ForgeLoc-R1 / baselines /af_tempsamp-r1 /trainer_state.json
sdzt's picture
Add baselines/af_tempsamp-r1
5f6e29b verified
Raw History Blame Contribute Delete
473 kB
{
"best_global_step": null,
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 4.0,
"eval_steps": 500,
"global_step": 956,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"completion_length": 22.390625,
"epoch": 0.0041841004184100415,
"grad_norm": 7.586400032043457,
"kl": 0.0,
"learning_rate": 1e-06,
"loss": 0.0,
"reward": 0.32652924954891205,
"reward_std": 0.17557984590530396,
"reward_std_all": 0.2862996757030487,
"rewards-max/forensics_tempsamp_iou_reward": 0.6046417355537415,
"rewards/forensics_tempsamp_iou_reward": 0.222829170525074,
"step": 1
},
{
"completion_length": 22.15625,
"epoch": 0.008368200836820083,
"grad_norm": 5.234766483306885,
"kl": 0.00051116943359375,
"learning_rate": 9.989539748953975e-07,
"loss": 0.0,
"reward": 0.3683357536792755,
"reward_std": 0.13757380843162537,
"reward_std_all": 0.2576821595430374,
"rewards-max/forensics_tempsamp_iou_reward": 0.9177520871162415,
"rewards/forensics_tempsamp_iou_reward": 0.28651174157857895,
"step": 2
},
{
"completion_length": 23.15625,
"epoch": 0.012552301255230125,
"grad_norm": 5.281967639923096,
"kl": 0.00266265869140625,
"learning_rate": 9.97907949790795e-07,
"loss": 0.0001,
"reward": 0.3160640448331833,
"reward_std": 0.12132176011800766,
"reward_std_all": 0.2726598381996155,
"rewards-max/forensics_tempsamp_iou_reward": 0.7190731763839722,
"rewards/forensics_tempsamp_iou_reward": 0.21248193085193634,
"step": 3
},
{
"completion_length": 20.15625,
"epoch": 0.016736401673640166,
"grad_norm": 5.507230281829834,
"kl": 0.00537109375,
"learning_rate": 9.968619246861925e-07,
"loss": 0.0002,
"reward": 0.28863998502492905,
"reward_std": 0.1915564090013504,
"reward_std_all": 0.31092436611652374,
"rewards-max/forensics_tempsamp_iou_reward": 0.653877317905426,
"rewards/forensics_tempsamp_iou_reward": 0.19376827776432037,
"step": 4
},
{
"completion_length": 21.640625,
"epoch": 0.02092050209205021,
"grad_norm": 4.8625383377075195,
"kl": 0.01629638671875,
"learning_rate": 9.9581589958159e-07,
"loss": 0.0007,
"reward": 0.23931505531072617,
"reward_std": 0.15258487313985825,
"reward_std_all": 0.32243405282497406,
"rewards-max/forensics_tempsamp_iou_reward": 0.7462074160575867,
"rewards/forensics_tempsamp_iou_reward": 0.14216922968626022,
"step": 5
},
{
"completion_length": 23.71875,
"epoch": 0.02510460251046025,
"grad_norm": 3.571162223815918,
"kl": 0.0460205078125,
"learning_rate": 9.947698744769873e-07,
"loss": 0.0018,
"reward": 0.3139348477125168,
"reward_std": 0.1908045932650566,
"reward_std_all": 0.3002701848745346,
"rewards-max/forensics_tempsamp_iou_reward": 0.8048895001411438,
"rewards/forensics_tempsamp_iou_reward": 0.21924392879009247,
"step": 6
},
{
"completion_length": 23.25,
"epoch": 0.029288702928870293,
"grad_norm": 3.347341775894165,
"kl": 0.0543212890625,
"learning_rate": 9.937238493723848e-07,
"loss": 0.0022,
"reward": 0.3601393550634384,
"reward_std": 0.1646009311079979,
"reward_std_all": 0.2766386419534683,
"rewards-max/forensics_tempsamp_iou_reward": 0.8728953301906586,
"rewards/forensics_tempsamp_iou_reward": 0.2707321047782898,
"step": 7
},
{
"completion_length": 22.125,
"epoch": 0.03347280334728033,
"grad_norm": 4.406038761138916,
"kl": 0.0810546875,
"learning_rate": 9.926778242677823e-07,
"loss": 0.0032,
"reward": 0.3438200205564499,
"reward_std": 0.11181015893816948,
"reward_std_all": 0.2602531537413597,
"rewards-max/forensics_tempsamp_iou_reward": 0.7579298615455627,
"rewards/forensics_tempsamp_iou_reward": 0.24873536825180054,
"step": 8
},
{
"completion_length": 22.96875,
"epoch": 0.03765690376569038,
"grad_norm": 2.576413154602051,
"kl": 0.104248046875,
"learning_rate": 9.916317991631798e-07,
"loss": 0.0042,
"reward": 0.32269978523254395,
"reward_std": 0.18512476235628128,
"reward_std_all": 0.30032122135162354,
"rewards-max/forensics_tempsamp_iou_reward": 0.9412148892879486,
"rewards/forensics_tempsamp_iou_reward": 0.23744626343250275,
"step": 9
},
{
"completion_length": 24.328125,
"epoch": 0.04184100418410042,
"grad_norm": 2.9359893798828125,
"kl": 0.18017578125,
"learning_rate": 9.905857740585774e-07,
"loss": 0.0072,
"reward": 0.2715313583612442,
"reward_std": 0.11592591553926468,
"reward_std_all": 0.2966170907020569,
"rewards-max/forensics_tempsamp_iou_reward": 0.7083293348550797,
"rewards/forensics_tempsamp_iou_reward": 0.17647871375083923,
"step": 10
},
{
"completion_length": 23.796875,
"epoch": 0.04602510460251046,
"grad_norm": 6.247899532318115,
"kl": 0.20556640625,
"learning_rate": 9.895397489539749e-07,
"loss": 0.0082,
"reward": 0.32468774914741516,
"reward_std": 0.1301351860165596,
"reward_std_all": 0.2775004357099533,
"rewards-max/forensics_tempsamp_iou_reward": 0.8256918787956238,
"rewards/forensics_tempsamp_iou_reward": 0.2337658777832985,
"step": 11
},
{
"completion_length": 25.203125,
"epoch": 0.0502092050209205,
"grad_norm": 4.460570812225342,
"kl": 0.20703125,
"learning_rate": 9.884937238493724e-07,
"loss": 0.0083,
"reward": 0.3567911684513092,
"reward_std": 0.16738688200712204,
"reward_std_all": 0.27143267542123795,
"rewards-max/forensics_tempsamp_iou_reward": 0.823586106300354,
"rewards/forensics_tempsamp_iou_reward": 0.2605914920568466,
"step": 12
},
{
"completion_length": 22.296875,
"epoch": 0.05439330543933055,
"grad_norm": 3.6588432788848877,
"kl": 0.22412109375,
"learning_rate": 9.8744769874477e-07,
"loss": 0.009,
"reward": 0.3232514262199402,
"reward_std": 0.17221737653017044,
"reward_std_all": 0.29511335492134094,
"rewards-max/forensics_tempsamp_iou_reward": 0.7741326987743378,
"rewards/forensics_tempsamp_iou_reward": 0.22805257141590118,
"step": 13
},
{
"completion_length": 25.4375,
"epoch": 0.058577405857740586,
"grad_norm": 3.558849334716797,
"kl": 0.17919921875,
"learning_rate": 9.864016736401674e-07,
"loss": 0.0072,
"reward": 0.2531444877386093,
"reward_std": 0.15694618225097656,
"reward_std_all": 0.31244952976703644,
"rewards-max/forensics_tempsamp_iou_reward": 0.4689566195011139,
"rewards/forensics_tempsamp_iou_reward": 0.15358345955610275,
"step": 14
},
{
"completion_length": 21.515625,
"epoch": 0.06276150627615062,
"grad_norm": 2.943216562271118,
"kl": 0.21484375,
"learning_rate": 9.85355648535565e-07,
"loss": 0.0086,
"reward": 0.31452177464962006,
"reward_std": 0.18208516389131546,
"reward_std_all": 0.29727230966091156,
"rewards-max/forensics_tempsamp_iou_reward": 0.7364770472049713,
"rewards/forensics_tempsamp_iou_reward": 0.21466580778360367,
"step": 15
},
{
"completion_length": 23.953125,
"epoch": 0.06694560669456066,
"grad_norm": 4.29056453704834,
"kl": 0.22412109375,
"learning_rate": 9.843096234309622e-07,
"loss": 0.009,
"reward": 0.29436418414115906,
"reward_std": 0.13953615725040436,
"reward_std_all": 0.2921256572008133,
"rewards-max/forensics_tempsamp_iou_reward": 0.6781406700611115,
"rewards/forensics_tempsamp_iou_reward": 0.1986214816570282,
"step": 16
},
{
"completion_length": 21.046875,
"epoch": 0.07112970711297072,
"grad_norm": 3.4024159908294678,
"kl": 0.236328125,
"learning_rate": 9.832635983263597e-07,
"loss": 0.0095,
"reward": 0.3372637778520584,
"reward_std": 0.18419578671455383,
"reward_std_all": 0.29400408267974854,
"rewards-max/forensics_tempsamp_iou_reward": 0.9562384784221649,
"rewards/forensics_tempsamp_iou_reward": 0.25259098410606384,
"step": 17
},
{
"completion_length": 21.828125,
"epoch": 0.07531380753138076,
"grad_norm": 3.3547465801239014,
"kl": 0.2431640625,
"learning_rate": 9.822175732217572e-07,
"loss": 0.0097,
"reward": 0.3338204622268677,
"reward_std": 0.14419721066951752,
"reward_std_all": 0.2832743674516678,
"rewards-max/forensics_tempsamp_iou_reward": 0.7712738513946533,
"rewards/forensics_tempsamp_iou_reward": 0.244672492146492,
"step": 18
},
{
"completion_length": 22.640625,
"epoch": 0.0794979079497908,
"grad_norm": 3.6062145233154297,
"kl": 0.23876953125,
"learning_rate": 9.811715481171548e-07,
"loss": 0.0096,
"reward": 0.2902381122112274,
"reward_std": 0.15652930736541748,
"reward_std_all": 0.3060254156589508,
"rewards-max/forensics_tempsamp_iou_reward": 0.6975690424442291,
"rewards/forensics_tempsamp_iou_reward": 0.19456204771995544,
"step": 19
},
{
"completion_length": 20.859375,
"epoch": 0.08368200836820083,
"grad_norm": 4.997276306152344,
"kl": 0.33984375,
"learning_rate": 9.801255230125523e-07,
"loss": 0.0136,
"reward": 0.32277560234069824,
"reward_std": 0.10894428193569183,
"reward_std_all": 0.26980940997600555,
"rewards-max/forensics_tempsamp_iou_reward": 0.7998393177986145,
"rewards/forensics_tempsamp_iou_reward": 0.22661571204662323,
"step": 20
},
{
"completion_length": 20.671875,
"epoch": 0.08786610878661087,
"grad_norm": 2.639582633972168,
"kl": 0.22314453125,
"learning_rate": 9.790794979079498e-07,
"loss": 0.0089,
"reward": 0.30074335634708405,
"reward_std": 0.20942280441522598,
"reward_std_all": 0.31690627336502075,
"rewards-max/forensics_tempsamp_iou_reward": 0.7201886177062988,
"rewards/forensics_tempsamp_iou_reward": 0.20385383069515228,
"step": 21
},
{
"completion_length": 20.1875,
"epoch": 0.09205020920502092,
"grad_norm": 3.4374496936798096,
"kl": 0.23291015625,
"learning_rate": 9.780334728033473e-07,
"loss": 0.0093,
"reward": 0.43312735855579376,
"reward_std": 0.19417206197977066,
"reward_std_all": 0.25543202459812164,
"rewards-max/forensics_tempsamp_iou_reward": 0.9062532484531403,
"rewards/forensics_tempsamp_iou_reward": 0.3543802946805954,
"step": 22
},
{
"completion_length": 21.21875,
"epoch": 0.09623430962343096,
"grad_norm": 3.527658462524414,
"kl": 0.2197265625,
"learning_rate": 9.769874476987448e-07,
"loss": 0.0088,
"reward": 0.389848455786705,
"reward_std": 0.15388014167547226,
"reward_std_all": 0.2567143738269806,
"rewards-max/forensics_tempsamp_iou_reward": 0.7790012359619141,
"rewards/forensics_tempsamp_iou_reward": 0.2948019728064537,
"step": 23
},
{
"completion_length": 19.765625,
"epoch": 0.100418410041841,
"grad_norm": 2.6915128231048584,
"kl": 0.20556640625,
"learning_rate": 9.759414225941423e-07,
"loss": 0.0082,
"reward": 0.37542735040187836,
"reward_std": 0.19585968554019928,
"reward_std_all": 0.28278176486492157,
"rewards-max/forensics_tempsamp_iou_reward": 0.7738228142261505,
"rewards/forensics_tempsamp_iou_reward": 0.2851048409938812,
"step": 24
},
{
"completion_length": 20.171875,
"epoch": 0.10460251046025104,
"grad_norm": 2.734421491622925,
"kl": 0.21484375,
"learning_rate": 9.748953974895398e-07,
"loss": 0.0086,
"reward": 0.3591455817222595,
"reward_std": 0.21876687556505203,
"reward_std_all": 0.3020075857639313,
"rewards-max/forensics_tempsamp_iou_reward": 0.8293378055095673,
"rewards/forensics_tempsamp_iou_reward": 0.27461325377225876,
"step": 25
},
{
"completion_length": 20.25,
"epoch": 0.1087866108786611,
"grad_norm": 2.7922000885009766,
"kl": 0.20703125,
"learning_rate": 9.738493723849371e-07,
"loss": 0.0083,
"reward": 0.3647250831127167,
"reward_std": 0.1600918173789978,
"reward_std_all": 0.2729649245738983,
"rewards-max/forensics_tempsamp_iou_reward": 0.8378541469573975,
"rewards/forensics_tempsamp_iou_reward": 0.27459298074245453,
"step": 26
},
{
"completion_length": 20.71875,
"epoch": 0.11297071129707113,
"grad_norm": 2.6484546661376953,
"kl": 0.20849609375,
"learning_rate": 9.728033472803347e-07,
"loss": 0.0083,
"reward": 0.3264050930738449,
"reward_std": 0.13920806348323822,
"reward_std_all": 0.2765499949455261,
"rewards-max/forensics_tempsamp_iou_reward": 0.7107179164886475,
"rewards/forensics_tempsamp_iou_reward": 0.22992640733718872,
"step": 27
},
{
"completion_length": 21.484375,
"epoch": 0.11715481171548117,
"grad_norm": 2.6283891201019287,
"kl": 0.2021484375,
"learning_rate": 9.717573221757322e-07,
"loss": 0.0081,
"reward": 0.4493083208799362,
"reward_std": 0.20125902444124222,
"reward_std_all": 0.25433163344860077,
"rewards-max/forensics_tempsamp_iou_reward": 0.8362603485584259,
"rewards/forensics_tempsamp_iou_reward": 0.3686801940202713,
"step": 28
},
{
"completion_length": 22.3125,
"epoch": 0.12133891213389121,
"grad_norm": 2.4925622940063477,
"kl": 0.212890625,
"learning_rate": 9.707112970711297e-07,
"loss": 0.0085,
"reward": 0.2992720305919647,
"reward_std": 0.11565325036644936,
"reward_std_all": 0.2809276580810547,
"rewards-max/forensics_tempsamp_iou_reward": 0.6758152544498444,
"rewards/forensics_tempsamp_iou_reward": 0.19789063185453415,
"step": 29
},
{
"completion_length": 22.1875,
"epoch": 0.12552301255230125,
"grad_norm": 2.4201557636260986,
"kl": 0.2236328125,
"learning_rate": 9.696652719665272e-07,
"loss": 0.0089,
"reward": 0.38816623389720917,
"reward_std": 0.18402977287769318,
"reward_std_all": 0.26735904812812805,
"rewards-max/forensics_tempsamp_iou_reward": 0.7545492649078369,
"rewards/forensics_tempsamp_iou_reward": 0.29266615211963654,
"step": 30
},
{
"completion_length": 21.4375,
"epoch": 0.1297071129707113,
"grad_norm": 3.0774667263031006,
"kl": 0.18212890625,
"learning_rate": 9.686192468619247e-07,
"loss": 0.0073,
"reward": 0.443069651722908,
"reward_std": 0.16818813979625702,
"reward_std_all": 0.23900970071554184,
"rewards-max/forensics_tempsamp_iou_reward": 0.9521180987358093,
"rewards/forensics_tempsamp_iou_reward": 0.36849141120910645,
"step": 31
},
{
"completion_length": 25.34375,
"epoch": 0.13389121338912133,
"grad_norm": 2.5716047286987305,
"kl": 0.203125,
"learning_rate": 9.675732217573222e-07,
"loss": 0.0081,
"reward": 0.3637811839580536,
"reward_std": 0.16480311006307602,
"reward_std_all": 0.27341093122959137,
"rewards-max/forensics_tempsamp_iou_reward": 0.7260869741439819,
"rewards/forensics_tempsamp_iou_reward": 0.26909536123275757,
"step": 32
},
{
"completion_length": 22.28125,
"epoch": 0.13807531380753138,
"grad_norm": 3.069857120513916,
"kl": 0.2158203125,
"learning_rate": 9.665271966527197e-07,
"loss": 0.0086,
"reward": 0.34146536886692047,
"reward_std": 0.16381171345710754,
"reward_std_all": 0.27423295378685,
"rewards-max/forensics_tempsamp_iou_reward": 0.7755928039550781,
"rewards/forensics_tempsamp_iou_reward": 0.24398396909236908,
"step": 33
},
{
"completion_length": 22.5625,
"epoch": 0.14225941422594143,
"grad_norm": 3.188498020172119,
"kl": 0.20361328125,
"learning_rate": 9.654811715481172e-07,
"loss": 0.0081,
"reward": 0.3494633287191391,
"reward_std": 0.14528657495975494,
"reward_std_all": 0.26684587448835373,
"rewards-max/forensics_tempsamp_iou_reward": 0.747743546962738,
"rewards/forensics_tempsamp_iou_reward": 0.2746489755809307,
"step": 34
},
{
"completion_length": 20.5625,
"epoch": 0.14644351464435146,
"grad_norm": 2.5757181644439697,
"kl": 0.193359375,
"learning_rate": 9.644351464435145e-07,
"loss": 0.0077,
"reward": 0.33570457994937897,
"reward_std": 0.1109573096036911,
"reward_std_all": 0.26059287786483765,
"rewards-max/forensics_tempsamp_iou_reward": 0.7379959523677826,
"rewards/forensics_tempsamp_iou_reward": 0.23691926896572113,
"step": 35
},
{
"completion_length": 20.484375,
"epoch": 0.1506276150627615,
"grad_norm": 2.52881121635437,
"kl": 0.19189453125,
"learning_rate": 9.63389121338912e-07,
"loss": 0.0077,
"reward": 0.43021704256534576,
"reward_std": 0.14606455713510513,
"reward_std_all": 0.23222054541110992,
"rewards-max/forensics_tempsamp_iou_reward": 0.8082605004310608,
"rewards/forensics_tempsamp_iou_reward": 0.3387262672185898,
"step": 36
},
{
"completion_length": 20.59375,
"epoch": 0.15481171548117154,
"grad_norm": 4.086231231689453,
"kl": 0.24072265625,
"learning_rate": 9.623430962343096e-07,
"loss": 0.0097,
"reward": 0.48338504135608673,
"reward_std": 0.1683856025338173,
"reward_std_all": 0.22045395523309708,
"rewards-max/forensics_tempsamp_iou_reward": 0.9292173385620117,
"rewards/forensics_tempsamp_iou_reward": 0.4046168029308319,
"step": 37
},
{
"completion_length": 20.703125,
"epoch": 0.1589958158995816,
"grad_norm": 2.329577922821045,
"kl": 0.20703125,
"learning_rate": 9.61297071129707e-07,
"loss": 0.0083,
"reward": 0.41878238320350647,
"reward_std": 0.13207846879959106,
"reward_std_all": 0.23971392214298248,
"rewards-max/forensics_tempsamp_iou_reward": 0.8918798565864563,
"rewards/forensics_tempsamp_iou_reward": 0.34818150103092194,
"step": 38
},
{
"completion_length": 22.125,
"epoch": 0.16317991631799164,
"grad_norm": 3.3772428035736084,
"kl": 0.2021484375,
"learning_rate": 9.602510460251046e-07,
"loss": 0.0081,
"reward": 0.34119078516960144,
"reward_std": 0.1733928769826889,
"reward_std_all": 0.2866531014442444,
"rewards-max/forensics_tempsamp_iou_reward": 0.8822580873966217,
"rewards/forensics_tempsamp_iou_reward": 0.2573893368244171,
"step": 39
},
{
"completion_length": 21.0,
"epoch": 0.16736401673640167,
"grad_norm": 2.882540464401245,
"kl": 0.22216796875,
"learning_rate": 9.592050209205021e-07,
"loss": 0.0089,
"reward": 0.3684040457010269,
"reward_std": 0.20962397009134293,
"reward_std_all": 0.29534460604190826,
"rewards-max/forensics_tempsamp_iou_reward": 0.8444999158382416,
"rewards/forensics_tempsamp_iou_reward": 0.27859051525592804,
"step": 40
},
{
"completion_length": 21.578125,
"epoch": 0.17154811715481172,
"grad_norm": 2.3474440574645996,
"kl": 0.21826171875,
"learning_rate": 9.581589958158996e-07,
"loss": 0.0087,
"reward": 0.27565792202949524,
"reward_std": 0.14532283693552017,
"reward_std_all": 0.3018806129693985,
"rewards-max/forensics_tempsamp_iou_reward": 0.6841050982475281,
"rewards/forensics_tempsamp_iou_reward": 0.17951739951968193,
"step": 41
},
{
"completion_length": 21.3125,
"epoch": 0.17573221757322174,
"grad_norm": 3.983370542526245,
"kl": 0.1826171875,
"learning_rate": 9.571129707112971e-07,
"loss": 0.0073,
"reward": 0.2732257917523384,
"reward_std": 0.128234650939703,
"reward_std_all": 0.29850438237190247,
"rewards-max/forensics_tempsamp_iou_reward": 0.5697210133075714,
"rewards/forensics_tempsamp_iou_reward": 0.17275435477495193,
"step": 42
},
{
"completion_length": 19.515625,
"epoch": 0.1799163179916318,
"grad_norm": 2.832289934158325,
"kl": 0.20556640625,
"learning_rate": 9.560669456066946e-07,
"loss": 0.0082,
"reward": 0.3517254590988159,
"reward_std": 0.18037428706884384,
"reward_std_all": 0.2857476770877838,
"rewards-max/forensics_tempsamp_iou_reward": 0.9071855247020721,
"rewards/forensics_tempsamp_iou_reward": 0.2747180610895157,
"step": 43
},
{
"completion_length": 20.5625,
"epoch": 0.18410041841004185,
"grad_norm": 2.7222142219543457,
"kl": 0.185546875,
"learning_rate": 9.55020920502092e-07,
"loss": 0.0074,
"reward": 0.38121676445007324,
"reward_std": 0.160438671708107,
"reward_std_all": 0.26082316786050797,
"rewards-max/forensics_tempsamp_iou_reward": 0.9112547934055328,
"rewards/forensics_tempsamp_iou_reward": 0.2993100583553314,
"step": 44
},
{
"completion_length": 20.90625,
"epoch": 0.18828451882845187,
"grad_norm": 2.500004768371582,
"kl": 0.2041015625,
"learning_rate": 9.539748953974895e-07,
"loss": 0.0082,
"reward": 0.4045710861682892,
"reward_std": 0.15461453795433044,
"reward_std_all": 0.2456449717283249,
"rewards-max/forensics_tempsamp_iou_reward": 0.8004147410392761,
"rewards/forensics_tempsamp_iou_reward": 0.3073156625032425,
"step": 45
},
{
"completion_length": 20.625,
"epoch": 0.19246861924686193,
"grad_norm": 2.2287964820861816,
"kl": 0.1826171875,
"learning_rate": 9.52928870292887e-07,
"loss": 0.0073,
"reward": 0.2748546749353409,
"reward_std": 0.17132844775915146,
"reward_std_all": 0.3189420700073242,
"rewards-max/forensics_tempsamp_iou_reward": 0.9095750749111176,
"rewards/forensics_tempsamp_iou_reward": 0.19107332825660706,
"step": 46
},
{
"completion_length": 20.484375,
"epoch": 0.19665271966527198,
"grad_norm": 2.7088253498077393,
"kl": 0.16357421875,
"learning_rate": 9.518828451882844e-07,
"loss": 0.0065,
"reward": 0.3824062794446945,
"reward_std": 0.20348912477493286,
"reward_std_all": 0.27947454154491425,
"rewards-max/forensics_tempsamp_iou_reward": 0.9206356704235077,
"rewards/forensics_tempsamp_iou_reward": 0.2945864647626877,
"step": 47
},
{
"completion_length": 22.5625,
"epoch": 0.200836820083682,
"grad_norm": 3.575388193130493,
"kl": 0.185546875,
"learning_rate": 9.508368200836819e-07,
"loss": 0.0074,
"reward": 0.30769166350364685,
"reward_std": 0.18517035990953445,
"reward_std_all": 0.30867528915405273,
"rewards-max/forensics_tempsamp_iou_reward": 0.8546599447727203,
"rewards/forensics_tempsamp_iou_reward": 0.2166326716542244,
"step": 48
},
{
"completion_length": 21.296875,
"epoch": 0.20502092050209206,
"grad_norm": 3.062904119491577,
"kl": 0.18994140625,
"learning_rate": 9.497907949790794e-07,
"loss": 0.0076,
"reward": 0.36563727259635925,
"reward_std": 0.19126105308532715,
"reward_std_all": 0.28109224140644073,
"rewards-max/forensics_tempsamp_iou_reward": 0.7904012799263,
"rewards/forensics_tempsamp_iou_reward": 0.27382370829582214,
"step": 49
},
{
"completion_length": 20.46875,
"epoch": 0.20920502092050208,
"grad_norm": 2.5525851249694824,
"kl": 0.20166015625,
"learning_rate": 9.487447698744769e-07,
"loss": 0.0081,
"reward": 0.345081090927124,
"reward_std": 0.15642611682415009,
"reward_std_all": 0.27373331785202026,
"rewards-max/forensics_tempsamp_iou_reward": 0.7981561422348022,
"rewards/forensics_tempsamp_iou_reward": 0.2621668726205826,
"step": 50
},
{
"completion_length": 20.328125,
"epoch": 0.21338912133891214,
"grad_norm": 2.0393307209014893,
"kl": 0.19580078125,
"learning_rate": 9.476987447698744e-07,
"loss": 0.0078,
"reward": 0.3610183149576187,
"reward_std": 0.20577482879161835,
"reward_std_all": 0.28537262976169586,
"rewards-max/forensics_tempsamp_iou_reward": 0.7973324954509735,
"rewards/forensics_tempsamp_iou_reward": 0.2715054154396057,
"step": 51
},
{
"completion_length": 21.4375,
"epoch": 0.2175732217573222,
"grad_norm": 3.2122840881347656,
"kl": 0.16064453125,
"learning_rate": 9.466527196652719e-07,
"loss": 0.0064,
"reward": 0.3488035351037979,
"reward_std": 0.126681886613369,
"reward_std_all": 0.26765692234039307,
"rewards-max/forensics_tempsamp_iou_reward": 0.8196521699428558,
"rewards/forensics_tempsamp_iou_reward": 0.257691890001297,
"step": 52
},
{
"completion_length": 24.125,
"epoch": 0.2217573221757322,
"grad_norm": 2.167557716369629,
"kl": 0.1611328125,
"learning_rate": 9.456066945606695e-07,
"loss": 0.0064,
"reward": 0.34760577976703644,
"reward_std": 0.15085458755493164,
"reward_std_all": 0.2770399749279022,
"rewards-max/forensics_tempsamp_iou_reward": 0.7991946935653687,
"rewards/forensics_tempsamp_iou_reward": 0.25365591049194336,
"step": 53
},
{
"completion_length": 21.859375,
"epoch": 0.22594142259414227,
"grad_norm": 3.0633938312530518,
"kl": 0.1796875,
"learning_rate": 9.44560669456067e-07,
"loss": 0.0072,
"reward": 0.44028422236442566,
"reward_std": 0.24309130012989044,
"reward_std_all": 0.2743610739707947,
"rewards-max/forensics_tempsamp_iou_reward": 0.9340702295303345,
"rewards/forensics_tempsamp_iou_reward": 0.36326864361763,
"step": 54
},
{
"completion_length": 23.3125,
"epoch": 0.2301255230125523,
"grad_norm": 3.2539944648742676,
"kl": 0.21533203125,
"learning_rate": 9.435146443514645e-07,
"loss": 0.0086,
"reward": 0.32758061587810516,
"reward_std": 0.16607342660427094,
"reward_std_all": 0.28696076571941376,
"rewards-max/forensics_tempsamp_iou_reward": 0.7835468649864197,
"rewards/forensics_tempsamp_iou_reward": 0.22972016781568527,
"step": 55
},
{
"completion_length": 23.734375,
"epoch": 0.23430962343096234,
"grad_norm": 2.7632741928100586,
"kl": 0.1982421875,
"learning_rate": 9.424686192468618e-07,
"loss": 0.0079,
"reward": 0.3129670172929764,
"reward_std": 0.1564241498708725,
"reward_std_all": 0.28786371648311615,
"rewards-max/forensics_tempsamp_iou_reward": 0.6681990623474121,
"rewards/forensics_tempsamp_iou_reward": 0.2126583531498909,
"step": 56
},
{
"completion_length": 22.25,
"epoch": 0.2384937238493724,
"grad_norm": 3.011981964111328,
"kl": 0.23046875,
"learning_rate": 9.414225941422593e-07,
"loss": 0.0092,
"reward": 0.42676813900470734,
"reward_std": 0.19988757371902466,
"reward_std_all": 0.2643730863928795,
"rewards-max/forensics_tempsamp_iou_reward": 0.9045308828353882,
"rewards/forensics_tempsamp_iou_reward": 0.35181228816509247,
"step": 57
},
{
"completion_length": 27.71875,
"epoch": 0.24267782426778242,
"grad_norm": 2.5664639472961426,
"kl": 0.16943359375,
"learning_rate": 9.403765690376568e-07,
"loss": 0.0068,
"reward": 0.35306112468242645,
"reward_std": 0.2114599049091339,
"reward_std_all": 0.29966430366039276,
"rewards-max/forensics_tempsamp_iou_reward": 0.8806408643722534,
"rewards/forensics_tempsamp_iou_reward": 0.2669929787516594,
"step": 58
},
{
"completion_length": 19.578125,
"epoch": 0.24686192468619247,
"grad_norm": 2.5698797702789307,
"kl": 0.197265625,
"learning_rate": 9.393305439330543e-07,
"loss": 0.0079,
"reward": 0.3327927142381668,
"reward_std": 0.18998057395219803,
"reward_std_all": 0.2902723699808121,
"rewards-max/forensics_tempsamp_iou_reward": 0.7480018734931946,
"rewards/forensics_tempsamp_iou_reward": 0.24441038817167282,
"step": 59
},
{
"completion_length": 22.890625,
"epoch": 0.2510460251046025,
"grad_norm": 2.887362003326416,
"kl": 0.17041015625,
"learning_rate": 9.382845188284518e-07,
"loss": 0.0068,
"reward": 0.3334708958864212,
"reward_std": 0.16507523506879807,
"reward_std_all": 0.2874464839696884,
"rewards-max/forensics_tempsamp_iou_reward": 0.8347360491752625,
"rewards/forensics_tempsamp_iou_reward": 0.24782820045948029,
"step": 60
},
{
"completion_length": 22.5,
"epoch": 0.25523012552301255,
"grad_norm": 2.0620532035827637,
"kl": 0.173828125,
"learning_rate": 9.372384937238493e-07,
"loss": 0.007,
"reward": 0.42092128098011017,
"reward_std": 0.19240941107273102,
"reward_std_all": 0.2560470253229141,
"rewards-max/forensics_tempsamp_iou_reward": 0.8779537677764893,
"rewards/forensics_tempsamp_iou_reward": 0.3297566920518875,
"step": 61
},
{
"completion_length": 22.09375,
"epoch": 0.2594142259414226,
"grad_norm": 3.0325570106506348,
"kl": 0.17138671875,
"learning_rate": 9.361924686192469e-07,
"loss": 0.0068,
"reward": 0.32446475327014923,
"reward_std": 0.15927113592624664,
"reward_std_all": 0.28787219524383545,
"rewards-max/forensics_tempsamp_iou_reward": 0.8384105563163757,
"rewards/forensics_tempsamp_iou_reward": 0.24168004095554352,
"step": 62
},
{
"completion_length": 22.71875,
"epoch": 0.26359832635983266,
"grad_norm": 2.446808338165283,
"kl": 0.1640625,
"learning_rate": 9.351464435146444e-07,
"loss": 0.0065,
"reward": 0.43146589398384094,
"reward_std": 0.1586010381579399,
"reward_std_all": 0.23731720447540283,
"rewards-max/forensics_tempsamp_iou_reward": 0.8308470547199249,
"rewards/forensics_tempsamp_iou_reward": 0.3464578837156296,
"step": 63
},
{
"completion_length": 24.09375,
"epoch": 0.26778242677824265,
"grad_norm": 2.342294216156006,
"kl": 0.18115234375,
"learning_rate": 9.341004184100419e-07,
"loss": 0.0072,
"reward": 0.40610258281230927,
"reward_std": 0.22527223080396652,
"reward_std_all": 0.2842797487974167,
"rewards-max/forensics_tempsamp_iou_reward": 0.8654806911945343,
"rewards/forensics_tempsamp_iou_reward": 0.3188140094280243,
"step": 64
},
{
"completion_length": 21.984375,
"epoch": 0.2719665271966527,
"grad_norm": 2.130748748779297,
"kl": 0.18017578125,
"learning_rate": 9.330543933054393e-07,
"loss": 0.0072,
"reward": 0.3678058087825775,
"reward_std": 0.17842144519090652,
"reward_std_all": 0.27484454959630966,
"rewards-max/forensics_tempsamp_iou_reward": 0.7750593423843384,
"rewards/forensics_tempsamp_iou_reward": 0.2718849629163742,
"step": 65
},
{
"completion_length": 22.40625,
"epoch": 0.27615062761506276,
"grad_norm": 2.604670763015747,
"kl": 0.16796875,
"learning_rate": 9.320083682008367e-07,
"loss": 0.0067,
"reward": 0.3262310028076172,
"reward_std": 0.13436225056648254,
"reward_std_all": 0.2714546173810959,
"rewards-max/forensics_tempsamp_iou_reward": 0.84375,
"rewards/forensics_tempsamp_iou_reward": 0.2329244315624237,
"step": 66
},
{
"completion_length": 24.125,
"epoch": 0.2803347280334728,
"grad_norm": 2.806077718734741,
"kl": 0.1728515625,
"learning_rate": 9.309623430962342e-07,
"loss": 0.0069,
"reward": 0.3901044875383377,
"reward_std": 0.11893605440855026,
"reward_std_all": 0.2361702024936676,
"rewards-max/forensics_tempsamp_iou_reward": 0.8212221264839172,
"rewards/forensics_tempsamp_iou_reward": 0.2957017570734024,
"step": 67
},
{
"completion_length": 21.75,
"epoch": 0.28451882845188287,
"grad_norm": 2.9567649364471436,
"kl": 0.20068359375,
"learning_rate": 9.299163179916317e-07,
"loss": 0.008,
"reward": 0.3170981705188751,
"reward_std": 0.15230406820774078,
"reward_std_all": 0.28708404302597046,
"rewards-max/forensics_tempsamp_iou_reward": 0.7831559777259827,
"rewards/forensics_tempsamp_iou_reward": 0.2246413677930832,
"step": 68
},
{
"completion_length": 25.109375,
"epoch": 0.28870292887029286,
"grad_norm": 2.6112756729125977,
"kl": 0.1640625,
"learning_rate": 9.288702928870292e-07,
"loss": 0.0066,
"reward": 0.39385800063610077,
"reward_std": 0.16494587063789368,
"reward_std_all": 0.2597482353448868,
"rewards-max/forensics_tempsamp_iou_reward": 0.8362612724304199,
"rewards/forensics_tempsamp_iou_reward": 0.29728274047374725,
"step": 69
},
{
"completion_length": 24.5625,
"epoch": 0.2928870292887029,
"grad_norm": 3.015868902206421,
"kl": 0.1884765625,
"learning_rate": 9.278242677824268e-07,
"loss": 0.0075,
"reward": 0.35404084622859955,
"reward_std": 0.1602024808526039,
"reward_std_all": 0.27648378908634186,
"rewards-max/forensics_tempsamp_iou_reward": 0.8639806807041168,
"rewards/forensics_tempsamp_iou_reward": 0.2686200439929962,
"step": 70
},
{
"completion_length": 22.296875,
"epoch": 0.29707112970711297,
"grad_norm": 3.8776068687438965,
"kl": 0.17529296875,
"learning_rate": 9.267782426778243e-07,
"loss": 0.007,
"reward": 0.3987846225500107,
"reward_std": 0.15969102829694748,
"reward_std_all": 0.2560664564371109,
"rewards-max/forensics_tempsamp_iou_reward": 0.8402622044086456,
"rewards/forensics_tempsamp_iou_reward": 0.31750574707984924,
"step": 71
},
{
"completion_length": 23.109375,
"epoch": 0.301255230125523,
"grad_norm": 2.4891602993011475,
"kl": 0.17041015625,
"learning_rate": 9.257322175732218e-07,
"loss": 0.0068,
"reward": 0.45526957511901855,
"reward_std": 0.170804463326931,
"reward_std_all": 0.23407982289791107,
"rewards-max/forensics_tempsamp_iou_reward": 0.8395125567913055,
"rewards/forensics_tempsamp_iou_reward": 0.368361160159111,
"step": 72
},
{
"completion_length": 21.640625,
"epoch": 0.3054393305439331,
"grad_norm": 2.499086856842041,
"kl": 0.185546875,
"learning_rate": 9.246861924686192e-07,
"loss": 0.0074,
"reward": 0.35304781794548035,
"reward_std": 0.1780683621764183,
"reward_std_all": 0.2834163010120392,
"rewards-max/forensics_tempsamp_iou_reward": 0.8075635135173798,
"rewards/forensics_tempsamp_iou_reward": 0.2588299810886383,
"step": 73
},
{
"completion_length": 21.140625,
"epoch": 0.30962343096234307,
"grad_norm": 2.864497423171997,
"kl": 0.20263671875,
"learning_rate": 9.236401673640167e-07,
"loss": 0.0081,
"reward": 0.41607198119163513,
"reward_std": 0.11527368053793907,
"reward_std_all": 0.2349308580160141,
"rewards-max/forensics_tempsamp_iou_reward": 0.9594540596008301,
"rewards/forensics_tempsamp_iou_reward": 0.35371387004852295,
"step": 74
},
{
"completion_length": 19.625,
"epoch": 0.3138075313807531,
"grad_norm": 3.3631751537323,
"kl": 0.193359375,
"learning_rate": 9.225941422594142e-07,
"loss": 0.0077,
"reward": 0.3692771941423416,
"reward_std": 0.17589308321475983,
"reward_std_all": 0.2745033800601959,
"rewards-max/forensics_tempsamp_iou_reward": 0.7776351869106293,
"rewards/forensics_tempsamp_iou_reward": 0.2821844816207886,
"step": 75
},
{
"completion_length": 25.265625,
"epoch": 0.3179916317991632,
"grad_norm": 3.2122628688812256,
"kl": 0.22412109375,
"learning_rate": 9.215481171548116e-07,
"loss": 0.009,
"reward": 0.41379180550575256,
"reward_std": 0.1993199661374092,
"reward_std_all": 0.26996180415153503,
"rewards-max/forensics_tempsamp_iou_reward": 0.9028154909610748,
"rewards/forensics_tempsamp_iou_reward": 0.3367422968149185,
"step": 76
},
{
"completion_length": 20.4375,
"epoch": 0.32217573221757323,
"grad_norm": 3.6009676456451416,
"kl": 0.21826171875,
"learning_rate": 9.205020920502091e-07,
"loss": 0.0087,
"reward": 0.4053776115179062,
"reward_std": 0.14346328005194664,
"reward_std_all": 0.23989441990852356,
"rewards-max/forensics_tempsamp_iou_reward": 0.8589051961898804,
"rewards/forensics_tempsamp_iou_reward": 0.31789630651474,
"step": 77
},
{
"completion_length": 22.46875,
"epoch": 0.3263598326359833,
"grad_norm": 2.690593957901001,
"kl": 0.18798828125,
"learning_rate": 9.194560669456066e-07,
"loss": 0.0075,
"reward": 0.33698494732379913,
"reward_std": 0.17444373667240143,
"reward_std_all": 0.2880709171295166,
"rewards-max/forensics_tempsamp_iou_reward": 0.7623375058174133,
"rewards/forensics_tempsamp_iou_reward": 0.24299560487270355,
"step": 78
},
{
"completion_length": 22.09375,
"epoch": 0.3305439330543933,
"grad_norm": 3.2760088443756104,
"kl": 0.1845703125,
"learning_rate": 9.184100418410042e-07,
"loss": 0.0074,
"reward": 0.3815077841281891,
"reward_std": 0.1497289165854454,
"reward_std_all": 0.2607156038284302,
"rewards-max/forensics_tempsamp_iou_reward": 0.8757800459861755,
"rewards/forensics_tempsamp_iou_reward": 0.28850650787353516,
"step": 79
},
{
"completion_length": 21.171875,
"epoch": 0.33472803347280333,
"grad_norm": 2.410973072052002,
"kl": 0.1982421875,
"learning_rate": 9.173640167364017e-07,
"loss": 0.0079,
"reward": 0.4528784453868866,
"reward_std": 0.17536590993404388,
"reward_std_all": 0.23991592228412628,
"rewards-max/forensics_tempsamp_iou_reward": 0.940642923116684,
"rewards/forensics_tempsamp_iou_reward": 0.37946097552776337,
"step": 80
},
{
"completion_length": 20.765625,
"epoch": 0.3389121338912134,
"grad_norm": 2.683236598968506,
"kl": 0.21875,
"learning_rate": 9.163179916317991e-07,
"loss": 0.0088,
"reward": 0.37575775384902954,
"reward_std": 0.17851759493350983,
"reward_std_all": 0.2725459188222885,
"rewards-max/forensics_tempsamp_iou_reward": 0.8715241849422455,
"rewards/forensics_tempsamp_iou_reward": 0.2849694490432739,
"step": 81
},
{
"completion_length": 22.75,
"epoch": 0.34309623430962344,
"grad_norm": 3.1503398418426514,
"kl": 0.2119140625,
"learning_rate": 9.152719665271966e-07,
"loss": 0.0085,
"reward": 0.3728176951408386,
"reward_std": 0.13261707872152328,
"reward_std_all": 0.2619427889585495,
"rewards-max/forensics_tempsamp_iou_reward": 0.8797271847724915,
"rewards/forensics_tempsamp_iou_reward": 0.286740317940712,
"step": 82
},
{
"completion_length": 20.15625,
"epoch": 0.3472803347280335,
"grad_norm": 3.081580638885498,
"kl": 0.20458984375,
"learning_rate": 9.142259414225941e-07,
"loss": 0.0082,
"reward": 0.38699860870838165,
"reward_std": 0.15666668117046356,
"reward_std_all": 0.26290811598300934,
"rewards-max/forensics_tempsamp_iou_reward": 0.7808124423027039,
"rewards/forensics_tempsamp_iou_reward": 0.2996341437101364,
"step": 83
},
{
"completion_length": 23.84375,
"epoch": 0.3514644351464435,
"grad_norm": 2.8324291706085205,
"kl": 0.20263671875,
"learning_rate": 9.131799163179916e-07,
"loss": 0.0081,
"reward": 0.36178258061408997,
"reward_std": 0.18250061571598053,
"reward_std_all": 0.28149572014808655,
"rewards-max/forensics_tempsamp_iou_reward": 0.9591636657714844,
"rewards/forensics_tempsamp_iou_reward": 0.27068886160850525,
"step": 84
},
{
"completion_length": 21.09375,
"epoch": 0.35564853556485354,
"grad_norm": 3.804971933364868,
"kl": 0.26171875,
"learning_rate": 9.121338912133891e-07,
"loss": 0.0105,
"reward": 0.44546370208263397,
"reward_std": 0.1519039273262024,
"reward_std_all": 0.2349248304963112,
"rewards-max/forensics_tempsamp_iou_reward": 0.8869717419147491,
"rewards/forensics_tempsamp_iou_reward": 0.374598428606987,
"step": 85
},
{
"completion_length": 20.75,
"epoch": 0.3598326359832636,
"grad_norm": 2.7997212409973145,
"kl": 0.24267578125,
"learning_rate": 9.110878661087865e-07,
"loss": 0.0097,
"reward": 0.3559633642435074,
"reward_std": 0.16267015039920807,
"reward_std_all": 0.27574947476387024,
"rewards-max/forensics_tempsamp_iou_reward": 0.7843071818351746,
"rewards/forensics_tempsamp_iou_reward": 0.26181748509407043,
"step": 86
},
{
"completion_length": 23.34375,
"epoch": 0.36401673640167365,
"grad_norm": 2.095106363296509,
"kl": 0.20849609375,
"learning_rate": 9.10041841004184e-07,
"loss": 0.0083,
"reward": 0.3718299865722656,
"reward_std": 0.11165961623191833,
"reward_std_all": 0.24783912301063538,
"rewards-max/forensics_tempsamp_iou_reward": 0.651852160692215,
"rewards/forensics_tempsamp_iou_reward": 0.27822286635637283,
"step": 87
},
{
"completion_length": 19.75,
"epoch": 0.3682008368200837,
"grad_norm": 2.4638006687164307,
"kl": 0.19921875,
"learning_rate": 9.089958158995816e-07,
"loss": 0.008,
"reward": 0.4299616515636444,
"reward_std": 0.13000230491161346,
"reward_std_all": 0.22978141903877258,
"rewards-max/forensics_tempsamp_iou_reward": 0.8977851867675781,
"rewards/forensics_tempsamp_iou_reward": 0.34900718927383423,
"step": 88
},
{
"completion_length": 19.890625,
"epoch": 0.3723849372384937,
"grad_norm": 2.41025972366333,
"kl": 0.21435546875,
"learning_rate": 9.07949790794979e-07,
"loss": 0.0086,
"reward": 0.4060441702604294,
"reward_std": 0.17378535866737366,
"reward_std_all": 0.25658953189849854,
"rewards-max/forensics_tempsamp_iou_reward": 0.7541440725326538,
"rewards/forensics_tempsamp_iou_reward": 0.3128507435321808,
"step": 89
},
{
"completion_length": 21.0,
"epoch": 0.37656903765690375,
"grad_norm": 2.7619025707244873,
"kl": 0.22802734375,
"learning_rate": 9.069037656903765e-07,
"loss": 0.0091,
"reward": 0.4481276720762253,
"reward_std": 0.1563880518078804,
"reward_std_all": 0.23036200553178787,
"rewards-max/forensics_tempsamp_iou_reward": 0.7761400938034058,
"rewards/forensics_tempsamp_iou_reward": 0.35051658749580383,
"step": 90
},
{
"completion_length": 22.015625,
"epoch": 0.3807531380753138,
"grad_norm": 2.7098612785339355,
"kl": 0.18798828125,
"learning_rate": 9.05857740585774e-07,
"loss": 0.0075,
"reward": 0.32770682871341705,
"reward_std": 0.13553054630756378,
"reward_std_all": 0.2789801359176636,
"rewards-max/forensics_tempsamp_iou_reward": 0.7694220244884491,
"rewards/forensics_tempsamp_iou_reward": 0.23242538422346115,
"step": 91
},
{
"completion_length": 22.046875,
"epoch": 0.38493723849372385,
"grad_norm": 2.859090805053711,
"kl": 0.20654296875,
"learning_rate": 9.048117154811715e-07,
"loss": 0.0083,
"reward": 0.42070935666561127,
"reward_std": 0.1405000127851963,
"reward_std_all": 0.24001451581716537,
"rewards-max/forensics_tempsamp_iou_reward": 0.8833147883415222,
"rewards/forensics_tempsamp_iou_reward": 0.3415044844150543,
"step": 92
},
{
"completion_length": 21.046875,
"epoch": 0.3891213389121339,
"grad_norm": 2.689154624938965,
"kl": 0.197265625,
"learning_rate": 9.03765690376569e-07,
"loss": 0.0079,
"reward": 0.43007929623126984,
"reward_std": 0.16231907904148102,
"reward_std_all": 0.24156006425619125,
"rewards-max/forensics_tempsamp_iou_reward": 0.8733600974082947,
"rewards/forensics_tempsamp_iou_reward": 0.34607982635498047,
"step": 93
},
{
"completion_length": 23.078125,
"epoch": 0.39330543933054396,
"grad_norm": 3.126600742340088,
"kl": 0.20263671875,
"learning_rate": 9.027196652719665e-07,
"loss": 0.0081,
"reward": 0.3452116698026657,
"reward_std": 0.15583550930023193,
"reward_std_all": 0.2789909839630127,
"rewards-max/forensics_tempsamp_iou_reward": 0.9051045775413513,
"rewards/forensics_tempsamp_iou_reward": 0.2600211948156357,
"step": 94
},
{
"completion_length": 21.109375,
"epoch": 0.39748953974895396,
"grad_norm": 1.9876948595046997,
"kl": 0.18994140625,
"learning_rate": 9.01673640167364e-07,
"loss": 0.0076,
"reward": 0.41506506502628326,
"reward_std": 0.12063340470194817,
"reward_std_all": 0.22870634496212006,
"rewards-max/forensics_tempsamp_iou_reward": 0.7716578841209412,
"rewards/forensics_tempsamp_iou_reward": 0.3191799074411392,
"step": 95
},
{
"completion_length": 22.0,
"epoch": 0.401673640167364,
"grad_norm": 2.872668981552124,
"kl": 0.2138671875,
"learning_rate": 9.006276150627615e-07,
"loss": 0.0086,
"reward": 0.3717724084854126,
"reward_std": 0.16219652444124222,
"reward_std_all": 0.2662489265203476,
"rewards-max/forensics_tempsamp_iou_reward": 0.8738818466663361,
"rewards/forensics_tempsamp_iou_reward": 0.27805814146995544,
"step": 96
},
{
"completion_length": 22.3125,
"epoch": 0.40585774058577406,
"grad_norm": 2.8319153785705566,
"kl": 0.19970703125,
"learning_rate": 8.99581589958159e-07,
"loss": 0.008,
"reward": 0.3316686749458313,
"reward_std": 0.13618012517690659,
"reward_std_all": 0.28182078897953033,
"rewards-max/forensics_tempsamp_iou_reward": 0.7416793406009674,
"rewards/forensics_tempsamp_iou_reward": 0.2386946827173233,
"step": 97
},
{
"completion_length": 24.65625,
"epoch": 0.4100418410041841,
"grad_norm": 2.5535874366760254,
"kl": 0.20458984375,
"learning_rate": 8.985355648535564e-07,
"loss": 0.0082,
"reward": 0.4091688096523285,
"reward_std": 0.1997481733560562,
"reward_std_all": 0.26823901385068893,
"rewards-max/forensics_tempsamp_iou_reward": 0.9526621699333191,
"rewards/forensics_tempsamp_iou_reward": 0.32271507382392883,
"step": 98
},
{
"completion_length": 22.671875,
"epoch": 0.41422594142259417,
"grad_norm": 3.4918320178985596,
"kl": 0.23779296875,
"learning_rate": 8.974895397489539e-07,
"loss": 0.0095,
"reward": 0.4746667146682739,
"reward_std": 0.17448637634515762,
"reward_std_all": 0.2271493673324585,
"rewards-max/forensics_tempsamp_iou_reward": 0.8673669993877411,
"rewards/forensics_tempsamp_iou_reward": 0.3896245062351227,
"step": 99
},
{
"completion_length": 23.265625,
"epoch": 0.41841004184100417,
"grad_norm": 2.3665590286254883,
"kl": 0.212890625,
"learning_rate": 8.964435146443514e-07,
"loss": 0.0085,
"reward": 0.35852548480033875,
"reward_std": 0.14135438948869705,
"reward_std_all": 0.26262067258358,
"rewards-max/forensics_tempsamp_iou_reward": 0.866885244846344,
"rewards/forensics_tempsamp_iou_reward": 0.2648075222969055,
"step": 100
},
{
"completion_length": 20.390625,
"epoch": 0.4225941422594142,
"grad_norm": 2.5446417331695557,
"kl": 0.2861328125,
"learning_rate": 8.953974895397489e-07,
"loss": 0.0114,
"reward": 0.4049398750066757,
"reward_std": 0.15363071858882904,
"reward_std_all": 0.24122053384780884,
"rewards-max/forensics_tempsamp_iou_reward": 0.7949315905570984,
"rewards/forensics_tempsamp_iou_reward": 0.3074842840433121,
"step": 101
},
{
"completion_length": 23.15625,
"epoch": 0.42677824267782427,
"grad_norm": 2.8922619819641113,
"kl": 0.27294921875,
"learning_rate": 8.943514644351464e-07,
"loss": 0.0109,
"reward": 0.38747841119766235,
"reward_std": 0.14295434206724167,
"reward_std_all": 0.2511216998100281,
"rewards-max/forensics_tempsamp_iou_reward": 0.7708611190319061,
"rewards/forensics_tempsamp_iou_reward": 0.2926473096013069,
"step": 102
},
{
"completion_length": 22.390625,
"epoch": 0.4309623430962343,
"grad_norm": 3.332198143005371,
"kl": 0.27197265625,
"learning_rate": 8.933054393305439e-07,
"loss": 0.0109,
"reward": 0.4021445959806442,
"reward_std": 0.12268586084246635,
"reward_std_all": 0.23634861409664154,
"rewards-max/forensics_tempsamp_iou_reward": 0.9228777587413788,
"rewards/forensics_tempsamp_iou_reward": 0.32669635117053986,
"step": 103
},
{
"completion_length": 25.4375,
"epoch": 0.4351464435146444,
"grad_norm": 2.0750114917755127,
"kl": 0.2080078125,
"learning_rate": 8.922594142259414e-07,
"loss": 0.0083,
"reward": 0.37635770440101624,
"reward_std": 0.14773592725396156,
"reward_std_all": 0.26692817360162735,
"rewards-max/forensics_tempsamp_iou_reward": 0.8644660413265228,
"rewards/forensics_tempsamp_iou_reward": 0.2949496954679489,
"step": 104
},
{
"completion_length": 21.40625,
"epoch": 0.4393305439330544,
"grad_norm": 2.6949427127838135,
"kl": 0.23046875,
"learning_rate": 8.91213389121339e-07,
"loss": 0.0092,
"reward": 0.4585359990596771,
"reward_std": 0.17895497381687164,
"reward_std_all": 0.23442470282316208,
"rewards-max/forensics_tempsamp_iou_reward": 0.8797975778579712,
"rewards/forensics_tempsamp_iou_reward": 0.3803655803203583,
"step": 105
},
{
"completion_length": 24.703125,
"epoch": 0.4435146443514644,
"grad_norm": 2.5802485942840576,
"kl": 0.2001953125,
"learning_rate": 8.901673640167363e-07,
"loss": 0.008,
"reward": 0.47577258944511414,
"reward_std": 0.1904684379696846,
"reward_std_all": 0.23487290740013123,
"rewards-max/forensics_tempsamp_iou_reward": 0.9099609553813934,
"rewards/forensics_tempsamp_iou_reward": 0.40353283286094666,
"step": 106
},
{
"completion_length": 22.296875,
"epoch": 0.4476987447698745,
"grad_norm": 2.643733501434326,
"kl": 0.21875,
"learning_rate": 8.891213389121338e-07,
"loss": 0.0088,
"reward": 0.4633331596851349,
"reward_std": 0.1954338625073433,
"reward_std_all": 0.24786019325256348,
"rewards-max/forensics_tempsamp_iou_reward": 0.9045187830924988,
"rewards/forensics_tempsamp_iou_reward": 0.3926125019788742,
"step": 107
},
{
"completion_length": 20.90625,
"epoch": 0.45188284518828453,
"grad_norm": 3.067986011505127,
"kl": 0.20947265625,
"learning_rate": 8.880753138075313e-07,
"loss": 0.0084,
"reward": 0.45437172055244446,
"reward_std": 0.17405963689088821,
"reward_std_all": 0.24038219451904297,
"rewards-max/forensics_tempsamp_iou_reward": 0.8637037873268127,
"rewards/forensics_tempsamp_iou_reward": 0.3765452802181244,
"step": 108
},
{
"completion_length": 20.71875,
"epoch": 0.4560669456066946,
"grad_norm": 2.71237850189209,
"kl": 0.18408203125,
"learning_rate": 8.870292887029288e-07,
"loss": 0.0074,
"reward": 0.42786090075969696,
"reward_std": 0.21435575187206268,
"reward_std_all": 0.27776917815208435,
"rewards-max/forensics_tempsamp_iou_reward": 0.8901926875114441,
"rewards/forensics_tempsamp_iou_reward": 0.3493655323982239,
"step": 109
},
{
"completion_length": 19.734375,
"epoch": 0.4602510460251046,
"grad_norm": 3.9000916481018066,
"kl": 0.197265625,
"learning_rate": 8.859832635983263e-07,
"loss": 0.0079,
"reward": 0.4886866509914398,
"reward_std": 0.16028254479169846,
"reward_std_all": 0.21461468189954758,
"rewards-max/forensics_tempsamp_iou_reward": 0.942463606595993,
"rewards/forensics_tempsamp_iou_reward": 0.4303586781024933,
"step": 110
},
{
"completion_length": 22.84375,
"epoch": 0.46443514644351463,
"grad_norm": 3.3757073879241943,
"kl": 0.177734375,
"learning_rate": 8.849372384937238e-07,
"loss": 0.0071,
"reward": 0.38874877989292145,
"reward_std": 0.14397985488176346,
"reward_std_all": 0.24822485446929932,
"rewards-max/forensics_tempsamp_iou_reward": 0.8374001085758209,
"rewards/forensics_tempsamp_iou_reward": 0.30239614844322205,
"step": 111
},
{
"completion_length": 21.265625,
"epoch": 0.4686192468619247,
"grad_norm": 2.9442272186279297,
"kl": 0.1826171875,
"learning_rate": 8.838912133891213e-07,
"loss": 0.0073,
"reward": 0.35087350010871887,
"reward_std": 0.17515070736408234,
"reward_std_all": 0.27819710969924927,
"rewards-max/forensics_tempsamp_iou_reward": 0.6912501454353333,
"rewards/forensics_tempsamp_iou_reward": 0.2518915757536888,
"step": 112
},
{
"completion_length": 22.0625,
"epoch": 0.47280334728033474,
"grad_norm": 2.623720645904541,
"kl": 0.22705078125,
"learning_rate": 8.828451882845189e-07,
"loss": 0.0091,
"reward": 0.4834417998790741,
"reward_std": 0.10748527944087982,
"reward_std_all": 0.19094017148017883,
"rewards-max/forensics_tempsamp_iou_reward": 0.9855321645736694,
"rewards/forensics_tempsamp_iou_reward": 0.40672165155410767,
"step": 113
},
{
"completion_length": 22.640625,
"epoch": 0.4769874476987448,
"grad_norm": 2.7359771728515625,
"kl": 0.1708984375,
"learning_rate": 8.817991631799164e-07,
"loss": 0.0068,
"reward": 0.41464781761169434,
"reward_std": 0.14165400713682175,
"reward_std_all": 0.23397576063871384,
"rewards-max/forensics_tempsamp_iou_reward": 0.8276611566543579,
"rewards/forensics_tempsamp_iou_reward": 0.3258593678474426,
