_reward": 0.6153021454811096,
      "rewards/forensics_strict_edge_reward": 0.07735111378133297,
      "rewards/forensics_tiou_reward": 0.5756059885025024,
      "step": 974
    },
    {
      "completion_length": 24.296875,
      "epoch": 2.9923195084485408,
      "grad_norm": 7.226021766662598,
      "kl": 0.5,
      "learning_rate": 2.507692307692307e-07,
      "loss": -0.0214,
      "reward": 0.7825815677642822,
      "reward_std": 0.03407133324071765,
      "reward_std_all": 0.03238355182111263,
      "rewards-max/forensics_iou_reward": 0.9750917851924896,
      "rewards-max/forensics_strict_edge_reward": 0.507667601108551,
      "rewards-max/forensics_tiou_reward": 0.965439110994339,
      "rewards/forensics_iou_reward": 0.7188751697540283,
      "rewards/forensics_strict_edge_reward": 0.05834922939538956,
      "rewards/forensics_tiou_reward": 0.6640764772891998,
      "step": 975
    },
    {
      "completion_length": 29.28125,
      "epoch": 2.9953917050691246,
      "grad_norm": 6.60842752456665,
      "kl": 0.416015625,
      "learning_rate": 2.5e-07,
      "loss": -0.0324,
      "reward": 0.7950111925601959,
      "reward_std": 0.011191420257091522,
      "reward_std_all": 0.012944920687004924,
      "rewards-max/forensics_iou_reward": 0.9929638206958771,
      "rewards-max/forensics_strict_edge_reward": 0.6618969887495041,
      "rewards-max/forensics_tiou_reward": 0.9903427958488464,
      "rewards/forensics_iou_reward": 0.7027642726898193,
      "rewards/forensics_strict_edge_reward": 0.07810438238084316,
      "rewards/forensics_tiou_reward": 0.6555088460445404,
      "step": 976
    },
    {
      "completion_length": 24.34375,
      "epoch": 2.998463901689708,
      "grad_norm": 7.439937591552734,
      "kl": 0.392578125,
      "learning_rate": 2.492307692307692e-07,
      "loss": 0.0661,
      "reward": 0.7256954610347748,
      "reward_std": 0.014200021978467703,
      "reward_std_all": 0.04899154044687748,
      "rewards-max/forensics_iou_reward": 0.9862622618675232,
      "rewards-max/forensics_strict_edge_reward": 0.6419725120067596,
      "rewards-max/forensics_tiou_reward": 0.9802044034004211,
      "rewards/forensics_iou_reward": 0.6540676355361938,
      "rewards/forensics_strict_edge_reward": 0.07204312831163406,
      "rewards/forensics_tiou_reward": 0.6094855070114136,
      "step": 977
    },
    {
      "completion_length": 29.600000381469727,
      "epoch": 3.0,
      "grad_norm": 3.6642234325408936,
      "kl": 0.40625,
      "learning_rate": 2.4846153846153847e-07,
      "loss": 0.023,
      "reward": 0.7376903891563416,
      "reward_std": 0.07607545703649521,
      "reward_std_all": 0.08741853386163712,
      "rewards-max/forensics_iou_reward": 0.9910556674003601,
      "rewards-max/forensics_strict_edge_reward": 0.07939393073320389,
      "rewards-max/forensics_tiou_reward": 0.9862706661224365,
      "rewards/forensics_iou_reward": 0.6032158732414246,
      "rewards/forensics_strict_edge_reward": 0.015887241810560226,
      "rewards/forensics_tiou_reward": 0.5456904172897339,
      "step": 978
    },
    {
      "completion_length": 23.234375,
      "epoch": 3.003072196620584,
      "grad_norm": 5.916833400726318,
      "kl": 0.35546875,
      "learning_rate": 2.476923076923077e-07,
      "loss": -0.0018,
      "reward": 0.7681535184383392,
      "reward_std": 0.016578633221797645,
      "reward_std_all": 0.02995722263585776,
      "rewards-max/forensics_iou_reward": 0.9746636152267456,
      "rewards-max/forensics_strict_edge_reward": 0.6094030141830444,
      "rewards-max/forensics_tiou_reward": 0.9677349925041199,
      "rewards/forensics_iou_reward": 0.716690331697464,
      "rewards/forensics_strict_edge_reward": 0.06966324150562286,
      "rewards/forensics_tiou_reward": 0.6643266081809998,
      "step": 979
    },
    {
      "completion_length": 24.328125,
      "epoch": 3.0061443932411676,
      "grad_norm": 8.826061248779297,
      "kl": 0.5390625,
      "learning_rate": 2.469230769230769e-07,
      "loss": 0.0025,
      "reward": 0.7317522466182709,
      "reward_std": 0.03173188120126724,
      "reward_std_all": 0.054574767127633095,
      "rewards-max/forensics_iou_reward": 0.9815940856933594,
      "rewards-max/forensics_strict_edge_reward": 0.66208615899086,
      "rewards-max/forensics_tiou_reward": 0.9765190184116364,
      "rewards/forensics_iou_reward": 0.5854824185371399,
      "rewards/forensics_strict_edge_reward": 0.047657616436481476,
      "rewards/forensics_tiou_reward": 0.538727417588234,
      "step": 980
    },
    {
      "completion_length": 24.875,
      "epoch": 3.0092165898617513,
      "grad_norm": 7.288759708404541,
      "kl": 0.7470703125,
      "learning_rate": 2.4615384615384616e-07,
      "loss": 0.1429,
      "reward": 0.7363468706607819,
      "reward_std": 0.07166535407304764,
      "reward_std_all": 0.07159323617815971,
      "rewards-max/forensics_iou_reward": 0.9622716307640076,
      "rewards-max/forensics_strict_edge_reward": 0.6234554499387741,
      "rewards-max/forensics_tiou_reward": 0.955809473991394,
      "rewards/forensics_iou_reward": 0.6420503258705139,
      "rewards/forensics_strict_edge_reward": 0.05050054099410772,
      "rewards/forensics_tiou_reward": 0.5914972722530365,
      "step": 981
    },
    {
      "completion_length": 24.96875,
      "epoch": 3.0122887864823347,
      "grad_norm": 7.292248725891113,
      "kl": 0.71484375,
      "learning_rate": 2.453846153846154e-07,
      "loss": 0.0067,
      "reward": 0.7590253353118896,
      "reward_std": 0.03768839128315449,
      "reward_std_all": 0.047270006500184536,
      "rewards-max/forensics_iou_reward": 0.9845766127109528,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9791500866413116,
      "rewards/forensics_iou_reward": 0.7245171070098877,
      "rewards/forensics_strict_edge_reward": 0.043590216897428036,
      "rewards/forensics_tiou_reward": 0.6778097450733185,
      "step": 982
    },
    {
      "completion_length": 25.5,
      "epoch": 3.0153609831029184,
      "grad_norm": 10.371917724609375,
      "kl": 0.509765625,
      "learning_rate": 2.446153846153846e-07,
      "loss": 0.045,
      "reward": 0.8009187877178192,
      "reward_std": 0.008270228281617165,
      "reward_std_all": 0.00920993136242032,
      "rewards-max/forensics_iou_reward": 0.9717459678649902,
      "rewards-max/forensics_strict_edge_reward": 0.4425395280122757,
      "rewards-max/forensics_tiou_reward": 0.9606614112854004,
      "rewards/forensics_iou_reward": 0.7291187644004822,
      "rewards/forensics_strict_edge_reward": 0.0745595321059227,
      "rewards/forensics_tiou_reward": 0.6831956505775452,
      "step": 983
    },
    {
      "completion_length": 24.5,
      "epoch": 3.0184331797235022,
      "grad_norm": 7.930580139160156,
      "kl": 0.4267578125,
      "learning_rate": 2.4384615384615386e-07,
      "loss": 0.0603,
      "reward": 0.7047976851463318,
      "reward_std": 0.06664736196398735,
      "reward_std_all": 0.08265900984406471,
      "rewards-max/forensics_iou_reward": 0.9697591364383698,
      "rewards-max/forensics_strict_edge_reward": 0.559220913797617,
      "rewards-max/forensics_tiou_reward": 0.9613493978977203,
      "rewards/forensics_iou_reward": 0.5794888138771057,
      "rewards/forensics_strict_edge_reward": 0.059416407253593206,
      "rewards/forensics_tiou_reward": 0.5283966809511185,
      "step": 984
    },
    {
      "completion_length": 24.5625,
      "epoch": 3.021505376344086,
      "grad_norm": 7.170724391937256,
      "kl": 0.57421875,
      "learning_rate": 2.4307692307692307e-07,
      "loss": 0.1442,
      "reward": 0.7683566510677338,
      "reward_std": 0.04118923097848892,
      "reward_std_all": 0.041209928691387177,
      "rewards-max/forensics_iou_reward": 0.985136091709137,
      "rewards-max/forensics_strict_edge_reward": 0.21488627046346664,
      "rewards-max/forensics_tiou_reward": 0.9774952530860901,
      "rewards/forensics_iou_reward": 0.6305156648159027,
      "rewards/forensics_strict_edge_reward": 0.02796788327395916,
      "rewards/forensics_tiou_reward": 0.5704061985015869,
      "step": 985
    },
    {
      "completion_length": 24.546875,
      "epoch": 3.02457757296467,
      "grad_norm": 7.894376754760742,
      "kl": 0.4443359375,
      "learning_rate": 2.423076923076923e-07,
      "loss": -0.0135,
      "reward": 0.7827657461166382,
      "reward_std": 0.024537439923733473,
      "reward_std_all": 0.02524331957101822,
      "rewards-max/forensics_iou_reward": 0.9967468082904816,
      "rewards-max/forensics_strict_edge_reward": 0.8418737053871155,
      "rewards-max/forensics_tiou_reward": 0.9959721565246582,
      "rewards/forensics_iou_reward": 0.7406724393367767,
      "rewards/forensics_strict_edge_reward": 0.11092953011393547,
      "rewards/forensics_tiou_reward": 0.6940978467464447,
      "step": 986
    },
    {
      "completion_length": 25.0,
      "epoch": 3.0276497695852536,
      "grad_norm": 6.274105548858643,
      "kl": 0.59765625,
      "learning_rate": 2.415384615384615e-07,
      "loss": 0.0434,
      "reward": 0.7642578184604645,
      "reward_std": 0.06595588475465775,
      "reward_std_all": 0.0597978625446558,
      "rewards-max/forensics_iou_reward": 0.9617250561714172,
      "rewards-max/forensics_strict_edge_reward": 0.5216934233903885,
      "rewards-max/forensics_tiou_reward": 0.9540719091892242,
      "rewards/forensics_iou_reward": 0.6664429306983948,
      "rewards/forensics_strict_edge_reward": 0.07761828973889351,
      "rewards/forensics_tiou_reward": 0.6138487160205841,
      "step": 987
    },
    {
      "completion_length": 21.96875,
      "epoch": 3.0307219662058373,
      "grad_norm": 8.131728172302246,
      "kl": 0.396484375,
      "learning_rate": 2.4076923076923076e-07,
      "loss": 0.0388,
      "reward": 0.6929543018341064,
      "reward_std": 0.06642000563442707,
      "reward_std_all": 0.09565155021846294,
      "rewards-max/forensics_iou_reward": 0.9829473495483398,
      "rewards-max/forensics_strict_edge_reward": 0.4272303581237793,
      "rewards-max/forensics_tiou_reward": 0.9767010807991028,
      "rewards/forensics_iou_reward": 0.6095143556594849,
      "rewards/forensics_strict_edge_reward": 0.07077623531222343,
      "rewards/forensics_tiou_reward": 0.5702269673347473,
      "step": 988
    },
    {
      "completion_length": 28.84375,
      "epoch": 3.033794162826421,
      "grad_norm": 6.2508769035339355,
      "kl": 0.419921875,
      "learning_rate": 2.4e-07,
      "loss": 0.0077,
      "reward": 0.6668916046619415,
      "reward_std": 0.08565603010356426,
      "reward_std_all": 0.10909663140773773,
      "rewards-max/forensics_iou_reward": 0.9693748950958252,
      "rewards-max/forensics_strict_edge_reward": 0.3701959550380707,
      "rewards-max/forensics_tiou_reward": 0.9624678492546082,
      "rewards/forensics_iou_reward": 0.5630552023649216,
      "rewards/forensics_strict_edge_reward": 0.03460230678319931,
      "rewards/forensics_tiou_reward": 0.5193289965391159,
      "step": 989
    },
    {
      "completion_length": 25.203125,
      "epoch": 3.0368663594470044,
      "grad_norm": 5.866549015045166,
      "kl": 0.51953125,
      "learning_rate": 2.392307692307692e-07,
      "loss": 0.0927,
      "reward": 0.7863594889640808,
      "reward_std": 0.03298621438443661,
      "reward_std_all": 0.030549170449376106,
      "rewards-max/forensics_iou_reward": 0.9878263175487518,
      "rewards-max/forensics_strict_edge_reward": 0.5928491055965424,
      "rewards-max/forensics_tiou_reward": 0.982584536075592,
      "rewards/forensics_iou_reward": 0.7087981998920441,
      "rewards/forensics_strict_edge_reward": 0.05096974968910217,
      "rewards/forensics_tiou_reward": 0.654115617275238,
      "step": 990
    },
    {
      "completion_length": 24.953125,
      "epoch": 3.0399385560675882,
      "grad_norm": 7.260572910308838,
      "kl": 0.4951171875,
      "learning_rate": 2.3846153846153846e-07,
      "loss": 0.0251,
      "reward": 0.7990249991416931,
      "reward_std": 0.012179388897493482,
      "reward_std_all": 0.01267201965674758,
      "rewards-max/forensics_iou_reward": 0.9749560952186584,
      "rewards-max/forensics_strict_edge_reward": 0.3330855369567871,
      "rewards-max/forensics_tiou_reward": 0.9664692878723145,
      "rewards/forensics_iou_reward": 0.7510571777820587,
      "rewards/forensics_strict_edge_reward": 0.03796636965125799,
      "rewards/forensics_tiou_reward": 0.7037824988365173,
      "step": 991
    },
    {
      "completion_length": 27.6875,
      "epoch": 3.043010752688172,
      "grad_norm": 8.968742370605469,
      "kl": 0.521484375,
      "learning_rate": 2.376923076923077e-07,
      "loss": 0.0286,
      "reward": 0.782916784286499,
      "reward_std": 0.032670303888153285,
      "reward_std_all": 0.032464431365951896,
      "rewards-max/forensics_iou_reward": 0.9712316989898682,
      "rewards-max/forensics_strict_edge_reward": 0.45919308066368103,
      "rewards-max/forensics_tiou_reward": 0.9586973488330841,
      "rewards/forensics_iou_reward": 0.7414284944534302,
      "rewards/forensics_strict_edge_reward": 0.053141118958592415,
      "rewards/forensics_tiou_reward": 0.6893167197704315,
      "step": 992
    },
    {
      "completion_length": 27.09375,
      "epoch": 3.046082949308756,
      "grad_norm": 7.9773454666137695,
      "kl": 0.404296875,
      "learning_rate": 2.369230769230769e-07,
      "loss": 0.0744,
      "reward": 0.7259176969528198,
      "reward_std": 0.09837922267615795,
      "reward_std_all": 0.0931142270565033,
      "rewards-max/forensics_iou_reward": 0.9834946095943451,
      "rewards-max/forensics_strict_edge_reward": 0.5417766571044922,
      "rewards-max/forensics_tiou_reward": 0.9774795770645142,
      "rewards/forensics_iou_reward": 0.6846732497215271,
      "rewards/forensics_strict_edge_reward": 0.09652024507522583,
      "rewards/forensics_tiou_reward": 0.6476216614246368,
      "step": 993
    },
    {
      "completion_length": 28.34375,
      "epoch": 3.0491551459293396,
      "grad_norm": 6.346601486206055,
      "kl": 0.4248046875,
      "learning_rate": 2.3615384615384615e-07,
      "loss": 0.0157,
      "reward": 0.7275756597518921,
      "reward_std": 0.037989383563399315,
      "reward_std_all": 0.05777405574917793,
      "rewards-max/forensics_iou_reward": 0.9820462167263031,
      "rewards-max/forensics_strict_edge_reward": 0.4704190194606781,
      "rewards-max/forensics_tiou_reward": 0.9754172265529633,
      "rewards/forensics_iou_reward": 0.6309175193309784,
      "rewards/forensics_strict_edge_reward": 0.05518621578812599,
      "rewards/forensics_tiou_reward": 0.5896501541137695,
      "step": 994
    },
    {
      "completion_length": 24.859375,
      "epoch": 3.0522273425499233,
      "grad_norm": 6.692965507507324,
      "kl": 0.41015625,
      "learning_rate": 2.3538461538461537e-07,
      "loss": 0.1281,
      "reward": 0.7631158530712128,
      "reward_std": 0.06009461730718613,
      "reward_std_all": 0.055910009890794754,
      "rewards-max/forensics_iou_reward": 0.9501354992389679,
      "rewards-max/forensics_strict_edge_reward": 0.5022076666355133,
      "rewards-max/forensics_tiou_reward": 0.9361652731895447,
      "rewards/forensics_iou_reward": 0.6554906964302063,
      "rewards/forensics_strict_edge_reward": 0.06790447980165482,
      "rewards/forensics_tiou_reward": 0.6071953773498535,
      "step": 995
    },
    {
      "completion_length": 24.28125,
      "epoch": 3.055299539170507,
      "grad_norm": 7.40266227722168,
      "kl": 0.4794921875,
      "learning_rate": 2.346153846153846e-07,
      "loss": 0.0789,
      "reward": 0.7553152441978455,
      "reward_std": 0.07532260194420815,
      "reward_std_all": 0.06848542205989361,
      "rewards-max/forensics_iou_reward": 0.9713511168956757,
      "rewards-max/forensics_strict_edge_reward": 0.7403165996074677,
      "rewards-max/forensics_tiou_reward": 0.9626804292201996,
      "rewards/forensics_iou_reward": 0.6619764268398285,
      "rewards/forensics_strict_edge_reward": 0.06603781133890152,
      "rewards/forensics_tiou_reward": 0.6103170812129974,
      "step": 996
    },
    {
      "completion_length": 25.5,
      "epoch": 3.0583717357910905,
      "grad_norm": 7.204547882080078,
      "kl": 0.455078125,
      "learning_rate": 2.3384615384615382e-07,
      "loss": 0.0945,
      "reward": 0.7371510565280914,
      "reward_std": 0.06632686406373978,
      "reward_std_all": 0.06787071749567986,
      "rewards-max/forensics_iou_reward": 0.9927390813827515,
      "rewards-max/forensics_strict_edge_reward": 0.5111399292945862,
      "rewards-max/forensics_tiou_reward": 0.9875231683254242,
      "rewards/forensics_iou_reward": 0.6606143414974213,
      "rewards/forensics_strict_edge_reward": 0.08736864104866982,
      "rewards/forensics_tiou_reward": 0.6097903847694397,
      "step": 997
    },
    {
      "completion_length": 26.875,
      "epoch": 3.0614439324116742,
      "grad_norm": 7.281313419342041,
      "kl": 0.443359375,
      "learning_rate": 2.3307692307692306e-07,
      "loss": 0.1128,
      "reward": 0.7330597937107086,
      "reward_std": 0.11611581966280937,
      "reward_std_all": 0.10362879186868668,
      "rewards-max/forensics_iou_reward": 0.9783222377300262,
      "rewards-max/forensics_strict_edge_reward": 0.7186079621315002,
      "rewards-max/forensics_tiou_reward": 0.9710113406181335,
      "rewards/forensics_iou_reward": 0.6288556456565857,
      "rewards/forensics_strict_edge_reward": 0.09807689860463142,
      "rewards/forensics_tiou_reward": 0.5834128856658936,
      "step": 998
    },
    {
      "completion_length": 25.078125,
      "epoch": 3.064516129032258,
      "grad_norm": 7.603899955749512,
      "kl": 0.423828125,
      "learning_rate": 2.323076923076923e-07,
      "loss": 0.0265,
      "reward": 0.7624679505825043,
      "reward_std": 0.010676795151084661,
      "reward_std_all": 0.027463514357805252,
      "rewards-max/forensics_iou_reward": 0.9948806166648865,
      "rewards-max/forensics_strict_edge_reward": 0.6221634447574615,
      "rewards-max/forensics_tiou_reward": 0.9923611283302307,
      "rewards/forensics_iou_reward": 0.6887555122375488,
      "rewards/forensics_strict_edge_reward": 0.0688934475183487,
      "rewards/forensics_tiou_reward": 0.6473195254802704,
      "step": 999
    },
    {
      "completion_length": 25.609375,
      "epoch": 3.067588325652842,
      "grad_norm": 6.900915145874023,
      "kl": 0.5,
      "learning_rate": 2.3153846153846151e-07,
      "loss": 0.0589,
      "reward": 0.7694089710712433,
      "reward_std": 0.06165958009660244,
      "reward_std_all": 0.05528590269386768,
      "rewards-max/forensics_iou_reward": 0.9783410131931305,
      "rewards-max/forensics_strict_edge_reward": 0.6824378371238708,
      "rewards-max/forensics_tiou_reward": 0.9751712679862976,
      "rewards/forensics_iou_reward": 0.7084122598171234,
      "rewards/forensics_strict_edge_reward": 0.09860186278820038,
      "rewards/forensics_tiou_reward": 0.6690333187580109,
      "step": 1000
    },
    {
      "completion_length": 26.171875,
      "epoch": 3.0706605222734256,
      "grad_norm": 7.098654747009277,
      "kl": 0.361328125,
      "learning_rate": 2.3076923076923078e-07,
      "loss": 0.0488,
      "reward": 0.7704761326313019,
      "reward_std": 0.03748927637934685,
      "reward_std_all": 0.03769118990749121,
      "rewards-max/forensics_iou_reward": 0.9362440705299377,
      "rewards-max/forensics_strict_edge_reward": 0.27784282714128494,
      "rewards-max/forensics_tiou_reward": 0.9159473478794098,
      "rewards/forensics_iou_reward": 0.6766576170921326,
      "rewards/forensics_strict_edge_reward": 0.04311363026499748,
      "rewards/forensics_tiou_reward": 0.6347230672836304,
      "step": 1001
    },
    {
      "completion_length": 26.21875,
      "epoch": 3.0737327188940093,
      "grad_norm": 6.975167274475098,
      "kl": 0.3642578125,
      "learning_rate": 2.3e-07,
      "loss": -0.0138,
      "reward": 0.7386361360549927,
      "reward_std": 0.014241697266697884,
      "reward_std_all": 0.04105275683104992,
      "rewards-max/forensics_iou_reward": 0.9801526963710785,
      "rewards-max/forensics_strict_edge_reward": 0.3328862637281418,
      "rewards-max/forensics_tiou_reward": 0.9745737612247467,
      "rewards/forensics_iou_reward": 0.6096757054328918,
      "rewards/forensics_strict_edge_reward": 0.05186568945646286,
      "rewards/forensics_tiou_reward": 0.5531857013702393,
      "step": 1002
    },
    {
      "completion_length": 25.609375,
      "epoch": 3.076804915514593,
      "grad_norm": 6.554322242736816,
      "kl": 0.6083984375,
      "learning_rate": 2.2923076923076924e-07,
      "loss": 0.0175,
      "reward": 0.7804866433143616,
      "reward_std": 0.045107392594218254,
      "reward_std_all": 0.04084593243896961,
      "rewards-max/forensics_iou_reward": 0.9824976623058319,
      "rewards-max/forensics_strict_edge_reward": 0.6660739481449127,
      "rewards-max/forensics_tiou_reward": 0.9776334464550018,
      "rewards/forensics_iou_reward": 0.7473064363002777,
      "rewards/forensics_strict_edge_reward": 0.10436438396573067,
      "rewards/forensics_tiou_reward": 0.7047043144702911,
      "step": 1003
    },
    {
      "completion_length": 25.90625,
      "epoch": 3.0798771121351765,
      "grad_norm": 6.534648895263672,
      "kl": 0.412109375,
      "learning_rate": 2.2846153846153845e-07,
      "loss": 0.0637,
      "reward": 0.7479470074176788,
      "reward_std": 0.09912565164268017,
      "reward_std_all": 0.08765736594796181,
      "rewards-max/forensics_iou_reward": 0.9951456189155579,
      "rewards-max/forensics_strict_edge_reward": 0.8704091012477875,
      "rewards-max/forensics_tiou_reward": 0.9929197132587433,
      "rewards/forensics_iou_reward": 0.6183401644229889,
      "rewards/forensics_strict_edge_reward": 0.10943583399057388,
      "rewards/forensics_tiou_reward": 0.5767926871776581,
      "step": 1004
    },
    {
      "completion_length": 23.234375,
      "epoch": 3.0829493087557602,
      "grad_norm": 7.628534317016602,
      "kl": 0.443359375,
      "learning_rate": 2.276923076923077e-07,
      "loss": 0.0085,
      "reward": 0.7531310617923737,
      "reward_std": 0.044086070731282234,
      "reward_std_all": 0.04743656516075134,
      "rewards-max/forensics_iou_reward": 0.989714503288269,
      "rewards-max/forensics_strict_edge_reward": 0.6656602919101715,
      "rewards-max/forensics_tiou_reward": 0.9862554371356964,
      "rewards/forensics_iou_reward": 0.6937745213508606,
      "rewards/forensics_strict_edge_reward": 0.10556994751095772,
      "rewards/forensics_tiou_reward": 0.6491920351982117,
      "step": 1005
    },
    {
      "completion_length": 23.171875,
      "epoch": 3.086021505376344,
      "grad_norm": 6.839277267456055,
      "kl": 0.4658203125,
      "learning_rate": 2.269230769230769e-07,
      "loss": 0.0661,
      "reward": 0.7737792134284973,
      "reward_std": 0.055359577760100365,
      "reward_std_all": 0.04990246519446373,
      "rewards-max/forensics_iou_reward": 0.9871645271778107,
      "rewards-max/forensics_strict_edge_reward": 0.7412298321723938,
      "rewards-max/forensics_tiou_reward": 0.9838035702705383,
      "rewards/forensics_iou_reward": 0.6926809549331665,
      "rewards/forensics_strict_edge_reward": 0.0647360272705555,
      "rewards/forensics_tiou_reward": 0.6456750333309174,
      "step": 1006
    },
    {
      "completion_length": 28.125,
      "epoch": 3.089093701996928,
      "grad_norm": 7.419963359832764,
      "kl": 0.4677734375,
      "learning_rate": 2.2615384615384614e-07,
      "loss": 0.0877,
      "reward": 0.782614141702652,
      "reward_std": 0.03744600620120764,
      "reward_std_all": 0.03468109481036663,
      "rewards-max/forensics_iou_reward": 0.9878717958927155,
      "rewards-max/forensics_strict_edge_reward": 0.5434702187776566,
      "rewards-max/forensics_tiou_reward": 0.9837975800037384,
      "rewards/forensics_iou_reward": 0.6849855184555054,
      "rewards/forensics_strict_edge_reward": 0.07125039398670197,
      "rewards/forensics_tiou_reward": 0.631306916475296,
      "step": 1007
    },
    {
      "completion_length": 23.21875,
      "epoch": 3.0921658986175116,
      "grad_norm": 9.168730735778809,
      "kl": 0.3681640625,
      "learning_rate": 2.2538461538461536e-07,
      "loss": -0.0325,
      "reward": 0.700215756893158,
      "reward_std": 0.0279210593434982,
      "reward_std_all": 0.0712285041809082,
      "rewards-max/forensics_iou_reward": 0.9773138463497162,
      "rewards-max/forensics_strict_edge_reward": 0.4985542297363281,
      "rewards-max/forensics_tiou_reward": 0.9680616557598114,
      "rewards/forensics_iou_reward": 0.6558068990707397,
      "rewards/forensics_strict_edge_reward": 0.06298770755529404,
      "rewards/forensics_tiou_reward": 0.6188927590847015,
      "step": 1008
    },
    {
      "completion_length": 24.546875,
      "epoch": 3.0952380952380953,
      "grad_norm": 7.769938945770264,
      "kl": 0.3623046875,
      "learning_rate": 2.246153846153846e-07,
      "loss": 0.0253,
      "reward": 0.7494067549705505,
      "reward_std": 0.0693744255695492,
      "reward_std_all": 0.06786808744072914,
      "rewards-max/forensics_iou_reward": 0.942810446023941,
      "rewards-max/forensics_strict_edge_reward": 0.3720540702342987,
      "rewards-max/forensics_tiou_reward": 0.9269519150257111,
      "rewards/forensics_iou_reward": 0.6840915083885193,
      "rewards/forensics_strict_edge_reward": 0.05878311488777399,
      "rewards/forensics_tiou_reward": 0.6395978331565857,
      "step": 1009
    },
    {
      "completion_length": 25.4375,
      "epoch": 3.098310291858679,
      "grad_norm": 9.954432487487793,
      "kl": 0.4521484375,
      "learning_rate": 2.238461538461538e-07,
      "loss": 0.0076,
      "reward": 0.7844268381595612,
      "reward_std": 0.03349410742521286,
      "reward_std_all": 0.03146562445908785,
      "rewards-max/forensics_iou_reward": 0.9829787313938141,
      "rewards-max/forensics_strict_edge_reward": 0.7246644794940948,
      "rewards-max/forensics_tiou_reward": 0.97719407081604,
      "rewards/forensics_iou_reward": 0.7083850800991058,
      "rewards/forensics_strict_edge_reward": 0.06496932171285152,
      "rewards/forensics_tiou_reward": 0.6559259295463562,
      "step": 1010
    },
    {
      "completion_length": 29.875,
      "epoch": 3.1013824884792625,
      "grad_norm": 6.07760763168335,
      "kl": 0.3583984375,
      "learning_rate": 2.2307692307692308e-07,
      "loss": -0.0883,
      "reward": 0.7783924341201782,
      "reward_std": 0.025393720716238022,
      "reward_std_all": 0.027024426497519016,
      "rewards-max/forensics_iou_reward": 0.9668844044208527,
      "rewards-max/forensics_strict_edge_reward": 0.4950350970029831,
      "rewards-max/forensics_tiou_reward": 0.9516211450099945,
      "rewards/forensics_iou_reward": 0.7040838897228241,
      "rewards/forensics_strict_edge_reward": 0.07528435066342354,
      "rewards/forensics_tiou_reward": 0.6609368920326233,
      "step": 1011
    },
    {
      "completion_length": 22.484375,
      "epoch": 3.1044546850998462,
      "grad_norm": 7.134568214416504,
      "kl": 0.40234375,
      "learning_rate": 2.2230769230769232e-07,
      "loss": 0.0237,
      "reward": 0.7938498854637146,
      "reward_std": 0.014975257450714707,
      "reward_std_all": 0.015532954595983028,
      "rewards-max/forensics_iou_reward": 0.9896911978721619,
      "rewards-max/forensics_strict_edge_reward": 0.7542104423046112,
      "rewards-max/forensics_tiou_reward": 0.9863120317459106,
      "rewards/forensics_iou_reward": 0.7539796531200409,
      "rewards/forensics_strict_edge_reward": 0.11503370851278305,
      "rewards/forensics_tiou_reward": 0.6915073990821838,
      "step": 1012
    },
    {
      "completion_length": 26.1875,
      "epoch": 3.10752688172043,
      "grad_norm": 7.515931129455566,
      "kl": 0.4404296875,
      "learning_rate": 2.2153846153846153e-07,
      "loss": -0.0423,
      "reward": 0.7596031427383423,
      "reward_std": 0.003206115507055074,
      "reward_std_all": 0.027310401666909456,
      "rewards-max/forensics_iou_reward": 0.9857815504074097,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9797409772872925,
      "rewards/forensics_iou_reward": 0.712631344795227,
      "rewards/forensics_strict_edge_reward": 0.09117892757058144,
      "rewards/forensics_tiou_reward": 0.6628924310207367,
      "step": 1013
    },
    {
      "completion_length": 27.171875,
      "epoch": 3.110599078341014,
      "grad_norm": 4.773848533630371,
      "kl": 0.4140625,
      "learning_rate": 2.2076923076923077e-07,
      "loss": -0.0395,
      "reward": 0.7863344550132751,
      "reward_std": 0.021579738706350327,
      "reward_std_all": 0.022221320308744907,
      "rewards-max/forensics_iou_reward": 0.9824478328227997,
      "rewards-max/forensics_strict_edge_reward": 0.5932882726192474,
      "rewards-max/forensics_tiou_reward": 0.9778445661067963,
      "rewards/forensics_iou_reward": 0.7489898800849915,
      "rewards/forensics_strict_edge_reward": 0.08283155038952827,
      "rewards/forensics_tiou_reward": 0.7118784785270691,
      "step": 1014
    },
    {
      "completion_length": 25.109375,
      "epoch": 3.1136712749615976,
      "grad_norm": 7.020647048950195,
      "kl": 0.3916015625,
      "learning_rate": 2.1999999999999998e-07,
      "loss": -0.0135,
      "reward": 0.787085622549057,
      "reward_std": 0.032597113167867064,
      "reward_std_all": 0.030192150734364986,
      "rewards-max/forensics_iou_reward": 0.989019900560379,
      "rewards-max/forensics_strict_edge_reward": 0.6102284789085388,
      "rewards-max/forensics_tiou_reward": 0.9836932420730591,
      "rewards/forensics_iou_reward": 0.7485058605670929,
      "rewards/forensics_strict_edge_reward": 0.07352333888411522,
      "rewards/forensics_tiou_reward": 0.7178189158439636,
      "step": 1015
    },
    {
      "completion_length": 28.109375,
      "epoch": 3.1167434715821813,
      "grad_norm": 8.387072563171387,
      "kl": 0.998046875,
      "learning_rate": 2.1923076923076922e-07,
      "loss": 0.1052,
      "reward": 0.778295248746872,
      "reward_std": 0.0418283436447382,
      "reward_std_all": 0.03885506093502045,
      "rewards-max/forensics_iou_reward": 0.9735457003116608,
      "rewards-max/forensics_strict_edge_reward": 0.5486612468957901,
      "rewards-max/forensics_tiou_reward": 0.9689018428325653,
      "rewards/forensics_iou_reward": 0.7002211511135101,
      "rewards/forensics_strict_edge_reward": 0.07397640869021416,
      "rewards/forensics_tiou_reward": 0.6548627614974976,
      "step": 1016
    },
    {
      "completion_length": 27.765625,
      "epoch": 3.119815668202765,
      "grad_norm": 7.429429531097412,
      "kl": 0.5546875,
      "learning_rate": 2.1846153846153844e-07,
      "loss": 0.1301,
      "reward": 0.728283703327179,
      "reward_std": 0.08128640800714493,
      "reward_std_all": 0.07942108809947968,
      "rewards-max/forensics_iou_reward": 0.9767085611820221,
      "rewards-max/forensics_strict_edge_reward": 0.472219854593277,
      "rewards-max/forensics_tiou_reward": 0.9673896133899689,
      "rewards/forensics_iou_reward": 0.6434434056282043,
      "rewards/forensics_strict_edge_reward": 0.059190426021814346,
      "rewards/forensics_tiou_reward": 0.6020358800888062,
      "step": 1017
    },
    {
      "completion_length": 28.78125,
      "epoch": 3.122887864823349,
      "grad_norm": 6.473794937133789,
      "kl": 0.380859375,
      "learning_rate": 2.1769230769230768e-07,
      "loss": 0.1017,
      "reward": 0.7698265314102173,
      "reward_std": 0.040189904160797596,
      "reward_std_all": 0.04173752944916487,
      "rewards-max/forensics_iou_reward": 0.9679162800312042,
      "rewards-max/forensics_strict_edge_reward": 0.349711574614048,
      "rewards-max/forensics_tiou_reward": 0.9557690322399139,
      "rewards/forensics_iou_reward": 0.681318461894989,
      "rewards/forensics_strict_edge_reward": 0.0395921035669744,
      "rewards/forensics_tiou_reward": 0.6381381750106812,
      "step": 1018
    },
    {
      "completion_length": 27.109375,
      "epoch": 3.1259600614439322,
      "grad_norm": 7.467581272125244,
      "kl": 0.4287109375,
      "learning_rate": 2.169230769230769e-07,
      "loss": 0.1385,
      "reward": 0.7374728322029114,
      "reward_std": 0.08044353500008583,
      "reward_std_all": 0.07647286541759968,
      "rewards-max/forensics_iou_reward": 0.9761205911636353,
      "rewards-max/forensics_strict_edge_reward": 0.5240353047847748,
      "rewards-max/forensics_tiou_reward": 0.9700237214565277,
      "rewards/forensics_iou_reward": 0.6290975213050842,
      "rewards/forensics_strict_edge_reward": 0.0701314490288496,
      "rewards/forensics_tiou_reward": 0.5896319448947906,
      "step": 1019
    },
    {
      "completion_length": 29.03125,
      "epoch": 3.129032258064516,
      "grad_norm": 8.18212890625,
      "kl": 0.640625,
      "learning_rate": 2.1615384615384613e-07,
      "loss": 0.0826,
      "reward": 0.7597865462303162,
      "reward_std": 0.05976407416164875,
      "reward_std_all": 0.056248873472213745,
      "rewards-max/forensics_iou_reward": 0.9897540509700775,
      "rewards-max/forensics_strict_edge_reward": 0.7069477140903473,
      "rewards-max/forensics_tiou_reward": 0.9875409007072449,
      "rewards/forensics_iou_reward": 0.6927394270896912,
      "rewards/forensics_strict_edge_reward": 0.07842533104121685,
      "rewards/forensics_tiou_reward": 0.6526601612567902,
      "step": 1020
    },
    {
      "completion_length": 29.15625,
      "epoch": 3.1321044546851,
      "grad_norm": 8.622498512268066,
      "kl": 0.4501953125,
      "learning_rate": 2.153846153846154e-07,
      "loss": 0.0132,
      "reward": 0.7584388852119446,
      "reward_std": 0.056656770408153534,
      "reward_std_all": 0.05424618534743786,
      "rewards-max/forensics_iou_reward": 0.9582662284374237,
      "rewards-max/forensics_strict_edge_reward": 0.5669038891792297,
      "rewards-max/forensics_tiou_reward": 0.9492726922035217,
      "rewards/forensics_iou_reward": 0.6415287852287292,
      "rewards/forensics_strict_edge_reward": 0.06738924235105515,
      "rewards/forensics_tiou_reward": 0.5972063839435577,
      "step": 1021
    },
    {
      "completion_length": 24.90625,
      "epoch": 3.1351766513056836,
      "grad_norm": 7.871446132659912,
      "kl": 0.568359375,
      "learning_rate": 2.146153846153846e-07,
      "loss": -0.016,
      "reward": 0.7395979166030884,
      "reward_std": 0.08149751112796366,
      "reward_std_all": 0.07728703645989299,
      "rewards-max/forensics_iou_reward": 0.9697956442832947,
      "rewards-max/forensics_strict_edge_reward": 0.36687174439430237,
      "rewards-max/forensics_tiou_reward": 0.9613049626350403,
      "rewards/forensics_iou_reward": 0.607691764831543,
      "rewards/forensics_strict_edge_reward": 0.04340003617107868,
      "rewards/forensics_tiou_reward": 0.5682023465633392,
      "step": 1022
    },
    {
      "completion_length": 27.09375,
      "epoch": 3.1382488479262673,
      "grad_norm": 8.895766258239746,
      "kl": 0.3515625,
      "learning_rate": 2.1384615384615385e-07,
      "loss": 0.0131,
      "reward": 0.7375945150852203,
      "reward_std": 0.03169739060103893,
      "reward_std_all": 0.05152367800474167,
      "rewards-max/forensics_iou_reward": 0.9965170323848724,
      "rewards-max/forensics_strict_edge_reward": 0.8704091012477875,
      "rewards-max/forensics_tiou_reward": 0.9932765364646912,
      "rewards/forensics_iou_reward": 0.6651434004306793,
      "rewards/forensics_strict_edge_reward": 0.12726494297385216,
      "rewards/forensics_tiou_reward": 0.6182529628276825,
      "step": 1023
    },
    {
      "completion_length": 26.40625,
      "epoch": 3.141321044546851,
      "grad_norm": 11.839967727661133,
      "kl": 0.4638671875,
      "learning_rate": 2.1307692307692307e-07,
      "loss": -0.0041,
      "reward": 0.7653831243515015,
      "reward_std": 0.04615476634353399,
      "reward_std_all": 0.045241646468639374,
      "rewards-max/forensics_iou_reward": 0.9939807653427124,
      "rewards-max/forensics_strict_edge_reward": 0.8079957664012909,
      "rewards-max/forensics_tiou_reward": 0.9927531778812408,
      "rewards/forensics_iou_reward": 0.6427915692329407,
      "rewards/forensics_strict_edge_reward": 0.06839395873248577,
      "rewards/forensics_tiou_reward": 0.5974441170692444,
      "step": 1024
    },
    {
      "completion_length": 23.984375,
      "epoch": 3.144393241167435,
      "grad_norm": 9.869505882263184,
      "kl": 0.69140625,
      "learning_rate": 2.123076923076923e-07,
      "loss": 0.1051,
      "reward": 0.752933919429779,
      "reward_std": 0.07549773994833231,
      "reward_std_all": 0.06943740602582693,
      "rewards-max/forensics_iou_reward": 0.9948854148387909,
      "rewards-max/forensics_strict_edge_reward": 0.6215263903141022,
      "rewards-max/forensics_tiou_reward": 0.9920713305473328,
      "rewards/forensics_iou_reward": 0.699832558631897,
      "rewards/forensics_strict_edge_reward": 0.08589852042496204,
      "rewards/forensics_tiou_reward": 0.6518026292324066,
      "step": 1025
    },
    {
      "completion_length": 29.3125,
      "epoch": 3.1474654377880182,
      "grad_norm": 7.846224308013916,
      "kl": 0.458984375,
      "learning_rate": 2.1153846153846152e-07,
      "loss": -0.0141,
      "reward": 0.7888756990432739,
      "reward_std": 0.012795851333066821,
      "reward_std_all": 0.016035446664318442,
      "rewards-max/forensics_iou_reward": 0.9928147792816162,
      "rewards-max/forensics_strict_edge_reward": 0.4946095645427704,
      "rewards-max/forensics_tiou_reward": 0.9895313382148743,
      "rewards/forensics_iou_reward": 0.6906624436378479,
      "rewards/forensics_strict_edge_reward": 0.050020042806863785,
      "rewards/forensics_tiou_reward": 0.6521636843681335,
      "step": 1026
    },
    {
      "completion_length": 27.046875,
      "epoch": 3.150537634408602,
      "grad_norm": 9.608138084411621,
      "kl": 0.3955078125,
      "learning_rate": 2.1076923076923076e-07,
      "loss": 0.0256,
      "reward": 0.7887849807739258,
      "reward_std": 0.015681067481637,
      "reward_std_all": 0.017606650479137897,
      "rewards-max/forensics_iou_reward": 1.0,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_iou_reward": 0.6931853592395782,
      "rewards/forensics_strict_edge_reward": 0.1115003153681755,
      "rewards/forensics_tiou_reward": 0.6526619493961334,
      "step": 1027
    },
    {
      "completion_length": 22.671875,
      "epoch": 3.153609831029186,
      "grad_norm": 10.515105247497559,
      "kl": 0.447265625,
      "learning_rate": 2.0999999999999997e-07,
      "loss": 0.0473,
      "reward": 0.7505306899547577,
      "reward_std": 0.08637614920735359,
      "reward_std_all": 0.07809262722730637,
      "rewards-max/forensics_iou_reward": 0.9870998561382294,
      "rewards-max/forensics_strict_edge_reward": 0.568440854549408,
      "rewards-max/forensics_tiou_reward": 0.9790469110012054,
      "rewards/forensics_iou_reward": 0.6622260808944702,
      "rewards/forensics_strict_edge_reward": 0.04705999791622162,
      "rewards/forensics_tiou_reward": 0.6150428652763367,
      "step": 1028
    },
    {
      "completion_length": 26.34375,
      "epoch": 3.1566820276497696,
      "grad_norm": 7.174704551696777,
      "kl": 0.470703125,
      "learning_rate": 2.0923076923076921e-07,
      "loss": -0.0719,
      "reward": 0.7439121007919312,
      "reward_std": 0.06332147493958473,
      "reward_std_all": 0.06270780693739653,
      "rewards-max/forensics_iou_reward": 0.9823261797428131,
      "rewards-max/forensics_strict_edge_reward": 0.41745882481336594,
      "rewards-max/forensics_tiou_reward": 0.9766715466976166,
      "rewards/forensics_iou_reward": 0.6413812041282654,
      "rewards/forensics_strict_edge_reward": 0.06187301315367222,
      "rewards/forensics_tiou_reward": 0.6064697206020355,
      "step": 1029
    },
    {
      "completion_length": 27.0625,
      "epoch": 3.1597542242703534,
      "grad_norm": 8.140495300292969,
      "kl": 0.3759765625,
      "learning_rate": 2.0846153846153843e-07,
      "loss": 0.0832,
      "reward": 0.7297684550285339,
      "reward_std": 0.06593501567840576,
      "reward_std_all": 0.07115629874169827,
      "rewards-max/forensics_iou_reward": 0.9849444925785065,
      "rewards-max/forensics_strict_edge_reward": 0.42674338072538376,
      "rewards-max/forensics_tiou_reward": 0.976837694644928,
      "rewards/forensics_iou_reward": 0.71002197265625,
      "rewards/forensics_strict_edge_reward": 0.06489610485732555,
      "rewards/forensics_tiou_reward": 0.6740626692771912,
      "step": 1030
    },
    {
      "completion_length": 26.453125,
      "epoch": 3.162826420890937,
      "grad_norm": 6.551144123077393,
      "kl": 0.4462890625,
      "learning_rate": 2.076923076923077e-07,
      "loss": -0.0054,
      "reward": 0.7794704139232635,
      "reward_std": 0.04477799573214725,
      "reward_std_all": 0.040927879861555994,
      "rewards-max/forensics_iou_reward": 0.9810730814933777,
      "rewards-max/forensics_strict_edge_reward": 0.44907623529434204,
      "rewards-max/forensics_tiou_reward": 0.9757269024848938,
      "rewards/forensics_iou_reward": 0.7494069933891296,
      "rewards/forensics_strict_edge_reward": 0.07352723926305771,
      "rewards/forensics_tiou_reward": 0.7053371071815491,
      "step": 1031
    },
    {
      "completion_length": 28.171875,
      "epoch": 3.165898617511521,
      "grad_norm": 6.447516918182373,
      "kl": 0.3818359375,
      "learning_rate": 2.0692307692307694e-07,
      "loss": 0.0177,
      "reward": 0.7565338611602783,
      "reward_std": 0.08628152497112751,
      "reward_std_all": 0.07636627927422523,
      "rewards-max/forensics_iou_reward": 0.9819722473621368,
      "rewards-max/forensics_strict_edge_reward": 0.5277433395385742,
      "rewards-max/forensics_tiou_reward": 0.9791130125522614,
      "rewards/forensics_iou_reward": 0.7233806252479553,
      "rewards/forensics_strict_edge_reward": 0.07503386214375496,
      "rewards/forensics_tiou_reward": 0.6845536231994629,
      "step": 1032
    },
    {
      "completion_length": 24.125,
      "epoch": 3.1689708141321047,
      "grad_norm": 7.391925811767578,
      "kl": 0.474609375,
      "learning_rate": 2.0615384615384615e-07,
      "loss": -0.0227,
      "reward": 0.7019506692886353,
      "reward_std": 0.0741586722433567,
      "reward_std_all": 0.09422062337398529,
      "rewards-max/forensics_iou_reward": 0.9910093247890472,
      "rewards-max/forensics_strict_edge_reward": 0.5422053933143616,
      "rewards-max/forensics_tiou_reward": 0.983845442533493,
      "rewards/forensics_iou_reward": 0.6384539008140564,
      "rewards/forensics_strict_edge_reward": 0.06357845664024353,
      "rewards/forensics_tiou_reward": 0.5865657925605774,
      "step": 1033
    },
    {
      "completion_length": 27.25,
      "epoch": 3.172043010752688,
      "grad_norm": 8.529468536376953,
      "kl": 0.671875,
      "learning_rate": 2.053846153846154e-07,
      "loss": -0.005,
      "reward": 0.7473758459091187,
      "reward_std": 0.039616454392671585,
      "reward_std_all": 0.049604080617427826,
      "rewards-max/forensics_iou_reward": 0.9821454286575317,
      "rewards-max/forensics_strict_edge_reward": 0.5592689514160156,
      "rewards-max/forensics_tiou_reward": 0.9765320718288422,
      "rewards/forensics_iou_reward": 0.6692870259284973,
      "rewards/forensics_strict_edge_reward": 0.07040742225944996,
      "rewards/forensics_tiou_reward": 0.6324097812175751,
      "step": 1034
    },
    {
      "completion_length": 24.484375,
      "epoch": 3.175115207373272,
      "grad_norm": 7.9500908851623535,
      "kl": 0.482421875,
      "learning_rate": 2.046153846153846e-07,
      "loss": 0.0893,
      "reward": 0.7733225524425507,
      "reward_std": 0.038981524761766195,
      "reward_std_all": 0.03995030885562301,
      "rewards-max/forensics_iou_reward": 0.9859951734542847,
      "rewards-max/forensics_strict_edge_reward": 0.6862294971942902,
      "rewards-max/forensics_tiou_reward": 0.981888085603714,
      "rewards/forensics_iou_reward": 0.7119763791561127,
      "rewards/forensics_strict_edge_reward": 0.06317173689603806,
      "rewards/forensics_tiou_reward": 0.6528559923171997,
      "step": 1035
    },
    {
      "completion_length": 27.84375,
      "epoch": 3.1781874039938556,
      "grad_norm": 8.093894958496094,
      "kl": 0.3818359375,
      "learning_rate": 2.0384615384615384e-07,
      "loss": 0.0508,
      "reward": 0.6993999779224396,
      "reward_std": 0.11963033676147461,
      "reward_std_all": 0.11412299796938896,
      "rewards-max/forensics_iou_reward": 0.9480316936969757,
      "rewards-max/forensics_strict_edge_reward": 0.33508339524269104,
      "rewards-max/forensics_tiou_reward": 0.9364649057388306,
      "rewards/forensics_iou_reward": 0.6038330495357513,
      "rewards/forensics_strict_edge_reward": 0.050831232219934464,
      "rewards/forensics_tiou_reward": 0.5670951306819916,
      "step": 1036
    },
    {
      "completion_length": 31.765625,
      "epoch": 3.1812596006144394,
      "grad_norm": 5.204221725463867,
      "kl": 0.453125,
      "learning_rate": 2.0307692307692306e-07,
      "loss": -0.0076,
      "reward": 0.7858206629753113,
      "reward_std": 0.02806304325349629,
      "reward_std_all": 0.027293398045003414,
      "rewards-max/forensics_iou_reward": 0.9882033467292786,
      "rewards-max/forensics_strict_edge_reward": 0.520082414150238,
      "rewards-max/forensics_tiou_reward": 0.9837484359741211,
      "rewards/forensics_iou_reward": 0.7116729617118835,
      "rewards/forensics_strict_edge_reward": 0.07348254695534706,
      "rewards/forensics_tiou_reward": 0.6732204556465149,
      "step": 1037
    },
    {
      "completion_length": 31.078125,
      "epoch": 3.184331797235023,
      "grad_norm": 9.184662818908691,
      "kl": 0.4375,
      "learning_rate": 2.023076923076923e-07,
      "loss": -0.0375,
      "reward": 0.8008069097995758,
      "reward_std": 0.008009727811440825,
      "reward_std_all": 0.009182505775243044,
      "rewards-max/forensics_iou_reward": 0.991525411605835,
      "rewards-max/forensics_strict_edge_reward": 0.8783542811870575,
      "rewards-max/forensics_tiou_reward": 0.9900294542312622,
      "rewards/forensics_iou_reward": 0.7227528095245361,
      "rewards/forensics_strict_edge_reward": 0.09305153042078018,
      "rewards/forensics_tiou_reward": 0.6802006363868713,
      "step": 1038
    },
    {
      "completion_length": 26.5,
      "epoch": 3.187403993855607,
      "grad_norm": 7.7712202072143555,
      "kl": 0.4990234375,
      "learning_rate": 2.015384615384615e-07,
      "loss": -0.0561,
      "reward": 0.7731567025184631,
      "reward_std": 0.018595385598018765,
      "reward_std_all": 0.02491007442586124,
      "rewards-max/forensics_iou_reward": 0.9833144843578339,
      "rewards-max/forensics_strict_edge_reward": 0.5353964418172836,
      "rewards-max/forensics_tiou_reward": 0.9805068373680115,
      "rewards/forensics_iou_reward": 0.6996163725852966,
      "rewards/forensics_strict_edge_reward": 0.07791389152407646,
      "rewards/forensics_tiou_reward": 0.6424263119697571,
      "step": 1039
    },
    {
      "completion_length": 29.65625,
      "epoch": 3.1904761904761907,
      "grad_norm": 8.268730163574219,
      "kl": 0.4072265625,
      "learning_rate": 2.0076923076923075e-07,
      "loss": -0.0517,
      "reward": 0.7423849999904633,
      "reward_std": 0.058065628283657134,
      "reward_std_all": 0.059244175790809095,
      "rewards-max/forensics_iou_reward": 0.9904618859291077,
      "rewards-max/forensics_strict_edge_reward": 0.48208603262901306,
      "rewards-max/forensics_tiou_reward": 0.9844858050346375,
      "rewards/forensics_iou_reward": 0.6906315982341766,
      "rewards/forensics_strict_edge_reward": 0.07641505263745785,
      "rewards/forensics_tiou_reward": 0.6518310010433197,
      "step": 1040
    },
    {
      "completion_length": 26.609375,
      "epoch": 3.193548387096774,
      "grad_norm": 8.264449119567871,
      "kl": 0.408203125,
      "learning_rate": 2e-07,
      "loss": 0.118,
      "reward": 0.790995866060257,
      "reward_std": 0.018258183728903532,
      "reward_std_all": 0.018604597076773643,
      "rewards-max/forensics_iou_reward": 0.9917679727077484,
      "rewards-max/forensics_strict_edge_reward": 0.718743309378624,
      "rewards-max/forensics_tiou_reward": 0.989079475402832,
      "rewards/forensics_iou_reward": 0.7106462121009827,
      "rewards/forensics_strict_edge_reward": 0.0623263418674469,
      "rewards/forensics_tiou_reward": 0.6531383693218231,
      "step": 1041
    },
    {
      "completion_length": 28.078125,
      "epoch": 3.196620583717358,
      "grad_norm": 8.303853988647461,
      "kl": 0.4853515625,
      "learning_rate": 1.9923076923076923e-07,
      "loss": 0.0236,
      "reward": 0.7174831032752991,
      "reward_std": 0.0588295292109251,
      "reward_std_all": 0.07846562936902046,
      "rewards-max/forensics_iou_reward": 0.9807600378990173,
      "rewards-max/forensics_strict_edge_reward": 0.34069936722517014,
      "rewards-max/forensics_tiou_reward": 0.9743953049182892,
      "rewards/forensics_iou_reward": 0.6468167304992676,
      "rewards/forensics_strict_edge_reward": 0.07602135138586164,
      "rewards/forensics_tiou_reward": 0.6011783629655838,
      "step": 1042
    },
    {
      "completion_length": 22.640625,
      "epoch": 3.1996927803379416,
      "grad_norm": 6.992489814758301,
      "kl": 0.603515625,
      "learning_rate": 1.9846153846153844e-07,
      "loss": 0.1104,
      "reward": 0.7240093350410461,
      "reward_std": 0.051291489973664284,
      "reward_std_all": 0.06728074513375759,
      "rewards-max/forensics_iou_reward": 0.9903733134269714,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9837347865104675,
      "rewards/forensics_iou_reward": 0.6439831256866455,
      "rewards/forensics_strict_edge_reward": 0.0871758721768856,
      "rewards/forensics_tiou_reward": 0.5930120646953583,
      "step": 1043
    },
    {
      "completion_length": 29.140625,
      "epoch": 3.2027649769585254,
      "grad_norm": 8.826003074645996,
      "kl": 0.447265625,
      "learning_rate": 1.9769230769230768e-07,
      "loss": 0.0181,
      "reward": 0.7923311293125153,
      "reward_std": 0.025005293311551213,
      "reward_std_all": 0.02341161179356277,
      "rewards-max/forensics_iou_reward": 0.9903997480869293,
      "rewards-max/forensics_strict_edge_reward": 0.648044228553772,
      "rewards-max/forensics_tiou_reward": 0.9873435497283936,
      "rewards/forensics_iou_reward": 0.7679758667945862,
      "rewards/forensics_strict_edge_reward": 0.10556339100003242,
      "rewards/forensics_tiou_reward": 0.7380603551864624,
      "step": 1044
    },
    {
      "completion_length": 24.640625,
      "epoch": 3.205837173579109,
      "grad_norm": 8.447744369506836,
      "kl": 0.408203125,
      "learning_rate": 1.9692307692307693e-07,
      "loss": -0.0424,
      "reward": 0.719939649105072,
      "reward_std": 0.0783865749835968,
      "reward_std_all": 0.08091773837804794,
      "rewards-max/forensics_iou_reward": 0.9772889912128448,
      "rewards-max/forensics_strict_edge_reward": 0.3330773785710335,
      "rewards-max/forensics_tiou_reward": 0.9640708267688751,
      "rewards/forensics_iou_reward": 0.616163820028305,
      "rewards/forensics_strict_edge_reward": 0.04876540368422866,
      "rewards/forensics_tiou_reward": 0.5608711540699005,
      "step": 1045
    },
    {
      "completion_length": 24.46875,
      "epoch": 3.208909370199693,
      "grad_norm": 7.147201061248779,
      "kl": 0.48828125,
      "learning_rate": 1.9615384615384614e-07,
      "loss": 0.0118,
      "reward": 0.7281540930271149,
      "reward_std": 0.04794033616781235,
      "reward_std_all": 0.06745501607656479,
      "rewards-max/forensics_iou_reward": 0.9630216062068939,
      "rewards-max/forensics_strict_edge_reward": 0.2644351199269295,
      "rewards-max/forensics_tiou_reward": 0.951213151216507,
      "rewards/forensics_iou_reward": 0.6726495027542114,
      "rewards/forensics_strict_edge_reward": 0.026669010519981384,
      "rewards/forensics_tiou_reward": 0.6318819224834442,
      "step": 1046
    },
    {
      "completion_length": 24.1875,
      "epoch": 3.2119815668202767,
      "grad_norm": 8.401223182678223,
      "kl": 0.67578125,
      "learning_rate": 1.9538461538461538e-07,
      "loss": -0.0062,
      "reward": 0.80002361536026,
      "reward_std": 0.009230719995684922,
      "reward_std_all": 0.010247891186736524,
      "rewards-max/forensics_iou_reward": 0.9840948283672333,
      "rewards-max/forensics_strict_edge_reward": 0.700235515832901,
      "rewards-max/forensics_tiou_reward": 0.9807149171829224,
      "rewards/forensics_iou_reward": 0.7537235617637634,
      "rewards/forensics_strict_edge_reward": 0.09247148782014847,
      "rewards/forensics_tiou_reward": 0.7062203288078308,
      "step": 1047
    },
    {
      "completion_length": 24.484375,
      "epoch": 3.21505376344086,
      "grad_norm": 8.688982009887695,
      "kl": 0.8173828125,
      "learning_rate": 1.946153846153846e-07,
      "loss": 0.0619,
      "reward": 0.7507733702659607,
      "reward_std": 0.0603354312479496,
      "reward_std_all": 0.0592024102807045,
      "rewards-max/forensics_iou_reward": 0.9928095638751984,
      "rewards-max/forensics_strict_edge_reward": 0.7490177750587463,
      "rewards-max/forensics_tiou_reward": 0.9882417917251587,
      "rewards/forensics_iou_reward": 0.6651866734027863,
      "rewards/forensics_strict_edge_reward": 0.0695846751332283,
      "rewards/forensics_tiou_reward": 0.6188412606716156,
      "step": 1048
    },
    {
      "completion_length": 27.6875,
      "epoch": 3.218125960061444,
      "grad_norm": 7.080418586730957,
      "kl": 0.453125,
      "learning_rate": 1.9384615384615383e-07,
      "loss": -0.0317,
      "reward": 0.7574599981307983,
      "reward_std": 0.08108993992209435,
      "reward_std_all": 0.07254640012979507,
      "rewards-max/forensics_iou_reward": 0.9539566934108734,
      "rewards-max/forensics_strict_edge_reward": 0.49832790344953537,
      "rewards-max/forensics_tiou_reward": 0.9403861165046692,
      "rewards/forensics_iou_reward": 0.6402067542076111,
      "rewards/forensics_strict_edge_reward": 0.06413864158093929,
      "rewards/forensics_tiou_reward": 0.6014067530632019,
      "step": 1049
    },
    {
      "completion_length": 28.53125,
      "epoch": 3.2211981566820276,
      "grad_norm": 7.388866424560547,
      "kl": 0.4169921875,
      "learning_rate": 1.9307692307692307e-07,
      "loss": 0.0575,
      "reward": 0.7662416994571686,
      "reward_std": 0.04386940971016884,
      "reward_std_all": 0.04277089540846646,
      "rewards-max/forensics_iou_reward": 0.982844352722168,
      "rewards-max/forensics_strict_edge_reward": 0.606867641210556,
      "rewards-max/forensics_tiou_reward": 0.9776348769664764,
      "rewards/forensics_iou_reward": 0.7692940831184387,
      "rewards/forensics_strict_edge_reward": 0.11865836381912231,
      "rewards/forensics_tiou_reward": 0.7316138744354248,
      "step": 1050
    },
    {
      "completion_length": 24.71875,
      "epoch": 3.2242703533026114,
      "grad_norm": 6.64885950088501,
      "kl": 0.4228515625,
      "learning_rate": 1.9230769230769231e-07,
      "loss": 0.015,
      "reward": 0.7308660745620728,
      "reward_std": 0.05245485343039036,
      "reward_std_all": 0.0700029581785202,
      "rewards-max/forensics_iou_reward": 0.9777873158454895,
      "rewards-max/forensics_strict_edge_reward": 0.5391099154949188,
      "rewards-max/forensics_tiou_reward": 0.9702726006507874,
      "rewards/forensics_iou_reward": 0.7115188539028168,
      "rewards/forensics_strict_edge_reward": 0.061882203444838524,
      "rewards/forensics_tiou_reward": 0.6731443703174591,
      "step": 1051
    },
    {
      "completion_length": 26.46875,
      "epoch": 3.227342549923195,
      "grad_norm": 6.175872325897217,
      "kl": 0.4384765625,
      "learning_rate": 1.9153846153846153e-07,
      "loss": -0.0511,
      "reward": 0.7571266293525696,
      "reward_std": 0.05247349105775356,
      "reward_std_all": 0.051962440833449364,
      "rewards-max/forensics_iou_reward": 0.983693927526474,
      "rewards-max/forensics_strict_edge_reward": 0.579438790678978,
      "rewards-max/forensics_tiou_reward": 0.9776958525180817,
      "rewards/forensics_iou_reward": 0.7159081995487213,
      "rewards/forensics_strict_edge_reward": 0.07499362155795097,
      "rewards/forensics_tiou_reward": 0.6795764565467834,
      "step": 1052
    },
    {
      "completion_length": 28.15625,
      "epoch": 3.230414746543779,
      "grad_norm": 6.533298492431641,
      "kl": 0.4189453125,
      "learning_rate": 1.9076923076923077e-07,
      "loss": -0.0816,
      "reward": 0.7756963670253754,
      "reward_std": 0.027857103385031223,
      "reward_std_all": 0.029508943669497967,
      "rewards-max/forensics_iou_reward": 0.992026686668396,
      "rewards-max/forensics_strict_edge_reward": 0.6112430393695831,
      "rewards-max/forensics_tiou_reward": 0.9902689158916473,
      "rewards/forensics_iou_reward": 0.7554246187210083,
      "rewards/forensics_strict_edge_reward": 0.09635201096534729,
      "rewards/forensics_tiou_reward": 0.7119395136833191,
      "step": 1053
    },
    {
      "completion_length": 24.875,
      "epoch": 3.2334869431643627,
      "grad_norm": 7.466168403625488,
      "kl": 0.59375,
      "learning_rate": 1.8999999999999998e-07,
      "loss": 0.0355,
      "reward": 0.7929236888885498,
      "reward_std": 0.020912674721330404,
      "reward_std_all": 0.02037665667012334,
      "rewards-max/forensics_iou_reward": 0.9723984897136688,
      "rewards-max/forensics_strict_edge_reward": 0.47550734877586365,
      "rewards-max/forensics_tiou_reward": 0.9636439085006714,
      "rewards/forensics_iou_reward": 0.7647059559822083,
      "rewards/forensics_strict_edge_reward": 0.06459818035364151,
      "rewards/forensics_tiou_reward": 0.722440093755722,
      "step": 1054
    },
    {
      "completion_length": 25.71875,
      "epoch": 3.236559139784946,
      "grad_norm": 8.041510581970215,
      "kl": 0.5068359375,
      "learning_rate": 1.8923076923076922e-07,
      "loss": 0.0026,
      "reward": 0.7628780901432037,
      "reward_std": 0.03717799484729767,
      "reward_std_all": 0.0394886489957571,
      "rewards-max/forensics_iou_reward": 0.9626972675323486,
      "rewards-max/forensics_strict_edge_reward": 0.6236754059791565,
      "rewards-max/forensics_tiou_reward": 0.9567852020263672,
      "rewards/forensics_iou_reward": 0.6713805198669434,
      "rewards/forensics_strict_edge_reward": 0.05512814782559872,
      "rewards/forensics_tiou_reward": 0.6248118281364441,
      "step": 1055
    },
    {
      "completion_length": 26.359375,
      "epoch": 3.23963133640553,