"step": 114
},
{
"completion_length": 21.953125,
"epoch": 0.4811715481171548,
"grad_norm": 2.9807729721069336,
"kl": 0.17041015625,
"learning_rate": 8.807531380753138e-07,
"loss": 0.0068,
"reward": 0.383725568652153,
"reward_std": 0.13195287436246872,
"reward_std_all": 0.24784056842327118,
"rewards-max/forensics_tempsamp_iou_reward": 0.8918665051460266,
"rewards/forensics_tempsamp_iou_reward": 0.3063836097717285,
"step": 115
},
{
"completion_length": 21.953125,
"epoch": 0.48535564853556484,
"grad_norm": 3.17545485496521,
"kl": 0.1884765625,
"learning_rate": 8.797071129707112e-07,
"loss": 0.0075,
"reward": 0.44285212457180023,
"reward_std": 0.15928660333156586,
"reward_std_all": 0.23400167375802994,
"rewards-max/forensics_tempsamp_iou_reward": 0.8989458382129669,
"rewards/forensics_tempsamp_iou_reward": 0.3684719651937485,
"step": 116
},
{
"completion_length": 19.96875,
"epoch": 0.4895397489539749,
"grad_norm": 2.4447708129882812,
"kl": 0.193359375,
"learning_rate": 8.786610878661087e-07,
"loss": 0.0077,
"reward": 0.4413929581642151,
"reward_std": 0.23515470325946808,
"reward_std_all": 0.27337145805358887,
"rewards-max/forensics_tempsamp_iou_reward": 0.8891858756542206,
"rewards/forensics_tempsamp_iou_reward": 0.36477820575237274,
"step": 117
},
{
"completion_length": 22.375,
"epoch": 0.49372384937238495,
"grad_norm": 2.5961410999298096,
"kl": 0.17919921875,
"learning_rate": 8.776150627615062e-07,
"loss": 0.0072,
"reward": 0.40608617663383484,
"reward_std": 0.1389121375977993,
"reward_std_all": 0.2508374750614166,
"rewards-max/forensics_tempsamp_iou_reward": 0.8445764482021332,
"rewards/forensics_tempsamp_iou_reward": 0.3192327469587326,
"step": 118
},
{
"completion_length": 20.46875,
"epoch": 0.497907949790795,
"grad_norm": 2.6434261798858643,
"kl": 0.21240234375,
"learning_rate": 8.765690376569037e-07,
"loss": 0.0085,
"reward": 0.40813907980918884,
"reward_std": 0.16808870434761047,
"reward_std_all": 0.24641027301549911,
"rewards-max/forensics_tempsamp_iou_reward": 0.8661004602909088,
"rewards/forensics_tempsamp_iou_reward": 0.31640492379665375,
"step": 119
},
{
"completion_length": 20.5625,
"epoch": 0.502092050209205,
"grad_norm": 2.6388192176818848,
"kl": 0.1982421875,
"learning_rate": 8.755230125523012e-07,
"loss": 0.0079,
"reward": 0.4793882369995117,
"reward_std": 0.12512316182255745,
"reward_std_all": 0.20260989665985107,
"rewards-max/forensics_tempsamp_iou_reward": 0.9372714757919312,
"rewards/forensics_tempsamp_iou_reward": 0.4141802340745926,
"step": 120
},
{
"completion_length": 21.28125,
"epoch": 0.5062761506276151,
"grad_norm": 3.487044095993042,
"kl": 0.2080078125,
"learning_rate": 8.744769874476987e-07,
"loss": 0.0083,
"reward": 0.48427271842956543,
"reward_std": 0.1616232693195343,
"reward_std_all": 0.21365190297365189,
"rewards-max/forensics_tempsamp_iou_reward": 0.9411173462867737,
"rewards/forensics_tempsamp_iou_reward": 0.41497400403022766,
"step": 121
},
{
"completion_length": 24.234375,
"epoch": 0.5104602510460251,
"grad_norm": 3.0354321002960205,
"kl": 0.1591796875,
"learning_rate": 8.734309623430963e-07,
"loss": 0.0064,
"reward": 0.39963117241859436,
"reward_std": 0.15710975974798203,
"reward_std_all": 0.25147461891174316,
"rewards-max/forensics_tempsamp_iou_reward": 0.8066351413726807,
"rewards/forensics_tempsamp_iou_reward": 0.3015369474887848,
"step": 122
},
{
"completion_length": 21.71875,
"epoch": 0.5146443514644351,
"grad_norm": 3.017091989517212,
"kl": 0.16552734375,
"learning_rate": 8.723849372384937e-07,
"loss": 0.0066,
"reward": 0.4131031781435013,
"reward_std": 0.18442408740520477,
"reward_std_all": 0.2597008943557739,
"rewards-max/forensics_tempsamp_iou_reward": 0.932875782251358,
"rewards/forensics_tempsamp_iou_reward": 0.343762144446373,
"step": 123
},
{
"completion_length": 20.46875,
"epoch": 0.5188284518828452,
"grad_norm": 3.7269277572631836,
"kl": 0.208984375,
"learning_rate": 8.713389121338912e-07,
"loss": 0.0083,
"reward": 0.4837358444929123,
"reward_std": 0.15381664782762527,
"reward_std_all": 0.21456818282604218,
"rewards-max/forensics_tempsamp_iou_reward": 0.9017967581748962,
"rewards/forensics_tempsamp_iou_reward": 0.41204752027988434,
"step": 124
},
{
"completion_length": 20.734375,
"epoch": 0.5230125523012552,
"grad_norm": 5.1426239013671875,
"kl": 0.2158203125,
"learning_rate": 8.702928870292887e-07,
"loss": 0.0086,
"reward": 0.49494320154190063,
"reward_std": 0.12166949734091759,
"reward_std_all": 0.19107121229171753,
"rewards-max/forensics_tempsamp_iou_reward": 0.9302825629711151,
"rewards/forensics_tempsamp_iou_reward": 0.42943650484085083,
"step": 125
},
{
"completion_length": 23.375,
"epoch": 0.5271966527196653,
"grad_norm": 3.0175423622131348,
"kl": 0.18994140625,
"learning_rate": 8.692468619246861e-07,
"loss": 0.0076,
"reward": 0.4020064175128937,
"reward_std": 0.121223334223032,
"reward_std_all": 0.23639969527721405,
"rewards-max/forensics_tempsamp_iou_reward": 0.8566511869430542,
"rewards/forensics_tempsamp_iou_reward": 0.3105523884296417,
"step": 126
},
{
"completion_length": 20.046875,
"epoch": 0.5313807531380753,
"grad_norm": 2.433946132659912,
"kl": 0.18505859375,
"learning_rate": 8.682008368200836e-07,
"loss": 0.0074,
"reward": 0.5017316937446594,
"reward_std": 0.14362675324082375,
"reward_std_all": 0.2073051929473877,
"rewards-max/forensics_tempsamp_iou_reward": 0.9106599688529968,
"rewards/forensics_tempsamp_iou_reward": 0.44401469826698303,
"step": 127
},
{
"completion_length": 21.265625,
"epoch": 0.5355648535564853,
"grad_norm": 2.952678680419922,
"kl": 0.22607421875,
"learning_rate": 8.671548117154811e-07,
"loss": 0.009,
"reward": 0.45628105103969574,
"reward_std": 0.13382942974567413,
"reward_std_all": 0.22084788978099823,
"rewards-max/forensics_tempsamp_iou_reward": 0.9668984413146973,
"rewards/forensics_tempsamp_iou_reward": 0.38444581627845764,
"step": 128
},
{
"completion_length": 19.78125,
"epoch": 0.5397489539748954,
"grad_norm": 2.743352174758911,
"kl": 0.20458984375,
"learning_rate": 8.661087866108786e-07,
"loss": 0.0082,
"reward": 0.42789220809936523,
"reward_std": 0.14278340339660645,
"reward_std_all": 0.2393144816160202,
"rewards-max/forensics_tempsamp_iou_reward": 0.9444932341575623,
"rewards/forensics_tempsamp_iou_reward": 0.36573757231235504,
"step": 129
},
{
"completion_length": 20.25,
"epoch": 0.5439330543933054,
"grad_norm": 2.849771738052368,
"kl": 0.177734375,
"learning_rate": 8.650627615062761e-07,
"loss": 0.0071,
"reward": 0.4272582530975342,
"reward_std": 0.19119562953710556,
"reward_std_all": 0.25316163897514343,
"rewards-max/forensics_tempsamp_iou_reward": 0.7757718861103058,
"rewards/forensics_tempsamp_iou_reward": 0.33690232038497925,
"step": 130
},
{
"completion_length": 24.1875,
"epoch": 0.5481171548117155,
"grad_norm": 2.4035611152648926,
"kl": 0.22705078125,
"learning_rate": 8.640167364016737e-07,
"loss": 0.0091,
"reward": 0.3635486662387848,
"reward_std": 0.11964528262615204,
"reward_std_all": 0.25798438489437103,
"rewards-max/forensics_tempsamp_iou_reward": 0.8915493488311768,
"rewards/forensics_tempsamp_iou_reward": 0.27745555341243744,
"step": 131
},
{
"completion_length": 22.96875,
"epoch": 0.5523012552301255,
"grad_norm": 2.8074169158935547,
"kl": 0.1943359375,
"learning_rate": 8.629707112970711e-07,
"loss": 0.0078,
"reward": 0.43903808295726776,
"reward_std": 0.14686980843544006,
"reward_std_all": 0.2310941368341446,
"rewards-max/forensics_tempsamp_iou_reward": 0.9122007191181183,
"rewards/forensics_tempsamp_iou_reward": 0.35912472009658813,
"step": 132
},
{
"completion_length": 23.40625,
"epoch": 0.5564853556485355,
"grad_norm": 2.4067201614379883,
"kl": 0.20751953125,
"learning_rate": 8.619246861924686e-07,
"loss": 0.0083,
"reward": 0.40644124150276184,
"reward_std": 0.13685572519898415,
"reward_std_all": 0.2431831732392311,
"rewards-max/forensics_tempsamp_iou_reward": 0.9134937822818756,
"rewards/forensics_tempsamp_iou_reward": 0.31430450081825256,
"step": 133
},
{
"completion_length": 21.421875,
"epoch": 0.5606694560669456,
"grad_norm": 2.9175379276275635,
"kl": 0.1953125,
"learning_rate": 8.608786610878661e-07,
"loss": 0.0078,
"reward": 0.44299888610839844,
"reward_std": 0.1378997266292572,
"reward_std_all": 0.2222016528248787,
"rewards-max/forensics_tempsamp_iou_reward": 0.9360418021678925,
"rewards/forensics_tempsamp_iou_reward": 0.3682139813899994,
"step": 134
},
{
"completion_length": 22.359375,
"epoch": 0.5648535564853556,
"grad_norm": 2.451629161834717,
"kl": 0.22216796875,
"learning_rate": 8.598326359832636e-07,
"loss": 0.0089,
"reward": 0.4005991518497467,
"reward_std": 0.1276058740913868,
"reward_std_all": 0.23929477483034134,
"rewards-max/forensics_tempsamp_iou_reward": 0.9119961261749268,
"rewards/forensics_tempsamp_iou_reward": 0.32354769110679626,
"step": 135
},
{
"completion_length": 20.421875,
"epoch": 0.5690376569037657,
"grad_norm": 2.153024435043335,
"kl": 0.1953125,
"learning_rate": 8.58786610878661e-07,
"loss": 0.0078,
"reward": 0.4740244299173355,
"reward_std": 0.1444147303700447,
"reward_std_all": 0.21636095643043518,
"rewards-max/forensics_tempsamp_iou_reward": 0.9031059443950653,
"rewards/forensics_tempsamp_iou_reward": 0.4018695056438446,
"step": 136
},
{
"completion_length": 21.3125,
"epoch": 0.5732217573221757,
"grad_norm": 3.050682544708252,
"kl": 0.2431640625,
"learning_rate": 8.577405857740585e-07,
"loss": 0.0097,
"reward": 0.4499470889568329,
"reward_std": 0.18046756833791733,
"reward_std_all": 0.23873861134052277,
"rewards-max/forensics_tempsamp_iou_reward": 0.9610866010189056,
"rewards/forensics_tempsamp_iou_reward": 0.37879204750061035,
"step": 137
},
{
"completion_length": 23.53125,
"epoch": 0.5774058577405857,
"grad_norm": 3.1136395931243896,
"kl": 0.23291015625,
"learning_rate": 8.56694560669456e-07,
"loss": 0.0093,
"reward": 0.5147728025913239,
"reward_std": 0.18133465945720673,
"reward_std_all": 0.21935952454805374,
"rewards-max/forensics_tempsamp_iou_reward": 0.929665595293045,
"rewards/forensics_tempsamp_iou_reward": 0.44274812936782837,
"step": 138
},
{
"completion_length": 22.125,
"epoch": 0.5815899581589958,
"grad_norm": 2.5090715885162354,
"kl": 0.234375,
"learning_rate": 8.556485355648536e-07,
"loss": 0.0094,
"reward": 0.5767271816730499,
"reward_std": 0.1422729156911373,
"reward_std_all": 0.16612739861011505,
"rewards-max/forensics_tempsamp_iou_reward": 0.8583837449550629,
"rewards/forensics_tempsamp_iou_reward": 0.5035188794136047,
"step": 139
},
{
"completion_length": 21.265625,
"epoch": 0.5857740585774058,
"grad_norm": 2.816964626312256,
"kl": 0.21630859375,
"learning_rate": 8.54602510460251e-07,
"loss": 0.0086,
"reward": 0.4988979548215866,
"reward_std": 0.14711756259202957,
"reward_std_all": 0.20451705902814865,
"rewards-max/forensics_tempsamp_iou_reward": 0.9052472710609436,
"rewards/forensics_tempsamp_iou_reward": 0.42210060358047485,
"step": 140
},
{
"completion_length": 23.8125,
"epoch": 0.5899581589958159,
"grad_norm": 3.428279161453247,
"kl": 0.2880859375,
"learning_rate": 8.535564853556485e-07,
"loss": 0.0115,
"reward": 0.5659472644329071,
"reward_std": 0.1879376620054245,
"reward_std_all": 0.20124821364879608,
"rewards-max/forensics_tempsamp_iou_reward": 0.9567589163780212,
"rewards/forensics_tempsamp_iou_reward": 0.5213198363780975,
"step": 141
},
{
"completion_length": 22.296875,
"epoch": 0.5941422594142259,
"grad_norm": 3.072437047958374,
"kl": 0.22998046875,
"learning_rate": 8.52510460251046e-07,
"loss": 0.0092,
"reward": 0.4164227694272995,
"reward_std": 0.12511048465967178,
"reward_std_all": 0.22945372760295868,
"rewards-max/forensics_tempsamp_iou_reward": 0.8135422170162201,
"rewards/forensics_tempsamp_iou_reward": 0.3279648721218109,
"step": 142
},
{
"completion_length": 24.78125,
"epoch": 0.5983263598326359,
"grad_norm": 3.051137685775757,
"kl": 0.22216796875,
"learning_rate": 8.514644351464435e-07,
"loss": 0.0089,
"reward": 0.44584985077381134,
"reward_std": 0.134488545358181,
"reward_std_all": 0.2198861613869667,
"rewards-max/forensics_tempsamp_iou_reward": 0.9396452903747559,
"rewards/forensics_tempsamp_iou_reward": 0.368113249540329,
"step": 143
},
{
"completion_length": 23.984375,
"epoch": 0.602510460251046,
"grad_norm": 3.862793207168579,
"kl": 0.185546875,
"learning_rate": 8.50418410041841e-07,
"loss": 0.0074,
"reward": 0.45022670924663544,
"reward_std": 0.14671873301267624,
"reward_std_all": 0.22519346326589584,
"rewards-max/forensics_tempsamp_iou_reward": 0.8508294522762299,
"rewards/forensics_tempsamp_iou_reward": 0.3635125309228897,
"step": 144
},
{
"completion_length": 25.453125,
"epoch": 0.606694560669456,
"grad_norm": 3.2549514770507812,
"kl": 0.2421875,
"learning_rate": 8.493723849372385e-07,
"loss": 0.0097,
"reward": 0.3441780060529709,
"reward_std": 0.14902973175048828,
"reward_std_all": 0.27569784224033356,
"rewards-max/forensics_tempsamp_iou_reward": 0.7128922641277313,
"rewards/forensics_tempsamp_iou_reward": 0.24711671471595764,
"step": 145
},
{
"completion_length": 22.140625,
"epoch": 0.6108786610878661,
"grad_norm": 3.3701300621032715,
"kl": 0.1845703125,
"learning_rate": 8.483263598326359e-07,
"loss": 0.0074,
"reward": 0.4855945408344269,
"reward_std": 0.1593681052327156,
"reward_std_all": 0.21955610811710358,
"rewards-max/forensics_tempsamp_iou_reward": 0.8708549737930298,
"rewards/forensics_tempsamp_iou_reward": 0.40526606142520905,
"step": 146
},
{
"completion_length": 20.25,
"epoch": 0.6150627615062761,
"grad_norm": 2.755064010620117,
"kl": 0.224609375,
"learning_rate": 8.472803347280334e-07,
"loss": 0.009,
"reward": 0.5241940319538116,
"reward_std": 0.1578264757990837,
"reward_std_all": 0.2001493126153946,
"rewards-max/forensics_tempsamp_iou_reward": 0.9042113125324249,
"rewards/forensics_tempsamp_iou_reward": 0.45731498301029205,
"step": 147
},
{
"completion_length": 21.046875,
"epoch": 0.6192468619246861,
"grad_norm": 2.699641466140747,
"kl": 0.19482421875,
"learning_rate": 8.462343096234308e-07,
"loss": 0.0078,
"reward": 0.42927534878253937,
"reward_std": 0.07328962162137032,
"reward_std_all": 0.20531512051820755,
"rewards-max/forensics_tempsamp_iou_reward": 0.9384458363056183,
"rewards/forensics_tempsamp_iou_reward": 0.3543698340654373,
"step": 148
},
{
"completion_length": 22.984375,
"epoch": 0.6234309623430963,
"grad_norm": 3.4765737056732178,
"kl": 0.27294921875,
"learning_rate": 8.451882845188284e-07,
"loss": 0.0109,
"reward": 0.4355625808238983,
"reward_std": 0.12133080512285233,
"reward_std_all": 0.21979010105133057,
"rewards-max/forensics_tempsamp_iou_reward": 0.8812152147293091,
"rewards/forensics_tempsamp_iou_reward": 0.350423663854599,
"step": 149
},
{
"completion_length": 21.140625,
"epoch": 0.6276150627615062,
"grad_norm": 2.752394437789917,
"kl": 0.21728515625,
"learning_rate": 8.441422594142259e-07,
"loss": 0.0087,
"reward": 0.41755253076553345,
"reward_std": 0.2179994136095047,
"reward_std_all": 0.2757028639316559,
"rewards-max/forensics_tempsamp_iou_reward": 0.9277379214763641,
"rewards/forensics_tempsamp_iou_reward": 0.3413289934396744,
"step": 150
},
{
"completion_length": 20.96875,
"epoch": 0.6317991631799164,
"grad_norm": 2.6615583896636963,
"kl": 0.2021484375,
"learning_rate": 8.430962343096234e-07,
"loss": 0.0081,
"reward": 0.3939850777387619,
"reward_std": 0.13101816177368164,
"reward_std_all": 0.2440369725227356,
"rewards-max/forensics_tempsamp_iou_reward": 0.8808875977993011,
"rewards/forensics_tempsamp_iou_reward": 0.31139564514160156,
"step": 151
},
{
"completion_length": 19.875,
"epoch": 0.6359832635983264,
"grad_norm": 2.1855051517486572,
"kl": 0.228515625,
"learning_rate": 8.420502092050209e-07,
"loss": 0.0091,
"reward": 0.48360365629196167,
"reward_std": 0.09144109860062599,
"reward_std_all": 0.19477061927318573,
"rewards-max/forensics_tempsamp_iou_reward": 0.8971797823905945,
"rewards/forensics_tempsamp_iou_reward": 0.4289720356464386,
"step": 152
},
{
"completion_length": 19.921875,
"epoch": 0.6401673640167364,
"grad_norm": 2.5410447120666504,
"kl": 0.23681640625,
"learning_rate": 8.410041841004184e-07,
"loss": 0.0095,
"reward": 0.532864898443222,
"reward_std": 0.17426315695047379,
"reward_std_all": 0.21076004952192307,
"rewards-max/forensics_tempsamp_iou_reward": 0.9394762217998505,
"rewards/forensics_tempsamp_iou_reward": 0.48673059046268463,
"step": 153
},
{
"completion_length": 21.0625,
"epoch": 0.6443514644351465,
"grad_norm": 3.2715723514556885,
"kl": 0.2236328125,
"learning_rate": 8.399581589958159e-07,
"loss": 0.0089,
"reward": 0.45571091771125793,
"reward_std": 0.1523030549287796,
"reward_std_all": 0.22705690562725067,
"rewards-max/forensics_tempsamp_iou_reward": 0.8981777727603912,
"rewards/forensics_tempsamp_iou_reward": 0.39293012022972107,
"step": 154
},
{
"completion_length": 20.015625,
"epoch": 0.6485355648535565,
"grad_norm": 2.7428181171417236,
"kl": 0.23779296875,
"learning_rate": 8.389121338912134e-07,
"loss": 0.0095,
"reward": 0.45074114203453064,
"reward_std": 0.1848168522119522,
"reward_std_all": 0.25254911184310913,
"rewards-max/forensics_tempsamp_iou_reward": 0.9553932845592499,
"rewards/forensics_tempsamp_iou_reward": 0.40142688155174255,
"step": 155
},
{
"completion_length": 20.65625,
"epoch": 0.6527196652719666,
"grad_norm": 2.396305799484253,
"kl": 0.2001953125,
"learning_rate": 8.378661087866107e-07,
"loss": 0.008,
"reward": 0.4004524052143097,
"reward_std": 0.13376879319548607,
"reward_std_all": 0.2451367825269699,
"rewards-max/forensics_tempsamp_iou_reward": 0.9098137617111206,
"rewards/forensics_tempsamp_iou_reward": 0.3227430135011673,
"step": 156
},
{
"completion_length": 21.578125,
"epoch": 0.6569037656903766,
"grad_norm": 2.7211360931396484,
"kl": 0.16845703125,
"learning_rate": 8.368200836820083e-07,
"loss": 0.0067,
"reward": 0.3646601438522339,
"reward_std": 0.11267167702317238,
"reward_std_all": 0.2500659078359604,
"rewards-max/forensics_tempsamp_iou_reward": 0.8038130402565002,
"rewards/forensics_tempsamp_iou_reward": 0.28284918516874313,
"step": 157
},
{
"completion_length": 20.265625,
"epoch": 0.6610878661087866,
"grad_norm": 3.1417579650878906,
"kl": 0.21484375,
"learning_rate": 8.357740585774058e-07,
"loss": 0.0086,
"reward": 0.47523047029972076,
"reward_std": 0.17171894013881683,
"reward_std_all": 0.231247179210186,
"rewards-max/forensics_tempsamp_iou_reward": 0.8490384519100189,
"rewards/forensics_tempsamp_iou_reward": 0.3956393450498581,
"step": 158
},
{
"completion_length": 21.46875,
"epoch": 0.6652719665271967,
"grad_norm": 3.0648622512817383,
"kl": 0.3203125,
"learning_rate": 8.347280334728033e-07,
"loss": 0.0128,
"reward": 0.4476076066493988,
"reward_std": 0.1520785465836525,
"reward_std_all": 0.23183970898389816,
"rewards-max/forensics_tempsamp_iou_reward": 0.879236102104187,
"rewards/forensics_tempsamp_iou_reward": 0.3722226917743683,
"step": 159
},
{
"completion_length": 19.90625,
"epoch": 0.6694560669456067,
"grad_norm": 2.613673686981201,
"kl": 0.23828125,
"learning_rate": 8.336820083682008e-07,
"loss": 0.0095,
"reward": 0.3872516304254532,
"reward_std": 0.18575584888458252,
"reward_std_all": 0.27996020019054413,
"rewards-max/forensics_tempsamp_iou_reward": 0.9223994612693787,
"rewards/forensics_tempsamp_iou_reward": 0.31443534791469574,
"step": 160
},
{
"completion_length": 19.796875,
"epoch": 0.6736401673640168,
"grad_norm": 3.170278310775757,
"kl": 0.20556640625,
"learning_rate": 8.326359832635983e-07,
"loss": 0.0082,
"reward": 0.47693106532096863,
"reward_std": 0.09521405771374702,
"reward_std_all": 0.19265932589769363,
"rewards-max/forensics_tempsamp_iou_reward": 0.9170872569084167,
"rewards/forensics_tempsamp_iou_reward": 0.4039805829524994,
"step": 161
},
{
"completion_length": 21.1875,
"epoch": 0.6778242677824268,
"grad_norm": 3.641517400741577,
"kl": 0.23876953125,
"learning_rate": 8.315899581589958e-07,
"loss": 0.0095,
"reward": 0.5118494629859924,
"reward_std": 0.18338937312364578,
"reward_std_all": 0.22490958124399185,
"rewards-max/forensics_tempsamp_iou_reward": 0.9177646934986115,
"rewards/forensics_tempsamp_iou_reward": 0.4537495821714401,
"step": 162
},
{
"completion_length": 21.0625,
"epoch": 0.6820083682008368,
"grad_norm": 3.0905048847198486,
"kl": 0.22998046875,
"learning_rate": 8.305439330543933e-07,
"loss": 0.0092,
"reward": 0.4999140650033951,
"reward_std": 0.14736276492476463,
"reward_std_all": 0.21067219227552414,
"rewards-max/forensics_tempsamp_iou_reward": 0.9354741871356964,
"rewards/forensics_tempsamp_iou_reward": 0.43387193977832794,
"step": 163
},
{
"completion_length": 20.34375,
"epoch": 0.6861924686192469,
"grad_norm": 2.8736541271209717,
"kl": 0.21630859375,
"learning_rate": 8.294979079497908e-07,
"loss": 0.0087,
"reward": 0.47392456233501434,
"reward_std": 0.1292835958302021,
"reward_std_all": 0.20665565133094788,
"rewards-max/forensics_tempsamp_iou_reward": 0.8792850971221924,
"rewards/forensics_tempsamp_iou_reward": 0.39171041548252106,
"step": 164
},
{
"completion_length": 22.46875,
"epoch": 0.6903765690376569,
"grad_norm": 2.429201364517212,
"kl": 0.19873046875,
"learning_rate": 8.284518828451882e-07,
"loss": 0.0079,
"reward": 0.47362270951271057,
"reward_std": 0.12597563862800598,
"reward_std_all": 0.2081482782959938,
"rewards-max/forensics_tempsamp_iou_reward": 0.9658733308315277,
"rewards/forensics_tempsamp_iou_reward": 0.4163615107536316,
"step": 165
},
{
"completion_length": 21.84375,
"epoch": 0.694560669456067,
"grad_norm": 2.4877402782440186,
"kl": 0.228515625,
"learning_rate": 8.274058577405857e-07,
"loss": 0.0091,
"reward": 0.4998297691345215,
"reward_std": 0.1216079369187355,
"reward_std_all": 0.19593335688114166,
"rewards-max/forensics_tempsamp_iou_reward": 0.9306803643703461,
"rewards/forensics_tempsamp_iou_reward": 0.4344746470451355,
"step": 166
},
{
"completion_length": 21.03125,
"epoch": 0.698744769874477,
"grad_norm": 3.3212668895721436,
"kl": 0.2060546875,
"learning_rate": 8.263598326359832e-07,
"loss": 0.0082,
"reward": 0.4406004846096039,
"reward_std": 0.16852135956287384,
"reward_std_all": 0.24621307104825974,
"rewards-max/forensics_tempsamp_iou_reward": 0.9169618785381317,
"rewards/forensics_tempsamp_iou_reward": 0.36687174439430237,
"step": 167
},
{
"completion_length": 20.75,
"epoch": 0.702928870292887,
"grad_norm": 2.7538318634033203,
"kl": 0.23193359375,
"learning_rate": 8.253138075313807e-07,
"loss": 0.0093,
"reward": 0.49902525544166565,
"reward_std": 0.1871483251452446,
"reward_std_all": 0.22284496575593948,
"rewards-max/forensics_tempsamp_iou_reward": 0.8337318599224091,
"rewards/forensics_tempsamp_iou_reward": 0.4203528165817261,
"step": 168
},
{
"completion_length": 21.21875,
"epoch": 0.7071129707112971,
"grad_norm": 2.9568936824798584,
"kl": 0.205078125,
"learning_rate": 8.242677824267782e-07,
"loss": 0.0082,
"reward": 0.4685553014278412,
"reward_std": 0.12919190898537636,
"reward_std_all": 0.21346978843212128,
"rewards-max/forensics_tempsamp_iou_reward": 0.9470005333423615,
"rewards/forensics_tempsamp_iou_reward": 0.40412504971027374,
"step": 169
},
{
"completion_length": 22.890625,
"epoch": 0.7112970711297071,
"grad_norm": 3.2359461784362793,
"kl": 0.23779296875,
"learning_rate": 8.232217573221757e-07,
"loss": 0.0095,
"reward": 0.44061385095119476,
"reward_std": 0.16264215111732483,
"reward_std_all": 0.23639892786741257,
"rewards-max/forensics_tempsamp_iou_reward": 0.9212930798530579,
"rewards/forensics_tempsamp_iou_reward": 0.3668813109397888,
"step": 170
},
{
"completion_length": 22.734375,
"epoch": 0.7154811715481172,
"grad_norm": 2.961820602416992,
"kl": 0.24755859375,
"learning_rate": 8.221757322175732e-07,
"loss": 0.0099,
"reward": 0.5017964839935303,
"reward_std": 0.1242625042796135,
"reward_std_all": 0.18973636627197266,
"rewards-max/forensics_tempsamp_iou_reward": 0.8366635739803314,
"rewards/forensics_tempsamp_iou_reward": 0.4209460914134979,
"step": 171
},
{
"completion_length": 20.78125,
"epoch": 0.7196652719665272,
"grad_norm": 3.2609643936157227,
"kl": 0.24658203125,
"learning_rate": 8.211297071129707e-07,
"loss": 0.0099,
"reward": 0.49299533665180206,
"reward_std": 0.08453327417373657,
"reward_std_all": 0.1793249100446701,
"rewards-max/forensics_tempsamp_iou_reward": 0.8321774303913116,
"rewards/forensics_tempsamp_iou_reward": 0.42463693022727966,
"step": 172
},
{
"completion_length": 20.609375,
"epoch": 0.7238493723849372,
"grad_norm": 2.9796805381774902,
"kl": 0.20849609375,
"learning_rate": 8.200836820083682e-07,
"loss": 0.0083,
"reward": 0.4336213022470474,
"reward_std": 0.14152024686336517,
"reward_std_all": 0.2349139228463173,
"rewards-max/forensics_tempsamp_iou_reward": 0.9082809388637543,
"rewards/forensics_tempsamp_iou_reward": 0.36374300718307495,
"step": 173
},
{
"completion_length": 23.3125,
"epoch": 0.7280334728033473,
"grad_norm": 3.01364803314209,
"kl": 0.24658203125,
"learning_rate": 8.190376569037657e-07,
"loss": 0.0099,
"reward": 0.4674176573753357,
"reward_std": 0.15620875358581543,
"reward_std_all": 0.2268999144434929,
"rewards-max/forensics_tempsamp_iou_reward": 0.9155757427215576,
"rewards/forensics_tempsamp_iou_reward": 0.39580968022346497,
"step": 174
},
{
"completion_length": 22.421875,
"epoch": 0.7322175732217573,
"grad_norm": 2.8095555305480957,
"kl": 0.2607421875,
"learning_rate": 8.179916317991632e-07,
"loss": 0.0104,
"reward": 0.4752761572599411,
"reward_std": 0.14566955715417862,
"reward_std_all": 0.21782489866018295,
"rewards-max/forensics_tempsamp_iou_reward": 0.8762764036655426,
"rewards/forensics_tempsamp_iou_reward": 0.39315688610076904,
"step": 175
},
{
"completion_length": 22.9375,
"epoch": 0.7364016736401674,
"grad_norm": 3.1036009788513184,
"kl": 0.25439453125,
"learning_rate": 8.169456066945606e-07,
"loss": 0.0102,
"reward": 0.3794126808643341,
"reward_std": 0.1754727065563202,
"reward_std_all": 0.27418793737888336,
"rewards-max/forensics_tempsamp_iou_reward": 0.8977140784263611,
"rewards/forensics_tempsamp_iou_reward": 0.29535171389579773,
"step": 176
},
{
"completion_length": 23.765625,
"epoch": 0.7405857740585774,
"grad_norm": 3.3552582263946533,
"kl": 0.23095703125,
"learning_rate": 8.158995815899581e-07,
"loss": 0.0092,
"reward": 0.4229155480861664,
"reward_std": 0.13358639553189278,
"reward_std_all": 0.23477252572774887,
"rewards-max/forensics_tempsamp_iou_reward": 0.8086326122283936,
"rewards/forensics_tempsamp_iou_reward": 0.33889272809028625,
"step": 177
},
{
"completion_length": 21.828125,
"epoch": 0.7447698744769874,
"grad_norm": 3.182995319366455,
"kl": 0.23486328125,
"learning_rate": 8.148535564853556e-07,
"loss": 0.0094,
"reward": 0.5248161852359772,
"reward_std": 0.17395954579114914,
"reward_std_all": 0.203041173517704,
"rewards-max/forensics_tempsamp_iou_reward": 0.9158969521522522,
"rewards/forensics_tempsamp_iou_reward": 0.46165670454502106,
"step": 178
},
{
"completion_length": 24.390625,
"epoch": 0.7489539748953975,
"grad_norm": 2.787008762359619,
"kl": 0.22607421875,
"learning_rate": 8.138075313807531e-07,
"loss": 0.0091,
"reward": 0.5045577585697174,
"reward_std": 0.14821110665798187,
"reward_std_all": 0.20557596534490585,
"rewards-max/forensics_tempsamp_iou_reward": 0.8753862380981445,
"rewards/forensics_tempsamp_iou_reward": 0.4298754185438156,
"step": 179
},
{
"completion_length": 25.546875,
"epoch": 0.7531380753138075,
"grad_norm": 3.3069863319396973,
"kl": 0.220703125,
"learning_rate": 8.127615062761506e-07,
"loss": 0.0088,
"reward": 0.45841725170612335,
"reward_std": 0.13112735003232956,
"reward_std_all": 0.213491752743721,
"rewards-max/forensics_tempsamp_iou_reward": 0.8984034061431885,
"rewards/forensics_tempsamp_iou_reward": 0.3781336098909378,
"step": 180
},
{
"completion_length": 24.171875,
"epoch": 0.7573221757322176,
"grad_norm": 2.972693681716919,
"kl": 0.2392578125,
"learning_rate": 8.117154811715481e-07,
"loss": 0.0096,
"reward": 0.4576316624879837,
"reward_std": 0.15145494043827057,
"reward_std_all": 0.2268020063638687,
"rewards-max/forensics_tempsamp_iou_reward": 0.8955655992031097,
"rewards/forensics_tempsamp_iou_reward": 0.38260869681835175,
"step": 181
},
{
"completion_length": 21.40625,
"epoch": 0.7615062761506276,
"grad_norm": 3.5101563930511475,
"kl": 0.19970703125,
"learning_rate": 8.106694560669455e-07,
"loss": 0.008,
"reward": 0.4825403094291687,
"reward_std": 0.14002035558223724,
"reward_std_all": 0.21266379207372665,
"rewards-max/forensics_tempsamp_iou_reward": 0.9040354192256927,
"rewards/forensics_tempsamp_iou_reward": 0.4063745439052582,
"step": 182
},
{
"completion_length": 21.375,
"epoch": 0.7656903765690377,
"grad_norm": 3.110072612762451,
"kl": 0.2373046875,
"learning_rate": 8.096234309623431e-07,
"loss": 0.0095,
"reward": 0.4870626777410507,
"reward_std": 0.23556938767433167,
"reward_std_all": 0.2564467042684555,
"rewards-max/forensics_tempsamp_iou_reward": 0.9442270696163177,
"rewards/forensics_tempsamp_iou_reward": 0.42147161066532135,
"step": 183
},
{
"completion_length": 22.0,
"epoch": 0.7698744769874477,
"grad_norm": 3.106229543685913,
"kl": 0.21484375,
"learning_rate": 8.085774058577406e-07,
"loss": 0.0086,
"reward": 0.5039902329444885,
"reward_std": 0.12779080495238304,
"reward_std_all": 0.2006426751613617,
"rewards-max/forensics_tempsamp_iou_reward": 0.9264122843742371,
"rewards/forensics_tempsamp_iou_reward": 0.45152513682842255,
"step": 184
},
{
"completion_length": 22.96875,
"epoch": 0.7740585774058577,
"grad_norm": 3.5228633880615234,
"kl": 0.23486328125,
"learning_rate": 8.075313807531381e-07,
"loss": 0.0094,
"reward": 0.43880653381347656,
"reward_std": 0.21230527758598328,
"reward_std_all": 0.25976429879665375,
"rewards-max/forensics_tempsamp_iou_reward": 0.8808355033397675,
"rewards/forensics_tempsamp_iou_reward": 0.3544783443212509,
"step": 185
},
{
"completion_length": 21.25,
"epoch": 0.7782426778242678,
"grad_norm": 2.6524996757507324,
"kl": 0.23486328125,
"learning_rate": 8.064853556485355e-07,
"loss": 0.0094,
"reward": 0.5435624122619629,
"reward_std": 0.126361683011055,
"reward_std_all": 0.17262475192546844,
"rewards-max/forensics_tempsamp_iou_reward": 0.8786827325820923,
"rewards/forensics_tempsamp_iou_reward": 0.4717957079410553,
"step": 186
},
{
"completion_length": 22.984375,
"epoch": 0.7824267782426778,
"grad_norm": 4.7811431884765625,
"kl": 0.2900390625,
"learning_rate": 8.05439330543933e-07,
"loss": 0.0116,
"reward": 0.46644362807273865,
"reward_std": 0.12555569410324097,
"reward_std_all": 0.21093381196260452,
"rewards-max/forensics_tempsamp_iou_reward": 0.8558894097805023,
"rewards/forensics_tempsamp_iou_reward": 0.3875201642513275,
"step": 187
},
{
"completion_length": 23.140625,
"epoch": 0.7866108786610879,
"grad_norm": 2.4787087440490723,
"kl": 0.2353515625,
"learning_rate": 8.043933054393305e-07,
"loss": 0.0094,
"reward": 0.5302537977695465,
"reward_std": 0.1639224886894226,
"reward_std_all": 0.19575748592615128,
"rewards-max/forensics_tempsamp_iou_reward": 0.9368042945861816,
"rewards/forensics_tempsamp_iou_reward": 0.4657113701105118,
"step": 188
},
{
"completion_length": 22.421875,
"epoch": 0.7907949790794979,
"grad_norm": 2.504493236541748,
"kl": 0.24951171875,
"learning_rate": 8.03347280334728e-07,
"loss": 0.01,
"reward": 0.4700623005628586,
"reward_std": 0.10663755983114243,
"reward_std_all": 0.19997362047433853,
"rewards-max/forensics_tempsamp_iou_reward": 0.911878228187561,
"rewards/forensics_tempsamp_iou_reward": 0.39281585812568665,
"step": 189
},
{
"completion_length": 21.203125,
"epoch": 0.7949790794979079,
"grad_norm": 3.0539772510528564,
"kl": 0.25048828125,
"learning_rate": 8.023012552301254e-07,
"loss": 0.01,
"reward": 0.5498219281435013,
"reward_std": 0.11988098174333572,
"reward_std_all": 0.17097638547420502,
"rewards-max/forensics_tempsamp_iou_reward": 0.9491911828517914,
"rewards/forensics_tempsamp_iou_reward": 0.5010037273168564,
"step": 190
},
{
"completion_length": 21.265625,
"epoch": 0.799163179916318,
"grad_norm": 3.0065865516662598,
"kl": 0.27734375,
"learning_rate": 8.01255230125523e-07,
"loss": 0.0111,
"reward": 0.5193320512771606,
"reward_std": 0.09360900148749352,
"reward_std_all": 0.17452287673950195,
"rewards-max/forensics_tempsamp_iou_reward": 0.916426807641983,
"rewards/forensics_tempsamp_iou_reward": 0.46101684868335724,
"step": 191
},
{
"completion_length": 20.40625,
"epoch": 0.803347280334728,
"grad_norm": 2.4014053344726562,
"kl": 0.23388671875,
"learning_rate": 8.002092050209205e-07,
"loss": 0.0093,
"reward": 0.6023570895195007,
"reward_std": 0.09932931140065193,
"reward_std_all": 0.13800369203090668,
"rewards-max/forensics_tempsamp_iou_reward": 0.9454897344112396,
"rewards/forensics_tempsamp_iou_reward": 0.5582804381847382,
"step": 192
},
{
"completion_length": 20.296875,
"epoch": 0.8075313807531381,
"grad_norm": 3.3778600692749023,
"kl": 0.19921875,
"learning_rate": 7.99163179916318e-07,
"loss": 0.008,
"reward": 0.48741570115089417,
"reward_std": 0.1268736720085144,
"reward_std_all": 0.20043056458234787,
"rewards-max/forensics_tempsamp_iou_reward": 0.9081154465675354,
"rewards/forensics_tempsamp_iou_reward": 0.41168293356895447,
"step": 193
},
{
"completion_length": 20.5,
"epoch": 0.8117154811715481,
"grad_norm": 3.2653019428253174,
"kl": 0.28564453125,
"learning_rate": 7.981171548117155e-07,
"loss": 0.0114,
"reward": 0.5302922129631042,
"reward_std": 0.15213659405708313,
"reward_std_all": 0.19866738468408585,
"rewards-max/forensics_tempsamp_iou_reward": 0.9435873925685883,
"rewards/forensics_tempsamp_iou_reward": 0.4792274385690689,
"step": 194
},
{
"completion_length": 20.828125,
"epoch": 0.8158995815899581,
"grad_norm": 2.576396942138672,
"kl": 0.259765625,
"learning_rate": 7.97071129707113e-07,
"loss": 0.0104,
"reward": 0.42537812888622284,
"reward_std": 0.14282989501953125,
"reward_std_all": 0.23934666067361832,
"rewards-max/forensics_tempsamp_iou_reward": 0.8956102132797241,
"rewards/forensics_tempsamp_iou_reward": 0.3497551828622818,
"step": 195
},
{
"completion_length": 21.640625,
"epoch": 0.8200836820083682,
"grad_norm": 3.430049180984497,
"kl": 0.36865234375,
"learning_rate": 7.960251046025104e-07,
"loss": 0.0147,
"reward": 0.5035016983747482,
"reward_std": 0.10397813841700554,
"reward_std_all": 0.18489725142717361,
"rewards-max/forensics_tempsamp_iou_reward": 0.9101535081863403,
"rewards/forensics_tempsamp_iou_reward": 0.43821047246456146,
"step": 196
},
{
"completion_length": 20.9375,
"epoch": 0.8242677824267782,
"grad_norm": 2.962965250015259,
"kl": 0.23974609375,
"learning_rate": 7.949790794979079e-07,
"loss": 0.0096,
"reward": 0.47257475554943085,
"reward_std": 0.15035942941904068,
"reward_std_all": 0.2228764072060585,
"rewards-max/forensics_tempsamp_iou_reward": 0.8841829597949982,
"rewards/forensics_tempsamp_iou_reward": 0.41789956390857697,
"step": 197
},
{
"completion_length": 24.4375,
"epoch": 0.8284518828451883,
"grad_norm": 3.26701283454895,
"kl": 0.32275390625,
"learning_rate": 7.939330543933053e-07,
"loss": 0.0129,
"reward": 0.4376286119222641,
"reward_std": 0.1243523582816124,
"reward_std_all": 0.21740268170833588,
"rewards-max/forensics_tempsamp_iou_reward": 0.9838087260723114,
"rewards/forensics_tempsamp_iou_reward": 0.3651266247034073,
"step": 198
},
{
"completion_length": 21.25,
"epoch": 0.8326359832635983,
"grad_norm": 2.820166826248169,
"kl": 0.20654296875,
"learning_rate": 7.928870292887028e-07,
"loss": 0.0083,
"reward": 0.4441964328289032,
"reward_std": 0.15912766009569168,
"reward_std_all": 0.2330639809370041,
"rewards-max/forensics_tempsamp_iou_reward": 0.9417475759983063,
"rewards/forensics_tempsamp_iou_reward": 0.3780969977378845,
"step": 199
},
{
"completion_length": 21.765625,
"epoch": 0.8368200836820083,
"grad_norm": 2.7870678901672363,
"kl": 0.2470703125,
"learning_rate": 7.918410041841004e-07,
"loss": 0.0099,
"reward": 0.41933462023735046,
"reward_std": 0.12255899608135223,
"reward_std_all": 0.22802068293094635,
"rewards-max/forensics_tempsamp_iou_reward": 0.9585937559604645,
"rewards/forensics_tempsamp_iou_reward": 0.34702616930007935,
"step": 200
},
{
"completion_length": 20.515625,
"epoch": 0.8410041841004184,
"grad_norm": 2.982102632522583,
"kl": 0.2412109375,
"learning_rate": 7.907949790794979e-07,
"loss": 0.0097,
"reward": 0.46187594532966614,
"reward_std": 0.17189842462539673,
"reward_std_all": 0.2359485849738121,
"rewards-max/forensics_tempsamp_iou_reward": 0.9705944359302521,
"rewards/forensics_tempsamp_iou_reward": 0.40182164311408997,
"step": 201
},
{
"completion_length": 22.296875,
"epoch": 0.8451882845188284,
"grad_norm": 2.991445779800415,
"kl": 0.19873046875,
"learning_rate": 7.897489539748954e-07,
"loss": 0.008,
"reward": 0.36798734962940216,
"reward_std": 0.15709343552589417,
"reward_std_all": 0.26785267889499664,
"rewards-max/forensics_tempsamp_iou_reward": 0.8682383894920349,
"rewards/forensics_tempsamp_iou_reward": 0.2782354801893234,
"step": 202
},
{
"completion_length": 23.265625,
"epoch": 0.8493723849372385,
"grad_norm": 2.9757676124572754,
"kl": 0.19091796875,
"learning_rate": 7.887029288702929e-07,
"loss": 0.0077,
"reward": 0.4394252300262451,
"reward_std": 0.18786875158548355,
"reward_std_all": 0.25210700929164886,
"rewards-max/forensics_tempsamp_iou_reward": 0.9387990534305573,
"rewards/forensics_tempsamp_iou_reward": 0.3632575571537018,
"step": 203
},
{
"completion_length": 19.65625,
"epoch": 0.8535564853556485,
"grad_norm": 2.7130942344665527,
"kl": 0.294921875,
"learning_rate": 7.876569037656904e-07,
"loss": 0.0118,
"reward": 0.5747864544391632,
"reward_std": 0.16425180435180664,
"reward_std_all": 0.18271279335021973,
"rewards-max/forensics_tempsamp_iou_reward": 0.9329926073551178,
"rewards/forensics_tempsamp_iou_reward": 0.5177406668663025,
"step": 204
},
{
"completion_length": 22.890625,
"epoch": 0.8577405857740585,
"grad_norm": 2.594869375228882,
"kl": 0.2060546875,
"learning_rate": 7.866108786610879e-07,
"loss": 0.0082,
"reward": 0.5173429846763611,
"reward_std": 0.1691846027970314,
"reward_std_all": 0.20683391392230988,
"rewards-max/forensics_tempsamp_iou_reward": 0.9226298630237579,
"rewards/forensics_tempsamp_iou_reward": 0.44503675401210785,
"step": 205
},
{
"completion_length": 22.8125,
"epoch": 0.8619246861924686,
"grad_norm": 3.7502756118774414,
"kl": 0.21484375,
"learning_rate": 7.855648535564853e-07,
"loss": 0.0086,
"reward": 0.4609629958868027,
"reward_std": 0.19262301921844482,
"reward_std_all": 0.24545274674892426,
"rewards-max/forensics_tempsamp_iou_reward": 0.9291132092475891,
"rewards/forensics_tempsamp_iou_reward": 0.3958514332771301,
"step": 206
},
{
"completion_length": 21.0,
"epoch": 0.8661087866108786,
"grad_norm": 2.974710702896118,
"kl": 0.28125,
"learning_rate": 7.845188284518827e-07,
"loss": 0.0113,
"reward": 0.3848852068185806,
"reward_std": 0.20623034238815308,
"reward_std_all": 0.28727981448173523,
"rewards-max/forensics_tempsamp_iou_reward": 0.8507529199123383,
"rewards/forensics_tempsamp_iou_reward": 0.2996257245540619,
"step": 207
},
{
"completion_length": 21.78125,
"epoch": 0.8702928870292888,
"grad_norm": 2.7488296031951904,
"kl": 0.265625,
"learning_rate": 7.834728033472802e-07,
"loss": 0.0106,
"reward": 0.4288987070322037,
"reward_std": 0.1736207976937294,
"reward_std_all": 0.24638701975345612,
"rewards-max/forensics_tempsamp_iou_reward": 0.9122807085514069,
"rewards/forensics_tempsamp_iou_reward": 0.34523868560791016,
"step": 208
},
{
"completion_length": 20.546875,
"epoch": 0.8744769874476988,
"grad_norm": 2.767073392868042,
"kl": 0.24267578125,
"learning_rate": 7.824267782426778e-07,
"loss": 0.0097,
"reward": 0.5557944178581238,
"reward_std": 0.17601989954710007,
"reward_std_all": 0.19384042918682098,
"rewards-max/forensics_tempsamp_iou_reward": 0.9826851487159729,
"rewards/forensics_tempsamp_iou_reward": 0.49782395362854004,
"step": 209
},
{
"completion_length": 20.1875,
"epoch": 0.8786610878661087,
"grad_norm": 2.8378989696502686,
"kl": 0.21826171875,
"learning_rate": 7.813807531380753e-07,
"loss": 0.0087,
"reward": 0.515391081571579,
"reward_std": 0.179767906665802,
"reward_std_all": 0.22289501875638962,
"rewards-max/forensics_tempsamp_iou_reward": 0.8967886567115784,
"rewards/forensics_tempsamp_iou_reward": 0.44681820273399353,
"step": 210
},
{
"completion_length": 19.40625,
"epoch": 0.8828451882845189,
"grad_norm": 2.2181572914123535,
"kl": 0.2666015625,
"learning_rate": 7.803347280334728e-07,
"loss": 0.0107,
"reward": 0.477676123380661,
"reward_std": 0.16649889945983887,
"reward_std_all": 0.22404198348522186,
"rewards-max/forensics_tempsamp_iou_reward": 0.9306403994560242,
"rewards/forensics_tempsamp_iou_reward": 0.4057765007019043,
"step": 211
},
{
"completion_length": 19.828125,
"epoch": 0.8870292887029289,
"grad_norm": 3.1561989784240723,
"kl": 0.2314453125,
"learning_rate": 7.792887029288703e-07,
"loss": 0.0093,
"reward": 0.4781081974506378,
"reward_std": 0.20230144262313843,
"reward_std_all": 0.24319110065698624,
"rewards-max/forensics_tempsamp_iou_reward": 0.8857801258563995,
"rewards/forensics_tempsamp_iou_reward": 0.40602242946624756,
"step": 212
},
{
"completion_length": 20.75,
"epoch": 0.891213389121339,
"grad_norm": 2.478482723236084,
"kl": 0.23681640625,
"learning_rate": 7.782426778242678e-07,
"loss": 0.0095,
"reward": 0.4585840106010437,
"reward_std": 0.19420409947633743,
"reward_std_all": 0.24980199337005615,
"rewards-max/forensics_tempsamp_iou_reward": 0.8904297649860382,
"rewards/forensics_tempsamp_iou_reward": 0.39127199351787567,
"step": 213
},
{
"completion_length": 21.609375,
"epoch": 0.895397489539749,
"grad_norm": 2.8684754371643066,
"kl": 0.24609375,
"learning_rate": 7.771966527196653e-07,
"loss": 0.0099,
"reward": 0.5294999480247498,
"reward_std": 0.16930413246154785,
"reward_std_all": 0.1997930333018303,
"rewards-max/forensics_tempsamp_iou_reward": 0.9805516302585602,
"rewards/forensics_tempsamp_iou_reward": 0.4737071096897125,
"step": 214
},
{
"completion_length": 22.578125,
"epoch": 0.899581589958159,
"grad_norm": 2.982945680618286,
"kl": 0.19677734375,
"learning_rate": 7.761506276150628e-07,
"loss": 0.0079,
"reward": 0.36478492617607117,
"reward_std": 0.10291404277086258,
"reward_std_all": 0.24725479632616043,
"rewards-max/forensics_tempsamp_iou_reward": 0.7124014645814896,
"rewards/forensics_tempsamp_iou_reward": 0.2808893695473671,
"step": 215
},
{
"completion_length": 19.515625,
"epoch": 0.9037656903765691,
"grad_norm": 3.543671131134033,
"kl": 0.2880859375,
"learning_rate": 7.751046025104601e-07,
"loss": 0.0115,
"reward": 0.48197421431541443,
"reward_std": 0.15266630053520203,
"reward_std_all": 0.21370842307806015,
"rewards-max/forensics_tempsamp_iou_reward": 0.9019806981086731,
"rewards/forensics_tempsamp_iou_reward": 0.4163857698440552,
"step": 216
},
{
"completion_length": 21.234375,
"epoch": 0.9079497907949791,
"grad_norm": 3.4962265491485596,
"kl": 0.328125,
"learning_rate": 7.740585774058576e-07,
"loss": 0.0131,
"reward": 0.45272061228752136,
"reward_std": 0.17142578214406967,
"reward_std_all": 0.23806852102279663,
"rewards-max/forensics_tempsamp_iou_reward": 0.8259552419185638,
"rewards/forensics_tempsamp_iou_reward": 0.3735013008117676,
"step": 217
},
{
"completion_length": 20.28125,
"epoch": 0.9121338912133892,
"grad_norm": 2.3328404426574707,
"kl": 0.20703125,
"learning_rate": 7.730125523012552e-07,
"loss": 0.0083,
"reward": 0.5424121022224426,
"reward_std": 0.13422730565071106,
"reward_std_all": 0.18518050014972687,
"rewards-max/forensics_tempsamp_iou_reward": 0.8995182812213898,
"rewards/forensics_tempsamp_iou_reward": 0.49242454767227173,
"step": 218
},
{
"completion_length": 21.796875,
"epoch": 0.9163179916317992,
"grad_norm": 2.6833362579345703,
"kl": 0.2294921875,
"learning_rate": 7.719665271966527e-07,
"loss": 0.0092,
"reward": 0.4486705958843231,
"reward_std": 0.1046195887029171,
"reward_std_all": 0.20982079207897186,
"rewards-max/forensics_tempsamp_iou_reward": 0.8844920098781586,
"rewards/forensics_tempsamp_iou_reward": 0.3761271834373474,
"step": 219
},
{
"completion_length": 20.75,
"epoch": 0.9205020920502092,
"grad_norm": 2.917285203933716,
"kl": 0.2783203125,
"learning_rate": 7.709205020920502e-07,
"loss": 0.0111,
"reward": 0.5231565237045288,
"reward_std": 0.14112459123134613,
"reward_std_all": 0.19468006491661072,
"rewards-max/forensics_tempsamp_iou_reward": 0.8755681216716766,
"rewards/forensics_tempsamp_iou_reward": 0.4544021040201187,
"step": 220
},
{
"completion_length": 24.9375,
"epoch": 0.9246861924686193,
"grad_norm": 2.7568986415863037,
"kl": 0.23779296875,
"learning_rate": 7.698744769874477e-07,
"loss": 0.0095,
"reward": 0.39048123359680176,
"reward_std": 0.21019791066646576,
"reward_std_all": 0.28257623314857483,
"rewards-max/forensics_tempsamp_iou_reward": 0.9461603760719299,
"rewards/forensics_tempsamp_iou_reward": 0.3105696588754654,
"step": 221
},
{
"completion_length": 22.0625,
"epoch": 0.9288702928870293,
"grad_norm": 2.8259966373443604,
"kl": 0.30419921875,
"learning_rate": 7.688284518828452e-07,
"loss": 0.0122,
"reward": 0.38780489563941956,
"reward_std": 0.1341216042637825,
"reward_std_all": 0.2435222789645195,
"rewards-max/forensics_tempsamp_iou_reward": 0.8021270334720612,
"rewards/forensics_tempsamp_iou_reward": 0.2885711193084717,
"step": 222
},
{
"completion_length": 21.171875,
"epoch": 0.9330543933054394,
"grad_norm": 2.7412943840026855,
"kl": 0.25146484375,
"learning_rate": 7.677824267782427e-07,
"loss": 0.0101,
"reward": 0.4320545047521591,
"reward_std": 0.18888843059539795,
"reward_std_all": 0.25703153014183044,
"rewards-max/forensics_tempsamp_iou_reward": 0.888391762971878,
"rewards/forensics_tempsamp_iou_reward": 0.3529551327228546,
"step": 223
},
{
"completion_length": 22.90625,
"epoch": 0.9372384937238494,
"grad_norm": 2.965860605239868,
"kl": 0.34375,
"learning_rate": 7.667364016736401e-07,
"loss": 0.0137,
"reward": 0.5466612279415131,
"reward_std": 0.09992274641990662,
"reward_std_all": 0.1602260321378708,
"rewards-max/forensics_tempsamp_iou_reward": 0.949523538351059,
"rewards/forensics_tempsamp_iou_reward": 0.48074495792388916,
"step": 224
},
{
"completion_length": 22.828125,
"epoch": 0.9414225941422594,
"grad_norm": 3.049088954925537,
"kl": 0.255859375,
"learning_rate": 7.656903765690376e-07,
"loss": 0.0103,
"reward": 0.5095895826816559,
"reward_std": 0.1619444638490677,
"reward_std_all": 0.21143320947885513,
"rewards-max/forensics_tempsamp_iou_reward": 0.8608218133449554,
"rewards/forensics_tempsamp_iou_reward": 0.43393535912036896,
"step": 225
},
{
"completion_length": 26.796875,
"epoch": 0.9456066945606695,
"grad_norm": 2.5555121898651123,
"kl": 0.26513671875,
"learning_rate": 7.64644351464435e-07,
"loss": 0.0106,
"reward": 0.38324546813964844,
"reward_std": 0.09288034588098526,
"reward_std_all": 0.2360876500606537,
"rewards-max/forensics_tempsamp_iou_reward": 0.8620824813842773,
"rewards/forensics_tempsamp_iou_reward": 0.2976876348257065,
"step": 226
},
{
"completion_length": 22.828125,
"epoch": 0.9497907949790795,
"grad_norm": 2.714383840560913,
"kl": 0.271484375,
"learning_rate": 7.635983263598326e-07,
"loss": 0.0108,
"reward": 0.4541253298521042,
"reward_std": 0.11566142737865448,
"reward_std_all": 0.21116336435079575,
"rewards-max/forensics_tempsamp_iou_reward": 0.9491065740585327,
"rewards/forensics_tempsamp_iou_reward": 0.38119879364967346,
"step": 227
},
{
"completion_length": 23.34375,
"epoch": 0.9539748953974896,
"grad_norm": 2.4098281860351562,
"kl": 0.26171875,
"learning_rate": 7.625523012552301e-07,
"loss": 0.0105,
"reward": 0.5027758330106735,
"reward_std": 0.185209259390831,
"reward_std_all": 0.22052696347236633,
"rewards-max/forensics_tempsamp_iou_reward": 0.9107955694198608,
"rewards/forensics_tempsamp_iou_reward": 0.42998918890953064,
"step": 228
},
{
"completion_length": 24.1875,
"epoch": 0.9581589958158996,
"grad_norm": 2.6245439052581787,
"kl": 0.25634765625,
"learning_rate": 7.615062761506276e-07,
"loss": 0.0102,
"reward": 0.4221508800983429,
"reward_std": 0.11530141159892082,
"reward_std_all": 0.23151814937591553,
"rewards-max/forensics_tempsamp_iou_reward": 0.9399924874305725,
"rewards/forensics_tempsamp_iou_reward": 0.3421252369880676,
"step": 229
},
{
"completion_length": 22.90625,
"epoch": 0.9623430962343096,
"grad_norm": 3.325984239578247,
"kl": 0.2890625,
"learning_rate": 7.604602510460251e-07,
"loss": 0.0116,
"reward": 0.4244838356971741,
"reward_std": 0.1300276182591915,
"reward_std_all": 0.2310565933585167,
"rewards-max/forensics_tempsamp_iou_reward": 0.9081459641456604,
"rewards/forensics_tempsamp_iou_reward": 0.34608910977840424,
"step": 230
},
{
"completion_length": 23.796875,
"epoch": 0.9665271966527197,
"grad_norm": 3.1329855918884277,
"kl": 0.291015625,
"learning_rate": 7.594142259414226e-07,
"loss": 0.0117,
"reward": 0.4453258365392685,
"reward_std": 0.1702342852950096,
"reward_std_all": 0.23973409831523895,
"rewards-max/forensics_tempsamp_iou_reward": 0.9037531912326813,
"rewards/forensics_tempsamp_iou_reward": 0.3643639385700226,
"step": 231
},
{
"completion_length": 24.703125,
"epoch": 0.9707112970711297,
"grad_norm": 2.8671205043792725,
"kl": 0.24267578125,
"learning_rate": 7.5836820083682e-07,
"loss": 0.0097,
"reward": 0.4461572468280792,
"reward_std": 0.18001391738653183,
"reward_std_all": 0.2440803274512291,
"rewards-max/forensics_tempsamp_iou_reward": 0.8942068815231323,
"rewards/forensics_tempsamp_iou_reward": 0.3725469410419464,
"step": 232
},
{
"completion_length": 26.515625,
"epoch": 0.9748953974895398,
"grad_norm": 2.6558263301849365,
"kl": 0.244140625,
"learning_rate": 7.573221757322175e-07,
"loss": 0.0098,
"reward": 0.4009731262922287,
"reward_std": 0.13910868018865585,
"reward_std_all": 0.2447948008775711,
"rewards-max/forensics_tempsamp_iou_reward": 0.8860389590263367,
"rewards/forensics_tempsamp_iou_reward": 0.31428056955337524,
"step": 233
},
{
"completion_length": 22.359375,
"epoch": 0.9790794979079498,
"grad_norm": 2.8703744411468506,
"kl": 0.27734375,
"learning_rate": 7.56276150627615e-07,
"loss": 0.0111,
"reward": 0.4656105935573578,
"reward_std": 0.13129105418920517,
"reward_std_all": 0.21583713591098785,
"rewards-max/forensics_tempsamp_iou_reward": 0.9010334014892578,
"rewards/forensics_tempsamp_iou_reward": 0.39821185171604156,
"step": 234
},
{
"completion_length": 22.953125,
"epoch": 0.9832635983263598,
"grad_norm": 2.4311509132385254,
"kl": 0.21484375,
"learning_rate": 7.552301255230126e-07,
"loss": 0.0086,
"reward": 0.5233648419380188,
"reward_std": 0.11986519396305084,
"reward_std_all": 0.17819354683160782,
"rewards-max/forensics_tempsamp_iou_reward": 0.9213725328445435,
"rewards/forensics_tempsamp_iou_reward": 0.45844584703445435,
"step": 235
},
{
"completion_length": 21.15625,
"epoch": 0.9874476987447699,
"grad_norm": 2.8223154544830322,
"kl": 0.2392578125,
"learning_rate": 7.5418410041841e-07,
"loss": 0.0095,
"reward": 0.49878184497356415,
"reward_std": 0.1817081719636917,
"reward_std_all": 0.22260653227567673,
"rewards-max/forensics_tempsamp_iou_reward": 0.9375977218151093,
"rewards/forensics_tempsamp_iou_reward": 0.42771054804325104,
"step": 236
},
{
"completion_length": 20.71875,
"epoch": 0.9916317991631799,
"grad_norm": 3.2318050861358643,
"kl": 0.2255859375,
"learning_rate": 7.531380753138075e-07,
"loss": 0.009,
"reward": 0.5772554576396942,
"reward_std": 0.11502983421087265,
"reward_std_all": 0.1570388525724411,
"rewards-max/forensics_tempsamp_iou_reward": 0.9452534019947052,
"rewards/forensics_tempsamp_iou_reward": 0.5242081880569458,
"step": 237
},
{
"completion_length": 21.28125,
"epoch": 0.99581589958159,
"grad_norm": 2.953636407852173,
"kl": 0.1982421875,
"learning_rate": 7.52092050209205e-07,
"loss": 0.0079,
"reward": 0.4944388121366501,
"reward_std": 0.143937349319458,
"reward_std_all": 0.2036518156528473,
"rewards-max/forensics_tempsamp_iou_reward": 0.9042218625545502,
"rewards/forensics_tempsamp_iou_reward": 0.43114589154720306,
"step": 238
},
{
"completion_length": 24.142858505249023,
"epoch": 1.0,
"grad_norm": 2.4333882331848145,
"kl": 0.181640625,
"learning_rate": 7.510460251046025e-07,
"loss": 0.0073,
"reward": 0.5391118973493576,
"reward_std": 0.11269606649875641,
"reward_std_all": 0.17295445501804352,
"rewards-max/forensics_tempsamp_iou_reward": 0.774907112121582,
"rewards/forensics_tempsamp_iou_reward": 0.48869478702545166,
"step": 239
},
{
"completion_length": 20.5,
"epoch": 1.00418410041841,
"grad_norm": 3.1751489639282227,
"kl": 0.21337890625,
"learning_rate": 7.5e-07,
"loss": 0.0085,
"reward": 0.5711886286735535,
"reward_std": 0.1397424414753914,
"reward_std_all": 0.17281435430049896,
"rewards-max/forensics_tempsamp_iou_reward": 0.9726583659648895,
"rewards/forensics_tempsamp_iou_reward": 0.5212274789810181,
"step": 240
},
{
"completion_length": 20.5,
"epoch": 1.00836820083682,
"grad_norm": 2.9954943656921387,
"kl": 0.28125,
"learning_rate": 7.489539748953974e-07,
"loss": 0.0113,
"reward": 0.585724949836731,
"reward_std": 0.12797259539365768,
"reward_std_all": 0.16164620965719223,
"rewards-max/forensics_tempsamp_iou_reward": 0.9676521420478821,
"rewards/forensics_tempsamp_iou_reward": 0.5425120890140533,
"step": 241
},
{
"completion_length": 23.140625,
"epoch": 1.0125523012552302,
"grad_norm": 2.822831153869629,
"kl": 0.2080078125,
"learning_rate": 7.479079497907949e-07,
"loss": 0.0083,
"reward": 0.5188894271850586,
"reward_std": 0.16484617441892624,
"reward_std_all": 0.2060597836971283,
"rewards-max/forensics_tempsamp_iou_reward": 0.8710611462593079,
"rewards/forensics_tempsamp_iou_reward": 0.4425643980503082,
"step": 242
},
{
"completion_length": 22.0,
"epoch": 1.0167364016736402,
"grad_norm": 2.8700313568115234,
"kl": 0.26171875,
"learning_rate": 7.468619246861925e-07,
"loss": 0.0105,
"reward": 0.5465741753578186,
"reward_std": 0.0680680051445961,
"reward_std_all": 0.14961640536785126,
"rewards-max/forensics_tempsamp_iou_reward": 0.892967015504837,
"rewards/forensics_tempsamp_iou_reward": 0.48684388399124146,
"step": 243
},
{
"completion_length": 21.21875,
"epoch": 1.0209205020920502,
"grad_norm": 2.9829611778259277,
"kl": 0.279296875,
"learning_rate": 7.4581589958159e-07,
"loss": 0.0112,
"reward": 0.5293379426002502,
"reward_std": 0.13763121888041496,
"reward_std_all": 0.1869383379817009,
"rewards-max/forensics_tempsamp_iou_reward": 0.8995577096939087,
"rewards/forensics_tempsamp_iou_reward": 0.46375414729118347,
"step": 244
},
{
"completion_length": 20.71875,
"epoch": 1.0251046025104602,
"grad_norm": 2.8357324600219727,
"kl": 0.23388671875,
"learning_rate": 7.447698744769874e-07,
"loss": 0.0094,
"reward": 0.5194398462772369,
"reward_std": 0.14051742106676102,
"reward_std_all": 0.18857649713754654,
"rewards-max/forensics_tempsamp_iou_reward": 0.9733692109584808,
"rewards/forensics_tempsamp_iou_reward": 0.4700639247894287,
"step": 245
},
{
"completion_length": 21.53125,
"epoch": 1.0292887029288702,
"grad_norm": 3.7512168884277344,
"kl": 0.22314453125,
"learning_rate": 7.437238493723849e-07,
"loss": 0.0089,
"reward": 0.5247024297714233,
"reward_std": 0.1112057976424694,
"reward_std_all": 0.1765020489692688,
"rewards-max/forensics_tempsamp_iou_reward": 0.8781596422195435,
"rewards/forensics_tempsamp_iou_reward": 0.4491224139928818,
"step": 246
},
{
"completion_length": 24.203125,
"epoch": 1.0334728033472804,
"grad_norm": 2.5578560829162598,
"kl": 0.25341796875,
"learning_rate": 7.426778242677824e-07,
"loss": 0.0101,
"reward": 0.5313106775283813,
"reward_std": 0.133583664894104,
"reward_std_all": 0.1826692670583725,
"rewards-max/forensics_tempsamp_iou_reward": 0.8455827534198761,
"rewards/forensics_tempsamp_iou_reward": 0.45618388056755066,
"step": 247
},
{
"completion_length": 22.03125,
"epoch": 1.0376569037656904,
"grad_norm": 2.791870355606079,
"kl": 0.21875,
"learning_rate": 7.416317991631799e-07,
"loss": 0.0087,
"reward": 0.4771214425563812,
"reward_std": 0.12686984986066818,
"reward_std_all": 0.2107955664396286,
"rewards-max/forensics_tempsamp_iou_reward": 0.9012301862239838,
"rewards/forensics_tempsamp_iou_reward": 0.411191388964653,
"step": 248
},
{
"completion_length": 23.890625,
"epoch": 1.0418410041841004,
"grad_norm": 2.552312135696411,
"kl": 0.22021484375,
"learning_rate": 7.405857740585773e-07,
"loss": 0.0088,
"reward": 0.5310852527618408,
"reward_std": 0.14674681425094604,
"reward_std_all": 0.19230832159519196,
"rewards-max/forensics_tempsamp_iou_reward": 0.915603905916214,
"rewards/forensics_tempsamp_iou_reward": 0.46750861406326294,
"step": 249
},
{
"completion_length": 22.515625,
"epoch": 1.0460251046025104,
"grad_norm": 3.9564101696014404,
"kl": 0.21240234375,
"learning_rate": 7.395397489539748e-07,
"loss": 0.0085,
"reward": 0.5883689522743225,
"reward_std": 0.16731204837560654,
"reward_std_all": 0.18058864772319794,
"rewards-max/forensics_tempsamp_iou_reward": 0.9943610429763794,
"rewards/forensics_tempsamp_iou_reward": 0.5530558228492737,
"step": 250
},
{
"completion_length": 21.46875,
"epoch": 1.0502092050209204,
"grad_norm": 3.151780605316162,
"kl": 0.2412109375,
"learning_rate": 7.384937238493723e-07,
"loss": 0.0096,
"reward": 0.5892803966999054,
"reward_std": 0.1454412043094635,
"reward_std_all": 0.1636819764971733,
"rewards-max/forensics_tempsamp_iou_reward": 0.9695702195167542,
"rewards/forensics_tempsamp_iou_reward": 0.5499061197042465,
"step": 251
},
{
"completion_length": 23.3125,
"epoch": 1.0543933054393306,
"grad_norm": 3.610527276992798,
"kl": 0.27001953125,
"learning_rate": 7.374476987447699e-07,
"loss": 0.0108,
"reward": 0.48310159146785736,
"reward_std": 0.12885990366339684,
"reward_std_all": 0.2007329985499382,
"rewards-max/forensics_tempsamp_iou_reward": 0.816636472940445,
"rewards/forensics_tempsamp_iou_reward": 0.3920867145061493,
"step": 252
},
{
"completion_length": 22.671875,
"epoch": 1.0585774058577406,
"grad_norm": 2.9160239696502686,
"kl": 0.2734375,
"learning_rate": 7.364016736401674e-07,
"loss": 0.0109,
"reward": 0.5183329582214355,
"reward_std": 0.18024856597185135,
"reward_std_all": 0.21802489459514618,
"rewards-max/forensics_tempsamp_iou_reward": 0.9295851588249207,
"rewards/forensics_tempsamp_iou_reward": 0.45799025893211365,
"step": 253
},
{
"completion_length": 20.125,
"epoch": 1.0627615062761506,
"grad_norm": 2.984461545944214,
"kl": 0.25244140625,
"learning_rate": 7.353556485355649e-07,
"loss": 0.0101,
"reward": 0.5660831928253174,
"reward_std": 0.1368194781243801,
"reward_std_all": 0.16809197515249252,
"rewards-max/forensics_tempsamp_iou_reward": 0.9190649092197418,
"rewards/forensics_tempsamp_iou_reward": 0.5205181390047073,
"step": 254
},
{
"completion_length": 22.4375,
"epoch": 1.0669456066945606,
"grad_norm": 3.1417200565338135,
"kl": 0.24365234375,
"learning_rate": 7.343096234309623e-07,
"loss": 0.0098,
"reward": 0.43124328553676605,
"reward_std": 0.1375356838107109,
"reward_std_all": 0.23387806862592697,
"rewards-max/forensics_tempsamp_iou_reward": 0.9054712057113647,
"rewards/forensics_tempsamp_iou_reward": 0.3502599895000458,
"step": 255
},
{
"completion_length": 21.46875,
"epoch": 1.0711297071129706,
"grad_norm": 2.845176935195923,
"kl": 0.23046875,
"learning_rate": 7.332635983263598e-07,
"loss": 0.0092,
"reward": 0.5145905911922455,
"reward_std": 0.11194794625043869,
"reward_std_all": 0.1824449747800827,
"rewards-max/forensics_tempsamp_iou_reward": 0.9760283827781677,
"rewards/forensics_tempsamp_iou_reward": 0.4643611013889313,
"step": 256
},
{
"completion_length": 21.796875,
"epoch": 1.0753138075313808,
"grad_norm": 2.875415086746216,
"kl": 0.18603515625,
"learning_rate": 7.322175732217572e-07,
"loss": 0.0075,
"reward": 0.5560543537139893,
"reward_std": 0.1196829341351986,
"reward_std_all": 0.16810527443885803,
"rewards-max/forensics_tempsamp_iou_reward": 0.9502838551998138,
"rewards/forensics_tempsamp_iou_reward": 0.4933578073978424,
"step": 257
},
{
"completion_length": 21.34375,
"epoch": 1.0794979079497908,
"grad_norm": 3.432624340057373,
"kl": 0.2490234375,
"learning_rate": 7.311715481171547e-07,
"loss": 0.01,
"reward": 0.44275417923927307,
"reward_std": 0.12824781239032745,
"reward_std_all": 0.2230178862810135,
"rewards-max/forensics_tempsamp_iou_reward": 0.8501627743244171,
"rewards/forensics_tempsamp_iou_reward": 0.36633414030075073,
"step": 258
},
{
"completion_length": 23.21875,
"epoch": 1.0836820083682008,
"grad_norm": 3.3391928672790527,
"kl": 0.2783203125,
"learning_rate": 7.301255230125522e-07,
"loss": 0.0112,
"reward": 0.44474785029888153,
"reward_std": 0.127333864569664,
"reward_std_all": 0.21916208416223526,
"rewards-max/forensics_tempsamp_iou_reward": 0.8645506203174591,
"rewards/forensics_tempsamp_iou_reward": 0.36465153098106384,
"step": 259
},
{
"completion_length": 23.890625,
"epoch": 1.0878661087866108,
"grad_norm": 3.082578420639038,
"kl": 0.2451171875,
"learning_rate": 7.290794979079497e-07,
"loss": 0.0098,
"reward": 0.5212238729000092,
"reward_std": 0.10541628301143646,
"reward_std_all": 0.17908016592264175,
"rewards-max/forensics_tempsamp_iou_reward": 0.9139945209026337,
"rewards/forensics_tempsamp_iou_reward": 0.44380728900432587,
"step": 260
},
{
"completion_length": 22.140625,
"epoch": 1.0920502092050208,
"grad_norm": 2.71238374710083,
"kl": 0.23876953125,
"learning_rate": 7.280334728033473e-07,
"loss": 0.0096,
"reward": 0.4448404759168625,
"reward_std": 0.14464901387691498,
"reward_std_all": 0.23661063611507416,
"rewards-max/forensics_tempsamp_iou_reward": 0.8963099420070648,
"rewards/forensics_tempsamp_iou_reward": 0.382612943649292,
"step": 261
},
{
"completion_length": 21.390625,
"epoch": 1.096234309623431,
"grad_norm": 3.261598825454712,
"kl": 0.2685546875,
"learning_rate": 7.269874476987448e-07,
"loss": 0.0107,
"reward": 0.4150901138782501,
"reward_std": 0.12200522422790527,
"reward_std_all": 0.23671426624059677,
"rewards-max/forensics_tempsamp_iou_reward": 0.8627020120620728,
"rewards/forensics_tempsamp_iou_reward": 0.34210868179798126,
"step": 262
},
{
"completion_length": 22.21875,
"epoch": 1.100418410041841,
"grad_norm": 4.187417030334473,
"kl": 0.2900390625,
"learning_rate": 7.259414225941423e-07,
"loss": 0.0116,
"reward": 0.4896872639656067,
"reward_std": 0.18468395620584488,
"reward_std_all": 0.23245997726917267,
"rewards-max/forensics_tempsamp_iou_reward": 0.9749446213245392,
"rewards/forensics_tempsamp_iou_reward": 0.434678316116333,
"step": 263
},
{
"completion_length": 24.65625,
"epoch": 1.104602510460251,
"grad_norm": 2.8712642192840576,
"kl": 0.2392578125,
"learning_rate": 7.248953974895398e-07,
"loss": 0.0096,
"reward": 0.4625113904476166,
"reward_std": 0.1280992068350315,
"reward_std_all": 0.2170550897717476,
"rewards-max/forensics_tempsamp_iou_reward": 0.8755057752132416,
"rewards/forensics_tempsamp_iou_reward": 0.37597933411598206,
"step": 264
},
{
"completion_length": 26.921875,
"epoch": 1.108786610878661,
"grad_norm": 2.9239959716796875,
"kl": 0.24169921875,
"learning_rate": 7.238493723849371e-07,
"loss": 0.0097,
"reward": 0.5131942927837372,
"reward_std": 0.11525341495871544,
"reward_std_all": 0.18406273424625397,
"rewards-max/forensics_tempsamp_iou_reward": 0.944828987121582,
"rewards/forensics_tempsamp_iou_reward": 0.4416406899690628,
"step": 265
},
{
"completion_length": 19.875,
"epoch": 1.112970711297071,
"grad_norm": 3.7060437202453613,
"kl": 0.2421875,
"learning_rate": 7.228033472803346e-07,
"loss": 0.0097,
"reward": 0.5995339453220367,
"reward_std": 0.13496682047843933,
"reward_std_all": 0.1582745462656021,
"rewards-max/forensics_tempsamp_iou_reward": 0.9685847461223602,
"rewards/forensics_tempsamp_iou_reward": 0.5870585739612579,
"step": 266
},
{
"completion_length": 21.859375,
"epoch": 1.1171548117154813,
"grad_norm": 2.865807056427002,
"kl": 0.21533203125,
"learning_rate": 7.217573221757321e-07,
"loss": 0.0086,
"reward": 0.5211834013462067,
"reward_std": 0.1269640475511551,
"reward_std_all": 0.18592801690101624,
"rewards-max/forensics_tempsamp_iou_reward": 0.908519446849823,
"rewards/forensics_tempsamp_iou_reward": 0.45511171221733093,
"step": 267
},
{
"completion_length": 24.359375,
"epoch": 1.1213389121338913,
"grad_norm": 3.3377957344055176,
"kl": 0.224609375,
"learning_rate": 7.207112970711296e-07,
"loss": 0.009,
"reward": 0.4746879041194916,
"reward_std": 0.17468178272247314,
"reward_std_all": 0.23382113873958588,
"rewards-max/forensics_tempsamp_iou_reward": 0.8943271934986115,
"rewards/forensics_tempsamp_iou_reward": 0.4027795195579529,
"step": 268
},
{
"completion_length": 20.3125,
"epoch": 1.1255230125523012,
"grad_norm": 3.0716865062713623,
"kl": 0.28515625,
"learning_rate": 7.196652719665272e-07,
"loss": 0.0114,
"reward": 0.5112700909376144,
"reward_std": 0.16749513149261475,
"reward_std_all": 0.21662598848342896,
"rewards-max/forensics_tempsamp_iou_reward": 0.9814382791519165,
"rewards/forensics_tempsamp_iou_reward": 0.4492391347885132,
"step": 269
},
{
"completion_length": 21.984375,
"epoch": 1.1297071129707112,
"grad_norm": 3.7890193462371826,
"kl": 0.25830078125,
"learning_rate": 7.186192468619247e-07,
"loss": 0.0104,
"reward": 0.4063350707292557,
"reward_std": 0.10845069959759712,
"reward_std_all": 0.23198890686035156,
"rewards-max/forensics_tempsamp_iou_reward": 0.8708868622779846,
"rewards/forensics_tempsamp_iou_reward": 0.31159640848636627,
"step": 270
},
{
"completion_length": 21.1875,
"epoch": 1.1338912133891212,
"grad_norm": 3.060938596725464,
"kl": 0.27099609375,
"learning_rate": 7.175732217573222e-07,
"loss": 0.0108,
"reward": 0.5395030975341797,
"reward_std": 0.07365496270358562,
"reward_std_all": 0.1550772786140442,
"rewards-max/forensics_tempsamp_iou_reward": 0.9208497107028961,
"rewards/forensics_tempsamp_iou_reward": 0.48623892664909363,
"step": 271
},
{
"completion_length": 21.703125,
"epoch": 1.1380753138075315,
"grad_norm": 3.056727886199951,
"kl": 0.2255859375,
"learning_rate": 7.165271966527197e-07,
"loss": 0.009,
"reward": 0.4906967580318451,
"reward_std": 0.17550615221261978,
"reward_std_all": 0.22429701685905457,
"rewards-max/forensics_tempsamp_iou_reward": 0.961403489112854,
"rewards/forensics_tempsamp_iou_reward": 0.4303285777568817,
"step": 272
},
{
"completion_length": 22.46875,
"epoch": 1.1422594142259415,
"grad_norm": 2.8386991024017334,
"kl": 0.2265625,
"learning_rate": 7.154811715481172e-07,
"loss": 0.0091,
"reward": 0.5909722447395325,
"reward_std": 0.1389872282743454,
"reward_std_all": 0.16256901621818542,
"rewards-max/forensics_tempsamp_iou_reward": 0.9788158237934113,
"rewards/forensics_tempsamp_iou_reward": 0.5412116944789886,
"step": 273
},
{
"completion_length": 20.75,
"epoch": 1.1464435146443515,
"grad_norm": 3.271554946899414,
"kl": 0.2822265625,
"learning_rate": 7.144351464435146e-07,
"loss": 0.0113,
"reward": 0.569398432970047,
"reward_std": 0.1256658136844635,
"reward_std_all": 0.16598975658416748,
"rewards-max/forensics_tempsamp_iou_reward": 0.9748988747596741,
"rewards/forensics_tempsamp_iou_reward": 0.5282865166664124,
"step": 274
},
{
"completion_length": 21.375,
"epoch": 1.1506276150627615,
"grad_norm": 3.6734707355499268,
"kl": 0.3564453125,
"learning_rate": 7.13389121338912e-07,
"loss": 0.0143,
"reward": 0.4940859377384186,
"reward_std": 0.12352230027318001,
"reward_std_all": 0.19736018776893616,
"rewards-max/forensics_tempsamp_iou_reward": 0.9744027853012085,
"rewards/forensics_tempsamp_iou_reward": 0.43466606736183167,
"step": 275
},
{
"completion_length": 21.703125,
"epoch": 1.1548117154811715,
"grad_norm": 2.669881820678711,
"kl": 0.24072265625,
"learning_rate": 7.123430962343095e-07,
"loss": 0.0096,
"reward": 0.4963110089302063,
"reward_std": 0.1382359266281128,
"reward_std_all": 0.2097702994942665,
"rewards-max/forensics_tempsamp_iou_reward": 0.9712812900543213,
"rewards/forensics_tempsamp_iou_reward": 0.4381861835718155,
"step": 276
},
{
"completion_length": 23.609375,
"epoch": 1.1589958158995817,
"grad_norm": 4.182821273803711,
"kl": 0.33203125,
"learning_rate": 7.11297071129707e-07,
"loss": 0.0133,
"reward": 0.49152572453022003,
"reward_std": 0.14198710024356842,
"reward_std_all": 0.21040713042020798,
"rewards-max/forensics_tempsamp_iou_reward": 0.9207504987716675,
"rewards/forensics_tempsamp_iou_reward": 0.43109850585460663,
"step": 277
},
{
"completion_length": 22.875,
"epoch": 1.1631799163179917,
"grad_norm": 2.991118907928467,
"kl": 0.3427734375,
"learning_rate": 7.102510460251046e-07,
"loss": 0.0137,
"reward": 0.5004332214593887,
"reward_std": 0.11465354263782501,
"reward_std_all": 0.1961006373167038,
"rewards-max/forensics_tempsamp_iou_reward": 0.8824974000453949,
"rewards/forensics_tempsamp_iou_reward": 0.4250725358724594,
"step": 278
},
{
"completion_length": 20.109375,
"epoch": 1.1673640167364017,
"grad_norm": 4.143142223358154,
"kl": 0.2763671875,
"learning_rate": 7.092050209205021e-07,
"loss": 0.0111,
"reward": 0.5680044591426849,
"reward_std": 0.11133379489183426,
"reward_std_all": 0.16162142902612686,
"rewards-max/forensics_tempsamp_iou_reward": 0.9112753868103027,
"rewards/forensics_tempsamp_iou_reward": 0.5280047357082367,
"step": 279
},
{
"completion_length": 21.984375,
"epoch": 1.1715481171548117,
"grad_norm": 3.288547992706299,
"kl": 0.271484375,
"learning_rate": 7.081589958158996e-07,
"loss": 0.0109,
"reward": 0.4785521775484085,
"reward_std": 0.16663628816604614,
"reward_std_all": 0.21978308260440826,
"rewards-max/forensics_tempsamp_iou_reward": 0.882258802652359,
"rewards/forensics_tempsamp_iou_reward": 0.4001775234937668,
"step": 280
},
{
"completion_length": 20.390625,
"epoch": 1.1757322175732217,
"grad_norm": 3.340106964111328,
"kl": 0.25048828125,
"learning_rate": 7.071129707112971e-07,
"loss": 0.01,
"reward": 0.4259641170501709,
"reward_std": 0.1054859384894371,
"reward_std_all": 0.22115624696016312,
"rewards-max/forensics_tempsamp_iou_reward": 0.8478197753429413,
"rewards/forensics_tempsamp_iou_reward": 0.3553828001022339,
"step": 281
},
{
"completion_length": 22.46875,
"epoch": 1.1799163179916319,
"grad_norm": 2.8793787956237793,
"kl": 0.23486328125,
"learning_rate": 7.060669456066946e-07,
"loss": 0.0094,
"reward": 0.5125273913145065,
"reward_std": 0.12388189882040024,
"reward_std_all": 0.18644868582487106,
"rewards-max/forensics_tempsamp_iou_reward": 0.92188560962677,
"rewards/forensics_tempsamp_iou_reward": 0.4622008502483368,
"step": 282
},
{
"completion_length": 21.234375,
"epoch": 1.1841004184100419,
"grad_norm": 2.8642048835754395,
"kl": 0.24072265625,
"learning_rate": 7.05020920502092e-07,
"loss": 0.0096,
"reward": 0.5907235741615295,
"reward_std": 0.1389547698199749,
"reward_std_all": 0.16323034465312958,
"rewards-max/forensics_tempsamp_iou_reward": 0.9475938975811005,
"rewards/forensics_tempsamp_iou_reward": 0.5342935025691986,
"step": 283
},
{
"completion_length": 19.4375,
"epoch": 1.1882845188284519,
"grad_norm": 2.781069755554199,
"kl": 0.3134765625,
"learning_rate": 7.039748953974895e-07,
"loss": 0.0125,
"reward": 0.56010302901268,
"reward_std": 0.12104970961809158,
"reward_std_all": 0.16579771786928177,
"rewards-max/forensics_tempsamp_iou_reward": 0.7700367569923401,
"rewards/forensics_tempsamp_iou_reward": 0.48219195008277893,
"step": 284
},
{
"completion_length": 22.109375,
"epoch": 1.1924686192468619,
"grad_norm": 3.619520425796509,
"kl": 0.3037109375,
"learning_rate": 7.029288702928869e-07,
"loss": 0.0121,
"reward": 0.5277376472949982,
"reward_std": 0.17717893794178963,
"reward_std_all": 0.21588369458913803,
"rewards-max/forensics_tempsamp_iou_reward": 0.9468532204627991,
"rewards/forensics_tempsamp_iou_reward": 0.4742322713136673,
"step": 285
},
{
"completion_length": 20.484375,
"epoch": 1.196652719665272,
"grad_norm": 2.9519553184509277,
"kl": 0.3046875,
"learning_rate": 7.018828451882844e-07,
"loss": 0.0122,
"reward": 0.5457387119531631,
"reward_std": 0.1211397834122181,
"reward_std_all": 0.176845483481884,
"rewards-max/forensics_tempsamp_iou_reward": 0.9710467755794525,
"rewards/forensics_tempsamp_iou_reward": 0.49382247030735016,
"step": 286
},
{
"completion_length": 20.796875,
"epoch": 1.200836820083682,
"grad_norm": 3.07560658454895,
"kl": 0.18017578125,
"learning_rate": 7.00836820083682e-07,
"loss": 0.0072,
"reward": 0.4588312804698944,
"reward_std": 0.12586675584316254,
"reward_std_all": 0.2182418629527092,
"rewards-max/forensics_tempsamp_iou_reward": 0.916353851556778,
"rewards/forensics_tempsamp_iou_reward": 0.38211433589458466,
"step": 287
},
{
"completion_length": 21.328125,
"epoch": 1.205020920502092,
"grad_norm": 3.0007317066192627,
"kl": 0.2373046875,
"learning_rate": 6.997907949790795e-07,
"loss": 0.0095,
"reward": 0.6140807271003723,
"reward_std": 0.1069311872124672,
"reward_std_all": 0.1367195099592209,
"rewards-max/forensics_tempsamp_iou_reward": 0.9812987446784973,
"rewards/forensics_tempsamp_iou_reward": 0.5816353261470795,
"step": 288
},
{
"completion_length": 21.921875,
"epoch": 1.209205020920502,
"grad_norm": 3.1988589763641357,
"kl": 0.3095703125,
"learning_rate": 6.98744769874477e-07,
"loss": 0.0124,
"reward": 0.5624093115329742,
"reward_std": 0.17472976446151733,
"reward_std_all": 0.1960398107767105,
"rewards-max/forensics_tempsamp_iou_reward": 0.9856593608856201,
"rewards/forensics_tempsamp_iou_reward": 0.5169031322002411,
"step": 289
},
{
"completion_length": 22.359375,
"epoch": 1.213389121338912,
"grad_norm": 2.425137758255005,
"kl": 0.248046875,
"learning_rate": 6.976987447698745e-07,
"loss": 0.0099,
"reward": 0.5196384191513062,
"reward_std": 0.09962593019008636,
"reward_std_all": 0.17626073211431503,
"rewards-max/forensics_tempsamp_iou_reward": 0.9767921566963196,
"rewards/forensics_tempsamp_iou_reward": 0.4731341749429703,
"step": 290
},
{
"completion_length": 20.1875,
"epoch": 1.217573221757322,
"grad_norm": 3.178602695465088,
"kl": 0.24072265625,
"learning_rate": 6.966527196652719e-07,
"loss": 0.0097,
"reward": 0.6012136340141296,
"reward_std": 0.13446548581123352,
"reward_std_all": 0.15242978185415268,
"rewards-max/forensics_tempsamp_iou_reward": 0.9009915292263031,
"rewards/forensics_tempsamp_iou_reward": 0.549381822347641,
"step": 291
},
{
"completion_length": 23.328125,
"epoch": 1.2217573221757323,
"grad_norm": 3.3315846920013428,
"kl": 0.3837890625,
"learning_rate": 6.956066945606694e-07,
"loss": 0.0153,
"reward": 0.48299558460712433,
"reward_std": 0.15616995468735695,
"reward_std_all": 0.22530879825353622,
"rewards-max/forensics_tempsamp_iou_reward": 0.8562081158161163,
"rewards/forensics_tempsamp_iou_reward": 0.40489886701107025,
"step": 292
},
{
"completion_length": 20.46875,
"epoch": 1.2259414225941423,
"grad_norm": 3.2294278144836426,
"kl": 0.2333984375,
"learning_rate": 6.945606694560669e-07,
"loss": 0.0093,
"reward": 0.4300323575735092,
"reward_std": 0.1455903798341751,
"reward_std_all": 0.24073757976293564,
"rewards-max/forensics_tempsamp_iou_reward": 0.9475255310535431,
"rewards/forensics_tempsamp_iou_reward": 0.37979552149772644,
"step": 293
},
{
"completion_length": 20.75,
"epoch": 1.2301255230125523,
"grad_norm": 2.909597873687744,
"kl": 0.31640625,
"learning_rate": 6.935146443514644e-07,
"loss": 0.0126,
"reward": 0.5279944688081741,
"reward_std": 0.15544330328702927,
"reward_std_all": 0.1972196400165558,
"rewards-max/forensics_tempsamp_iou_reward": 0.9067757427692413,
"rewards/forensics_tempsamp_iou_reward": 0.4618963450193405,
"step": 294
},
{
"completion_length": 22.015625,
"epoch": 1.2343096234309623,
"grad_norm": 2.742448568344116,
"kl": 0.26171875,
"learning_rate": 6.924686192468618e-07,
"loss": 0.0104,
"reward": 0.5155168026685715,
"reward_std": 0.09776774048805237,
"reward_std_all": 0.18247465044260025,
"rewards-max/forensics_tempsamp_iou_reward": 0.9375448822975159,
"rewards/forensics_tempsamp_iou_reward": 0.4561128318309784,
"step": 295
},
{
"completion_length": 20.734375,
"epoch": 1.2384937238493725,
"grad_norm": 3.130018949508667,
"kl": 0.3154296875,
"learning_rate": 6.914225941422594e-07,
"loss": 0.0126,
"reward": 0.5508215129375458,
"reward_std": 0.15014388412237167,
"reward_std_all": 0.18593864142894745,
"rewards-max/forensics_tempsamp_iou_reward": 0.9429559707641602,
"rewards/forensics_tempsamp_iou_reward": 0.4929106682538986,
"step": 296
},
{
"completion_length": 21.890625,
"epoch": 1.2426778242677825,
"grad_norm": 3.4748306274414062,
"kl": 0.24951171875,
"learning_rate": 6.903765690376569e-07,
"loss": 0.01,
"reward": 0.4116181284189224,
"reward_std": 0.1305130124092102,
"reward_std_all": 0.24434350430965424,
"rewards-max/forensics_tempsamp_iou_reward": 0.8631412088871002,
"rewards/forensics_tempsamp_iou_reward": 0.3464584946632385,
"step": 297
},
{
"completion_length": 20.265625,
"epoch": 1.2468619246861925,
"grad_norm": 3.09808611869812,
"kl": 0.2587890625,
"learning_rate": 6.893305439330544e-07,
"loss": 0.0104,
"reward": 0.5428947508335114,
"reward_std": 0.13900281116366386,
"reward_std_all": 0.19075483828783035,
"rewards-max/forensics_tempsamp_iou_reward": 0.9551098644733429,
"rewards/forensics_tempsamp_iou_reward": 0.5080538094043732,
"step": 298
},
{
"completion_length": 22.5625,
"epoch": 1.2510460251046025,
"grad_norm": 3.1037702560424805,
"kl": 0.27685546875,
"learning_rate": 6.882845188284518e-07,
"loss": 0.0111,
"reward": 0.5402005016803741,
"reward_std": 0.1421683169901371,
"reward_std_all": 0.1821429580450058,
"rewards-max/forensics_tempsamp_iou_reward": 0.9429636597633362,
"rewards/forensics_tempsamp_iou_reward": 0.4698293209075928,
"step": 299
},
{
"completion_length": 22.1875,
"epoch": 1.2552301255230125,
"grad_norm": 3.6114649772644043,
"kl": 0.3193359375,
"learning_rate": 6.872384937238493e-07,
"loss": 0.0128,
"reward": 0.46428269147872925,
"reward_std": 0.16988356411457062,
"reward_std_all": 0.2248026356101036,
"rewards-max/forensics_tempsamp_iou_reward": 0.8381713926792145,
"rewards/forensics_tempsamp_iou_reward": 0.373076930642128,
"step": 300
},
{
"completion_length": 22.453125,
"epoch": 1.2594142259414225,
"grad_norm": 2.650183916091919,
"kl": 0.23095703125,
"learning_rate": 6.861924686192468e-07,
"loss": 0.0093,
"reward": 0.5035516619682312,
"reward_std": 0.118218794465065,
"reward_std_all": 0.19156968593597412,
"rewards-max/forensics_tempsamp_iou_reward": 0.9772674143314362,
"rewards/forensics_tempsamp_iou_reward": 0.4334537833929062,
"step": 301
},
{
"completion_length": 21.609375,
"epoch": 1.2635983263598327,
"grad_norm": 3.0134894847869873,
"kl": 0.32763671875,
"learning_rate": 6.851464435146443e-07,
"loss": 0.0131,
"reward": 0.5361500978469849,
"reward_std": 0.15954600274562836,
"reward_std_all": 0.1976267769932747,
"rewards-max/forensics_tempsamp_iou_reward": 0.9787305295467377,
"rewards/forensics_tempsamp_iou_reward": 0.47835153341293335,
"step": 302
},
{
"completion_length": 23.890625,
"epoch": 1.2677824267782427,
"grad_norm": 3.1754026412963867,
"kl": 0.234375,
"learning_rate": 6.841004184100418e-07,
"loss": 0.0094,
"reward": 0.48732106387615204,
"reward_std": 0.17709441483020782,
"reward_std_all": 0.22626961022615433,
"rewards-max/forensics_tempsamp_iou_reward": 0.8968303203582764,
"rewards/forensics_tempsamp_iou_reward": 0.42034168541431427,
"step": 303
},
{
"completion_length": 21.265625,
"epoch": 1.2719665271966527,
"grad_norm": 3.391558885574341,
"kl": 0.2880859375,
"learning_rate": 6.830543933054394e-07,
"loss": 0.0116,
"reward": 0.6146261990070343,
"reward_std": 0.1080276370048523,
"reward_std_all": 0.13155054301023483,
"rewards-max/forensics_tempsamp_iou_reward": 0.9086852371692657,
"rewards/forensics_tempsamp_iou_reward": 0.5885582268238068,
"step": 304
},
{
"completion_length": 20.890625,
"epoch": 1.2761506276150627,
"grad_norm": 2.5862009525299072,
"kl": 0.20703125,
"learning_rate": 6.820083682008368e-07,
"loss": 0.0083,
"reward": 0.5176709443330765,
"reward_std": 0.13535484671592712,
"reward_std_all": 0.19168037921190262,
"rewards-max/forensics_tempsamp_iou_reward": 0.945205956697464,
"rewards/forensics_tempsamp_iou_reward": 0.4570407271385193,
"step": 305
},
{
"completion_length": 21.078125,
"epoch": 1.280334728033473,
"grad_norm": 3.119452476501465,
"kl": 0.30859375,
"learning_rate": 6.809623430962343e-07,
"loss": 0.0123,
"reward": 0.5251602232456207,
"reward_std": 0.10563767328858376,
"reward_std_all": 0.17099104821681976,
"rewards-max/forensics_tempsamp_iou_reward": 0.988253653049469,
"rewards/forensics_tempsamp_iou_reward": 0.48570452630519867,
"step": 306
},
{
"completion_length": 25.59375,
"epoch": 1.284518828451883,
"grad_norm": 5.29395055770874,
"kl": 0.26025390625,
"learning_rate": 6.799163179916317e-07,
"loss": 0.0104,
"reward": 0.4762590825557709,
"reward_std": 0.14060115814208984,
"reward_std_all": 0.2083079293370247,
"rewards-max/forensics_tempsamp_iou_reward": 0.9342305362224579,
"rewards/forensics_tempsamp_iou_reward": 0.4001456946134567,
"step": 307
},
{
"completion_length": 22.984375,
"epoch": 1.288702928870293,
"grad_norm": 3.499300479888916,
"kl": 0.2890625,
"learning_rate": 6.788702928870292e-07,
"loss": 0.0115,
"reward": 0.5399060845375061,
"reward_std": 0.1343313716351986,
"reward_std_all": 0.1829609051346779,
"rewards-max/forensics_tempsamp_iou_reward": 0.950259655714035,
"rewards/forensics_tempsamp_iou_reward": 0.47720131278038025,
"step": 308
},
{
"completion_length": 22.921875,
"epoch": 1.292887029288703,
"grad_norm": 3.6338305473327637,
"kl": 0.2998046875,
"learning_rate": 6.778242677824267e-07,
"loss": 0.012,
"reward": 0.5521591305732727,
"reward_std": 0.1446193903684616,
"reward_std_all": 0.18509434908628464,
"rewards-max/forensics_tempsamp_iou_reward": 0.9218709170818329,
"rewards/forensics_tempsamp_iou_reward": 0.49739110469818115,
"step": 309
},
{
"completion_length": 21.703125,
"epoch": 1.297071129707113,
"grad_norm": 3.3475983142852783,
"kl": 0.3154296875,
"learning_rate": 6.767782426778242e-07,
"loss": 0.0126,
"reward": 0.5879004597663879,
"reward_std": 0.09220810234546661,
"reward_std_all": 0.14379049092531204,
"rewards-max/forensics_tempsamp_iou_reward": 0.9442806839942932,
"rewards/forensics_tempsamp_iou_reward": 0.5551051199436188,
"step": 310
},
{
"completion_length": 20.921875,
"epoch": 1.301255230125523,
"grad_norm": 3.5438578128814697,
"kl": 0.310546875,
"learning_rate": 6.757322175732217e-07,
"loss": 0.0124,
"reward": 0.5613609850406647,
"reward_std": 0.15211957693099976,
"reward_std_all": 0.18297743052244186,
"rewards-max/forensics_tempsamp_iou_reward": 0.9610278904438019,
"rewards/forensics_tempsamp_iou_reward": 0.5173511356115341,
"step": 311
},
{
"completion_length": 25.53125,
"epoch": 1.3054393305439331,
"grad_norm": 3.077789068222046,
"kl": 0.21533203125,
"learning_rate": 6.746861924686192e-07,
"loss": 0.0086,
"reward": 0.4502028077840805,
"reward_std": 0.14705242216587067,
"reward_std_all": 0.22507282346487045,
"rewards-max/forensics_tempsamp_iou_reward": 0.8812744617462158,
"rewards/forensics_tempsamp_iou_reward": 0.3710065484046936,
"step": 312
},
{
"completion_length": 20.609375,
"epoch": 1.3096234309623431,
"grad_norm": 3.184544324874878,
"kl": 0.25439453125,
"learning_rate": 6.736401673640168e-07,
"loss": 0.0102,
"reward": 0.5610341429710388,
"reward_std": 0.1378471851348877,
"reward_std_all": 0.17896975576877594,
"rewards-max/forensics_tempsamp_iou_reward": 0.9615353345870972,
"rewards/forensics_tempsamp_iou_reward": 0.5183050036430359,
"step": 313
},
{
"completion_length": 20.515625,
"epoch": 1.3138075313807531,
"grad_norm": 3.7001430988311768,
"kl": 0.2802734375,
"learning_rate": 6.725941422594143e-07,
"loss": 0.0112,
"reward": 0.6408420503139496,
"reward_std": 0.11505874991416931,
"reward_std_all": 0.13059363886713982,
"rewards-max/forensics_tempsamp_iou_reward": 0.9749207496643066,
"rewards/forensics_tempsamp_iou_reward": 0.6107769310474396,
"step": 314
},
{
"completion_length": 20.890625,
"epoch": 1.3179916317991631,
"grad_norm": 3.603107452392578,
"kl": 0.26806640625,
"learning_rate": 6.715481171548117e-07,
"loss": 0.0107,
"reward": 0.593597799539566,
"reward_std": 0.1315619945526123,
"reward_std_all": 0.16681238263845444,
"rewards-max/forensics_tempsamp_iou_reward": 0.915211021900177,
"rewards/forensics_tempsamp_iou_reward": 0.555583581328392,
"step": 315
},
{
"completion_length": 21.484375,
"epoch": 1.3221757322175733,
"grad_norm": 2.862846851348877,
"kl": 0.279296875,
"learning_rate": 6.705020920502091e-07,
"loss": 0.0112,
"reward": 0.5540410578250885,
"reward_std": 0.10073474049568176,
"reward_std_all": 0.15980281680822372,
"rewards-max/forensics_tempsamp_iou_reward": 0.9107109904289246,
"rewards/forensics_tempsamp_iou_reward": 0.5020983815193176,
"step": 316
},
{
"completion_length": 21.25,
"epoch": 1.3263598326359833,
"grad_norm": 2.9675302505493164,
"kl": 0.3291015625,
"learning_rate": 6.694560669456066e-07,
"loss": 0.0132,
"reward": 0.5798704326152802,
"reward_std": 0.07980705797672272,
"reward_std_all": 0.1361624002456665,
"rewards-max/forensics_tempsamp_iou_reward": 0.9520018994808197,
"rewards/forensics_tempsamp_iou_reward": 0.5284975469112396,
"step": 317
},
{
"completion_length": 20.78125,
"epoch": 1.3305439330543933,
"grad_norm": 3.180077314376831,
"kl": 0.2607421875,
"learning_rate": 6.684100418410041e-07,
"loss": 0.0104,
"reward": 0.536300778388977,
"reward_std": 0.14843681082129478,
"reward_std_all": 0.19522221386432648,
"rewards-max/forensics_tempsamp_iou_reward": 0.9527053236961365,
"rewards/forensics_tempsamp_iou_reward": 0.4755610227584839,
"step": 318
},
{
"completion_length": 23.0625,
"epoch": 1.3347280334728033,
"grad_norm": 3.382484197616577,
"kl": 0.3173828125,
"learning_rate": 6.673640167364016e-07,
"loss": 0.0127,
"reward": 0.4867607355117798,
"reward_std": 0.12162896990776062,
"reward_std_all": 0.2012038603425026,
"rewards-max/forensics_tempsamp_iou_reward": 0.9470567405223846,
"rewards/forensics_tempsamp_iou_reward": 0.42015112936496735,
"step": 319
},
{
"completion_length": 21.578125,
"epoch": 1.3389121338912133,
"grad_norm": 3.611802816390991,
"kl": 0.2744140625,
"learning_rate": 6.663179916317991e-07,
"loss": 0.011,
"reward": 0.5253792405128479,
"reward_std": 0.11473264917731285,
"reward_std_all": 0.18041078001260757,
"rewards-max/forensics_tempsamp_iou_reward": 0.9639557003974915,
"rewards/forensics_tempsamp_iou_reward": 0.4606862962245941,
"step": 320
},
{
"completion_length": 21.0625,
"epoch": 1.3430962343096233,
"grad_norm": 3.4029507637023926,
"kl": 0.294921875,
"learning_rate": 6.652719665271967e-07,
"loss": 0.0118,
"reward": 0.586474597454071,
"reward_std": 0.10689892247319221,
"reward_std_all": 0.1457955166697502,
"rewards-max/forensics_tempsamp_iou_reward": 0.9670183062553406,
"rewards/forensics_tempsamp_iou_reward": 0.5485915988683701,
"step": 321
},
{
"completion_length": 20.265625,
"epoch": 1.3472803347280335,
"grad_norm": 3.7310972213745117,
"kl": 0.275390625,
"learning_rate": 6.642259414225942e-07,
"loss": 0.011,
"reward": 0.5661682784557343,
"reward_std": 0.14247343316674232,
"reward_std_all": 0.17878852039575577,
"rewards-max/forensics_tempsamp_iou_reward": 0.9735225141048431,
"rewards/forensics_tempsamp_iou_reward": 0.5128143876791,
"step": 322
},
{
"completion_length": 20.671875,
"epoch": 1.3514644351464435,
"grad_norm": 4.179111003875732,
"kl": 0.310546875,
"learning_rate": 6.631799163179917e-07,
"loss": 0.0124,
"reward": 0.5244480222463608,
"reward_std": 0.15478388220071793,
"reward_std_all": 0.20325811207294464,
"rewards-max/forensics_tempsamp_iou_reward": 0.9203829169273376,
"rewards/forensics_tempsamp_iou_reward": 0.47193287312984467,
"step": 323
},
{
"completion_length": 20.453125,
"epoch": 1.3556485355648535,
"grad_norm": 3.9537546634674072,
"kl": 0.2783203125,
"learning_rate": 6.621338912133892e-07,
"loss": 0.0111,
"reward": 0.6158213913440704,
"reward_std": 0.12697765976190567,
"reward_std_all": 0.1438097283244133,
"rewards-max/forensics_tempsamp_iou_reward": 0.9432215392589569,
"rewards/forensics_tempsamp_iou_reward": 0.5775099396705627,
"step": 324
},
{
"completion_length": 19.90625,
"epoch": 1.3598326359832635,
"grad_norm": 3.414336681365967,
"kl": 0.4365234375,
"learning_rate": 6.610878661087865e-07,
"loss": 0.0175,
"reward": 0.5226276516914368,
"reward_std": 0.12506340444087982,
"reward_std_all": 0.18961552530527115,
"rewards-max/forensics_tempsamp_iou_reward": 0.900643914937973,
"rewards/forensics_tempsamp_iou_reward": 0.4551932364702225,
"step": 325
},
{
"completion_length": 20.40625,
"epoch": 1.3640167364016738,
"grad_norm": 3.1939404010772705,
"kl": 0.24169921875,
"learning_rate": 6.60041841004184e-07,
"loss": 0.0097,
"reward": 0.5194676369428635,
"reward_std": 0.08024582639336586,
"reward_std_all": 0.17020665109157562,
"rewards-max/forensics_tempsamp_iou_reward": 0.954453706741333,
"rewards/forensics_tempsamp_iou_reward": 0.4680538773536682,
"step": 326
},
{
"completion_length": 20.84375,
"epoch": 1.3682008368200838,
"grad_norm": 4.032566070556641,
"kl": 0.3037109375,
"learning_rate": 6.589958158995815e-07,
"loss": 0.0122,
"reward": 0.42827875912189484,
"reward_std": 0.12545044347643852,
"reward_std_all": 0.2317737191915512,
"rewards-max/forensics_tempsamp_iou_reward": 0.880994588136673,
"rewards/forensics_tempsamp_iou_reward": 0.3565494865179062,
"step": 327
},
{
"completion_length": 21.78125,
"epoch": 1.3723849372384938,
"grad_norm": 4.159017086029053,
"kl": 0.30078125,
"learning_rate": 6.57949790794979e-07,
"loss": 0.012,
"reward": 0.4997859448194504,
"reward_std": 0.10792533680796623,
"reward_std_all": 0.19243212044239044,
"rewards-max/forensics_tempsamp_iou_reward": 0.9028030335903168,
"rewards/forensics_tempsamp_iou_reward": 0.43827466666698456,
"step": 328
},
{
"completion_length": 20.6875,
"epoch": 1.3765690376569037,
"grad_norm": 3.291788339614868,
"kl": 0.2685546875,
"learning_rate": 6.569037656903765e-07,
"loss": 0.0107,
"reward": 0.5304462313652039,
"reward_std": 0.13829558342695236,
"reward_std_all": 0.18609736859798431,
"rewards-max/forensics_tempsamp_iou_reward": 0.9135085642337799,
"rewards/forensics_tempsamp_iou_reward": 0.47071924805641174,
"step": 329
},
{
"completion_length": 20.53125,
"epoch": 1.3807531380753137,
"grad_norm": 3.1831247806549072,
"kl": 0.25634765625,
"learning_rate": 6.558577405857741e-07,
"loss": 0.0103,
"reward": 0.5685157477855682,
"reward_std": 0.09351160004734993,
"reward_std_all": 0.15180335938930511,
"rewards-max/forensics_tempsamp_iou_reward": 0.9740835428237915,
"rewards/forensics_tempsamp_iou_reward": 0.5233237445354462,
"step": 330
},
{
"completion_length": 22.0625,
"epoch": 1.3849372384937237,
"grad_norm": 2.412064790725708,
"kl": 0.2431640625,
"learning_rate": 6.548117154811716e-07,
"loss": 0.0097,
"reward": 0.5367299765348434,
"reward_std": 0.1467502862215042,
"reward_std_all": 0.19418880343437195,
"rewards-max/forensics_tempsamp_iou_reward": 0.9786568880081177,
"rewards/forensics_tempsamp_iou_reward": 0.49881957471370697,
"step": 331
},
{
"completion_length": 20.65625,
"epoch": 1.389121338912134,
"grad_norm": 2.992009401321411,
"kl": 0.2431640625,
"learning_rate": 6.537656903765691e-07,
"loss": 0.0097,
"reward": 0.4984583854675293,
"reward_std": 0.1943698599934578,
"reward_std_all": 0.2320670336484909,
"rewards-max/forensics_tempsamp_iou_reward": 0.8981521725654602,
"rewards/forensics_tempsamp_iou_reward": 0.4372841566801071,
"step": 332
},
{
"completion_length": 20.8125,
"epoch": 1.393305439330544,
"grad_norm": 2.812589168548584,
"kl": 0.21435546875,
"learning_rate": 6.527196652719665e-07,
"loss": 0.0086,
"reward": 0.5211863815784454,
"reward_std": 0.0984501764178276,
"reward_std_all": 0.17661236226558685,
"rewards-max/forensics_tempsamp_iou_reward": 0.8700154721736908,
"rewards/forensics_tempsamp_iou_reward": 0.4483562409877777,
"step": 333
},
{
"completion_length": 24.859375,
"epoch": 1.397489539748954,
"grad_norm": 2.4910166263580322,
"kl": 0.23095703125,
"learning_rate": 6.51673640167364e-07,
"loss": 0.0093,
"reward": 0.48905569314956665,
"reward_std": 0.11482776701450348,
"reward_std_all": 0.19650709629058838,
"rewards-max/forensics_tempsamp_iou_reward": 0.8994240164756775,
"rewards/forensics_tempsamp_iou_reward": 0.42242710292339325,
"step": 334
},
{
"completion_length": 23.984375,
"epoch": 1.401673640167364,
"grad_norm": 2.94777774810791,
"kl": 0.255859375,
"learning_rate": 6.506276150627614e-07,
"loss": 0.0102,
"reward": 0.45374396443367004,
"reward_std": 0.14729144424200058,
"reward_std_all": 0.22216801345348358,
"rewards-max/forensics_tempsamp_iou_reward": 0.9415111243724823,
"rewards/forensics_tempsamp_iou_reward": 0.37539181113243103,
"step": 335
},
{
"completion_length": 21.5,
"epoch": 1.4058577405857742,
"grad_norm": 3.003829002380371,
"kl": 0.25439453125,
"learning_rate": 6.495815899581589e-07,
"loss": 0.0102,
"reward": 0.47673608362674713,
"reward_std": 0.12292925268411636,
"reward_std_all": 0.2083888277411461,
"rewards-max/forensics_tempsamp_iou_reward": 0.9167440831661224,
"rewards/forensics_tempsamp_iou_reward": 0.4088769257068634,
"step": 336
},
{
"completion_length": 24.671875,
"epoch": 1.4100418410041842,
"grad_norm": 3.128627061843872,
"kl": 0.287109375,
"learning_rate": 6.485355648535564e-07,
"loss": 0.0115,
"reward": 0.5121713131666183,
"reward_std": 0.17998922616243362,
"reward_std_all": 0.21732670813798904,
"rewards-max/forensics_tempsamp_iou_reward": 0.9010821878910065,
"rewards/forensics_tempsamp_iou_reward": 0.435995489358902,
"step": 337
},
{
"completion_length": 22.125,
"epoch": 1.4142259414225942,
"grad_norm": 2.9469475746154785,
"kl": 0.28125,
"learning_rate": 6.47489539748954e-07,
"loss": 0.0113,
"reward": 0.4938007891178131,
"reward_std": 0.14521430432796478,
"reward_std_all": 0.20711195468902588,
"rewards-max/forensics_tempsamp_iou_reward": 0.9135783910751343,
"rewards/forensics_tempsamp_iou_reward": 0.4292701184749603,
"step": 338
},
{
"completion_length": 24.546875,
"epoch": 1.4184100418410042,
"grad_norm": 3.092914581298828,
"kl": 0.2119140625,
"learning_rate": 6.464435146443515e-07,
"loss": 0.0085,
"reward": 0.5353133380413055,
"reward_std": 0.10966141521930695,
"reward_std_all": 0.16906414926052094,
"rewards-max/forensics_tempsamp_iou_reward": 0.9189062416553497,
"rewards/forensics_tempsamp_iou_reward": 0.4746718257665634,
"step": 339
},
{
"completion_length": 22.609375,
"epoch": 1.4225941422594142,
"grad_norm": 3.0244030952453613,
"kl": 0.240234375,
"learning_rate": 6.45397489539749e-07,
"loss": 0.0096,
"reward": 0.5749362111091614,
"reward_std": 0.10388772562146187,
"reward_std_all": 0.1549764797091484,
"rewards-max/forensics_tempsamp_iou_reward": 0.9525250792503357,
"rewards/forensics_tempsamp_iou_reward": 0.5335897654294968,
"step": 340
},
{
"completion_length": 20.8125,
"epoch": 1.4267782426778242,
"grad_norm": 3.67155385017395,
"kl": 0.25390625,
"learning_rate": 6.443514644351464e-07,
"loss": 0.0101,
"reward": 0.5270803868770599,
"reward_std": 0.12610367685556412,
"reward_std_all": 0.18013404309749603,
"rewards-max/forensics_tempsamp_iou_reward": 0.8892752230167389,
"rewards/forensics_tempsamp_iou_reward": 0.46469251811504364,
"step": 341
},
{
"completion_length": 24.1875,
"epoch": 1.4309623430962344,
"grad_norm": 3.485036849975586,
"kl": 0.228515625,
"learning_rate": 6.433054393305439e-07,
"loss": 0.0091,
"reward": 0.5251704454421997,
"reward_std": 0.15428511798381805,
"reward_std_all": 0.19650664925575256,
"rewards-max/forensics_tempsamp_iou_reward": 0.9364333748817444,
"rewards/forensics_tempsamp_iou_reward": 0.4581688344478607,
"step": 342
},
{
"completion_length": 22.578125,
"epoch": 1.4351464435146444,
"grad_norm": 2.9246315956115723,
"kl": 0.240234375,
"learning_rate": 6.422594142259414e-07,
"loss": 0.0096,
"reward": 0.5290128588676453,
"reward_std": 0.16317344456911087,
"reward_std_all": 0.20027300715446472,
"rewards-max/forensics_tempsamp_iou_reward": 0.8654080629348755,
"rewards/forensics_tempsamp_iou_reward": 0.4621415287256241,
"step": 343
},
{
"completion_length": 28.09375,
"epoch": 1.4393305439330544,
"grad_norm": 2.9637181758880615,
"kl": 0.25390625,
"learning_rate": 6.412133891213389e-07,
"loss": 0.0102,