      "grad_norm": 6.838957786560059,
      "kl": 0.46875,
      "learning_rate": 1.8846153846153846e-07,
      "loss": 0.0712,
      "reward": 0.700291633605957,
      "reward_std": 0.08064216747879982,
      "reward_std_all": 0.09743802808225155,
      "rewards-max/forensics_iou_reward": 0.9715568423271179,
      "rewards-max/forensics_strict_edge_reward": 0.5711847841739655,
      "rewards-max/forensics_tiou_reward": 0.9683049917221069,
      "rewards/forensics_iou_reward": 0.6115922927856445,
      "rewards/forensics_strict_edge_reward": 0.07507575862109661,
      "rewards/forensics_tiou_reward": 0.5783737897872925,
      "step": 1056
    },
    {
      "completion_length": 28.015625,
      "epoch": 3.2427035330261136,
      "grad_norm": 5.552458763122559,
      "kl": 0.505859375,
      "learning_rate": 1.8769230769230767e-07,
      "loss": 0.0129,
      "reward": 0.7972827851772308,
      "reward_std": 0.012984842527657747,
      "reward_std_all": 0.013383457437157631,
      "rewards-max/forensics_iou_reward": 0.991962194442749,
      "rewards-max/forensics_strict_edge_reward": 0.6420580893754959,
      "rewards-max/forensics_tiou_reward": 0.9876406490802765,
      "rewards/forensics_iou_reward": 0.7105191051959991,
      "rewards/forensics_strict_edge_reward": 0.10646840929985046,
      "rewards/forensics_tiou_reward": 0.6818161308765411,
      "step": 1057
    },
    {
      "completion_length": 25.234375,
      "epoch": 3.2457757296466974,
      "grad_norm": 7.506042957305908,
      "kl": 0.400390625,
      "learning_rate": 1.8692307692307692e-07,
      "loss": 0.0283,
      "reward": 0.7361609637737274,
      "reward_std": 0.05748863331973553,
      "reward_std_all": 0.06158639770001173,
      "rewards-max/forensics_iou_reward": 0.9673402607440948,
      "rewards-max/forensics_strict_edge_reward": 0.5076928734779358,
      "rewards-max/forensics_tiou_reward": 0.9599605202674866,
      "rewards/forensics_iou_reward": 0.6647777557373047,
      "rewards/forensics_strict_edge_reward": 0.07335198111832142,
      "rewards/forensics_tiou_reward": 0.6181727647781372,
      "step": 1058
    },
    {
      "completion_length": 26.90625,
      "epoch": 3.248847926267281,
      "grad_norm": 6.5224175453186035,
      "kl": 0.40234375,
      "learning_rate": 1.8615384615384613e-07,
      "loss": 0.0092,
      "reward": 0.7850067317485809,
      "reward_std": 0.024931437277700752,
      "reward_std_all": 0.02504317951388657,
      "rewards-max/forensics_iou_reward": 0.972608894109726,
      "rewards-max/forensics_strict_edge_reward": 0.6468778848648071,
      "rewards-max/forensics_tiou_reward": 0.9637349247932434,
      "rewards/forensics_iou_reward": 0.6926351487636566,
      "rewards/forensics_strict_edge_reward": 0.059790292754769325,
      "rewards/forensics_tiou_reward": 0.6417053639888763,
      "step": 1059
    },
    {
      "completion_length": 25.453125,
      "epoch": 3.251920122887865,
      "grad_norm": 10.889005661010742,
      "kl": 0.4140625,
      "learning_rate": 1.853846153846154e-07,
      "loss": 0.07,
      "reward": 0.7493980824947357,
      "reward_std": 0.06591774523258209,
      "reward_std_all": 0.06322929821908474,
      "rewards-max/forensics_iou_reward": 0.958084374666214,
      "rewards-max/forensics_strict_edge_reward": 0.4251216650009155,
      "rewards-max/forensics_tiou_reward": 0.9515638947486877,
      "rewards/forensics_iou_reward": 0.6473794281482697,
      "rewards/forensics_strict_edge_reward": 0.047101229429244995,
      "rewards/forensics_tiou_reward": 0.6015772223472595,
      "step": 1060
    },
    {
      "completion_length": 25.171875,
      "epoch": 3.2549923195084487,
      "grad_norm": 7.885561943054199,
      "kl": 0.46484375,
      "learning_rate": 1.846153846153846e-07,
      "loss": 0.1081,
      "reward": 0.7160089910030365,
      "reward_std": 0.07554802671074867,
      "reward_std_all": 0.08549358695745468,
      "rewards-max/forensics_iou_reward": 0.9876825511455536,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9858591258525848,
      "rewards/forensics_iou_reward": 0.6838690340518951,
      "rewards/forensics_strict_edge_reward": 0.13684392347931862,
      "rewards/forensics_tiou_reward": 0.6583611071109772,
      "step": 1061
    },
    {
      "completion_length": 23.4375,
      "epoch": 3.258064516129032,
      "grad_norm": 8.66284465789795,
      "kl": 0.4130859375,
      "learning_rate": 1.8384615384615385e-07,
      "loss": 0.0044,
      "reward": 0.7160170078277588,
      "reward_std": 0.09306665882468224,
      "reward_std_all": 0.09368164092302322,
      "rewards-max/forensics_iou_reward": 0.9851068556308746,
      "rewards-max/forensics_strict_edge_reward": 0.5816373080015182,
      "rewards-max/forensics_tiou_reward": 0.9809419810771942,
      "rewards/forensics_iou_reward": 0.6292309165000916,
      "rewards/forensics_strict_edge_reward": 0.07675454393029213,
      "rewards/forensics_tiou_reward": 0.5848373770713806,
      "step": 1062
    },
    {
      "completion_length": 22.71875,
      "epoch": 3.261136712749616,
      "grad_norm": 6.993076801300049,
      "kl": 0.4833984375,
      "learning_rate": 1.8307692307692306e-07,
      "loss": -0.0024,
      "reward": 0.6989222168922424,
      "reward_std": 0.07111570797860622,
      "reward_std_all": 0.08856989070773125,
      "rewards-max/forensics_iou_reward": 0.9790030717849731,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9743696749210358,
      "rewards/forensics_iou_reward": 0.6808501183986664,
      "rewards/forensics_strict_edge_reward": 0.08351433835923672,
      "rewards/forensics_tiou_reward": 0.6378214955329895,
      "step": 1063
    },
    {
      "completion_length": 28.09375,
      "epoch": 3.2642089093701996,
      "grad_norm": 6.480101585388184,
      "kl": 0.48046875,
      "learning_rate": 1.823076923076923e-07,
      "loss": 0.0254,
      "reward": 0.7572597861289978,
      "reward_std": 0.06860145996324718,
      "reward_std_all": 0.06545137939974666,
      "rewards-max/forensics_iou_reward": 0.9673108160495758,
      "rewards-max/forensics_strict_edge_reward": 0.3744131475687027,
      "rewards-max/forensics_tiou_reward": 0.9563605189323425,
      "rewards/forensics_iou_reward": 0.723648190498352,
      "rewards/forensics_strict_edge_reward": 0.05462896078824997,
      "rewards/forensics_tiou_reward": 0.6790939271450043,
      "step": 1064
    },
    {
      "completion_length": 27.75,
      "epoch": 3.2672811059907834,
      "grad_norm": 8.274223327636719,
      "kl": 0.4072265625,
      "learning_rate": 1.8153846153846152e-07,
      "loss": 0.0514,
      "reward": 0.7734546959400177,
      "reward_std": 0.05139384791254997,
      "reward_std_all": 0.04721183329820633,
      "rewards-max/forensics_iou_reward": 0.9839463829994202,
      "rewards-max/forensics_strict_edge_reward": 0.5805842876434326,
      "rewards-max/forensics_tiou_reward": 0.9744916260242462,
      "rewards/forensics_iou_reward": 0.703456461429596,
      "rewards/forensics_strict_edge_reward": 0.11915182322263718,
      "rewards/forensics_tiou_reward": 0.6622760593891144,
      "step": 1065
    },
    {
      "completion_length": 23.921875,
      "epoch": 3.270353302611367,
      "grad_norm": 8.140983581542969,
      "kl": 0.43359375,
      "learning_rate": 1.8076923076923076e-07,
      "loss": -0.0418,
      "reward": 0.7446096837520599,
      "reward_std": 0.02907710801810026,
      "reward_std_all": 0.049464608542621136,
      "rewards-max/forensics_iou_reward": 0.9884068071842194,
      "rewards-max/forensics_strict_edge_reward": 0.5152584314346313,
      "rewards-max/forensics_tiou_reward": 0.984317272901535,
      "rewards/forensics_iou_reward": 0.688468724489212,
      "rewards/forensics_strict_edge_reward": 0.06004638783633709,
      "rewards/forensics_tiou_reward": 0.6428828239440918,
      "step": 1066
    },
    {
      "completion_length": 28.953125,
      "epoch": 3.273425499231951,
      "grad_norm": 7.050485610961914,
      "kl": 0.4482421875,
      "learning_rate": 1.8e-07,
      "loss": 0.0324,
      "reward": 0.7854641377925873,
      "reward_std": 0.025595704093575478,
      "reward_std_all": 0.02545812726020813,
      "rewards-max/forensics_iou_reward": 0.993786096572876,
      "rewards-max/forensics_strict_edge_reward": 0.7958523333072662,
      "rewards-max/forensics_tiou_reward": 0.9910898208618164,
      "rewards/forensics_iou_reward": 0.674685537815094,
      "rewards/forensics_strict_edge_reward": 0.07174002751708031,
      "rewards/forensics_tiou_reward": 0.6320539712905884,
      "step": 1067
    },
    {
      "completion_length": 26.0,
      "epoch": 3.2764976958525347,
      "grad_norm": 6.795102119445801,
      "kl": 0.50390625,
      "learning_rate": 1.792307692307692e-07,
      "loss": 0.0127,
      "reward": 0.7689521610736847,
      "reward_std": 0.050650669261813164,
      "reward_std_all": 0.0476206187158823,
      "rewards-max/forensics_iou_reward": 0.9923709034919739,
      "rewards-max/forensics_strict_edge_reward": 0.824172168970108,
      "rewards-max/forensics_tiou_reward": 0.9903604090213776,
      "rewards/forensics_iou_reward": 0.6923601627349854,
      "rewards/forensics_strict_edge_reward": 0.12118304148316383,
      "rewards/forensics_tiou_reward": 0.6501531898975372,
      "step": 1068
    },
    {
      "completion_length": 28.78125,
      "epoch": 3.279569892473118,
      "grad_norm": 8.467233657836914,
      "kl": 0.560546875,
      "learning_rate": 1.7846153846153845e-07,
      "loss": 0.0281,
      "reward": 0.6982365250587463,
      "reward_std": 0.06405562534928322,
      "reward_std_all": 0.0856895912438631,
      "rewards-max/forensics_iou_reward": 0.9795202910900116,
      "rewards-max/forensics_strict_edge_reward": 0.37861546874046326,
      "rewards-max/forensics_tiou_reward": 0.9709680676460266,
      "rewards/forensics_iou_reward": 0.6379250884056091,
      "rewards/forensics_strict_edge_reward": 0.06538334488868713,
      "rewards/forensics_tiou_reward": 0.6013721227645874,
      "step": 1069
    },
    {
      "completion_length": 28.828125,
      "epoch": 3.282642089093702,
      "grad_norm": 7.9794769287109375,
      "kl": 0.630859375,
      "learning_rate": 1.776923076923077e-07,
      "loss": 0.1194,
      "reward": 0.7448290288448334,
      "reward_std": 0.06197015568614006,
      "reward_std_all": 0.06349692493677139,
      "rewards-max/forensics_iou_reward": 0.9920773506164551,
      "rewards-max/forensics_strict_edge_reward": 0.5014478266239166,
      "rewards-max/forensics_tiou_reward": 0.9900397956371307,
      "rewards/forensics_iou_reward": 0.6744428277015686,
      "rewards/forensics_strict_edge_reward": 0.08463536016643047,
      "rewards/forensics_tiou_reward": 0.6326887905597687,
      "step": 1070
    },
    {
      "completion_length": 29.578125,
      "epoch": 3.2857142857142856,
      "grad_norm": 9.50998306274414,
      "kl": 0.7109375,
      "learning_rate": 1.7692307692307693e-07,
      "loss": 0.1048,
      "reward": 0.7784841656684875,
      "reward_std": 0.0463956231251359,
      "reward_std_all": 0.04213985800743103,
      "rewards-max/forensics_iou_reward": 0.9710927605628967,
      "rewards-max/forensics_strict_edge_reward": 0.49672752618789673,
      "rewards-max/forensics_tiou_reward": 0.961968868970871,
      "rewards/forensics_iou_reward": 0.6801996231079102,
      "rewards/forensics_strict_edge_reward": 0.05433645099401474,
      "rewards/forensics_tiou_reward": 0.631569892168045,
      "step": 1071
    },
    {
      "completion_length": 24.953125,
      "epoch": 3.2887864823348694,
      "grad_norm": 9.922019004821777,
      "kl": 0.595703125,
      "learning_rate": 1.7615384615384615e-07,
      "loss": -0.0687,
      "reward": 0.7778218686580658,
      "reward_std": 0.04327624989673495,
      "reward_std_all": 0.04014733387157321,
      "rewards-max/forensics_iou_reward": 0.9917639791965485,
      "rewards-max/forensics_strict_edge_reward": 0.7727466821670532,
      "rewards-max/forensics_tiou_reward": 0.9886415004730225,
      "rewards/forensics_iou_reward": 0.680878221988678,
      "rewards/forensics_strict_edge_reward": 0.08076279051601887,
      "rewards/forensics_tiou_reward": 0.6356077790260315,
      "step": 1072
    },
    {
      "completion_length": 26.703125,
      "epoch": 3.291858678955453,
      "grad_norm": 8.025321006774902,
      "kl": 0.6005859375,
      "learning_rate": 1.7538461538461539e-07,
      "loss": -0.0107,
      "reward": 0.7909823656082153,
      "reward_std": 0.027063144138082862,
      "reward_std_all": 0.025485098245553672,
      "rewards-max/forensics_iou_reward": 0.9850991070270538,
      "rewards-max/forensics_strict_edge_reward": 0.5325971990823746,
      "rewards-max/forensics_tiou_reward": 0.9742029309272766,
      "rewards/forensics_iou_reward": 0.7845702767372131,
      "rewards/forensics_strict_edge_reward": 0.08983162045478821,
      "rewards/forensics_tiou_reward": 0.7512580156326294,
      "step": 1073
    },
    {
      "completion_length": 29.359375,
      "epoch": 3.294930875576037,
      "grad_norm": 7.291997909545898,
      "kl": 0.39453125,
      "learning_rate": 1.746153846153846e-07,
      "loss": -0.0061,
      "reward": 0.7866594195365906,
      "reward_std": 0.021706941537559032,
      "reward_std_all": 0.02255667746067047,
      "rewards-max/forensics_iou_reward": 0.9804339706897736,
      "rewards-max/forensics_strict_edge_reward": 0.3803492709994316,
      "rewards-max/forensics_tiou_reward": 0.9742229282855988,
      "rewards/forensics_iou_reward": 0.6901757121086121,
      "rewards/forensics_strict_edge_reward": 0.04558616690337658,
      "rewards/forensics_tiou_reward": 0.6402861475944519,
      "step": 1074
    },
    {
      "completion_length": 30.953125,
      "epoch": 3.2980030721966207,
      "grad_norm": 8.784178733825684,
      "kl": 0.5732421875,
      "learning_rate": 1.7384615384615384e-07,
      "loss": 0.0159,
      "reward": 0.6709096431732178,
      "reward_std": 0.06434333324432373,
      "reward_std_all": 0.10384934395551682,
      "rewards-max/forensics_iou_reward": 0.9987309575080872,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9984519779682159,
      "rewards/forensics_iou_reward": 0.5425575077533722,
      "rewards/forensics_strict_edge_reward": 0.09284545481204987,
      "rewards/forensics_tiou_reward": 0.5102788805961609,
      "step": 1075
    },
    {
      "completion_length": 23.03125,
      "epoch": 3.3010752688172045,
      "grad_norm": 9.97095775604248,
      "kl": 0.90234375,
      "learning_rate": 1.7307692307692305e-07,
      "loss": -0.0384,
      "reward": 0.7807434499263763,
      "reward_std": 0.025261192116886377,
      "reward_std_all": 0.026261855382472277,
      "rewards-max/forensics_iou_reward": 0.9791841208934784,
      "rewards-max/forensics_strict_edge_reward": 0.5939172208309174,
      "rewards-max/forensics_tiou_reward": 0.9720333516597748,
      "rewards/forensics_iou_reward": 0.7126226127147675,
      "rewards/forensics_strict_edge_reward": 0.10776074603199959,
      "rewards/forensics_tiou_reward": 0.6660502254962921,
      "step": 1076
    },
    {
      "completion_length": 24.296875,
      "epoch": 3.3041474654377883,
      "grad_norm": 6.686517238616943,
      "kl": 0.5078125,
      "learning_rate": 1.723076923076923e-07,
      "loss": -0.0154,
      "reward": 0.7026849091053009,
      "reward_std": 0.058397695422172546,
      "reward_std_all": 0.08262510225176811,
      "rewards-max/forensics_iou_reward": 0.9852255284786224,
      "rewards-max/forensics_strict_edge_reward": 0.6526533365249634,
      "rewards-max/forensics_tiou_reward": 0.9802602231502533,
      "rewards/forensics_iou_reward": 0.6597192287445068,
      "rewards/forensics_strict_edge_reward": 0.07585480436682701,
      "rewards/forensics_tiou_reward": 0.6211938858032227,
      "step": 1077
    },
    {
      "completion_length": 27.59375,
      "epoch": 3.3072196620583716,
      "grad_norm": 6.984361171722412,
      "kl": 0.4716796875,
      "learning_rate": 1.7153846153846153e-07,
      "loss": -0.0135,
      "reward": 0.722588986158371,
      "reward_std": 0.06231528893113136,
      "reward_std_all": 0.06854032538831234,
      "rewards-max/forensics_iou_reward": 0.9866388738155365,
      "rewards-max/forensics_strict_edge_reward": 0.639672726392746,
      "rewards-max/forensics_tiou_reward": 0.9819864630699158,
      "rewards/forensics_iou_reward": 0.6154719591140747,
      "rewards/forensics_strict_edge_reward": 0.06294012255966663,
      "rewards/forensics_tiou_reward": 0.5639501512050629,
      "step": 1078
    },
    {
      "completion_length": 31.34375,
      "epoch": 3.3102918586789554,
      "grad_norm": 6.703818321228027,
      "kl": 0.431640625,
      "learning_rate": 1.7076923076923075e-07,
      "loss": 0.0751,
      "reward": 0.7896661460399628,
      "reward_std": 0.028285576030611992,
      "reward_std_all": 0.026329943910241127,
      "rewards-max/forensics_iou_reward": 0.997572124004364,
      "rewards-max/forensics_strict_edge_reward": 0.7388437688350677,
      "rewards-max/forensics_tiou_reward": 0.9958896040916443,
      "rewards/forensics_iou_reward": 0.7012903392314911,
      "rewards/forensics_strict_edge_reward": 0.08312248811125755,
      "rewards/forensics_tiou_reward": 0.6623771786689758,
      "step": 1079
    },
    {
      "completion_length": 23.890625,
      "epoch": 3.313364055299539,
      "grad_norm": 7.139591693878174,
      "kl": 0.37109375,
      "learning_rate": 1.7000000000000001e-07,
      "loss": 0.1245,
      "reward": 0.6881309151649475,
      "reward_std": 0.03851966839283705,
      "reward_std_all": 0.08035657368600368,
      "rewards-max/forensics_iou_reward": 0.9980511367321014,
      "rewards-max/forensics_strict_edge_reward": 0.7055691182613373,
      "rewards-max/forensics_tiou_reward": 0.9969762861728668,
      "rewards/forensics_iou_reward": 0.5750068426132202,
      "rewards/forensics_strict_edge_reward": 0.06702491082251072,
      "rewards/forensics_tiou_reward": 0.5246913731098175,
      "step": 1080
    },
    {
      "completion_length": 31.5625,
      "epoch": 3.316436251920123,
      "grad_norm": 6.582668304443359,
      "kl": 0.533203125,
      "learning_rate": 1.6923076923076923e-07,
      "loss": 0.08,
      "reward": 0.6880720257759094,
      "reward_std": 0.10299630090594292,
      "reward_std_all": 0.11634768173098564,
      "rewards-max/forensics_iou_reward": 0.9843316674232483,
      "rewards-max/forensics_strict_edge_reward": 0.6832874715328217,
      "rewards-max/forensics_tiou_reward": 0.9795048236846924,
      "rewards/forensics_iou_reward": 0.6110751032829285,
      "rewards/forensics_strict_edge_reward": 0.08533358946442604,
      "rewards/forensics_tiou_reward": 0.571584552526474,
      "step": 1081
    },
    {
      "completion_length": 26.21875,
      "epoch": 3.3195084485407067,
      "grad_norm": 10.412839889526367,
      "kl": 0.701171875,
      "learning_rate": 1.6846153846153847e-07,
      "loss": 0.0084,
      "reward": 0.7478054761886597,
      "reward_std": 0.0207892507314682,
      "reward_std_all": 0.04276322387158871,
      "rewards-max/forensics_iou_reward": 0.9853213131427765,
      "rewards-max/forensics_strict_edge_reward": 0.4228300601243973,
      "rewards-max/forensics_tiou_reward": 0.9809961915016174,
      "rewards/forensics_iou_reward": 0.6542787849903107,
      "rewards/forensics_strict_edge_reward": 0.06568540446460247,
      "rewards/forensics_tiou_reward": 0.6127846837043762,
      "step": 1082
    },
    {
      "completion_length": 23.296875,
      "epoch": 3.3225806451612905,
      "grad_norm": 9.65318775177002,
      "kl": 0.775390625,
      "learning_rate": 1.6769230769230768e-07,
      "loss": 0.0735,
      "reward": 0.698220431804657,
      "reward_std": 0.07433444075286388,
      "reward_std_all": 0.09401332214474678,
      "rewards-max/forensics_iou_reward": 0.9704428315162659,
      "rewards-max/forensics_strict_edge_reward": 0.4934397041797638,
      "rewards-max/forensics_tiou_reward": 0.9639711976051331,
      "rewards/forensics_iou_reward": 0.6227498054504395,
      "rewards/forensics_strict_edge_reward": 0.08924033865332603,
      "rewards/forensics_tiou_reward": 0.5704892575740814,
      "step": 1083
    },
    {
      "completion_length": 29.859375,
      "epoch": 3.3256528417818743,
      "grad_norm": 6.165536403656006,
      "kl": 0.53515625,
      "learning_rate": 1.6692307692307692e-07,
      "loss": 0.0081,
      "reward": 0.7446637749671936,
      "reward_std": 0.05407932959496975,
      "reward_std_all": 0.05639292672276497,
      "rewards-max/forensics_iou_reward": 0.9673053026199341,
      "rewards-max/forensics_strict_edge_reward": 0.5702531039714813,
      "rewards-max/forensics_tiou_reward": 0.962407648563385,
      "rewards/forensics_iou_reward": 0.6439452171325684,
      "rewards/forensics_strict_edge_reward": 0.06930912658572197,
      "rewards/forensics_tiou_reward": 0.6094746887683868,
      "step": 1084
    },
    {
      "completion_length": 24.046875,
      "epoch": 3.3287250384024576,
      "grad_norm": 7.609842300415039,
      "kl": 0.4296875,
      "learning_rate": 1.6615384615384613e-07,
      "loss": -0.0971,
      "reward": 0.7790640592575073,
      "reward_std": 0.028662284836173058,
      "reward_std_all": 0.029471985064446926,
      "rewards-max/forensics_iou_reward": 0.9976768791675568,
      "rewards-max/forensics_strict_edge_reward": 0.8483634889125824,
      "rewards-max/forensics_tiou_reward": 0.9966205954551697,
      "rewards/forensics_iou_reward": 0.692878782749176,
      "rewards/forensics_strict_edge_reward": 0.10660970024764538,
      "rewards/forensics_tiou_reward": 0.6373497247695923,
      "step": 1085
    },
    {
      "completion_length": 32.796875,
      "epoch": 3.3317972350230414,
      "grad_norm": 5.461789608001709,
      "kl": 0.4814453125,
      "learning_rate": 1.6538461538461538e-07,
      "loss": 0.0028,
      "reward": 0.7525030970573425,
      "reward_std": 0.019639155303593725,
      "reward_std_all": 0.03960986970923841,
      "rewards-max/forensics_iou_reward": 0.9860257208347321,
      "rewards-max/forensics_strict_edge_reward": 0.5689917057752609,
      "rewards-max/forensics_tiou_reward": 0.9839287102222443,
      "rewards/forensics_iou_reward": 0.7449445128440857,
      "rewards/forensics_strict_edge_reward": 0.06650775484740734,
      "rewards/forensics_tiou_reward": 0.7108752727508545,
      "step": 1086
    },
    {
      "completion_length": 30.546875,
      "epoch": 3.334869431643625,
      "grad_norm": 6.131428241729736,
      "kl": 0.4814453125,
      "learning_rate": 1.646153846153846e-07,
      "loss": 0.1261,
      "reward": 0.7344861328601837,
      "reward_std": 0.05431508645415306,
      "reward_std_all": 0.06014478579163551,
      "rewards-max/forensics_iou_reward": 0.9739693403244019,
      "rewards-max/forensics_strict_edge_reward": 0.6538934260606766,
      "rewards-max/forensics_tiou_reward": 0.9669073522090912,
      "rewards/forensics_iou_reward": 0.5997445732355118,
      "rewards/forensics_strict_edge_reward": 0.04884199798107147,
      "rewards/forensics_tiou_reward": 0.5529559552669525,
      "step": 1087
    },
    {
      "completion_length": 26.1875,
      "epoch": 3.337941628264209,
      "grad_norm": 6.417045593261719,
      "kl": 0.3955078125,
      "learning_rate": 1.6384615384615383e-07,
      "loss": -0.0436,
      "reward": 0.7805551886558533,
      "reward_std": 0.03326637018471956,
      "reward_std_all": 0.032095192931592464,
      "rewards-max/forensics_iou_reward": 0.984319269657135,
      "rewards-max/forensics_strict_edge_reward": 0.6900639235973358,
      "rewards-max/forensics_tiou_reward": 0.9815137982368469,
      "rewards/forensics_iou_reward": 0.705759197473526,
      "rewards/forensics_strict_edge_reward": 0.1365128755569458,
      "rewards/forensics_tiou_reward": 0.6684047877788544,
      "step": 1088
    },
    {
      "completion_length": 24.4375,
      "epoch": 3.3410138248847927,
      "grad_norm": 6.646300792694092,
      "kl": 0.4150390625,
      "learning_rate": 1.6307692307692307e-07,
      "loss": -0.0376,
      "reward": 0.794459730386734,
      "reward_std": 0.02208735141903162,
      "reward_std_all": 0.020616136956959963,
      "rewards-max/forensics_iou_reward": 0.9943638145923615,
      "rewards-max/forensics_strict_edge_reward": 0.7744058072566986,
      "rewards-max/forensics_tiou_reward": 0.9925158023834229,
      "rewards/forensics_iou_reward": 0.7949036955833435,
      "rewards/forensics_strict_edge_reward": 0.11460821703076363,
      "rewards/forensics_tiou_reward": 0.7517006397247314,
      "step": 1089
    },
    {
      "completion_length": 28.765625,
      "epoch": 3.3440860215053765,
      "grad_norm": 6.331485748291016,
      "kl": 0.4443359375,
      "learning_rate": 1.623076923076923e-07,
      "loss": 0.0597,
      "reward": 0.7951314151287079,
      "reward_std": 0.01977091422304511,
      "reward_std_all": 0.01896520797163248,
      "rewards-max/forensics_iou_reward": 0.9769188761711121,
      "rewards-max/forensics_strict_edge_reward": 0.4283051937818527,
      "rewards-max/forensics_tiou_reward": 0.9690129160881042,
      "rewards/forensics_iou_reward": 0.7781157493591309,
      "rewards/forensics_strict_edge_reward": 0.051612554118037224,
      "rewards/forensics_tiou_reward": 0.7389595806598663,
      "step": 1090
    },
    {
      "completion_length": 25.84375,
      "epoch": 3.3471582181259603,
      "grad_norm": 7.430399417877197,
      "kl": 0.4384765625,
      "learning_rate": 1.6153846153846155e-07,
      "loss": 0.1082,
      "reward": 0.767246425151825,
      "reward_std": 0.05028242524713278,
      "reward_std_all": 0.047760944813489914,
      "rewards-max/forensics_iou_reward": 0.9929057955741882,
      "rewards-max/forensics_strict_edge_reward": 0.6037173122167587,
      "rewards-max/forensics_tiou_reward": 0.9884902238845825,
      "rewards/forensics_iou_reward": 0.6988672614097595,
      "rewards/forensics_strict_edge_reward": 0.08890825882554054,
      "rewards/forensics_tiou_reward": 0.6516116857528687,
      "step": 1091
    },
    {
      "completion_length": 22.515625,
      "epoch": 3.3502304147465436,
      "grad_norm": 8.465216636657715,
      "kl": 0.41015625,
      "learning_rate": 1.6076923076923076e-07,
      "loss": -0.054,
      "reward": 0.776777058839798,
      "reward_std": 0.051983846351504326,
      "reward_std_all": 0.046743905171751976,
      "rewards-max/forensics_iou_reward": 0.9828855693340302,
      "rewards-max/forensics_strict_edge_reward": 0.8032653331756592,
      "rewards-max/forensics_tiou_reward": 0.9777909815311432,
      "rewards/forensics_iou_reward": 0.7608469426631927,
      "rewards/forensics_strict_edge_reward": 0.144864059984684,
      "rewards/forensics_tiou_reward": 0.7325153946876526,
      "step": 1092
    },
    {
      "completion_length": 28.28125,
      "epoch": 3.3533026113671274,
      "grad_norm": 8.70473575592041,
      "kl": 0.490234375,
      "learning_rate": 1.6e-07,
      "loss": -0.0029,
      "reward": 0.7874051928520203,
      "reward_std": 0.014900804497301579,
      "reward_std_all": 0.01773271942511201,
      "rewards-max/forensics_iou_reward": 0.9902157187461853,
      "rewards-max/forensics_strict_edge_reward": 0.8205508291721344,
      "rewards-max/forensics_tiou_reward": 0.9876305162906647,
      "rewards/forensics_iou_reward": 0.6999723017215729,
      "rewards/forensics_strict_edge_reward": 0.10749869048595428,
      "rewards/forensics_tiou_reward": 0.6544391512870789,
      "step": 1093
    },
    {
      "completion_length": 27.984375,
      "epoch": 3.356374807987711,
      "grad_norm": 7.301880836486816,
      "kl": 0.451171875,
      "learning_rate": 1.5923076923076922e-07,
      "loss": 0.0316,
      "reward": 0.7605892717838287,
      "reward_std": 0.06499271094799042,
      "reward_std_all": 0.05988718755543232,
      "rewards-max/forensics_iou_reward": 0.977055162191391,
      "rewards-max/forensics_strict_edge_reward": 0.5197639167308807,
      "rewards-max/forensics_tiou_reward": 0.9701134860515594,
      "rewards/forensics_iou_reward": 0.6519083082675934,
      "rewards/forensics_strict_edge_reward": 0.05299276113510132,
      "rewards/forensics_tiou_reward": 0.6059150099754333,
      "step": 1094
    },
    {
      "completion_length": 27.28125,
      "epoch": 3.359447004608295,
      "grad_norm": 6.780945777893066,
      "kl": 0.462890625,
      "learning_rate": 1.5846153846153846e-07,
      "loss": 0.0318,
      "reward": 0.7282673120498657,
      "reward_std": 0.036453296430408955,
      "reward_std_all": 0.05351732298731804,
      "rewards-max/forensics_iou_reward": 0.9908323287963867,
      "rewards-max/forensics_strict_edge_reward": 0.3544929772615433,
      "rewards-max/forensics_tiou_reward": 0.9860191941261292,
      "rewards/forensics_iou_reward": 0.6278634965419769,
      "rewards/forensics_strict_edge_reward": 0.03268325701355934,
      "rewards/forensics_tiou_reward": 0.5825276076793671,
      "step": 1095
    },
    {
      "completion_length": 30.90625,
      "epoch": 3.3625192012288787,
      "grad_norm": 7.631364822387695,
      "kl": 0.392578125,
      "learning_rate": 1.5769230769230767e-07,
      "loss": 0.0846,
      "reward": 0.7534418106079102,
      "reward_std": 0.05533131584525108,
      "reward_std_all": 0.05529942736029625,
      "rewards-max/forensics_iou_reward": 0.980601578950882,
      "rewards-max/forensics_strict_edge_reward": 0.5309537649154663,
      "rewards-max/forensics_tiou_reward": 0.9750866889953613,
      "rewards/forensics_iou_reward": 0.6548973917961121,