"reward": 0.44897834956645966,
"reward_std": 0.08345292881131172,
"reward_std_all": 0.19571767002344131,
"rewards-max/forensics_tempsamp_iou_reward": 0.891797661781311,
"rewards/forensics_tempsamp_iou_reward": 0.38138405978679657,
"step": 344
},
{
"completion_length": 23.28125,
"epoch": 1.4435146443514644,
"grad_norm": 2.927433490753174,
"kl": 0.22265625,
"learning_rate": 6.401673640167363e-07,
"loss": 0.0089,
"reward": 0.5147290080785751,
"reward_std": 0.153611958026886,
"reward_std_all": 0.19498346000909805,
"rewards-max/forensics_tempsamp_iou_reward": 0.9696350693702698,
"rewards/forensics_tempsamp_iou_reward": 0.45621126890182495,
"step": 345
},
{
"completion_length": 21.734375,
"epoch": 1.4476987447698746,
"grad_norm": 2.7401397228240967,
"kl": 0.291015625,
"learning_rate": 6.391213389121338e-07,
"loss": 0.0116,
"reward": 0.528337150812149,
"reward_std": 0.12121102958917618,
"reward_std_all": 0.18277867883443832,
"rewards-max/forensics_tempsamp_iou_reward": 0.9608431458473206,
"rewards/forensics_tempsamp_iou_reward": 0.48570357263088226,
"step": 346
},
{
"completion_length": 26.859375,
"epoch": 1.4518828451882846,
"grad_norm": 2.8602402210235596,
"kl": 0.1943359375,
"learning_rate": 6.380753138075314e-07,
"loss": 0.0078,
"reward": 0.46167486906051636,
"reward_std": 0.12390781193971634,
"reward_std_all": 0.20860695838928223,
"rewards-max/forensics_tempsamp_iou_reward": 0.7998866140842438,
"rewards/forensics_tempsamp_iou_reward": 0.36780914664268494,
"step": 347
},
{
"completion_length": 22.09375,
"epoch": 1.4560669456066946,
"grad_norm": 3.293510675430298,
"kl": 0.2529296875,
"learning_rate": 6.370292887029289e-07,
"loss": 0.0101,
"reward": 0.5725951045751572,
"reward_std": 0.12188725918531418,
"reward_std_all": 0.16459089517593384,
"rewards-max/forensics_tempsamp_iou_reward": 0.8510983288288116,
"rewards/forensics_tempsamp_iou_reward": 0.5058462470769882,
"step": 348
},
{
"completion_length": 23.640625,
"epoch": 1.4602510460251046,
"grad_norm": 3.4766147136688232,
"kl": 0.2509765625,
"learning_rate": 6.359832635983263e-07,
"loss": 0.01,
"reward": 0.5143277049064636,
"reward_std": 0.08670084923505783,
"reward_std_all": 0.1711454913020134,
"rewards-max/forensics_tempsamp_iou_reward": 0.8770862221717834,
"rewards/forensics_tempsamp_iou_reward": 0.4546923488378525,
"step": 349
},
{
"completion_length": 20.984375,
"epoch": 1.4644351464435146,
"grad_norm": 3.2458701133728027,
"kl": 0.24755859375,
"learning_rate": 6.349372384937238e-07,
"loss": 0.0099,
"reward": 0.623320683836937,
"reward_std": 0.12768685817718506,
"reward_std_all": 0.1483868658542633,
"rewards-max/forensics_tempsamp_iou_reward": 0.9258260428905487,
"rewards/forensics_tempsamp_iou_reward": 0.6005743443965912,
"step": 350
},
{
"completion_length": 20.640625,
"epoch": 1.4686192468619246,
"grad_norm": 2.77729868888855,
"kl": 0.24609375,
"learning_rate": 6.338912133891213e-07,
"loss": 0.0098,
"reward": 0.5813420414924622,
"reward_std": 0.11786995828151703,
"reward_std_all": 0.15745988488197327,
"rewards-max/forensics_tempsamp_iou_reward": 0.9448424279689789,
"rewards/forensics_tempsamp_iou_reward": 0.5398286581039429,
"step": 351
},
{
"completion_length": 21.0,
"epoch": 1.4728033472803348,
"grad_norm": 3.0658316612243652,
"kl": 0.228515625,
"learning_rate": 6.328451882845188e-07,
"loss": 0.0091,
"reward": 0.48085737228393555,
"reward_std": 0.20863647758960724,
"reward_std_all": 0.2467048019170761,
"rewards-max/forensics_tempsamp_iou_reward": 0.9378328621387482,
"rewards/forensics_tempsamp_iou_reward": 0.42760127782821655,
"step": 352
},
{
"completion_length": 22.9375,
"epoch": 1.4769874476987448,
"grad_norm": 3.1865079402923584,
"kl": 0.4033203125,
"learning_rate": 6.317991631799163e-07,
"loss": 0.0161,
"reward": 0.4672323614358902,
"reward_std": 0.1357121467590332,
"reward_std_all": 0.22270023822784424,
"rewards-max/forensics_tempsamp_iou_reward": 0.9078212976455688,
"rewards/forensics_tempsamp_iou_reward": 0.39956575632095337,
"step": 353
},
{
"completion_length": 20.921875,
"epoch": 1.4811715481171548,
"grad_norm": 3.2084298133850098,
"kl": 0.2587890625,
"learning_rate": 6.307531380753138e-07,
"loss": 0.0104,
"reward": 0.4562839865684509,
"reward_std": 0.1585327461361885,
"reward_std_all": 0.2395022287964821,
"rewards-max/forensics_tempsamp_iou_reward": 0.8762037754058838,
"rewards/forensics_tempsamp_iou_reward": 0.3885035514831543,
"step": 354
},
{
"completion_length": 23.6875,
"epoch": 1.4853556485355648,
"grad_norm": 3.1232759952545166,
"kl": 0.26318359375,
"learning_rate": 6.297071129707112e-07,
"loss": 0.0105,
"reward": 0.5853065848350525,
"reward_std": 0.10011300444602966,
"reward_std_all": 0.14566251635551453,
"rewards-max/forensics_tempsamp_iou_reward": 0.9328036606311798,
"rewards/forensics_tempsamp_iou_reward": 0.5524810552597046,
"step": 355
},
{
"completion_length": 20.578125,
"epoch": 1.489539748953975,
"grad_norm": 3.6004607677459717,
"kl": 0.24853515625,
"learning_rate": 6.286610878661088e-07,
"loss": 0.0099,
"reward": 0.526012510061264,
"reward_std": 0.1489264965057373,
"reward_std_all": 0.19912638515233994,
"rewards-max/forensics_tempsamp_iou_reward": 0.8813231289386749,
"rewards/forensics_tempsamp_iou_reward": 0.46852099895477295,
"step": 356
},
{
"completion_length": 23.96875,
"epoch": 1.493723849372385,
"grad_norm": 3.1752448081970215,
"kl": 0.25244140625,
"learning_rate": 6.276150627615063e-07,
"loss": 0.0101,
"reward": 0.5113407671451569,
"reward_std": 0.15550552308559418,
"reward_std_all": 0.2064043954014778,
"rewards-max/forensics_tempsamp_iou_reward": 0.9327881336212158,
"rewards/forensics_tempsamp_iou_reward": 0.4503702223300934,
"step": 357
},
{
"completion_length": 24.484375,
"epoch": 1.497907949790795,
"grad_norm": 2.978180170059204,
"kl": 0.22802734375,
"learning_rate": 6.265690376569037e-07,
"loss": 0.0091,
"reward": 0.5049306005239487,
"reward_std": 0.16909684985876083,
"reward_std_all": 0.22063501179218292,
"rewards-max/forensics_tempsamp_iou_reward": 0.95748171210289,
"rewards/forensics_tempsamp_iou_reward": 0.4641716480255127,
"step": 358
},
{
"completion_length": 22.15625,
"epoch": 1.502092050209205,
"grad_norm": 2.799163341522217,
"kl": 0.232421875,
"learning_rate": 6.255230125523012e-07,
"loss": 0.0093,
"reward": 0.5889742076396942,
"reward_std": 0.08084183931350708,
"reward_std_all": 0.134225782006979,
"rewards-max/forensics_tempsamp_iou_reward": 0.9875971078872681,
"rewards/forensics_tempsamp_iou_reward": 0.5571427643299103,
"step": 359
},
{
"completion_length": 19.8125,
"epoch": 1.506276150627615,
"grad_norm": 3.440596580505371,
"kl": 0.20556640625,
"learning_rate": 6.244769874476987e-07,
"loss": 0.0082,
"reward": 0.5559490621089935,
"reward_std": 0.1288694441318512,
"reward_std_all": 0.17573144286870956,
"rewards-max/forensics_tempsamp_iou_reward": 0.9034629762172699,
"rewards/forensics_tempsamp_iou_reward": 0.5036945044994354,
"step": 360
},
{
"completion_length": 23.15625,
"epoch": 1.510460251046025,
"grad_norm": 3.1645030975341797,
"kl": 0.318359375,
"learning_rate": 6.234309623430962e-07,
"loss": 0.0127,
"reward": 0.558051735162735,
"reward_std": 0.16260144114494324,
"reward_std_all": 0.1959943026304245,
"rewards-max/forensics_tempsamp_iou_reward": 0.9305509626865387,
"rewards/forensics_tempsamp_iou_reward": 0.5158680975437164,
"step": 361
},
{
"completion_length": 21.421875,
"epoch": 1.514644351464435,
"grad_norm": 3.053607225418091,
"kl": 0.26611328125,
"learning_rate": 6.223849372384937e-07,
"loss": 0.0107,
"reward": 0.5495312213897705,
"reward_std": 0.14096727967262268,
"reward_std_all": 0.1826690509915352,
"rewards-max/forensics_tempsamp_iou_reward": 0.905597448348999,
"rewards/forensics_tempsamp_iou_reward": 0.5064651817083359,
"step": 362
},
{
"completion_length": 22.6875,
"epoch": 1.5188284518828452,
"grad_norm": 2.7557969093322754,
"kl": 0.275390625,
"learning_rate": 6.213389121338912e-07,
"loss": 0.011,
"reward": 0.5712511539459229,
"reward_std": 0.17238947749137878,
"reward_std_all": 0.18776467442512512,
"rewards-max/forensics_tempsamp_iou_reward": 0.9488110542297363,
"rewards/forensics_tempsamp_iou_reward": 0.5403575301170349,
"step": 363
},
{
"completion_length": 23.25,
"epoch": 1.5230125523012552,
"grad_norm": 3.8760111331939697,
"kl": 0.283203125,
"learning_rate": 6.202928870292888e-07,
"loss": 0.0113,
"reward": 0.5436842441558838,
"reward_std": 0.16962844133377075,
"reward_std_all": 0.20606636255979538,
"rewards-max/forensics_tempsamp_iou_reward": 0.9487777054309845,
"rewards/forensics_tempsamp_iou_reward": 0.5147917717695236,
"step": 364
},
{
"completion_length": 21.875,
"epoch": 1.5271966527196654,
"grad_norm": 2.5433971881866455,
"kl": 0.27978515625,
"learning_rate": 6.192468619246862e-07,
"loss": 0.0112,
"reward": 0.5830157995223999,
"reward_std": 0.06932118348777294,
"reward_std_all": 0.1295211873948574,
"rewards-max/forensics_tempsamp_iou_reward": 0.9099680781364441,
"rewards/forensics_tempsamp_iou_reward": 0.5384480357170105,
"step": 365
},
{
"completion_length": 20.421875,
"epoch": 1.5313807531380754,
"grad_norm": 3.4305620193481445,
"kl": 0.26513671875,
"learning_rate": 6.182008368200836e-07,
"loss": 0.0106,
"reward": 0.46273747086524963,
"reward_std": 0.16242001950740814,
"reward_std_all": 0.22893159836530685,
"rewards-max/forensics_tempsamp_iou_reward": 0.854321300983429,
"rewards/forensics_tempsamp_iou_reward": 0.3902307450771332,
"step": 366
},
{
"completion_length": 22.828125,
"epoch": 1.5355648535564854,
"grad_norm": 2.952786445617676,
"kl": 0.25048828125,
"learning_rate": 6.171548117154811e-07,
"loss": 0.01,
"reward": 0.5426499545574188,
"reward_std": 0.10433517023921013,
"reward_std_all": 0.16760308295488358,
"rewards-max/forensics_tempsamp_iou_reward": 0.9118170440196991,
"rewards/forensics_tempsamp_iou_reward": 0.4877493381500244,
"step": 367
},
{
"completion_length": 21.765625,
"epoch": 1.5397489539748954,
"grad_norm": 2.811307668685913,
"kl": 0.29296875,
"learning_rate": 6.161087866108786e-07,
"loss": 0.0117,
"reward": 0.6009742915630341,
"reward_std": 0.15101142972707748,
"reward_std_all": 0.1662329062819481,
"rewards-max/forensics_tempsamp_iou_reward": 0.9200770556926727,
"rewards/forensics_tempsamp_iou_reward": 0.5724839866161346,
"step": 368
},
{
"completion_length": 20.375,
"epoch": 1.5439330543933054,
"grad_norm": 2.784069538116455,
"kl": 0.240234375,
"learning_rate": 6.150627615062761e-07,
"loss": 0.0096,
"reward": 0.5003524720668793,
"reward_std": 0.14700151234865189,
"reward_std_all": 0.20730440318584442,
"rewards-max/forensics_tempsamp_iou_reward": 0.9881933927536011,
"rewards/forensics_tempsamp_iou_reward": 0.45821841061115265,
"step": 369
},
{
"completion_length": 20.234375,
"epoch": 1.5481171548117154,
"grad_norm": 3.0649898052215576,
"kl": 0.17529296875,
"learning_rate": 6.140167364016736e-07,
"loss": 0.007,
"reward": 0.6301942765712738,
"reward_std": 0.09829603880643845,
"reward_std_all": 0.12470857799053192,
"rewards-max/forensics_tempsamp_iou_reward": 0.9325909912586212,
"rewards/forensics_tempsamp_iou_reward": 0.5880740880966187,
"step": 370
},
{
"completion_length": 23.34375,
"epoch": 1.5523012552301254,
"grad_norm": 3.4420974254608154,
"kl": 0.2177734375,
"learning_rate": 6.129707112970711e-07,
"loss": 0.0087,
"reward": 0.4185475707054138,
"reward_std": 0.12168850377202034,
"reward_std_all": 0.2305411398410797,
"rewards-max/forensics_tempsamp_iou_reward": 0.875639945268631,
"rewards/forensics_tempsamp_iou_reward": 0.3260466307401657,
"step": 371
},
{
"completion_length": 25.40625,
"epoch": 1.5564853556485354,
"grad_norm": 3.0281929969787598,
"kl": 0.26806640625,
"learning_rate": 6.119246861924686e-07,
"loss": 0.0107,
"reward": 0.4927195757627487,
"reward_std": 0.11981699243187904,
"reward_std_all": 0.19839461892843246,
"rewards-max/forensics_tempsamp_iou_reward": 0.9366100430488586,
"rewards/forensics_tempsamp_iou_reward": 0.4279094636440277,
"step": 372
},
{
"completion_length": 22.25,
"epoch": 1.5606694560669456,
"grad_norm": 3.5088553428649902,
"kl": 0.2802734375,
"learning_rate": 6.108786610878662e-07,
"loss": 0.0112,
"reward": 0.564223438501358,
"reward_std": 0.13937027007341385,
"reward_std_all": 0.1748919114470482,
"rewards-max/forensics_tempsamp_iou_reward": 0.9191728830337524,
"rewards/forensics_tempsamp_iou_reward": 0.5071009248495102,
"step": 373
},
{
"completion_length": 22.28125,
"epoch": 1.5648535564853556,
"grad_norm": 3.413996934890747,
"kl": 0.228515625,
"learning_rate": 6.098326359832637e-07,
"loss": 0.0091,
"reward": 0.596574455499649,
"reward_std": 0.09222955256700516,
"reward_std_all": 0.1349511593580246,
"rewards-max/forensics_tempsamp_iou_reward": 0.9311298131942749,
"rewards/forensics_tempsamp_iou_reward": 0.5464615821838379,
"step": 374
},
{
"completion_length": 22.0625,
"epoch": 1.5690376569037658,
"grad_norm": 4.050120830535889,
"kl": 0.3154296875,
"learning_rate": 6.08786610878661e-07,
"loss": 0.0126,
"reward": 0.4948773682117462,
"reward_std": 0.10316507518291473,
"reward_std_all": 0.19568412750959396,
"rewards-max/forensics_tempsamp_iou_reward": 0.9218202531337738,
"rewards/forensics_tempsamp_iou_reward": 0.45185090601444244,
"step": 375
},
{
"completion_length": 19.6875,
"epoch": 1.5732217573221758,
"grad_norm": 3.2636094093322754,
"kl": 0.3076171875,
"learning_rate": 6.077405857740585e-07,
"loss": 0.0123,
"reward": 0.5862186253070831,
"reward_std": 0.09244918450713158,
"reward_std_all": 0.13693944737315178,
"rewards-max/forensics_tempsamp_iou_reward": 0.9582740068435669,
"rewards/forensics_tempsamp_iou_reward": 0.5377525687217712,
"step": 376
},
{
"completion_length": 22.046875,
"epoch": 1.5774058577405858,
"grad_norm": 3.2574808597564697,
"kl": 0.2861328125,
"learning_rate": 6.06694560669456e-07,
"loss": 0.0114,
"reward": 0.4846600294113159,
"reward_std": 0.12442874908447266,
"reward_std_all": 0.20208978652954102,
"rewards-max/forensics_tempsamp_iou_reward": 0.854726105928421,
"rewards/forensics_tempsamp_iou_reward": 0.4023599922657013,
"step": 377
},
{
"completion_length": 23.359375,
"epoch": 1.5815899581589958,
"grad_norm": 2.912583589553833,
"kl": 0.3232421875,
"learning_rate": 6.056485355648535e-07,
"loss": 0.0129,
"reward": 0.5384095907211304,
"reward_std": 0.12341487407684326,
"reward_std_all": 0.18313508480787277,
"rewards-max/forensics_tempsamp_iou_reward": 0.9351534247398376,
"rewards/forensics_tempsamp_iou_reward": 0.47417062520980835,
"step": 378
},
{
"completion_length": 24.296875,
"epoch": 1.5857740585774058,
"grad_norm": 2.6893746852874756,
"kl": 0.3017578125,
"learning_rate": 6.04602510460251e-07,
"loss": 0.0121,
"reward": 0.5525435209274292,
"reward_std": 0.128608837723732,
"reward_std_all": 0.17467942833900452,
"rewards-max/forensics_tempsamp_iou_reward": 0.9034557938575745,
"rewards/forensics_tempsamp_iou_reward": 0.4853852242231369,
"step": 379
},
{
"completion_length": 20.34375,
"epoch": 1.5899581589958158,
"grad_norm": 3.355776786804199,
"kl": 0.3046875,
"learning_rate": 6.035564853556485e-07,
"loss": 0.0122,
"reward": 0.5400529503822327,
"reward_std": 0.11306949704885483,
"reward_std_all": 0.17080723494291306,
"rewards-max/forensics_tempsamp_iou_reward": 0.9294924736022949,
"rewards/forensics_tempsamp_iou_reward": 0.48490583896636963,
"step": 380
},
{
"completion_length": 22.40625,
"epoch": 1.5941422594142258,
"grad_norm": 3.7972824573516846,
"kl": 0.27734375,
"learning_rate": 6.02510460251046e-07,
"loss": 0.0111,
"reward": 0.48923438787460327,
"reward_std": 0.14301390945911407,
"reward_std_all": 0.20845472067594528,
"rewards-max/forensics_tempsamp_iou_reward": 0.9273065328598022,
"rewards/forensics_tempsamp_iou_reward": 0.4185422509908676,
"step": 381
},
{
"completion_length": 23.203125,
"epoch": 1.5983263598326358,
"grad_norm": 2.8281216621398926,
"kl": 0.25,
"learning_rate": 6.014644351464436e-07,
"loss": 0.01,
"reward": 0.5641220808029175,
"reward_std": 0.15153378248214722,
"reward_std_all": 0.18153223395347595,
"rewards-max/forensics_tempsamp_iou_reward": 0.9450465142726898,
"rewards/forensics_tempsamp_iou_reward": 0.5337520092725754,
"step": 382
},
{
"completion_length": 23.34375,
"epoch": 1.602510460251046,
"grad_norm": 3.825131416320801,
"kl": 0.3095703125,
"learning_rate": 6.00418410041841e-07,
"loss": 0.0124,
"reward": 0.5354636758565903,
"reward_std": 0.14929747954010963,
"reward_std_all": 0.18867459148168564,
"rewards-max/forensics_tempsamp_iou_reward": 0.9500509798526764,
"rewards/forensics_tempsamp_iou_reward": 0.4771190732717514,
"step": 383
},
{
"completion_length": 22.5625,
"epoch": 1.606694560669456,
"grad_norm": 2.818605661392212,
"kl": 0.2734375,
"learning_rate": 5.993723849372385e-07,
"loss": 0.0109,
"reward": 0.5781461000442505,
"reward_std": 0.12505726516246796,
"reward_std_all": 0.16013674437999725,
"rewards-max/forensics_tempsamp_iou_reward": 0.9833612740039825,
"rewards/forensics_tempsamp_iou_reward": 0.5261702537536621,
"step": 384
},
{
"completion_length": 21.984375,
"epoch": 1.6108786610878663,
"grad_norm": 2.927767515182495,
"kl": 0.22705078125,
"learning_rate": 5.983263598326359e-07,
"loss": 0.0091,
"reward": 0.49611757695674896,
"reward_std": 0.1084052249789238,
"reward_std_all": 0.19418121129274368,
"rewards-max/forensics_tempsamp_iou_reward": 0.9641205966472626,
"rewards/forensics_tempsamp_iou_reward": 0.4547435939311981,
"step": 385
},
{
"completion_length": 24.40625,
"epoch": 1.6150627615062763,
"grad_norm": 2.8825912475585938,
"kl": 0.30322265625,
"learning_rate": 5.972803347280334e-07,
"loss": 0.0121,
"reward": 0.5375231802463531,
"reward_std": 0.09768886119127274,
"reward_std_all": 0.16525086760520935,
"rewards-max/forensics_tempsamp_iou_reward": 0.9639983773231506,
"rewards/forensics_tempsamp_iou_reward": 0.47527778148651123,
"step": 386
},
{
"completion_length": 20.375,
"epoch": 1.6192468619246863,
"grad_norm": 3.301826238632202,
"kl": 0.25048828125,
"learning_rate": 5.962343096234309e-07,
"loss": 0.01,
"reward": 0.5264537632465363,
"reward_std": 0.13642678037285805,
"reward_std_all": 0.19409923255443573,
"rewards-max/forensics_tempsamp_iou_reward": 0.9222158491611481,
"rewards/forensics_tempsamp_iou_reward": 0.4730764478445053,
"step": 387
},
{
"completion_length": 23.453125,
"epoch": 1.6234309623430963,
"grad_norm": 3.354816198348999,
"kl": 0.248046875,
"learning_rate": 5.951882845188284e-07,
"loss": 0.0099,
"reward": 0.510815441608429,
"reward_std": 0.1063709482550621,
"reward_std_all": 0.18238592147827148,
"rewards-max/forensics_tempsamp_iou_reward": 0.9551570117473602,
"rewards/forensics_tempsamp_iou_reward": 0.4475674629211426,
"step": 388
},
{
"completion_length": 21.21875,
"epoch": 1.6276150627615062,
"grad_norm": 3.3492400646209717,
"kl": 0.2939453125,
"learning_rate": 5.941422594142259e-07,
"loss": 0.0118,
"reward": 0.5469623506069183,
"reward_std": 0.1488061547279358,
"reward_std_all": 0.18974897265434265,
"rewards-max/forensics_tempsamp_iou_reward": 0.9180749654769897,
"rewards/forensics_tempsamp_iou_reward": 0.4970465451478958,
"step": 389
},
{
"completion_length": 23.96875,
"epoch": 1.6317991631799162,
"grad_norm": 2.6403024196624756,
"kl": 0.26171875,
"learning_rate": 5.930962343096235e-07,
"loss": 0.0105,
"reward": 0.4734335243701935,
"reward_std": 0.11224086582660675,
"reward_std_all": 0.2110137715935707,
"rewards-max/forensics_tempsamp_iou_reward": 0.8869847655296326,
"rewards/forensics_tempsamp_iou_reward": 0.4101007729768753,
"step": 390
},
{
"completion_length": 23.390625,
"epoch": 1.6359832635983262,
"grad_norm": 2.9583475589752197,
"kl": 0.25830078125,
"learning_rate": 5.92050209205021e-07,
"loss": 0.0103,
"reward": 0.5097901672124863,
"reward_std": 0.11264177784323692,
"reward_std_all": 0.1926453858613968,
"rewards-max/forensics_tempsamp_iou_reward": 0.9174661040306091,
"rewards/forensics_tempsamp_iou_reward": 0.45810218155384064,
"step": 391
},
{
"completion_length": 21.84375,
"epoch": 1.6401673640167362,
"grad_norm": 3.8469200134277344,
"kl": 0.2705078125,
"learning_rate": 5.910041841004184e-07,
"loss": 0.0108,
"reward": 0.5841392278671265,
"reward_std": 0.11175201833248138,
"reward_std_all": 0.14915192127227783,
"rewards-max/forensics_tempsamp_iou_reward": 0.9093350768089294,
"rewards/forensics_tempsamp_iou_reward": 0.5242405831813812,
"step": 392
},
{
"completion_length": 24.59375,
"epoch": 1.6443514644351465,
"grad_norm": 4.544254779815674,
"kl": 0.2880859375,
"learning_rate": 5.899581589958159e-07,
"loss": 0.0115,
"reward": 0.5557644069194794,
"reward_std": 0.12359505891799927,
"reward_std_all": 0.16821598261594772,
"rewards-max/forensics_tempsamp_iou_reward": 0.9202987551689148,
"rewards/forensics_tempsamp_iou_reward": 0.4940621256828308,
"step": 393
},
{
"completion_length": 20.0625,
"epoch": 1.6485355648535565,
"grad_norm": 3.386193037033081,
"kl": 0.2578125,
"learning_rate": 5.889121338912134e-07,
"loss": 0.0103,
"reward": 0.5790053904056549,
"reward_std": 0.14417418837547302,
"reward_std_all": 0.17081739008426666,
"rewards-max/forensics_tempsamp_iou_reward": 0.9559610486030579,
"rewards/forensics_tempsamp_iou_reward": 0.5398641675710678,
"step": 394
},
{
"completion_length": 22.0,
"epoch": 1.6527196652719667,
"grad_norm": 2.94695782661438,
"kl": 0.2744140625,
"learning_rate": 5.878661087866108e-07,
"loss": 0.011,
"reward": 0.6380205154418945,
"reward_std": 0.13628315180540085,
"reward_std_all": 0.14265017956495285,
"rewards-max/forensics_tempsamp_iou_reward": 0.9589721262454987,
"rewards/forensics_tempsamp_iou_reward": 0.5973404347896576,
"step": 395
},
{
"completion_length": 21.5,
"epoch": 1.6569037656903767,
"grad_norm": 3.776149272918701,
"kl": 0.3330078125,
"learning_rate": 5.868200836820083e-07,
"loss": 0.0133,
"reward": 0.5083327889442444,
"reward_std": 0.11447755619883537,
"reward_std_all": 0.18514567613601685,
"rewards-max/forensics_tempsamp_iou_reward": 0.96346715092659,
"rewards/forensics_tempsamp_iou_reward": 0.45623552799224854,
"step": 396
},
{
"completion_length": 21.75,
"epoch": 1.6610878661087867,
"grad_norm": 3.3434698581695557,
"kl": 0.3486328125,
"learning_rate": 5.857740585774058e-07,
"loss": 0.0139,
"reward": 0.48931169509887695,
"reward_std": 0.09483235329389572,
"reward_std_all": 0.19083571434020996,
"rewards-max/forensics_tempsamp_iou_reward": 0.9375819265842438,
"rewards/forensics_tempsamp_iou_reward": 0.4380386620759964,
"step": 397
},
{
"completion_length": 21.390625,
"epoch": 1.6652719665271967,
"grad_norm": 3.416506052017212,
"kl": 0.251953125,
"learning_rate": 5.847280334728033e-07,
"loss": 0.0101,
"reward": 0.4901869148015976,
"reward_std": 0.17198879271745682,
"reward_std_all": 0.22545018792152405,
"rewards-max/forensics_tempsamp_iou_reward": 0.8941802382469177,
"rewards/forensics_tempsamp_iou_reward": 0.427117183804512,
"step": 398
},
{
"completion_length": 21.984375,
"epoch": 1.6694560669456067,
"grad_norm": 3.029048204421997,
"kl": 0.2578125,
"learning_rate": 5.836820083682009e-07,
"loss": 0.0103,
"reward": 0.5413128137588501,
"reward_std": 0.1337764635682106,
"reward_std_all": 0.18257711827754974,
"rewards-max/forensics_tempsamp_iou_reward": 0.9002233445644379,
"rewards/forensics_tempsamp_iou_reward": 0.47358980774879456,
"step": 399
},
{
"completion_length": 23.6875,
"epoch": 1.6736401673640167,
"grad_norm": 2.7494845390319824,
"kl": 0.322265625,
"learning_rate": 5.826359832635983e-07,
"loss": 0.0129,
"reward": 0.5964899659156799,
"reward_std": 0.12628758698701859,
"reward_std_all": 0.1504451185464859,
"rewards-max/forensics_tempsamp_iou_reward": 0.9130244255065918,
"rewards/forensics_tempsamp_iou_reward": 0.5391490310430527,
"step": 400
},
{
"completion_length": 20.28125,
"epoch": 1.6778242677824267,
"grad_norm": 3.6101090908050537,
"kl": 0.2978515625,
"learning_rate": 5.815899581589958e-07,
"loss": 0.0119,
"reward": 0.6021172106266022,
"reward_std": 0.12798482552170753,
"reward_std_all": 0.1512243077158928,
"rewards-max/forensics_tempsamp_iou_reward": 0.9038149118423462,
"rewards/forensics_tempsamp_iou_reward": 0.5534736514091492,
"step": 401
},
{
"completion_length": 23.296875,
"epoch": 1.6820083682008367,
"grad_norm": 3.803753614425659,
"kl": 0.3671875,
"learning_rate": 5.805439330543933e-07,
"loss": 0.0147,
"reward": 0.466086208820343,
"reward_std": 0.12830642610788345,
"reward_std_all": 0.2120107263326645,
"rewards-max/forensics_tempsamp_iou_reward": 0.9481686949729919,
"rewards/forensics_tempsamp_iou_reward": 0.41129299998283386,
"step": 402
},
{
"completion_length": 22.90625,
"epoch": 1.6861924686192469,
"grad_norm": 2.885190010070801,
"kl": 0.3359375,
"learning_rate": 5.794979079497908e-07,
"loss": 0.0134,
"reward": 0.5489423871040344,
"reward_std": 0.13436608761548996,
"reward_std_all": 0.18138045817613602,
"rewards-max/forensics_tempsamp_iou_reward": 0.9200722277164459,
"rewards/forensics_tempsamp_iou_reward": 0.4934311956167221,
"step": 403
},
{
"completion_length": 23.0625,
"epoch": 1.6903765690376569,
"grad_norm": 3.325322389602661,
"kl": 0.3681640625,
"learning_rate": 5.784518828451883e-07,
"loss": 0.0148,
"reward": 0.5254388451576233,
"reward_std": 0.15727196261286736,
"reward_std_all": 0.20417528599500656,
"rewards-max/forensics_tempsamp_iou_reward": 0.8972099721431732,
"rewards/forensics_tempsamp_iou_reward": 0.46347519755363464,
"step": 404
},
{
"completion_length": 21.9375,
"epoch": 1.694560669456067,
"grad_norm": 3.46321177482605,
"kl": 0.3076171875,
"learning_rate": 5.774058577405857e-07,
"loss": 0.0123,
"reward": 0.5031396895647049,
"reward_std": 0.13882622122764587,
"reward_std_all": 0.198733851313591,
"rewards-max/forensics_tempsamp_iou_reward": 0.8837114572525024,
"rewards/forensics_tempsamp_iou_reward": 0.4275573641061783,
"step": 405
},
{
"completion_length": 21.546875,
"epoch": 1.698744769874477,
"grad_norm": 3.7751352787017822,
"kl": 0.31640625,
"learning_rate": 5.763598326359832e-07,
"loss": 0.0126,
"reward": 0.554825559258461,
"reward_std": 0.08042900264263153,
"reward_std_all": 0.15189321339130402,
"rewards-max/forensics_tempsamp_iou_reward": 0.9412012696266174,
"rewards/forensics_tempsamp_iou_reward": 0.5061057507991791,
"step": 406
},
{
"completion_length": 20.5,
"epoch": 1.702928870292887,
"grad_norm": 3.231557607650757,
"kl": 0.271484375,
"learning_rate": 5.753138075313807e-07,
"loss": 0.0109,
"reward": 0.5318292379379272,
"reward_std": 0.15926489233970642,
"reward_std_all": 0.19516776502132416,
"rewards-max/forensics_tempsamp_iou_reward": 0.8954205811023712,
"rewards/forensics_tempsamp_iou_reward": 0.46559223532676697,
"step": 407
},
{
"completion_length": 23.046875,
"epoch": 1.707112970711297,
"grad_norm": 3.114687919616699,
"kl": 0.26953125,
"learning_rate": 5.742677824267782e-07,
"loss": 0.0108,
"reward": 0.6280832290649414,
"reward_std": 0.10963032394647598,
"reward_std_all": 0.1312325783073902,
"rewards-max/forensics_tempsamp_iou_reward": 0.9585627019405365,
"rewards/forensics_tempsamp_iou_reward": 0.5936050713062286,
"step": 408
},
{
"completion_length": 20.9375,
"epoch": 1.711297071129707,
"grad_norm": 4.042527675628662,
"kl": 0.31640625,
"learning_rate": 5.732217573221757e-07,
"loss": 0.0127,
"reward": 0.5288809239864349,
"reward_std": 0.16797205805778503,
"reward_std_all": 0.20527436584234238,
"rewards-max/forensics_tempsamp_iou_reward": 0.9421538710594177,
"rewards/forensics_tempsamp_iou_reward": 0.4615977108478546,
"step": 409
},
{
"completion_length": 21.78125,
"epoch": 1.715481171548117,
"grad_norm": 3.1821837425231934,
"kl": 0.32421875,
"learning_rate": 5.721757322175732e-07,
"loss": 0.013,
"reward": 0.5727097392082214,
"reward_std": 0.15024232119321823,
"reward_std_all": 0.1787293329834938,
"rewards-max/forensics_tempsamp_iou_reward": 0.9505433738231659,
"rewards/forensics_tempsamp_iou_reward": 0.5215087532997131,
"step": 410
},
{
"completion_length": 21.40625,
"epoch": 1.719665271966527,
"grad_norm": 2.942908525466919,
"kl": 0.2392578125,
"learning_rate": 5.711297071129707e-07,
"loss": 0.0096,
"reward": 0.6200471520423889,
"reward_std": 0.10134653374552727,
"reward_std_all": 0.12796632200479507,
"rewards-max/forensics_tempsamp_iou_reward": 0.9717857539653778,
"rewards/forensics_tempsamp_iou_reward": 0.5925179421901703,
"step": 411
},
{
"completion_length": 25.859375,
"epoch": 1.723849372384937,
"grad_norm": 3.4081146717071533,
"kl": 0.279296875,
"learning_rate": 5.700836820083682e-07,
"loss": 0.0112,
"reward": 0.47493238747119904,
"reward_std": 0.11251130513846874,
"reward_std_all": 0.20488633960485458,
"rewards-max/forensics_tempsamp_iou_reward": 0.9203941822052002,
"rewards/forensics_tempsamp_iou_reward": 0.41798490285873413,
"step": 412
},
{
"completion_length": 22.546875,
"epoch": 1.7280334728033473,
"grad_norm": 3.1668639183044434,
"kl": 0.265625,
"learning_rate": 5.690376569037657e-07,
"loss": 0.0106,
"reward": 0.5502183437347412,
"reward_std": 0.12322250753641129,
"reward_std_all": 0.17774056643247604,
"rewards-max/forensics_tempsamp_iou_reward": 0.9476392567157745,
"rewards/forensics_tempsamp_iou_reward": 0.49982406198978424,
"step": 413
},
{
"completion_length": 20.9375,
"epoch": 1.7322175732217573,
"grad_norm": 3.1288223266601562,
"kl": 0.2724609375,
"learning_rate": 5.679916317991632e-07,
"loss": 0.0109,
"reward": 0.5998818278312683,
"reward_std": 0.11111967265605927,
"reward_std_all": 0.14279919862747192,
"rewards-max/forensics_tempsamp_iou_reward": 0.9459256827831268,
"rewards/forensics_tempsamp_iou_reward": 0.542967677116394,
"step": 414
},
{
"completion_length": 22.453125,
"epoch": 1.7364016736401675,
"grad_norm": 4.356881141662598,
"kl": 0.390625,
"learning_rate": 5.669456066945606e-07,
"loss": 0.0156,
"reward": 0.5479208827018738,
"reward_std": 0.08354900032281876,
"reward_std_all": 0.15276940166950226,
"rewards-max/forensics_tempsamp_iou_reward": 0.9872075021266937,
"rewards/forensics_tempsamp_iou_reward": 0.5079075396060944,
"step": 415
},
{
"completion_length": 20.71875,
"epoch": 1.7405857740585775,
"grad_norm": 3.9545772075653076,
"kl": 0.265625,
"learning_rate": 5.65899581589958e-07,
"loss": 0.0106,
"reward": 0.5004356354475021,
"reward_std": 0.12149204313755035,
"reward_std_all": 0.19336996972560883,
"rewards-max/forensics_tempsamp_iou_reward": 0.9315469861030579,
"rewards/forensics_tempsamp_iou_reward": 0.44384393095970154,
"step": 416
},
{
"completion_length": 21.515625,
"epoch": 1.7447698744769875,
"grad_norm": 3.299649953842163,
"kl": 0.34765625,
"learning_rate": 5.648535564853556e-07,
"loss": 0.0139,
"reward": 0.5504034012556076,
"reward_std": 0.10780133306980133,
"reward_std_all": 0.16645995527505875,
"rewards-max/forensics_tempsamp_iou_reward": 0.96255162358284,
"rewards/forensics_tempsamp_iou_reward": 0.5029471218585968,
"step": 417
},
{
"completion_length": 23.6875,
"epoch": 1.7489539748953975,
"grad_norm": 3.00467848777771,
"kl": 0.2578125,
"learning_rate": 5.638075313807531e-07,
"loss": 0.0103,
"reward": 0.47396570444107056,
"reward_std": 0.12441272288560867,
"reward_std_all": 0.2041047215461731,
"rewards-max/forensics_tempsamp_iou_reward": 0.9530105888843536,
"rewards/forensics_tempsamp_iou_reward": 0.43055765330791473,
"step": 418
},
{
"completion_length": 19.5625,
"epoch": 1.7531380753138075,
"grad_norm": 4.586589336395264,
"kl": 0.3291015625,
"learning_rate": 5.627615062761506e-07,
"loss": 0.0132,
"reward": 0.5532050132751465,
"reward_std": 0.15434157848358154,
"reward_std_all": 0.1914910152554512,
"rewards-max/forensics_tempsamp_iou_reward": 0.949573278427124,
"rewards/forensics_tempsamp_iou_reward": 0.5302557349205017,
"step": 419
},
{
"completion_length": 20.65625,
"epoch": 1.7573221757322175,
"grad_norm": 3.032069206237793,
"kl": 0.27099609375,
"learning_rate": 5.617154811715481e-07,
"loss": 0.0109,
"reward": 0.5159889161586761,
"reward_std": 0.1204436644911766,
"reward_std_all": 0.19084187597036362,
"rewards-max/forensics_tempsamp_iou_reward": 0.9567921459674835,
"rewards/forensics_tempsamp_iou_reward": 0.4694535732269287,
"step": 420
},
{
"completion_length": 20.265625,
"epoch": 1.7615062761506275,
"grad_norm": 3.0054852962493896,
"kl": 0.2919921875,
"learning_rate": 5.606694560669456e-07,
"loss": 0.0117,
"reward": 0.5832904279232025,
"reward_std": 0.11177239194512367,
"reward_std_all": 0.1509362868964672,
"rewards-max/forensics_tempsamp_iou_reward": 0.8631621301174164,
"rewards/forensics_tempsamp_iou_reward": 0.5379594713449478,
"step": 421
},
{
"completion_length": 21.875,
"epoch": 1.7656903765690377,
"grad_norm": 3.5174832344055176,
"kl": 0.22705078125,
"learning_rate": 5.596234309623431e-07,
"loss": 0.0091,
"reward": 0.5442539751529694,
"reward_std": 0.16536672040820122,
"reward_std_all": 0.19736886769533157,
"rewards-max/forensics_tempsamp_iou_reward": 0.9676527082920074,
"rewards/forensics_tempsamp_iou_reward": 0.49452511966228485,
"step": 422
},
{
"completion_length": 21.96875,
"epoch": 1.7698744769874477,
"grad_norm": 3.8021018505096436,
"kl": 0.244140625,
"learning_rate": 5.585774058577406e-07,
"loss": 0.0098,
"reward": 0.5342560112476349,
"reward_std": 0.16498998552560806,
"reward_std_all": 0.20063729584217072,
"rewards-max/forensics_tempsamp_iou_reward": 0.9472317695617676,
"rewards/forensics_tempsamp_iou_reward": 0.4827561676502228,
"step": 423
},
{
"completion_length": 23.328125,
"epoch": 1.7740585774058577,
"grad_norm": 3.814397096633911,
"kl": 0.21484375,
"learning_rate": 5.575313807531381e-07,
"loss": 0.0086,
"reward": 0.48156821727752686,
"reward_std": 0.1491038277745247,
"reward_std_all": 0.21688498556613922,
"rewards-max/forensics_tempsamp_iou_reward": 0.9167681038379669,
"rewards/forensics_tempsamp_iou_reward": 0.41011907160282135,
"step": 424
},
{
"completion_length": 21.640625,
"epoch": 1.778242677824268,
"grad_norm": 3.1427371501922607,
"kl": 0.2685546875,
"learning_rate": 5.564853556485354e-07,
"loss": 0.0107,
"reward": 0.5231329202651978,
"reward_std": 0.13268467783927917,
"reward_std_all": 0.18697848170995712,
"rewards-max/forensics_tempsamp_iou_reward": 0.9090035855770111,
"rewards/forensics_tempsamp_iou_reward": 0.45528803765773773,
"step": 425
},
{
"completion_length": 20.875,
"epoch": 1.782426778242678,
"grad_norm": 2.7749886512756348,
"kl": 0.24560546875,
"learning_rate": 5.55439330543933e-07,
"loss": 0.0098,
"reward": 0.5939431190490723,
"reward_std": 0.07563382387161255,
"reward_std_all": 0.13147753477096558,
"rewards-max/forensics_tempsamp_iou_reward": 0.9581961929798126,
"rewards/forensics_tempsamp_iou_reward": 0.560116171836853,
"step": 426
},
{
"completion_length": 20.21875,
"epoch": 1.786610878661088,
"grad_norm": 3.2387099266052246,
"kl": 0.3056640625,
"learning_rate": 5.543933054393305e-07,
"loss": 0.0122,
"reward": 0.5932926535606384,
"reward_std": 0.11566894873976707,
"reward_std_all": 0.15323252230882645,
"rewards-max/forensics_tempsamp_iou_reward": 0.9390993118286133,
"rewards/forensics_tempsamp_iou_reward": 0.5426921099424362,
"step": 427
},
{
"completion_length": 21.9375,
"epoch": 1.790794979079498,
"grad_norm": 3.4454848766326904,
"kl": 0.310546875,
"learning_rate": 5.53347280334728e-07,
"loss": 0.0124,
"reward": 0.5442227125167847,
"reward_std": 0.10179883241653442,
"reward_std_all": 0.16180025786161423,
"rewards-max/forensics_tempsamp_iou_reward": 0.8627505004405975,
"rewards/forensics_tempsamp_iou_reward": 0.4989946484565735,
"step": 428
},
{
"completion_length": 23.984375,
"epoch": 1.794979079497908,
"grad_norm": 3.1043126583099365,
"kl": 0.2529296875,
"learning_rate": 5.523012552301255e-07,
"loss": 0.0101,
"reward": 0.5024720132350922,
"reward_std": 0.11131147667765617,
"reward_std_all": 0.19542790949344635,
"rewards-max/forensics_tempsamp_iou_reward": 0.9665012955665588,
"rewards/forensics_tempsamp_iou_reward": 0.456233412027359,
"step": 429
},
{
"completion_length": 22.140625,
"epoch": 1.799163179916318,
"grad_norm": 3.055474042892456,
"kl": 0.27392578125,
"learning_rate": 5.51255230125523e-07,
"loss": 0.0109,
"reward": 0.492337241768837,
"reward_std": 0.1281743273139,
"reward_std_all": 0.2036646082997322,
"rewards-max/forensics_tempsamp_iou_reward": 0.8997687101364136,
"rewards/forensics_tempsamp_iou_reward": 0.4165388196706772,
"step": 430
},
{
"completion_length": 22.421875,
"epoch": 1.803347280334728,
"grad_norm": 3.221771240234375,
"kl": 0.25634765625,
"learning_rate": 5.502092050209205e-07,
"loss": 0.0102,
"reward": 0.5609057247638702,
"reward_std": 0.11431120336055756,
"reward_std_all": 0.16229231655597687,
"rewards-max/forensics_tempsamp_iou_reward": 0.9410938620567322,
"rewards/forensics_tempsamp_iou_reward": 0.5150580406188965,
"step": 431
},
{
"completion_length": 22.953125,
"epoch": 1.8075313807531381,
"grad_norm": 3.5820353031158447,
"kl": 0.37255859375,
"learning_rate": 5.49163179916318e-07,
"loss": 0.0149,
"reward": 0.47058916091918945,
"reward_std": 0.15132106095552444,
"reward_std_all": 0.21447279304265976,
"rewards-max/forensics_tempsamp_iou_reward": 0.8761255443096161,
"rewards/forensics_tempsamp_iou_reward": 0.3966824859380722,
"step": 432
},
{
"completion_length": 20.34375,
"epoch": 1.8117154811715481,
"grad_norm": 2.376065492630005,
"kl": 0.2734375,
"learning_rate": 5.481171548117156e-07,
"loss": 0.0109,
"reward": 0.5881136953830719,
"reward_std": 0.14078053086996078,
"reward_std_all": 0.17036908864974976,
"rewards-max/forensics_tempsamp_iou_reward": 0.9732219874858856,
"rewards/forensics_tempsamp_iou_reward": 0.5663430094718933,
"step": 433
},
{
"completion_length": 26.875,
"epoch": 1.8158995815899581,
"grad_norm": 3.2515807151794434,
"kl": 0.3291015625,
"learning_rate": 5.47071129707113e-07,
"loss": 0.0132,
"reward": 0.3998873680830002,
"reward_std": 0.14542201906442642,
"reward_std_all": 0.24365831911563873,
"rewards-max/forensics_tempsamp_iou_reward": 0.7606239020824432,
"rewards/forensics_tempsamp_iou_reward": 0.3101573586463928,
"step": 434
},
{
"completion_length": 23.015625,
"epoch": 1.8200836820083683,
"grad_norm": 2.6752495765686035,
"kl": 0.2783203125,
"learning_rate": 5.460251046025104e-07,
"loss": 0.0111,
"reward": 0.5873515903949738,
"reward_std": 0.1258254572749138,
"reward_std_all": 0.16005094349384308,
"rewards-max/forensics_tempsamp_iou_reward": 0.918854832649231,
"rewards/forensics_tempsamp_iou_reward": 0.5368918180465698,
"step": 435
},
{
"completion_length": 22.828125,
"epoch": 1.8242677824267783,
"grad_norm": 3.321962356567383,
"kl": 0.24853515625,
"learning_rate": 5.449790794979079e-07,
"loss": 0.0099,
"reward": 0.5557071566581726,
"reward_std": 0.12795208394527435,
"reward_std_all": 0.167966790497303,
"rewards-max/forensics_tempsamp_iou_reward": 0.9491123259067535,
"rewards/forensics_tempsamp_iou_reward": 0.5051940977573395,
"step": 436
},
{
"completion_length": 20.75,
"epoch": 1.8284518828451883,
"grad_norm": 2.2569704055786133,
"kl": 0.2890625,
"learning_rate": 5.439330543933054e-07,
"loss": 0.0116,
"reward": 0.6097114086151123,
"reward_std": 0.12603801116347313,
"reward_std_all": 0.14560042694211006,
"rewards-max/forensics_tempsamp_iou_reward": 0.9402079880237579,
"rewards/forensics_tempsamp_iou_reward": 0.5699800848960876,
"step": 437
},
{
"completion_length": 20.5,
"epoch": 1.8326359832635983,
"grad_norm": 2.913205862045288,
"kl": 0.3046875,
"learning_rate": 5.428870292887029e-07,
"loss": 0.0122,
"reward": 0.5253362655639648,
"reward_std": 0.14599011838436127,
"reward_std_all": 0.19011997431516647,
"rewards-max/forensics_tempsamp_iou_reward": 0.9371514618396759,
"rewards/forensics_tempsamp_iou_reward": 0.46212902665138245,
"step": 438
},
{
"completion_length": 22.75,
"epoch": 1.8368200836820083,
"grad_norm": 2.666419506072998,
"kl": 0.25,
"learning_rate": 5.418410041841004e-07,
"loss": 0.01,
"reward": 0.60310959815979,
"reward_std": 0.10429032146930695,
"reward_std_all": 0.1392216756939888,
"rewards-max/forensics_tempsamp_iou_reward": 0.9403590857982635,
"rewards/forensics_tempsamp_iou_reward": 0.5529260039329529,
"step": 439
},
{
"completion_length": 24.03125,
"epoch": 1.8410041841004183,
"grad_norm": 3.8333585262298584,
"kl": 0.283203125,
"learning_rate": 5.407949790794979e-07,
"loss": 0.0114,
"reward": 0.5513620674610138,
"reward_std": 0.1304529644548893,
"reward_std_all": 0.1741430088877678,
"rewards-max/forensics_tempsamp_iou_reward": 0.9436557590961456,
"rewards/forensics_tempsamp_iou_reward": 0.5042534172534943,
"step": 440
},
{
"completion_length": 20.171875,
"epoch": 1.8451882845188283,
"grad_norm": 2.9518706798553467,
"kl": 0.2744140625,
"learning_rate": 5.397489539748954e-07,
"loss": 0.011,
"reward": 0.48424582183361053,
"reward_std": 0.16240251809358597,
"reward_std_all": 0.22120697796344757,
"rewards-max/forensics_tempsamp_iou_reward": 0.9440165758132935,
"rewards/forensics_tempsamp_iou_reward": 0.43020015954971313,
"step": 441
},
{
"completion_length": 25.875,
"epoch": 1.8493723849372385,
"grad_norm": 3.221921682357788,
"kl": 0.36328125,
"learning_rate": 5.387029288702928e-07,
"loss": 0.0145,
"reward": 0.43141648173332214,
"reward_std": 0.19094012677669525,
"reward_std_all": 0.25816071033477783,
"rewards-max/forensics_tempsamp_iou_reward": 0.9141843914985657,
"rewards/forensics_tempsamp_iou_reward": 0.3569919019937515,
"step": 442
},
{
"completion_length": 22.09375,
"epoch": 1.8535564853556485,
"grad_norm": 2.613770008087158,
"kl": 0.2685546875,
"learning_rate": 5.376569037656904e-07,
"loss": 0.0107,
"reward": 0.5525793731212616,
"reward_std": 0.12158074975013733,
"reward_std_all": 0.16931996494531631,
"rewards-max/forensics_tempsamp_iou_reward": 0.9561307430267334,
"rewards/forensics_tempsamp_iou_reward": 0.4914204478263855,
"step": 443
},
{
"completion_length": 22.53125,
"epoch": 1.8577405857740585,
"grad_norm": 2.8747165203094482,
"kl": 0.26318359375,
"learning_rate": 5.366108786610879e-07,
"loss": 0.0105,
"reward": 0.5028133988380432,
"reward_std": 0.14544931054115295,
"reward_std_all": 0.20456069707870483,
"rewards-max/forensics_tempsamp_iou_reward": 0.9426700174808502,
"rewards/forensics_tempsamp_iou_reward": 0.4413255900144577,
"step": 444
},
{
"completion_length": 23.109375,
"epoch": 1.8619246861924688,
"grad_norm": 2.7452476024627686,
"kl": 0.22021484375,
"learning_rate": 5.355648535564853e-07,
"loss": 0.0088,
"reward": 0.5050263702869415,
"reward_std": 0.12905443459749222,
"reward_std_all": 0.19627108424901962,
"rewards-max/forensics_tempsamp_iou_reward": 0.9400934278964996,
"rewards/forensics_tempsamp_iou_reward": 0.43221279978752136,
"step": 445
},
{
"completion_length": 19.875,
"epoch": 1.8661087866108788,
"grad_norm": 4.670612812042236,
"kl": 0.24658203125,
"learning_rate": 5.345188284518828e-07,
"loss": 0.0099,
"reward": 0.4934425950050354,
"reward_std": 0.1461126208305359,
"reward_std_all": 0.2159237191081047,
"rewards-max/forensics_tempsamp_iou_reward": 0.9654726088047028,
"rewards/forensics_tempsamp_iou_reward": 0.43040281534194946,
"step": 446
},
{
"completion_length": 20.75,
"epoch": 1.8702928870292888,
"grad_norm": 3.132246732711792,
"kl": 0.2353515625,
"learning_rate": 5.334728033472803e-07,
"loss": 0.0094,
"reward": 0.5927966833114624,
"reward_std": 0.14627771079540253,
"reward_std_all": 0.16178085654973984,
"rewards-max/forensics_tempsamp_iou_reward": 0.9736230373382568,
"rewards/forensics_tempsamp_iou_reward": 0.5663591921329498,
"step": 447
},
{
"completion_length": 22.8125,
"epoch": 1.8744769874476988,
"grad_norm": 3.8249096870422363,
"kl": 0.306640625,
"learning_rate": 5.324267782426778e-07,
"loss": 0.0123,
"reward": 0.584292933344841,
"reward_std": 0.12335709109902382,
"reward_std_all": 0.1578608974814415,
"rewards-max/forensics_tempsamp_iou_reward": 0.9871288537979126,
"rewards/forensics_tempsamp_iou_reward": 0.5450064539909363,
"step": 448
},
{
"completion_length": 21.796875,
"epoch": 1.8786610878661087,
"grad_norm": 3.0004217624664307,
"kl": 0.3154296875,
"learning_rate": 5.313807531380753e-07,
"loss": 0.0126,
"reward": 0.57233065366745,
"reward_std": 0.2047274485230446,
"reward_std_all": 0.20986869931221008,
"rewards-max/forensics_tempsamp_iou_reward": 0.9619010388851166,
"rewards/forensics_tempsamp_iou_reward": 0.5329872071743011,
"step": 449
},
{
"completion_length": 20.4375,
"epoch": 1.8828451882845187,
"grad_norm": 2.603494167327881,
"kl": 0.3125,
"learning_rate": 5.303347280334727e-07,
"loss": 0.0125,
"reward": 0.6558628380298615,
"reward_std": 0.1298680156469345,
"reward_std_all": 0.13532085716724396,
"rewards-max/forensics_tempsamp_iou_reward": 0.90593221783638,
"rewards/forensics_tempsamp_iou_reward": 0.6307059526443481,
"step": 450
},
{
"completion_length": 23.75,
"epoch": 1.8870292887029287,
"grad_norm": 3.076854705810547,
"kl": 0.447265625,
"learning_rate": 5.292887029288703e-07,
"loss": 0.0179,
"reward": 0.6129073798656464,
"reward_std": 0.12284231185913086,
"reward_std_all": 0.14373405277729034,
"rewards-max/forensics_tempsamp_iou_reward": 0.8199488520622253,
"rewards/forensics_tempsamp_iou_reward": 0.5506384819746017,
"step": 451
},
{
"completion_length": 22.6875,
"epoch": 1.891213389121339,
"grad_norm": 2.872957468032837,
"kl": 0.2451171875,
"learning_rate": 5.282426778242678e-07,
"loss": 0.0098,
"reward": 0.5826454758644104,
"reward_std": 0.1818544641137123,
"reward_std_all": 0.19081025570631027,
"rewards-max/forensics_tempsamp_iou_reward": 0.9699686169624329,
"rewards/forensics_tempsamp_iou_reward": 0.5459464490413666,
"step": 452
},
{
"completion_length": 21.625,
"epoch": 1.895397489539749,
"grad_norm": 4.242298126220703,
"kl": 0.44921875,
"learning_rate": 5.271966527196653e-07,
"loss": 0.018,
"reward": 0.5471889823675156,
"reward_std": 0.13824594765901566,
"reward_std_all": 0.18502464145421982,
"rewards-max/forensics_tempsamp_iou_reward": 0.9139624834060669,
"rewards/forensics_tempsamp_iou_reward": 0.49499279260635376,
"step": 453
},
{
"completion_length": 20.6875,
"epoch": 1.899581589958159,
"grad_norm": 3.273439645767212,
"kl": 0.22607421875,
"learning_rate": 5.261506276150628e-07,
"loss": 0.0091,
"reward": 0.5386363565921783,
"reward_std": 0.11030194163322449,
"reward_std_all": 0.16997869312763214,
"rewards-max/forensics_tempsamp_iou_reward": 0.9229929745197296,
"rewards/forensics_tempsamp_iou_reward": 0.47879981994628906,
"step": 454
},
{
"completion_length": 21.484375,
"epoch": 1.9037656903765692,
"grad_norm": 3.1429545879364014,
"kl": 0.2646484375,
"learning_rate": 5.251046025104602e-07,
"loss": 0.0106,
"reward": 0.5328640639781952,
"reward_std": 0.14986685290932655,
"reward_std_all": 0.18998189270496368,
"rewards-max/forensics_tempsamp_iou_reward": 0.8901421129703522,
"rewards/forensics_tempsamp_iou_reward": 0.4843981862068176,
"step": 455
},
{
"completion_length": 20.25,
"epoch": 1.9079497907949792,
"grad_norm": 2.8016180992126465,
"kl": 0.2724609375,
"learning_rate": 5.240585774058577e-07,
"loss": 0.0109,
"reward": 0.5673164129257202,
"reward_std": 0.14675039798021317,
"reward_std_all": 0.17588364332914352,
"rewards-max/forensics_tempsamp_iou_reward": 0.9503170251846313,
"rewards/forensics_tempsamp_iou_reward": 0.519074946641922,
"step": 456
},
{
"completion_length": 21.0,
"epoch": 1.9121338912133892,
"grad_norm": 4.167629241943359,
"kl": 0.2958984375,
"learning_rate": 5.230125523012552e-07,
"loss": 0.0118,
"reward": 0.6233984231948853,
"reward_std": 0.10975459590554237,
"reward_std_all": 0.13220065832138062,
"rewards-max/forensics_tempsamp_iou_reward": 0.9678176045417786,
"rewards/forensics_tempsamp_iou_reward": 0.5929979085922241,
"step": 457
},
{
"completion_length": 25.734375,
"epoch": 1.9163179916317992,
"grad_norm": 2.3692169189453125,
"kl": 0.3125,
"learning_rate": 5.219665271966526e-07,
"loss": 0.0125,
"reward": 0.4761795401573181,
"reward_std": 0.12024709209799767,
"reward_std_all": 0.20399682223796844,
"rewards-max/forensics_tempsamp_iou_reward": 0.9236446619033813,
"rewards/forensics_tempsamp_iou_reward": 0.4099622666835785,
"step": 458
},
{
"completion_length": 21.40625,
"epoch": 1.9205020920502092,
"grad_norm": 2.773545980453491,
"kl": 0.208984375,
"learning_rate": 5.209205020920501e-07,
"loss": 0.0083,
"reward": 0.48215606808662415,