      "rewards/forensics_strict_edge_reward": 0.08144411444664001,
      "rewards/forensics_tiou_reward": 0.6180128753185272,
      "step": 1096
    },
    {
      "completion_length": 22.4375,
      "epoch": 3.3655913978494625,
      "grad_norm": 8.191710472106934,
      "kl": 0.5,
      "learning_rate": 1.569230769230769e-07,
      "loss": -0.0009,
      "reward": 0.7598457932472229,
      "reward_std": 0.05423922371119261,
      "reward_std_all": 0.05232718866318464,
      "rewards-max/forensics_iou_reward": 0.9879296123981476,
      "rewards-max/forensics_strict_edge_reward": 0.707136332988739,
      "rewards-max/forensics_tiou_reward": 0.9836361408233643,
      "rewards/forensics_iou_reward": 0.7112272679805756,
      "rewards/forensics_strict_edge_reward": 0.09656264632940292,
      "rewards/forensics_tiou_reward": 0.6686388552188873,
      "step": 1097
    },
    {
      "completion_length": 28.890625,
      "epoch": 3.3686635944700463,
      "grad_norm": 6.925998687744141,
      "kl": 0.47265625,
      "learning_rate": 1.5615384615384612e-07,
      "loss": -0.0205,
      "reward": 0.7680497765541077,
      "reward_std": 0.06626790575683117,
      "reward_std_all": 0.05897415243089199,
      "rewards-max/forensics_iou_reward": 0.9905396401882172,
      "rewards-max/forensics_strict_edge_reward": 0.4686945676803589,
      "rewards-max/forensics_tiou_reward": 0.9877768456935883,
      "rewards/forensics_iou_reward": 0.7346534430980682,
      "rewards/forensics_strict_edge_reward": 0.0656915120780468,
      "rewards/forensics_tiou_reward": 0.7101370394229889,
      "step": 1098
    },
    {
      "completion_length": 23.03125,
      "epoch": 3.3717357910906296,
      "grad_norm": 7.257787704467773,
      "kl": 0.388671875,
      "learning_rate": 1.5538461538461536e-07,
      "loss": 0.0225,
      "reward": 0.7884264886379242,
      "reward_std": 0.018282102420926094,
      "reward_std_all": 0.0198843227699399,
      "rewards-max/forensics_iou_reward": 0.9737476408481598,
      "rewards-max/forensics_strict_edge_reward": 0.6000697016716003,
      "rewards-max/forensics_tiou_reward": 0.963005781173706,
      "rewards/forensics_iou_reward": 0.7102325856685638,
      "rewards/forensics_strict_edge_reward": 0.07386320736259222,
      "rewards/forensics_tiou_reward": 0.6590862274169922,
      "step": 1099
    },
    {
      "completion_length": 25.71875,
      "epoch": 3.3748079877112134,
      "grad_norm": 7.254058837890625,
      "kl": 0.3916015625,
      "learning_rate": 1.5461538461538463e-07,
      "loss": 0.0358,
      "reward": 0.6842878758907318,
      "reward_std": 0.10928772017359734,
      "reward_std_all": 0.12278278544545174,
      "rewards-max/forensics_iou_reward": 0.9966104328632355,
      "rewards-max/forensics_strict_edge_reward": 0.7311047911643982,
      "rewards-max/forensics_tiou_reward": 0.9950420260429382,
      "rewards/forensics_iou_reward": 0.6026614308357239,
      "rewards/forensics_strict_edge_reward": 0.10062294080853462,
      "rewards/forensics_tiou_reward": 0.5680877268314362,
      "step": 1100
    },
    {
      "completion_length": 23.515625,
      "epoch": 3.377880184331797,
      "grad_norm": 7.3461103439331055,
      "kl": 0.4638671875,
      "learning_rate": 1.5384615384615385e-07,
      "loss": 0.114,
      "reward": 0.7445500195026398,
      "reward_std": 0.0692579634487629,
      "reward_std_all": 0.0668349601328373,
      "rewards-max/forensics_iou_reward": 0.9911997318267822,
      "rewards-max/forensics_strict_edge_reward": 0.41768769919872284,
      "rewards-max/forensics_tiou_reward": 0.9843576848506927,
      "rewards/forensics_iou_reward": 0.6496102213859558,
      "rewards/forensics_strict_edge_reward": 0.04616732709109783,
      "rewards/forensics_tiou_reward": 0.5987949669361115,
      "step": 1101
    },
    {
      "completion_length": 23.703125,
      "epoch": 3.380952380952381,
      "grad_norm": 7.732852935791016,
      "kl": 0.3720703125,
      "learning_rate": 1.5307692307692309e-07,
      "loss": 0.1004,
      "reward": 0.7314266860485077,
      "reward_std": 0.0935765840113163,
      "reward_std_all": 0.08730396255850792,
      "rewards-max/forensics_iou_reward": 0.972972959280014,
      "rewards-max/forensics_strict_edge_reward": 0.7351278513669968,
      "rewards-max/forensics_tiou_reward": 0.9633916020393372,
      "rewards/forensics_iou_reward": 0.604025661945343,
      "rewards/forensics_strict_edge_reward": 0.08649182692170143,
      "rewards/forensics_tiou_reward": 0.5408596247434616,
      "step": 1102
    },
    {
      "completion_length": 27.0,
      "epoch": 3.3840245775729647,
      "grad_norm": 8.39577579498291,
      "kl": 0.376953125,
      "learning_rate": 1.523076923076923e-07,
      "loss": -0.0814,
      "reward": 0.7819955945014954,
      "reward_std": 0.010755672032246366,
      "reward_std_all": 0.017375926254317164,
      "rewards-max/forensics_iou_reward": 0.9793495833873749,
      "rewards-max/forensics_strict_edge_reward": 0.5879279673099518,
      "rewards-max/forensics_tiou_reward": 0.9716211259365082,
      "rewards/forensics_iou_reward": 0.7849910259246826,
      "rewards/forensics_strict_edge_reward": 0.10704412683844566,
      "rewards/forensics_tiou_reward": 0.7487460076808929,
      "step": 1103
    },
    {
      "completion_length": 28.546875,
      "epoch": 3.3870967741935485,
      "grad_norm": 7.2399210929870605,
      "kl": 0.5068359375,
      "learning_rate": 1.5153846153846154e-07,
      "loss": 0.0157,
      "reward": 0.7490850389003754,
      "reward_std": 0.08787078410387039,
      "reward_std_all": 0.07918502017855644,
      "rewards-max/forensics_iou_reward": 0.9907073974609375,
      "rewards-max/forensics_strict_edge_reward": 0.5592742264270782,
      "rewards-max/forensics_tiou_reward": 0.9877656996250153,
      "rewards/forensics_iou_reward": 0.6879648864269257,
      "rewards/forensics_strict_edge_reward": 0.08865310624241829,
      "rewards/forensics_tiou_reward": 0.644681841135025,
      "step": 1104
    },
    {
      "completion_length": 28.40625,
      "epoch": 3.3901689708141323,
      "grad_norm": 7.5307416915893555,
      "kl": 0.4599609375,
      "learning_rate": 1.5076923076923075e-07,
      "loss": 0.1391,
      "reward": 0.7392435371875763,
      "reward_std": 0.07354377582669258,
      "reward_std_all": 0.07128679566085339,
      "rewards-max/forensics_iou_reward": 0.9780433475971222,
      "rewards-max/forensics_strict_edge_reward": 0.4190879166126251,
      "rewards-max/forensics_tiou_reward": 0.9696860015392303,
      "rewards/forensics_iou_reward": 0.6056898832321167,
      "rewards/forensics_strict_edge_reward": 0.05372561141848564,
      "rewards/forensics_tiou_reward": 0.5566785335540771,
      "step": 1105
    },
    {
      "completion_length": 27.125,
      "epoch": 3.3932411674347156,
      "grad_norm": 7.229068279266357,
      "kl": 0.50390625,
      "learning_rate": 1.5e-07,
      "loss": -0.0829,
      "reward": 0.7578931152820587,
      "reward_std": 0.037582749500870705,
      "reward_std_all": 0.0413898266851902,
      "rewards-max/forensics_iou_reward": 0.9798786342144012,
      "rewards-max/forensics_strict_edge_reward": 0.5740107297897339,
      "rewards-max/forensics_tiou_reward": 0.9745809435844421,
      "rewards/forensics_iou_reward": 0.7196042537689209,
      "rewards/forensics_strict_edge_reward": 0.08425082825124264,
      "rewards/forensics_tiou_reward": 0.6796986162662506,
      "step": 1106
    },
    {
      "completion_length": 21.421875,
      "epoch": 3.3963133640552994,
      "grad_norm": 8.2166166305542,
      "kl": 0.388671875,
      "learning_rate": 1.492307692307692e-07,
      "loss": -0.0338,
      "reward": 0.7537747323513031,
      "reward_std": 0.013025313965044916,
      "reward_std_all": 0.03588818293064833,
      "rewards-max/forensics_iou_reward": 0.9911634624004364,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.988217830657959,
      "rewards/forensics_iou_reward": 0.721148818731308,
      "rewards/forensics_strict_edge_reward": 0.06478360295295715,
      "rewards/forensics_tiou_reward": 0.6727853119373322,
      "step": 1107
    },
    {
      "completion_length": 25.296875,
      "epoch": 3.399385560675883,
      "grad_norm": 7.123376369476318,
      "kl": 0.4482421875,
      "learning_rate": 1.4846153846153845e-07,
      "loss": 0.0999,
      "reward": 0.7734609544277191,
      "reward_std": 0.04799369536340237,
      "reward_std_all": 0.04453946463763714,
      "rewards-max/forensics_iou_reward": 0.9945550262928009,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.9932770431041718,
      "rewards/forensics_iou_reward": 0.7244323492050171,
      "rewards/forensics_strict_edge_reward": 0.09914151951670647,
      "rewards/forensics_tiou_reward": 0.6818541586399078,
      "step": 1108
    },
    {
      "completion_length": 29.703125,
      "epoch": 3.402457757296467,
      "grad_norm": 8.904394149780273,
      "kl": 0.5546875,
      "learning_rate": 1.476923076923077e-07,
      "loss": 0.1054,
      "reward": 0.7618148922920227,
      "reward_std": 0.07728124223649502,
      "reward_std_all": 0.06857937946915627,
      "rewards-max/forensics_iou_reward": 0.9618291258811951,
      "rewards-max/forensics_strict_edge_reward": 0.3184485696256161,
      "rewards-max/forensics_tiou_reward": 0.9478241503238678,
      "rewards/forensics_iou_reward": 0.7338723242282867,
      "rewards/forensics_strict_edge_reward": 0.06418597465381026,
      "rewards/forensics_tiou_reward": 0.695182591676712,
      "step": 1109
    },
    {
      "completion_length": 22.125,
      "epoch": 3.4055299539170507,
      "grad_norm": 8.319511413574219,
      "kl": 0.4228515625,
      "learning_rate": 1.4692307692307693e-07,
      "loss": 0.0936,
      "reward": 0.7458566129207611,
      "reward_std": 0.07483906485140324,
      "reward_std_all": 0.07081066071987152,
      "rewards-max/forensics_iou_reward": 0.9814021289348602,
      "rewards-max/forensics_strict_edge_reward": 0.5877082049846649,
      "rewards-max/forensics_tiou_reward": 0.9730795323848724,
      "rewards/forensics_iou_reward": 0.5931103229522705,
      "rewards/forensics_strict_edge_reward": 0.06202586181461811,
      "rewards/forensics_tiou_reward": 0.5395562052726746,
      "step": 1110
    },
    {
      "completion_length": 23.28125,
      "epoch": 3.4086021505376345,
      "grad_norm": 10.078051567077637,
      "kl": 0.40625,
      "learning_rate": 1.4615384615384617e-07,
      "loss": 0.1431,
      "reward": 0.757105141878128,
      "reward_std": 0.07216768711805344,
      "reward_std_all": 0.06591044925153255,
      "rewards-max/forensics_iou_reward": 0.9880228042602539,
      "rewards-max/forensics_strict_edge_reward": 0.5733275562524796,
      "rewards-max/forensics_tiou_reward": 0.9833307564258575,
      "rewards/forensics_iou_reward": 0.6441778242588043,
      "rewards/forensics_strict_edge_reward": 0.08111143857240677,
      "rewards/forensics_tiou_reward": 0.6062460839748383,
      "step": 1111
    },
    {
      "completion_length": 24.28125,
      "epoch": 3.4116743471582183,
      "grad_norm": 9.470738410949707,
      "kl": 0.64453125,
      "learning_rate": 1.4538461538461538e-07,
      "loss": -0.0241,
      "reward": 0.7894587218761444,
      "reward_std": 0.030067639774642885,
      "reward_std_all": 0.027821660973131657,
      "rewards-max/forensics_iou_reward": 0.9753424525260925,
      "rewards-max/forensics_strict_edge_reward": 0.7744058072566986,
      "rewards-max/forensics_tiou_reward": 0.9710721075534821,
      "rewards/forensics_iou_reward": 0.7679699659347534,
      "rewards/forensics_strict_edge_reward": 0.10523643717169762,
      "rewards/forensics_tiou_reward": 0.726906955242157,
      "step": 1112
    },
    {
      "completion_length": 25.53125,
      "epoch": 3.4147465437788016,
      "grad_norm": 6.035399436950684,
      "kl": 0.568359375,
      "learning_rate": 1.4461538461538462e-07,
      "loss": 0.0219,
      "reward": 0.7566595077514648,
      "reward_std": 0.008800828596577048,
      "reward_std_all": 0.03195745870471001,
      "rewards-max/forensics_iou_reward": 0.9745784401893616,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9681399166584015,
      "rewards/forensics_iou_reward": 0.7335238456726074,
      "rewards/forensics_strict_edge_reward": 0.10090233013033867,
      "rewards/forensics_tiou_reward": 0.7026822566986084,
      "step": 1113
    },
    {
      "completion_length": 27.125,
      "epoch": 3.4178187403993854,
      "grad_norm": 6.918250560760498,
      "kl": 0.3701171875,
      "learning_rate": 1.4384615384615384e-07,
      "loss": 0.024,
      "reward": 0.783165693283081,
      "reward_std": 0.02616029419004917,
      "reward_std_all": 0.026410856284201145,
      "rewards-max/forensics_iou_reward": 0.9792123138904572,
      "rewards-max/forensics_strict_edge_reward": 0.538444846868515,
      "rewards-max/forensics_tiou_reward": 0.9720596969127655,
      "rewards/forensics_iou_reward": 0.7053832709789276,
      "rewards/forensics_strict_edge_reward": 0.06840762868523598,
      "rewards/forensics_tiou_reward": 0.656433641910553,
      "step": 1114
    },
    {
      "completion_length": 26.75,
      "epoch": 3.420890937019969,
      "grad_norm": 6.906099319458008,
      "kl": 0.662109375,
      "learning_rate": 1.4307692307692308e-07,
      "loss": 0.0575,
      "reward": 0.7850867211818695,
      "reward_std": 0.030502932146191597,
      "reward_std_all": 0.02897796221077442,
      "rewards-max/forensics_iou_reward": 0.988088458776474,
      "rewards-max/forensics_strict_edge_reward": 0.6988634467124939,
      "rewards-max/forensics_tiou_reward": 0.984470933675766,
      "rewards/forensics_iou_reward": 0.73287034034729,
      "rewards/forensics_strict_edge_reward": 0.0851174034178257,
      "rewards/forensics_tiou_reward": 0.6923068463802338,
      "step": 1115
    },
    {
      "completion_length": 23.859375,
      "epoch": 3.423963133640553,
      "grad_norm": 9.240121841430664,
      "kl": 0.5654296875,
      "learning_rate": 1.423076923076923e-07,
      "loss": 0.0046,
      "reward": 0.7136178910732269,
      "reward_std": 0.061196668073534966,
      "reward_std_all": 0.07227685488760471,
      "rewards-max/forensics_iou_reward": 0.9931839108467102,
      "rewards-max/forensics_strict_edge_reward": 0.6556849330663681,
      "rewards-max/forensics_tiou_reward": 0.9899217486381531,
      "rewards/forensics_iou_reward": 0.5681791305541992,
      "rewards/forensics_strict_edge_reward": 0.046667350456118584,
      "rewards/forensics_tiou_reward": 0.5172559022903442,
      "step": 1116
    },
    {
      "completion_length": 28.4375,
      "epoch": 3.4270353302611367,
      "grad_norm": 7.551198959350586,
      "kl": 0.5849609375,
      "learning_rate": 1.4153846153846153e-07,
      "loss": -0.0424,
      "reward": 0.7518392205238342,
      "reward_std": 0.07718358188867569,
      "reward_std_all": 0.07072221487760544,
      "rewards-max/forensics_iou_reward": 0.9640061557292938,
      "rewards-max/forensics_strict_edge_reward": 0.45518364012241364,
      "rewards-max/forensics_tiou_reward": 0.9501952826976776,
      "rewards/forensics_iou_reward": 0.6684503853321075,
      "rewards/forensics_strict_edge_reward": 0.06861225701868534,
      "rewards/forensics_tiou_reward": 0.6264676153659821,
      "step": 1117
    },
    {
      "completion_length": 27.96875,
      "epoch": 3.4301075268817205,
      "grad_norm": 7.5418171882629395,
      "kl": 0.45703125,
      "learning_rate": 1.4076923076923074e-07,
      "loss": 0.035,
      "reward": 0.780129611492157,
      "reward_std": 0.02953591151162982,
      "reward_std_all": 0.02912720199674368,
      "rewards-max/forensics_iou_reward": 0.9925206899642944,
      "rewards-max/forensics_strict_edge_reward": 0.8074871003627777,
      "rewards-max/forensics_tiou_reward": 0.9902585744857788,
      "rewards/forensics_iou_reward": 0.7566474378108978,
      "rewards/forensics_strict_edge_reward": 0.13634922355413437,
      "rewards/forensics_tiou_reward": 0.7162146866321564,
      "step": 1118
    },
    {
      "completion_length": 23.46875,
      "epoch": 3.4331797235023043,
      "grad_norm": 7.890410423278809,
      "kl": 0.4501953125,
      "learning_rate": 1.4e-07,
      "loss": 0.0604,
      "reward": 0.7514182329177856,
      "reward_std": 0.0484981881454587,
      "reward_std_all": 0.0514600221067667,
      "rewards-max/forensics_iou_reward": 0.9797474145889282,
      "rewards-max/forensics_strict_edge_reward": 0.5950735807418823,
      "rewards-max/forensics_tiou_reward": 0.9721653759479523,
      "rewards/forensics_iou_reward": 0.6598895788192749,
      "rewards/forensics_strict_edge_reward": 0.04523547552525997,
      "rewards/forensics_tiou_reward": 0.6060895621776581,
      "step": 1119
    },
    {
      "completion_length": 24.078125,
      "epoch": 3.436251920122888,
      "grad_norm": 8.122788429260254,
      "kl": 0.537109375,
      "learning_rate": 1.3923076923076922e-07,
      "loss": 0.0179,
      "reward": 0.7989329099655151,
      "reward_std": 0.011845230241306126,
      "reward_std_all": 0.012513570953160524,
      "rewards-max/forensics_iou_reward": 0.9964685440063477,
      "rewards-max/forensics_strict_edge_reward": 0.5724648237228394,
      "rewards-max/forensics_tiou_reward": 0.9934418499469757,
      "rewards/forensics_iou_reward": 0.7230261862277985,
      "rewards/forensics_strict_edge_reward": 0.061436112970113754,
      "rewards/forensics_tiou_reward": 0.6711916923522949,
      "step": 1120
    },
    {
      "completion_length": 26.140625,
      "epoch": 3.439324116743472,
      "grad_norm": 6.956057071685791,
      "kl": 0.498046875,
      "learning_rate": 1.3846153846153846e-07,
      "loss": 0.0366,
      "reward": 0.7089639902114868,
      "reward_std": 0.0687506478279829,
      "reward_std_all": 0.0869840495288372,
      "rewards-max/forensics_iou_reward": 0.9678021669387817,
      "rewards-max/forensics_strict_edge_reward": 0.5439433008432388,
      "rewards-max/forensics_tiou_reward": 0.960863322019577,
      "rewards/forensics_iou_reward": 0.6482253670692444,
      "rewards/forensics_strict_edge_reward": 0.05451199598610401,
      "rewards/forensics_tiou_reward": 0.6101591289043427,
      "step": 1121
    },
    {
      "completion_length": 23.6875,
      "epoch": 3.442396313364055,
      "grad_norm": 7.334460735321045,
      "kl": 0.359375,
      "learning_rate": 1.376923076923077e-07,
      "loss": 0.0044,
      "reward": 0.742872029542923,
      "reward_std": 0.031365727074444294,
      "reward_std_all": 0.05025816150009632,
      "rewards-max/forensics_iou_reward": 0.977874219417572,
      "rewards-max/forensics_strict_edge_reward": 0.5971094369888306,
      "rewards-max/forensics_tiou_reward": 0.9700478613376617,
      "rewards/forensics_iou_reward": 0.6837020814418793,
      "rewards/forensics_strict_edge_reward": 0.05132394842803478,
      "rewards/forensics_tiou_reward": 0.6269822418689728,
      "step": 1122
    },
    {
      "completion_length": 28.109375,
      "epoch": 3.445468509984639,
      "grad_norm": 7.009228706359863,
      "kl": 0.4521484375,
      "learning_rate": 1.3692307692307692e-07,
      "loss": 0.1253,
      "reward": 0.7423431575298309,
      "reward_std": 0.09887971729040146,
      "reward_std_all": 0.0890200138092041,
      "rewards-max/forensics_iou_reward": 0.9736487567424774,
      "rewards-max/forensics_strict_edge_reward": 0.518597200512886,
      "rewards-max/forensics_tiou_reward": 0.9659492373466492,
      "rewards/forensics_iou_reward": 0.6540337800979614,
      "rewards/forensics_strict_edge_reward": 0.09017261117696762,
      "rewards/forensics_tiou_reward": 0.6241414546966553,
      "step": 1123
    },
    {
      "completion_length": 29.640625,
      "epoch": 3.4485407066052227,
      "grad_norm": 8.717985153198242,
      "kl": 0.390625,
      "learning_rate": 1.3615384615384616e-07,
      "loss": -0.0976,
      "reward": 0.7742205262184143,
      "reward_std": 0.036317138001322746,
      "reward_std_all": 0.03584001958370209,
      "rewards-max/forensics_iou_reward": 0.980040043592453,
      "rewards-max/forensics_strict_edge_reward": 0.7094893157482147,
      "rewards-max/forensics_tiou_reward": 0.9752779006958008,
      "rewards/forensics_iou_reward": 0.7033205926418304,
      "rewards/forensics_strict_edge_reward": 0.09420157596468925,
      "rewards/forensics_tiou_reward": 0.6731121242046356,
      "step": 1124
    },
    {
      "completion_length": 23.4375,
      "epoch": 3.4516129032258065,
      "grad_norm": 6.739078998565674,
      "kl": 0.59375,
      "learning_rate": 1.3538461538461537e-07,
      "loss": 0.0375,
      "reward": 0.7641699016094208,
      "reward_std": 0.07090982608497143,
      "reward_std_all": 0.06353054568171501,
      "rewards-max/forensics_iou_reward": 0.9846417307853699,
      "rewards-max/forensics_strict_edge_reward": 0.6438986212015152,
      "rewards-max/forensics_tiou_reward": 0.9765571057796478,
      "rewards/forensics_iou_reward": 0.7192274034023285,
      "rewards/forensics_strict_edge_reward": 0.0714888870716095,
      "rewards/forensics_tiou_reward": 0.6758274137973785,
      "step": 1125
    },
    {
      "completion_length": 26.484375,
      "epoch": 3.4546850998463903,
      "grad_norm": 8.974876403808594,
      "kl": 0.76171875,
      "learning_rate": 1.346153846153846e-07,
      "loss": 0.0025,
      "reward": 0.7963403165340424,
      "reward_std": 0.017270168289542198,
      "reward_std_all": 0.0169609310105443,
      "rewards-max/forensics_iou_reward": 0.9978306591510773,
      "rewards-max/forensics_strict_edge_reward": 0.7639648914337158,
      "rewards-max/forensics_tiou_reward": 0.9973179697990417,
      "rewards/forensics_iou_reward": 0.7358863055706024,
      "rewards/forensics_strict_edge_reward": 0.11313435062766075,
      "rewards/forensics_tiou_reward": 0.7005436420440674,
      "step": 1126
    },
    {
      "completion_length": 26.515625,
      "epoch": 3.457757296466974,
      "grad_norm": 7.674606800079346,
      "kl": 0.3798828125,
      "learning_rate": 1.3384615384615383e-07,
      "loss": -0.0135,
      "reward": 0.7394278347492218,
      "reward_std": 0.056006452068686485,
      "reward_std_all": 0.059036729857325554,
      "rewards-max/forensics_iou_reward": 0.9718821346759796,
      "rewards-max/forensics_strict_edge_reward": 0.421891450881958,
      "rewards-max/forensics_tiou_reward": 0.961302250623703,
      "rewards/forensics_iou_reward": 0.604438990354538,
      "rewards/forensics_strict_edge_reward": 0.05704299546778202,
      "rewards/forensics_tiou_reward": 0.5606463551521301,
      "step": 1127
    },
    {
      "completion_length": 27.15625,
      "epoch": 3.460829493087558,
      "grad_norm": 6.834789752960205,
      "kl": 0.400390625,
      "learning_rate": 1.3307692307692307e-07,
      "loss": 0.067,
      "reward": 0.7743286490440369,
      "reward_std": 0.04604235850274563,
      "reward_std_all": 0.04287070035934448,
      "rewards-max/forensics_iou_reward": 0.9668478071689606,
      "rewards-max/forensics_strict_edge_reward": 0.2428314909338951,
      "rewards-max/forensics_tiou_reward": 0.9620649516582489,
      "rewards/forensics_iou_reward": 0.6943214535713196,
      "rewards/forensics_strict_edge_reward": 0.02811894193291664,
      "rewards/forensics_tiou_reward": 0.6491120159626007,
      "step": 1128
    },
    {
      "completion_length": 23.65625,
      "epoch": 3.463901689708141,
      "grad_norm": 6.379636764526367,
      "kl": 0.388671875,
      "learning_rate": 1.323076923076923e-07,
      "loss": -0.0139,
      "reward": 0.7859700918197632,
      "reward_std": 0.036578839644789696,
      "reward_std_all": 0.03334597032517195,
      "rewards-max/forensics_iou_reward": 0.9901803731918335,
      "rewards-max/forensics_strict_edge_reward": 0.6505971103906631,
      "rewards-max/forensics_tiou_reward": 0.9879055023193359,
      "rewards/forensics_iou_reward": 0.8061608970165253,
      "rewards/forensics_strict_edge_reward": 0.08917032182216644,
      "rewards/forensics_tiou_reward": 0.7623607516288757,
      "step": 1129
    },
    {
      "completion_length": 28.21875,
      "epoch": 3.466973886328725,
      "grad_norm": 7.553085803985596,
      "kl": 0.51953125,
      "learning_rate": 1.3153846153846155e-07,
      "loss": 0.1166,
      "reward": 0.740267276763916,
      "reward_std": 0.05666336230933666,
      "reward_std_all": 0.0622405931353569,
      "rewards-max/forensics_iou_reward": 0.9955008625984192,
      "rewards-max/forensics_strict_edge_reward": 0.43677301704883575,
      "rewards-max/forensics_tiou_reward": 0.9935351610183716,
      "rewards/forensics_iou_reward": 0.6858179569244385,
      "rewards/forensics_strict_edge_reward": 0.056435972452163696,
      "rewards/forensics_tiou_reward": 0.639618307352066,
      "step": 1130
    },
    {
      "completion_length": 24.96875,
      "epoch": 3.4700460829493087,
      "grad_norm": 6.891150951385498,
      "kl": 0.3720703125,
      "learning_rate": 1.3076923076923076e-07,
      "loss": -0.0184,
      "reward": 0.7381088733673096,
      "reward_std": 0.043669043108820915,
      "reward_std_all": 0.05294694006443024,
      "rewards-max/forensics_iou_reward": 0.9875717759132385,
      "rewards-max/forensics_strict_edge_reward": 0.4991503059864044,
      "rewards-max/forensics_tiou_reward": 0.9787333905696869,
      "rewards/forensics_iou_reward": 0.699487030506134,
      "rewards/forensics_strict_edge_reward": 0.05764610134065151,
      "rewards/forensics_tiou_reward": 0.6428026556968689,
      "step": 1131
    },
    {
      "completion_length": 27.34375,
      "epoch": 3.4731182795698925,
      "grad_norm": 10.079678535461426,
      "kl": 0.671875,
      "learning_rate": 1.3e-07,
      "loss": 0.0587,
      "reward": 0.7645174264907837,
      "reward_std": 0.05960292834788561,
      "reward_std_all": 0.055268154479563236,
      "rewards-max/forensics_iou_reward": 0.9897233843803406,
      "rewards-max/forensics_strict_edge_reward": 0.6488787531852722,
      "rewards-max/forensics_tiou_reward": 0.9863467514514923,
      "rewards/forensics_iou_reward": 0.6520050466060638,
      "rewards/forensics_strict_edge_reward": 0.06079443171620369,
      "rewards/forensics_tiou_reward": 0.5957573652267456,
      "step": 1132
    },
    {
      "completion_length": 25.71875,
      "epoch": 3.4761904761904763,
      "grad_norm": 7.5734333992004395,
      "kl": 0.4306640625,
      "learning_rate": 1.2923076923076924e-07,
      "loss": 0.1144,
      "reward": 0.7775674164295197,
      "reward_std": 0.04016283806413412,
      "reward_std_all": 0.03887409344315529,
      "rewards-max/forensics_iou_reward": 0.9765636622905731,
      "rewards-max/forensics_strict_edge_reward": 0.6317795068025589,
      "rewards-max/forensics_tiou_reward": 0.9586326479911804,
      "rewards/forensics_iou_reward": 0.692677766084671,
      "rewards/forensics_strict_edge_reward": 0.060214689001441,
      "rewards/forensics_tiou_reward": 0.6467545032501221,
      "step": 1133
    },
    {
      "completion_length": 25.1875,
      "epoch": 3.47926267281106,
      "grad_norm": 8.687816619873047,
      "kl": 0.5693359375,
      "learning_rate": 1.2846153846153845e-07,
      "loss": 0.0966,
      "reward": 0.7682445645332336,
      "reward_std": 0.056360467337071896,
      "reward_std_all": 0.051540978252887726,
      "rewards-max/forensics_iou_reward": 0.9883861541748047,
      "rewards-max/forensics_strict_edge_reward": 0.6515887677669525,
      "rewards-max/forensics_tiou_reward": 0.9842771887779236,
      "rewards/forensics_iou_reward": 0.7147473096847534,
      "rewards/forensics_strict_edge_reward": 0.08582212589681149,
      "rewards/forensics_tiou_reward": 0.6663647592067719,
      "step": 1134
    },
    {
      "completion_length": 29.875,
      "epoch": 3.482334869431644,
      "grad_norm": 7.556921482086182,
      "kl": 0.4111328125,
      "learning_rate": 1.276923076923077e-07,
      "loss": -0.0107,
      "reward": 0.6950176060199738,
      "reward_std": 0.11234788130968809,
      "reward_std_all": 0.11153657361865044,
      "rewards-max/forensics_iou_reward": 0.9868914186954498,
      "rewards-max/forensics_strict_edge_reward": 0.49879422783851624,
      "rewards-max/forensics_tiou_reward": 0.9805754721164703,
      "rewards/forensics_iou_reward": 0.6441963016986847,
      "rewards/forensics_strict_edge_reward": 0.04208218585699797,
      "rewards/forensics_tiou_reward": 0.5910807251930237,
      "step": 1135
    },
    {
      "completion_length": 26.125,
      "epoch": 3.485407066052227,
      "grad_norm": 7.896363735198975,
      "kl": 0.51171875,
      "learning_rate": 1.269230769230769e-07,
      "loss": 0.0899,
      "reward": 0.7728703320026398,
      "reward_std": 0.05311843007802963,
      "reward_std_all": 0.04850807785987854,
      "rewards-max/forensics_iou_reward": 0.9603466093540192,
      "rewards-max/forensics_strict_edge_reward": 0.5795845240354538,
      "rewards-max/forensics_tiou_reward": 0.950873076915741,
      "rewards/forensics_iou_reward": 0.6495124995708466,
      "rewards/forensics_strict_edge_reward": 0.04353044554591179,
      "rewards/forensics_tiou_reward": 0.605604887008667,
      "step": 1136
    },
    {
      "completion_length": 22.65625,
      "epoch": 3.488479262672811,
      "grad_norm": 8.740720748901367,
      "kl": 0.4833984375,
      "learning_rate": 1.2615384615384615e-07,
      "loss": -0.0368,
      "reward": 0.7341316640377045,
      "reward_std": 0.06468484550714493,
      "reward_std_all": 0.06835051625967026,
      "rewards-max/forensics_iou_reward": 0.9893521964550018,
      "rewards-max/forensics_strict_edge_reward": 0.6942952275276184,
      "rewards-max/forensics_tiou_reward": 0.985969752073288,
      "rewards/forensics_iou_reward": 0.6725343465805054,
      "rewards/forensics_strict_edge_reward": 0.06727999448776245,
      "rewards/forensics_tiou_reward": 0.6271106749773026,
      "step": 1137
    },
    {
      "completion_length": 26.640625,
      "epoch": 3.4915514592933947,
      "grad_norm": 5.895510196685791,
      "kl": 0.580078125,
      "learning_rate": 1.2538461538461536e-07,
      "loss": 0.0883,
      "reward": 0.7507723569869995,
      "reward_std": 0.05471338331699371,
      "reward_std_all": 0.05472390167415142,
      "rewards-max/forensics_iou_reward": 0.9802233576774597,
      "rewards-max/forensics_strict_edge_reward": 0.47454652190208435,
      "rewards-max/forensics_tiou_reward": 0.97264364361763,
      "rewards/forensics_iou_reward": 0.6738108694553375,
      "rewards/forensics_strict_edge_reward": 0.04844534210860729,
      "rewards/forensics_tiou_reward": 0.6375199854373932,
      "step": 1138
    },
    {
      "completion_length": 26.703125,
      "epoch": 3.4946236559139785,
      "grad_norm": 8.705241203308105,
      "kl": 0.6748046875,
      "learning_rate": 1.246153846153846e-07,
      "loss": 0.064,
      "reward": 0.7359722554683685,
      "reward_std": 0.03692600969225168,
      "reward_std_all": 0.05715970089659095,
      "rewards-max/forensics_iou_reward": 0.9663240313529968,
      "rewards-max/forensics_strict_edge_reward": 0.6065583229064941,
      "rewards-max/forensics_tiou_reward": 0.9578388929367065,
      "rewards/forensics_iou_reward": 0.6762520670890808,
      "rewards/forensics_strict_edge_reward": 0.07022503763437271,
      "rewards/forensics_tiou_reward": 0.6372121572494507,
      "step": 1139
    },
    {
      "completion_length": 26.90625,
      "epoch": 3.4976958525345623,
      "grad_norm": 8.359652519226074,
      "kl": 0.3505859375,
      "learning_rate": 1.2384615384615384e-07,
      "loss": -0.0635,
      "reward": 0.7969483733177185,
      "reward_std": 0.01396703626960516,
      "reward_std_all": 0.014210904017090797,
      "rewards-max/forensics_iou_reward": 0.989171713590622,
      "rewards-max/forensics_strict_edge_reward": 0.8062050342559814,
      "rewards-max/forensics_tiou_reward": 0.9859639108181,
      "rewards/forensics_iou_reward": 0.7384995818138123,
      "rewards/forensics_strict_edge_reward": 0.128652673214674,
      "rewards/forensics_tiou_reward": 0.6934216916561127,
      "step": 1140
    },
    {
      "completion_length": 22.875,
      "epoch": 3.500768049155146,
      "grad_norm": 9.760985374450684,
      "kl": 0.7421875,
      "learning_rate": 1.2307692307692308e-07,
      "loss": 0.0344,
      "reward": 0.7508893609046936,
      "reward_std": 0.01937787141650915,
      "reward_std_all": 0.04069592524319887,
      "rewards-max/forensics_iou_reward": 0.9920634925365448,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9901116192340851,
      "rewards/forensics_iou_reward": 0.7250449061393738,
      "rewards/forensics_strict_edge_reward": 0.12368110194802284,
      "rewards/forensics_tiou_reward": 0.6838651895523071,
      "step": 1141
    },
    {
      "completion_length": 26.78125,
      "epoch": 3.50384024577573,
      "grad_norm": 8.9514799118042,
      "kl": 0.580078125,
      "learning_rate": 1.223076923076923e-07,
      "loss": 0.1009,
      "reward": 0.7883354425430298,
      "reward_std": 0.031226662918925285,
      "reward_std_all": 0.02891156729310751,
      "rewards-max/forensics_iou_reward": 0.9738134145736694,
      "rewards-max/forensics_strict_edge_reward": 0.5948129594326019,
      "rewards-max/forensics_tiou_reward": 0.9693354964256287,
      "rewards/forensics_iou_reward": 0.7420416474342346,
      "rewards/forensics_strict_edge_reward": 0.07072626985609531,
      "rewards/forensics_tiou_reward": 0.7034810781478882,
      "step": 1142
    },
    {
      "completion_length": 26.46875,
      "epoch": 3.506912442396313,
      "grad_norm": 8.72061538696289,
      "kl": 0.396484375,
      "learning_rate": 1.2153846153846154e-07,
      "loss": -0.0066,
      "reward": 0.7871252000331879,
      "reward_std": 0.0320423151133582,
      "reward_std_all": 0.02986375754699111,
      "rewards-max/forensics_iou_reward": 0.9824220836162567,
      "rewards-max/forensics_strict_edge_reward": 0.7000330090522766,
      "rewards-max/forensics_tiou_reward": 0.9753574728965759,
      "rewards/forensics_iou_reward": 0.7411963939666748,
      "rewards/forensics_strict_edge_reward": 0.08043651096522808,
      "rewards/forensics_tiou_reward": 0.7020160853862762,
      "step": 1143
    },
    {
      "completion_length": 23.25,
      "epoch": 3.509984639016897,
      "grad_norm": 18.9714298248291,
      "kl": 1.3408203125,
      "learning_rate": 1.2076923076923075e-07,
      "loss": 0.0921,
      "reward": 0.7229646742343903,
      "reward_std": 0.07729904167354107,
      "reward_std_all": 0.08042877167463303,
      "rewards-max/forensics_iou_reward": 0.9901695251464844,
      "rewards-max/forensics_strict_edge_reward": 0.5294047147035599,
      "rewards-max/forensics_tiou_reward": 0.9866597950458527,
      "rewards/forensics_iou_reward": 0.6898859739303589,
      "rewards/forensics_strict_edge_reward": 0.06812400370836258,
      "rewards/forensics_tiou_reward": 0.6316751837730408,
      "step": 1144
    },
    {
      "completion_length": 26.140625,
      "epoch": 3.5130568356374807,
      "grad_norm": 8.23950481414795,
      "kl": 0.4775390625,
      "learning_rate": 1.2e-07,
      "loss": 0.0084,
      "reward": 0.7903056740760803,
      "reward_std": 0.023670130176469684,
      "reward_std_all": 0.022605532314628363,
      "rewards-max/forensics_iou_reward": 0.9873503148555756,
      "rewards-max/forensics_strict_edge_reward": 0.562157079577446,
      "rewards-max/forensics_tiou_reward": 0.9828484952449799,
      "rewards/forensics_iou_reward": 0.7210978269577026,
      "rewards/forensics_strict_edge_reward": 0.0902373269200325,
      "rewards/forensics_tiou_reward": 0.6663078963756561,
      "step": 1145
    },
    {
      "completion_length": 26.046875,
      "epoch": 3.5161290322580645,
      "grad_norm": 8.120950698852539,
      "kl": 0.7529296875,
      "learning_rate": 1.1923076923076923e-07,
      "loss": -0.0336,
      "reward": 0.7129751741886139,
      "reward_std": 0.10774189606308937,
      "reward_std_all": 0.10315546207129955,
      "rewards-max/forensics_iou_reward": 0.8988340198993683,
      "rewards-max/forensics_strict_edge_reward": 0.3981185480952263,
      "rewards-max/forensics_tiou_reward": 0.8641709685325623,
      "rewards/forensics_iou_reward": 0.5874934792518616,
      "rewards/forensics_strict_edge_reward": 0.06203046254813671,
      "rewards/forensics_tiou_reward": 0.5464404225349426,
      "step": 1146
    },
    {
      "completion_length": 29.328125,
      "epoch": 3.5192012288786483,
      "grad_norm": 6.442105770111084,
      "kl": 0.5302734375,
      "learning_rate": 1.1846153846153846e-07,
      "loss": 0.0379,
      "reward": 0.7804501056671143,
      "reward_std": 0.04357650876045227,
      "reward_std_all": 0.039721750654280186,
      "rewards-max/forensics_iou_reward": 0.9765461981296539,
      "rewards-max/forensics_strict_edge_reward": 0.3017593175172806,
      "rewards-max/forensics_tiou_reward": 0.964298278093338,
      "rewards/forensics_iou_reward": 0.7122507095336914,
      "rewards/forensics_strict_edge_reward": 0.03823007270693779,
      "rewards/forensics_tiou_reward": 0.6720596551895142,
      "step": 1147
    },
    {
      "completion_length": 29.0,
      "epoch": 3.522273425499232,
      "grad_norm": 6.082571506500244,
      "kl": 0.3984375,
      "learning_rate": 1.1769230769230768e-07,
      "loss": 0.0205,
      "reward": 0.7095761001110077,
      "reward_std": 0.09663297981023788,
      "reward_std_all": 0.09787118062376976,
      "rewards-max/forensics_iou_reward": 0.9964562654495239,
      "rewards-max/forensics_strict_edge_reward": 0.4251216650009155,
      "rewards-max/forensics_tiou_reward": 0.9933028817176819,
      "rewards/forensics_iou_reward": 0.6706033945083618,
      "rewards/forensics_strict_edge_reward": 0.056315623223781586,
      "rewards/forensics_tiou_reward": 0.6322911381721497,
      "step": 1148
    },
    {
      "completion_length": 26.25,
      "epoch": 3.525345622119816,
      "grad_norm": 8.558990478515625,
      "kl": 0.4462890625,
      "learning_rate": 1.1692307692307691e-07,
      "loss": 0.0098,
      "reward": 0.7048350870609283,
      "reward_std": 0.07295691967010498,
      "reward_std_all": 0.09052587300539017,
      "rewards-max/forensics_iou_reward": 0.9744569957256317,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9676699936389923,
      "rewards/forensics_iou_reward": 0.6238828897476196,
      "rewards/forensics_strict_edge_reward": 0.0868326760828495,
      "rewards/forensics_tiou_reward": 0.593207836151123,
      "step": 1149
    },
    {
      "completion_length": 25.09375,
      "epoch": 3.528417818740399,
      "grad_norm": 8.940905570983887,
      "kl": 0.6015625,
      "learning_rate": 1.1615384615384615e-07,
      "loss": 0.0473,
      "reward": 0.7752281427383423,
      "reward_std": 0.0298363845795393,
      "reward_std_all": 0.031203716062009335,
      "rewards-max/forensics_iou_reward": 0.9589266180992126,
      "rewards-max/forensics_strict_edge_reward": 0.30594509094953537,
      "rewards-max/forensics_tiou_reward": 0.9460338056087494,
      "rewards/forensics_iou_reward": 0.6761914491653442,
      "rewards/forensics_strict_edge_reward": 0.03322631632909179,
      "rewards/forensics_tiou_reward": 0.6308017671108246,
      "step": 1150
    },
    {
      "completion_length": 25.734375,
      "epoch": 3.531490015360983,
      "grad_norm": 6.389458656311035,
      "kl": 0.4716796875,
      "learning_rate": 1.1538461538461539e-07,
      "loss": 0.0694,
      "reward": 0.7789717316627502,
      "reward_std": 0.041895680129528046,
      "reward_std_all": 0.038893493823707104,
      "rewards-max/forensics_iou_reward": 0.9926804602146149,
      "rewards-max/forensics_strict_edge_reward": 0.740818202495575,
      "rewards-max/forensics_tiou_reward": 0.9898041188716888,
      "rewards/forensics_iou_reward": 0.7292014360427856,
      "rewards/forensics_strict_edge_reward": 0.10085485875606537,
      "rewards/forensics_tiou_reward": 0.6933073103427887,
      "step": 1151
    },
    {
      "completion_length": 23.46875,
      "epoch": 3.5345622119815667,
      "grad_norm": 7.276225566864014,
      "kl": 0.375,
      "learning_rate": 1.1461538461538462e-07,
      "loss": -0.0311,
      "reward": 0.7880038022994995,
      "reward_std": 0.021127174608409405,
      "reward_std_all": 0.022385863587260246,
      "rewards-max/forensics_iou_reward": 0.9882949590682983,
      "rewards-max/forensics_strict_edge_reward": 0.7697611749172211,
      "rewards-max/forensics_tiou_reward": 0.9819979071617126,
      "rewards/forensics_iou_reward": 0.6977759301662445,
      "rewards/forensics_strict_edge_reward": 0.060632433742284775,
      "rewards/forensics_tiou_reward": 0.637777715921402,
      "step": 1152
    },
    {
      "completion_length": 26.359375,
      "epoch": 3.5376344086021505,
      "grad_norm": 8.09441089630127,
      "kl": 0.4794921875,
      "learning_rate": 1.1384615384615384e-07,
      "loss": 0.0388,
      "reward": 0.7442405223846436,
      "reward_std": 0.06815915368497372,
      "reward_std_all": 0.0659500751644373,
      "rewards-max/forensics_iou_reward": 0.9514763355255127,
      "rewards-max/forensics_strict_edge_reward": 0.38581259548664093,
      "rewards-max/forensics_tiou_reward": 0.940301388502121,
      "rewards/forensics_iou_reward": 0.6172578930854797,
      "rewards/forensics_strict_edge_reward": 0.04418962262570858,
      "rewards/forensics_tiou_reward": 0.5703544318675995,
      "step": 1153
    },
    {
      "completion_length": 24.34375,
      "epoch": 3.5407066052227343,
      "grad_norm": 9.500958442687988,
      "kl": 0.58203125,
      "learning_rate": 1.1307692307692307e-07,
      "loss": -0.0025,
      "reward": 0.7646931707859039,
      "reward_std": 0.05335509777069092,
      "reward_std_all": 0.050790103152394295,
      "rewards-max/forensics_iou_reward": 0.9837942719459534,
      "rewards-max/forensics_strict_edge_reward": 0.5794400870800018,
      "rewards-max/forensics_tiou_reward": 0.9795729517936707,
      "rewards/forensics_iou_reward": 0.7301929891109467,
      "rewards/forensics_strict_edge_reward": 0.09706412628293037,
      "rewards/forensics_tiou_reward": 0.681698203086853,
      "step": 1154
    },
    {
      "completion_length": 25.046875,
      "epoch": 3.543778801843318,
      "grad_norm": 13.08638858795166,
      "kl": 0.7138671875,
      "learning_rate": 1.123076923076923e-07,
      "loss": 0.0749,
      "reward": 0.76771479845047,
      "reward_std": 0.06729681044816971,
      "reward_std_all": 0.059892311692237854,
      "rewards-max/forensics_iou_reward": 0.9829059839248657,
      "rewards-max/forensics_strict_edge_reward": 0.6858413815498352,
      "rewards-max/forensics_tiou_reward": 0.9789076447486877,
      "rewards/forensics_iou_reward": 0.7192035019397736,
      "rewards/forensics_strict_edge_reward": 0.11762207001447678,
      "rewards/forensics_tiou_reward": 0.6810902953147888,
      "step": 1155
    },
    {
      "completion_length": 26.4375,
      "epoch": 3.546850998463902,
      "grad_norm": 6.6865973472595215,
      "kl": 0.580078125,
      "learning_rate": 1.1153846153846154e-07,
      "loss": 0.052,
      "reward": 0.7836866676807404,
      "reward_std": 0.03873224463313818,
      "reward_std_all": 0.0356993954628706,
      "rewards-max/forensics_iou_reward": 0.9496459364891052,
      "rewards-max/forensics_strict_edge_reward": 0.3907497674226761,
      "rewards-max/forensics_tiou_reward": 0.9362497329711914,
      "rewards/forensics_iou_reward": 0.7328427731990814,
      "rewards/forensics_strict_edge_reward": 0.06245681643486023,
      "rewards/forensics_tiou_reward": 0.6816442906856537,
      "step": 1156
    },
    {
      "completion_length": 29.0,
      "epoch": 3.549923195084485,
      "grad_norm": 7.372470855712891,
      "kl": 0.447265625,
      "learning_rate": 1.1076923076923077e-07,
      "loss": -0.0432,
      "reward": 0.806294709444046,
      "reward_std": 0.0009607442771084607,
      "reward_std_all": 0.0029330262914299965,
      "rewards-max/forensics_iou_reward": 0.9963545799255371,
      "rewards-max/forensics_strict_edge_reward": 0.8414767682552338,
      "rewards-max/forensics_tiou_reward": 0.9948465824127197,
      "rewards/forensics_iou_reward": 0.8344284296035767,
      "rewards/forensics_strict_edge_reward": 0.08648436516523361,
      "rewards/forensics_tiou_reward": 0.79686439037323,
      "step": 1157
    },
    {
      "completion_length": 26.34375,
      "epoch": 3.5529953917050694,
      "grad_norm": 6.6681976318359375,
      "kl": 0.578125,
      "learning_rate": 1.0999999999999999e-07,
      "loss": -0.0012,
      "reward": 0.7684138119220734,
      "reward_std": 0.059884512796998024,
      "reward_std_all": 0.05434391647577286,
      "rewards-max/forensics_iou_reward": 0.9938511848449707,
      "rewards-max/forensics_strict_edge_reward": 0.6983798742294312,
      "rewards-max/forensics_tiou_reward": 0.9917446374893188,
      "rewards/forensics_iou_reward": 0.729174941778183,
      "rewards/forensics_strict_edge_reward": 0.1284804791212082,
      "rewards/forensics_tiou_reward": 0.6925285160541534,
      "step": 1158
    },
    {
      "completion_length": 25.359375,
      "epoch": 3.5560675883256527,
      "grad_norm": 9.703442573547363,
      "kl": 0.408203125,
      "learning_rate": 1.0923076923076922e-07,
      "loss": -0.0418,
      "reward": 0.7760465741157532,
      "reward_std": 0.029413506388664246,
      "reward_std_all": 0.03044022247195244,
      "rewards-max/forensics_iou_reward": 0.9774546027183533,
      "rewards-max/forensics_strict_edge_reward": 0.6105617582798004,
      "rewards-max/forensics_tiou_reward": 0.9686627984046936,
      "rewards/forensics_iou_reward": 0.7416501045227051,
      "rewards/forensics_strict_edge_reward": 0.09436486661434174,
      "rewards/forensics_tiou_reward": 0.6914886236190796,
      "step": 1159
    },
    {
      "completion_length": 28.546875,
      "epoch": 3.5591397849462365,
      "grad_norm": 7.626988887786865,
      "kl": 0.5,
      "learning_rate": 1.0846153846153845e-07,
      "loss": 0.0487,
      "reward": 0.734889566898346,
      "reward_std": 0.07075918838381767,
      "reward_std_all": 0.07152531109750271,
      "rewards-max/forensics_iou_reward": 0.9932081997394562,
      "rewards-max/forensics_strict_edge_reward": 0.6581914722919464,
      "rewards-max/forensics_tiou_reward": 0.9903469085693359,
      "rewards/forensics_iou_reward": 0.6509180963039398,
      "rewards/forensics_strict_edge_reward": 0.10892269015312195,
      "rewards/forensics_tiou_reward": 0.6172857880592346,
      "step": 1160
    },
    {
      "completion_length": 22.65625,
      "epoch": 3.5622119815668203,
      "grad_norm": 8.027215957641602,
      "kl": 0.3701171875,
      "learning_rate": 1.076923076923077e-07,
      "loss": -0.0786,
      "reward": 0.7823931872844696,
      "reward_std": 0.02344598388299346,
      "reward_std_all": 0.024759800173342228,
      "rewards-max/forensics_iou_reward": 0.9758575856685638,
      "rewards-max/forensics_strict_edge_reward": 0.6065583229064941,
      "rewards-max/forensics_tiou_reward": 0.9671388268470764,
      "rewards/forensics_iou_reward": 0.7218974232673645,
      "rewards/forensics_strict_edge_reward": 0.07693955488502979,
      "rewards/forensics_tiou_reward": 0.6598919332027435,
      "step": 1161
    },
    {
      "completion_length": 26.234375,
      "epoch": 3.565284178187404,
      "grad_norm": 6.008260250091553,
      "kl": 0.552734375,
      "learning_rate": 1.0692307692307693e-07,
      "loss": 0.0378,
      "reward": 0.794983446598053,
      "reward_std": 0.017402217723429203,
      "reward_std_all": 0.017038672231137753,
      "rewards-max/forensics_iou_reward": 0.9868862926959991,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9837886989116669,
      "rewards/forensics_iou_reward": 0.7644649147987366,
      "rewards/forensics_strict_edge_reward": 0.10596181452274323,
      "rewards/forensics_tiou_reward": 0.7255379557609558,
      "step": 1162
    },
    {
      "completion_length": 23.796875,
      "epoch": 3.568356374807988,
      "grad_norm": 9.193981170654297,
      "kl": 0.3515625,
      "learning_rate": 1.0615384615384615e-07,
      "loss": 0.0239,
      "reward": 0.7822127342224121,
      "reward_std": 0.04165598750114441,
      "reward_std_all": 0.03781292587518692,
      "rewards-max/forensics_iou_reward": 0.9823146164417267,
      "rewards-max/forensics_strict_edge_reward": 0.5829301476478577,
      "rewards-max/forensics_tiou_reward": 0.9784929752349854,
      "rewards/forensics_iou_reward": 0.6946513056755066,
      "rewards/forensics_strict_edge_reward": 0.11022185161709785,
      "rewards/forensics_tiou_reward": 0.6489090323448181,
      "step": 1163
    },
    {
      "completion_length": 26.84375,
      "epoch": 3.571428571428571,
      "grad_norm": 7.914222717285156,
      "kl": 0.4228515625,
      "learning_rate": 1.0538461538461538e-07,
      "loss": 0.0763,
      "reward": 0.7250556945800781,
      "reward_std": 0.0845441147685051,
      "reward_std_all": 0.08266850560903549,
      "rewards-max/forensics_iou_reward": 0.9252782464027405,
      "rewards-max/forensics_strict_edge_reward": 0.26107610762119293,
      "rewards-max/forensics_tiou_reward": 0.9051661193370819,
      "rewards/forensics_iou_reward": 0.5918158143758774,
      "rewards/forensics_strict_edge_reward": 0.04649627115577459,
      "rewards/forensics_tiou_reward": 0.548767015337944,
      "step": 1164
    },
    {
      "completion_length": 31.15625,
      "epoch": 3.5745007680491554,
      "grad_norm": 9.091349601745605,
      "kl": 0.4013671875,
      "learning_rate": 1.0461538461538461e-07,
      "loss": 0.0331,
      "reward": 0.7517987787723541,
      "reward_std": 0.06939378194510937,
      "reward_std_all": 0.06513902544975281,
      "rewards-max/forensics_iou_reward": 0.9819999933242798,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.977897584438324,
      "rewards/forensics_iou_reward": 0.6596614718437195,
      "rewards/forensics_strict_edge_reward": 0.06914530508220196,
      "rewards/forensics_tiou_reward": 0.6123790740966797,
      "step": 1165
    },
    {
      "completion_length": 24.28125,
      "epoch": 3.5775729646697387,
      "grad_norm": 8.419288635253906,
      "kl": 0.458984375,
      "learning_rate": 1.0384615384615385e-07,
      "loss": -0.0143,
      "reward": 0.7647444903850555,
      "reward_std": 0.06437924318015575,
      "reward_std_all": 0.05847938172519207,
      "rewards-max/forensics_iou_reward": 0.9771671295166016,
      "rewards-max/forensics_strict_edge_reward": 0.48699167370796204,
      "rewards-max/forensics_tiou_reward": 0.9676006138324738,
      "rewards/forensics_iou_reward": 0.7210003733634949,
      "rewards/forensics_strict_edge_reward": 0.07725103944540024,
      "rewards/forensics_tiou_reward": 0.6710876822471619,
      "step": 1166
    },
    {
      "completion_length": 25.53125,
      "epoch": 3.5806451612903225,
      "grad_norm": 6.360409259796143,
      "kl": 0.5146484375,
      "learning_rate": 1.0307692307692307e-07,
      "loss": 0.0832,
      "reward": 0.7459350526332855,
      "reward_std": 0.03203282877802849,
      "reward_std_all": 0.04668245278298855,
      "rewards-max/forensics_iou_reward": 0.99120232462883,
      "rewards-max/forensics_strict_edge_reward": 0.6942952275276184,
      "rewards-max/forensics_tiou_reward": 0.9897506237030029,
      "rewards/forensics_iou_reward": 0.6882838904857635,
      "rewards/forensics_strict_edge_reward": 0.11654295586049557,
      "rewards/forensics_tiou_reward": 0.6379052698612213,
      "step": 1167
    },
    {
      "completion_length": 30.234375,
      "epoch": 3.5837173579109063,
      "grad_norm": 5.983321666717529,
      "kl": 0.52734375,
      "learning_rate": 1.023076923076923e-07,
      "loss": 0.0264,
      "reward": 0.7652247548103333,
      "reward_std": 0.0679200254380703,
      "reward_std_all": 0.06072890758514404,
      "rewards-max/forensics_iou_reward": 0.9948706328868866,
      "rewards-max/forensics_strict_edge_reward": 0.6495180428028107,
      "rewards-max/forensics_tiou_reward": 0.9928975105285645,
      "rewards/forensics_iou_reward": 0.7002209722995758,
      "rewards/forensics_strict_edge_reward": 0.12777682021260262,
      "rewards/forensics_tiou_reward": 0.6666069030761719,
      "step": 1168
    },
    {
      "completion_length": 27.796875,
      "epoch": 3.58678955453149,
      "grad_norm": 7.445324420928955,
      "kl": 0.45703125,
      "learning_rate": 1.0153846153846153e-07,
      "loss": -0.054,
      "reward": 0.7632412612438202,
      "reward_std": 0.005262294318526983,
      "reward_std_all": 0.02535327710211277,
      "rewards-max/forensics_iou_reward": 0.9940425157546997,
      "rewards-max/forensics_strict_edge_reward": 0.42116816341876984,
      "rewards-max/forensics_tiou_reward": 0.9899214208126068,
      "rewards/forensics_iou_reward": 0.760101318359375,
      "rewards/forensics_strict_edge_reward": 0.057686975225806236,
      "rewards/forensics_tiou_reward": 0.7307034134864807,
      "step": 1169
    },
    {
      "completion_length": 25.375,
      "epoch": 3.589861751152074,
      "grad_norm": 8.87951374053955,
      "kl": 0.40625,
      "learning_rate": 1.0076923076923076e-07,
      "loss": 0.0118,
      "reward": 0.7669236660003662,
      "reward_std": 0.03705655783414841,
      "reward_std_all": 0.038720520911738276,
      "rewards-max/forensics_iou_reward": 0.9886437058448792,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9839856624603271,
      "rewards/forensics_iou_reward": 0.7299742698669434,
      "rewards/forensics_strict_edge_reward": 0.10479724779725075,
      "rewards/forensics_tiou_reward": 0.6853578090667725,
      "step": 1170
    },
    {
      "completion_length": 23.96875,
      "epoch": 3.592933947772657,
      "grad_norm": 7.416459083557129,
      "kl": 0.353515625,
      "learning_rate": 1e-07,
      "loss": 0.0099,
      "reward": 0.7666527330875397,
      "reward_std": 0.023212053580209613,
      "reward_std_all": 0.03157123678829521,
      "rewards-max/forensics_iou_reward": 0.975488543510437,
      "rewards-max/forensics_strict_edge_reward": 0.44073912501335144,
      "rewards-max/forensics_tiou_reward": 0.9641883373260498,
      "rewards/forensics_iou_reward": 0.6930252909660339,
      "rewards/forensics_strict_edge_reward": 0.06155227869749069,
      "rewards/forensics_tiou_reward": 0.6322470605373383,
      "step": 1171
    },
    {
      "completion_length": 27.40625,
      "epoch": 3.5960061443932414,
      "grad_norm": 7.525160789489746,
      "kl": 0.4169921875,
      "learning_rate": 9.923076923076922e-08,
      "loss": 0.0098,
      "reward": 0.7323644459247589,
      "reward_std": 0.06476440373808146,
      "reward_std_all": 0.06686747260391712,
      "rewards-max/forensics_iou_reward": 0.978963166475296,
      "rewards-max/forensics_strict_edge_reward": 0.36683474481105804,
      "rewards-max/forensics_tiou_reward": 0.9679731130599976,
      "rewards/forensics_iou_reward": 0.675243467092514,
      "rewards/forensics_strict_edge_reward": 0.037358587607741356,
      "rewards/forensics_tiou_reward": 0.6342249810695648,
      "step": 1172
    },
    {
      "completion_length": 25.203125,
      "epoch": 3.5990783410138247,
      "grad_norm": 9.668485641479492,
      "kl": 0.40625,
      "learning_rate": 9.846153846153846e-08,
      "loss": 0.0569,
      "reward": 0.7569408714771271,
      "reward_std": 0.0759811345487833,
      "reward_std_all": 0.06872181408107281,
      "rewards-max/forensics_iou_reward": 0.9726207852363586,
      "rewards-max/forensics_strict_edge_reward": 0.5256026089191437,
      "rewards-max/forensics_tiou_reward": 0.954020231962204,
      "rewards/forensics_iou_reward": 0.7149190902709961,
      "rewards/forensics_strict_edge_reward": 0.07596099004149437,
      "rewards/forensics_tiou_reward": 0.6779325604438782,
      "step": 1173
    },
    {
      "completion_length": 23.203125,
      "epoch": 3.6021505376344085,
      "grad_norm": 9.321643829345703,
      "kl": 0.3623046875,
      "learning_rate": 9.769230769230769e-08,
      "loss": -0.0023,
      "reward": 0.7571873366832733,
      "reward_std": 0.023882487788796425,
      "reward_std_all": 0.036466084420681,
      "rewards-max/forensics_iou_reward": 0.9933687150478363,
      "rewards-max/forensics_strict_edge_reward": 0.824172168970108,
      "rewards-max/forensics_tiou_reward": 0.9896753430366516,
      "rewards/forensics_iou_reward": 0.7435938119888306,
      "rewards/forensics_strict_edge_reward": 0.12451464682817459,
      "rewards/forensics_tiou_reward": 0.7006298899650574,
      "step": 1174
    },
    {
      "completion_length": 24.5,
      "epoch": 3.6052227342549923,
      "grad_norm": 8.48859691619873,
      "kl": 0.478515625,
      "learning_rate": 9.692307692307692e-08,
      "loss": 0.0119,
      "reward": 0.7896716892719269,
      "reward_std": 0.016271673026494682,
      "reward_std_all": 0.017720842268317938,
      "rewards-max/forensics_iou_reward": 0.9945659339427948,
      "rewards-max/forensics_strict_edge_reward": 0.7165639698505402,
      "rewards-max/forensics_tiou_reward": 0.9925587475299835,
      "rewards/forensics_iou_reward": 0.7257782518863678,
      "rewards/forensics_strict_edge_reward": 0.09881296381354332,
      "rewards/forensics_tiou_reward": 0.6828382611274719,
      "step": 1175
    },
    {
      "completion_length": 26.078125,
      "epoch": 3.608294930875576,
      "grad_norm": 8.051315307617188,
      "kl": 0.5859375,
      "learning_rate": 9.615384615384616e-08,
      "loss": -0.0588,
      "reward": 0.8036300539970398,
      "reward_std": 0.003755968762561679,
      "reward_std_all": 0.005446471623145044,
      "rewards-max/forensics_iou_reward": 0.970013439655304,
      "rewards-max/forensics_strict_edge_reward": 0.5050012171268463,
      "rewards-max/forensics_tiou_reward": 0.9597599506378174,
      "rewards/forensics_iou_reward": 0.7083956599235535,
      "rewards/forensics_strict_edge_reward": 0.08814247697591782,
      "rewards/forensics_tiou_reward": 0.6649402379989624,
      "step": 1176
    },
    {
      "completion_length": 26.46875,
      "epoch": 3.61136712749616,
      "grad_norm": 6.944526672363281,
      "kl": 0.4384765625,
      "learning_rate": 9.538461538461538e-08,
      "loss": -0.0066,
      "reward": 0.7917308211326599,
      "reward_std": 0.01925147883594036,
      "reward_std_all": 0.01952230604365468,
      "rewards-max/forensics_iou_reward": 0.9897131025791168,
      "rewards-max/forensics_strict_edge_reward": 0.6575055122375488,
      "rewards-max/forensics_tiou_reward": 0.9886057078838348,
      "rewards/forensics_iou_reward": 0.6962171494960785,
      "rewards/forensics_strict_edge_reward": 0.08163745328783989,
      "rewards/forensics_tiou_reward": 0.6561339199542999,
      "step": 1177
    },
    {
      "completion_length": 25.578125,
      "epoch": 3.6144393241167436,
      "grad_norm": 7.2410502433776855,
      "kl": 0.5908203125,
      "learning_rate": 9.461538461538461e-08,
      "loss": 0.2011,
      "reward": 0.7097683548927307,
      "reward_std": 0.12444831430912018,