"reward_std": 0.09199458360671997,
"reward_std_all": 0.18726100772619247,
"rewards-max/forensics_tempsamp_iou_reward": 0.9463438987731934,
"rewards/forensics_tempsamp_iou_reward": 0.41545623540878296,
"step": 459
},
{
"completion_length": 22.296875,
"epoch": 1.9246861924686192,
"grad_norm": 2.793351888656616,
"kl": 0.27587890625,
"learning_rate": 5.198744769874477e-07,
"loss": 0.011,
"reward": 0.5714404284954071,
"reward_std": 0.10337398946285248,
"reward_std_all": 0.148724764585495,
"rewards-max/forensics_tempsamp_iou_reward": 0.9184659719467163,
"rewards/forensics_tempsamp_iou_reward": 0.5074762105941772,
"step": 460
},
{
"completion_length": 22.015625,
"epoch": 1.9288702928870292,
"grad_norm": 3.3379158973693848,
"kl": 0.3984375,
"learning_rate": 5.188284518828452e-07,
"loss": 0.0159,
"reward": 0.4983537048101425,
"reward_std": 0.13100981712341309,
"reward_std_all": 0.1983979269862175,
"rewards-max/forensics_tempsamp_iou_reward": 0.9149859547615051,
"rewards/forensics_tempsamp_iou_reward": 0.439119428396225,
"step": 461
},
{
"completion_length": 21.8125,
"epoch": 1.9330543933054394,
"grad_norm": 2.8016436100006104,
"kl": 0.23779296875,
"learning_rate": 5.177824267782427e-07,
"loss": 0.0095,
"reward": 0.5352634489536285,
"reward_std": 0.11016078293323517,
"reward_std_all": 0.17343104630708694,
"rewards-max/forensics_tempsamp_iou_reward": 0.961770236492157,
"rewards/forensics_tempsamp_iou_reward": 0.47944924235343933,
"step": 462
},
{
"completion_length": 21.21875,
"epoch": 1.9372384937238494,
"grad_norm": 2.716845750808716,
"kl": 0.291015625,
"learning_rate": 5.167364016736402e-07,
"loss": 0.0117,
"reward": 0.6007946133613586,
"reward_std": 0.09640981629490852,
"reward_std_all": 0.13308996707201004,
"rewards-max/forensics_tempsamp_iou_reward": 0.9332673251628876,
"rewards/forensics_tempsamp_iou_reward": 0.5604954361915588,
"step": 463
},
{
"completion_length": 20.78125,
"epoch": 1.9414225941422594,
"grad_norm": 2.9168403148651123,
"kl": 0.244140625,
"learning_rate": 5.156903765690377e-07,
"loss": 0.0097,
"reward": 0.6338576376438141,
"reward_std": 0.10844780877232552,
"reward_std_all": 0.12880702316761017,
"rewards-max/forensics_tempsamp_iou_reward": 0.976862758398056,
"rewards/forensics_tempsamp_iou_reward": 0.5980263948440552,
"step": 464
},
{
"completion_length": 23.140625,
"epoch": 1.9456066945606696,
"grad_norm": 3.2749388217926025,
"kl": 0.23193359375,
"learning_rate": 5.146443514644351e-07,
"loss": 0.0093,
"reward": 0.45984455943107605,
"reward_std": 0.09087621793150902,
"reward_std_all": 0.20245179533958435,
"rewards-max/forensics_tempsamp_iou_reward": 0.8685316443443298,
"rewards/forensics_tempsamp_iou_reward": 0.3895958364009857,
"step": 465
},
{
"completion_length": 21.765625,
"epoch": 1.9497907949790796,
"grad_norm": 2.7607250213623047,
"kl": 0.2880859375,
"learning_rate": 5.135983263598326e-07,
"loss": 0.0115,
"reward": 0.5344310998916626,
"reward_std": 0.11682819575071335,
"reward_std_all": 0.18420138210058212,
"rewards-max/forensics_tempsamp_iou_reward": 0.9541963636875153,
"rewards/forensics_tempsamp_iou_reward": 0.4871354103088379,
"step": 466
},
{
"completion_length": 24.703125,
"epoch": 1.9539748953974896,
"grad_norm": 2.915860652923584,
"kl": 0.275390625,
"learning_rate": 5.1255230125523e-07,
"loss": 0.011,
"reward": 0.4468015283346176,
"reward_std": 0.11242418736219406,
"reward_std_all": 0.2142244428396225,
"rewards-max/forensics_tempsamp_iou_reward": 0.9491353631019592,
"rewards/forensics_tempsamp_iou_reward": 0.4059915989637375,
"step": 467
},
{
"completion_length": 24.375,
"epoch": 1.9581589958158996,
"grad_norm": 3.1319339275360107,
"kl": 0.298828125,
"learning_rate": 5.115062761506275e-07,
"loss": 0.012,
"reward": 0.4815996587276459,
"reward_std": 0.13114353269338608,
"reward_std_all": 0.2062152475118637,
"rewards-max/forensics_tempsamp_iou_reward": 0.8731112778186798,
"rewards/forensics_tempsamp_iou_reward": 0.3992263376712799,
"step": 468
},
{
"completion_length": 21.140625,
"epoch": 1.9623430962343096,
"grad_norm": 3.254326105117798,
"kl": 0.263671875,
"learning_rate": 5.104602510460251e-07,
"loss": 0.0105,
"reward": 0.5237514227628708,
"reward_std": 0.13830158859491348,
"reward_std_all": 0.19145987927913666,
"rewards-max/forensics_tempsamp_iou_reward": 0.912243127822876,
"rewards/forensics_tempsamp_iou_reward": 0.4668334424495697,
"step": 469
},
{
"completion_length": 23.59375,
"epoch": 1.9665271966527196,
"grad_norm": 3.091493606567383,
"kl": 0.296875,
"learning_rate": 5.094142259414226e-07,
"loss": 0.0119,
"reward": 0.5293968617916107,
"reward_std": 0.196629136800766,
"reward_std_all": 0.2145216092467308,
"rewards-max/forensics_tempsamp_iou_reward": 0.977480560541153,
"rewards/forensics_tempsamp_iou_reward": 0.47412094473838806,
"step": 470
},
{
"completion_length": 20.90625,
"epoch": 1.9707112970711296,
"grad_norm": 3.1458654403686523,
"kl": 0.28369140625,
"learning_rate": 5.083682008368201e-07,
"loss": 0.0114,
"reward": 0.5642709136009216,
"reward_std": 0.10369719006121159,
"reward_std_all": 0.1560986563563347,
"rewards-max/forensics_tempsamp_iou_reward": 0.9119617342948914,
"rewards/forensics_tempsamp_iou_reward": 0.5158179998397827,
"step": 471
},
{
"completion_length": 21.578125,
"epoch": 1.9748953974895398,
"grad_norm": 4.239656925201416,
"kl": 0.27734375,
"learning_rate": 5.073221757322176e-07,
"loss": 0.0111,
"reward": 0.5274409204721451,
"reward_std": 0.13930770754814148,
"reward_std_all": 0.18624810129404068,
"rewards-max/forensics_tempsamp_iou_reward": 0.9430566132068634,
"rewards/forensics_tempsamp_iou_reward": 0.47138290107250214,
"step": 472
},
{
"completion_length": 22.203125,
"epoch": 1.9790794979079498,
"grad_norm": 3.1280148029327393,
"kl": 0.2958984375,
"learning_rate": 5.062761506276151e-07,
"loss": 0.0118,
"reward": 0.5172931998968124,
"reward_std": 0.11964775994420052,
"reward_std_all": 0.18217886984348297,
"rewards-max/forensics_tempsamp_iou_reward": 0.9137568473815918,
"rewards/forensics_tempsamp_iou_reward": 0.45173387229442596,
"step": 473
},
{
"completion_length": 21.484375,
"epoch": 1.9832635983263598,
"grad_norm": 3.204549551010132,
"kl": 0.2939453125,
"learning_rate": 5.052301255230126e-07,
"loss": 0.0118,
"reward": 0.5657247304916382,
"reward_std": 0.11450528725981712,
"reward_std_all": 0.15904781222343445,
"rewards-max/forensics_tempsamp_iou_reward": 0.9595403969287872,
"rewards/forensics_tempsamp_iou_reward": 0.522627666592598,
"step": 474
},
{
"completion_length": 22.890625,
"epoch": 1.98744769874477,
"grad_norm": 8.935075759887695,
"kl": 0.5283203125,
"learning_rate": 5.041841004184099e-07,
"loss": 0.0211,
"reward": 0.5141766965389252,
"reward_std": 0.14085077494382858,
"reward_std_all": 0.19774504005908966,
"rewards-max/forensics_tempsamp_iou_reward": 0.9821454286575317,
"rewards/forensics_tempsamp_iou_reward": 0.45345550775527954,
"step": 475
},
{
"completion_length": 21.40625,
"epoch": 1.99163179916318,
"grad_norm": 3.4929451942443848,
"kl": 0.3046875,
"learning_rate": 5.031380753138074e-07,
"loss": 0.0122,
"reward": 0.5624554455280304,
"reward_std": 0.11267445608973503,
"reward_std_all": 0.16363517194986343,
"rewards-max/forensics_tempsamp_iou_reward": 0.91632479429245,
"rewards/forensics_tempsamp_iou_reward": 0.5127173662185669,
"step": 476
},
{
"completion_length": 22.265625,
"epoch": 1.99581589958159,
"grad_norm": 3.77622389793396,
"kl": 0.3212890625,
"learning_rate": 5.02092050209205e-07,
"loss": 0.0128,
"reward": 0.5260196030139923,
"reward_std": 0.12025714665651321,
"reward_std_all": 0.17813877761363983,
"rewards-max/forensics_tempsamp_iou_reward": 0.9867202937602997,
"rewards/forensics_tempsamp_iou_reward": 0.48433177173137665,
"step": 477
},
{
"completion_length": 20.57142925262451,
"epoch": 2.0,
"grad_norm": 2.6754987239837646,
"kl": 0.236328125,
"learning_rate": 5.010460251046025e-07,
"loss": 0.01,
"reward": 0.5669224709272385,
"reward_std": 0.16646206378936768,
"reward_std_all": 0.1965046152472496,
"rewards-max/forensics_tempsamp_iou_reward": 0.8889205753803253,
"rewards/forensics_tempsamp_iou_reward": 0.5265335142612457,
"step": 478
},
{
"completion_length": 22.0,
"epoch": 2.00418410041841,
"grad_norm": 3.792766571044922,
"kl": 0.2998046875,
"learning_rate": 5e-07,
"loss": 0.012,
"reward": 0.528780609369278,
"reward_std": 0.15000252425670624,
"reward_std_all": 0.19722583889961243,
"rewards-max/forensics_tempsamp_iou_reward": 0.9500171542167664,
"rewards/forensics_tempsamp_iou_reward": 0.47117628157138824,
"step": 479
},
{
"completion_length": 22.09375,
"epoch": 2.00836820083682,
"grad_norm": 3.29897403717041,
"kl": 0.3095703125,
"learning_rate": 4.989539748953975e-07,
"loss": 0.0124,
"reward": 0.4755638837814331,
"reward_std": 0.17233595997095108,
"reward_std_all": 0.22341988235712051,
"rewards-max/forensics_tempsamp_iou_reward": 0.9522050321102142,
"rewards/forensics_tempsamp_iou_reward": 0.4041341096162796,
"step": 480
},
{
"completion_length": 22.125,
"epoch": 2.01255230125523,
"grad_norm": 2.9443392753601074,
"kl": 0.28125,
"learning_rate": 4.97907949790795e-07,
"loss": 0.0112,
"reward": 0.571324348449707,
"reward_std": 0.1376635879278183,
"reward_std_all": 0.174566850066185,
"rewards-max/forensics_tempsamp_iou_reward": 0.9679237902164459,
"rewards/forensics_tempsamp_iou_reward": 0.5243099629878998,
"step": 481
},
{
"completion_length": 23.703125,
"epoch": 2.01673640167364,
"grad_norm": 3.0099031925201416,
"kl": 0.23291015625,
"learning_rate": 4.968619246861924e-07,
"loss": 0.0093,
"reward": 0.609778881072998,
"reward_std": 0.13441113010048866,
"reward_std_all": 0.15035565197467804,
"rewards-max/forensics_tempsamp_iou_reward": 0.9328465461730957,
"rewards/forensics_tempsamp_iou_reward": 0.5585941970348358,
"step": 482
},
{
"completion_length": 22.671875,
"epoch": 2.02092050209205,
"grad_norm": 2.7159383296966553,
"kl": 0.25537109375,
"learning_rate": 4.958158995815899e-07,
"loss": 0.0102,
"reward": 0.6106895804405212,
"reward_std": 0.1266840547323227,
"reward_std_all": 0.14766421914100647,
"rewards-max/forensics_tempsamp_iou_reward": 0.9399745166301727,
"rewards/forensics_tempsamp_iou_reward": 0.574670135974884,
"step": 483
},
{
"completion_length": 25.484375,
"epoch": 2.0251046025104604,
"grad_norm": 3.209160566329956,
"kl": 0.3193359375,
"learning_rate": 4.947698744769874e-07,
"loss": 0.0128,
"reward": 0.5774664282798767,
"reward_std": 0.1114666759967804,
"reward_std_all": 0.15103068947792053,
"rewards-max/forensics_tempsamp_iou_reward": 0.9001769423484802,
"rewards/forensics_tempsamp_iou_reward": 0.508263349533081,
"step": 484
},
{
"completion_length": 23.25,
"epoch": 2.0292887029288704,
"grad_norm": 3.198000192642212,
"kl": 0.2294921875,
"learning_rate": 4.93723849372385e-07,
"loss": 0.0092,
"reward": 0.5582730770111084,
"reward_std": 0.11445305496454239,
"reward_std_all": 0.16582338511943817,
"rewards-max/forensics_tempsamp_iou_reward": 0.9597097933292389,
"rewards/forensics_tempsamp_iou_reward": 0.5310408771038055,
"step": 485
},
{
"completion_length": 21.609375,
"epoch": 2.0334728033472804,
"grad_norm": 2.94328236579895,
"kl": 0.2373046875,
"learning_rate": 4.926778242677825e-07,
"loss": 0.0095,
"reward": 0.6050435602664948,
"reward_std": 0.14068204164505005,
"reward_std_all": 0.16343745589256287,
"rewards-max/forensics_tempsamp_iou_reward": 0.9756313860416412,
"rewards/forensics_tempsamp_iou_reward": 0.5862843692302704,
"step": 486
},
{
"completion_length": 21.171875,
"epoch": 2.0376569037656904,
"grad_norm": 2.6958675384521484,
"kl": 0.32421875,
"learning_rate": 4.916317991631799e-07,
"loss": 0.013,
"reward": 0.5775099694728851,
"reward_std": 0.10749755427241325,
"reward_std_all": 0.15292714536190033,
"rewards-max/forensics_tempsamp_iou_reward": 0.9381996691226959,
"rewards/forensics_tempsamp_iou_reward": 0.5350232720375061,
"step": 487
},
{
"completion_length": 21.90625,
"epoch": 2.0418410041841004,
"grad_norm": 3.1186277866363525,
"kl": 0.23095703125,
"learning_rate": 4.905857740585774e-07,
"loss": 0.0092,
"reward": 0.6758381128311157,
"reward_std": 0.10789644345641136,
"reward_std_all": 0.11297116801142693,
"rewards-max/forensics_tempsamp_iou_reward": 0.9739825129508972,
"rewards/forensics_tempsamp_iou_reward": 0.6676480770111084,
"step": 488
},
{
"completion_length": 23.125,
"epoch": 2.0460251046025104,
"grad_norm": 3.481931686401367,
"kl": 0.29296875,
"learning_rate": 4.895397489539749e-07,
"loss": 0.0117,
"reward": 0.5391671657562256,
"reward_std": 0.13834134489297867,
"reward_std_all": 0.18995241075754166,
"rewards-max/forensics_tempsamp_iou_reward": 0.9630107879638672,
"rewards/forensics_tempsamp_iou_reward": 0.49039027094841003,
"step": 489
},
{
"completion_length": 20.65625,
"epoch": 2.0502092050209204,
"grad_norm": 3.3845105171203613,
"kl": 0.3134765625,
"learning_rate": 4.884937238493724e-07,
"loss": 0.0125,
"reward": 0.5739368796348572,
"reward_std": 0.1121911071240902,
"reward_std_all": 0.1563386246562004,
"rewards-max/forensics_tempsamp_iou_reward": 0.9504444897174835,
"rewards/forensics_tempsamp_iou_reward": 0.5036379843950272,
"step": 490
},
{
"completion_length": 22.859375,
"epoch": 2.0543933054393304,
"grad_norm": 3.8901784420013428,
"kl": 0.251953125,
"learning_rate": 4.874476987447699e-07,
"loss": 0.01,
"reward": 0.48689959943294525,
"reward_std": 0.1829364225268364,
"reward_std_all": 0.23224548995494843,
"rewards-max/forensics_tempsamp_iou_reward": 0.9393362104892731,
"rewards/forensics_tempsamp_iou_reward": 0.42621734738349915,
"step": 491
},
{
"completion_length": 20.859375,
"epoch": 2.0585774058577404,
"grad_norm": 2.8517088890075684,
"kl": 0.26953125,
"learning_rate": 4.864016736401673e-07,
"loss": 0.0108,
"reward": 0.6521345376968384,
"reward_std": 0.08067351393401623,
"reward_std_all": 0.1119704321026802,
"rewards-max/forensics_tempsamp_iou_reward": 0.9885670840740204,
"rewards/forensics_tempsamp_iou_reward": 0.6550649702548981,
"step": 492
},
{
"completion_length": 22.203125,
"epoch": 2.062761506276151,
"grad_norm": 3.0588760375976562,
"kl": 0.24609375,
"learning_rate": 4.853556485355648e-07,
"loss": 0.0098,
"reward": 0.5665013790130615,
"reward_std": 0.13902686536312103,
"reward_std_all": 0.17121627926826477,
"rewards-max/forensics_tempsamp_iou_reward": 0.9404776096343994,
"rewards/forensics_tempsamp_iou_reward": 0.5235993564128876,
"step": 493
},
{
"completion_length": 19.53125,
"epoch": 2.066945606694561,
"grad_norm": 2.6643495559692383,
"kl": 0.25537109375,
"learning_rate": 4.843096234309624e-07,
"loss": 0.0102,
"reward": 0.6419840753078461,
"reward_std": 0.07620410621166229,
"reward_std_all": 0.11127152666449547,
"rewards-max/forensics_tempsamp_iou_reward": 0.9642808139324188,
"rewards/forensics_tempsamp_iou_reward": 0.6260183751583099,
"step": 494
},
{
"completion_length": 22.203125,
"epoch": 2.071129707112971,
"grad_norm": 3.1962788105010986,
"kl": 0.3056640625,
"learning_rate": 4.832635983263599e-07,
"loss": 0.0122,
"reward": 0.582617461681366,
"reward_std": 0.11593329906463623,
"reward_std_all": 0.15557840839028358,
"rewards-max/forensics_tempsamp_iou_reward": 0.9371537864208221,
"rewards/forensics_tempsamp_iou_reward": 0.542485699057579,
"step": 495
},
{
"completion_length": 21.9375,
"epoch": 2.075313807531381,
"grad_norm": 2.644209861755371,
"kl": 0.2509765625,
"learning_rate": 4.822175732217573e-07,
"loss": 0.01,
"reward": 0.560307115316391,
"reward_std": 0.13747172057628632,
"reward_std_all": 0.17550893872976303,
"rewards-max/forensics_tempsamp_iou_reward": 0.953134149312973,
"rewards/forensics_tempsamp_iou_reward": 0.5179655253887177,
"step": 496
},
{
"completion_length": 24.15625,
"epoch": 2.079497907949791,
"grad_norm": 3.2032887935638428,
"kl": 0.25048828125,
"learning_rate": 4.811715481171548e-07,
"loss": 0.01,
"reward": 0.5432860553264618,
"reward_std": 0.11237609386444092,
"reward_std_all": 0.17327538132667542,
"rewards-max/forensics_tempsamp_iou_reward": 0.974264919757843,
"rewards/forensics_tempsamp_iou_reward": 0.5117965638637543,
"step": 497
},
{
"completion_length": 22.53125,
"epoch": 2.083682008368201,
"grad_norm": 3.5050599575042725,
"kl": 0.24609375,
"learning_rate": 4.801255230125523e-07,
"loss": 0.0098,
"reward": 0.6659961938858032,
"reward_std": 0.08856941014528275,
"reward_std_all": 0.10445315390825272,
"rewards-max/forensics_tempsamp_iou_reward": 0.981409341096878,
"rewards/forensics_tempsamp_iou_reward": 0.6611562371253967,
"step": 498
},
{
"completion_length": 20.8125,
"epoch": 2.087866108786611,
"grad_norm": 3.236025094985962,
"kl": 0.283203125,
"learning_rate": 4.790794979079498e-07,
"loss": 0.0113,
"reward": 0.5856389403343201,
"reward_std": 0.10613448917865753,
"reward_std_all": 0.15263331681489944,
"rewards-max/forensics_tempsamp_iou_reward": 0.9682587683200836,
"rewards/forensics_tempsamp_iou_reward": 0.5584262013435364,
"step": 499
},
{
"completion_length": 21.296875,
"epoch": 2.092050209205021,
"grad_norm": 2.704141855239868,
"kl": 0.3447265625,
"learning_rate": 4.780334728033473e-07,
"loss": 0.0138,
"reward": 0.6242341995239258,
"reward_std": 0.09436363726854324,
"reward_std_all": 0.12756076455116272,
"rewards-max/forensics_tempsamp_iou_reward": 0.9677453935146332,
"rewards/forensics_tempsamp_iou_reward": 0.6100756824016571,
"step": 500
},
{
"completion_length": 21.59375,
"epoch": 2.096234309623431,
"grad_norm": 3.020782470703125,
"kl": 0.251953125,
"learning_rate": 4.769874476987447e-07,
"loss": 0.0101,
"reward": 0.5585637092590332,
"reward_std": 0.0874299630522728,
"reward_std_all": 0.15757396072149277,
"rewards-max/forensics_tempsamp_iou_reward": 0.9815067946910858,
"rewards/forensics_tempsamp_iou_reward": 0.5330643951892853,
"step": 501
},
{
"completion_length": 20.9375,
"epoch": 2.100418410041841,
"grad_norm": 3.884214162826538,
"kl": 0.23095703125,
"learning_rate": 4.759414225941422e-07,
"loss": 0.0093,
"reward": 0.5556460618972778,
"reward_std": 0.11866497620940208,
"reward_std_all": 0.16729051619768143,
"rewards-max/forensics_tempsamp_iou_reward": 0.9655066430568695,
"rewards/forensics_tempsamp_iou_reward": 0.5154403299093246,
"step": 502
},
{
"completion_length": 19.515625,
"epoch": 2.104602510460251,
"grad_norm": 4.139841556549072,
"kl": 0.287109375,
"learning_rate": 4.748953974895397e-07,
"loss": 0.0115,
"reward": 0.555654913187027,
"reward_std": 0.12610401213169098,
"reward_std_all": 0.17361660301685333,
"rewards-max/forensics_tempsamp_iou_reward": 0.9665434658527374,
"rewards/forensics_tempsamp_iou_reward": 0.4989585876464844,
"step": 503
},
{
"completion_length": 21.109375,
"epoch": 2.1087866108786613,
"grad_norm": 3.776219367980957,
"kl": 0.2939453125,
"learning_rate": 4.738493723849372e-07,
"loss": 0.0118,
"reward": 0.5656493306159973,
"reward_std": 0.13528533279895782,
"reward_std_all": 0.17266159504652023,
"rewards-max/forensics_tempsamp_iou_reward": 0.8980980813503265,
"rewards/forensics_tempsamp_iou_reward": 0.5194644182920456,
"step": 504
},
{
"completion_length": 22.484375,
"epoch": 2.1129707112970713,
"grad_norm": 3.0239086151123047,
"kl": 0.2529296875,
"learning_rate": 4.7280334728033473e-07,
"loss": 0.0101,
"reward": 0.5851441621780396,
"reward_std": 0.13803792372345924,
"reward_std_all": 0.1703181341290474,
"rewards-max/forensics_tempsamp_iou_reward": 0.9951456487178802,
"rewards/forensics_tempsamp_iou_reward": 0.5532107651233673,
"step": 505
},
{
"completion_length": 21.6875,
"epoch": 2.1171548117154813,
"grad_norm": 3.4693751335144043,
"kl": 0.3037109375,
"learning_rate": 4.7175732217573224e-07,
"loss": 0.0122,
"reward": 0.5268757939338684,
"reward_std": 0.09143941849470139,
"reward_std_all": 0.16690731793642044,
"rewards-max/forensics_tempsamp_iou_reward": 0.9181396961212158,
"rewards/forensics_tempsamp_iou_reward": 0.4611597955226898,
"step": 506
},
{
"completion_length": 20.640625,
"epoch": 2.1213389121338913,
"grad_norm": 4.171563625335693,
"kl": 0.2822265625,
"learning_rate": 4.7071129707112965e-07,
"loss": 0.0113,
"reward": 0.6214130520820618,
"reward_std": 0.1184162124991417,
"reward_std_all": 0.14420395344495773,
"rewards-max/forensics_tempsamp_iou_reward": 0.9561403691768646,
"rewards/forensics_tempsamp_iou_reward": 0.5935178399085999,
"step": 507
},
{
"completion_length": 25.640625,
"epoch": 2.1255230125523012,
"grad_norm": 3.092456340789795,
"kl": 0.2470703125,
"learning_rate": 4.6966527196652716e-07,
"loss": 0.0099,
"reward": 0.5098638534545898,
"reward_std": 0.099769726395607,
"reward_std_all": 0.17380887269973755,
"rewards-max/forensics_tempsamp_iou_reward": 0.9550364017486572,
"rewards/forensics_tempsamp_iou_reward": 0.4459356963634491,
"step": 508
},
{
"completion_length": 21.40625,
"epoch": 2.1297071129707112,
"grad_norm": 3.0372512340545654,
"kl": 0.3369140625,
"learning_rate": 4.686192468619247e-07,
"loss": 0.0135,
"reward": 0.5740071535110474,
"reward_std": 0.12156225740909576,
"reward_std_all": 0.16509831696748734,
"rewards-max/forensics_tempsamp_iou_reward": 0.9608973562717438,
"rewards/forensics_tempsamp_iou_reward": 0.5449798703193665,
"step": 509
},
{
"completion_length": 19.59375,
"epoch": 2.1338912133891212,
"grad_norm": 3.5194509029388428,
"kl": 0.287109375,
"learning_rate": 4.675732217573222e-07,
"loss": 0.0115,
"reward": 0.5647676587104797,
"reward_std": 0.12858236581087112,
"reward_std_all": 0.17601685225963593,
"rewards-max/forensics_tempsamp_iou_reward": 0.9666218161582947,
"rewards/forensics_tempsamp_iou_reward": 0.5379217714071274,
"step": 510
},
{
"completion_length": 20.140625,
"epoch": 2.1380753138075312,
"grad_norm": 3.499208927154541,
"kl": 0.24951171875,
"learning_rate": 4.6652719665271965e-07,
"loss": 0.01,
"reward": 0.6758279204368591,
"reward_std": 0.07582109421491623,
"reward_std_all": 0.0923115611076355,
"rewards-max/forensics_tempsamp_iou_reward": 0.9591671228408813,
"rewards/forensics_tempsamp_iou_reward": 0.6562155187129974,
"step": 511
},
{
"completion_length": 21.765625,
"epoch": 2.1422594142259412,
"grad_norm": 3.7643320560455322,
"kl": 0.2822265625,
"learning_rate": 4.654811715481171e-07,
"loss": 0.0113,
"reward": 0.5811596214771271,
"reward_std": 0.13345350325107574,
"reward_std_all": 0.1702062487602234,
"rewards-max/forensics_tempsamp_iou_reward": 0.931604653596878,
"rewards/forensics_tempsamp_iou_reward": 0.5382580608129501,
"step": 512
},
{
"completion_length": 23.28125,
"epoch": 2.1464435146443517,
"grad_norm": 3.2541987895965576,
"kl": 0.2900390625,
"learning_rate": 4.644351464435146e-07,
"loss": 0.0116,
"reward": 0.6011123359203339,
"reward_std": 0.09554769471287727,
"reward_std_all": 0.13309592381119728,
"rewards-max/forensics_tempsamp_iou_reward": 0.9680828750133514,
"rewards/forensics_tempsamp_iou_reward": 0.5529173612594604,
"step": 513
},
{
"completion_length": 25.109375,
"epoch": 2.1506276150627617,
"grad_norm": 3.5106441974639893,
"kl": 0.291015625,
"learning_rate": 4.6338912133891213e-07,
"loss": 0.0116,
"reward": 0.5765882432460785,
"reward_std": 0.0779513381421566,
"reward_std_all": 0.14294026792049408,
"rewards-max/forensics_tempsamp_iou_reward": 0.9327703714370728,
"rewards/forensics_tempsamp_iou_reward": 0.5419337749481201,
"step": 514
},
{
"completion_length": 19.15625,
"epoch": 2.1548117154811717,
"grad_norm": 3.504999876022339,
"kl": 0.318359375,
"learning_rate": 4.623430962343096e-07,
"loss": 0.0127,
"reward": 0.5865446925163269,
"reward_std": 0.06817492097616196,
"reward_std_all": 0.13275668397545815,
"rewards-max/forensics_tempsamp_iou_reward": 0.9910408556461334,
"rewards/forensics_tempsamp_iou_reward": 0.5617014169692993,
"step": 515
},
{
"completion_length": 22.609375,
"epoch": 2.1589958158995817,
"grad_norm": 3.6981213092803955,
"kl": 0.2568359375,
"learning_rate": 4.612970711297071e-07,
"loss": 0.0103,
"reward": 0.5686238706111908,
"reward_std": 0.11488286778330803,
"reward_std_all": 0.16204971820116043,
"rewards-max/forensics_tempsamp_iou_reward": 0.956998348236084,
"rewards/forensics_tempsamp_iou_reward": 0.5312823355197906,
"step": 516
},
{
"completion_length": 23.9375,
"epoch": 2.1631799163179917,
"grad_norm": 2.9771616458892822,
"kl": 0.2802734375,
"learning_rate": 4.6025104602510457e-07,
"loss": 0.0112,
"reward": 0.6537542641162872,
"reward_std": 0.12509287893772125,
"reward_std_all": 0.13201312348246574,
"rewards-max/forensics_tempsamp_iou_reward": 0.9942606687545776,
"rewards/forensics_tempsamp_iou_reward": 0.6436310708522797,
"step": 517
},
{
"completion_length": 23.71875,
"epoch": 2.1673640167364017,
"grad_norm": 3.405677318572998,
"kl": 0.244140625,
"learning_rate": 4.592050209205021e-07,
"loss": 0.0098,
"reward": 0.5914618968963623,
"reward_std": 0.09758695587515831,
"reward_std_all": 0.13968484103679657,
"rewards-max/forensics_tempsamp_iou_reward": 0.8436784744262695,
"rewards/forensics_tempsamp_iou_reward": 0.5318981409072876,
"step": 518
},
{
"completion_length": 21.046875,
"epoch": 2.1715481171548117,
"grad_norm": 3.344404935836792,
"kl": 0.265625,
"learning_rate": 4.5815899581589954e-07,
"loss": 0.0106,
"reward": 0.6047070324420929,
"reward_std": 0.08198874071240425,
"reward_std_all": 0.13079963624477386,
"rewards-max/forensics_tempsamp_iou_reward": 0.9676753282546997,
"rewards/forensics_tempsamp_iou_reward": 0.5661070942878723,
"step": 519
},
{
"completion_length": 24.4375,
"epoch": 2.1757322175732217,
"grad_norm": 3.8077242374420166,
"kl": 0.26953125,
"learning_rate": 4.5711297071129705e-07,
"loss": 0.0108,
"reward": 0.6622301936149597,
"reward_std": 0.06738485395908356,
"reward_std_all": 0.09360724687576294,
"rewards-max/forensics_tempsamp_iou_reward": 0.9570926129817963,
"rewards/forensics_tempsamp_iou_reward": 0.6445083320140839,
"step": 520
},
{
"completion_length": 27.46875,
"epoch": 2.1799163179916317,
"grad_norm": 3.175279378890991,
"kl": 0.23828125,
"learning_rate": 4.5606694560669456e-07,
"loss": 0.0095,
"reward": 0.550952672958374,
"reward_std": 0.10678141191601753,
"reward_std_all": 0.1745479330420494,
"rewards-max/forensics_tempsamp_iou_reward": 0.9357796907424927,
"rewards/forensics_tempsamp_iou_reward": 0.5054082572460175,
"step": 521
},
{
"completion_length": 24.21875,
"epoch": 2.1841004184100417,
"grad_norm": 3.401611328125,
"kl": 0.2578125,
"learning_rate": 4.55020920502092e-07,
"loss": 0.0103,
"reward": 0.5410476922988892,
"reward_std": 0.1224898248910904,
"reward_std_all": 0.1765742227435112,
"rewards-max/forensics_tempsamp_iou_reward": 0.9758546948432922,
"rewards/forensics_tempsamp_iou_reward": 0.5063414424657822,
"step": 522
},
{
"completion_length": 23.5,
"epoch": 2.1882845188284517,
"grad_norm": 3.292405605316162,
"kl": 0.267578125,
"learning_rate": 4.539748953974895e-07,
"loss": 0.0107,
"reward": 0.5575979351997375,
"reward_std": 0.11540326103568077,
"reward_std_all": 0.168237142264843,
"rewards-max/forensics_tempsamp_iou_reward": 0.9628800749778748,
"rewards/forensics_tempsamp_iou_reward": 0.5234864950180054,
"step": 523
},
{
"completion_length": 21.265625,
"epoch": 2.192468619246862,
"grad_norm": 3.698896646499634,
"kl": 0.2666015625,
"learning_rate": 4.52928870292887e-07,
"loss": 0.0106,
"reward": 0.5588782429695129,
"reward_std": 0.11483526974916458,
"reward_std_all": 0.1608031988143921,
"rewards-max/forensics_tempsamp_iou_reward": 0.9636680483818054,
"rewards/forensics_tempsamp_iou_reward": 0.4913494288921356,
"step": 524
},
{
"completion_length": 22.0625,
"epoch": 2.196652719665272,
"grad_norm": 3.9618735313415527,
"kl": 0.302734375,
"learning_rate": 4.518828451882845e-07,
"loss": 0.0121,
"reward": 0.5377824902534485,
"reward_std": 0.10877719521522522,
"reward_std_all": 0.1736266016960144,
"rewards-max/forensics_tempsamp_iou_reward": 0.9312700927257538,
"rewards/forensics_tempsamp_iou_reward": 0.4781617969274521,
"step": 525
},
{
"completion_length": 22.6875,
"epoch": 2.200836820083682,
"grad_norm": 2.9874160289764404,
"kl": 0.21923828125,
"learning_rate": 4.50836820083682e-07,
"loss": 0.0088,
"reward": 0.590328574180603,
"reward_std": 0.12378329411149025,
"reward_std_all": 0.1508851796388626,
"rewards-max/forensics_tempsamp_iou_reward": 0.9494439661502838,
"rewards/forensics_tempsamp_iou_reward": 0.5352127850055695,
"step": 526
},
{
"completion_length": 26.234375,
"epoch": 2.205020920502092,
"grad_norm": 3.493013620376587,
"kl": 0.271484375,
"learning_rate": 4.497907949790795e-07,
"loss": 0.0108,
"reward": 0.5391711592674255,
"reward_std": 0.10783146694302559,
"reward_std_all": 0.16941095888614655,
"rewards-max/forensics_tempsamp_iou_reward": 0.9261752963066101,
"rewards/forensics_tempsamp_iou_reward": 0.4756401479244232,
"step": 527
},
{
"completion_length": 21.375,
"epoch": 2.209205020920502,
"grad_norm": 4.3152289390563965,
"kl": 0.23193359375,
"learning_rate": 4.4874476987447694e-07,
"loss": 0.0093,
"reward": 0.5383678674697876,
"reward_std": 0.13210804015398026,
"reward_std_all": 0.18454814702272415,
"rewards-max/forensics_tempsamp_iou_reward": 0.9285714328289032,
"rewards/forensics_tempsamp_iou_reward": 0.4982212632894516,
"step": 528
},
{
"completion_length": 23.015625,
"epoch": 2.213389121338912,
"grad_norm": 3.8792519569396973,
"kl": 0.3046875,
"learning_rate": 4.4769874476987445e-07,
"loss": 0.0122,
"reward": 0.6066391170024872,
"reward_std": 0.11306080967187881,
"reward_std_all": 0.1467704214155674,
"rewards-max/forensics_tempsamp_iou_reward": 0.9883177876472473,
"rewards/forensics_tempsamp_iou_reward": 0.5869813561439514,
"step": 529
},
{
"completion_length": 23.0,
"epoch": 2.217573221757322,
"grad_norm": 5.31892728805542,
"kl": 0.5205078125,
"learning_rate": 4.4665271966527197e-07,
"loss": 0.0208,
"reward": 0.6351097226142883,
"reward_std": 0.07888621836900711,
"reward_std_all": 0.11654597893357277,
"rewards-max/forensics_tempsamp_iou_reward": 0.9841755926609039,
"rewards/forensics_tempsamp_iou_reward": 0.6363240778446198,
"step": 530
},
{
"completion_length": 24.25,
"epoch": 2.221757322175732,
"grad_norm": 3.9110560417175293,
"kl": 0.271484375,
"learning_rate": 4.456066945606695e-07,
"loss": 0.0108,
"reward": 0.5977832078933716,
"reward_std": 0.1437695547938347,
"reward_std_all": 0.1707897186279297,
"rewards-max/forensics_tempsamp_iou_reward": 0.9750616550445557,
"rewards/forensics_tempsamp_iou_reward": 0.5789860188961029,
"step": 531
},
{
"completion_length": 24.90625,
"epoch": 2.225941422594142,
"grad_norm": 3.7554092407226562,
"kl": 0.310546875,
"learning_rate": 4.445606694560669e-07,
"loss": 0.0124,
"reward": 0.5044494867324829,
"reward_std": 0.15146023035049438,
"reward_std_all": 0.21185988187789917,
"rewards-max/forensics_tempsamp_iou_reward": 0.9625877141952515,
"rewards/forensics_tempsamp_iou_reward": 0.45700716972351074,
"step": 532
},
{
"completion_length": 23.734375,
"epoch": 2.2301255230125525,
"grad_norm": 3.143817901611328,
"kl": 0.25732421875,
"learning_rate": 4.435146443514644e-07,
"loss": 0.0103,
"reward": 0.5997700393199921,
"reward_std": 0.1236688494682312,
"reward_std_all": 0.14943524450063705,
"rewards-max/forensics_tempsamp_iou_reward": 0.9410814642906189,
"rewards/forensics_tempsamp_iou_reward": 0.5547880530357361,
"step": 533
},
{
"completion_length": 20.140625,
"epoch": 2.2343096234309625,
"grad_norm": 3.6995086669921875,
"kl": 0.341796875,
"learning_rate": 4.424686192468619e-07,
"loss": 0.0137,
"reward": 0.5761308968067169,
"reward_std": 0.11953693628311157,
"reward_std_all": 0.16295315325260162,
"rewards-max/forensics_tempsamp_iou_reward": 0.9862031936645508,
"rewards/forensics_tempsamp_iou_reward": 0.5420914888381958,
"step": 534
},
{
"completion_length": 23.890625,
"epoch": 2.2384937238493725,
"grad_norm": 3.1931474208831787,
"kl": 0.25244140625,
"learning_rate": 4.414225941422594e-07,
"loss": 0.0101,
"reward": 0.5862003266811371,
"reward_std": 0.10501349717378616,
"reward_std_all": 0.15020647644996643,
"rewards-max/forensics_tempsamp_iou_reward": 0.95271235704422,
"rewards/forensics_tempsamp_iou_reward": 0.5443427264690399,
"step": 535
},
{
"completion_length": 21.46875,
"epoch": 2.2426778242677825,
"grad_norm": 3.723048448562622,
"kl": 0.283203125,
"learning_rate": 4.403765690376569e-07,
"loss": 0.0113,
"reward": 0.6106281578540802,
"reward_std": 0.12569432705640793,
"reward_std_all": 0.14507876336574554,
"rewards-max/forensics_tempsamp_iou_reward": 0.9429206252098083,
"rewards/forensics_tempsamp_iou_reward": 0.5706012845039368,
"step": 536
},
{
"completion_length": 23.25,
"epoch": 2.2468619246861925,
"grad_norm": 3.5991172790527344,
"kl": 0.3408203125,
"learning_rate": 4.3933054393305435e-07,
"loss": 0.0137,
"reward": 0.5168288052082062,
"reward_std": 0.14052055776119232,
"reward_std_all": 0.19934479892253876,
"rewards-max/forensics_tempsamp_iou_reward": 0.9271969795227051,
"rewards/forensics_tempsamp_iou_reward": 0.45720839500427246,
"step": 537
},
{
"completion_length": 22.6875,
"epoch": 2.2510460251046025,
"grad_norm": 3.147806406021118,
"kl": 0.24462890625,
"learning_rate": 4.3828451882845186e-07,
"loss": 0.0098,
"reward": 0.5826697051525116,
"reward_std": 0.11525051295757294,
"reward_std_all": 0.1552778035402298,
"rewards-max/forensics_tempsamp_iou_reward": 0.9473684132099152,
"rewards/forensics_tempsamp_iou_reward": 0.5314865112304688,
"step": 538
},
{
"completion_length": 21.625,
"epoch": 2.2552301255230125,
"grad_norm": 3.566258668899536,
"kl": 0.3671875,
"learning_rate": 4.3723849372384937e-07,
"loss": 0.0147,
"reward": 0.5459695160388947,
"reward_std": 0.15519853681325912,
"reward_std_all": 0.19275039434432983,
"rewards-max/forensics_tempsamp_iou_reward": 0.9428260922431946,
"rewards/forensics_tempsamp_iou_reward": 0.49625739455223083,
"step": 539
},
{
"completion_length": 23.265625,
"epoch": 2.2594142259414225,
"grad_norm": 3.825098991394043,
"kl": 0.22265625,
"learning_rate": 4.3619246861924683e-07,
"loss": 0.0089,
"reward": 0.5198812931776047,
"reward_std": 0.13256766647100449,
"reward_std_all": 0.18919435143470764,
"rewards-max/forensics_tempsamp_iou_reward": 0.9647331535816193,
"rewards/forensics_tempsamp_iou_reward": 0.4615863412618637,
"step": 540
},
{
"completion_length": 20.59375,
"epoch": 2.2635983263598325,
"grad_norm": 3.49696683883667,
"kl": 0.2294921875,
"learning_rate": 4.3514644351464434e-07,
"loss": 0.0092,
"reward": 0.6078659892082214,
"reward_std": 0.06907740421593189,
"reward_std_all": 0.1201673075556755,
"rewards-max/forensics_tempsamp_iou_reward": 0.9748539924621582,
"rewards/forensics_tempsamp_iou_reward": 0.5803891718387604,
"step": 541
},
{
"completion_length": 20.859375,
"epoch": 2.2677824267782425,
"grad_norm": 2.928506374359131,
"kl": 0.33203125,
"learning_rate": 4.341004184100418e-07,
"loss": 0.0133,
"reward": 0.7041153311729431,
"reward_std": 0.0809868648648262,
"reward_std_all": 0.0856545977294445,
"rewards-max/forensics_tempsamp_iou_reward": 0.9395063817501068,
"rewards/forensics_tempsamp_iou_reward": 0.6820828914642334,
"step": 542
},
{
"completion_length": 21.28125,
"epoch": 2.2719665271966525,
"grad_norm": 2.931365489959717,
"kl": 0.2919921875,
"learning_rate": 4.330543933054393e-07,
"loss": 0.0117,
"reward": 0.6141932308673859,
"reward_std": 0.10747497901320457,
"reward_std_all": 0.13449930772185326,
"rewards-max/forensics_tempsamp_iou_reward": 0.9909341931343079,
"rewards/forensics_tempsamp_iou_reward": 0.591324508190155,
"step": 543
},
{
"completion_length": 20.90625,
"epoch": 2.276150627615063,
"grad_norm": 3.0525314807891846,
"kl": 0.2822265625,
"learning_rate": 4.3200836820083683e-07,
"loss": 0.0113,
"reward": 0.6380703747272491,
"reward_std": 0.0922907404601574,
"reward_std_all": 0.11644239351153374,
"rewards-max/forensics_tempsamp_iou_reward": 0.9300973415374756,
"rewards/forensics_tempsamp_iou_reward": 0.624352902173996,
"step": 544
},
{
"completion_length": 19.796875,
"epoch": 2.280334728033473,
"grad_norm": 3.1044037342071533,
"kl": 0.216796875,
"learning_rate": 4.309623430962343e-07,
"loss": 0.0087,
"reward": 0.6277871131896973,
"reward_std": 0.09796342998743057,
"reward_std_all": 0.12168305367231369,
"rewards-max/forensics_tempsamp_iou_reward": 0.9768962562084198,
"rewards/forensics_tempsamp_iou_reward": 0.6006993353366852,
"step": 545
},
{
"completion_length": 21.953125,
"epoch": 2.284518828451883,
"grad_norm": 3.5345754623413086,
"kl": 0.3154296875,
"learning_rate": 4.299163179916318e-07,
"loss": 0.0126,
"reward": 0.6088225543498993,
"reward_std": 0.18452230840921402,
"reward_std_all": 0.18421754240989685,
"rewards-max/forensics_tempsamp_iou_reward": 0.9822118878364563,
"rewards/forensics_tempsamp_iou_reward": 0.5987941324710846,
"step": 546
},
{
"completion_length": 23.78125,
"epoch": 2.288702928870293,
"grad_norm": 3.161405324935913,
"kl": 0.328125,
"learning_rate": 4.2887029288702926e-07,
"loss": 0.0131,
"reward": 0.4428938776254654,
"reward_std": 0.11442334577441216,
"reward_std_all": 0.22064705193042755,
"rewards-max/forensics_tempsamp_iou_reward": 0.920212060213089,
"rewards/forensics_tempsamp_iou_reward": 0.380876287817955,
"step": 547
},
{
"completion_length": 23.046875,
"epoch": 2.292887029288703,
"grad_norm": 3.1039507389068604,
"kl": 0.2705078125,
"learning_rate": 4.278242677824268e-07,
"loss": 0.0108,
"reward": 0.5948722958564758,
"reward_std": 0.12378187477588654,
"reward_std_all": 0.1514744572341442,
"rewards-max/forensics_tempsamp_iou_reward": 0.948803186416626,
"rewards/forensics_tempsamp_iou_reward": 0.552727684378624,
"step": 548
},
{
"completion_length": 19.8125,
"epoch": 2.297071129707113,
"grad_norm": 3.5297865867614746,
"kl": 0.36474609375,
"learning_rate": 4.2677824267782424e-07,
"loss": 0.0146,
"reward": 0.6146714687347412,
"reward_std": 0.1349342241883278,
"reward_std_all": 0.14949403703212738,
"rewards-max/forensics_tempsamp_iou_reward": 0.9512049853801727,
"rewards/forensics_tempsamp_iou_reward": 0.5900232195854187,
"step": 549
},
{
"completion_length": 23.5,
"epoch": 2.301255230125523,
"grad_norm": 3.302309989929199,
"kl": 0.275390625,
"learning_rate": 4.2573221757322175e-07,
"loss": 0.011,
"reward": 0.628202497959137,
"reward_std": 0.10261930525302887,
"reward_std_all": 0.12708501145243645,
"rewards-max/forensics_tempsamp_iou_reward": 0.9407891929149628,
"rewards/forensics_tempsamp_iou_reward": 0.5886779427528381,
"step": 550
},
{
"completion_length": 21.609375,
"epoch": 2.305439330543933,
"grad_norm": 3.209219455718994,
"kl": 0.32861328125,
"learning_rate": 4.2468619246861926e-07,
"loss": 0.0131,
"reward": 0.5277585238218307,
"reward_std": 0.10281521826982498,
"reward_std_all": 0.17370499670505524,
"rewards-max/forensics_tempsamp_iou_reward": 0.967717707157135,
"rewards/forensics_tempsamp_iou_reward": 0.4712243676185608,
"step": 551
},
{
"completion_length": 21.921875,
"epoch": 2.309623430962343,
"grad_norm": 2.9841504096984863,
"kl": 0.2548828125,
"learning_rate": 4.236401673640167e-07,
"loss": 0.0102,
"reward": 0.6286369860172272,
"reward_std": 0.08519545942544937,
"reward_std_all": 0.11660626158118248,
"rewards-max/forensics_tempsamp_iou_reward": 0.8917027711868286,
"rewards/forensics_tempsamp_iou_reward": 0.5810851156711578,
"step": 552
},
{
"completion_length": 21.859375,
"epoch": 2.3138075313807533,
"grad_norm": 3.446737766265869,
"kl": 0.2978515625,
"learning_rate": 4.225941422594142e-07,
"loss": 0.0119,
"reward": 0.6222237646579742,
"reward_std": 0.11559063196182251,
"reward_std_all": 0.14099738001823425,
"rewards-max/forensics_tempsamp_iou_reward": 0.9861379265785217,
"rewards/forensics_tempsamp_iou_reward": 0.599914014339447,
"step": 553
},
{
"completion_length": 21.484375,
"epoch": 2.3179916317991633,
"grad_norm": 3.7323813438415527,
"kl": 0.2646484375,
"learning_rate": 4.215481171548117e-07,
"loss": 0.0106,
"reward": 0.5726643204689026,
"reward_std": 0.09893467091023922,
"reward_std_all": 0.14931995794177055,
"rewards-max/forensics_tempsamp_iou_reward": 0.9721894860267639,
"rewards/forensics_tempsamp_iou_reward": 0.5309462249279022,
"step": 554
},
{
"completion_length": 20.59375,
"epoch": 2.3221757322175733,
"grad_norm": 3.201817035675049,
"kl": 0.296875,
"learning_rate": 4.205020920502092e-07,
"loss": 0.0119,
"reward": 0.6240266263484955,
"reward_std": 0.09946726262569427,
"reward_std_all": 0.12944738194346428,
"rewards-max/forensics_tempsamp_iou_reward": 0.9646796584129333,
"rewards/forensics_tempsamp_iou_reward": 0.608906626701355,
"step": 555
},
{
"completion_length": 21.84375,
"epoch": 2.3263598326359833,
"grad_norm": 3.2397797107696533,
"kl": 0.3125,
"learning_rate": 4.194560669456067e-07,
"loss": 0.0125,
"reward": 0.6130309104919434,
"reward_std": 0.16433283686637878,
"reward_std_all": 0.17436052113771439,
"rewards-max/forensics_tempsamp_iou_reward": 0.9827275276184082,
"rewards/forensics_tempsamp_iou_reward": 0.5907942354679108,
"step": 556
},
{
"completion_length": 22.078125,
"epoch": 2.3305439330543933,
"grad_norm": 3.664954900741577,
"kl": 0.369140625,
"learning_rate": 4.184100418410041e-07,
"loss": 0.0148,
"reward": 0.5592598617076874,
"reward_std": 0.14954201877117157,
"reward_std_all": 0.1822916492819786,
"rewards-max/forensics_tempsamp_iou_reward": 0.9342721998691559,
"rewards/forensics_tempsamp_iou_reward": 0.5119351744651794,
"step": 557
},
{
"completion_length": 21.5625,
"epoch": 2.3347280334728033,
"grad_norm": 3.38332200050354,
"kl": 0.2666015625,
"learning_rate": 4.1736401673640164e-07,
"loss": 0.0107,
"reward": 0.5458851754665375,
"reward_std": 0.1944860890507698,
"reward_std_all": 0.2122049331665039,
"rewards-max/forensics_tempsamp_iou_reward": 0.9831523895263672,
"rewards/forensics_tempsamp_iou_reward": 0.505381166934967,
"step": 558
},
{
"completion_length": 20.71875,
"epoch": 2.3389121338912133,
"grad_norm": 3.0790648460388184,
"kl": 0.3154296875,
"learning_rate": 4.1631799163179915e-07,
"loss": 0.0126,
"reward": 0.6217442452907562,
"reward_std": 0.10764461755752563,
"reward_std_all": 0.1337558850646019,
"rewards-max/forensics_tempsamp_iou_reward": 0.9672962129116058,
"rewards/forensics_tempsamp_iou_reward": 0.59091517329216,
"step": 559
},
{
"completion_length": 22.515625,
"epoch": 2.3430962343096233,
"grad_norm": 3.506368637084961,
"kl": 0.31640625,
"learning_rate": 4.1527196652719666e-07,
"loss": 0.0126,
"reward": 0.6175608336925507,
"reward_std": 0.11754832416772842,
"reward_std_all": 0.1426597312092781,
"rewards-max/forensics_tempsamp_iou_reward": 0.9941544234752655,
"rewards/forensics_tempsamp_iou_reward": 0.587405800819397,
"step": 560
},
{
"completion_length": 21.1875,
"epoch": 2.3472803347280333,
"grad_norm": 3.928934335708618,
"kl": 0.4306640625,
"learning_rate": 4.142259414225941e-07,
"loss": 0.0173,
"reward": 0.618482768535614,
"reward_std": 0.12189881503582001,
"reward_std_all": 0.14341089129447937,
"rewards-max/forensics_tempsamp_iou_reward": 0.962807297706604,
"rewards/forensics_tempsamp_iou_reward": 0.6035981476306915,
"step": 561
},
{
"completion_length": 21.796875,
"epoch": 2.3514644351464433,
"grad_norm": 3.7327990531921387,
"kl": 0.328125,
"learning_rate": 4.131799163179916e-07,
"loss": 0.0131,
"reward": 0.4992970675230026,
"reward_std": 0.14693597704172134,
"reward_std_all": 0.21423710137605667,
"rewards-max/forensics_tempsamp_iou_reward": 0.9252076148986816,
"rewards/forensics_tempsamp_iou_reward": 0.4557834267616272,
"step": 562
},
{
"completion_length": 21.3125,
"epoch": 2.3556485355648533,
"grad_norm": 3.7771222591400146,
"kl": 0.2705078125,
"learning_rate": 4.121338912133891e-07,
"loss": 0.0108,
"reward": 0.5986967980861664,
"reward_std": 0.1097685918211937,
"reward_std_all": 0.14170190691947937,
"rewards-max/forensics_tempsamp_iou_reward": 0.9337850213050842,
"rewards/forensics_tempsamp_iou_reward": 0.5607682764530182,
"step": 563
},
{
"completion_length": 21.921875,
"epoch": 2.3598326359832638,
"grad_norm": 3.3176543712615967,
"kl": 0.37109375,
"learning_rate": 4.110878661087866e-07,
"loss": 0.0148,
"reward": 0.6745978593826294,
"reward_std": 0.07579115778207779,
"reward_std_all": 0.09410944953560829,
"rewards-max/forensics_tempsamp_iou_reward": 0.967901885509491,
"rewards/forensics_tempsamp_iou_reward": 0.6678906679153442,
"step": 564
},
{
"completion_length": 23.1875,
"epoch": 2.3640167364016738,
"grad_norm": 3.794670820236206,
"kl": 0.3427734375,
"learning_rate": 4.100418410041841e-07,
"loss": 0.0137,
"reward": 0.6093064248561859,
"reward_std": 0.0924217663705349,
"reward_std_all": 0.1288432590663433,
"rewards-max/forensics_tempsamp_iou_reward": 0.9438683688640594,
"rewards/forensics_tempsamp_iou_reward": 0.5847324430942535,
"step": 565
},
{
"completion_length": 20.78125,
"epoch": 2.3682008368200838,
"grad_norm": 3.376857042312622,
"kl": 0.3251953125,
"learning_rate": 4.089958158995816e-07,
"loss": 0.013,
"reward": 0.6339181065559387,
"reward_std": 0.06673239544034004,
"reward_std_all": 0.10592527687549591,
"rewards-max/forensics_tempsamp_iou_reward": 0.9856680333614349,
"rewards/forensics_tempsamp_iou_reward": 0.6104268133640289,
"step": 566
},
{
"completion_length": 23.1875,
"epoch": 2.3723849372384938,
"grad_norm": 3.5088016986846924,
"kl": 0.3115234375,
"learning_rate": 4.0794979079497904e-07,
"loss": 0.0125,
"reward": 0.6061556339263916,
"reward_std": 0.12539448030292988,
"reward_std_all": 0.15184886008501053,
"rewards-max/forensics_tempsamp_iou_reward": 0.9487417042255402,
"rewards/forensics_tempsamp_iou_reward": 0.5766194760799408,
"step": 567
},
{
"completion_length": 21.578125,
"epoch": 2.3765690376569037,
"grad_norm": 3.9436769485473633,
"kl": 0.26904296875,
"learning_rate": 4.0690376569037656e-07,
"loss": 0.0108,
"reward": 0.5664979517459869,
"reward_std": 0.1384132206439972,
"reward_std_all": 0.18296413123607635,
"rewards-max/forensics_tempsamp_iou_reward": 0.9785057604312897,
"rewards/forensics_tempsamp_iou_reward": 0.5367894470691681,
"step": 568
},
{
"completion_length": 24.609375,
"epoch": 2.3807531380753137,
"grad_norm": 3.9234373569488525,
"kl": 0.24951171875,
"learning_rate": 4.0585774058577407e-07,
"loss": 0.01,
"reward": 0.5926119089126587,
"reward_std": 0.08051668480038643,
"reward_std_all": 0.1367916688323021,
"rewards-max/forensics_tempsamp_iou_reward": 0.9734136164188385,
"rewards/forensics_tempsamp_iou_reward": 0.5680616796016693,
"step": 569
},
{
"completion_length": 23.015625,
"epoch": 2.3849372384937237,
"grad_norm": 2.94470477104187,
"kl": 0.3056640625,
"learning_rate": 4.0481171548117153e-07,
"loss": 0.0122,
"reward": 0.618526965379715,
"reward_std": 0.13014677166938782,
"reward_std_all": 0.14734162390232086,
"rewards-max/forensics_tempsamp_iou_reward": 0.9839504063129425,
"rewards/forensics_tempsamp_iou_reward": 0.6027808487415314,
"step": 570
},
{
"completion_length": 22.0,
"epoch": 2.3891213389121337,
"grad_norm": 4.663936138153076,
"kl": 0.2939453125,
"learning_rate": 4.0376569037656904e-07,
"loss": 0.0117,
"reward": 0.6009610891342163,
"reward_std": 0.10298660770058632,
"reward_std_all": 0.13647223263978958,
"rewards-max/forensics_tempsamp_iou_reward": 0.9146077334880829,
"rewards/forensics_tempsamp_iou_reward": 0.5657795816659927,
"step": 571
},
{
"completion_length": 25.359375,
"epoch": 2.393305439330544,
"grad_norm": 3.1127328872680664,
"kl": 0.349609375,
"learning_rate": 4.027196652719665e-07,
"loss": 0.014,
"reward": 0.5708608627319336,
"reward_std": 0.1459970325231552,
"reward_std_all": 0.17439260333776474,
"rewards-max/forensics_tempsamp_iou_reward": 0.9481077492237091,
"rewards/forensics_tempsamp_iou_reward": 0.5342836827039719,
"step": 572
},
{
"completion_length": 29.0625,
"epoch": 2.397489539748954,
"grad_norm": 3.647199869155884,
"kl": 0.34375,
"learning_rate": 4.01673640167364e-07,
"loss": 0.0138,
"reward": 0.5511965751647949,
"reward_std": 0.14144006371498108,
"reward_std_all": 0.18118397891521454,
"rewards-max/forensics_tempsamp_iou_reward": 0.9728695154190063,
"rewards/forensics_tempsamp_iou_reward": 0.5194943249225616,
"step": 573
},
{
"completion_length": 21.75,
"epoch": 2.401673640167364,
"grad_norm": 3.6597237586975098,
"kl": 0.3623046875,
"learning_rate": 4.006276150627615e-07,
"loss": 0.0145,
"reward": 0.6547446846961975,
"reward_std": 0.1102772131562233,
"reward_std_all": 0.12399977445602417,
"rewards-max/forensics_tempsamp_iou_reward": 0.9748409390449524,
"rewards/forensics_tempsamp_iou_reward": 0.6328780651092529,
"step": 574
},
{
"completion_length": 22.671875,
"epoch": 2.405857740585774,
"grad_norm": 3.2060394287109375,
"kl": 0.345703125,
"learning_rate": 3.99581589958159e-07,
"loss": 0.0138,
"reward": 0.6319500803947449,
"reward_std": 0.13010414689779282,
"reward_std_all": 0.14020727574825287,
"rewards-max/forensics_tempsamp_iou_reward": 0.9194819927215576,
"rewards/forensics_tempsamp_iou_reward": 0.6041719317436218,
"step": 575
},
{
"completion_length": 21.109375,
"epoch": 2.410041841004184,
"grad_norm": 4.133090972900391,
"kl": 0.345703125,
"learning_rate": 3.985355648535565e-07,
"loss": 0.0138,
"reward": 0.643950879573822,
"reward_std": 0.09225236251950264,
"reward_std_all": 0.11465983837842941,
"rewards-max/forensics_tempsamp_iou_reward": 0.965807318687439,
"rewards/forensics_tempsamp_iou_reward": 0.6279866397380829,
"step": 576
},
{
"completion_length": 23.53125,
"epoch": 2.414225941422594,
"grad_norm": 3.8716824054718018,
"kl": 0.2890625,
"learning_rate": 3.9748953974895396e-07,
"loss": 0.0115,
"reward": 0.6282783150672913,
"reward_std": 0.12395740672945976,
"reward_std_all": 0.1374230608344078,
"rewards-max/forensics_tempsamp_iou_reward": 0.9338120818138123,
"rewards/forensics_tempsamp_iou_reward": 0.5839888453483582,
"step": 577
},
{
"completion_length": 23.0625,
"epoch": 2.418410041841004,
"grad_norm": 4.578693866729736,
"kl": 0.2841796875,
"learning_rate": 3.964435146443514e-07,
"loss": 0.0114,
"reward": 0.5828222632408142,
"reward_std": 0.16624310612678528,
"reward_std_all": 0.18840254843235016,
"rewards-max/forensics_tempsamp_iou_reward": 0.939368486404419,
"rewards/forensics_tempsamp_iou_reward": 0.5367421209812164,
"step": 578
},
{
"completion_length": 23.359375,
"epoch": 2.422594142259414,
"grad_norm": 3.437190294265747,
"kl": 0.28515625,
"learning_rate": 3.9539748953974893e-07,
"loss": 0.0114,
"reward": 0.5219451487064362,
"reward_std": 0.08328575640916824,
"reward_std_all": 0.17019959539175034,
"rewards-max/forensics_tempsamp_iou_reward": 0.9537524878978729,
"rewards/forensics_tempsamp_iou_reward": 0.4814227521419525,
"step": 579
},
{
"completion_length": 22.359375,
"epoch": 2.426778242677824,
"grad_norm": 4.090584754943848,
"kl": 0.3359375,
"learning_rate": 3.9435146443514645e-07,
"loss": 0.0135,
"reward": 0.6488107442855835,
"reward_std": 0.09318359941244125,
"reward_std_all": 0.1167636550962925,
"rewards-max/forensics_tempsamp_iou_reward": 0.9266680181026459,
"rewards/forensics_tempsamp_iou_reward": 0.611078530550003,
"step": 580
},
{
"completion_length": 21.359375,
"epoch": 2.430962343096234,
"grad_norm": 3.430753469467163,
"kl": 0.2998046875,
"learning_rate": 3.9330543933054396e-07,
"loss": 0.012,
"reward": 0.5340698659420013,
"reward_std": 0.08405968174338341,
"reward_std_all": 0.15923526138067245,
"rewards-max/forensics_tempsamp_iou_reward": 0.9798002243041992,
"rewards/forensics_tempsamp_iou_reward": 0.4890493154525757,
"step": 581
},
{
"completion_length": 21.359375,
"epoch": 2.435146443514644,
"grad_norm": 3.8079559803009033,
"kl": 0.310546875,
"learning_rate": 3.9225941422594137e-07,
"loss": 0.0124,
"reward": 0.5744971632957458,
"reward_std": 0.11740944907069206,
"reward_std_all": 0.1611509546637535,
"rewards-max/forensics_tempsamp_iou_reward": 0.9757384657859802,
"rewards/forensics_tempsamp_iou_reward": 0.5399361550807953,
"step": 582
},
{
"completion_length": 20.921875,
"epoch": 2.439330543933054,
"grad_norm": 3.9568724632263184,
"kl": 0.341796875,
"learning_rate": 3.912133891213389e-07,
"loss": 0.0136,
"reward": 0.6260944604873657,
"reward_std": 0.12139134109020233,
"reward_std_all": 0.140584796667099,
"rewards-max/forensics_tempsamp_iou_reward": 0.9700054228305817,
"rewards/forensics_tempsamp_iou_reward": 0.5975183844566345,
"step": 583
},
{
"completion_length": 22.421875,
"epoch": 2.4435146443514646,
"grad_norm": 3.730532646179199,
"kl": 0.3125,
"learning_rate": 3.901673640167364e-07,
"loss": 0.0125,
"reward": 0.5787149667739868,
"reward_std": 0.08937370032072067,
"reward_std_all": 0.14426712691783905,
"rewards-max/forensics_tempsamp_iou_reward": 0.9915654063224792,
"rewards/forensics_tempsamp_iou_reward": 0.5410738289356232,
"step": 584
},
{
"completion_length": 21.921875,
"epoch": 2.4476987447698746,
"grad_norm": 3.6607253551483154,
"kl": 0.30859375,
"learning_rate": 3.891213389121339e-07,
"loss": 0.0123,
"reward": 0.6206243336200714,
"reward_std": 0.1054820790886879,
"reward_std_all": 0.13273431360721588,
"rewards-max/forensics_tempsamp_iou_reward": 0.9709268510341644,
"rewards/forensics_tempsamp_iou_reward": 0.5909880101680756,
"step": 585
},
{
"completion_length": 22.046875,
"epoch": 2.4518828451882846,
"grad_norm": 3.5657899379730225,
"kl": 0.28662109375,
"learning_rate": 3.880753138075314e-07,
"loss": 0.0115,
"reward": 0.5220147669315338,
"reward_std": 0.15818699821829796,
"reward_std_all": 0.19933411478996277,
"rewards-max/forensics_tempsamp_iou_reward": 0.8984393775463104,
"rewards/forensics_tempsamp_iou_reward": 0.4608219861984253,
"step": 586
},
{
"completion_length": 25.171875,
"epoch": 2.4560669456066946,
"grad_norm": 3.606109142303467,
"kl": 0.24658203125,
"learning_rate": 3.870292887029288e-07,
"loss": 0.0099,
"reward": 0.5487775206565857,
"reward_std": 0.16070281714200974,
"reward_std_all": 0.19035983085632324,
"rewards-max/forensics_tempsamp_iou_reward": 0.9105849266052246,
"rewards/forensics_tempsamp_iou_reward": 0.5006779730319977,
"step": 587
},
{
"completion_length": 22.5,
"epoch": 2.4602510460251046,
"grad_norm": 3.2609238624572754,
"kl": 0.28955078125,
"learning_rate": 3.8598326359832634e-07,
"loss": 0.0116,
"reward": 0.6798648834228516,
"reward_std": 0.06305016577243805,
"reward_std_all": 0.08500618115067482,
"rewards-max/forensics_tempsamp_iou_reward": 0.9949755370616913,
"rewards/forensics_tempsamp_iou_reward": 0.7078424394130707,
"step": 588
},
{
"completion_length": 21.078125,
"epoch": 2.4644351464435146,
"grad_norm": 3.352112293243408,
"kl": 0.26416015625,
"learning_rate": 3.8493723849372385e-07,
"loss": 0.0106,
"reward": 0.6774241328239441,
"reward_std": 0.0749996118247509,
"reward_std_all": 0.09413795173168182,
"rewards-max/forensics_tempsamp_iou_reward": 0.970442533493042,
"rewards/forensics_tempsamp_iou_reward": 0.6562419235706329,
"step": 589
},
{
"completion_length": 22.40625,
"epoch": 2.4686192468619246,
"grad_norm": 3.1631429195404053,
"kl": 0.34765625,
"learning_rate": 3.8389121338912136e-07,
"loss": 0.0139,
"reward": 0.5758270919322968,
"reward_std": 0.11041048541665077,
"reward_std_all": 0.1557772420346737,
"rewards-max/forensics_tempsamp_iou_reward": 0.9594098031520844,
"rewards/forensics_tempsamp_iou_reward": 0.5224132537841797,
"step": 590
},
{
"completion_length": 20.625,
"epoch": 2.4728033472803346,
"grad_norm": 3.717522621154785,
"kl": 0.3203125,
"learning_rate": 3.828451882845188e-07,
"loss": 0.0128,
"reward": 0.6219297647476196,
"reward_std": 0.12045402824878693,
"reward_std_all": 0.14257729053497314,
"rewards-max/forensics_tempsamp_iou_reward": 0.965234100818634,
"rewards/forensics_tempsamp_iou_reward": 0.5911360383033752,
"step": 591
},
{
"completion_length": 20.09375,
"epoch": 2.476987447698745,
"grad_norm": 3.996687650680542,
"kl": 0.2822265625,
"learning_rate": 3.817991631799163e-07,
"loss": 0.0113,
"reward": 0.5946872532367706,
"reward_std": 0.10051785781979561,
"reward_std_all": 0.14273889362812042,
"rewards-max/forensics_tempsamp_iou_reward": 0.952566921710968,
"rewards/forensics_tempsamp_iou_reward": 0.556111752986908,
"step": 592
},
{
"completion_length": 24.015625,
"epoch": 2.481171548117155,
"grad_norm": 2.6333556175231934,
"kl": 0.3583984375,
"learning_rate": 3.807531380753138e-07,
"loss": 0.0144,
"reward": 0.553569570183754,
"reward_std": 0.1041041798889637,
"reward_std_all": 0.1614965945482254,
"rewards-max/forensics_tempsamp_iou_reward": 0.9655989706516266,
"rewards/forensics_tempsamp_iou_reward": 0.5204084366559982,
"step": 593
},
{
"completion_length": 21.453125,
"epoch": 2.485355648535565,
"grad_norm": 3.493680238723755,
"kl": 0.255859375,
"learning_rate": 3.797071129707113e-07,
"loss": 0.0102,
"reward": 0.631303995847702,
"reward_std": 0.15271760523319244,
"reward_std_all": 0.1570901796221733,
"rewards-max/forensics_tempsamp_iou_reward": 0.9922886192798615,
"rewards/forensics_tempsamp_iou_reward": 0.6133964359760284,
"step": 594
},
{
"completion_length": 21.859375,
"epoch": 2.489539748953975,
"grad_norm": 3.3746178150177,
"kl": 0.30078125,
"learning_rate": 3.7866108786610877e-07,
"loss": 0.012,
"reward": 0.6086874902248383,
"reward_std": 0.11932839453220367,
"reward_std_all": 0.13933353126049042,
"rewards-max/forensics_tempsamp_iou_reward": 0.9725795388221741,
"rewards/forensics_tempsamp_iou_reward": 0.5968716740608215,
"step": 595
},
{
"completion_length": 19.515625,
"epoch": 2.493723849372385,
"grad_norm": 3.763921022415161,
"kl": 0.3359375,
"learning_rate": 3.776150627615063e-07,
"loss": 0.0134,
"reward": 0.6531428694725037,
"reward_std": 0.1040758565068245,
"reward_std_all": 0.12160872668027878,
"rewards-max/forensics_tempsamp_iou_reward": 0.9791970252990723,
"rewards/forensics_tempsamp_iou_reward": 0.6394948363304138,
"step": 596
},
{
"completion_length": 19.359375,
"epoch": 2.497907949790795,
"grad_norm": 5.3168768882751465,
"kl": 0.3046875,
"learning_rate": 3.7656903765690374e-07,
"loss": 0.0122,
"reward": 0.6011612415313721,
"reward_std": 0.07981645688414574,
"reward_std_all": 0.12980560958385468,
"rewards-max/forensics_tempsamp_iou_reward": 0.9398858547210693,
"rewards/forensics_tempsamp_iou_reward": 0.5653946399688721,
"step": 597
},
{
"completion_length": 20.734375,
"epoch": 2.502092050209205,
"grad_norm": 4.101611614227295,
"kl": 0.3994140625,
"learning_rate": 3.7552301255230125e-07,
"loss": 0.016,
"reward": 0.6108323931694031,
"reward_std": 0.11586211621761322,
"reward_std_all": 0.14521579444408417,
"rewards-max/forensics_tempsamp_iou_reward": 0.9810636341571808,
"rewards/forensics_tempsamp_iou_reward": 0.5943670570850372,
"step": 598
},
{
"completion_length": 21.296875,
"epoch": 2.506276150627615,
"grad_norm": 3.5795226097106934,
"kl": 0.26171875,
"learning_rate": 3.744769874476987e-07,
"loss": 0.0105,
"reward": 0.5883554816246033,
"reward_std": 0.054660266265273094,
"reward_std_all": 0.12066063284873962,
"rewards-max/forensics_tempsamp_iou_reward": 0.9561880826950073,
"rewards/forensics_tempsamp_iou_reward": 0.5732365697622299,
"step": 599
},
{
"completion_length": 20.75,
"epoch": 2.510460251046025,
"grad_norm": 4.246886253356934,
"kl": 0.3056640625,
"learning_rate": 3.734309623430962e-07,
"loss": 0.0122,
"reward": 0.6504931449890137,
"reward_std": 0.07914502173662186,
"reward_std_all": 0.11035866662859917,
"rewards-max/forensics_tempsamp_iou_reward": 0.9899076521396637,
"rewards/forensics_tempsamp_iou_reward": 0.6406541764736176,
"step": 600
},
{
"completion_length": 23.609375,
"epoch": 2.514644351464435,
"grad_norm": 3.5316014289855957,
"kl": 0.3056640625,
"learning_rate": 3.723849372384937e-07,
"loss": 0.0122,
"reward": 0.5983056426048279,
"reward_std": 0.09721928834915161,
"reward_std_all": 0.14281054586172104,
"rewards-max/forensics_tempsamp_iou_reward": 0.9583944082260132,
"rewards/forensics_tempsamp_iou_reward": 0.564404159784317,
"step": 601
},
{
"completion_length": 20.5625,
"epoch": 2.518828451882845,
"grad_norm": 4.299693584442139,
"kl": 0.244140625,
"learning_rate": 3.713389121338912e-07,
"loss": 0.0098,
"reward": 0.6924675703048706,
"reward_std": 0.05647977441549301,
"reward_std_all": 0.08162900805473328,
"rewards-max/forensics_tempsamp_iou_reward": 0.9895201325416565,
"rewards/forensics_tempsamp_iou_reward": 0.7212570905685425,
"step": 602
},
{
"completion_length": 25.0,
"epoch": 2.523012552301255,
"grad_norm": 3.2921907901763916,
"kl": 0.3017578125,
"learning_rate": 3.7029288702928866e-07,
"loss": 0.0121,
"reward": 0.5553880631923676,
"reward_std": 0.13558777421712875,
"reward_std_all": 0.17536525428295135,
"rewards-max/forensics_tempsamp_iou_reward": 0.9583052396774292,
"rewards/forensics_tempsamp_iou_reward": 0.5114840567111969,
"step": 603
},
{
"completion_length": 24.09375,
"epoch": 2.5271966527196654,
"grad_norm": 3.459507942199707,
"kl": 0.25244140625,
"learning_rate": 3.6924686192468617e-07,
"loss": 0.0101,
"reward": 0.6129578351974487,
"reward_std": 0.059796277433633804,
"reward_std_all": 0.11675035208463669,
"rewards-max/forensics_tempsamp_iou_reward": 0.970708817243576,
"rewards/forensics_tempsamp_iou_reward": 0.593353271484375,
"step": 604
},
{
"completion_length": 22.515625,
"epoch": 2.5313807531380754,
"grad_norm": 3.117269992828369,
"kl": 0.28515625,
"learning_rate": 3.682008368200837e-07,
"loss": 0.0114,
"reward": 0.5921386480331421,
"reward_std": 0.08722283504903316,
"reward_std_all": 0.14364532753825188,
"rewards-max/forensics_tempsamp_iou_reward": 0.9639187157154083,
"rewards/forensics_tempsamp_iou_reward": 0.5762631893157959,
"step": 605
},
{
"completion_length": 24.078125,
"epoch": 2.5355648535564854,
"grad_norm": 3.675764560699463,
"kl": 0.302734375,
"learning_rate": 3.6715481171548114e-07,
"loss": 0.0121,
"reward": 0.652106523513794,
"reward_std": 0.0809782836586237,
"reward_std_all": 0.10578892379999161,
"rewards-max/forensics_tempsamp_iou_reward": 0.9382214546203613,
"rewards/forensics_tempsamp_iou_reward": 0.6209999918937683,
"step": 606
},
{
"completion_length": 20.984375,
"epoch": 2.5397489539748954,
"grad_norm": 2.884434938430786,
"kl": 0.3203125,
"learning_rate": 3.661087866108786e-07,
"loss": 0.0128,
"reward": 0.6349229514598846,
"reward_std": 0.07303816825151443,
"reward_std_all": 0.11201514676213264,
"rewards-max/forensics_tempsamp_iou_reward": 0.9800192713737488,
"rewards/forensics_tempsamp_iou_reward": 0.6400167346000671,
"step": 607
},
{
"completion_length": 21.1875,
"epoch": 2.5439330543933054,
"grad_norm": 3.877326488494873,
"kl": 0.2705078125,
"learning_rate": 3.650627615062761e-07,
"loss": 0.0108,
"reward": 0.600864827632904,
"reward_std": 0.0735573060810566,
"reward_std_all": 0.12726978585124016,
"rewards-max/forensics_tempsamp_iou_reward": 0.9924250841140747,
"rewards/forensics_tempsamp_iou_reward": 0.5724217891693115,
"step": 608
},
{
"completion_length": 24.328125,
"epoch": 2.5481171548117154,
"grad_norm": 3.6900267601013184,
"kl": 0.2900390625,
"learning_rate": 3.6401673640167363e-07,
"loss": 0.0116,
"reward": 0.5705026090145111,
"reward_std": 0.09698760136961937,
"reward_std_all": 0.1493896171450615,
"rewards-max/forensics_tempsamp_iou_reward": 0.9475453794002533,
"rewards/forensics_tempsamp_iou_reward": 0.5248421728610992,
"step": 609
},
{
"completion_length": 24.328125,
"epoch": 2.5523012552301254,
"grad_norm": 3.5189170837402344,
"kl": 0.294921875,
"learning_rate": 3.6297071129707114e-07,
"loss": 0.0118,
"reward": 0.621088832616806,
"reward_std": 0.10967160761356354,
"reward_std_all": 0.13371721655130386,
"rewards-max/forensics_tempsamp_iou_reward": 0.9550910592079163,
"rewards/forensics_tempsamp_iou_reward": 0.5795931220054626,
"step": 610
},
{
"completion_length": 21.328125,
"epoch": 2.5564853556485354,
"grad_norm": 4.430899620056152,
"kl": 0.326171875,
"learning_rate": 3.6192468619246855e-07,
"loss": 0.0131,
"reward": 0.5574095845222473,
"reward_std": 0.11256799101829529,
"reward_std_all": 0.15865596383810043,
"rewards-max/forensics_tempsamp_iou_reward": 0.9490938186645508,
"rewards/forensics_tempsamp_iou_reward": 0.5126940906047821,
"step": 611
},
{
"completion_length": 23.0625,
"epoch": 2.560669456066946,
"grad_norm": 3.95123291015625,
"kl": 0.251953125,
"learning_rate": 3.6087866108786606e-07,
"loss": 0.0101,
"reward": 0.5465464890003204,
"reward_std": 0.12461555004119873,
"reward_std_all": 0.1684238314628601,
"rewards-max/forensics_tempsamp_iou_reward": 0.8883891999721527,
"rewards/forensics_tempsamp_iou_reward": 0.48112303018569946,
"step": 612
},
{
"completion_length": 24.625,
"epoch": 2.564853556485356,
"grad_norm": 3.2312681674957275,
"kl": 0.265625,
"learning_rate": 3.598326359832636e-07,
"loss": 0.0106,
"reward": 0.6207885146141052,
"reward_std": 0.08389360085129738,
"reward_std_all": 0.12023128941655159,
"rewards-max/forensics_tempsamp_iou_reward": 0.8980579376220703,
"rewards/forensics_tempsamp_iou_reward": 0.5798745602369308,
"step": 613
},
{
"completion_length": 22.015625,
"epoch": 2.569037656903766,
"grad_norm": 4.9787774085998535,
"kl": 0.3349609375,
"learning_rate": 3.587866108786611e-07,
"loss": 0.0134,
"reward": 0.5192758738994598,
"reward_std": 0.14669708162546158,
"reward_std_all": 0.19666026532649994,
"rewards-max/forensics_tempsamp_iou_reward": 0.9236807823181152,
"rewards/forensics_tempsamp_iou_reward": 0.45599226653575897,
"step": 614
},
{
"completion_length": 21.359375,
"epoch": 2.573221757322176,
"grad_norm": 4.248890399932861,
"kl": 0.3876953125,
"learning_rate": 3.577405857740586e-07,
"loss": 0.0155,
"reward": 0.5512374639511108,
"reward_std": 0.11371862143278122,
"reward_std_all": 0.16622334718704224,
"rewards-max/forensics_tempsamp_iou_reward": 0.9838961064815521,
"rewards/forensics_tempsamp_iou_reward": 0.5205874890089035,
"step": 615
},
{
"completion_length": 22.078125,
"epoch": 2.577405857740586,
"grad_norm": 3.808805465698242,
"kl": 0.294921875,
"learning_rate": 3.56694560669456e-07,
"loss": 0.0118,
"reward": 0.6537117660045624,
"reward_std": 0.11417174711823463,
"reward_std_all": 0.12522833421826363,
"rewards-max/forensics_tempsamp_iou_reward": 0.9498840570449829,
"rewards/forensics_tempsamp_iou_reward": 0.6431044340133667,
"step": 616
},
{
"completion_length": 21.625,
"epoch": 2.581589958158996,
"grad_norm": 3.0728094577789307,
"kl": 0.2861328125,
"learning_rate": 3.556485355648535e-07,
"loss": 0.0114,
"reward": 0.6461372077465057,
"reward_std": 0.1428610235452652,
"reward_std_all": 0.14559144526720047,
"rewards-max/forensics_tempsamp_iou_reward": 0.9730086922645569,
"rewards/forensics_tempsamp_iou_reward": 0.6414053440093994,
"step": 617
},
{
"completion_length": 21.203125,
"epoch": 2.585774058577406,
"grad_norm": 4.41464376449585,
"kl": 0.3359375,
"learning_rate": 3.5460251046025103e-07,
"loss": 0.0134,
"reward": 0.4938051998615265,
"reward_std": 0.1361643448472023,
"reward_std_all": 0.2017139121890068,
"rewards-max/forensics_tempsamp_iou_reward": 0.9110865890979767,
"rewards/forensics_tempsamp_iou_reward": 0.42506635189056396,
"step": 618
},
{
"completion_length": 20.09375,
"epoch": 2.589958158995816,
"grad_norm": 4.39471960067749,
"kl": 0.24755859375,
"learning_rate": 3.5355648535564855e-07,
"loss": 0.0099,
"reward": 0.6933718621730804,
"reward_std": 0.09064193442463875,
"reward_std_all": 0.0957554318010807,
"rewards-max/forensics_tempsamp_iou_reward": 0.9426442384719849,
"rewards/forensics_tempsamp_iou_reward": 0.676351010799408,
"step": 619
},
{
"completion_length": 21.9375,
"epoch": 2.594142259414226,
"grad_norm": 4.550974369049072,
"kl": 0.4794921875,
"learning_rate": 3.52510460251046e-07,
"loss": 0.0192,
"reward": 0.6326103806495667,
"reward_std": 0.11123710125684738,
"reward_std_all": 0.13188401609659195,
"rewards-max/forensics_tempsamp_iou_reward": 0.9769256114959717,
"rewards/forensics_tempsamp_iou_reward": 0.6091565191745758,
"step": 620
},
{
"completion_length": 25.59375,
"epoch": 2.598326359832636,
"grad_norm": 3.063918352127075,
"kl": 0.2822265625,
"learning_rate": 3.5146443514644347e-07,
"loss": 0.0113,
"reward": 0.6112503856420517,
"reward_std": 0.09763641655445099,
"reward_std_all": 0.13666915893554688,
"rewards-max/forensics_tempsamp_iou_reward": 0.9227200150489807,
"rewards/forensics_tempsamp_iou_reward": 0.5754857063293457,
"step": 621
},
{
"completion_length": 21.65625,
"epoch": 2.602510460251046,
"grad_norm": 3.187391519546509,
"kl": 0.3623046875,
"learning_rate": 3.50418410041841e-07,
"loss": 0.0145,
"reward": 0.5709642469882965,
"reward_std": 0.14726316928863525,
"reward_std_all": 0.18460176140069962,
"rewards-max/forensics_tempsamp_iou_reward": 0.9375,
"rewards/forensics_tempsamp_iou_reward": 0.551268070936203,
"step": 622
},
{
"completion_length": 22.21875,
"epoch": 2.606694560669456,
"grad_norm": 3.8583481311798096,
"kl": 0.28515625,
"learning_rate": 3.493723849372385e-07,
"loss": 0.0114,
"reward": 0.5181092023849487,
"reward_std": 0.14150075614452362,
"reward_std_all": 0.19191660732030869,
"rewards-max/forensics_tempsamp_iou_reward": 0.9244455993175507,
"rewards/forensics_tempsamp_iou_reward": 0.450262188911438,
"step": 623
},
{
"completion_length": 22.46875,
"epoch": 2.6108786610878663,
"grad_norm": 3.9330837726593018,
"kl": 0.26953125,
"learning_rate": 3.4832635983263595e-07,
"loss": 0.0108,
"reward": 0.6138279736042023,
"reward_std": 0.053839173167943954,
"reward_std_all": 0.11412803083658218,
"rewards-max/forensics_tempsamp_iou_reward": 0.9282636046409607,
"rewards/forensics_tempsamp_iou_reward": 0.5864170491695404,
"step": 624
},
{
"completion_length": 21.59375,
"epoch": 2.6150627615062763,
"grad_norm": 3.766523599624634,
"kl": 0.2880859375,
"learning_rate": 3.4728033472803346e-07,
"loss": 0.0115,
"reward": 0.5538625121116638,
"reward_std": 0.11352476477622986,
"reward_std_all": 0.17088905721902847,
"rewards-max/forensics_tempsamp_iou_reward": 0.9452923238277435,
"rewards/forensics_tempsamp_iou_reward": 0.5121610909700394,
"step": 625
},
{
"completion_length": 21.34375,
"epoch": 2.6192468619246863,
"grad_norm": 3.4199180603027344,
"kl": 0.26708984375,
"learning_rate": 3.462343096234309e-07,
"loss": 0.0107,
"reward": 0.49646054208278656,
"reward_std": 0.11342882737517357,
"reward_std_all": 0.19075243920087814,
"rewards-max/forensics_tempsamp_iou_reward": 0.9263237714767456,
"rewards/forensics_tempsamp_iou_reward": 0.43894895911216736,
"step": 626
},
{
"completion_length": 26.1875,
"epoch": 2.6234309623430963,
"grad_norm": 9.291690826416016,
"kl": 0.6103515625,
"learning_rate": 3.4518828451882844e-07,
"loss": 0.0245,
"reward": 0.5446388423442841,
"reward_std": 0.13391615450382233,
"reward_std_all": 0.17761249095201492,
"rewards-max/forensics_tempsamp_iou_reward": 0.9302096366882324,
"rewards/forensics_tempsamp_iou_reward": 0.48792779445648193,
"step": 627
},
{
"completion_length": 24.875,
"epoch": 2.6276150627615062,
"grad_norm": 3.4347984790802,
"kl": 0.3232421875,
"learning_rate": 3.441422594142259e-07,
"loss": 0.0129,
"reward": 0.598111480474472,
"reward_std": 0.09453686140477657,
"reward_std_all": 0.1398836150765419,
"rewards-max/forensics_tempsamp_iou_reward": 0.9658273160457611,
"rewards/forensics_tempsamp_iou_reward": 0.58257757127285,
"step": 628
},
{
"completion_length": 24.203125,
"epoch": 2.6317991631799162,
"grad_norm": 3.3421034812927246,
"kl": 0.4208984375,
"learning_rate": 3.430962343096234e-07,
"loss": 0.0168,
"reward": 0.5302221477031708,
"reward_std": 0.16839954257011414,
"reward_std_all": 0.1987835094332695,
"rewards-max/forensics_tempsamp_iou_reward": 0.9747619032859802,
"rewards/forensics_tempsamp_iou_reward": 0.47473499178886414,
"step": 629
},
{
"completion_length": 20.53125,
"epoch": 2.6359832635983262,
"grad_norm": 3.4060797691345215,
"kl": 0.345703125,
"learning_rate": 3.420502092050209e-07,
"loss": 0.0138,
"reward": 0.6363090872764587,
"reward_std": 0.0998586006462574,
"reward_std_all": 0.1248692199587822,
"rewards-max/forensics_tempsamp_iou_reward": 0.9777970612049103,
"rewards/forensics_tempsamp_iou_reward": 0.624195009469986,
"step": 630
},
{
"completion_length": 21.078125,
"epoch": 2.6401673640167362,
"grad_norm": 4.069526195526123,
"kl": 0.341796875,
"learning_rate": 3.410041841004184e-07,
"loss": 0.0137,
"reward": 0.6160279512405396,
"reward_std": 0.09099408984184265,
"reward_std_all": 0.13295096531510353,
"rewards-max/forensics_tempsamp_iou_reward": 0.9549190998077393,
"rewards/forensics_tempsamp_iou_reward": 0.5813497602939606,
"step": 631
},
{
"completion_length": 27.109375,
"epoch": 2.6443514644351467,
"grad_norm": 2.9064435958862305,
"kl": 0.2548828125,
"learning_rate": 3.3995815899581584e-07,
"loss": 0.0102,
"reward": 0.6268495321273804,
"reward_std": 0.1233019009232521,
"reward_std_all": 0.14461533352732658,
"rewards-max/forensics_tempsamp_iou_reward": 0.9729066789150238,
"rewards/forensics_tempsamp_iou_reward": 0.6055904626846313,
"step": 632
},
{
"completion_length": 23.890625,
"epoch": 2.6485355648535567,
"grad_norm": 3.475452423095703,
"kl": 0.32421875,
"learning_rate": 3.3891213389121336e-07,
"loss": 0.0129,
"reward": 0.6015328764915466,
"reward_std": 0.08729894831776619,
"reward_std_all": 0.1311866194009781,
"rewards-max/forensics_tempsamp_iou_reward": 0.9307331144809723,
"rewards/forensics_tempsamp_iou_reward": 0.5512396097183228,
"step": 633
},
{
"completion_length": 23.90625,
"epoch": 2.6527196652719667,
"grad_norm": 3.282285451889038,
"kl": 0.26171875,
"learning_rate": 3.3786610878661087e-07,
"loss": 0.0104,
"reward": 0.553039163351059,
"reward_std": 0.10746150836348534,
"reward_std_all": 0.16339002549648285,
"rewards-max/forensics_tempsamp_iou_reward": 0.9853858351707458,
"rewards/forensics_tempsamp_iou_reward": 0.5177697837352753,
"step": 634
},
{
"completion_length": 24.28125,
"epoch": 2.6569037656903767,
"grad_norm": 3.6057136058807373,
"kl": 0.3076171875,
"learning_rate": 3.368200836820084e-07,
"loss": 0.0123,
"reward": 0.5872741341590881,
"reward_std": 0.16256285458803177,
"reward_std_all": 0.17678164690732956,
"rewards-max/forensics_tempsamp_iou_reward": 0.9540030658245087,
"rewards/forensics_tempsamp_iou_reward": 0.5483111590147018,
"step": 635
},
{
"completion_length": 23.78125,
"epoch": 2.6610878661087867,
"grad_norm": 2.985278367996216,
"kl": 0.22509765625,
"learning_rate": 3.3577405857740584e-07,
"loss": 0.009,
"reward": 0.6360033750534058,
"reward_std": 0.06235573999583721,
"reward_std_all": 0.10871709883213043,
"rewards-max/forensics_tempsamp_iou_reward": 0.9843358397483826,
"rewards/forensics_tempsamp_iou_reward": 0.6227855980396271,
"step": 636
},
{
"completion_length": 21.046875,
"epoch": 2.6652719665271967,
"grad_norm": 3.139543056488037,
"kl": 0.2919921875,
"learning_rate": 3.347280334728033e-07,
"loss": 0.0117,
"reward": 0.6337639093399048,
"reward_std": 0.13411591947078705,
"reward_std_all": 0.14293218404054642,
"rewards-max/forensics_tempsamp_iou_reward": 0.9872037768363953,
"rewards/forensics_tempsamp_iou_reward": 0.6411427557468414,
"step": 637
},
{
"completion_length": 22.09375,
"epoch": 2.6694560669456067,
"grad_norm": 3.1471338272094727,
"kl": 0.28515625,
"learning_rate": 3.336820083682008e-07,
"loss": 0.0114,
"reward": 0.5781482458114624,
"reward_std": 0.08277568221092224,
"reward_std_all": 0.147479847073555,
"rewards-max/forensics_tempsamp_iou_reward": 0.9407638609409332,
"rewards/forensics_tempsamp_iou_reward": 0.5599465668201447,
"step": 638
},
{
"completion_length": 20.96875,
"epoch": 2.6736401673640167,
"grad_norm": 4.228304862976074,
"kl": 0.3193359375,
"learning_rate": 3.3263598326359833e-07,
"loss": 0.0128,
"reward": 0.6204688251018524,
"reward_std": 0.11957770586013794,
"reward_std_all": 0.1375497803092003,
"rewards-max/forensics_tempsamp_iou_reward": 0.9643837511539459,
"rewards/forensics_tempsamp_iou_reward": 0.590234637260437,
"step": 639
},
{
"completion_length": 26.828125,
"epoch": 2.6778242677824267,
"grad_norm": 3.1941349506378174,
"kl": 0.359375,
"learning_rate": 3.3158995815899584e-07,
"loss": 0.0144,
"reward": 0.4796629101037979,
"reward_std": 0.15943820774555206,
"reward_std_all": 0.22280504554510117,
"rewards-max/forensics_tempsamp_iou_reward": 0.9727468192577362,
"rewards/forensics_tempsamp_iou_reward": 0.4297988712787628,
"step": 640
},
{
"completion_length": 22.15625,
"epoch": 2.6820083682008367,
"grad_norm": 3.9240517616271973,
"kl": 0.32421875,
"learning_rate": 3.3054393305439325e-07,
"loss": 0.013,
"reward": 0.5953053832054138,
"reward_std": 0.15687179565429688,
"reward_std_all": 0.17552541196346283,
"rewards-max/forensics_tempsamp_iou_reward": 0.9616459608078003,
"rewards/forensics_tempsamp_iou_reward": 0.5484549701213837,
"step": 641
},
{
"completion_length": 22.40625,
"epoch": 2.6861924686192467,
"grad_norm": 4.222926139831543,
"kl": 0.3193359375,
"learning_rate": 3.2949790794979076e-07,
"loss": 0.0128,
"reward": 0.6463072299957275,
"reward_std": 0.07709444314241409,
"reward_std_all": 0.10876954719424248,
"rewards-max/forensics_tempsamp_iou_reward": 0.9676842391490936,
"rewards/forensics_tempsamp_iou_reward": 0.6299434006214142,
"step": 642
},
{
"completion_length": 24.03125,
"epoch": 2.6903765690376567,
"grad_norm": 3.6438992023468018,
"kl": 0.275390625,
"learning_rate": 3.2845188284518827e-07,
"loss": 0.011,
"reward": 0.4901754558086395,
"reward_std": 0.13876695930957794,
"reward_std_all": 0.2109624296426773,
"rewards-max/forensics_tempsamp_iou_reward": 0.9527312815189362,
"rewards/forensics_tempsamp_iou_reward": 0.4270303547382355,
"step": 643
},
{
"completion_length": 25.296875,
"epoch": 2.694560669456067,
"grad_norm": 3.2288429737091064,
"kl": 0.2861328125,
"learning_rate": 3.274058577405858e-07,
"loss": 0.0114,
"reward": 0.6050442159175873,
"reward_std": 0.06301477551460266,
"reward_std_all": 0.1183616891503334,
"rewards-max/forensics_tempsamp_iou_reward": 0.9714860916137695,
"rewards/forensics_tempsamp_iou_reward": 0.568668395280838,
"step": 644
},
{
"completion_length": 20.03125,
"epoch": 2.698744769874477,
"grad_norm": 3.4631550312042236,
"kl": 0.322265625,
"learning_rate": 3.2635983263598325e-07,
"loss": 0.0128,
"reward": 0.6552241146564484,
"reward_std": 0.10063625872135162,
"reward_std_all": 0.12001919746398926,
"rewards-max/forensics_tempsamp_iou_reward": 0.9733690917491913,
"rewards/forensics_tempsamp_iou_reward": 0.6468003392219543,
"step": 645
},
{
"completion_length": 21.5,
"epoch": 2.702928870292887,
"grad_norm": 3.544546127319336,
"kl": 0.3115234375,
"learning_rate": 3.253138075313807e-07,
"loss": 0.0124,
"reward": 0.6037134826183319,
"reward_std": 0.08348578587174416,
"reward_std_all": 0.12954752147197723,
"rewards-max/forensics_tempsamp_iou_reward": 0.9208795428276062,
"rewards/forensics_tempsamp_iou_reward": 0.5574637353420258,
"step": 646
},
{
"completion_length": 22.0625,
"epoch": 2.707112970711297,
"grad_norm": 3.317963123321533,
"kl": 0.30859375,
"learning_rate": 3.242677824267782e-07,
"loss": 0.0123,
"reward": 0.5965796113014221,
"reward_std": 0.11765872687101364,
"reward_std_all": 0.15702896565198898,
"rewards-max/forensics_tempsamp_iou_reward": 0.9500021934509277,
"rewards/forensics_tempsamp_iou_reward": 0.5737986862659454,
"step": 647
},
{
"completion_length": 22.390625,
"epoch": 2.711297071129707,
"grad_norm": 3.735795736312866,
"kl": 0.318359375,
"learning_rate": 3.2322175732217573e-07,
"loss": 0.0127,
"reward": 0.6172704100608826,
"reward_std": 0.130568228662014,
"reward_std_all": 0.15113970637321472,
"rewards-max/forensics_tempsamp_iou_reward": 0.9932178556919098,
"rewards/forensics_tempsamp_iou_reward": 0.602893590927124,
"step": 648
},
{
"completion_length": 24.4375,
"epoch": 2.715481171548117,
"grad_norm": 3.3417537212371826,
"kl": 0.3037109375,
"learning_rate": 3.221757322175732e-07,
"loss": 0.0122,
"reward": 0.5941672921180725,
"reward_std": 0.10577447339892387,
"reward_std_all": 0.15219411253929138,
"rewards-max/forensics_tempsamp_iou_reward": 0.9407255053520203,
"rewards/forensics_tempsamp_iou_reward": 0.5527370274066925,
"step": 649
},
{
"completion_length": 23.328125,
"epoch": 2.719665271966527,
"grad_norm": 3.1483654975891113,
"kl": 0.29296875,
"learning_rate": 3.211297071129707e-07,
"loss": 0.0117,
"reward": 0.6355357468128204,
"reward_std": 0.10250138118863106,
"reward_std_all": 0.12426773831248283,
"rewards-max/forensics_tempsamp_iou_reward": 0.9367874562740326,
"rewards/forensics_tempsamp_iou_reward": 0.5908394753932953,
"step": 650
},
{
"completion_length": 21.109375,
"epoch": 2.723849372384937,
"grad_norm": 3.58354115486145,
"kl": 0.28271484375,
"learning_rate": 3.2008368200836816e-07,
"loss": 0.0113,
"reward": 0.6496289372444153,
"reward_std": 0.07985902205109596,
"reward_std_all": 0.10757678747177124,
"rewards-max/forensics_tempsamp_iou_reward": 0.9342105388641357,
"rewards/forensics_tempsamp_iou_reward": 0.6093459129333496,
"step": 651
},
{
"completion_length": 22.421875,
"epoch": 2.7280334728033475,
"grad_norm": 3.7148451805114746,
"kl": 0.2451171875,
"learning_rate": 3.190376569037657e-07,
"loss": 0.0098,
"reward": 0.6515793204307556,
"reward_std": 0.09036925435066223,
"reward_std_all": 0.10805292055010796,
"rewards-max/forensics_tempsamp_iou_reward": 0.9748037457466125,
"rewards/forensics_tempsamp_iou_reward": 0.6441369354724884,
"step": 652
},
{
"completion_length": 21.546875,
"epoch": 2.7322175732217575,
"grad_norm": 4.749652862548828,
"kl": 0.373046875,
"learning_rate": 3.1799163179916314e-07,
"loss": 0.0149,
"reward": 0.5287840217351913,
"reward_std": 0.12280981987714767,
"reward_std_all": 0.18334315717220306,
"rewards-max/forensics_tempsamp_iou_reward": 0.9201511740684509,
"rewards/forensics_tempsamp_iou_reward": 0.4791155606508255,
"step": 653
},
{
"completion_length": 20.59375,
"epoch": 2.7364016736401675,
"grad_norm": 2.983733654022217,
"kl": 0.2841796875,
"learning_rate": 3.1694560669456065e-07,
"loss": 0.0114,
"reward": 0.6214417815208435,
"reward_std": 0.09164834022521973,
"reward_std_all": 0.12362902611494064,
"rewards-max/forensics_tempsamp_iou_reward": 0.9572416841983795,
"rewards/forensics_tempsamp_iou_reward": 0.5967545509338379,
"step": 654
},
{
"completion_length": 21.359375,
"epoch": 2.7405857740585775,
"grad_norm": 4.272718906402588,
"kl": 0.26904296875,
"learning_rate": 3.1589958158995816e-07,
"loss": 0.0107,
"reward": 0.480203777551651,
"reward_std": 0.12658098712563515,
"reward_std_all": 0.20671354234218597,
"rewards-max/forensics_tempsamp_iou_reward": 0.8999539613723755,
"rewards/forensics_tempsamp_iou_reward": 0.41230161488056183,
"step": 655
},
{
"completion_length": 20.375,
"epoch": 2.7447698744769875,
"grad_norm": 3.3441245555877686,
"kl": 0.279296875,
"learning_rate": 3.148535564853556e-07,
"loss": 0.0112,
"reward": 0.6781189143657684,
"reward_std": 0.09590207785367966,
"reward_std_all": 0.10573500394821167,
"rewards-max/forensics_tempsamp_iou_reward": 0.9887225031852722,
"rewards/forensics_tempsamp_iou_reward": 0.6720601916313171,
"step": 656
},
{
"completion_length": 21.734375,
"epoch": 2.7489539748953975,
"grad_norm": 3.0810024738311768,
"kl": 0.3330078125,
"learning_rate": 3.1380753138075313e-07,
"loss": 0.0133,
"reward": 0.6262463331222534,
"reward_std": 0.09520746394991875,
"reward_std_all": 0.12559575214982033,
"rewards-max/forensics_tempsamp_iou_reward": 0.9814895689487457,
"rewards/forensics_tempsamp_iou_reward": 0.6393488943576813,
"step": 657
},
{
"completion_length": 22.4375,
"epoch": 2.7531380753138075,
"grad_norm": 4.032228946685791,
"kl": 0.2587890625,
"learning_rate": 3.127615062761506e-07,
"loss": 0.0104,
"reward": 0.5769912600517273,
"reward_std": 0.1142331175506115,
"reward_std_all": 0.15640505403280258,
"rewards-max/forensics_tempsamp_iou_reward": 0.9485458135604858,
"rewards/forensics_tempsamp_iou_reward": 0.530319482088089,
"step": 658
},
{
"completion_length": 23.40625,
"epoch": 2.7573221757322175,
"grad_norm": 4.1430864334106445,
"kl": 0.28515625,
"learning_rate": 3.117154811715481e-07,
"loss": 0.0114,
"reward": 0.6032213568687439,
"reward_std": 0.10318268090486526,
"reward_std_all": 0.14400455355644226,
"rewards-max/forensics_tempsamp_iou_reward": 0.9325861632823944,
"rewards/forensics_tempsamp_iou_reward": 0.5570269823074341,
"step": 659
},
{
"completion_length": 20.6875,
"epoch": 2.7615062761506275,
"grad_norm": 3.195777416229248,
"kl": 0.3486328125,
"learning_rate": 3.106694560669456e-07,
"loss": 0.0139,
"reward": 0.6231753826141357,
"reward_std": 0.13026998937129974,
"reward_std_all": 0.14974753931164742,
"rewards-max/forensics_tempsamp_iou_reward": 0.9480767846107483,
"rewards/forensics_tempsamp_iou_reward": 0.5994877219200134,
"step": 660
},
{
"completion_length": 22.578125,
"epoch": 2.7656903765690375,
"grad_norm": 3.4554266929626465,
"kl": 0.296875,
"learning_rate": 3.096234309623431e-07,
"loss": 0.0119,
"reward": 0.5466829240322113,
"reward_std": 0.0872900914400816,
"reward_std_all": 0.1599983051419258,
"rewards-max/forensics_tempsamp_iou_reward": 0.9298961460590363,
"rewards/forensics_tempsamp_iou_reward": 0.5151547491550446,
"step": 661
},
{
"completion_length": 21.03125,
"epoch": 2.7698744769874475,
"grad_norm": 4.219101428985596,
"kl": 0.3046875,
"learning_rate": 3.0857740585774054e-07,
"loss": 0.0122,
"reward": 0.5788836926221848,
"reward_std": 0.1065061017870903,
"reward_std_all": 0.15737732499837875,
"rewards-max/forensics_tempsamp_iou_reward": 0.931291252374649,
"rewards/forensics_tempsamp_iou_reward": 0.5323945879936218,
"step": 662
},
{
"completion_length": 20.234375,
"epoch": 2.7740585774058575,
"grad_norm": 2.779808759689331,
"kl": 0.3076171875,
"learning_rate": 3.0753138075313805e-07,
"loss": 0.0123,
"reward": 0.6500290036201477,
"reward_std": 0.0450700968503952,
"reward_std_all": 0.09148093685507774,
"rewards-max/forensics_tempsamp_iou_reward": 0.9794057607650757,
"rewards/forensics_tempsamp_iou_reward": 0.6590697765350342,
"step": 663
},
{
"completion_length": 23.015625,
"epoch": 2.778242677824268,
"grad_norm": 3.7051169872283936,
"kl": 0.310546875,
"learning_rate": 3.0648535564853557e-07,
"loss": 0.0124,
"reward": 0.5672456920146942,
"reward_std": 0.11589067801833153,
"reward_std_all": 0.16354955732822418,
"rewards-max/forensics_tempsamp_iou_reward": 0.973960280418396,
"rewards/forensics_tempsamp_iou_reward": 0.5417978167533875,
"step": 664
},
{
"completion_length": 21.53125,
"epoch": 2.782426778242678,
"grad_norm": 3.765585422515869,
"kl": 0.3984375,
"learning_rate": 3.054393305439331e-07,
"loss": 0.0159,
"reward": 0.4947871118783951,
"reward_std": 0.18791787326335907,
"reward_std_all": 0.22878380864858627,
"rewards-max/forensics_tempsamp_iou_reward": 0.8838200867176056,
"rewards/forensics_tempsamp_iou_reward": 0.4321691244840622,
"step": 665
},
{
"completion_length": 24.40625,
"epoch": 2.786610878661088,
"grad_norm": 3.454228639602661,
"kl": 0.3173828125,
"learning_rate": 3.043933054393305e-07,
"loss": 0.0127,
"reward": 0.5908461213111877,
"reward_std": 0.07185960561037064,
"reward_std_all": 0.13205718994140625,
"rewards-max/forensics_tempsamp_iou_reward": 0.967840313911438,
"rewards/forensics_tempsamp_iou_reward": 0.5683613419532776,
"step": 666
},
{
"completion_length": 23.125,
"epoch": 2.790794979079498,
"grad_norm": 3.69488263130188,
"kl": 0.27734375,
"learning_rate": 3.03347280334728e-07,
"loss": 0.0111,
"reward": 0.594020813703537,
"reward_std": 0.11345468461513519,
"reward_std_all": 0.15316064655780792,
"rewards-max/forensics_tempsamp_iou_reward": 0.9614655077457428,
"rewards/forensics_tempsamp_iou_reward": 0.5709936916828156,
"step": 667
},
{
"completion_length": 21.71875,
"epoch": 2.794979079497908,
"grad_norm": 3.485447645187378,
"kl": 0.255859375,
"learning_rate": 3.023012552301255e-07,
"loss": 0.0103,
"reward": 0.6052148938179016,
"reward_std": 0.07050754688680172,
"reward_std_all": 0.12643496692180634,
"rewards-max/forensics_tempsamp_iou_reward": 0.9770126640796661,
"rewards/forensics_tempsamp_iou_reward": 0.5867641568183899,
"step": 668
},
{
"completion_length": 23.1875,
"epoch": 2.799163179916318,
"grad_norm": 3.393594264984131,
"kl": 0.3603515625,
"learning_rate": 3.01255230125523e-07,
"loss": 0.0144,
"reward": 0.5533435642719269,
"reward_std": 0.14479156583547592,
"reward_std_all": 0.1821710243821144,
"rewards-max/forensics_tempsamp_iou_reward": 0.9871684014797211,
"rewards/forensics_tempsamp_iou_reward": 0.5232573747634888,
"step": 669
},
{
"completion_length": 21.3125,
"epoch": 2.803347280334728,
"grad_norm": 3.9777650833129883,
"kl": 0.283203125,
"learning_rate": 3.002092050209205e-07,
"loss": 0.0113,
"reward": 0.5523435920476913,
"reward_std": 0.11308852583169937,
"reward_std_all": 0.17046084254980087,
"rewards-max/forensics_tempsamp_iou_reward": 0.9706152677536011,
"rewards/forensics_tempsamp_iou_reward": 0.5304816663265228,
"step": 670
},
{
"completion_length": 20.109375,
"epoch": 2.8075313807531384,
"grad_norm": 3.0195436477661133,
"kl": 0.3408203125,
"learning_rate": 2.9916317991631794e-07,
"loss": 0.0137,
"reward": 0.6468496918678284,
"reward_std": 0.1087062880396843,
"reward_std_all": 0.12447264790534973,
"rewards-max/forensics_tempsamp_iou_reward": 0.9774772822856903,
"rewards/forensics_tempsamp_iou_reward": 0.6146302223205566,
"step": 671
},
{
"completion_length": 20.8125,
"epoch": 2.8117154811715483,
"grad_norm": 3.5465030670166016,
"kl": 0.3837890625,
"learning_rate": 2.9811715481171546e-07,
"loss": 0.0153,
"reward": 0.6483493447303772,
"reward_std": 0.1183515414595604,
"reward_std_all": 0.1270548328757286,
"rewards-max/forensics_tempsamp_iou_reward": 0.9265790581703186,
"rewards/forensics_tempsamp_iou_reward": 0.6139612793922424,
"step": 672
},
{
"completion_length": 21.921875,
"epoch": 2.8158995815899583,
"grad_norm": 3.733797788619995,
"kl": 0.31640625,
"learning_rate": 2.9707112970711297e-07,
"loss": 0.0126,
"reward": 0.5737069249153137,
"reward_std": 0.14129740744829178,
"reward_std_all": 0.17360208928585052,
"rewards-max/forensics_tempsamp_iou_reward": 0.92955282330513,
"rewards/forensics_tempsamp_iou_reward": 0.528062105178833,
"step": 673
},
{
"completion_length": 21.546875,
"epoch": 2.8200836820083683,
"grad_norm": 3.3256096839904785,
"kl": 0.20654296875,
"learning_rate": 2.960251046025105e-07,
"loss": 0.0083,
"reward": 0.5754133760929108,
"reward_std": 0.11449195817112923,
"reward_std_all": 0.1541481465101242,
"rewards-max/forensics_tempsamp_iou_reward": 0.8952365815639496,
"rewards/forensics_tempsamp_iou_reward": 0.5164327174425125,
"step": 674
},
{
"completion_length": 20.9375,
"epoch": 2.8242677824267783,
"grad_norm": 3.9653918743133545,
"kl": 0.33203125,
"learning_rate": 2.9497907949790794e-07,
"loss": 0.0133,
"reward": 0.6418025493621826,
"reward_std": 0.11213788762688637,
"reward_std_all": 0.13075289130210876,
"rewards-max/forensics_tempsamp_iou_reward": 0.9335792064666748,
"rewards/forensics_tempsamp_iou_reward": 0.6213106513023376,
"step": 675
},
{
"completion_length": 19.5,
"epoch": 2.8284518828451883,
"grad_norm": 3.815150499343872,
"kl": 0.408203125,
"learning_rate": 2.939330543933054e-07,
"loss": 0.0163,
"reward": 0.6728692650794983,
"reward_std": 0.0463024377822876,
"reward_std_all": 0.08540932834148407,
"rewards-max/forensics_tempsamp_iou_reward": 0.9868012368679047,
"rewards/forensics_tempsamp_iou_reward": 0.6868253946304321,
"step": 676
},
{
"completion_length": 24.609375,
"epoch": 2.8326359832635983,
"grad_norm": 4.0053791999816895,
"kl": 0.3603515625,
"learning_rate": 2.928870292887029e-07,
"loss": 0.0145,
"reward": 0.5784385204315186,
"reward_std": 0.15461038425564766,
"reward_std_all": 0.18128690123558044,
"rewards-max/forensics_tempsamp_iou_reward": 0.929152250289917,
"rewards/forensics_tempsamp_iou_reward": 0.5364931672811508,
"step": 677
},
{
"completion_length": 22.71875,
"epoch": 2.8368200836820083,
"grad_norm": 3.3843791484832764,
"kl": 0.2734375,
"learning_rate": 2.9184100418410043e-07,
"loss": 0.011,
"reward": 0.6264186799526215,
"reward_std": 0.1128426045179367,
"reward_std_all": 0.13805611059069633,
"rewards-max/forensics_tempsamp_iou_reward": 0.9420257210731506,
"rewards/forensics_tempsamp_iou_reward": 0.6081046760082245,
"step": 678
},
{
"completion_length": 23.296875,
"epoch": 2.8410041841004183,
"grad_norm": 3.397686243057251,
"kl": 0.2939453125,
"learning_rate": 2.907949790794979e-07,
"loss": 0.0118,
"reward": 0.5616505742073059,
"reward_std": 0.14982762932777405,
"reward_std_all": 0.17481151223182678,
"rewards-max/forensics_tempsamp_iou_reward": 0.9624746739864349,
"rewards/forensics_tempsamp_iou_reward": 0.5299995243549347,
"step": 679
},
{
"completion_length": 20.828125,
"epoch": 2.8451882845188283,
"grad_norm": 3.220405101776123,
"kl": 0.27734375,
"learning_rate": 2.897489539748954e-07,
"loss": 0.0111,
"reward": 0.651875376701355,
"reward_std": 0.08418696001172066,
"reward_std_all": 0.10472263023257256,
"rewards-max/forensics_tempsamp_iou_reward": 0.9505556225776672,
"rewards/forensics_tempsamp_iou_reward": 0.6304767429828644,
"step": 680
},
{
"completion_length": 21.453125,
"epoch": 2.8493723849372383,
"grad_norm": 3.9489519596099854,
"kl": 0.3310546875,
"learning_rate": 2.8870292887029286e-07,
"loss": 0.0132,
"reward": 0.5360816717147827,
"reward_std": 0.11968819051980972,
"reward_std_all": 0.17961417138576508,
"rewards-max/forensics_tempsamp_iou_reward": 0.831722766160965,
"rewards/forensics_tempsamp_iou_reward": 0.46346694231033325,
"step": 681
},
{
"completion_length": 21.34375,
"epoch": 2.8535564853556483,
"grad_norm": 3.158705949783325,
"kl": 0.23681640625,
"learning_rate": 2.8765690376569037e-07,
"loss": 0.0095,
"reward": 0.6761603951454163,
"reward_std": 0.08116457238793373,
"reward_std_all": 0.10020026564598083,
"rewards-max/forensics_tempsamp_iou_reward": 0.9223038256168365,
"rewards/forensics_tempsamp_iou_reward": 0.6466088593006134,
"step": 682
},
{
"completion_length": 20.703125,
"epoch": 2.8577405857740583,
"grad_norm": 4.656108856201172,
"kl": 0.2666015625,
"learning_rate": 2.8661087866108783e-07,
"loss": 0.0107,
"reward": 0.5771493315696716,
"reward_std": 0.0970553383231163,
"reward_std_all": 0.15434294193983078,
"rewards-max/forensics_tempsamp_iou_reward": 0.9683290123939514,
"rewards/forensics_tempsamp_iou_reward": 0.5389849841594696,
"step": 683
},
{
"completion_length": 22.578125,
"epoch": 2.8619246861924688,
"grad_norm": 3.3651957511901855,
"kl": 0.2626953125,
"learning_rate": 2.8556485355648535e-07,
"loss": 0.0105,
"reward": 0.5878520309925079,
"reward_std": 0.09226076304912567,
"reward_std_all": 0.1397608071565628,
"rewards-max/forensics_tempsamp_iou_reward": 0.9403260052204132,
"rewards/forensics_tempsamp_iou_reward": 0.531660407781601,
"step": 684
},
{
"completion_length": 21.015625,
"epoch": 2.8661087866108788,
"grad_norm": 3.351952314376831,
"kl": 0.294921875,
"learning_rate": 2.8451882845188286e-07,
"loss": 0.0118,
"reward": 0.5914384126663208,
"reward_std": 0.11222663149237633,
"reward_std_all": 0.1517285332083702,
"rewards-max/forensics_tempsamp_iou_reward": 0.9892851710319519,
"rewards/forensics_tempsamp_iou_reward": 0.561022624373436,
"step": 685
},
{
"completion_length": 21.75,
"epoch": 2.8702928870292888,
"grad_norm": 3.468640089035034,
"kl": 0.3134765625,
"learning_rate": 2.834728033472803e-07,
"loss": 0.0125,
"reward": 0.6339369416236877,
"reward_std": 0.0777743011713028,
"reward_std_all": 0.1159452497959137,
"rewards-max/forensics_tempsamp_iou_reward": 0.9510228335857391,
"rewards/forensics_tempsamp_iou_reward": 0.6255444586277008,
"step": 686
},
{
"completion_length": 19.578125,
"epoch": 2.8744769874476988,
"grad_norm": 3.6239097118377686,
"kl": 0.287109375,
"learning_rate": 2.824267782426778e-07,
"loss": 0.0115,
"reward": 0.6374494433403015,
"reward_std": 0.09540623053908348,
"reward_std_all": 0.1207464411854744,
"rewards-max/forensics_tempsamp_iou_reward": 0.9640681147575378,
"rewards/forensics_tempsamp_iou_reward": 0.623212605714798,
"step": 687
},
{
"completion_length": 22.171875,
"epoch": 2.8786610878661087,
"grad_norm": 3.4448964595794678,
"kl": 0.27490234375,
"learning_rate": 2.813807531380753e-07,
"loss": 0.011,
"reward": 0.6615195274353027,
"reward_std": 0.07832407206296921,
"reward_std_all": 0.10419698059558868,
"rewards-max/forensics_tempsamp_iou_reward": 0.9928825497627258,
"rewards/forensics_tempsamp_iou_reward": 0.645116001367569,
"step": 688
},
{
"completion_length": 22.109375,
"epoch": 2.8828451882845187,
"grad_norm": 3.4994771480560303,
"kl": 0.2958984375,
"learning_rate": 2.803347280334728e-07,
"loss": 0.0119,
"reward": 0.6433141827583313,
"reward_std": 0.08056185394525528,