      "reward_std_all": 0.11545337736606598,
      "rewards-max/forensics_iou_reward": 0.9763971865177155,
      "rewards-max/forensics_strict_edge_reward": 0.6966203600168228,
      "rewards-max/forensics_tiou_reward": 0.9706141948699951,
      "rewards/forensics_iou_reward": 0.6292419731616974,
      "rewards/forensics_strict_edge_reward": 0.09620027989149094,
      "rewards/forensics_tiou_reward": 0.5881876051425934,
      "step": 1178
    },
    {
      "completion_length": 26.921875,
      "epoch": 3.6175115207373274,
      "grad_norm": 6.486563205718994,
      "kl": 0.4541015625,
      "learning_rate": 9.384615384615384e-08,
      "loss": 0.0514,
      "reward": 0.7625922560691833,
      "reward_std": 0.07707376405596733,
      "reward_std_all": 0.0682651437819004,
      "rewards-max/forensics_iou_reward": 0.9922744035720825,
      "rewards-max/forensics_strict_edge_reward": 0.6604506373405457,
      "rewards-max/forensics_tiou_reward": 0.9903771877288818,
      "rewards/forensics_iou_reward": 0.7645038664340973,
      "rewards/forensics_strict_edge_reward": 0.10601511970162392,
      "rewards/forensics_tiou_reward": 0.7316338419914246,
      "step": 1179
    },
    {
      "completion_length": 25.890625,
      "epoch": 3.6205837173579107,
      "grad_norm": 6.944598197937012,
      "kl": 0.4677734375,
      "learning_rate": 9.307692307692306e-08,
      "loss": 0.0349,
      "reward": 0.7242847084999084,
      "reward_std": 0.043963370844721794,
      "reward_std_all": 0.06487051397562027,
      "rewards-max/forensics_iou_reward": 0.9863977432250977,
      "rewards-max/forensics_strict_edge_reward": 0.44064828753471375,
      "rewards-max/forensics_tiou_reward": 0.9799864590167999,
      "rewards/forensics_iou_reward": 0.6911218464374542,
      "rewards/forensics_strict_edge_reward": 0.04190992098301649,
      "rewards/forensics_tiou_reward": 0.6554870903491974,
      "step": 1180
    },
    {
      "completion_length": 24.96875,
      "epoch": 3.6236559139784945,
      "grad_norm": 12.013938903808594,
      "kl": 0.4326171875,
      "learning_rate": 9.23076923076923e-08,
      "loss": -0.0455,
      "reward": 0.7631859183311462,
      "reward_std": 0.014725966146215796,
      "reward_std_all": 0.028800694504752755,
      "rewards-max/forensics_iou_reward": 0.9943064749240875,
      "rewards-max/forensics_strict_edge_reward": 0.5540220141410828,
      "rewards-max/forensics_tiou_reward": 0.9915203750133514,
      "rewards/forensics_iou_reward": 0.7213583588600159,
      "rewards/forensics_strict_edge_reward": 0.07161496207118034,
      "rewards/forensics_tiou_reward": 0.6758513152599335,
      "step": 1181
    },
    {
      "completion_length": 23.25,
      "epoch": 3.6267281105990783,
      "grad_norm": 6.9476494789123535,
      "kl": 0.5595703125,
      "learning_rate": 9.153846153846153e-08,
      "loss": 0.0509,
      "reward": 0.7494361698627472,
      "reward_std": 0.04732985328882933,
      "reward_std_all": 0.05277233477681875,
      "rewards-max/forensics_iou_reward": 0.9819387793540955,
      "rewards-max/forensics_strict_edge_reward": 0.6118685156106949,
      "rewards-max/forensics_tiou_reward": 0.9791035652160645,
      "rewards/forensics_iou_reward": 0.6745662987232208,
      "rewards/forensics_strict_edge_reward": 0.09218202531337738,
      "rewards/forensics_tiou_reward": 0.6280657351016998,
      "step": 1182
    },
    {
      "completion_length": 25.703125,
      "epoch": 3.629800307219662,
      "grad_norm": 8.974733352661133,
      "kl": 0.357421875,
      "learning_rate": 9.076923076923076e-08,
      "loss": -0.0432,
      "reward": 0.7954808175563812,
      "reward_std": 0.012455033604055643,
      "reward_std_all": 0.013654401525855064,
      "rewards-max/forensics_iou_reward": 0.9976803362369537,
      "rewards-max/forensics_strict_edge_reward": 0.8518882095813751,
      "rewards-max/forensics_tiou_reward": 0.9971277415752411,
      "rewards/forensics_iou_reward": 0.737562358379364,
      "rewards/forensics_strict_edge_reward": 0.09190384298563004,
      "rewards/forensics_tiou_reward": 0.6987516582012177,
      "step": 1183
    },
    {
      "completion_length": 29.359375,
      "epoch": 3.632872503840246,
      "grad_norm": 9.123453140258789,
      "kl": 0.4404296875,
      "learning_rate": 9e-08,
      "loss": 0.039,
      "reward": 0.7788182497024536,
      "reward_std": 0.038156239315867424,
      "reward_std_all": 0.0360535578802228,
      "rewards-max/forensics_iou_reward": 0.9815849959850311,
      "rewards-max/forensics_strict_edge_reward": 0.5175835937261581,
      "rewards-max/forensics_tiou_reward": 0.9733245670795441,
      "rewards/forensics_iou_reward": 0.7430989146232605,
      "rewards/forensics_strict_edge_reward": 0.0732487291097641,
      "rewards/forensics_tiou_reward": 0.7066691219806671,
      "step": 1184
    },
    {
      "completion_length": 28.890625,
      "epoch": 3.6359447004608296,
      "grad_norm": 6.572298049926758,
      "kl": 0.4208984375,
      "learning_rate": 8.923076923076923e-08,
      "loss": 0.0599,
      "reward": 0.7446985840797424,
      "reward_std": 0.054633695632219315,
      "reward_std_all": 0.05662590265274048,
      "rewards-max/forensics_iou_reward": 0.9880546033382416,
      "rewards-max/forensics_strict_edge_reward": 0.8135445415973663,
      "rewards-max/forensics_tiou_reward": 0.9846440851688385,
      "rewards/forensics_iou_reward": 0.6363034248352051,
      "rewards/forensics_strict_edge_reward": 0.08078434690833092,
      "rewards/forensics_tiou_reward": 0.5978136658668518,
      "step": 1185
    },
    {
      "completion_length": 28.6875,
      "epoch": 3.6390168970814134,
      "grad_norm": 8.981417655944824,
      "kl": 0.6044921875,
      "learning_rate": 8.846153846153847e-08,
      "loss": 0.0219,
      "reward": 0.74368616938591,
      "reward_std": 0.057265305891633034,
      "reward_std_all": 0.06052161566913128,
      "rewards-max/forensics_iou_reward": 0.9896959364414215,
      "rewards-max/forensics_strict_edge_reward": 0.5716506689786911,
      "rewards-max/forensics_tiou_reward": 0.9841425120830536,
      "rewards/forensics_iou_reward": 0.6763516068458557,
      "rewards/forensics_strict_edge_reward": 0.08706767484545708,
      "rewards/forensics_tiou_reward": 0.6414717137813568,
      "step": 1186
    },
    {
      "completion_length": 27.0,
      "epoch": 3.6420890937019967,
      "grad_norm": 6.466035842895508,
      "kl": 0.400390625,
      "learning_rate": 8.769230769230769e-08,
      "loss": 0.0439,
      "reward": 0.7905909419059753,
      "reward_std": 0.016807834850624204,
      "reward_std_all": 0.017962872050702572,
      "rewards-max/forensics_iou_reward": 0.9600293040275574,
      "rewards-max/forensics_strict_edge_reward": 0.5087610185146332,
      "rewards-max/forensics_tiou_reward": 0.9472560882568359,
      "rewards/forensics_iou_reward": 0.7332083582878113,
      "rewards/forensics_strict_edge_reward": 0.08771160617470741,
      "rewards/forensics_tiou_reward": 0.6757290959358215,
      "step": 1187
    },
    {
      "completion_length": 28.890625,
      "epoch": 3.6451612903225805,
      "grad_norm": 7.321418762207031,
      "kl": 0.4423828125,
      "learning_rate": 8.692307692307692e-08,
      "loss": 0.0576,
      "reward": 0.718064695596695,
      "reward_std": 0.06104767136275768,
      "reward_std_all": 0.07567376270890236,
      "rewards-max/forensics_iou_reward": 0.9973849356174469,
      "rewards-max/forensics_strict_edge_reward": 0.9232408702373505,
      "rewards-max/forensics_tiou_reward": 0.996575117111206,
      "rewards/forensics_iou_reward": 0.6718636751174927,
      "rewards/forensics_strict_edge_reward": 0.08696508221328259,
      "rewards/forensics_tiou_reward": 0.6264233887195587,
      "step": 1188
    },
    {
      "completion_length": 22.828125,
      "epoch": 3.6482334869431643,
      "grad_norm": 7.87805700302124,
      "kl": 0.4384765625,
      "learning_rate": 8.615384615384615e-08,
      "loss": -0.0199,
      "reward": 0.7991810142993927,
      "reward_std": 0.011874030926264822,
      "reward_std_all": 0.01250491791870445,
      "rewards-max/forensics_iou_reward": 0.9822018146514893,
      "rewards-max/forensics_strict_edge_reward": 0.401033416390419,
      "rewards-max/forensics_tiou_reward": 0.9756213426589966,
      "rewards/forensics_iou_reward": 0.7588871121406555,
      "rewards/forensics_strict_edge_reward": 0.043730854988098145,
      "rewards/forensics_tiou_reward": 0.7116553485393524,
      "step": 1189
    },
    {
      "completion_length": 22.453125,
      "epoch": 3.651305683563748,
      "grad_norm": 8.49077320098877,
      "kl": 0.46484375,
      "learning_rate": 8.538461538461537e-08,
      "loss": -0.0146,
      "reward": 0.7819002866744995,
      "reward_std": 0.02010939223691821,
      "reward_std_all": 0.02526191994547844,
      "rewards-max/forensics_iou_reward": 0.9731045067310333,
      "rewards-max/forensics_strict_edge_reward": 0.6706924736499786,
      "rewards-max/forensics_tiou_reward": 0.9672761261463165,
      "rewards/forensics_iou_reward": 0.6781282126903534,
      "rewards/forensics_strict_edge_reward": 0.06220058724284172,
      "rewards/forensics_tiou_reward": 0.6235299408435822,
      "step": 1190
    },
    {
      "completion_length": 22.421875,
      "epoch": 3.654377880184332,
      "grad_norm": 6.990849494934082,
      "kl": 0.484375,
      "learning_rate": 8.461538461538461e-08,
      "loss": 0.014,
      "reward": 0.7939363121986389,
      "reward_std": 0.022327749989926815,
      "reward_std_all": 0.021005162969231606,
      "rewards-max/forensics_iou_reward": 0.9900915324687958,
      "rewards-max/forensics_strict_edge_reward": 0.5702531039714813,
      "rewards-max/forensics_tiou_reward": 0.9850035607814789,
      "rewards/forensics_iou_reward": 0.7832557260990143,
      "rewards/forensics_strict_edge_reward": 0.09261507168412209,
      "rewards/forensics_tiou_reward": 0.7347503900527954,
      "step": 1191
    },
    {
      "completion_length": 22.203125,
      "epoch": 3.6574500768049156,
      "grad_norm": 9.509905815124512,
      "kl": 0.4365234375,
      "learning_rate": 8.384615384615384e-08,
      "loss": -0.0224,
      "reward": 0.7755665183067322,
      "reward_std": 0.037534041330218315,
      "reward_std_all": 0.03622298128902912,
      "rewards-max/forensics_iou_reward": 0.9931674897670746,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9917069971561432,
      "rewards/forensics_iou_reward": 0.7012860774993896,
      "rewards/forensics_strict_edge_reward": 0.127749502658844,
      "rewards/forensics_tiou_reward": 0.663925439119339,
      "step": 1192
    },
    {
      "completion_length": 26.125,
      "epoch": 3.6605222734254994,
      "grad_norm": 7.910709381103516,
      "kl": 0.58203125,
      "learning_rate": 8.307692307692307e-08,
      "loss": 0.0382,
      "reward": 0.7840073108673096,
      "reward_std": 0.0381127092987299,
      "reward_std_all": 0.03513123653829098,
      "rewards-max/forensics_iou_reward": 0.9890809655189514,
      "rewards-max/forensics_strict_edge_reward": 0.5603262558579445,
      "rewards-max/forensics_tiou_reward": 0.9847725033760071,
      "rewards/forensics_iou_reward": 0.7245456874370575,
      "rewards/forensics_strict_edge_reward": 0.06651190295815468,
      "rewards/forensics_tiou_reward": 0.6944655478000641,
      "step": 1193
    },
    {
      "completion_length": 24.265625,
      "epoch": 3.6635944700460827,
      "grad_norm": 10.377342224121094,
      "kl": 0.4853515625,
      "learning_rate": 8.23076923076923e-08,
      "loss": 0.0728,
      "reward": 0.7731499075889587,
      "reward_std": 0.056841726414859295,
      "reward_std_all": 0.0510779544711113,
      "rewards-max/forensics_iou_reward": 0.9834072291851044,
      "rewards-max/forensics_strict_edge_reward": 0.5422053933143616,
      "rewards-max/forensics_tiou_reward": 0.977730393409729,
      "rewards/forensics_iou_reward": 0.6862276494503021,
      "rewards/forensics_strict_edge_reward": 0.07134545408189297,
      "rewards/forensics_tiou_reward": 0.6407122910022736,
      "step": 1194
    },
    {
      "completion_length": 23.015625,
      "epoch": 3.6666666666666665,
      "grad_norm": 7.397069931030273,
      "kl": 0.3857421875,
      "learning_rate": 8.153846153846153e-08,
      "loss": 0.0553,
      "reward": 0.7491303086280823,
      "reward_std": 0.056747231632471085,
      "reward_std_all": 0.05744069814682007,
      "rewards-max/forensics_iou_reward": 0.9641986191272736,
      "rewards-max/forensics_strict_edge_reward": 0.424427792429924,
      "rewards-max/forensics_tiou_reward": 0.9431972503662109,
      "rewards/forensics_iou_reward": 0.6482219696044922,
      "rewards/forensics_strict_edge_reward": 0.07258268818259239,
      "rewards/forensics_tiou_reward": 0.5989575684070587,
      "step": 1195
    },
    {
      "completion_length": 25.40625,
      "epoch": 3.6697388632872503,
      "grad_norm": 6.780534744262695,
      "kl": 0.396484375,
      "learning_rate": 8.076923076923077e-08,
      "loss": 0.021,
      "reward": 0.7508625388145447,
      "reward_std": 0.0954190194606781,
      "reward_std_all": 0.0841558612883091,
      "rewards-max/forensics_iou_reward": 0.9850271344184875,
      "rewards-max/forensics_strict_edge_reward": 0.7373577654361725,
      "rewards-max/forensics_tiou_reward": 0.9825437963008881,
      "rewards/forensics_iou_reward": 0.6531355679035187,
      "rewards/forensics_strict_edge_reward": 0.088094811886549,
      "rewards/forensics_tiou_reward": 0.6172506213188171,
      "step": 1196
    },
    {
      "completion_length": 28.703125,
      "epoch": 3.672811059907834,
      "grad_norm": 6.245475769042969,
      "kl": 0.455078125,
      "learning_rate": 8e-08,
      "loss": 0.0384,
      "reward": 0.7708496153354645,
      "reward_std": 0.05735739879310131,
      "reward_std_all": 0.05212130956351757,
      "rewards-max/forensics_iou_reward": 0.983649492263794,
      "rewards-max/forensics_strict_edge_reward": 0.6599850654602051,
      "rewards-max/forensics_tiou_reward": 0.9779987335205078,
      "rewards/forensics_iou_reward": 0.7037177979946136,
      "rewards/forensics_strict_edge_reward": 0.08131053671240807,
      "rewards/forensics_tiou_reward": 0.665470153093338,
      "step": 1197
    },
    {
      "completion_length": 24.453125,
      "epoch": 3.675883256528418,
      "grad_norm": 8.04312801361084,
      "kl": 0.505859375,
      "learning_rate": 7.923076923076923e-08,
      "loss": -0.0198,
      "reward": 0.7110533118247986,
      "reward_std": 0.062190817669034004,
      "reward_std_all": 0.08053384721279144,
      "rewards-max/forensics_iou_reward": 0.9593962728977203,
      "rewards-max/forensics_strict_edge_reward": 0.5154729187488556,
      "rewards-max/forensics_tiou_reward": 0.9499149024486542,
      "rewards/forensics_iou_reward": 0.6211718618869781,
      "rewards/forensics_strict_edge_reward": 0.0761859305202961,
      "rewards/forensics_tiou_reward": 0.5822030305862427,
      "step": 1198
    },
    {
      "completion_length": 27.09375,
      "epoch": 3.6789554531490016,
      "grad_norm": 7.79232931137085,
      "kl": 0.5458984375,
      "learning_rate": 7.846153846153846e-08,
      "loss": -0.0081,
      "reward": 0.7690949440002441,
      "reward_std": 0.06471827626228333,
      "reward_std_all": 0.057811446487903595,
      "rewards-max/forensics_iou_reward": 0.9775857031345367,
      "rewards-max/forensics_strict_edge_reward": 0.5408801138401031,
      "rewards-max/forensics_tiou_reward": 0.9711766541004181,
      "rewards/forensics_iou_reward": 0.7371327579021454,
      "rewards/forensics_strict_edge_reward": 0.09361907094717026,
      "rewards/forensics_tiou_reward": 0.7005361616611481,
      "step": 1199
    },
    {
      "completion_length": 22.265625,
      "epoch": 3.6820276497695854,
      "grad_norm": 9.371967315673828,
      "kl": 0.642578125,
      "learning_rate": 7.769230769230768e-08,
      "loss": 0.1004,
      "reward": 0.7482981383800507,
      "reward_std": 0.06894212029874325,
      "reward_std_all": 0.06648287549614906,
      "rewards-max/forensics_iou_reward": 0.9955165088176727,
      "rewards-max/forensics_strict_edge_reward": 0.7331632375717163,
      "rewards-max/forensics_tiou_reward": 0.9941051602363586,
      "rewards/forensics_iou_reward": 0.6803631782531738,
      "rewards/forensics_strict_edge_reward": 0.12402403354644775,
      "rewards/forensics_tiou_reward": 0.6326343417167664,
      "step": 1200
    },
    {
      "completion_length": 24.234375,
      "epoch": 3.6850998463901687,
      "grad_norm": 7.013523578643799,
      "kl": 0.580078125,
      "learning_rate": 7.692307692307692e-08,
      "loss": 0.1497,
      "reward": 0.7717065215110779,
      "reward_std": 0.06055422127246857,
      "reward_std_all": 0.05395332118496299,
      "rewards-max/forensics_iou_reward": 0.9853898286819458,
      "rewards-max/forensics_strict_edge_reward": 0.661531001329422,
      "rewards-max/forensics_tiou_reward": 0.9825833737850189,
      "rewards/forensics_iou_reward": 0.7159932255744934,
      "rewards/forensics_strict_edge_reward": 0.09817219898104668,
      "rewards/forensics_tiou_reward": 0.6717409491539001,
      "step": 1201
    },
    {
      "completion_length": 22.765625,
      "epoch": 3.688172043010753,
      "grad_norm": 7.56277322769165,
      "kl": 0.529296875,
      "learning_rate": 7.615384615384615e-08,
      "loss": 0.0692,
      "reward": 0.7685839533805847,
      "reward_std": 0.05980696715414524,
      "reward_std_all": 0.05431240238249302,
      "rewards-max/forensics_iou_reward": 0.9788296520709991,
      "rewards-max/forensics_strict_edge_reward": 0.7033758163452148,
      "rewards-max/forensics_tiou_reward": 0.9710254967212677,
      "rewards/forensics_iou_reward": 0.6748419404029846,
      "rewards/forensics_strict_edge_reward": 0.0825746487826109,
      "rewards/forensics_tiou_reward": 0.6259800791740417,
      "step": 1202
    },
    {
      "completion_length": 26.15625,
      "epoch": 3.6912442396313363,
      "grad_norm": 8.174914360046387,
      "kl": 0.6328125,
      "learning_rate": 7.538461538461538e-08,
      "loss": 0.059,
      "reward": 0.7538063526153564,
      "reward_std": 0.04664459079504013,
      "reward_std_all": 0.048470595851540565,
      "rewards-max/forensics_iou_reward": 0.975085973739624,
      "rewards-max/forensics_strict_edge_reward": 0.6660739481449127,
      "rewards-max/forensics_tiou_reward": 0.9679714143276215,
      "rewards/forensics_iou_reward": 0.6609161496162415,
      "rewards/forensics_strict_edge_reward": 0.07010985910892487,
      "rewards/forensics_tiou_reward": 0.622758537530899,
      "step": 1203
    },
    {
      "completion_length": 24.40625,
      "epoch": 3.69431643625192,
      "grad_norm": 6.846460819244385,
      "kl": 0.3779296875,
      "learning_rate": 7.46153846153846e-08,
      "loss": 0.1808,
      "reward": 0.7313268780708313,
      "reward_std": 0.040936847217381,
      "reward_std_all": 0.05997008457779884,
      "rewards-max/forensics_iou_reward": 0.9875469207763672,
      "rewards-max/forensics_strict_edge_reward": 0.7253158986568451,
      "rewards-max/forensics_tiou_reward": 0.9857767522335052,
      "rewards/forensics_iou_reward": 0.6283256411552429,
      "rewards/forensics_strict_edge_reward": 0.061443500220775604,
      "rewards/forensics_tiou_reward": 0.5623424351215363,
      "step": 1204
    },
    {
      "completion_length": 22.09375,
      "epoch": 3.697388632872504,
      "grad_norm": 8.921612739562988,
      "kl": 0.583984375,
      "learning_rate": 7.384615384615384e-08,
      "loss": 0.1187,
      "reward": 0.7709879279136658,
      "reward_std": 0.049853174947202206,
      "reward_std_all": 0.04650441836565733,
      "rewards-max/forensics_iou_reward": 0.9916897416114807,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9902153015136719,
      "rewards/forensics_iou_reward": 0.7759255766868591,
      "rewards/forensics_strict_edge_reward": 0.109074916690588,
      "rewards/forensics_tiou_reward": 0.7287041246891022,
      "step": 1205
    },
    {
      "completion_length": 26.734375,
      "epoch": 3.7004608294930876,
      "grad_norm": 10.103010177612305,
      "kl": 0.419921875,
      "learning_rate": 7.307692307692308e-08,
      "loss": 0.0131,
      "reward": 0.8020296394824982,
      "reward_std": 0.007051538210362196,
      "reward_std_all": 0.008333152625709772,
      "rewards-max/forensics_iou_reward": 0.9863293170928955,
      "rewards-max/forensics_strict_edge_reward": 0.6093219816684723,
      "rewards-max/forensics_tiou_reward": 0.9756523072719574,
      "rewards/forensics_iou_reward": 0.7523626685142517,
      "rewards/forensics_strict_edge_reward": 0.05707604065537453,
      "rewards/forensics_tiou_reward": 0.7029235661029816,
      "step": 1206
    },
    {
      "completion_length": 26.28125,
      "epoch": 3.7035330261136714,
      "grad_norm": 10.227730751037598,
      "kl": 0.646484375,
      "learning_rate": 7.230769230769231e-08,
      "loss": 0.07,
      "reward": 0.7652566134929657,
      "reward_std": 0.04520853655412793,
      "reward_std_all": 0.04443807993084192,
      "rewards-max/forensics_iou_reward": 1.0,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_iou_reward": 0.7164926528930664,
      "rewards/forensics_strict_edge_reward": 0.13269872218370438,
      "rewards/forensics_tiou_reward": 0.6734485626220703,
      "step": 1207
    },
    {
      "completion_length": 28.296875,
      "epoch": 3.7066052227342547,
      "grad_norm": 8.227707862854004,
      "kl": 0.44140625,
      "learning_rate": 7.153846153846154e-08,
      "loss": -0.0091,
      "reward": 0.740957647562027,
      "reward_std": 0.03942619008012116,
      "reward_std_all": 0.056217608973383904,
      "rewards-max/forensics_iou_reward": 0.9824278056621552,
      "rewards-max/forensics_strict_edge_reward": 0.6234369277954102,
      "rewards-max/forensics_tiou_reward": 0.9795488119125366,
      "rewards/forensics_iou_reward": 0.7011848390102386,
      "rewards/forensics_strict_edge_reward": 0.0635660421103239,
      "rewards/forensics_tiou_reward": 0.6565888822078705,
      "step": 1208
    },
    {
      "completion_length": 26.359375,
      "epoch": 3.709677419354839,
      "grad_norm": 10.821630477905273,
      "kl": 1.037109375,
      "learning_rate": 7.076923076923076e-08,
      "loss": 0.1254,
      "reward": 0.7634647786617279,
      "reward_std": 0.046654338017106056,
      "reward_std_all": 0.04568750038743019,
      "rewards-max/forensics_iou_reward": 1.0,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_iou_reward": 0.6697306036949158,
      "rewards/forensics_strict_edge_reward": 0.10208340361714363,
      "rewards/forensics_tiou_reward": 0.6269328594207764,
      "step": 1209
    },
    {
      "completion_length": 27.125,
      "epoch": 3.7127496159754223,
      "grad_norm": 7.830411434173584,
      "kl": 0.3408203125,
      "learning_rate": 7e-08,
      "loss": -0.0767,
      "reward": 0.7669989764690399,
      "reward_std": 0.04848597198724747,
      "reward_std_all": 0.04637716244906187,
      "rewards-max/forensics_iou_reward": 0.9836479723453522,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9779529273509979,
      "rewards/forensics_iou_reward": 0.68788081407547,
      "rewards/forensics_strict_edge_reward": 0.07767381519079208,
      "rewards/forensics_tiou_reward": 0.6456928551197052,
      "step": 1210
    },
    {
      "completion_length": 25.75,
      "epoch": 3.715821812596006,
      "grad_norm": 7.495970726013184,
      "kl": 0.4267578125,
      "learning_rate": 6.923076923076923e-08,
      "loss": -0.032,
      "reward": 0.7856367826461792,
      "reward_std": 0.02780949790030718,
      "reward_std_all": 0.02698243409395218,
      "rewards-max/forensics_iou_reward": 0.9855433404445648,
      "rewards-max/forensics_strict_edge_reward": 0.5190997421741486,
      "rewards-max/forensics_tiou_reward": 0.9735495448112488,
      "rewards/forensics_iou_reward": 0.7789583504199982,
      "rewards/forensics_strict_edge_reward": 0.07222607359290123,
      "rewards/forensics_tiou_reward": 0.7299333810806274,
      "step": 1211
    },
    {
      "completion_length": 31.515625,
      "epoch": 3.71889400921659,
      "grad_norm": 5.699143409729004,
      "kl": 0.494140625,
      "learning_rate": 6.846153846153846e-08,
      "loss": 0.0481,
      "reward": 0.7563722729682922,
      "reward_std": 0.06909128837287426,
      "reward_std_all": 0.06404301524162292,
      "rewards-max/forensics_iou_reward": 0.9870090186595917,
      "rewards-max/forensics_strict_edge_reward": 0.6837711930274963,
      "rewards-max/forensics_tiou_reward": 0.9834642708301544,
      "rewards/forensics_iou_reward": 0.6763970851898193,
      "rewards/forensics_strict_edge_reward": 0.051253922283649445,
      "rewards/forensics_tiou_reward": 0.6412186324596405,
      "step": 1212
    },
    {
      "completion_length": 28.203125,
      "epoch": 3.7219662058371736,
      "grad_norm": 8.07038688659668,
      "kl": 0.3515625,
      "learning_rate": 6.769230769230769e-08,
      "loss": -0.0799,
      "reward": 0.7411088347434998,
      "reward_std": 0.05458119697868824,
      "reward_std_all": 0.058819087222218513,
      "rewards-max/forensics_iou_reward": 0.9861119389533997,
      "rewards-max/forensics_strict_edge_reward": 0.4574581980705261,
      "rewards-max/forensics_tiou_reward": 0.9792424142360687,
      "rewards/forensics_iou_reward": 0.7171503603458405,
      "rewards/forensics_strict_edge_reward": 0.06893611326813698,
      "rewards/forensics_tiou_reward": 0.6799814403057098,
      "step": 1213
    },
    {
      "completion_length": 26.140625,
      "epoch": 3.7250384024577574,
      "grad_norm": 7.959125995635986,
      "kl": 0.369140625,
      "learning_rate": 6.692307692307691e-08,
      "loss": -0.0374,
      "reward": 0.7304100692272186,
      "reward_std": 0.03329411894083023,
      "reward_std_all": 0.055620032362639904,
      "rewards-max/forensics_iou_reward": 0.9915430247783661,
      "rewards-max/forensics_strict_edge_reward": 0.5397044718265533,
      "rewards-max/forensics_tiou_reward": 0.9886043071746826,
      "rewards/forensics_iou_reward": 0.6418476998806,
      "rewards/forensics_strict_edge_reward": 0.09177186340093613,
      "rewards/forensics_tiou_reward": 0.6001218855381012,
      "step": 1214
    },
    {
      "completion_length": 25.59375,
      "epoch": 3.7281105990783407,
      "grad_norm": 10.213120460510254,
      "kl": 0.51171875,
      "learning_rate": 6.615384615384615e-08,
      "loss": 0.0874,
      "reward": 0.7427855134010315,
      "reward_std": 0.07803923822939396,
      "reward_std_all": 0.07367841899394989,
      "rewards-max/forensics_iou_reward": 0.987726241350174,
      "rewards-max/forensics_strict_edge_reward": 0.7299392521381378,
      "rewards-max/forensics_tiou_reward": 0.9831154048442841,
      "rewards/forensics_iou_reward": 0.7175578773021698,
      "rewards/forensics_strict_edge_reward": 0.10746435448527336,
      "rewards/forensics_tiou_reward": 0.6766146123409271,
      "step": 1215
    },
    {
      "completion_length": 23.6875,
      "epoch": 3.731182795698925,
      "grad_norm": 8.135313987731934,
      "kl": 0.4736328125,
      "learning_rate": 6.538461538461538e-08,
      "loss": 0.053,
      "reward": 0.7445524036884308,
      "reward_std": 0.0497029609978199,
      "reward_std_all": 0.05833658576011658,
      "rewards-max/forensics_iou_reward": 0.9914222955703735,
      "rewards-max/forensics_strict_edge_reward": 0.6934256851673126,
      "rewards-max/forensics_tiou_reward": 0.9872600436210632,
      "rewards/forensics_iou_reward": 0.6756462752819061,
      "rewards/forensics_strict_edge_reward": 0.09687073156237602,
      "rewards/forensics_tiou_reward": 0.63813117146492,
      "step": 1216
    },
    {
      "completion_length": 29.421875,
      "epoch": 3.7342549923195083,
      "grad_norm": 7.258788108825684,
      "kl": 0.4462890625,
      "learning_rate": 6.461538461538462e-08,
      "loss": 0.0292,
      "reward": 0.7374012172222137,
      "reward_std": 0.03935924731194973,
      "reward_std_all": 0.05913357809185982,
      "rewards-max/forensics_iou_reward": 0.9398348927497864,
      "rewards-max/forensics_strict_edge_reward": 0.2217167615890503,
      "rewards-max/forensics_tiou_reward": 0.9271933436393738,
      "rewards/forensics_iou_reward": 0.6542287170886993,
      "rewards/forensics_strict_edge_reward": 0.026916603557765484,
      "rewards/forensics_tiou_reward": 0.6048651337623596,
      "step": 1217
    },
    {
      "completion_length": 23.546875,
      "epoch": 3.737327188940092,
      "grad_norm": 10.271416664123535,
      "kl": 0.4365234375,
      "learning_rate": 6.384615384615385e-08,
      "loss": 0.0062,
      "reward": 0.7983713746070862,
      "reward_std": 0.013738906709477305,
      "reward_std_all": 0.013810974429361522,
      "rewards-max/forensics_iou_reward": 0.9789610207080841,
      "rewards-max/forensics_strict_edge_reward": 0.7029964029788971,
      "rewards-max/forensics_tiou_reward": 0.97340327501297,
      "rewards/forensics_iou_reward": 0.7607623636722565,
      "rewards/forensics_strict_edge_reward": 0.0911457221955061,
      "rewards/forensics_tiou_reward": 0.7220290899276733,
      "step": 1218
    },
    {
      "completion_length": 23.71875,
      "epoch": 3.740399385560676,
      "grad_norm": 8.652264595031738,
      "kl": 0.32421875,
      "learning_rate": 6.307692307692307e-08,
      "loss": -0.063,
      "reward": 0.7810418605804443,
      "reward_std": 0.038270230405032635,
      "reward_std_all": 0.03546732757240534,