"reward_std_all": 0.10732344910502434,
"rewards-max/forensics_tempsamp_iou_reward": 0.9338937103748322,
"rewards/forensics_tempsamp_iou_reward": 0.6009943187236786,
"step": 689
},
{
"completion_length": 25.0625,
"epoch": 2.8870292887029287,
"grad_norm": 3.5537502765655518,
"kl": 0.3193359375,
"learning_rate": 2.792887029288703e-07,
"loss": 0.0128,
"reward": 0.4828842431306839,
"reward_std": 0.1329953446984291,
"reward_std_all": 0.20989003032445908,
"rewards-max/forensics_tempsamp_iou_reward": 0.9323753714561462,
"rewards/forensics_tempsamp_iou_reward": 0.4173436462879181,
"step": 690
},
{
"completion_length": 23.953125,
"epoch": 2.891213389121339,
"grad_norm": 3.1917836666107178,
"kl": 0.333984375,
"learning_rate": 2.782426778242677e-07,
"loss": 0.0133,
"reward": 0.5488016605377197,
"reward_std": 0.13292908668518066,
"reward_std_all": 0.17824088782072067,
"rewards-max/forensics_tempsamp_iou_reward": 0.927255243062973,
"rewards/forensics_tempsamp_iou_reward": 0.49816279113292694,
"step": 691
},
{
"completion_length": 20.78125,
"epoch": 2.895397489539749,
"grad_norm": 4.214409828186035,
"kl": 0.3671875,
"learning_rate": 2.7719665271966524e-07,
"loss": 0.0147,
"reward": 0.6530353426933289,
"reward_std": 0.08994311466813087,
"reward_std_all": 0.10948063433170319,
"rewards-max/forensics_tempsamp_iou_reward": 0.9855879843235016,
"rewards/forensics_tempsamp_iou_reward": 0.6269619762897491,
"step": 692
},
{
"completion_length": 22.0,
"epoch": 2.899581589958159,
"grad_norm": 3.4495558738708496,
"kl": 0.302734375,
"learning_rate": 2.7615062761506275e-07,
"loss": 0.0121,
"reward": 0.5167496502399445,
"reward_std": 0.10491203516721725,
"reward_std_all": 0.17432709783315659,
"rewards-max/forensics_tempsamp_iou_reward": 0.8236111104488373,
"rewards/forensics_tempsamp_iou_reward": 0.44425031542778015,
"step": 693
},
{
"completion_length": 21.84375,
"epoch": 2.903765690376569,
"grad_norm": 4.102766036987305,
"kl": 0.330078125,
"learning_rate": 2.7510460251046026e-07,
"loss": 0.0132,
"reward": 0.6449437439441681,
"reward_std": 0.10337377712130547,
"reward_std_all": 0.11898164078593254,
"rewards-max/forensics_tempsamp_iou_reward": 0.920066237449646,
"rewards/forensics_tempsamp_iou_reward": 0.5972423553466797,
"step": 694
},
{
"completion_length": 19.9375,
"epoch": 2.907949790794979,
"grad_norm": 3.452371597290039,
"kl": 0.3095703125,
"learning_rate": 2.740585774058578e-07,
"loss": 0.0124,
"reward": 0.669518232345581,
"reward_std": 0.059802427887916565,
"reward_std_all": 0.09012790396809578,
"rewards-max/forensics_tempsamp_iou_reward": 0.9670390784740448,
"rewards/forensics_tempsamp_iou_reward": 0.6490434408187866,
"step": 695
},
{
"completion_length": 22.34375,
"epoch": 2.912133891213389,
"grad_norm": 3.2574613094329834,
"kl": 0.3134765625,
"learning_rate": 2.730125523012552e-07,
"loss": 0.0125,
"reward": 0.6863498389720917,
"reward_std": 0.07419846393167973,
"reward_std_all": 0.0866105780005455,
"rewards-max/forensics_tempsamp_iou_reward": 0.9320692718029022,
"rewards/forensics_tempsamp_iou_reward": 0.6666356921195984,
"step": 696
},
{
"completion_length": 22.953125,
"epoch": 2.916317991631799,
"grad_norm": 3.942436456680298,
"kl": 0.302734375,
"learning_rate": 2.719665271966527e-07,
"loss": 0.0121,
"reward": 0.6359078884124756,
"reward_std": 0.13104232028126717,
"reward_std_all": 0.14168337732553482,
"rewards-max/forensics_tempsamp_iou_reward": 0.9528480172157288,
"rewards/forensics_tempsamp_iou_reward": 0.5908394753932953,
"step": 697
},
{
"completion_length": 20.65625,
"epoch": 2.920502092050209,
"grad_norm": 4.0867085456848145,
"kl": 0.326171875,
"learning_rate": 2.709205020920502e-07,
"loss": 0.013,
"reward": 0.5968901515007019,
"reward_std": 0.06478001922369003,
"reward_std_all": 0.1268930658698082,
"rewards-max/forensics_tempsamp_iou_reward": 0.9499081969261169,
"rewards/forensics_tempsamp_iou_reward": 0.5722161531448364,
"step": 698
},
{
"completion_length": 22.078125,
"epoch": 2.924686192468619,
"grad_norm": 3.630370616912842,
"kl": 0.3759765625,
"learning_rate": 2.698744769874477e-07,
"loss": 0.0151,
"reward": 0.5815798342227936,
"reward_std": 0.153439462184906,
"reward_std_all": 0.17557479441165924,
"rewards-max/forensics_tempsamp_iou_reward": 0.9622410535812378,
"rewards/forensics_tempsamp_iou_reward": 0.569472461938858,
"step": 699
},
{
"completion_length": 23.390625,
"epoch": 2.928870292887029,
"grad_norm": 3.158795118331909,
"kl": 0.2890625,
"learning_rate": 2.688284518828452e-07,
"loss": 0.0116,
"reward": 0.6689203381538391,
"reward_std": 0.10134642757475376,
"reward_std_all": 0.1091356985270977,
"rewards-max/forensics_tempsamp_iou_reward": 0.980546772480011,
"rewards/forensics_tempsamp_iou_reward": 0.6497800350189209,
"step": 700
},
{
"completion_length": 23.65625,
"epoch": 2.933054393305439,
"grad_norm": 3.7011148929595947,
"kl": 0.31640625,
"learning_rate": 2.6778242677824264e-07,
"loss": 0.0127,
"reward": 0.6055537462234497,
"reward_std": 0.07891785353422165,
"reward_std_all": 0.12893827259540558,
"rewards-max/forensics_tempsamp_iou_reward": 0.9592277407646179,
"rewards/forensics_tempsamp_iou_reward": 0.566539466381073,
"step": 701
},
{
"completion_length": 22.578125,
"epoch": 2.937238493723849,
"grad_norm": 4.090435981750488,
"kl": 0.34375,
"learning_rate": 2.6673640167364015e-07,
"loss": 0.0138,
"reward": 0.6124224960803986,
"reward_std": 0.06575345247983932,
"reward_std_all": 0.11713037639856339,
"rewards-max/forensics_tempsamp_iou_reward": 0.9695390462875366,
"rewards/forensics_tempsamp_iou_reward": 0.5986289381980896,
"step": 702
},
{
"completion_length": 21.59375,
"epoch": 2.941422594142259,
"grad_norm": 3.6748275756835938,
"kl": 0.2802734375,
"learning_rate": 2.6569037656903767e-07,
"loss": 0.0112,
"reward": 0.6305477321147919,
"reward_std": 0.12961489334702492,
"reward_std_all": 0.14274130761623383,
"rewards-max/forensics_tempsamp_iou_reward": 0.9671422839164734,
"rewards/forensics_tempsamp_iou_reward": 0.6042807400226593,
"step": 703
},
{
"completion_length": 21.296875,
"epoch": 2.9456066945606696,
"grad_norm": 3.5739998817443848,
"kl": 0.322265625,
"learning_rate": 2.6464435146443513e-07,
"loss": 0.0129,
"reward": 0.596469521522522,
"reward_std": 0.11608601734042168,
"reward_std_all": 0.14587794244289398,
"rewards-max/forensics_tempsamp_iou_reward": 0.9587695896625519,
"rewards/forensics_tempsamp_iou_reward": 0.5401234924793243,
"step": 704
},
{
"completion_length": 21.015625,
"epoch": 2.9497907949790796,
"grad_norm": 4.094165325164795,
"kl": 0.318359375,
"learning_rate": 2.6359832635983264e-07,
"loss": 0.0127,
"reward": 0.5684386491775513,
"reward_std": 0.11727627739310265,
"reward_std_all": 0.16452713310718536,
"rewards-max/forensics_tempsamp_iou_reward": 0.9521668553352356,
"rewards/forensics_tempsamp_iou_reward": 0.5330586433410645,
"step": 705
},
{
"completion_length": 22.234375,
"epoch": 2.9539748953974896,
"grad_norm": 5.375185966491699,
"kl": 0.40625,
"learning_rate": 2.625523012552301e-07,
"loss": 0.0163,
"reward": 0.5887442827224731,
"reward_std": 0.13168835639953613,
"reward_std_all": 0.1722474992275238,
"rewards-max/forensics_tempsamp_iou_reward": 0.961458295583725,
"rewards/forensics_tempsamp_iou_reward": 0.5811972916126251,
"step": 706
},
{
"completion_length": 21.234375,
"epoch": 2.9581589958158996,
"grad_norm": 3.897305965423584,
"kl": 0.3232421875,
"learning_rate": 2.615062761506276e-07,
"loss": 0.0129,
"reward": 0.6451468169689178,
"reward_std": 0.09464456886053085,
"reward_std_all": 0.12092753499746323,
"rewards-max/forensics_tempsamp_iou_reward": 0.9453286528587341,
"rewards/forensics_tempsamp_iou_reward": 0.6168397963047028,
"step": 707
},
{
"completion_length": 23.234375,
"epoch": 2.9623430962343096,
"grad_norm": 3.662595748901367,
"kl": 0.2255859375,
"learning_rate": 2.6046025104602507e-07,
"loss": 0.009,
"reward": 0.6400285363197327,
"reward_std": 0.10431942716240883,
"reward_std_all": 0.1277022361755371,
"rewards-max/forensics_tempsamp_iou_reward": 0.9609699249267578,
"rewards/forensics_tempsamp_iou_reward": 0.6339298188686371,
"step": 708
},
{
"completion_length": 22.65625,
"epoch": 2.9665271966527196,
"grad_norm": 3.5984089374542236,
"kl": 0.2900390625,
"learning_rate": 2.594142259414226e-07,
"loss": 0.0116,
"reward": 0.656071126461029,
"reward_std": 0.11980466544628143,
"reward_std_all": 0.12637963145971298,
"rewards-max/forensics_tempsamp_iou_reward": 0.9771292805671692,
"rewards/forensics_tempsamp_iou_reward": 0.6446283161640167,
"step": 709
},
{
"completion_length": 20.609375,
"epoch": 2.9707112970711296,
"grad_norm": 3.784977436065674,
"kl": 0.310546875,
"learning_rate": 2.583682008368201e-07,
"loss": 0.0124,
"reward": 0.6403099596500397,
"reward_std": 0.10296735912561417,
"reward_std_all": 0.126574095338583,
"rewards-max/forensics_tempsamp_iou_reward": 0.9225796461105347,
"rewards/forensics_tempsamp_iou_reward": 0.5988414585590363,
"step": 710
},
{
"completion_length": 21.765625,
"epoch": 2.97489539748954,
"grad_norm": 3.3981339931488037,
"kl": 0.2958984375,
"learning_rate": 2.5732217573221756e-07,
"loss": 0.0118,
"reward": 0.634176641702652,
"reward_std": 0.1130896657705307,
"reward_std_all": 0.12960295006632805,
"rewards-max/forensics_tempsamp_iou_reward": 0.9666057825088501,
"rewards/forensics_tempsamp_iou_reward": 0.6109161674976349,
"step": 711
},
{
"completion_length": 22.1875,
"epoch": 2.97907949790795,
"grad_norm": 3.4436068534851074,
"kl": 0.3193359375,
"learning_rate": 2.56276150627615e-07,
"loss": 0.0128,
"reward": 0.6055036783218384,
"reward_std": 0.09790243208408356,
"reward_std_all": 0.13956987857818604,
"rewards-max/forensics_tempsamp_iou_reward": 0.9429652988910675,
"rewards/forensics_tempsamp_iou_reward": 0.5602905750274658,
"step": 712
},
{
"completion_length": 20.28125,
"epoch": 2.98326359832636,
"grad_norm": 3.6943657398223877,
"kl": 0.30078125,
"learning_rate": 2.5523012552301253e-07,
"loss": 0.012,
"reward": 0.6998597681522369,
"reward_std": 0.079256153665483,
"reward_std_all": 0.08939945884048939,
"rewards-max/forensics_tempsamp_iou_reward": 0.9616349041461945,
"rewards/forensics_tempsamp_iou_reward": 0.7206786274909973,
"step": 713
},
{
"completion_length": 23.75,
"epoch": 2.98744769874477,
"grad_norm": 3.5994458198547363,
"kl": 0.3671875,
"learning_rate": 2.5418410041841004e-07,
"loss": 0.0147,
"reward": 0.646806389093399,
"reward_std": 0.12965572252869606,
"reward_std_all": 0.13592582195997238,
"rewards-max/forensics_tempsamp_iou_reward": 0.9978025257587433,
"rewards/forensics_tempsamp_iou_reward": 0.6350214779376984,
"step": 714
},
{
"completion_length": 23.921875,
"epoch": 2.99163179916318,
"grad_norm": 3.51706862449646,
"kl": 0.4326171875,
"learning_rate": 2.5313807531380756e-07,
"loss": 0.0173,
"reward": 0.6008530259132385,
"reward_std": 0.09877537190914154,
"reward_std_all": 0.14456713944673538,
"rewards-max/forensics_tempsamp_iou_reward": 0.9877625107765198,
"rewards/forensics_tempsamp_iou_reward": 0.5790881216526031,
"step": 715
},
{
"completion_length": 20.71875,
"epoch": 2.99581589958159,
"grad_norm": 4.372291088104248,
"kl": 0.47265625,
"learning_rate": 2.5209205020920496e-07,
"loss": 0.0189,
"reward": 0.6095146536827087,
"reward_std": 0.1033933237195015,
"reward_std_all": 0.13486167043447495,
"rewards-max/forensics_tempsamp_iou_reward": 0.9836122393608093,
"rewards/forensics_tempsamp_iou_reward": 0.5973573327064514,
"step": 716
},
{
"completion_length": 18.92857265472412,
"epoch": 3.0,
"grad_norm": 3.2266385555267334,
"kl": 0.26806640625,
"learning_rate": 2.510460251046025e-07,
"loss": 0.0105,
"reward": 0.5538999140262604,
"reward_std": 0.11601581797003746,
"reward_std_all": 0.1815042346715927,
"rewards-max/forensics_tempsamp_iou_reward": 0.78663569688797,
"rewards/forensics_tempsamp_iou_reward": 0.47257116436958313,
"step": 717
},
{
"completion_length": 20.703125,
"epoch": 3.00418410041841,
"grad_norm": 3.344970941543579,
"kl": 0.2685546875,
"learning_rate": 2.5e-07,
"loss": 0.0107,
"reward": 0.5846231877803802,
"reward_std": 0.1060704905539751,
"reward_std_all": 0.15154192969202995,
"rewards-max/forensics_tempsamp_iou_reward": 0.9550267457962036,
"rewards/forensics_tempsamp_iou_reward": 0.5447432398796082,
"step": 718
},
{
"completion_length": 22.125,
"epoch": 3.00836820083682,
"grad_norm": 3.355693817138672,
"kl": 0.34375,
"learning_rate": 2.489539748953975e-07,
"loss": 0.0137,
"reward": 0.6836667656898499,
"reward_std": 0.10838402435183525,
"reward_std_all": 0.1099388338625431,
"rewards-max/forensics_tempsamp_iou_reward": 0.9662972986698151,
"rewards/forensics_tempsamp_iou_reward": 0.6672153174877167,
"step": 719
},
{
"completion_length": 20.34375,
"epoch": 3.01255230125523,
"grad_norm": 5.427825450897217,
"kl": 0.3720703125,
"learning_rate": 2.4790794979079496e-07,
"loss": 0.0149,
"reward": 0.7192235589027405,
"reward_std": 0.07394707947969437,
"reward_std_all": 0.07617325335741043,
"rewards-max/forensics_tempsamp_iou_reward": 0.9762239754199982,
"rewards/forensics_tempsamp_iou_reward": 0.7277051210403442,
"step": 720
},
{
"completion_length": 20.609375,
"epoch": 3.01673640167364,
"grad_norm": 3.1223161220550537,
"kl": 0.294921875,
"learning_rate": 2.468619246861925e-07,
"loss": 0.0118,
"reward": 0.5876792967319489,
"reward_std": 0.05431168153882027,
"reward_std_all": 0.12367380037903786,
"rewards-max/forensics_tempsamp_iou_reward": 0.9891836941242218,
"rewards/forensics_tempsamp_iou_reward": 0.5526955723762512,
"step": 721
},
{
"completion_length": 21.03125,
"epoch": 3.02092050209205,
"grad_norm": 3.543445348739624,
"kl": 0.2509765625,
"learning_rate": 2.4581589958158993e-07,
"loss": 0.0101,
"reward": 0.7062680125236511,
"reward_std": 0.053665364161133766,
"reward_std_all": 0.06921328976750374,
"rewards-max/forensics_tempsamp_iou_reward": 0.9865110814571381,
"rewards/forensics_tempsamp_iou_reward": 0.7233990132808685,
"step": 722
},
{
"completion_length": 21.546875,
"epoch": 3.0251046025104604,
"grad_norm": 3.5032708644866943,
"kl": 0.3212890625,
"learning_rate": 2.4476987447698745e-07,
"loss": 0.0129,
"reward": 0.6799125969409943,
"reward_std": 0.11842907965183258,
"reward_std_all": 0.11783576384186745,
"rewards-max/forensics_tempsamp_iou_reward": 0.9796621799468994,
"rewards/forensics_tempsamp_iou_reward": 0.6800848841667175,
"step": 723
},
{
"completion_length": 20.859375,
"epoch": 3.0292887029288704,
"grad_norm": 4.13994026184082,
"kl": 0.4033203125,
"learning_rate": 2.4372384937238496e-07,
"loss": 0.0161,
"reward": 0.6362946629524231,
"reward_std": 0.08204414695501328,
"reward_std_all": 0.11399751529097557,
"rewards-max/forensics_tempsamp_iou_reward": 0.9358419179916382,
"rewards/forensics_tempsamp_iou_reward": 0.6019563376903534,
"step": 724
},
{
"completion_length": 22.15625,
"epoch": 3.0334728033472804,
"grad_norm": 3.317507028579712,
"kl": 0.306640625,
"learning_rate": 2.426778242677824e-07,
"loss": 0.0123,
"reward": 0.6947639882564545,
"reward_std": 0.07408966310322285,
"reward_std_all": 0.08364052884280682,
"rewards-max/forensics_tempsamp_iou_reward": 0.9715147018432617,
"rewards/forensics_tempsamp_iou_reward": 0.6944238543510437,
"step": 725
},
{
"completion_length": 24.28125,
"epoch": 3.0376569037656904,
"grad_norm": 4.005174160003662,
"kl": 0.259765625,
"learning_rate": 2.4163179916317993e-07,
"loss": 0.0104,
"reward": 0.5772732496261597,
"reward_std": 0.09833129867911339,
"reward_std_all": 0.15045075118541718,
"rewards-max/forensics_tempsamp_iou_reward": 0.9592714607715607,
"rewards/forensics_tempsamp_iou_reward": 0.5311418771743774,
"step": 726
},
{
"completion_length": 21.9375,
"epoch": 3.0418410041841004,
"grad_norm": 4.011869430541992,
"kl": 0.3095703125,
"learning_rate": 2.405857740585774e-07,
"loss": 0.0124,
"reward": 0.6681427359580994,
"reward_std": 0.059497736394405365,
"reward_std_all": 0.08749701455235481,
"rewards-max/forensics_tempsamp_iou_reward": 0.9302968382835388,
"rewards/forensics_tempsamp_iou_reward": 0.6431573629379272,
"step": 727
},
{
"completion_length": 21.765625,
"epoch": 3.0460251046025104,
"grad_norm": 4.180687427520752,
"kl": 0.4052734375,
"learning_rate": 2.395397489539749e-07,
"loss": 0.0162,
"reward": 0.5807349979877472,
"reward_std": 0.1354065127670765,
"reward_std_all": 0.17196105420589447,
"rewards-max/forensics_tempsamp_iou_reward": 0.9943639039993286,
"rewards/forensics_tempsamp_iou_reward": 0.5863382518291473,
"step": 728
},
{
"completion_length": 21.28125,
"epoch": 3.0502092050209204,
"grad_norm": 3.395458698272705,
"kl": 0.28515625,
"learning_rate": 2.3849372384937237e-07,
"loss": 0.0114,
"reward": 0.678011417388916,
"reward_std": 0.06397948414087296,
"reward_std_all": 0.09065898507833481,
"rewards-max/forensics_tempsamp_iou_reward": 0.9389802813529968,
"rewards/forensics_tempsamp_iou_reward": 0.6715332567691803,
"step": 729
},
{
"completion_length": 22.546875,
"epoch": 3.0543933054393304,
"grad_norm": 3.632216453552246,
"kl": 0.2890625,
"learning_rate": 2.3744769874476985e-07,
"loss": 0.0115,
"reward": 0.629254013299942,
"reward_std": 0.12401673570275307,
"reward_std_all": 0.13482706248760223,
"rewards-max/forensics_tempsamp_iou_reward": 0.9881870150566101,
"rewards/forensics_tempsamp_iou_reward": 0.6148384213447571,
"step": 730
},
{
"completion_length": 23.875,
"epoch": 3.0585774058577404,
"grad_norm": 3.663996696472168,
"kl": 0.3515625,
"learning_rate": 2.3640167364016736e-07,
"loss": 0.014,
"reward": 0.6245902180671692,
"reward_std": 0.08336102962493896,
"reward_std_all": 0.12143489718437195,
"rewards-max/forensics_tempsamp_iou_reward": 0.9590642750263214,
"rewards/forensics_tempsamp_iou_reward": 0.6059985160827637,
"step": 731
},
{
"completion_length": 19.953125,
"epoch": 3.062761506276151,
"grad_norm": 4.117970943450928,
"kl": 0.2978515625,
"learning_rate": 2.3535564853556482e-07,
"loss": 0.0119,
"reward": 0.6723162233829498,
"reward_std": 0.09220881760120392,
"reward_std_all": 0.10484597831964493,
"rewards-max/forensics_tempsamp_iou_reward": 0.9772844314575195,
"rewards/forensics_tempsamp_iou_reward": 0.6721116602420807,
"step": 732
},
{
"completion_length": 22.296875,
"epoch": 3.066945606694561,
"grad_norm": 3.5191380977630615,
"kl": 0.2783203125,
"learning_rate": 2.3430962343096234e-07,
"loss": 0.0111,
"reward": 0.5579967498779297,
"reward_std": 0.1326952800154686,
"reward_std_all": 0.1793874353170395,
"rewards-max/forensics_tempsamp_iou_reward": 0.9435531198978424,
"rewards/forensics_tempsamp_iou_reward": 0.5146861970424652,
"step": 733
},
{
"completion_length": 22.578125,
"epoch": 3.071129707112971,
"grad_norm": 11.259969711303711,
"kl": 0.6640625,
"learning_rate": 2.3326359832635982e-07,
"loss": 0.0265,
"reward": 0.6339172124862671,
"reward_std": 0.16327060759067535,
"reward_std_all": 0.1620386615395546,
"rewards-max/forensics_tempsamp_iou_reward": 0.9798733294010162,
"rewards/forensics_tempsamp_iou_reward": 0.6224926114082336,
"step": 734
},
{
"completion_length": 21.15625,
"epoch": 3.075313807531381,
"grad_norm": 4.784045696258545,
"kl": 0.3095703125,
"learning_rate": 2.322175732217573e-07,
"loss": 0.0124,
"reward": 0.6453885138034821,
"reward_std": 0.0967586413025856,
"reward_std_all": 0.12264003232121468,
"rewards-max/forensics_tempsamp_iou_reward": 0.9441777765750885,
"rewards/forensics_tempsamp_iou_reward": 0.6308363080024719,
"step": 735
},
{
"completion_length": 20.203125,
"epoch": 3.079497907949791,
"grad_norm": 3.950704336166382,
"kl": 0.322265625,
"learning_rate": 2.311715481171548e-07,
"loss": 0.0129,
"reward": 0.6549343466758728,
"reward_std": 0.06794585473835468,
"reward_std_all": 0.09819387644529343,
"rewards-max/forensics_tempsamp_iou_reward": 0.979929119348526,
"rewards/forensics_tempsamp_iou_reward": 0.6518147587776184,
"step": 736
},
{
"completion_length": 22.53125,
"epoch": 3.083682008368201,
"grad_norm": 3.837066411972046,
"kl": 0.3203125,
"learning_rate": 2.3012552301255228e-07,
"loss": 0.0128,
"reward": 0.6322376132011414,
"reward_std": 0.06509104371070862,
"reward_std_all": 0.10629544407129288,
"rewards-max/forensics_tempsamp_iou_reward": 0.9799407124519348,
"rewards/forensics_tempsamp_iou_reward": 0.6290172934532166,
"step": 737
},
{
"completion_length": 22.140625,
"epoch": 3.087866108786611,
"grad_norm": 3.6684818267822266,
"kl": 0.26953125,
"learning_rate": 2.2907949790794977e-07,
"loss": 0.0108,
"reward": 0.6515555381774902,
"reward_std": 0.09042157977819443,
"reward_std_all": 0.11330331116914749,
"rewards-max/forensics_tempsamp_iou_reward": 0.9365082383155823,
"rewards/forensics_tempsamp_iou_reward": 0.6143901944160461,
"step": 738
},
{
"completion_length": 21.484375,
"epoch": 3.092050209205021,
"grad_norm": 4.233645915985107,
"kl": 0.28857421875,
"learning_rate": 2.2803347280334728e-07,
"loss": 0.0115,
"reward": 0.6375695466995239,
"reward_std": 0.11295890808105469,
"reward_std_all": 0.12908057123422623,
"rewards-max/forensics_tempsamp_iou_reward": 0.9613657891750336,
"rewards/forensics_tempsamp_iou_reward": 0.5981525778770447,
"step": 739
},
{
"completion_length": 23.9375,
"epoch": 3.096234309623431,
"grad_norm": 3.285170555114746,
"kl": 0.3095703125,
"learning_rate": 2.2698744769874474e-07,
"loss": 0.0124,
"reward": 0.6761277616024017,
"reward_std": 0.091676976531744,
"reward_std_all": 0.10454169288277626,
"rewards-max/forensics_tempsamp_iou_reward": 0.9796631038188934,
"rewards/forensics_tempsamp_iou_reward": 0.6792941093444824,
"step": 740
},
{
"completion_length": 23.171875,
"epoch": 3.100418410041841,
"grad_norm": 3.0277533531188965,
"kl": 0.33984375,
"learning_rate": 2.2594142259414225e-07,
"loss": 0.0136,
"reward": 0.6272063255310059,
"reward_std": 0.12404296919703484,
"reward_std_all": 0.14134949073195457,
"rewards-max/forensics_tempsamp_iou_reward": 0.9653592109680176,
"rewards/forensics_tempsamp_iou_reward": 0.6082949042320251,
"step": 741
},
{
"completion_length": 23.40625,
"epoch": 3.104602510460251,
"grad_norm": 3.4852957725524902,
"kl": 0.3076171875,
"learning_rate": 2.2489539748953974e-07,
"loss": 0.0123,
"reward": 0.5783243477344513,
"reward_std": 0.10470348969101906,
"reward_std_all": 0.1531965583562851,
"rewards-max/forensics_tempsamp_iou_reward": 0.925396054983139,
"rewards/forensics_tempsamp_iou_reward": 0.5537781119346619,
"step": 742
},
{
"completion_length": 30.234375,
"epoch": 3.1087866108786613,
"grad_norm": 3.4999167919158936,
"kl": 0.30615234375,
"learning_rate": 2.2384937238493723e-07,
"loss": 0.0122,
"reward": 0.5420901030302048,
"reward_std": 0.0940316841006279,
"reward_std_all": 0.16386041417717934,
"rewards-max/forensics_tempsamp_iou_reward": 0.9354508519172668,
"rewards/forensics_tempsamp_iou_reward": 0.48391857743263245,
"step": 743
},
{
"completion_length": 22.328125,
"epoch": 3.1129707112970713,
"grad_norm": 4.006590366363525,
"kl": 0.322265625,
"learning_rate": 2.2280334728033474e-07,
"loss": 0.0129,
"reward": 0.5581739544868469,
"reward_std": 0.1575053296983242,
"reward_std_all": 0.18496271967887878,
"rewards-max/forensics_tempsamp_iou_reward": 0.901408463716507,
"rewards/forensics_tempsamp_iou_reward": 0.5090890824794769,
"step": 744
},
{
"completion_length": 20.546875,
"epoch": 3.1171548117154813,
"grad_norm": 5.6236677169799805,
"kl": 0.494140625,
"learning_rate": 2.217573221757322e-07,
"loss": 0.0198,
"reward": 0.5990922451019287,
"reward_std": 0.0981099121272564,
"reward_std_all": 0.14326899498701096,
"rewards-max/forensics_tempsamp_iou_reward": 0.9886930584907532,
"rewards/forensics_tempsamp_iou_reward": 0.6033336520195007,
"step": 745
},
{
"completion_length": 22.71875,
"epoch": 3.1213389121338913,
"grad_norm": 3.952052116394043,
"kl": 0.3115234375,
"learning_rate": 2.207112970711297e-07,
"loss": 0.0125,
"reward": 0.5807205438613892,
"reward_std": 0.16175507754087448,
"reward_std_all": 0.179731123149395,
"rewards-max/forensics_tempsamp_iou_reward": 0.9304267168045044,
"rewards/forensics_tempsamp_iou_reward": 0.5332713425159454,
"step": 746
},
{
"completion_length": 24.4375,
"epoch": 3.1255230125523012,
"grad_norm": 3.351746082305908,
"kl": 0.30126953125,
"learning_rate": 2.1966527196652717e-07,
"loss": 0.0121,
"reward": 0.5960185825824738,
"reward_std": 0.09022612124681473,
"reward_std_all": 0.1363883912563324,
"rewards-max/forensics_tempsamp_iou_reward": 0.95392245054245,
"rewards/forensics_tempsamp_iou_reward": 0.5697579979896545,
"step": 747
},
{
"completion_length": 23.0625,
"epoch": 3.1297071129707112,
"grad_norm": 5.303876876831055,
"kl": 0.4013671875,
"learning_rate": 2.1861924686192469e-07,
"loss": 0.0161,
"reward": 0.606615275144577,
"reward_std": 0.13077545166015625,
"reward_std_all": 0.15798992663621902,
"rewards-max/forensics_tempsamp_iou_reward": 0.9732144773006439,
"rewards/forensics_tempsamp_iou_reward": 0.5911003351211548,
"step": 748
},
{
"completion_length": 25.1875,
"epoch": 3.1338912133891212,
"grad_norm": 3.562755823135376,
"kl": 0.2919921875,
"learning_rate": 2.1757322175732217e-07,
"loss": 0.0117,
"reward": 0.6336377859115601,
"reward_std": 0.11412113532423973,
"reward_std_all": 0.13134998828172684,
"rewards-max/forensics_tempsamp_iou_reward": 0.9415683746337891,
"rewards/forensics_tempsamp_iou_reward": 0.5961367189884186,
"step": 749
},
{
"completion_length": 24.734375,
"epoch": 3.1380753138075312,
"grad_norm": 3.8136088848114014,
"kl": 0.302734375,
"learning_rate": 2.1652719665271966e-07,
"loss": 0.0121,
"reward": 0.5912362337112427,
"reward_std": 0.13315610587596893,
"reward_std_all": 0.16550002619624138,
"rewards-max/forensics_tempsamp_iou_reward": 0.9898410439491272,
"rewards/forensics_tempsamp_iou_reward": 0.5684628933668137,
"step": 750
},
{
"completion_length": 20.625,
"epoch": 3.1422594142259412,
"grad_norm": 3.5951905250549316,
"kl": 0.337890625,
"learning_rate": 2.1548117154811714e-07,
"loss": 0.0135,
"reward": 0.7135409116744995,
"reward_std": 0.04869871027767658,
"reward_std_all": 0.0686940960586071,
"rewards-max/forensics_tempsamp_iou_reward": 0.9651147425174713,
"rewards/forensics_tempsamp_iou_reward": 0.7279137372970581,
"step": 751
},
{
"completion_length": 19.890625,
"epoch": 3.1464435146443517,
"grad_norm": 3.5075230598449707,
"kl": 0.3173828125,
"learning_rate": 2.1443514644351463e-07,
"loss": 0.0127,
"reward": 0.6222443282604218,
"reward_std": 0.10181065276265144,
"reward_std_all": 0.13676784932613373,
"rewards-max/forensics_tempsamp_iou_reward": 0.960991770029068,
"rewards/forensics_tempsamp_iou_reward": 0.6066408455371857,
"step": 752
},
{
"completion_length": 21.8125,
"epoch": 3.1506276150627617,
"grad_norm": 3.4528603553771973,
"kl": 0.291015625,
"learning_rate": 2.1338912133891212e-07,
"loss": 0.0116,
"reward": 0.7110661566257477,
"reward_std": 0.05064530484378338,
"reward_std_all": 0.06611946783959866,
"rewards-max/forensics_tempsamp_iou_reward": 0.9645794034004211,
"rewards/forensics_tempsamp_iou_reward": 0.7195700109004974,
"step": 753
},
{
"completion_length": 22.15625,
"epoch": 3.1548117154811717,
"grad_norm": 3.4200613498687744,
"kl": 0.255859375,
"learning_rate": 2.1234309623430963e-07,
"loss": 0.0102,
"reward": 0.6254252195358276,
"reward_std": 0.06749210692942142,
"reward_std_all": 0.11069216579198837,
"rewards-max/forensics_tempsamp_iou_reward": 0.9870235025882721,
"rewards/forensics_tempsamp_iou_reward": 0.608613133430481,
"step": 754
},
{
"completion_length": 20.609375,
"epoch": 3.1589958158995817,
"grad_norm": 14.072388648986816,
"kl": 1.0556640625,
"learning_rate": 2.112970711297071e-07,
"loss": 0.0424,
"reward": 0.6118759214878082,
"reward_std": 0.12075775116682053,
"reward_std_all": 0.1472245454788208,
"rewards-max/forensics_tempsamp_iou_reward": 0.9913006722927094,
"rewards/forensics_tempsamp_iou_reward": 0.5987991988658905,
"step": 755
},
{
"completion_length": 22.625,
"epoch": 3.1631799163179917,
"grad_norm": 4.017125606536865,
"kl": 0.34765625,
"learning_rate": 2.102510460251046e-07,
"loss": 0.0139,
"reward": 0.5698726773262024,
"reward_std": 0.15303561091423035,
"reward_std_all": 0.17580442130565643,
"rewards-max/forensics_tempsamp_iou_reward": 0.8927938938140869,
"rewards/forensics_tempsamp_iou_reward": 0.5066039711236954,
"step": 756
},
{
"completion_length": 22.671875,
"epoch": 3.1673640167364017,
"grad_norm": 3.4704678058624268,
"kl": 0.3017578125,
"learning_rate": 2.0920502092050206e-07,
"loss": 0.0121,
"reward": 0.6488735377788544,
"reward_std": 0.0856090821325779,
"reward_std_all": 0.11062776669859886,
"rewards-max/forensics_tempsamp_iou_reward": 0.9533072710037231,
"rewards/forensics_tempsamp_iou_reward": 0.6325498819351196,
"step": 757
},
{
"completion_length": 22.390625,
"epoch": 3.1715481171548117,
"grad_norm": 4.06320858001709,
"kl": 0.3828125,
"learning_rate": 2.0815899581589958e-07,
"loss": 0.0153,
"reward": 0.6143332719802856,
"reward_std": 0.14523445069789886,
"reward_std_all": 0.15815269947052002,
"rewards-max/forensics_tempsamp_iou_reward": 0.9526066482067108,
"rewards/forensics_tempsamp_iou_reward": 0.5966981649398804,
"step": 758
},
{
"completion_length": 22.375,
"epoch": 3.1757322175732217,
"grad_norm": 3.7127935886383057,
"kl": 0.373046875,
"learning_rate": 2.0711297071129706e-07,
"loss": 0.015,
"reward": 0.5580626428127289,
"reward_std": 0.08382205292582512,
"reward_std_all": 0.15066587179899216,
"rewards-max/forensics_tempsamp_iou_reward": 0.9163011908531189,
"rewards/forensics_tempsamp_iou_reward": 0.49272435903549194,
"step": 759
},
{
"completion_length": 24.28125,
"epoch": 3.1799163179916317,
"grad_norm": 3.87017560005188,
"kl": 0.3115234375,
"learning_rate": 2.0606694560669455e-07,
"loss": 0.0125,
"reward": 0.5540201365947723,
"reward_std": 0.10194896906614304,
"reward_std_all": 0.1577274054288864,
"rewards-max/forensics_tempsamp_iou_reward": 0.9542191922664642,
"rewards/forensics_tempsamp_iou_reward": 0.5118034482002258,
"step": 760
},
{
"completion_length": 20.984375,
"epoch": 3.1841004184100417,
"grad_norm": 3.4642255306243896,
"kl": 0.3486328125,
"learning_rate": 2.0502092050209206e-07,
"loss": 0.0139,
"reward": 0.6411330997943878,
"reward_std": 0.09847749024629593,
"reward_std_all": 0.12355808913707733,
"rewards-max/forensics_tempsamp_iou_reward": 0.9736931622028351,
"rewards/forensics_tempsamp_iou_reward": 0.6093690097332001,
"step": 761
},
{
"completion_length": 22.5,
"epoch": 3.1882845188284517,
"grad_norm": 3.1608047485351562,
"kl": 0.2861328125,
"learning_rate": 2.0397489539748952e-07,
"loss": 0.0114,
"reward": 0.6543476581573486,
"reward_std": 0.04467239137738943,
"reward_std_all": 0.09355160593986511,
"rewards-max/forensics_tempsamp_iou_reward": 0.9724310636520386,
"rewards/forensics_tempsamp_iou_reward": 0.6791562438011169,
"step": 762
},
{
"completion_length": 20.28125,
"epoch": 3.192468619246862,
"grad_norm": 4.536342144012451,
"kl": 0.3486328125,
"learning_rate": 2.0292887029288703e-07,
"loss": 0.0139,
"reward": 0.6270537078380585,
"reward_std": 0.10918480902910233,
"reward_std_all": 0.13200626149773598,
"rewards-max/forensics_tempsamp_iou_reward": 0.9463223218917847,
"rewards/forensics_tempsamp_iou_reward": 0.5994368195533752,
"step": 763
},
{
"completion_length": 20.78125,
"epoch": 3.196652719665272,
"grad_norm": 4.7726640701293945,
"kl": 0.28271484375,
"learning_rate": 2.0188284518828452e-07,
"loss": 0.0113,
"reward": 0.6481688320636749,
"reward_std": 0.05229376442730427,
"reward_std_all": 0.09901183098554611,
"rewards-max/forensics_tempsamp_iou_reward": 0.9744581282138824,
"rewards/forensics_tempsamp_iou_reward": 0.6329520344734192,
"step": 764
},
{
"completion_length": 20.65625,
"epoch": 3.200836820083682,
"grad_norm": 17.79173469543457,
"kl": 1.1015625,
"learning_rate": 2.00836820083682e-07,
"loss": 0.0442,
"reward": 0.6486473679542542,
"reward_std": 0.12094975262880325,
"reward_std_all": 0.13539263606071472,
"rewards-max/forensics_tempsamp_iou_reward": 0.9498660862445831,
"rewards/forensics_tempsamp_iou_reward": 0.6194221079349518,
"step": 765
},
{
"completion_length": 25.203125,
"epoch": 3.205020920502092,
"grad_norm": 3.20879864692688,
"kl": 0.3408203125,
"learning_rate": 1.997907949790795e-07,
"loss": 0.0136,
"reward": 0.5690047442913055,
"reward_std": 0.1276581510901451,
"reward_std_all": 0.16508369147777557,
"rewards-max/forensics_tempsamp_iou_reward": 0.9478484690189362,
"rewards/forensics_tempsamp_iou_reward": 0.5124613642692566,
"step": 766
},
{
"completion_length": 22.1875,
"epoch": 3.209205020920502,
"grad_norm": 4.168340682983398,
"kl": 0.3974609375,
"learning_rate": 1.9874476987447698e-07,
"loss": 0.0159,
"reward": 0.6168730854988098,
"reward_std": 0.08977312035858631,
"reward_std_all": 0.1315452679991722,
"rewards-max/forensics_tempsamp_iou_reward": 0.9722222089767456,
"rewards/forensics_tempsamp_iou_reward": 0.5887182056903839,
"step": 767
},
{
"completion_length": 20.296875,
"epoch": 3.213389121338912,
"grad_norm": 4.305205345153809,
"kl": 0.33984375,
"learning_rate": 1.9769874476987447e-07,
"loss": 0.0136,
"reward": 0.6404357254505157,
"reward_std": 0.12486452609300613,
"reward_std_all": 0.13539637252688408,
"rewards-max/forensics_tempsamp_iou_reward": 0.9754133522510529,
"rewards/forensics_tempsamp_iou_reward": 0.6196500658988953,
"step": 768
},
{
"completion_length": 20.640625,
"epoch": 3.217573221757322,
"grad_norm": 4.005370616912842,
"kl": 0.3603515625,
"learning_rate": 1.9665271966527198e-07,
"loss": 0.0144,
"reward": 0.6412510871887207,
"reward_std": 0.09808545559644699,
"reward_std_all": 0.11955130845308304,
"rewards-max/forensics_tempsamp_iou_reward": 0.9634779393672943,
"rewards/forensics_tempsamp_iou_reward": 0.6011396646499634,
"step": 769
},
{
"completion_length": 23.359375,
"epoch": 3.221757322175732,
"grad_norm": 3.715865135192871,
"kl": 0.28125,
"learning_rate": 1.9560669456066944e-07,
"loss": 0.0112,
"reward": 0.6583346724510193,
"reward_std": 0.12643462792038918,
"reward_std_all": 0.13130265101790428,
"rewards-max/forensics_tempsamp_iou_reward": 0.9704283177852631,
"rewards/forensics_tempsamp_iou_reward": 0.6655941009521484,
"step": 770
},
{
"completion_length": 20.359375,
"epoch": 3.225941422594142,
"grad_norm": 3.992439031600952,
"kl": 0.2509765625,
"learning_rate": 1.9456066945606695e-07,
"loss": 0.0101,
"reward": 0.6155928671360016,
"reward_std": 0.09275078028440475,
"reward_std_all": 0.1311618909239769,
"rewards-max/forensics_tempsamp_iou_reward": 0.9920135736465454,
"rewards/forensics_tempsamp_iou_reward": 0.6085658669471741,
"step": 771
},
{
"completion_length": 22.421875,
"epoch": 3.2301255230125525,
"grad_norm": 3.2192718982696533,
"kl": 0.3125,
"learning_rate": 1.935146443514644e-07,
"loss": 0.0126,
"reward": 0.6770308017730713,
"reward_std": 0.08688431978225708,
"reward_std_all": 0.100521270185709,
"rewards-max/forensics_tempsamp_iou_reward": 0.9648048281669617,
"rewards/forensics_tempsamp_iou_reward": 0.6705867946147919,
"step": 772
},
{
"completion_length": 21.828125,
"epoch": 3.2343096234309625,
"grad_norm": 3.528331756591797,
"kl": 0.31640625,
"learning_rate": 1.9246861924686192e-07,
"loss": 0.0127,
"reward": 0.6334876716136932,
"reward_std": 0.08350181765854359,
"reward_std_all": 0.12671619653701782,
"rewards-max/forensics_tempsamp_iou_reward": 0.9776051044464111,
"rewards/forensics_tempsamp_iou_reward": 0.6204764246940613,
"step": 773
},
{
"completion_length": 21.75,
"epoch": 3.2384937238493725,
"grad_norm": 3.9489474296569824,
"kl": 0.306640625,
"learning_rate": 1.914225941422594e-07,
"loss": 0.0123,
"reward": 0.6717178225517273,
"reward_std": 0.07564113289117813,
"reward_std_all": 0.09470974281430244,
"rewards-max/forensics_tempsamp_iou_reward": 0.9671399295330048,
"rewards/forensics_tempsamp_iou_reward": 0.666453629732132,
"step": 774
},
{
"completion_length": 21.109375,
"epoch": 3.2426778242677825,
"grad_norm": 4.061252117156982,
"kl": 0.375,
"learning_rate": 1.903765690376569e-07,
"loss": 0.015,
"reward": 0.6251223087310791,
"reward_std": 0.0969926118850708,
"reward_std_all": 0.12808020412921906,
"rewards-max/forensics_tempsamp_iou_reward": 0.9534385502338409,
"rewards/forensics_tempsamp_iou_reward": 0.6138964742422104,
"step": 775
},
{
"completion_length": 20.921875,
"epoch": 3.2468619246861925,
"grad_norm": 3.9319894313812256,
"kl": 0.3525390625,
"learning_rate": 1.8933054393305438e-07,
"loss": 0.0141,
"reward": 0.6499039828777313,
"reward_std": 0.0961885005235672,
"reward_std_all": 0.1179603710770607,
"rewards-max/forensics_tempsamp_iou_reward": 0.9878613650798798,
"rewards/forensics_tempsamp_iou_reward": 0.6486380994319916,
"step": 776
},
{
"completion_length": 20.46875,
"epoch": 3.2510460251046025,
"grad_norm": 4.164841651916504,
"kl": 0.23779296875,
"learning_rate": 1.8828451882845187e-07,
"loss": 0.0095,
"reward": 0.6653175055980682,
"reward_std": 0.08478635549545288,
"reward_std_all": 0.10747140645980835,
"rewards-max/forensics_tempsamp_iou_reward": 0.9343592822551727,
"rewards/forensics_tempsamp_iou_reward": 0.6388111710548401,
"step": 777
},
{
"completion_length": 22.234375,
"epoch": 3.2552301255230125,
"grad_norm": 3.645949125289917,
"kl": 0.314453125,
"learning_rate": 1.8723849372384936e-07,
"loss": 0.0125,
"reward": 0.6394545435905457,
"reward_std": 0.1274275779724121,
"reward_std_all": 0.14135175198316574,
"rewards-max/forensics_tempsamp_iou_reward": 0.9423768520355225,
"rewards/forensics_tempsamp_iou_reward": 0.6121837198734283,
"step": 778
},
{
"completion_length": 22.796875,
"epoch": 3.2594142259414225,
"grad_norm": 4.148622989654541,
"kl": 0.384765625,
"learning_rate": 1.8619246861924684e-07,
"loss": 0.0154,
"reward": 0.6289184987545013,
"reward_std": 0.12398971244692802,
"reward_std_all": 0.14005599170923233,
"rewards-max/forensics_tempsamp_iou_reward": 0.9732567071914673,
"rewards/forensics_tempsamp_iou_reward": 0.6044072210788727,
"step": 779
},
{
"completion_length": 22.4375,
"epoch": 3.2635983263598325,
"grad_norm": 3.8817296028137207,
"kl": 0.34765625,
"learning_rate": 1.8514644351464433e-07,
"loss": 0.0139,
"reward": 0.7424093186855316,
"reward_std": 0.06755030155181885,
"reward_std_all": 0.06357282679527998,
"rewards-max/forensics_tempsamp_iou_reward": 0.9896551072597504,
"rewards/forensics_tempsamp_iou_reward": 0.7820667624473572,
"step": 780
},
{
"completion_length": 21.0625,
"epoch": 3.2677824267782425,
"grad_norm": 3.6882715225219727,
"kl": 0.4462890625,
"learning_rate": 1.8410041841004184e-07,
"loss": 0.0179,
"reward": 0.6733382940292358,
"reward_std": 0.08647704869508743,
"reward_std_all": 0.10742238536477089,
"rewards-max/forensics_tempsamp_iou_reward": 0.9706634879112244,
"rewards/forensics_tempsamp_iou_reward": 0.6868650615215302,
"step": 781
},
{
"completion_length": 19.546875,
"epoch": 3.2719665271966525,
"grad_norm": 4.044732570648193,
"kl": 0.2783203125,
"learning_rate": 1.830543933054393e-07,
"loss": 0.0111,
"reward": 0.6354348361492157,
"reward_std": 0.12276003137230873,
"reward_std_all": 0.13607020676136017,
"rewards-max/forensics_tempsamp_iou_reward": 0.9508273899555206,
"rewards/forensics_tempsamp_iou_reward": 0.6101344227790833,
"step": 782
},
{
"completion_length": 19.96875,
"epoch": 3.276150627615063,
"grad_norm": 4.052115440368652,
"kl": 0.34375,
"learning_rate": 1.8200836820083681e-07,
"loss": 0.0137,
"reward": 0.5606111884117126,
"reward_std": 0.14057610929012299,
"reward_std_all": 0.18063056468963623,
"rewards-max/forensics_tempsamp_iou_reward": 0.9417475759983063,
"rewards/forensics_tempsamp_iou_reward": 0.5283100455999374,
"step": 783
},
{
"completion_length": 23.421875,
"epoch": 3.280334728033473,
"grad_norm": 4.058852195739746,
"kl": 0.376953125,
"learning_rate": 1.8096234309623427e-07,
"loss": 0.0151,
"reward": 0.5622908473014832,
"reward_std": 0.12172862887382507,
"reward_std_all": 0.16132927685976028,
"rewards-max/forensics_tempsamp_iou_reward": 0.9461227059364319,
"rewards/forensics_tempsamp_iou_reward": 0.5036291927099228,
"step": 784
},
{
"completion_length": 21.625,
"epoch": 3.284518828451883,
"grad_norm": 3.987161874771118,
"kl": 0.3310546875,
"learning_rate": 1.799163179916318e-07,
"loss": 0.0133,
"reward": 0.5973768830299377,
"reward_std": 0.09175760298967361,
"reward_std_all": 0.13659456372261047,
"rewards-max/forensics_tempsamp_iou_reward": 0.9675747156143188,
"rewards/forensics_tempsamp_iou_reward": 0.5544911026954651,
"step": 785
},
{
"completion_length": 23.65625,
"epoch": 3.288702928870293,
"grad_norm": 4.253826141357422,
"kl": 0.3515625,
"learning_rate": 1.788702928870293e-07,
"loss": 0.0141,
"reward": 0.6378292143344879,
"reward_std": 0.0876577328890562,
"reward_std_all": 0.11786075681447983,
"rewards-max/forensics_tempsamp_iou_reward": 0.9351857304573059,
"rewards/forensics_tempsamp_iou_reward": 0.5959987342357635,
"step": 786
},
{
"completion_length": 21.34375,
"epoch": 3.292887029288703,
"grad_norm": 4.179370880126953,
"kl": 0.552734375,
"learning_rate": 1.7782426778242676e-07,
"loss": 0.0221,
"reward": 0.6464307308197021,
"reward_std": 0.13747487217187881,
"reward_std_all": 0.1413891613483429,
"rewards-max/forensics_tempsamp_iou_reward": 0.9812059998512268,
"rewards/forensics_tempsamp_iou_reward": 0.6536993980407715,
"step": 787
},
{
"completion_length": 23.359375,
"epoch": 3.297071129707113,
"grad_norm": 4.240195274353027,
"kl": 0.375,
"learning_rate": 1.7677824267782427e-07,
"loss": 0.015,
"reward": 0.6370860040187836,
"reward_std": 0.07600258849561214,
"reward_std_all": 0.11318891495466232,
"rewards-max/forensics_tempsamp_iou_reward": 0.984773725271225,
"rewards/forensics_tempsamp_iou_reward": 0.615294873714447,
"step": 788
},
{
"completion_length": 20.90625,
"epoch": 3.301255230125523,
"grad_norm": 4.230740547180176,
"kl": 0.33984375,
"learning_rate": 1.7573221757322173e-07,
"loss": 0.0136,
"reward": 0.6768605709075928,
"reward_std": 0.08403268828988075,
"reward_std_all": 0.09663784876465797,
"rewards-max/forensics_tempsamp_iou_reward": 0.9410400688648224,
"rewards/forensics_tempsamp_iou_reward": 0.643295407295227,
"step": 789
},
{
"completion_length": 21.1875,
"epoch": 3.305439330543933,
"grad_norm": 3.945233106613159,
"kl": 0.4013671875,
"learning_rate": 1.7468619246861925e-07,
"loss": 0.016,
"reward": 0.6974157989025116,
"reward_std": 0.05494868755340576,
"reward_std_all": 0.07585224509239197,
"rewards-max/forensics_tempsamp_iou_reward": 0.9610726535320282,
"rewards/forensics_tempsamp_iou_reward": 0.704634040594101,
"step": 790
},
{
"completion_length": 24.375,
"epoch": 3.309623430962343,
"grad_norm": 3.6892590522766113,
"kl": 0.318359375,
"learning_rate": 1.7364016736401673e-07,
"loss": 0.0127,
"reward": 0.6716443002223969,
"reward_std": 0.07675696164369583,
"reward_std_all": 0.09705029055476189,
"rewards-max/forensics_tempsamp_iou_reward": 0.9505298435688019,
"rewards/forensics_tempsamp_iou_reward": 0.6622949540615082,
"step": 791
},
{
"completion_length": 20.890625,
"epoch": 3.3138075313807533,
"grad_norm": 3.525341510772705,
"kl": 0.3115234375,
"learning_rate": 1.7259414225941422e-07,
"loss": 0.0125,
"reward": 0.6552841067314148,
"reward_std": 0.06523806229233742,
"reward_std_all": 0.09872150793671608,
"rewards-max/forensics_tempsamp_iou_reward": 0.9681074619293213,
"rewards/forensics_tempsamp_iou_reward": 0.6392230093479156,
"step": 792
},
{
"completion_length": 21.328125,
"epoch": 3.3179916317991633,
"grad_norm": 4.299458980560303,
"kl": 0.380859375,
"learning_rate": 1.715481171548117e-07,
"loss": 0.0152,
"reward": 0.6417192816734314,
"reward_std": 0.10312698036432266,
"reward_std_all": 0.12292610481381416,
"rewards-max/forensics_tempsamp_iou_reward": 0.9840085208415985,
"rewards/forensics_tempsamp_iou_reward": 0.6345581114292145,
"step": 793
},
{
"completion_length": 22.875,
"epoch": 3.3221757322175733,
"grad_norm": 4.441270351409912,
"kl": 0.31640625,
"learning_rate": 1.705020920502092e-07,
"loss": 0.0127,
"reward": 0.6351883411407471,
"reward_std": 0.11452756449580193,
"reward_std_all": 0.13075106218457222,
"rewards-max/forensics_tempsamp_iou_reward": 0.9608781635761261,
"rewards/forensics_tempsamp_iou_reward": 0.6078509986400604,
"step": 794
},
{
"completion_length": 23.25,
"epoch": 3.3263598326359833,
"grad_norm": 3.254671812057495,
"kl": 0.2666015625,
"learning_rate": 1.6945606694560668e-07,
"loss": 0.0107,
"reward": 0.6661116480827332,
"reward_std": 0.06928987987339497,
"reward_std_all": 0.09942147508263588,
"rewards-max/forensics_tempsamp_iou_reward": 0.9748062193393707,
"rewards/forensics_tempsamp_iou_reward": 0.6512874960899353,
"step": 795
},
{
"completion_length": 19.78125,
"epoch": 3.3305439330543933,
"grad_norm": 5.331660270690918,
"kl": 0.3740234375,
"learning_rate": 1.684100418410042e-07,
"loss": 0.0149,
"reward": 0.6273001134395599,
"reward_std": 0.08264034241437912,
"reward_std_all": 0.11942427977919579,
"rewards-max/forensics_tempsamp_iou_reward": 0.9645191431045532,
"rewards/forensics_tempsamp_iou_reward": 0.6234716475009918,
"step": 796
},
{
"completion_length": 22.890625,
"epoch": 3.3347280334728033,
"grad_norm": 4.19597053527832,
"kl": 0.2978515625,
"learning_rate": 1.6736401673640165e-07,
"loss": 0.0119,
"reward": 0.6242906749248505,
"reward_std": 0.13337792828679085,
"reward_std_all": 0.14776550233364105,
"rewards-max/forensics_tempsamp_iou_reward": 0.9691494405269623,
"rewards/forensics_tempsamp_iou_reward": 0.6077776849269867,
"step": 797
},
{
"completion_length": 20.25,
"epoch": 3.3389121338912133,
"grad_norm": 3.9296882152557373,
"kl": 0.3466796875,
"learning_rate": 1.6631799163179916e-07,
"loss": 0.0139,
"reward": 0.6987778544425964,
"reward_std": 0.06789674982428551,
"reward_std_all": 0.07885409891605377,
"rewards-max/forensics_tempsamp_iou_reward": 0.9944781959056854,
"rewards/forensics_tempsamp_iou_reward": 0.7179518342018127,
"step": 798
},
{
"completion_length": 21.421875,
"epoch": 3.3430962343096233,
"grad_norm": 3.9196512699127197,
"kl": 0.33203125,
"learning_rate": 1.6527196652719662e-07,
"loss": 0.0133,
"reward": 0.6919412016868591,
"reward_std": 0.09147605672478676,
"reward_std_all": 0.0969744324684143,
"rewards-max/forensics_tempsamp_iou_reward": 0.9598599076271057,
"rewards/forensics_tempsamp_iou_reward": 0.6896673738956451,
"step": 799
},
{
"completion_length": 21.921875,
"epoch": 3.3472803347280333,
"grad_norm": 3.90858793258667,
"kl": 0.3193359375,
"learning_rate": 1.6422594142259414e-07,
"loss": 0.0128,
"reward": 0.6605148613452911,
"reward_std": 0.08226402848958969,
"reward_std_all": 0.09989386424422264,
"rewards-max/forensics_tempsamp_iou_reward": 0.9685268700122833,
"rewards/forensics_tempsamp_iou_reward": 0.6443490386009216,
"step": 800
},
{
"completion_length": 21.46875,
"epoch": 3.3514644351464433,
"grad_norm": 3.6850767135620117,
"kl": 0.2255859375,
"learning_rate": 1.6317991631799162e-07,
"loss": 0.009,
"reward": 0.632679283618927,
"reward_std": 0.05364008620381355,
"reward_std_all": 0.10270534455776215,
"rewards-max/forensics_tempsamp_iou_reward": 0.9508960843086243,
"rewards/forensics_tempsamp_iou_reward": 0.6086204648017883,
"step": 801
},
{
"completion_length": 23.796875,
"epoch": 3.3556485355648533,
"grad_norm": 3.467078447341919,
"kl": 0.263671875,
"learning_rate": 1.621338912133891e-07,
"loss": 0.0105,
"reward": 0.6170517802238464,
"reward_std": 0.13563894480466843,
"reward_std_all": 0.15156283229589462,
"rewards-max/forensics_tempsamp_iou_reward": 0.9765863120555878,
"rewards/forensics_tempsamp_iou_reward": 0.5890210866928101,
"step": 802
},
{
"completion_length": 20.171875,
"epoch": 3.3598326359832638,
"grad_norm": 4.613120079040527,
"kl": 0.2998046875,
"learning_rate": 1.610878661087866e-07,
"loss": 0.012,
"reward": 0.635022908449173,
"reward_std": 0.09811140224337578,
"reward_std_all": 0.12472343817353249,
"rewards-max/forensics_tempsamp_iou_reward": 0.9871906042098999,
"rewards/forensics_tempsamp_iou_reward": 0.6356608867645264,
"step": 803
},
{
"completion_length": 22.125,
"epoch": 3.3640167364016738,
"grad_norm": 3.5271105766296387,
"kl": 0.3017578125,
"learning_rate": 1.6004184100418408e-07,
"loss": 0.0121,
"reward": 0.6615295708179474,
"reward_std": 0.0921594388782978,
"reward_std_all": 0.10868504643440247,