      "rewards-max/forensics_iou_reward": 0.989350438117981,
      "rewards-max/forensics_strict_edge_reward": 0.6546298861503601,
      "rewards-max/forensics_tiou_reward": 0.9870820641517639,
      "rewards/forensics_iou_reward": 0.7031590342521667,
      "rewards/forensics_strict_edge_reward": 0.10927844792604446,
      "rewards/forensics_tiou_reward": 0.6618118584156036,
      "step": 1219
    },
    {
      "completion_length": 22.234375,
      "epoch": 3.7434715821812596,
      "grad_norm": 8.568329811096191,
      "kl": 0.50390625,
      "learning_rate": 6.23076923076923e-08,
      "loss": -0.038,
      "reward": 0.7303427755832672,
      "reward_std": 0.0512523353099823,
      "reward_std_all": 0.06611401215195656,
      "rewards-max/forensics_iou_reward": 0.9748882055282593,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9684414565563202,
      "rewards/forensics_iou_reward": 0.651930183172226,
      "rewards/forensics_strict_edge_reward": 0.09218992292881012,
      "rewards/forensics_tiou_reward": 0.6083322167396545,
      "step": 1220
    },
    {
      "completion_length": 27.0625,
      "epoch": 3.7465437788018434,
      "grad_norm": 7.769948482513428,
      "kl": 0.384765625,
      "learning_rate": 6.153846153846154e-08,
      "loss": -0.0742,
      "reward": 0.7420053780078888,
      "reward_std": 0.026786171656567603,
      "reward_std_all": 0.04350569751113653,
      "rewards-max/forensics_iou_reward": 0.9957493543624878,
      "rewards-max/forensics_strict_edge_reward": 0.8617841005325317,
      "rewards-max/forensics_tiou_reward": 0.9932664930820465,
      "rewards/forensics_iou_reward": 0.6982798278331757,
      "rewards/forensics_strict_edge_reward": 0.10271776840090752,
      "rewards/forensics_tiou_reward": 0.6624684929847717,
      "step": 1221
    },
    {
      "completion_length": 23.6875,
      "epoch": 3.749615975422427,
      "grad_norm": 9.379670143127441,
      "kl": 0.400390625,
      "learning_rate": 6.076923076923077e-08,
      "loss": -0.0446,
      "reward": 0.7739603817462921,
      "reward_std": 0.029099556151777506,
      "reward_std_all": 0.03137672459706664,
      "rewards-max/forensics_iou_reward": 0.993099182844162,
      "rewards-max/forensics_strict_edge_reward": 0.6824378371238708,
      "rewards-max/forensics_tiou_reward": 0.9908031821250916,
      "rewards/forensics_iou_reward": 0.7106501758098602,
      "rewards/forensics_strict_edge_reward": 0.054465487599372864,
      "rewards/forensics_tiou_reward": 0.6682343482971191,
      "step": 1222
    },
    {
      "completion_length": 23.59375,
      "epoch": 3.752688172043011,
      "grad_norm": 9.25857925415039,
      "kl": 0.466796875,
      "learning_rate": 6e-08,
      "loss": -0.0472,
      "reward": 0.799436628818512,
      "reward_std": 0.01252474298235029,
      "reward_std_all": 0.012505853082984686,
      "rewards-max/forensics_iou_reward": 0.9867661893367767,
      "rewards-max/forensics_strict_edge_reward": 0.7029964029788971,
      "rewards-max/forensics_tiou_reward": 0.9831684827804565,
      "rewards/forensics_iou_reward": 0.7321898341178894,
      "rewards/forensics_strict_edge_reward": 0.13831373676657677,
      "rewards/forensics_tiou_reward": 0.6898584961891174,
      "step": 1223
    },
    {
      "completion_length": 24.390625,
      "epoch": 3.7557603686635943,
      "grad_norm": 6.880444526672363,
      "kl": 0.4169921875,
      "learning_rate": 5.923076923076923e-08,
      "loss": 0.0568,
      "reward": 0.7842468023300171,
      "reward_std": 0.033430224284529686,
      "reward_std_all": 0.03125045821070671,
      "rewards-max/forensics_iou_reward": 0.993858128786087,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9911783039569855,
      "rewards/forensics_iou_reward": 0.7305087447166443,
      "rewards/forensics_strict_edge_reward": 0.1425328627228737,
      "rewards/forensics_tiou_reward": 0.6894195675849915,
      "step": 1224
    },
    {
      "completion_length": 25.296875,
      "epoch": 3.758832565284178,
      "grad_norm": 7.456821441650391,
      "kl": 0.4287109375,
      "learning_rate": 5.8461538461538455e-08,
      "loss": 0.0603,
      "reward": 0.7480155229568481,
      "reward_std": 0.061946989968419075,
      "reward_std_all": 0.062385598197579384,
      "rewards-max/forensics_iou_reward": 0.9713936150074005,
      "rewards-max/forensics_strict_edge_reward": 0.4419635832309723,
      "rewards-max/forensics_tiou_reward": 0.9496957659721375,
      "rewards/forensics_iou_reward": 0.6983695924282074,
      "rewards/forensics_strict_edge_reward": 0.0588364377617836,
      "rewards/forensics_tiou_reward": 0.6548425257205963,
      "step": 1225
    },
    {
      "completion_length": 25.984375,
      "epoch": 3.761904761904762,
      "grad_norm": 7.578303813934326,
      "kl": 0.359375,
      "learning_rate": 5.7692307692307695e-08,
      "loss": 0.0408,
      "reward": 0.7952819764614105,
      "reward_std": 0.002976364456117153,
      "reward_std_all": 0.008741655154153705,
      "rewards-max/forensics_iou_reward": 0.9892043471336365,
      "rewards-max/forensics_strict_edge_reward": 0.6515887677669525,
      "rewards-max/forensics_tiou_reward": 0.9857140779495239,
      "rewards/forensics_iou_reward": 0.7082273364067078,
      "rewards/forensics_strict_edge_reward": 0.08877100422978401,
      "rewards/forensics_tiou_reward": 0.6618707180023193,
      "step": 1226
    },
    {
      "completion_length": 25.984375,
      "epoch": 3.7649769585253456,
      "grad_norm": 6.5124430656433105,
      "kl": 0.373046875,
      "learning_rate": 5.692307692307692e-08,
      "loss": 0.1182,
      "reward": 0.7230077981948853,
      "reward_std": 0.09323636355111375,
      "reward_std_all": 0.09202855452895164,
      "rewards-max/forensics_iou_reward": 0.9759032130241394,
      "rewards-max/forensics_strict_edge_reward": 0.48703832924366,
      "rewards-max/forensics_tiou_reward": 0.9656966626644135,
      "rewards/forensics_iou_reward": 0.6240063905715942,
      "rewards/forensics_strict_edge_reward": 0.08038639836013317,
      "rewards/forensics_tiou_reward": 0.571930468082428,
      "step": 1227
    },
    {
      "completion_length": 24.703125,
      "epoch": 3.7680491551459294,
      "grad_norm": 9.403243064880371,
      "kl": 0.5322265625,
      "learning_rate": 5.615384615384615e-08,
      "loss": -0.0164,
      "reward": 0.7443765997886658,
      "reward_std": 0.02671147044748068,
      "reward_std_all": 0.048059738241136074,
      "rewards-max/forensics_iou_reward": 0.987280011177063,
      "rewards-max/forensics_strict_edge_reward": 0.5269662886857986,
      "rewards-max/forensics_tiou_reward": 0.979849100112915,
      "rewards/forensics_iou_reward": 0.7122468948364258,
      "rewards/forensics_strict_edge_reward": 0.07305393740534782,
      "rewards/forensics_tiou_reward": 0.6537478566169739,
      "step": 1228
    },
    {
      "completion_length": 26.4375,
      "epoch": 3.771121351766513,
      "grad_norm": 7.0354485511779785,
      "kl": 0.3828125,
      "learning_rate": 5.538461538461538e-08,
      "loss": 0.1503,
      "reward": 0.7569281160831451,
      "reward_std": 0.07563486602157354,
      "reward_std_all": 0.06824102438986301,
      "rewards-max/forensics_iou_reward": 0.9741411805152893,
      "rewards-max/forensics_strict_edge_reward": 0.5043487846851349,
      "rewards-max/forensics_tiou_reward": 0.9648866951465607,
      "rewards/forensics_iou_reward": 0.6807650625705719,
      "rewards/forensics_strict_edge_reward": 0.06178411468863487,
      "rewards/forensics_tiou_reward": 0.6337049305438995,
      "step": 1229
    },
    {
      "completion_length": 22.203125,
      "epoch": 3.774193548387097,
      "grad_norm": 9.027506828308105,
      "kl": 0.4150390625,
      "learning_rate": 5.461538461538461e-08,
      "loss": -0.0664,
      "reward": 0.7633076310157776,
      "reward_std": 0.03352766763418913,
      "reward_std_all": 0.039272772148251534,
      "rewards-max/forensics_iou_reward": 0.9934675395488739,
      "rewards-max/forensics_strict_edge_reward": 0.7789089381694794,
      "rewards-max/forensics_tiou_reward": 0.9907897114753723,
      "rewards/forensics_iou_reward": 0.706658661365509,
      "rewards/forensics_strict_edge_reward": 0.09831000864505768,
      "rewards/forensics_tiou_reward": 0.6719037592411041,
      "step": 1230
    },
    {
      "completion_length": 27.40625,
      "epoch": 3.7772657450076803,
      "grad_norm": 6.44115686416626,
      "kl": 0.5244140625,
      "learning_rate": 5.384615384615385e-08,
      "loss": 0.0965,
      "reward": 0.7939364314079285,
      "reward_std": 0.0172613556496799,
      "reward_std_all": 0.017480879556387663,
      "rewards-max/forensics_iou_reward": 0.9940384924411774,
      "rewards-max/forensics_strict_edge_reward": 0.6065583229064941,
      "rewards-max/forensics_tiou_reward": 0.9927070140838623,
      "rewards/forensics_iou_reward": 0.7117989659309387,
      "rewards/forensics_strict_edge_reward": 0.06058591231703758,
      "rewards/forensics_tiou_reward": 0.6647423505783081,
      "step": 1231
    },
    {
      "completion_length": 23.296875,
      "epoch": 3.780337941628264,
      "grad_norm": 6.885991096496582,
      "kl": 0.3759765625,
      "learning_rate": 5.307692307692308e-08,
      "loss": -0.0735,
      "reward": 0.719552218914032,
      "reward_std": 0.06998690479667857,
      "reward_std_all": 0.08512369357049465,
      "rewards-max/forensics_iou_reward": 0.9431107044219971,
      "rewards-max/forensics_strict_edge_reward": 0.4369872733950615,
      "rewards-max/forensics_tiou_reward": 0.9102686941623688,
      "rewards/forensics_iou_reward": 0.6383413672447205,
      "rewards/forensics_strict_edge_reward": 0.06549964752048254,
      "rewards/forensics_tiou_reward": 0.5983628034591675,
      "step": 1232
    },
    {
      "completion_length": 29.1875,
      "epoch": 3.783410138248848,
      "grad_norm": 7.472944259643555,
      "kl": 0.689453125,
      "learning_rate": 5.2307692307692304e-08,
      "loss": -0.0236,
      "reward": 0.8032988011837006,
      "reward_std": 0.004292209399864078,
      "reward_std_all": 0.006000764085911214,
      "rewards-max/forensics_iou_reward": 0.9718080461025238,
      "rewards-max/forensics_strict_edge_reward": 0.4466862380504608,
      "rewards-max/forensics_tiou_reward": 0.9633758366107941,
      "rewards/forensics_iou_reward": 0.7210010886192322,
      "rewards/forensics_strict_edge_reward": 0.07492118328809738,
      "rewards/forensics_tiou_reward": 0.6638201177120209,
      "step": 1233
    },
    {
      "completion_length": 23.546875,
      "epoch": 3.7864823348694316,
      "grad_norm": 8.339098930358887,
      "kl": 0.400390625,
      "learning_rate": 5.153846153846154e-08,
      "loss": -0.0424,
      "reward": 0.7424915730953217,
      "reward_std": 0.038816249929368496,
      "reward_std_all": 0.05441976897418499,
      "rewards-max/forensics_iou_reward": 0.9904661178588867,
      "rewards-max/forensics_strict_edge_reward": 0.5686750710010529,
      "rewards-max/forensics_tiou_reward": 0.9849963188171387,
      "rewards/forensics_iou_reward": 0.6707884073257446,
      "rewards/forensics_strict_edge_reward": 0.08635738492012024,
      "rewards/forensics_tiou_reward": 0.6322676539421082,
      "step": 1234
    },
    {
      "completion_length": 27.765625,
      "epoch": 3.7895545314900154,
      "grad_norm": 7.094878673553467,
      "kl": 0.48828125,
      "learning_rate": 5.0769230769230764e-08,
      "loss": 0.0518,
      "reward": 0.7367727756500244,
      "reward_std": 0.04296353738754988,
      "reward_std_all": 0.061443161219358444,
      "rewards-max/forensics_iou_reward": 0.98210808634758,
      "rewards-max/forensics_strict_edge_reward": 0.6004842221736908,
      "rewards-max/forensics_tiou_reward": 0.9759362936019897,
      "rewards/forensics_iou_reward": 0.7049349546432495,
      "rewards/forensics_strict_edge_reward": 0.06594127044081688,
      "rewards/forensics_tiou_reward": 0.664282888174057,
      "step": 1235
    },
    {
      "completion_length": 26.078125,
      "epoch": 3.792626728110599,
      "grad_norm": 10.055532455444336,
      "kl": 0.4365234375,
      "learning_rate": 5e-08,
      "loss": 0.0388,
      "reward": 0.73286372423172,
      "reward_std": 0.05349611397832632,
      "reward_std_all": 0.0609426312148571,
      "rewards-max/forensics_iou_reward": 0.960179328918457,
      "rewards-max/forensics_strict_edge_reward": 0.45834553241729736,
      "rewards-max/forensics_tiou_reward": 0.9509620070457458,
      "rewards/forensics_iou_reward": 0.6954581439495087,
      "rewards/forensics_strict_edge_reward": 0.09000327065587044,
      "rewards/forensics_tiou_reward": 0.6561473906040192,
      "step": 1236
    },
    {
      "completion_length": 22.9375,
      "epoch": 3.795698924731183,
      "grad_norm": 8.332404136657715,
      "kl": 0.412109375,
      "learning_rate": 4.923076923076923e-08,
      "loss": 0.1021,
      "reward": 0.7614081501960754,
      "reward_std": 0.07539365254342556,
      "reward_std_all": 0.06735048070549965,
      "rewards-max/forensics_iou_reward": 0.982134222984314,
      "rewards-max/forensics_strict_edge_reward": 0.5641956627368927,
      "rewards-max/forensics_tiou_reward": 0.9740285873413086,
      "rewards/forensics_iou_reward": 0.6654229164123535,
      "rewards/forensics_strict_edge_reward": 0.09094125777482986,
      "rewards/forensics_tiou_reward": 0.6035566329956055,
      "step": 1237
    },
    {
      "completion_length": 25.15625,
      "epoch": 3.7987711213517663,
      "grad_norm": 5.3084893226623535,
      "kl": 0.537109375,
      "learning_rate": 4.846153846153846e-08,
      "loss": 0.009,
      "reward": 0.784979522228241,
      "reward_std": 0.036879544612020254,
      "reward_std_all": 0.03366420045495033,
      "rewards-max/forensics_iou_reward": 0.9851683378219604,
      "rewards-max/forensics_strict_edge_reward": 0.4882577657699585,
      "rewards-max/forensics_tiou_reward": 0.9816502332687378,
      "rewards/forensics_iou_reward": 0.7407818734645844,
      "rewards/forensics_strict_edge_reward": 0.08872827887535095,
      "rewards/forensics_tiou_reward": 0.7040762603282928,
      "step": 1238
    },
    {
      "completion_length": 28.484375,
      "epoch": 3.80184331797235,
      "grad_norm": 6.512976169586182,
      "kl": 0.3427734375,
      "learning_rate": 4.769230769230769e-08,
      "loss": -0.0148,
      "reward": 0.751527339220047,
      "reward_std": 0.05762442201375961,
      "reward_std_all": 0.05838781222701073,
      "rewards-max/forensics_iou_reward": 0.9715935289859772,
      "rewards-max/forensics_strict_edge_reward": 0.4120611250400543,
      "rewards-max/forensics_tiou_reward": 0.9641596674919128,
      "rewards/forensics_iou_reward": 0.6682238280773163,
      "rewards/forensics_strict_edge_reward": 0.04742329567670822,
      "rewards/forensics_tiou_reward": 0.6250264942646027,
      "step": 1239
    },
    {
      "completion_length": 20.78125,
      "epoch": 3.804915514592934,
      "grad_norm": 8.106745719909668,
      "kl": 0.4560546875,
      "learning_rate": 4.692307692307692e-08,
      "loss": 0.0346,
      "reward": 0.7563101649284363,
      "reward_std": 0.07443435117602348,
      "reward_std_all": 0.06780761294066906,
      "rewards-max/forensics_iou_reward": 0.995803564786911,
      "rewards-max/forensics_strict_edge_reward": 0.8326062560081482,
      "rewards-max/forensics_tiou_reward": 0.994359940290451,
      "rewards/forensics_iou_reward": 0.6792579293251038,
      "rewards/forensics_strict_edge_reward": 0.13841725885868073,
      "rewards/forensics_tiou_reward": 0.6356660723686218,
      "step": 1240
    },
    {
      "completion_length": 24.390625,
      "epoch": 3.8079877112135176,
      "grad_norm": 7.998985290527344,
      "kl": 0.52734375,
      "learning_rate": 4.615384615384615e-08,
      "loss": 0.0398,
      "reward": 0.7179443836212158,
      "reward_std": 0.05665894038975239,
      "reward_std_all": 0.07584808021783829,
      "rewards-max/forensics_iou_reward": 0.9383740127086639,
      "rewards-max/forensics_strict_edge_reward": 0.4803053140640259,
      "rewards-max/forensics_tiou_reward": 0.9237698614597321,
      "rewards/forensics_iou_reward": 0.648086667060852,
      "rewards/forensics_strict_edge_reward": 0.040663774125277996,
      "rewards/forensics_tiou_reward": 0.5984965860843658,
      "step": 1241
    },
    {
      "completion_length": 23.59375,
      "epoch": 3.8110599078341014,
      "grad_norm": 9.431788444519043,
      "kl": 0.892578125,
      "learning_rate": 4.538461538461538e-08,
      "loss": 0.0112,
      "reward": 0.7621477842330933,
      "reward_std": 0.03080986812710762,
      "reward_std_all": 0.036770597100257874,
      "rewards-max/forensics_iou_reward": 0.954829603433609,
      "rewards-max/forensics_strict_edge_reward": 0.5433452129364014,
      "rewards-max/forensics_tiou_reward": 0.9443707764148712,
      "rewards/forensics_iou_reward": 0.6723513603210449,
      "rewards/forensics_strict_edge_reward": 0.0788927748799324,
      "rewards/forensics_tiou_reward": 0.6240044832229614,
      "step": 1242
    },
    {
      "completion_length": 25.609375,
      "epoch": 3.814132104454685,
      "grad_norm": 25.818925857543945,
      "kl": 1.875,
      "learning_rate": 4.461538461538461e-08,
      "loss": 0.1348,
      "reward": 0.7270419001579285,
      "reward_std": 0.10564156621694565,
      "reward_std_all": 0.09736570343375206,
      "rewards-max/forensics_iou_reward": 1.0,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_iou_reward": 0.6681531667709351,
      "rewards/forensics_strict_edge_reward": 0.1016266755759716,
      "rewards/forensics_tiou_reward": 0.6352595686912537,
      "step": 1243
    },
    {
      "completion_length": 25.765625,
      "epoch": 3.817204301075269,
      "grad_norm": 9.116508483886719,
      "kl": 0.529296875,
      "learning_rate": 4.3846153846153846e-08,
      "loss": -0.025,
      "reward": 0.7739415764808655,
      "reward_std": 0.03951177094131708,
      "reward_std_all": 0.038210324477404356,
      "rewards-max/forensics_iou_reward": 0.9869106411933899,
      "rewards-max/forensics_strict_edge_reward": 0.5590172111988068,
      "rewards-max/forensics_tiou_reward": 0.982048362493515,
      "rewards/forensics_iou_reward": 0.6829221546649933,
      "rewards/forensics_strict_edge_reward": 0.054662372916936874,
      "rewards/forensics_tiou_reward": 0.6335845291614532,
      "step": 1244
    },
    {
      "completion_length": 31.734375,
      "epoch": 3.8202764976958523,
      "grad_norm": 6.753821849822998,
      "kl": 0.5078125,
      "learning_rate": 4.307692307692307e-08,
      "loss": 0.0968,
      "reward": 0.7143785655498505,
      "reward_std": 0.1053585410118103,
      "reward_std_all": 0.10125116631388664,
      "rewards-max/forensics_iou_reward": 0.9729137420654297,
      "rewards-max/forensics_strict_edge_reward": 0.4899434596300125,
      "rewards-max/forensics_tiou_reward": 0.9692675173282623,
      "rewards/forensics_iou_reward": 0.6368169188499451,
      "rewards/forensics_strict_edge_reward": 0.07785201072692871,
      "rewards/forensics_tiou_reward": 0.6060594022274017,
      "step": 1245
    },
    {
      "completion_length": 22.921875,
      "epoch": 3.823348694316436,
      "grad_norm": 7.119022846221924,
      "kl": 0.3916015625,
      "learning_rate": 4.230769230769231e-08,
      "loss": 0.0197,
      "reward": 0.7651278674602509,
      "reward_std": 0.057311322540044785,
      "reward_std_all": 0.053143588826060295,
      "rewards-max/forensics_iou_reward": 0.9932576417922974,
      "rewards-max/forensics_strict_edge_reward": 0.7654097378253937,
      "rewards-max/forensics_tiou_reward": 0.9885859489440918,
      "rewards/forensics_iou_reward": 0.7299968302249908,
      "rewards/forensics_strict_edge_reward": 0.09935259819030762,
      "rewards/forensics_tiou_reward": 0.6793597936630249,
      "step": 1246
    },
    {
      "completion_length": 28.046875,
      "epoch": 3.82642089093702,
      "grad_norm": 7.152834415435791,
      "kl": 0.470703125,
      "learning_rate": 4.1538461538461534e-08,
      "loss": 0.1082,
      "reward": 0.7271172404289246,
      "reward_std": 0.08024409599602222,
      "reward_std_all": 0.0807306133210659,
      "rewards-max/forensics_iou_reward": 0.9761704206466675,
      "rewards-max/forensics_strict_edge_reward": 0.6342070251703262,
      "rewards-max/forensics_tiou_reward": 0.9684655666351318,
      "rewards/forensics_iou_reward": 0.6180866956710815,
      "rewards/forensics_strict_edge_reward": 0.07338318042457104,
      "rewards/forensics_tiou_reward": 0.5853772759437561,
      "step": 1247
    },
    {
      "completion_length": 32.125,
      "epoch": 3.8294930875576036,
      "grad_norm": 9.149994850158691,
      "kl": 0.6171875,
      "learning_rate": 4.076923076923077e-08,
      "loss": 0.0816,
      "reward": 0.7678101658821106,
      "reward_std": 0.059105779975652695,
      "reward_std_all": 0.05360282398760319,
      "rewards-max/forensics_iou_reward": 0.9878973662853241,
      "rewards-max/forensics_strict_edge_reward": 0.4721640944480896,
      "rewards-max/forensics_tiou_reward": 0.9848775565624237,
      "rewards/forensics_iou_reward": 0.6690184772014618,
      "rewards/forensics_strict_edge_reward": 0.06386620551347733,
      "rewards/forensics_tiou_reward": 0.6242747604846954,
      "step": 1248
    },
    {
      "completion_length": 24.515625,
      "epoch": 3.8325652841781874,
      "grad_norm": 11.5474853515625,
      "kl": 0.36328125,
      "learning_rate": 4e-08,
      "loss": -0.009,
      "reward": 0.7419515550136566,
      "reward_std": 0.07348304567858577,
      "reward_std_all": 0.07170685823075473,
      "rewards-max/forensics_iou_reward": 0.9924527108669281,
      "rewards-max/forensics_strict_edge_reward": 0.7958523333072662,
      "rewards-max/forensics_tiou_reward": 0.9888960421085358,
      "rewards/forensics_iou_reward": 0.6730825304985046,
      "rewards/forensics_strict_edge_reward": 0.16744904965162277,
      "rewards/forensics_tiou_reward": 0.6407282948493958,
      "step": 1249
    },
    {
      "completion_length": 26.609375,
      "epoch": 3.835637480798771,
      "grad_norm": 7.9636921882629395,
      "kl": 0.5556640625,
      "learning_rate": 3.923076923076923e-08,
      "loss": 0.0377,
      "reward": 0.7542312741279602,
      "reward_std": 0.05328051187098026,
      "reward_std_all": 0.053738897666335106,
      "rewards-max/forensics_iou_reward": 0.98578742146492,
      "rewards-max/forensics_strict_edge_reward": 0.3930101543664932,
      "rewards-max/forensics_tiou_reward": 0.9813163578510284,
      "rewards/forensics_iou_reward": 0.7098894119262695,
      "rewards/forensics_strict_edge_reward": 0.061531201004981995,
      "rewards/forensics_tiou_reward": 0.6780003905296326,
      "step": 1250
    },
    {
      "completion_length": 26.546875,
      "epoch": 3.838709677419355,
      "grad_norm": 8.873424530029297,
      "kl": 0.3740234375,
      "learning_rate": 3.846153846153846e-08,
      "loss": 0.0675,
      "reward": 0.7318228483200073,
      "reward_std": 0.08302488178014755,
      "reward_std_all": 0.07957905158400536,
      "rewards-max/forensics_iou_reward": 0.9945054948329926,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9935164749622345,
      "rewards/forensics_iou_reward": 0.652546763420105,
      "rewards/forensics_strict_edge_reward": 0.11493687704205513,
      "rewards/forensics_tiou_reward": 0.6098190248012543,
      "step": 1251
    },
    {
      "completion_length": 24.03125,
      "epoch": 3.8417818740399383,
      "grad_norm": 8.103692054748535,
      "kl": 0.388671875,
      "learning_rate": 3.769230769230769e-08,
      "loss": -0.0004,
      "reward": 0.6825259923934937,
      "reward_std": 0.08152804709970951,
      "reward_std_all": 0.10401057824492455,
      "rewards-max/forensics_iou_reward": 0.9779271483421326,
      "rewards-max/forensics_strict_edge_reward": 0.6656907647848129,
      "rewards-max/forensics_tiou_reward": 0.9717424511909485,
      "rewards/forensics_iou_reward": 0.6100410223007202,
      "rewards/forensics_strict_edge_reward": 0.06324651464819908,
      "rewards/forensics_tiou_reward": 0.5719729661941528,
      "step": 1252
    },
    {
      "completion_length": 27.109375,
      "epoch": 3.8448540706605225,
      "grad_norm": 8.3251371383667,
      "kl": 0.373046875,
      "learning_rate": 3.692307692307692e-08,
      "loss": -0.0247,
      "reward": 0.8006815016269684,
      "reward_std": 0.009054087131517008,
      "reward_std_all": 0.010155932512134314,
      "rewards-max/forensics_iou_reward": 0.9957108497619629,
      "rewards-max/forensics_strict_edge_reward": 0.5750632882118225,
      "rewards-max/forensics_tiou_reward": 0.9934584200382233,
      "rewards/forensics_iou_reward": 0.7541320025920868,
      "rewards/forensics_strict_edge_reward": 0.06278879009187222,
      "rewards/forensics_tiou_reward": 0.706445962190628,
      "step": 1253
    },
    {
      "completion_length": 22.75,
      "epoch": 3.847926267281106,
      "grad_norm": 7.442593574523926,
      "kl": 0.3837890625,
      "learning_rate": 3.6153846153846155e-08,
      "loss": 0.048,
      "reward": 0.7709001004695892,
      "reward_std": 0.051181694492697716,
      "reward_std_all": 0.04781987890601158,
      "rewards-max/forensics_iou_reward": 1.0,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_iou_reward": 0.6407262682914734,
      "rewards/forensics_strict_edge_reward": 0.08657289296388626,
      "rewards/forensics_tiou_reward": 0.5927143096923828,
      "step": 1254
    },
    {
      "completion_length": 29.484375,
      "epoch": 3.8509984639016897,
      "grad_norm": 9.9683198928833,
      "kl": 0.46875,
      "learning_rate": 3.538461538461538e-08,
      "loss": 0.0889,
      "reward": 0.7411674857139587,
      "reward_std": 0.07493280433118343,
      "reward_std_all": 0.07192209921777248,
      "rewards-max/forensics_iou_reward": 0.9709838032722473,
      "rewards-max/forensics_strict_edge_reward": 0.661531001329422,
      "rewards-max/forensics_tiou_reward": 0.9657701253890991,
      "rewards/forensics_iou_reward": 0.6374633610248566,
      "rewards/forensics_strict_edge_reward": 0.05908288434147835,
      "rewards/forensics_tiou_reward": 0.5901962518692017,
      "step": 1255
    },
    {
      "completion_length": 24.09375,
      "epoch": 3.8540706605222734,
      "grad_norm": 7.870098114013672,
      "kl": 0.3818359375,
      "learning_rate": 3.4615384615384616e-08,
      "loss": 0.0974,
      "reward": 0.7022591233253479,
      "reward_std": 0.1423775739967823,
      "reward_std_all": 0.12990288063883781,
      "rewards-max/forensics_iou_reward": 0.9735554456710815,
      "rewards-max/forensics_strict_edge_reward": 0.47726282477378845,
      "rewards-max/forensics_tiou_reward": 0.9670383930206299,
      "rewards/forensics_iou_reward": 0.6455926299095154,
      "rewards/forensics_strict_edge_reward": 0.05980610474944115,
      "rewards/forensics_tiou_reward": 0.6053817570209503,
      "step": 1256
    },
    {
      "completion_length": 25.859375,
      "epoch": 3.857142857142857,
      "grad_norm": 10.071964263916016,
      "kl": 0.642578125,
      "learning_rate": 3.384615384615384e-08,
      "loss": -0.0277,
      "reward": 0.711991161108017,
      "reward_std": 0.0397915244102478,
      "reward_std_all": 0.06569714471697807,
      "rewards-max/forensics_iou_reward": 0.9787059426307678,
      "rewards-max/forensics_strict_edge_reward": 0.7181241810321808,
      "rewards-max/forensics_tiou_reward": 0.970451295375824,
      "rewards/forensics_iou_reward": 0.6486910283565521,
      "rewards/forensics_strict_edge_reward": 0.09308744966983795,
      "rewards/forensics_tiou_reward": 0.6104685962200165,
      "step": 1257
    },
    {
      "completion_length": 29.265625,
      "epoch": 3.860215053763441,
      "grad_norm": 7.525156021118164,
      "kl": 0.529296875,
      "learning_rate": 3.3076923076923076e-08,
      "loss": 0.0257,
      "reward": 0.7173311710357666,
      "reward_std": 0.028651023749262094,
      "reward_std_all": 0.05595529451966286,
      "rewards-max/forensics_iou_reward": 0.9828219413757324,
      "rewards-max/forensics_strict_edge_reward": 0.6531622409820557,
      "rewards-max/forensics_tiou_reward": 0.9766981899738312,
      "rewards/forensics_iou_reward": 0.6428150534629822,
      "rewards/forensics_strict_edge_reward": 0.08504936471581459,
      "rewards/forensics_tiou_reward": 0.5978164076805115,
      "step": 1258
    },
    {
      "completion_length": 24.40625,
      "epoch": 3.8632872503840243,
      "grad_norm": 7.48286247253418,
      "kl": 0.533203125,
      "learning_rate": 3.230769230769231e-08,
      "loss": 0.038,
      "reward": 0.7957748472690582,
      "reward_std": 0.019082820042967796,
      "reward_std_all": 0.0182142099365592,
      "rewards-max/forensics_iou_reward": 0.9813411235809326,
      "rewards-max/forensics_strict_edge_reward": 0.5218044221401215,
      "rewards-max/forensics_tiou_reward": 0.9714827239513397,
      "rewards/forensics_iou_reward": 0.7826870083808899,
      "rewards/forensics_strict_edge_reward": 0.06685378216207027,
      "rewards/forensics_tiou_reward": 0.7303859889507294,
      "step": 1259
    },
    {
      "completion_length": 23.71875,
      "epoch": 3.8663594470046085,
      "grad_norm": 7.054775714874268,
      "kl": 0.4306640625,
      "learning_rate": 3.153846153846154e-08,
      "loss": 0.0154,
      "reward": 0.7559601664543152,
      "reward_std": 0.06142742931842804,
      "reward_std_all": 0.0587842317763716,
      "rewards-max/forensics_iou_reward": 0.9838976263999939,
      "rewards-max/forensics_strict_edge_reward": 0.7265062034130096,
      "rewards-max/forensics_tiou_reward": 0.9801516830921173,
      "rewards/forensics_iou_reward": 0.6530215442180634,
      "rewards/forensics_strict_edge_reward": 0.08252526819705963,
      "rewards/forensics_tiou_reward": 0.6077467799186707,
      "step": 1260
    },
    {
      "completion_length": 28.265625,
      "epoch": 3.869431643625192,
      "grad_norm": 7.230008125305176,
      "kl": 0.5244140625,
      "learning_rate": 3.076923076923077e-08,
      "loss": -0.0775,
      "reward": 0.75831738114357,
      "reward_std": 0.005757824692409486,
      "reward_std_all": 0.02939959615468979,
      "rewards-max/forensics_iou_reward": 0.9885170161724091,
      "rewards-max/forensics_strict_edge_reward": 0.7663538753986359,
      "rewards-max/forensics_tiou_reward": 0.9828600287437439,
      "rewards/forensics_iou_reward": 0.7400254011154175,
      "rewards/forensics_strict_edge_reward": 0.10050118342041969,
      "rewards/forensics_tiou_reward": 0.7043453752994537,
      "step": 1261
    },
    {
      "completion_length": 24.84375,
      "epoch": 3.8725038402457757,
      "grad_norm": 7.557322978973389,
      "kl": 0.4208984375,
      "learning_rate": 3e-08,
      "loss": -0.0312,
      "reward": 0.7493489980697632,
      "reward_std": 0.0744931511580944,
      "reward_std_all": 0.06983663514256477,
      "rewards-max/forensics_iou_reward": 0.9970435500144958,
      "rewards-max/forensics_strict_edge_reward": 0.7331126630306244,
      "rewards-max/forensics_tiou_reward": 0.9960009753704071,
      "rewards/forensics_iou_reward": 0.7163170278072357,
      "rewards/forensics_strict_edge_reward": 0.1213802918791771,
      "rewards/forensics_tiou_reward": 0.6824984550476074,
      "step": 1262
    },
    {
      "completion_length": 25.671875,
      "epoch": 3.8755760368663594,
      "grad_norm": 8.663727760314941,
      "kl": 0.5595703125,
      "learning_rate": 2.9230769230769228e-08,
      "loss": 0.0429,
      "reward": 0.7647138833999634,
      "reward_std": 0.046313680708408356,
      "reward_std_all": 0.04526020027697086,
      "rewards-max/forensics_iou_reward": 0.9899952411651611,
      "rewards-max/forensics_strict_edge_reward": 0.5343262553215027,
      "rewards-max/forensics_tiou_reward": 0.9837369620800018,
      "rewards/forensics_iou_reward": 0.6882166564464569,
      "rewards/forensics_strict_edge_reward": 0.08227892220020294,
      "rewards/forensics_tiou_reward": 0.6459015607833862,
      "step": 1263
    },
    {
      "completion_length": 28.859375,
      "epoch": 3.878648233486943,
      "grad_norm": 7.868238925933838,
      "kl": 0.541015625,
      "learning_rate": 2.846153846153846e-08,
      "loss": 0.0549,
      "reward": 0.7756660282611847,
      "reward_std": 0.04800320602953434,
      "reward_std_all": 0.04369573947042227,
      "rewards-max/forensics_iou_reward": 0.9847379624843597,
      "rewards-max/forensics_strict_edge_reward": 0.6952967792749405,
      "rewards-max/forensics_tiou_reward": 0.9800395667552948,
      "rewards/forensics_iou_reward": 0.7491486072540283,
      "rewards/forensics_strict_edge_reward": 0.12330295518040657,
      "rewards/forensics_tiou_reward": 0.7161702513694763,
      "step": 1264
    },
    {
      "completion_length": 29.625,
      "epoch": 3.881720430107527,
      "grad_norm": 7.952967643737793,
      "kl": 0.4150390625,
      "learning_rate": 2.769230769230769e-08,
      "loss": -0.0531,
      "reward": 0.789887547492981,
      "reward_std": 0.028121270705014467,
      "reward_std_all": 0.026537287514656782,
      "rewards-max/forensics_iou_reward": 0.9896198809146881,
      "rewards-max/forensics_strict_edge_reward": 0.7789089381694794,
      "rewards-max/forensics_tiou_reward": 0.987237811088562,
      "rewards/forensics_iou_reward": 0.7378876209259033,
      "rewards/forensics_strict_edge_reward": 0.09458699077367783,
      "rewards/forensics_tiou_reward": 0.6846173703670502,
      "step": 1265
    },
    {
      "completion_length": 23.59375,
      "epoch": 3.8847926267281108,
      "grad_norm": 8.793116569519043,
      "kl": 0.4560546875,
      "learning_rate": 2.6923076923076925e-08,
      "loss": -0.055,
      "reward": 0.7799775004386902,
      "reward_std": 0.00535190524533391,
      "reward_std_all": 0.017450836952775717,
      "rewards-max/forensics_iou_reward": 0.9847313165664673,
      "rewards-max/forensics_strict_edge_reward": 0.5805842876434326,
      "rewards-max/forensics_tiou_reward": 0.9761391878128052,
      "rewards/forensics_iou_reward": 0.6767424643039703,
      "rewards/forensics_strict_edge_reward": 0.06866970285773277,
      "rewards/forensics_tiou_reward": 0.6182726323604584,
      "step": 1266
    },
    {
      "completion_length": 24.5625,
      "epoch": 3.8878648233486945,
      "grad_norm": 9.741647720336914,
      "kl": 0.533203125,
      "learning_rate": 2.6153846153846152e-08,
      "loss": 0.0435,
      "reward": 0.7994741201400757,
      "reward_std": 0.009392267791554332,
      "reward_std_all": 0.01047359500080347,
      "rewards-max/forensics_iou_reward": 0.9763250946998596,
      "rewards-max/forensics_strict_edge_reward": 0.4281903952360153,
      "rewards-max/forensics_tiou_reward": 0.9677214324474335,
      "rewards/forensics_iou_reward": 0.7507173418998718,
      "rewards/forensics_strict_edge_reward": 0.058636732399463654,
      "rewards/forensics_tiou_reward": 0.6953473091125488,
      "step": 1267
    },
    {
      "completion_length": 26.9375,
      "epoch": 3.890937019969278,
      "grad_norm": 7.644545555114746,
      "kl": 0.326171875,
      "learning_rate": 2.5384615384615382e-08,
      "loss": 0.03,
      "reward": 0.7751221060752869,
      "reward_std": 0.05026019550859928,
      "reward_std_all": 0.04596443846821785,
      "rewards-max/forensics_iou_reward": 0.9760926961898804,
      "rewards-max/forensics_strict_edge_reward": 0.5809257328510284,
      "rewards-max/forensics_tiou_reward": 0.9663850665092468,
      "rewards/forensics_iou_reward": 0.6935788989067078,
      "rewards/forensics_strict_edge_reward": 0.105190210044384,
      "rewards/forensics_tiou_reward": 0.6562299430370331,
      "step": 1268
    },
    {
      "completion_length": 24.8125,
      "epoch": 3.8940092165898617,
      "grad_norm": 8.889426231384277,
      "kl": 0.427734375,
      "learning_rate": 2.4615384615384616e-08,
      "loss": 0.1454,
      "reward": 0.7478497922420502,
      "reward_std": 0.0680491030216217,
      "reward_std_all": 0.06761755608022213,
      "rewards-max/forensics_iou_reward": 0.9855090379714966,
      "rewards-max/forensics_strict_edge_reward": 0.6476292610168457,
      "rewards-max/forensics_tiou_reward": 0.9796042442321777,
      "rewards/forensics_iou_reward": 0.6512080430984497,
      "rewards/forensics_strict_edge_reward": 0.06630830280482769,
      "rewards/forensics_tiou_reward": 0.6075236201286316,
      "step": 1269
    },
    {
      "completion_length": 23.921875,
      "epoch": 3.8970814132104454,
      "grad_norm": 11.861905097961426,
      "kl": 0.416015625,
      "learning_rate": 2.3846153846153846e-08,
      "loss": 0.0807,
      "reward": 0.792768657207489,
      "reward_std": 0.019651522859930992,
      "reward_std_all": 0.01911342144012451,
      "rewards-max/forensics_iou_reward": 0.9889443218708038,
      "rewards-max/forensics_strict_edge_reward": 0.6762872636318207,
      "rewards-max/forensics_tiou_reward": 0.984826922416687,
      "rewards/forensics_iou_reward": 0.7492072284221649,
      "rewards/forensics_strict_edge_reward": 0.09531090408563614,
      "rewards/forensics_tiou_reward": 0.6963847577571869,
      "step": 1270
    },
    {
      "completion_length": 29.21875,
      "epoch": 3.900153609831029,
      "grad_norm": 7.2019805908203125,
      "kl": 0.431640625,
      "learning_rate": 2.3076923076923076e-08,
      "loss": -0.0513,
      "reward": 0.6941502690315247,
      "reward_std": 0.029131892137229443,
      "reward_std_all": 0.0704517550766468,
      "rewards-max/forensics_iou_reward": 0.9764253199100494,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9705285131931305,
      "rewards/forensics_iou_reward": 0.6773020923137665,
      "rewards/forensics_strict_edge_reward": 0.10911547392606735,
      "rewards/forensics_tiou_reward": 0.647040069103241,
      "step": 1271
    },
    {
      "completion_length": 26.15625,
      "epoch": 3.903225806451613,
      "grad_norm": 7.651074409484863,
      "kl": 0.384765625,
      "learning_rate": 2.2307692307692306e-08,
      "loss": -0.0013,
      "reward": 0.7723029255867004,
      "reward_std": 0.0463812667876482,
      "reward_std_all": 0.043668026104569435,
      "rewards-max/forensics_iou_reward": 0.992118239402771,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9902341663837433,
      "rewards/forensics_iou_reward": 0.6991950273513794,
      "rewards/forensics_strict_edge_reward": 0.08461781591176987,
      "rewards/forensics_tiou_reward": 0.6560357809066772,
      "step": 1272
    },
    {
      "completion_length": 27.59375,
      "epoch": 3.9062980030721968,
      "grad_norm": 5.882526397705078,
      "kl": 0.3779296875,
      "learning_rate": 2.1538461538461537e-08,
      "loss": 0.0341,
      "reward": 0.7924835681915283,
      "reward_std": 0.015789429657161236,
      "reward_std_all": 0.016640803776681423,
      "rewards-max/forensics_iou_reward": 0.9914875626564026,
      "rewards-max/forensics_strict_edge_reward": 0.863698273897171,
      "rewards-max/forensics_tiou_reward": 0.9892625510692596,
      "rewards/forensics_iou_reward": 0.7474527060985565,
      "rewards/forensics_strict_edge_reward": 0.0912555381655693,
      "rewards/forensics_tiou_reward": 0.7108218669891357,
      "step": 1273
    },
    {
      "completion_length": 28.34375,
      "epoch": 3.9093701996927805,
      "grad_norm": 6.675223350524902,
      "kl": 0.43359375,
      "learning_rate": 2.0769230769230767e-08,
      "loss": 0.1536,
      "reward": 0.7446359097957611,
      "reward_std": 0.09287238866090775,
      "reward_std_all": 0.08341910690069199,
      "rewards-max/forensics_iou_reward": 0.9928489029407501,
      "rewards-max/forensics_strict_edge_reward": 0.5864547789096832,
      "rewards-max/forensics_tiou_reward": 0.9893723130226135,
      "rewards/forensics_iou_reward": 0.6953976452350616,
      "rewards/forensics_strict_edge_reward": 0.1028989851474762,
      "rewards/forensics_tiou_reward": 0.6436453759670258,
      "step": 1274
    },
    {
      "completion_length": 24.65625,
      "epoch": 3.912442396313364,
      "grad_norm": 8.239100456237793,
      "kl": 0.515625,
      "learning_rate": 2e-08,
      "loss": 0.0215,
      "reward": 0.7860360443592072,
      "reward_std": 0.03575849998742342,
      "reward_std_all": 0.032748361118137836,
      "rewards-max/forensics_iou_reward": 0.989527016878128,
      "rewards-max/forensics_strict_edge_reward": 0.7589184939861298,
      "rewards-max/forensics_tiou_reward": 0.9839328825473785,
      "rewards/forensics_iou_reward": 0.7620241940021515,
      "rewards/forensics_strict_edge_reward": 0.09533197432756424,
      "rewards/forensics_tiou_reward": 0.7201164960861206,
      "step": 1275
    },
    {
      "completion_length": 27.0625,
      "epoch": 3.9155145929339477,
      "grad_norm": 7.211334228515625,
      "kl": 0.4619140625,
      "learning_rate": 1.923076923076923e-08,
      "loss": 0.0284,
      "reward": 0.7792826592922211,
      "reward_std": 0.030376846785657108,
      "reward_std_all": 0.030425073578953743,
      "rewards-max/forensics_iou_reward": 0.9950313866138458,
      "rewards-max/forensics_strict_edge_reward": 0.6857411116361618,
      "rewards-max/forensics_tiou_reward": 0.992514967918396,
      "rewards/forensics_iou_reward": 0.7476252913475037,
      "rewards/forensics_strict_edge_reward": 0.10790565609931946,
      "rewards/forensics_tiou_reward": 0.7118062376976013,
      "step": 1276
    },
    {
      "completion_length": 22.515625,
      "epoch": 3.9185867895545314,
      "grad_norm": 7.5101165771484375,
      "kl": 0.3857421875,
      "learning_rate": 1.846153846153846e-08,
      "loss": -0.0576,
      "reward": 0.7703525424003601,
      "reward_std": 0.04464499652385712,
      "reward_std_all": 0.043113719671964645,
      "rewards-max/forensics_iou_reward": 0.9868656396865845,
      "rewards-max/forensics_strict_edge_reward": 0.8142738044261932,
      "rewards-max/forensics_tiou_reward": 0.981727808713913,
      "rewards/forensics_iou_reward": 0.7521150708198547,
      "rewards/forensics_strict_edge_reward": 0.12853964790701866,
      "rewards/forensics_tiou_reward": 0.7184391915798187,
      "step": 1277
    },
    {
      "completion_length": 27.6875,
      "epoch": 3.921658986175115,
      "grad_norm": 9.128427505493164,
      "kl": 0.65625,
      "learning_rate": 1.769230769230769e-08,
      "loss": 0.0772,
      "reward": 0.7661449015140533,
      "reward_std": 0.025026529096066952,
      "reward_std_all": 0.03395839408040047,
      "rewards-max/forensics_iou_reward": 0.9815683662891388,
      "rewards-max/forensics_strict_edge_reward": 0.44864383339881897,
      "rewards-max/forensics_tiou_reward": 0.9753150045871735,
      "rewards/forensics_iou_reward": 0.7050826847553253,
      "rewards/forensics_strict_edge_reward": 0.08739429339766502,
      "rewards/forensics_tiou_reward": 0.6624475121498108,
      "step": 1278
    },
    {
      "completion_length": 23.703125,
      "epoch": 3.924731182795699,
      "grad_norm": 7.103725433349609,
      "kl": 0.4462890625,
      "learning_rate": 1.692307692307692e-08,
      "loss": 0.0882,
      "reward": 0.7527323663234711,
      "reward_std": 0.06109754089266062,
      "reward_std_all": 0.05899907648563385,
      "rewards-max/forensics_iou_reward": 0.9900662004947662,
      "rewards-max/forensics_strict_edge_reward": 0.6310813128948212,
      "rewards-max/forensics_tiou_reward": 0.9875774085521698,
      "rewards/forensics_iou_reward": 0.6815388798713684,
      "rewards/forensics_strict_edge_reward": 0.06984969973564148,
      "rewards/forensics_tiou_reward": 0.6455139219760895,
      "step": 1279
    },
    {
      "completion_length": 27.890625,
      "epoch": 3.9278033794162828,
      "grad_norm": 6.838144779205322,
      "kl": 0.3427734375,
      "learning_rate": 1.6153846153846155e-08,
      "loss": -0.0797,
      "reward": 0.8061814904212952,
      "reward_std": 0.0014264812925830483,
      "reward_std_all": 0.0031203142134472728,
      "rewards-max/forensics_iou_reward": 0.9864586293697357,
      "rewards-max/forensics_strict_edge_reward": 0.6555143892765045,
      "rewards-max/forensics_tiou_reward": 0.9819636642932892,
      "rewards/forensics_iou_reward": 0.8236380219459534,
      "rewards/forensics_strict_edge_reward": 0.08771225064992905,
      "rewards/forensics_tiou_reward": 0.7907333076000214,
      "step": 1280
    },
    {
      "completion_length": 31.21875,
      "epoch": 3.9308755760368665,
      "grad_norm": 7.591765403747559,
      "kl": 0.4453125,
      "learning_rate": 1.5384615384615385e-08,
      "loss": -0.031,
      "reward": 0.797736644744873,
      "reward_std": 0.009499955514911562,
      "reward_std_all": 0.010936662554740906,
      "rewards-max/forensics_iou_reward": 0.9807026386260986,
      "rewards-max/forensics_strict_edge_reward": 0.5692913830280304,
      "rewards-max/forensics_tiou_reward": 0.9755512773990631,
      "rewards/forensics_iou_reward": 0.7272993922233582,
      "rewards/forensics_strict_edge_reward": 0.08895356208086014,
      "rewards/forensics_tiou_reward": 0.6961434185504913,
      "step": 1281
    },
    {
      "completion_length": 28.078125,
      "epoch": 3.93394777265745,
      "grad_norm": 6.519299030303955,
      "kl": 0.4462890625,
      "learning_rate": 1.4615384615384614e-08,
      "loss": 0.0393,
      "reward": 0.7804326415061951,
      "reward_std": 0.03618492791429162,
      "reward_std_all": 0.033970441203564405,
      "rewards-max/forensics_iou_reward": 0.990688681602478,
      "rewards-max/forensics_strict_edge_reward": 0.6267848908901215,
      "rewards-max/forensics_tiou_reward": 0.9888553321361542,
      "rewards/forensics_iou_reward": 0.720567137002945,
      "rewards/forensics_strict_edge_reward": 0.11310983076691628,
      "rewards/forensics_tiou_reward": 0.6880472004413605,
      "step": 1282
    },
    {
      "completion_length": 22.59375,
      "epoch": 3.9370199692780337,
      "grad_norm": 6.661041736602783,
      "kl": 0.498046875,
      "learning_rate": 1.3846153846153846e-08,
      "loss": 0.0496,
      "reward": 0.7774759531021118,
      "reward_std": 0.03458297625184059,
      "reward_std_all": 0.0336165688931942,
      "rewards-max/forensics_iou_reward": 0.9761724472045898,
      "rewards-max/forensics_strict_edge_reward": 0.5922519117593765,
      "rewards-max/forensics_tiou_reward": 0.9661056399345398,
      "rewards/forensics_iou_reward": 0.7615812420845032,
      "rewards/forensics_strict_edge_reward": 0.11011308431625366,
      "rewards/forensics_tiou_reward": 0.7127462029457092,
      "step": 1283
    },
    {
      "completion_length": 26.234375,
      "epoch": 3.9400921658986174,
      "grad_norm": 8.513860702514648,
      "kl": 0.73828125,
      "learning_rate": 1.3076923076923076e-08,
      "loss": 0.0182,
      "reward": 0.779093861579895,
      "reward_std": 0.04282620060257614,
      "reward_std_all": 0.03945325291715562,
      "rewards-max/forensics_iou_reward": 0.9914754629135132,
      "rewards-max/forensics_strict_edge_reward": 0.7617588341236115,
      "rewards-max/forensics_tiou_reward": 0.9897921085357666,
      "rewards/forensics_iou_reward": 0.7390690743923187,
      "rewards/forensics_strict_edge_reward": 0.10240215063095093,
      "rewards/forensics_tiou_reward": 0.7057301998138428,
      "step": 1284
    },
    {
      "completion_length": 23.765625,
      "epoch": 3.943164362519201,
      "grad_norm": 9.85459041595459,
      "kl": 0.375,
      "learning_rate": 1.2307692307692308e-08,
      "loss": 0.0797,
      "reward": 0.7394066452980042,
      "reward_std": 0.09591308608651161,
      "reward_std_all": 0.08765043318271637,
      "rewards-max/forensics_iou_reward": 0.9006916880607605,
      "rewards-max/forensics_strict_edge_reward": 0.3744408041238785,
      "rewards-max/forensics_tiou_reward": 0.864306628704071,
      "rewards/forensics_iou_reward": 0.5973353385925293,
      "rewards/forensics_strict_edge_reward": 0.02907026745378971,
      "rewards/forensics_tiou_reward": 0.5461249649524689,
      "step": 1285
    },
    {
      "completion_length": 28.734375,
      "epoch": 3.946236559139785,
      "grad_norm": 7.160484313964844,
      "kl": 0.43359375,
      "learning_rate": 1.1538461538461538e-08,
      "loss": 0.0047,
      "reward": 0.7554255723953247,
      "reward_std": 0.05267322307918221,
      "reward_std_all": 0.05285278521478176,
      "rewards-max/forensics_iou_reward": 0.9899066388607025,
      "rewards-max/forensics_strict_edge_reward": 0.6425930857658386,
      "rewards-max/forensics_tiou_reward": 0.9862965941429138,
      "rewards/forensics_iou_reward": 0.6993840932846069,
      "rewards/forensics_strict_edge_reward": 0.05542263761162758,
      "rewards/forensics_tiou_reward": 0.6556322276592255,
      "step": 1286
    },
    {
      "completion_length": 23.21875,
      "epoch": 3.9493087557603688,
      "grad_norm": 8.00512981414795,
      "kl": 0.4169921875,
      "learning_rate": 1.0769230769230768e-08,
      "loss": -0.0546,
      "reward": 0.7859277725219727,
      "reward_std": 0.027161755540873855,
      "reward_std_all": 0.02794949896633625,
      "rewards-max/forensics_iou_reward": 0.9915020167827606,
      "rewards-max/forensics_strict_edge_reward": 0.5987008213996887,
      "rewards-max/forensics_tiou_reward": 0.9891421794891357,
      "rewards/forensics_iou_reward": 0.7653530836105347,
      "rewards/forensics_strict_edge_reward": 0.07210754603147507,
      "rewards/forensics_tiou_reward": 0.7303850650787354,
      "step": 1287
    },
    {
      "completion_length": 26.15625,
      "epoch": 3.9523809523809526,
      "grad_norm": 7.311583995819092,
      "kl": 0.4169921875,
      "learning_rate": 1e-08,
      "loss": -0.0371,
      "reward": 0.7497846782207489,
      "reward_std": 0.02122989750932902,
      "reward_std_all": 0.03865074133500457,
      "rewards-max/forensics_iou_reward": 0.984404057264328,
      "rewards-max/forensics_strict_edge_reward": 0.645659938454628,
      "rewards-max/forensics_tiou_reward": 0.9790441989898682,
      "rewards/forensics_iou_reward": 0.7222978174686432,
      "rewards/forensics_strict_edge_reward": 0.10279514081776142,
      "rewards/forensics_tiou_reward": 0.6887359321117401,
      "step": 1288
    },
    {
      "completion_length": 23.890625,
      "epoch": 3.955453149001536,
      "grad_norm": 8.03634262084961,
      "kl": 0.685546875,
      "learning_rate": 9.23076923076923e-09,
      "loss": 0.0463,
      "reward": 0.7079987823963165,
      "reward_std": 0.10709340125322342,
      "reward_std_all": 0.10515939816832542,
      "rewards-max/forensics_iou_reward": 0.9910714328289032,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9890037477016449,
      "rewards/forensics_iou_reward": 0.6298055946826935,
      "rewards/forensics_strict_edge_reward": 0.11254160106182098,
      "rewards/forensics_tiou_reward": 0.5807459652423859,
      "step": 1289
    },
    {
      "completion_length": 26.859375,
      "epoch": 3.9585253456221197,
      "grad_norm": 8.424270629882812,
      "kl": 0.44921875,
      "learning_rate": 8.46153846153846e-09,
      "loss": 0.1049,
      "reward": 0.7274856269359589,
      "reward_std": 0.04485565796494484,
      "reward_std_all": 0.059049224480986595,
      "rewards-max/forensics_iou_reward": 0.9733558893203735,
      "rewards-max/forensics_strict_edge_reward": 0.7249340116977692,
      "rewards-max/forensics_tiou_reward": 0.9679769575595856,
      "rewards/forensics_iou_reward": 0.6743337512016296,
      "rewards/forensics_strict_edge_reward": 0.08757226169109344,
      "rewards/forensics_tiou_reward": 0.6350674629211426,
      "step": 1290
    },
    {
      "completion_length": 26.421875,
      "epoch": 3.9615975422427034,
      "grad_norm": 8.105977058410645,
      "kl": 0.4990234375,
      "learning_rate": 7.692307692307693e-09,
      "loss": -0.0533,
      "reward": 0.7997050583362579,
      "reward_std": 0.007939512084703892,
      "reward_std_all": 0.009454551502130926,
      "rewards-max/forensics_iou_reward": 0.9888308048248291,
      "rewards-max/forensics_strict_edge_reward": 0.7309097647666931,
      "rewards-max/forensics_tiou_reward": 0.9848265051841736,
      "rewards/forensics_iou_reward": 0.7584745287895203,
      "rewards/forensics_strict_edge_reward": 0.11518585123121738,
      "rewards/forensics_tiou_reward": 0.7159454822540283,
      "step": 1291
    },
    {
      "completion_length": 25.3125,
      "epoch": 3.964669738863287,
      "grad_norm": 8.008488655090332,
      "kl": 0.55078125,
      "learning_rate": 6.923076923076923e-09,
      "loss": -0.0197,
      "reward": 0.7753922641277313,
      "reward_std": 0.02229646686464548,
      "reward_std_all": 0.026923757046461105,
      "rewards-max/forensics_iou_reward": 0.9887852966785431,
      "rewards-max/forensics_strict_edge_reward": 0.40398500859737396,
      "rewards-max/forensics_tiou_reward": 0.986075758934021,
      "rewards/forensics_iou_reward": 0.7131806015968323,
      "rewards/forensics_strict_edge_reward": 0.06022339686751366,
      "rewards/forensics_tiou_reward": 0.6700348258018494,
      "step": 1292
    },
    {
      "completion_length": 26.78125,
      "epoch": 3.967741935483871,
      "grad_norm": 8.720592498779297,
      "kl": 0.658203125,
      "learning_rate": 6.153846153846154e-09,
      "loss": 0.0553,
      "reward": 0.6880545318126678,
      "reward_std": 0.07555629685521126,
      "reward_std_all": 0.09712865948677063,
      "rewards-max/forensics_iou_reward": 0.9851727783679962,
      "rewards-max/forensics_strict_edge_reward": 0.4887978136539459,
      "rewards-max/forensics_tiou_reward": 0.9808536469936371,
      "rewards/forensics_iou_reward": 0.583142876625061,
      "rewards/forensics_strict_edge_reward": 0.06082826480269432,
      "rewards/forensics_tiou_reward": 0.5357453525066376,
      "step": 1293
    },
    {
      "completion_length": 24.09375,
      "epoch": 3.9708141321044548,
      "grad_norm": 7.810125827789307,
      "kl": 0.560546875,
      "learning_rate": 5.384615384615384e-09,
      "loss": -0.132,
      "reward": 0.7321552038192749,
      "reward_std": 0.04307556385174394,
      "reward_std_all": 0.06292231380939484,
      "rewards-max/forensics_iou_reward": 0.9692998826503754,
      "rewards-max/forensics_strict_edge_reward": 0.4724559336900711,
      "rewards-max/forensics_tiou_reward": 0.9570040702819824,
      "rewards/forensics_iou_reward": 0.594041496515274,
      "rewards/forensics_strict_edge_reward": 0.06314768828451633,
      "rewards/forensics_tiou_reward": 0.5385124534368515,
      "step": 1294
    },
    {
      "completion_length": 29.78125,
      "epoch": 3.9738863287250386,
      "grad_norm": 6.37498664855957,
      "kl": 0.4462890625,
      "learning_rate": 4.615384615384615e-09,
      "loss": 0.0018,
      "reward": 0.7923102676868439,
      "reward_std": 0.017176686320453882,
      "reward_std_all": 0.017748462967574596,
      "rewards-max/forensics_iou_reward": 0.9868473708629608,
      "rewards-max/forensics_strict_edge_reward": 0.5025823712348938,
      "rewards-max/forensics_tiou_reward": 0.9801545441150665,
      "rewards/forensics_iou_reward": 0.7224027812480927,
      "rewards/forensics_strict_edge_reward": 0.08199204504489899,
      "rewards/forensics_tiou_reward": 0.6575127840042114,
      "step": 1295
    },
    {
      "completion_length": 28.125,
      "epoch": 3.976958525345622,
      "grad_norm": 7.244144916534424,
      "kl": 0.48828125,
      "learning_rate": 3.846153846153846e-09,
      "loss": -0.0095,
      "reward": 0.8022579252719879,
      "reward_std": 0.00816559954546392,
      "reward_std_all": 0.008866927470080554,
      "rewards-max/forensics_iou_reward": 0.993796169757843,
      "rewards-max/forensics_strict_edge_reward": 0.5918685793876648,
      "rewards-max/forensics_tiou_reward": 0.991547167301178,
      "rewards/forensics_iou_reward": 0.800312876701355,
      "rewards/forensics_strict_edge_reward": 0.10583692044019699,
      "rewards/forensics_tiou_reward": 0.7611507773399353,
      "step": 1296
    },
    {
      "completion_length": 28.65625,
      "epoch": 3.980030721966206,
      "grad_norm": 7.307015419006348,
      "kl": 0.50390625,
      "learning_rate": 3.076923076923077e-09,
      "loss": 0.0243,
      "reward": 0.7293961346149445,
      "reward_std": 0.10506448522210121,
      "reward_std_all": 0.09671583026647568,
      "rewards-max/forensics_iou_reward": 0.994165301322937,
      "rewards-max/forensics_strict_edge_reward": 0.8029135167598724,
      "rewards-max/forensics_tiou_reward": 0.9910762012004852,
      "rewards/forensics_iou_reward": 0.6546715199947357,
      "rewards/forensics_strict_edge_reward": 0.11347842961549759,
      "rewards/forensics_tiou_reward": 0.6299288272857666,
      "step": 1297
    },
    {
      "completion_length": 25.296875,
      "epoch": 3.9831029185867894,
      "grad_norm": 10.354448318481445,
      "kl": 0.3955078125,
      "learning_rate": 2.3076923076923076e-09,
      "loss": 0.0539,
      "reward": 0.7320325374603271,
      "reward_std": 0.09881016612052917,
      "reward_std_all": 0.09098826348781586,
      "rewards-max/forensics_iou_reward": 0.9814172685146332,
      "rewards-max/forensics_strict_edge_reward": 0.7727466821670532,
      "rewards-max/forensics_tiou_reward": 0.9783138632774353,
      "rewards/forensics_iou_reward": 0.7032745480537415,
      "rewards/forensics_strict_edge_reward": 0.10484788939356804,
      "rewards/forensics_tiou_reward": 0.6728948056697845,
      "step": 1298
    },
    {
      "completion_length": 25.828125,
      "epoch": 3.986175115207373,
      "grad_norm": 8.834495544433594,
      "kl": 0.404296875,
      "learning_rate": 1.5384615384615385e-09,
      "loss": 0.0319,
      "reward": 0.7874384820461273,
      "reward_std": 0.025088281370699406,
      "reward_std_all": 0.024480942636728287,
      "rewards-max/forensics_iou_reward": 0.9778969883918762,
      "rewards-max/forensics_strict_edge_reward": 0.6785503029823303,
      "rewards-max/forensics_tiou_reward": 0.9630824625492096,
      "rewards/forensics_iou_reward": 0.7209479212760925,
      "rewards/forensics_strict_edge_reward": 0.11796383559703827,
      "rewards/forensics_tiou_reward": 0.691141813993454,
      "step": 1299
    },
    {
      "completion_length": 26.84375,
      "epoch": 3.989247311827957,
      "grad_norm": 7.229986190795898,
      "kl": 0.4521484375,
      "learning_rate": 7.692307692307692e-10,
      "loss": 0.1331,
      "reward": 0.7579329609870911,
      "reward_std": 0.08300960436463356,
      "reward_std_all": 0.07352462038397789,
      "rewards-max/forensics_iou_reward": 0.9967861473560333,
      "rewards-max/forensics_strict_edge_reward": 0.8838155269622803,
      "rewards-max/forensics_tiou_reward": 0.9958995282649994,
      "rewards/forensics_iou_reward": 0.7128270864486694,
      "rewards/forensics_strict_edge_reward": 0.0988265834748745,
      "rewards/forensics_tiou_reward": 0.6729466617107391,
      "step": 1300
    }
  ],
  "logging_steps": 1.0,
  "max_steps": 1300,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 1300,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}