"rewards-max/forensics_tempsamp_iou_reward": 0.9528551697731018,
"rewards/forensics_tempsamp_iou_reward": 0.642003208398819,
"step": 804
},
{
"completion_length": 21.734375,
"epoch": 3.3682008368200838,
"grad_norm": 3.80499529838562,
"kl": 0.3779296875,
"learning_rate": 1.5899581589958157e-07,
"loss": 0.0151,
"reward": 0.6159859001636505,
"reward_std": 0.1465453952550888,
"reward_std_all": 0.15863753110170364,
"rewards-max/forensics_tempsamp_iou_reward": 0.9658940136432648,
"rewards/forensics_tempsamp_iou_reward": 0.5930553376674652,
"step": 805
},
{
"completion_length": 20.296875,
"epoch": 3.3723849372384938,
"grad_norm": 3.9131264686584473,
"kl": 0.34375,
"learning_rate": 1.5794979079497908e-07,
"loss": 0.0137,
"reward": 0.72427698969841,
"reward_std": 0.055761683732271194,
"reward_std_all": 0.0641359444707632,
"rewards-max/forensics_tempsamp_iou_reward": 0.9391067624092102,
"rewards/forensics_tempsamp_iou_reward": 0.7392351627349854,
"step": 806
},
{
"completion_length": 21.390625,
"epoch": 3.3765690376569037,
"grad_norm": 3.4950249195098877,
"kl": 0.3447265625,
"learning_rate": 1.5690376569037657e-07,
"loss": 0.0138,
"reward": 0.6337325572967529,
"reward_std": 0.16656769812107086,
"reward_std_all": 0.1620955914258957,
"rewards-max/forensics_tempsamp_iou_reward": 0.9535194039344788,
"rewards/forensics_tempsamp_iou_reward": 0.6259340345859528,
"step": 807
},
{
"completion_length": 22.25,
"epoch": 3.3807531380753137,
"grad_norm": 3.893449544906616,
"kl": 0.23828125,
"learning_rate": 1.5585774058577405e-07,
"loss": 0.0095,
"reward": 0.6049911677837372,
"reward_std": 0.10281386971473694,
"reward_std_all": 0.14086105674505234,
"rewards-max/forensics_tempsamp_iou_reward": 0.9741097688674927,
"rewards/forensics_tempsamp_iou_reward": 0.5747244656085968,
"step": 808
},
{
"completion_length": 20.3125,
"epoch": 3.3849372384937237,
"grad_norm": 3.4609315395355225,
"kl": 0.3896484375,
"learning_rate": 1.5481171548117154e-07,
"loss": 0.0156,
"reward": 0.7296910285949707,
"reward_std": 0.061876729130744934,
"reward_std_all": 0.06820149347186089,
"rewards-max/forensics_tempsamp_iou_reward": 0.985884815454483,
"rewards/forensics_tempsamp_iou_reward": 0.7738637030124664,
"step": 809
},
{
"completion_length": 21.125,
"epoch": 3.3891213389121337,
"grad_norm": 5.362493991851807,
"kl": 0.337890625,
"learning_rate": 1.5376569037656903e-07,
"loss": 0.0135,
"reward": 0.6523925065994263,
"reward_std": 0.08768301084637642,
"reward_std_all": 0.11435410380363464,
"rewards-max/forensics_tempsamp_iou_reward": 0.9341956973075867,
"rewards/forensics_tempsamp_iou_reward": 0.620883047580719,
"step": 810
},
{
"completion_length": 22.15625,
"epoch": 3.393305439330544,
"grad_norm": 3.0813992023468018,
"kl": 0.349609375,
"learning_rate": 1.5271966527196654e-07,
"loss": 0.014,
"reward": 0.6113428175449371,
"reward_std": 0.1392645463347435,
"reward_std_all": 0.15578829497098923,
"rewards-max/forensics_tempsamp_iou_reward": 0.9592002630233765,
"rewards/forensics_tempsamp_iou_reward": 0.5753114223480225,
"step": 811
},
{
"completion_length": 23.609375,
"epoch": 3.397489539748954,
"grad_norm": 3.9397873878479004,
"kl": 0.42578125,
"learning_rate": 1.51673640167364e-07,
"loss": 0.017,
"reward": 0.6355010271072388,
"reward_std": 0.10102912038564682,
"reward_std_all": 0.1227574348449707,
"rewards-max/forensics_tempsamp_iou_reward": 0.9623938798904419,
"rewards/forensics_tempsamp_iou_reward": 0.6059442460536957,
"step": 812
},
{
"completion_length": 33.765625,
"epoch": 3.401673640167364,
"grad_norm": 3.4341578483581543,
"kl": 0.3056640625,
"learning_rate": 1.506276150627615e-07,
"loss": 0.0122,
"reward": 0.5277880877256393,
"reward_std": 0.08638235181570053,
"reward_std_all": 0.1697065755724907,
"rewards-max/forensics_tempsamp_iou_reward": 0.9399747848510742,
"rewards/forensics_tempsamp_iou_reward": 0.4961450695991516,
"step": 813
},
{
"completion_length": 23.53125,
"epoch": 3.405857740585774,
"grad_norm": 3.7504491806030273,
"kl": 0.3427734375,
"learning_rate": 1.4958158995815897e-07,
"loss": 0.0137,
"reward": 0.6580074429512024,
"reward_std": 0.07495803013443947,
"reward_std_all": 0.10404791682958603,
"rewards-max/forensics_tempsamp_iou_reward": 0.9754177927970886,
"rewards/forensics_tempsamp_iou_reward": 0.660801112651825,
"step": 814
},
{
"completion_length": 23.546875,
"epoch": 3.410041841004184,
"grad_norm": 4.124816417694092,
"kl": 0.2744140625,
"learning_rate": 1.4853556485355648e-07,
"loss": 0.011,
"reward": 0.5859216153621674,
"reward_std": 0.10529472678899765,
"reward_std_all": 0.1492040529847145,
"rewards-max/forensics_tempsamp_iou_reward": 0.9490167796611786,
"rewards/forensics_tempsamp_iou_reward": 0.5427637696266174,
"step": 815
},
{
"completion_length": 21.609375,
"epoch": 3.414225941422594,
"grad_norm": 3.8353333473205566,
"kl": 0.2607421875,
"learning_rate": 1.4748953974895397e-07,
"loss": 0.0104,
"reward": 0.734916478395462,
"reward_std": 0.04889093618839979,
"reward_std_all": 0.054084425792098045,
"rewards-max/forensics_tempsamp_iou_reward": 0.9795462191104889,
"rewards/forensics_tempsamp_iou_reward": 0.7707877457141876,
"step": 816
},
{
"completion_length": 23.375,
"epoch": 3.418410041841004,
"grad_norm": 4.465324878692627,
"kl": 0.4248046875,
"learning_rate": 1.4644351464435146e-07,
"loss": 0.017,
"reward": 0.5698880851268768,
"reward_std": 0.09592804685235023,
"reward_std_all": 0.15443124249577522,
"rewards-max/forensics_tempsamp_iou_reward": 0.9769332110881805,
"rewards/forensics_tempsamp_iou_reward": 0.5336625576019287,
"step": 817
},
{
"completion_length": 20.15625,
"epoch": 3.422594142259414,
"grad_norm": 3.742234945297241,
"kl": 0.3095703125,
"learning_rate": 1.4539748953974894e-07,
"loss": 0.0124,
"reward": 0.6583942472934723,
"reward_std": 0.06204892322421074,
"reward_std_all": 0.098881296813488,
"rewards-max/forensics_tempsamp_iou_reward": 0.9540451467037201,
"rewards/forensics_tempsamp_iou_reward": 0.6547232270240784,
"step": 818
},
{
"completion_length": 21.71875,
"epoch": 3.426778242677824,
"grad_norm": 3.354708671569824,
"kl": 0.2275390625,
"learning_rate": 1.4435146443514643e-07,
"loss": 0.0091,
"reward": 0.5758991539478302,
"reward_std": 0.0949534922838211,
"reward_std_all": 0.14693261682987213,
"rewards-max/forensics_tempsamp_iou_reward": 0.9845248460769653,
"rewards/forensics_tempsamp_iou_reward": 0.5343460440635681,
"step": 819
},
{
"completion_length": 21.78125,
"epoch": 3.430962343096234,
"grad_norm": 3.5545547008514404,
"kl": 0.41015625,
"learning_rate": 1.4330543933054392e-07,
"loss": 0.0164,
"reward": 0.6104190051555634,
"reward_std": 0.07246749475598335,
"reward_std_all": 0.12132774293422699,
"rewards-max/forensics_tempsamp_iou_reward": 0.9440257251262665,
"rewards/forensics_tempsamp_iou_reward": 0.5868684649467468,
"step": 820
},
{
"completion_length": 22.078125,
"epoch": 3.435146443514644,
"grad_norm": 3.898585796356201,
"kl": 0.353515625,
"learning_rate": 1.4225941422594143e-07,
"loss": 0.0141,
"reward": 0.6758208572864532,
"reward_std": 0.12352289631962776,
"reward_std_all": 0.12468074262142181,
"rewards-max/forensics_tempsamp_iou_reward": 0.9790610074996948,
"rewards/forensics_tempsamp_iou_reward": 0.6753322184085846,
"step": 821
},
{
"completion_length": 20.296875,
"epoch": 3.439330543933054,
"grad_norm": 3.959928512573242,
"kl": 0.357421875,
"learning_rate": 1.412133891213389e-07,
"loss": 0.0143,
"reward": 0.6967234909534454,
"reward_std": 0.09415392577648163,
"reward_std_all": 0.09660749137401581,
"rewards-max/forensics_tempsamp_iou_reward": 0.9928517043590546,
"rewards/forensics_tempsamp_iou_reward": 0.7018492519855499,
"step": 822
},
{
"completion_length": 23.65625,
"epoch": 3.4435146443514646,
"grad_norm": 6.045643329620361,
"kl": 0.576171875,
"learning_rate": 1.401673640167364e-07,
"loss": 0.0231,
"reward": 0.6831910312175751,
"reward_std": 0.09665711969137192,
"reward_std_all": 0.10580698400735855,
"rewards-max/forensics_tempsamp_iou_reward": 0.9421649277210236,
"rewards/forensics_tempsamp_iou_reward": 0.6625005304813385,
"step": 823
},
{
"completion_length": 24.84375,
"epoch": 3.4476987447698746,
"grad_norm": 4.104768753051758,
"kl": 0.3017578125,
"learning_rate": 1.3912133891213386e-07,
"loss": 0.0121,
"reward": 0.5551433861255646,
"reward_std": 0.10165431350469589,
"reward_std_all": 0.16163042187690735,
"rewards-max/forensics_tempsamp_iou_reward": 0.9442363679409027,
"rewards/forensics_tempsamp_iou_reward": 0.519254595041275,
"step": 824
},
{
"completion_length": 21.875,
"epoch": 3.4518828451882846,
"grad_norm": 4.0856828689575195,
"kl": 0.3154296875,
"learning_rate": 1.3807531380753137e-07,
"loss": 0.0126,
"reward": 0.5745992660522461,
"reward_std": 0.08147388696670532,
"reward_std_all": 0.1448260173201561,
"rewards-max/forensics_tempsamp_iou_reward": 0.9600443840026855,
"rewards/forensics_tempsamp_iou_reward": 0.5470623075962067,
"step": 825
},
{
"completion_length": 23.046875,
"epoch": 3.4560669456066946,
"grad_norm": 3.8943042755126953,
"kl": 0.3818359375,
"learning_rate": 1.370292887029289e-07,
"loss": 0.0153,
"reward": 0.6511403322219849,
"reward_std": 0.1386517398059368,
"reward_std_all": 0.14288148283958435,
"rewards-max/forensics_tempsamp_iou_reward": 0.9922429025173187,
"rewards/forensics_tempsamp_iou_reward": 0.6498256921768188,
"step": 826
},
{
"completion_length": 23.140625,
"epoch": 3.4602510460251046,
"grad_norm": 4.598392486572266,
"kl": 0.2880859375,
"learning_rate": 1.3598326359832635e-07,
"loss": 0.0115,
"reward": 0.6921956837177277,
"reward_std": 0.08118071407079697,
"reward_std_all": 0.09027332067489624,
"rewards-max/forensics_tempsamp_iou_reward": 0.9908731877803802,
"rewards/forensics_tempsamp_iou_reward": 0.6935982704162598,
"step": 827
},
{
"completion_length": 21.703125,
"epoch": 3.4644351464435146,
"grad_norm": 3.978278875350952,
"kl": 0.3779296875,
"learning_rate": 1.3493723849372386e-07,
"loss": 0.0151,
"reward": 0.5874050259590149,
"reward_std": 0.0863924827426672,
"reward_std_all": 0.14106756448745728,
"rewards-max/forensics_tempsamp_iou_reward": 0.9928381145000458,
"rewards/forensics_tempsamp_iou_reward": 0.5638235211372375,
"step": 828
},
{
"completion_length": 22.4375,
"epoch": 3.4686192468619246,
"grad_norm": 4.491161346435547,
"kl": 0.44140625,
"learning_rate": 1.3389121338912132e-07,
"loss": 0.0176,
"reward": 0.5980063378810883,
"reward_std": 0.14087598770856857,
"reward_std_all": 0.16294340789318085,
"rewards-max/forensics_tempsamp_iou_reward": 0.9687154293060303,
"rewards/forensics_tempsamp_iou_reward": 0.5721759498119354,
"step": 829
},
{
"completion_length": 20.59375,
"epoch": 3.4728033472803346,
"grad_norm": 4.093445777893066,
"kl": 0.3330078125,
"learning_rate": 1.3284518828451883e-07,
"loss": 0.0133,
"reward": 0.6794809401035309,
"reward_std": 0.09550528973340988,
"reward_std_all": 0.10510892048478127,
"rewards-max/forensics_tempsamp_iou_reward": 0.9866962432861328,
"rewards/forensics_tempsamp_iou_reward": 0.6940502822399139,
"step": 830
},
{
"completion_length": 21.1875,
"epoch": 3.476987447698745,
"grad_norm": 4.283040523529053,
"kl": 0.3642578125,
"learning_rate": 1.3179916317991632e-07,
"loss": 0.0146,
"reward": 0.7098208069801331,
"reward_std": 0.11304492503404617,
"reward_std_all": 0.103957649320364,
"rewards-max/forensics_tempsamp_iou_reward": 0.9705897569656372,
"rewards/forensics_tempsamp_iou_reward": 0.7056213021278381,
"step": 831
},
{
"completion_length": 20.125,
"epoch": 3.481171548117155,
"grad_norm": 4.537107467651367,
"kl": 0.3017578125,
"learning_rate": 1.307531380753138e-07,
"loss": 0.0121,
"reward": 0.5707978010177612,
"reward_std": 0.10767390578985214,
"reward_std_all": 0.1542801558971405,
"rewards-max/forensics_tempsamp_iou_reward": 0.9631167650222778,
"rewards/forensics_tempsamp_iou_reward": 0.5506298840045929,
"step": 832
},
{
"completion_length": 23.953125,
"epoch": 3.485355648535565,
"grad_norm": 4.162929058074951,
"kl": 0.3857421875,
"learning_rate": 1.297071129707113e-07,
"loss": 0.0154,
"reward": 0.6474619805812836,
"reward_std": 0.11433020234107971,
"reward_std_all": 0.12563523277640343,
"rewards-max/forensics_tempsamp_iou_reward": 0.9426724910736084,
"rewards/forensics_tempsamp_iou_reward": 0.6198407411575317,
"step": 833
},
{
"completion_length": 22.609375,
"epoch": 3.489539748953975,
"grad_norm": 3.640709400177002,
"kl": 0.3447265625,
"learning_rate": 1.2866108786610878e-07,
"loss": 0.0138,
"reward": 0.6736491918563843,
"reward_std": 0.12510839104652405,
"reward_std_all": 0.1221449263393879,
"rewards-max/forensics_tempsamp_iou_reward": 0.9629924297332764,
"rewards/forensics_tempsamp_iou_reward": 0.6719247102737427,
"step": 834
},
{
"completion_length": 21.65625,
"epoch": 3.493723849372385,
"grad_norm": 4.614521026611328,
"kl": 0.3525390625,
"learning_rate": 1.2761506276150627e-07,
"loss": 0.0141,
"reward": 0.6293576061725616,
"reward_std": 0.07331142202019691,
"reward_std_all": 0.11570650711655617,
"rewards-max/forensics_tempsamp_iou_reward": 0.9617400467395782,
"rewards/forensics_tempsamp_iou_reward": 0.6172356605529785,
"step": 835
},
{
"completion_length": 20.90625,
"epoch": 3.497907949790795,
"grad_norm": 4.076084136962891,
"kl": 0.3271484375,
"learning_rate": 1.2656903765690378e-07,
"loss": 0.0131,
"reward": 0.6890542805194855,
"reward_std": 0.0751022957265377,
"reward_std_all": 0.0847693681716919,
"rewards-max/forensics_tempsamp_iou_reward": 0.9919674098491669,
"rewards/forensics_tempsamp_iou_reward": 0.6836678683757782,
"step": 836
},
{
"completion_length": 20.625,
"epoch": 3.502092050209205,
"grad_norm": 4.051024913787842,
"kl": 0.322265625,
"learning_rate": 1.2552301255230124e-07,
"loss": 0.0129,
"reward": 0.681665688753128,
"reward_std": 0.06644447520375252,
"reward_std_all": 0.0844716839492321,
"rewards-max/forensics_tempsamp_iou_reward": 0.9309855103492737,
"rewards/forensics_tempsamp_iou_reward": 0.6691074073314667,
"step": 837
},
{
"completion_length": 21.8125,
"epoch": 3.506276150627615,
"grad_norm": 3.919538974761963,
"kl": 0.37109375,
"learning_rate": 1.2447698744769875e-07,
"loss": 0.0148,
"reward": 0.5657317042350769,
"reward_std": 0.130879994481802,
"reward_std_all": 0.17565935105085373,
"rewards-max/forensics_tempsamp_iou_reward": 0.9392009079456329,
"rewards/forensics_tempsamp_iou_reward": 0.5273265838623047,
"step": 838
},
{
"completion_length": 19.765625,
"epoch": 3.510460251046025,
"grad_norm": 3.440661907196045,
"kl": 0.33984375,
"learning_rate": 1.2343096234309624e-07,
"loss": 0.0136,
"reward": 0.6140101850032806,
"reward_std": 0.11734334379434586,
"reward_std_all": 0.14753156900405884,
"rewards-max/forensics_tempsamp_iou_reward": 0.9621542990207672,
"rewards/forensics_tempsamp_iou_reward": 0.621772438287735,
"step": 839
},
{
"completion_length": 20.578125,
"epoch": 3.514644351464435,
"grad_norm": 3.7148289680480957,
"kl": 0.3251953125,
"learning_rate": 1.2238493723849372e-07,
"loss": 0.013,
"reward": 0.67657670378685,
"reward_std": 0.10036706924438477,
"reward_std_all": 0.1064986065030098,
"rewards-max/forensics_tempsamp_iou_reward": 0.9582748711109161,
"rewards/forensics_tempsamp_iou_reward": 0.6586194932460785,
"step": 840
},
{
"completion_length": 21.34375,
"epoch": 3.518828451882845,
"grad_norm": 4.432254791259766,
"kl": 0.3046875,
"learning_rate": 1.213389121338912e-07,
"loss": 0.0122,
"reward": 0.6838335394859314,
"reward_std": 0.10014142841100693,
"reward_std_all": 0.10262815654277802,
"rewards-max/forensics_tempsamp_iou_reward": 0.9398931860923767,
"rewards/forensics_tempsamp_iou_reward": 0.6824215054512024,
"step": 841
},
{
"completion_length": 21.59375,
"epoch": 3.523012552301255,
"grad_norm": 4.3462724685668945,
"kl": 0.3173828125,
"learning_rate": 1.202928870292887e-07,
"loss": 0.0127,
"reward": 0.6221111714839935,
"reward_std": 0.0806998759508133,
"reward_std_all": 0.1207215003669262,
"rewards-max/forensics_tempsamp_iou_reward": 0.9621982872486115,
"rewards/forensics_tempsamp_iou_reward": 0.589678943157196,
"step": 842
},
{
"completion_length": 22.421875,
"epoch": 3.5271966527196654,
"grad_norm": 5.280770778656006,
"kl": 0.341796875,
"learning_rate": 1.1924686192468618e-07,
"loss": 0.0137,
"reward": 0.6277736723423004,
"reward_std": 0.11322294175624847,
"reward_std_all": 0.1351558342576027,
"rewards-max/forensics_tempsamp_iou_reward": 0.9858317375183105,
"rewards/forensics_tempsamp_iou_reward": 0.613071471452713,
"step": 843
},
{
"completion_length": 22.640625,
"epoch": 3.5313807531380754,
"grad_norm": 4.432629108428955,
"kl": 0.3486328125,
"learning_rate": 1.1820083682008368e-07,
"loss": 0.0139,
"reward": 0.5972003042697906,
"reward_std": 0.056999696418643,
"reward_std_all": 0.12109339237213135,
"rewards-max/forensics_tempsamp_iou_reward": 0.9717356860637665,
"rewards/forensics_tempsamp_iou_reward": 0.5770926773548126,
"step": 844
},
{
"completion_length": 20.6875,
"epoch": 3.5355648535564854,
"grad_norm": 4.250466346740723,
"kl": 0.3564453125,
"learning_rate": 1.1715481171548117e-07,
"loss": 0.0143,
"reward": 0.6727270483970642,
"reward_std": 0.09002546221017838,
"reward_std_all": 0.10493815317749977,
"rewards-max/forensics_tempsamp_iou_reward": 0.9786175191402435,
"rewards/forensics_tempsamp_iou_reward": 0.6747300922870636,
"step": 845
},
{
"completion_length": 24.640625,
"epoch": 3.5397489539748954,
"grad_norm": 4.596809387207031,
"kl": 0.283203125,
"learning_rate": 1.1610878661087866e-07,
"loss": 0.0113,
"reward": 0.5835047960281372,
"reward_std": 0.14407125115394592,
"reward_std_all": 0.16906961798667908,
"rewards-max/forensics_tempsamp_iou_reward": 0.9814073741436005,
"rewards/forensics_tempsamp_iou_reward": 0.5278753936290741,
"step": 846
},
{
"completion_length": 22.796875,
"epoch": 3.5439330543933054,
"grad_norm": 5.065432071685791,
"kl": 0.31640625,
"learning_rate": 1.1506276150627614e-07,
"loss": 0.0126,
"reward": 0.6426012217998505,
"reward_std": 0.06428413838148117,
"reward_std_all": 0.10125140100717545,
"rewards-max/forensics_tempsamp_iou_reward": 0.9765044748783112,
"rewards/forensics_tempsamp_iou_reward": 0.6342350840568542,
"step": 847
},
{
"completion_length": 21.40625,
"epoch": 3.5481171548117154,
"grad_norm": 3.5955798625946045,
"kl": 0.3505859375,
"learning_rate": 1.1401673640167364e-07,
"loss": 0.014,
"reward": 0.6187633872032166,
"reward_std": 0.09586426243185997,
"reward_std_all": 0.1315687820315361,
"rewards-max/forensics_tempsamp_iou_reward": 0.9570682048797607,
"rewards/forensics_tempsamp_iou_reward": 0.5891831517219543,
"step": 848
},
{
"completion_length": 19.90625,
"epoch": 3.5523012552301254,
"grad_norm": 4.300892353057861,
"kl": 0.3837890625,
"learning_rate": 1.1297071129707113e-07,
"loss": 0.0153,
"reward": 0.664683610200882,
"reward_std": 0.10903355479240417,
"reward_std_all": 0.12203646078705788,
"rewards-max/forensics_tempsamp_iou_reward": 0.9739622175693512,
"rewards/forensics_tempsamp_iou_reward": 0.6630917489528656,
"step": 849
},
{
"completion_length": 24.53125,
"epoch": 3.5564853556485354,
"grad_norm": 4.232895851135254,
"kl": 0.330078125,
"learning_rate": 1.1192468619246861e-07,
"loss": 0.0132,
"reward": 0.6686679720878601,
"reward_std": 0.06159411743283272,
"reward_std_all": 0.08546257764101028,
"rewards-max/forensics_tempsamp_iou_reward": 0.976051926612854,
"rewards/forensics_tempsamp_iou_reward": 0.6523406505584717,
"step": 850
},
{
"completion_length": 19.796875,
"epoch": 3.560669456066946,
"grad_norm": 5.342323303222656,
"kl": 0.2822265625,
"learning_rate": 1.108786610878661e-07,
"loss": 0.0113,
"reward": 0.5962255299091339,
"reward_std": 0.08196714520454407,
"reward_std_all": 0.14021573960781097,
"rewards-max/forensics_tempsamp_iou_reward": 0.9526103734970093,
"rewards/forensics_tempsamp_iou_reward": 0.574753612279892,
"step": 851
},
{
"completion_length": 23.34375,
"epoch": 3.564853556485356,
"grad_norm": 4.45583963394165,
"kl": 0.361328125,
"learning_rate": 1.0983263598326359e-07,
"loss": 0.0144,
"reward": 0.6801021993160248,
"reward_std": 0.052176643162965775,
"reward_std_all": 0.08118237555027008,
"rewards-max/forensics_tempsamp_iou_reward": 0.9806548058986664,
"rewards/forensics_tempsamp_iou_reward": 0.669297844171524,
"step": 852
},
{
"completion_length": 23.078125,
"epoch": 3.569037656903766,
"grad_norm": 4.955367565155029,
"kl": 0.3642578125,
"learning_rate": 1.0878661087866109e-07,
"loss": 0.0146,
"reward": 0.6168101727962494,
"reward_std": 0.043585218489170074,
"reward_std_all": 0.11041291803121567,
"rewards-max/forensics_tempsamp_iou_reward": 0.9632745385169983,
"rewards/forensics_tempsamp_iou_reward": 0.6054674088954926,
"step": 853
},
{
"completion_length": 22.53125,
"epoch": 3.573221757322176,
"grad_norm": 3.29134202003479,
"kl": 0.2880859375,
"learning_rate": 1.0774058577405857e-07,
"loss": 0.0115,
"reward": 0.6577629446983337,
"reward_std": 0.09408236294984818,
"reward_std_all": 0.11660861223936081,
"rewards-max/forensics_tempsamp_iou_reward": 0.9718573987483978,
"rewards/forensics_tempsamp_iou_reward": 0.6664155423641205,
"step": 854
},
{
"completion_length": 19.59375,
"epoch": 3.577405857740586,
"grad_norm": 3.3040413856506348,
"kl": 0.375,
"learning_rate": 1.0669456066945606e-07,
"loss": 0.015,
"reward": 0.6617259979248047,
"reward_std": 0.13675031810998917,
"reward_std_all": 0.13459690660238266,
"rewards-max/forensics_tempsamp_iou_reward": 0.9681629836559296,
"rewards/forensics_tempsamp_iou_reward": 0.6753084659576416,
"step": 855
},
{
"completion_length": 21.53125,
"epoch": 3.581589958158996,
"grad_norm": 3.7292537689208984,
"kl": 0.341796875,
"learning_rate": 1.0564853556485355e-07,
"loss": 0.0137,
"reward": 0.5261081904172897,
"reward_std": 0.11724435165524483,
"reward_std_all": 0.18619540333747864,
"rewards-max/forensics_tempsamp_iou_reward": 0.9708141386508942,
"rewards/forensics_tempsamp_iou_reward": 0.4842481464147568,
"step": 856
},
{
"completion_length": 20.921875,
"epoch": 3.585774058577406,
"grad_norm": 3.5547780990600586,
"kl": 0.3798828125,
"learning_rate": 1.0460251046025103e-07,
"loss": 0.0152,
"reward": 0.7444186806678772,
"reward_std": 0.04969438910484314,
"reward_std_all": 0.0505848154425621,
"rewards-max/forensics_tempsamp_iou_reward": 0.9739950597286224,
"rewards/forensics_tempsamp_iou_reward": 0.791414201259613,
"step": 857
},
{
"completion_length": 24.796875,
"epoch": 3.589958158995816,
"grad_norm": 4.149171829223633,
"kl": 0.27001953125,
"learning_rate": 1.0355648535564853e-07,
"loss": 0.0108,
"reward": 0.6001919806003571,
"reward_std": 0.08086299896240234,
"reward_std_all": 0.12855257466435432,
"rewards-max/forensics_tempsamp_iou_reward": 0.9969726204872131,
"rewards/forensics_tempsamp_iou_reward": 0.5676166415214539,
"step": 858
},
{
"completion_length": 24.21875,
"epoch": 3.594142259414226,
"grad_norm": 3.358888864517212,
"kl": 0.326171875,
"learning_rate": 1.0251046025104603e-07,
"loss": 0.013,
"reward": 0.6689631938934326,
"reward_std": 0.1225070022046566,
"reward_std_all": 0.12270345911383629,
"rewards-max/forensics_tempsamp_iou_reward": 0.9736599326133728,
"rewards/forensics_tempsamp_iou_reward": 0.6586126685142517,
"step": 859
},
{
"completion_length": 24.34375,
"epoch": 3.598326359832636,
"grad_norm": 3.8331406116485596,
"kl": 0.4462890625,
"learning_rate": 1.0146443514644352e-07,
"loss": 0.0179,
"reward": 0.5955730080604553,
"reward_std": 0.11277160421013832,
"reward_std_all": 0.15407584607601166,
"rewards-max/forensics_tempsamp_iou_reward": 0.9355087280273438,
"rewards/forensics_tempsamp_iou_reward": 0.5481024086475372,
"step": 860
},
{
"completion_length": 20.078125,
"epoch": 3.602510460251046,
"grad_norm": 4.363016128540039,
"kl": 0.3466796875,
"learning_rate": 1.00418410041841e-07,
"loss": 0.0139,
"reward": 0.6251027882099152,
"reward_std": 0.10415956377983093,
"reward_std_all": 0.13112883269786835,
"rewards-max/forensics_tempsamp_iou_reward": 0.992061048746109,
"rewards/forensics_tempsamp_iou_reward": 0.6316745579242706,
"step": 861
},
{
"completion_length": 22.09375,
"epoch": 3.606694560669456,
"grad_norm": 4.2111921310424805,
"kl": 0.388671875,
"learning_rate": 9.937238493723849e-08,
"loss": 0.0155,
"reward": 0.6085086762905121,
"reward_std": 0.06425469624809921,
"reward_std_all": 0.11892937123775482,
"rewards-max/forensics_tempsamp_iou_reward": 0.9771874248981476,
"rewards/forensics_tempsamp_iou_reward": 0.6094637215137482,
"step": 862
},
{
"completion_length": 21.53125,
"epoch": 3.6108786610878663,
"grad_norm": 4.712471961975098,
"kl": 0.333984375,
"learning_rate": 9.832635983263599e-08,
"loss": 0.0134,
"reward": 0.6818680167198181,
"reward_std": 0.06774722039699554,
"reward_std_all": 0.08361323922872543,
"rewards-max/forensics_tempsamp_iou_reward": 0.9925242960453033,
"rewards/forensics_tempsamp_iou_reward": 0.6939801871776581,
"step": 863
},
{
"completion_length": 20.890625,
"epoch": 3.6150627615062763,
"grad_norm": 3.7780978679656982,
"kl": 0.33984375,
"learning_rate": 9.728033472803348e-08,
"loss": 0.0136,
"reward": 0.7248209118843079,
"reward_std": 0.06424570828676224,
"reward_std_all": 0.06756417453289032,
"rewards-max/forensics_tempsamp_iou_reward": 0.9944478571414948,
"rewards/forensics_tempsamp_iou_reward": 0.762656569480896,
"step": 864
},
{
"completion_length": 21.5,
"epoch": 3.6192468619246863,
"grad_norm": 4.112601280212402,
"kl": 0.30859375,
"learning_rate": 9.623430962343096e-08,
"loss": 0.0124,
"reward": 0.5658475756645203,
"reward_std": 0.11173034831881523,
"reward_std_all": 0.16711435467004776,
"rewards-max/forensics_tempsamp_iou_reward": 0.9705889523029327,
"rewards/forensics_tempsamp_iou_reward": 0.5507297813892365,
"step": 865
},
{
"completion_length": 22.234375,
"epoch": 3.6234309623430963,
"grad_norm": 11.49528694152832,
"kl": 1.02734375,
"learning_rate": 9.518828451882845e-08,
"loss": 0.0412,
"reward": 0.7016259729862213,
"reward_std": 0.07068635895848274,
"reward_std_all": 0.07928229868412018,
"rewards-max/forensics_tempsamp_iou_reward": 0.9338837563991547,
"rewards/forensics_tempsamp_iou_reward": 0.6790214776992798,
"step": 866
},
{
"completion_length": 22.65625,
"epoch": 3.6276150627615062,
"grad_norm": 4.144462585449219,
"kl": 0.43359375,
"learning_rate": 9.414225941422594e-08,
"loss": 0.0174,
"reward": 0.6526122987270355,
"reward_std": 0.1031029187142849,
"reward_std_all": 0.11729295924305916,
"rewards-max/forensics_tempsamp_iou_reward": 0.9354166686534882,
"rewards/forensics_tempsamp_iou_reward": 0.6360883116722107,
"step": 867
},
{
"completion_length": 23.484375,
"epoch": 3.6317991631799162,
"grad_norm": 4.610730171203613,
"kl": 0.2783203125,
"learning_rate": 9.309623430962342e-08,
"loss": 0.0111,
"reward": 0.5348545163869858,
"reward_std": 0.09525278210639954,
"reward_std_all": 0.1708744391798973,
"rewards-max/forensics_tempsamp_iou_reward": 0.9570224583148956,
"rewards/forensics_tempsamp_iou_reward": 0.49258580803871155,
"step": 868
},
{
"completion_length": 22.3125,
"epoch": 3.6359832635983262,
"grad_norm": 3.448965311050415,
"kl": 0.3046875,
"learning_rate": 9.205020920502092e-08,
"loss": 0.0122,
"reward": 0.6603048145771027,
"reward_std": 0.07442989572882652,
"reward_std_all": 0.10458794981241226,
"rewards-max/forensics_tempsamp_iou_reward": 0.9821615219116211,
"rewards/forensics_tempsamp_iou_reward": 0.6638823449611664,
"step": 869
},
{
"completion_length": 22.4375,
"epoch": 3.6401673640167362,
"grad_norm": 4.518397331237793,
"kl": 0.3525390625,
"learning_rate": 9.100418410041841e-08,
"loss": 0.0141,
"reward": 0.6292495727539062,
"reward_std": 0.1196896992623806,
"reward_std_all": 0.13897276669740677,
"rewards-max/forensics_tempsamp_iou_reward": 0.9841166436672211,
"rewards/forensics_tempsamp_iou_reward": 0.6071669161319733,
"step": 870
},
{
"completion_length": 21.796875,
"epoch": 3.6443514644351467,
"grad_norm": 4.568434238433838,
"kl": 0.51171875,
"learning_rate": 8.99581589958159e-08,
"loss": 0.0205,
"reward": 0.6341868042945862,
"reward_std": 0.10673506185412407,
"reward_std_all": 0.127846822142601,
"rewards-max/forensics_tempsamp_iou_reward": 0.9818832874298096,
"rewards/forensics_tempsamp_iou_reward": 0.6260814964771271,
"step": 871
},
{
"completion_length": 22.125,
"epoch": 3.6485355648535567,
"grad_norm": 3.746765375137329,
"kl": 0.279296875,
"learning_rate": 8.891213389121338e-08,
"loss": 0.0111,
"reward": 0.5889076590538025,
"reward_std": 0.12324992567300797,
"reward_std_all": 0.15862033516168594,
"rewards-max/forensics_tempsamp_iou_reward": 0.9385161101818085,
"rewards/forensics_tempsamp_iou_reward": 0.5462877154350281,
"step": 872
},
{
"completion_length": 21.296875,
"epoch": 3.6527196652719667,
"grad_norm": 3.882101058959961,
"kl": 0.28369140625,
"learning_rate": 8.786610878661087e-08,
"loss": 0.0113,
"reward": 0.6332324743270874,
"reward_std": 0.11829734593629837,
"reward_std_all": 0.13712478429079056,
"rewards-max/forensics_tempsamp_iou_reward": 0.9448001086711884,
"rewards/forensics_tempsamp_iou_reward": 0.589026004076004,
"step": 873
},
{
"completion_length": 21.625,
"epoch": 3.6569037656903767,
"grad_norm": 4.076632022857666,
"kl": 0.3330078125,
"learning_rate": 8.682008368200837e-08,
"loss": 0.0133,
"reward": 0.6285557448863983,
"reward_std": 0.12293579801917076,
"reward_std_all": 0.13984419405460358,
"rewards-max/forensics_tempsamp_iou_reward": 0.9393890798091888,
"rewards/forensics_tempsamp_iou_reward": 0.5969720780849457,
"step": 874
},
{
"completion_length": 22.734375,
"epoch": 3.6610878661087867,
"grad_norm": 4.228011131286621,
"kl": 0.330078125,
"learning_rate": 8.577405857740585e-08,
"loss": 0.0132,
"reward": 0.6476071774959564,
"reward_std": 0.0983026996254921,
"reward_std_all": 0.11606080085039139,
"rewards-max/forensics_tempsamp_iou_reward": 0.9674106240272522,
"rewards/forensics_tempsamp_iou_reward": 0.6329072713851929,
"step": 875
},
{
"completion_length": 22.609375,
"epoch": 3.6652719665271967,
"grad_norm": 4.175085544586182,
"kl": 0.3193359375,
"learning_rate": 8.472803347280334e-08,
"loss": 0.0128,
"reward": 0.6007784903049469,
"reward_std": 0.04079297371208668,
"reward_std_all": 0.11412037909030914,
"rewards-max/forensics_tempsamp_iou_reward": 0.986151784658432,
"rewards/forensics_tempsamp_iou_reward": 0.6144733428955078,
"step": 876
},
{
"completion_length": 21.421875,
"epoch": 3.6694560669456067,
"grad_norm": 5.059027671813965,
"kl": 0.302734375,
"learning_rate": 8.368200836820083e-08,
"loss": 0.0121,
"reward": 0.5792494118213654,
"reward_std": 0.10658379271626472,
"reward_std_all": 0.15338629484176636,
"rewards-max/forensics_tempsamp_iou_reward": 0.9485167264938354,
"rewards/forensics_tempsamp_iou_reward": 0.5477468520402908,
"step": 877
},
{
"completion_length": 21.046875,
"epoch": 3.6736401673640167,
"grad_norm": 4.322465896606445,
"kl": 0.3310546875,
"learning_rate": 8.263598326359831e-08,
"loss": 0.0132,
"reward": 0.6435152590274811,
"reward_std": 0.0684956181794405,
"reward_std_all": 0.11175505071878433,
"rewards-max/forensics_tempsamp_iou_reward": 0.9922863245010376,
"rewards/forensics_tempsamp_iou_reward": 0.6641860902309418,
"step": 878
},
{
"completion_length": 23.109375,
"epoch": 3.6778242677824267,
"grad_norm": 5.045461177825928,
"kl": 0.4267578125,
"learning_rate": 8.158995815899581e-08,
"loss": 0.0171,
"reward": 0.624357283115387,
"reward_std": 0.11959496513009071,
"reward_std_all": 0.14752297848463058,
"rewards-max/forensics_tempsamp_iou_reward": 0.9900000095367432,
"rewards/forensics_tempsamp_iou_reward": 0.5970511436462402,
"step": 879
},
{
"completion_length": 19.28125,
"epoch": 3.6820083682008367,
"grad_norm": 3.732182025909424,
"kl": 0.31982421875,
"learning_rate": 8.05439330543933e-08,
"loss": 0.0128,
"reward": 0.6283750236034393,
"reward_std": 0.1254291608929634,
"reward_std_all": 0.13875902444124222,
"rewards-max/forensics_tempsamp_iou_reward": 0.9807409346103668,
"rewards/forensics_tempsamp_iou_reward": 0.6000624299049377,
"step": 880
},
{
"completion_length": 21.046875,
"epoch": 3.6861924686192467,
"grad_norm": 4.160340785980225,
"kl": 0.2978515625,
"learning_rate": 7.949790794979078e-08,
"loss": 0.0119,
"reward": 0.6586472392082214,
"reward_std": 0.12278763204813004,
"reward_std_all": 0.1280941367149353,
"rewards-max/forensics_tempsamp_iou_reward": 0.9781724512577057,
"rewards/forensics_tempsamp_iou_reward": 0.6408777236938477,
"step": 881
},
{
"completion_length": 21.390625,
"epoch": 3.6903765690376567,
"grad_norm": 3.8401405811309814,
"kl": 0.2861328125,
"learning_rate": 7.845188284518828e-08,
"loss": 0.0115,
"reward": 0.6424459218978882,
"reward_std": 0.06791802868247032,
"reward_std_all": 0.10529877990484238,
"rewards-max/forensics_tempsamp_iou_reward": 0.9656766057014465,
"rewards/forensics_tempsamp_iou_reward": 0.6252138316631317,
"step": 882
},
{
"completion_length": 21.875,
"epoch": 3.694560669456067,
"grad_norm": 3.230269432067871,
"kl": 0.37890625,
"learning_rate": 7.740585774058577e-08,
"loss": 0.0151,
"reward": 0.6136822998523712,
"reward_std": 0.1165110394358635,
"reward_std_all": 0.14451851695775986,
"rewards-max/forensics_tempsamp_iou_reward": 0.9451298117637634,
"rewards/forensics_tempsamp_iou_reward": 0.5949263870716095,
"step": 883
},
{
"completion_length": 24.859375,
"epoch": 3.698744769874477,
"grad_norm": 3.4204626083374023,
"kl": 0.3232421875,
"learning_rate": 7.635983263598327e-08,
"loss": 0.013,
"reward": 0.489309161901474,
"reward_std": 0.1342516541481018,
"reward_std_all": 0.20387941598892212,
"rewards-max/forensics_tempsamp_iou_reward": 0.9561607837677002,
"rewards/forensics_tempsamp_iou_reward": 0.4241843670606613,
"step": 884
},
{
"completion_length": 21.40625,
"epoch": 3.702928870292887,
"grad_norm": 3.8464226722717285,
"kl": 0.26318359375,
"learning_rate": 7.531380753138076e-08,
"loss": 0.0105,
"reward": 0.5625461786985397,
"reward_std": 0.11839500814676285,
"reward_std_all": 0.1642950475215912,
"rewards-max/forensics_tempsamp_iou_reward": 0.9012194871902466,
"rewards/forensics_tempsamp_iou_reward": 0.5051288902759552,
"step": 885
},
{
"completion_length": 20.859375,
"epoch": 3.707112970711297,
"grad_norm": 4.646881580352783,
"kl": 0.3115234375,
"learning_rate": 7.426778242677824e-08,
"loss": 0.0125,
"reward": 0.6516750454902649,
"reward_std": 0.10469171777367592,
"reward_std_all": 0.11744630336761475,
"rewards-max/forensics_tempsamp_iou_reward": 0.9340282082557678,
"rewards/forensics_tempsamp_iou_reward": 0.6220161318778992,
"step": 886
},
{
"completion_length": 23.71875,
"epoch": 3.711297071129707,
"grad_norm": 3.382333517074585,
"kl": 0.3212890625,
"learning_rate": 7.322175732217573e-08,
"loss": 0.0129,
"reward": 0.614057183265686,
"reward_std": 0.13860861212015152,
"reward_std_all": 0.1556924358010292,
"rewards-max/forensics_tempsamp_iou_reward": 0.9102033078670502,
"rewards/forensics_tempsamp_iou_reward": 0.5775401592254639,
"step": 887
},
{
"completion_length": 23.953125,
"epoch": 3.715481171548117,
"grad_norm": 4.209962844848633,
"kl": 0.2900390625,
"learning_rate": 7.217573221757322e-08,
"loss": 0.0116,
"reward": 0.6490301489830017,
"reward_std": 0.08278108015656471,
"reward_std_all": 0.118263840675354,
"rewards-max/forensics_tempsamp_iou_reward": 0.9599843323230743,
"rewards/forensics_tempsamp_iou_reward": 0.6540341079235077,
"step": 888
},
{
"completion_length": 22.34375,
"epoch": 3.719665271966527,
"grad_norm": 3.909090995788574,
"kl": 0.3017578125,
"learning_rate": 7.112970711297071e-08,
"loss": 0.0121,
"reward": 0.6490744650363922,
"reward_std": 0.06246555596590042,
"reward_std_all": 0.09956159815192223,
"rewards-max/forensics_tempsamp_iou_reward": 0.9742327034473419,
"rewards/forensics_tempsamp_iou_reward": 0.6297095715999603,
"step": 889
},
{
"completion_length": 23.40625,
"epoch": 3.723849372384937,
"grad_norm": 3.492238759994507,
"kl": 0.314453125,
"learning_rate": 7.00836820083682e-08,
"loss": 0.0126,
"reward": 0.6650658249855042,
"reward_std": 0.08865484595298767,
"reward_std_all": 0.10694406181573868,
"rewards-max/forensics_tempsamp_iou_reward": 0.9582949578762054,
"rewards/forensics_tempsamp_iou_reward": 0.6619082689285278,
"step": 890
},
{
"completion_length": 21.5625,
"epoch": 3.7280334728033475,
"grad_norm": 4.770296573638916,
"kl": 0.31640625,
"learning_rate": 6.903765690376569e-08,
"loss": 0.0126,
"reward": 0.7003653049468994,
"reward_std": 0.054564766585826874,
"reward_std_all": 0.07290589809417725,
"rewards-max/forensics_tempsamp_iou_reward": 0.9710017740726471,
"rewards/forensics_tempsamp_iou_reward": 0.7218624949455261,
"step": 891
},
{
"completion_length": 22.515625,
"epoch": 3.7322175732217575,
"grad_norm": 4.825583457946777,
"kl": 0.328125,
"learning_rate": 6.799163179916317e-08,
"loss": 0.0131,
"reward": 0.6601502001285553,
"reward_std": 0.12047639861702919,
"reward_std_all": 0.1263686642050743,
"rewards-max/forensics_tempsamp_iou_reward": 0.9704990983009338,
"rewards/forensics_tempsamp_iou_reward": 0.6417716443538666,
"step": 892
},
{
"completion_length": 22.078125,
"epoch": 3.7364016736401675,
"grad_norm": 4.086126327514648,
"kl": 0.291015625,
"learning_rate": 6.694560669456066e-08,
"loss": 0.0116,
"reward": 0.6292308568954468,
"reward_std": 0.07189259864389896,
"reward_std_all": 0.11350950971245766,
"rewards-max/forensics_tempsamp_iou_reward": 0.9663520455360413,
"rewards/forensics_tempsamp_iou_reward": 0.5952286720275879,
"step": 893
},
{
"completion_length": 20.484375,
"epoch": 3.7405857740585775,
"grad_norm": 3.608555316925049,
"kl": 0.3291015625,
"learning_rate": 6.589958158995816e-08,
"loss": 0.0131,
"reward": 0.7177779674530029,
"reward_std": 0.07934238016605377,
"reward_std_all": 0.08059995993971825,
"rewards-max/forensics_tempsamp_iou_reward": 0.9687996208667755,
"rewards/forensics_tempsamp_iou_reward": 0.7273205518722534,
"step": 894
},
{
"completion_length": 20.125,
"epoch": 3.7447698744769875,
"grad_norm": 4.225025177001953,
"kl": 0.28369140625,
"learning_rate": 6.485355648535565e-08,
"loss": 0.0114,
"reward": 0.690964013338089,
"reward_std": 0.07204379141330719,
"reward_std_all": 0.08468317613005638,
"rewards-max/forensics_tempsamp_iou_reward": 0.976727157831192,
"rewards/forensics_tempsamp_iou_reward": 0.7257002592086792,
"step": 895
},
{
"completion_length": 23.859375,
"epoch": 3.7489539748953975,
"grad_norm": 3.7152416706085205,
"kl": 0.3134765625,
"learning_rate": 6.380753138075313e-08,
"loss": 0.0126,
"reward": 0.6602135896682739,
"reward_std": 0.07447915710508823,
"reward_std_all": 0.10165398567914963,
"rewards-max/forensics_tempsamp_iou_reward": 0.9853972792625427,
"rewards/forensics_tempsamp_iou_reward": 0.6636320650577545,
"step": 896
},
{
"completion_length": 21.265625,
"epoch": 3.7531380753138075,
"grad_norm": 3.719637155532837,
"kl": 0.326171875,
"learning_rate": 6.276150627615062e-08,
"loss": 0.0131,
"reward": 0.6757879555225372,
"reward_std": 0.089620441198349,
"reward_std_all": 0.10160594806075096,
"rewards-max/forensics_tempsamp_iou_reward": 0.9759129583835602,
"rewards/forensics_tempsamp_iou_reward": 0.6741079688072205,
"step": 897
},
{
"completion_length": 21.109375,
"epoch": 3.7573221757322175,
"grad_norm": 3.746856927871704,
"kl": 0.294921875,
"learning_rate": 6.171548117154812e-08,
"loss": 0.0118,
"reward": 0.6774363815784454,
"reward_std": 0.09569645300507545,
"reward_std_all": 0.10574596002697945,
"rewards-max/forensics_tempsamp_iou_reward": 0.9797998368740082,
"rewards/forensics_tempsamp_iou_reward": 0.6944358050823212,
"step": 898
},
{
"completion_length": 22.71875,
"epoch": 3.7615062761506275,
"grad_norm": 3.7529585361480713,
"kl": 0.36328125,
"learning_rate": 6.06694560669456e-08,
"loss": 0.0145,
"reward": 0.6413888931274414,
"reward_std": 0.09500409290194511,
"reward_std_all": 0.12369924038648605,
"rewards-max/forensics_tempsamp_iou_reward": 0.9692308902740479,
"rewards/forensics_tempsamp_iou_reward": 0.622542530298233,
"step": 899
},
{
"completion_length": 20.328125,
"epoch": 3.7656903765690375,
"grad_norm": 3.503220319747925,
"kl": 0.28173828125,
"learning_rate": 5.962343096234309e-08,
"loss": 0.0113,
"reward": 0.6272989511489868,
"reward_std": 0.100553959608078,
"reward_std_all": 0.12711863219738007,
"rewards-max/forensics_tempsamp_iou_reward": 0.989947110414505,
"rewards/forensics_tempsamp_iou_reward": 0.6149181425571442,
"step": 900
},
{
"completion_length": 21.046875,
"epoch": 3.7698744769874475,
"grad_norm": 3.893921136856079,
"kl": 0.345703125,
"learning_rate": 5.8577405857740584e-08,
"loss": 0.0138,
"reward": 0.6684218049049377,
"reward_std": 0.06327111832797527,
"reward_std_all": 0.0920338612049818,
"rewards-max/forensics_tempsamp_iou_reward": 0.9931188225746155,
"rewards/forensics_tempsamp_iou_reward": 0.6727066338062286,
"step": 901
},
{
"completion_length": 20.140625,
"epoch": 3.7740585774058575,
"grad_norm": 4.65335750579834,
"kl": 0.287109375,
"learning_rate": 5.753138075313807e-08,
"loss": 0.0115,
"reward": 0.5845436304807663,
"reward_std": 0.10295115411281586,
"reward_std_all": 0.14803073927760124,
"rewards-max/forensics_tempsamp_iou_reward": 0.8995528221130371,
"rewards/forensics_tempsamp_iou_reward": 0.5465314239263535,
"step": 902
},
{
"completion_length": 21.9375,
"epoch": 3.778242677824268,
"grad_norm": 3.5435492992401123,
"kl": 0.33203125,
"learning_rate": 5.6485355648535564e-08,
"loss": 0.0133,
"reward": 0.6758415699005127,
"reward_std": 0.10125298798084259,
"reward_std_all": 0.10691889002919197,
"rewards-max/forensics_tempsamp_iou_reward": 0.9851135611534119,
"rewards/forensics_tempsamp_iou_reward": 0.6775215864181519,
"step": 903
},
{
"completion_length": 24.375,
"epoch": 3.782426778242678,
"grad_norm": 3.5319008827209473,
"kl": 0.2841796875,
"learning_rate": 5.543933054393305e-08,
"loss": 0.0114,
"reward": 0.6745407581329346,
"reward_std": 0.09156706184148788,
"reward_std_all": 0.10529224388301373,
"rewards-max/forensics_tempsamp_iou_reward": 0.9788776934146881,
"rewards/forensics_tempsamp_iou_reward": 0.6653129458427429,
"step": 904
},
{
"completion_length": 20.171875,
"epoch": 3.786610878661088,
"grad_norm": 4.539848804473877,
"kl": 0.2890625,
"learning_rate": 5.439330543933054e-08,
"loss": 0.0116,
"reward": 0.6128018200397491,
"reward_std": 0.06480086408555508,
"reward_std_all": 0.11817321553826332,
"rewards-max/forensics_tempsamp_iou_reward": 0.9829583466053009,
"rewards/forensics_tempsamp_iou_reward": 0.5862661898136139,
"step": 905
},
{
"completion_length": 21.390625,
"epoch": 3.790794979079498,
"grad_norm": 4.122274875640869,
"kl": 0.3388671875,
"learning_rate": 5.334728033472803e-08,
"loss": 0.0135,
"reward": 0.6799066364765167,
"reward_std": 0.12105746194720268,
"reward_std_all": 0.11911188066005707,
"rewards-max/forensics_tempsamp_iou_reward": 0.9469323456287384,
"rewards/forensics_tempsamp_iou_reward": 0.6715587973594666,
"step": 906
},
{
"completion_length": 26.671875,
"epoch": 3.794979079497908,
"grad_norm": 3.6313772201538086,
"kl": 0.306640625,
"learning_rate": 5.2301255230125516e-08,
"loss": 0.0123,
"reward": 0.5989182889461517,
"reward_std": 0.11509798467159271,
"reward_std_all": 0.14705341309309006,
"rewards-max/forensics_tempsamp_iou_reward": 0.9383317828178406,
"rewards/forensics_tempsamp_iou_reward": 0.5682941675186157,
"step": 907
},
{
"completion_length": 20.90625,
"epoch": 3.799163179916318,
"grad_norm": 3.448683500289917,
"kl": 0.3232421875,
"learning_rate": 5.1255230125523015e-08,
"loss": 0.0129,
"reward": 0.6090238094329834,
"reward_std": 0.07477807626128197,
"reward_std_all": 0.12511344999074936,
"rewards-max/forensics_tempsamp_iou_reward": 0.9866025149822235,
"rewards/forensics_tempsamp_iou_reward": 0.5885888636112213,
"step": 908
},
{
"completion_length": 22.6875,
"epoch": 3.803347280334728,
"grad_norm": 4.049531936645508,
"kl": 0.29296875,
"learning_rate": 5.02092050209205e-08,
"loss": 0.0117,
"reward": 0.6942583322525024,
"reward_std": 0.09567267447710037,
"reward_std_all": 0.09872192144393921,
"rewards-max/forensics_tempsamp_iou_reward": 0.9756803512573242,
"rewards/forensics_tempsamp_iou_reward": 0.7123798727989197,
"step": 909
},
{
"completion_length": 20.765625,
"epoch": 3.8075313807531384,
"grad_norm": 3.564295768737793,
"kl": 0.34375,
"learning_rate": 4.9163179916317995e-08,
"loss": 0.0138,
"reward": 0.7101831138134003,
"reward_std": 0.07434308901429176,
"reward_std_all": 0.07908477634191513,
"rewards-max/forensics_tempsamp_iou_reward": 0.9816188216209412,
"rewards/forensics_tempsamp_iou_reward": 0.7502680420875549,
"step": 910
},
{
"completion_length": 24.265625,
"epoch": 3.8117154811715483,
"grad_norm": 4.3151350021362305,
"kl": 0.3251953125,
"learning_rate": 4.811715481171548e-08,
"loss": 0.013,
"reward": 0.6212729513645172,
"reward_std": 0.10283070430159569,
"reward_std_all": 0.13012075424194336,
"rewards-max/forensics_tempsamp_iou_reward": 0.9383687078952789,
"rewards/forensics_tempsamp_iou_reward": 0.5997264683246613,
"step": 911
},
{
"completion_length": 21.140625,
"epoch": 3.8158995815899583,
"grad_norm": 3.866506576538086,
"kl": 0.3017578125,
"learning_rate": 4.707112970711297e-08,
"loss": 0.0121,
"reward": 0.6634983718395233,
"reward_std": 0.061267681419849396,
"reward_std_all": 0.08980569988489151,
"rewards-max/forensics_tempsamp_iou_reward": 0.9707653522491455,
"rewards/forensics_tempsamp_iou_reward": 0.6450428366661072,
"step": 912
},
{
"completion_length": 23.46875,
"epoch": 3.8200836820083683,
"grad_norm": 4.311226844787598,
"kl": 0.357421875,
"learning_rate": 4.602510460251046e-08,
"loss": 0.0143,
"reward": 0.6951288878917694,
"reward_std": 0.07971426844596863,
"reward_std_all": 0.0869264304637909,
"rewards-max/forensics_tempsamp_iou_reward": 0.9606405794620514,
"rewards/forensics_tempsamp_iou_reward": 0.6808315515518188,
"step": 913
},
{
"completion_length": 20.671875,
"epoch": 3.8242677824267783,
"grad_norm": 3.3142428398132324,
"kl": 0.283203125,
"learning_rate": 4.497907949790795e-08,
"loss": 0.0113,
"reward": 0.6467280387878418,
"reward_std": 0.09999274834990501,
"reward_std_all": 0.1211121566593647,
"rewards-max/forensics_tempsamp_iou_reward": 0.9792348742485046,
"rewards/forensics_tempsamp_iou_reward": 0.6356964707374573,
"step": 914
},
{
"completion_length": 20.9375,
"epoch": 3.8284518828451883,
"grad_norm": 3.08097243309021,
"kl": 0.31640625,
"learning_rate": 4.393305439330543e-08,
"loss": 0.0126,
"reward": 0.6532867848873138,
"reward_std": 0.13906599581241608,
"reward_std_all": 0.14295721799135208,
"rewards-max/forensics_tempsamp_iou_reward": 0.9749281108379364,
"rewards/forensics_tempsamp_iou_reward": 0.6484959125518799,
"step": 915
},
{
"completion_length": 23.296875,
"epoch": 3.8326359832635983,
"grad_norm": 3.5442888736724854,
"kl": 0.3427734375,
"learning_rate": 4.2887029288702926e-08,
"loss": 0.0137,
"reward": 0.6217423677444458,
"reward_std": 0.15066073089838028,
"reward_std_all": 0.15930892527103424,
"rewards-max/forensics_tempsamp_iou_reward": 0.9339231550693512,
"rewards/forensics_tempsamp_iou_reward": 0.6054432690143585,
"step": 916
},
{
"completion_length": 22.96875,
"epoch": 3.8368200836820083,
"grad_norm": 33.69537353515625,
"kl": 1.0283203125,
"learning_rate": 4.184100418410041e-08,
"loss": 0.041,
"reward": 0.6946248412132263,
"reward_std": 0.07816535979509354,
"reward_std_all": 0.08712641149759293,
"rewards-max/forensics_tempsamp_iou_reward": 0.9700241386890411,
"rewards/forensics_tempsamp_iou_reward": 0.6885211765766144,
"step": 917
},
{
"completion_length": 23.515625,
"epoch": 3.8410041841004183,
"grad_norm": 3.3551807403564453,
"kl": 0.3759765625,
"learning_rate": 4.0794979079497906e-08,
"loss": 0.0151,
"reward": 0.6493045091629028,
"reward_std": 0.13101234659552574,
"reward_std_all": 0.13917552679777145,
"rewards-max/forensics_tempsamp_iou_reward": 0.985296756029129,
"rewards/forensics_tempsamp_iou_reward": 0.652436226606369,
"step": 918
},
{
"completion_length": 21.625,
"epoch": 3.8451882845188283,
"grad_norm": 4.112758159637451,
"kl": 0.4072265625,
"learning_rate": 3.974895397489539e-08,
"loss": 0.0163,
"reward": 0.5832429528236389,
"reward_std": 0.06439216434955597,
"reward_std_all": 0.1308731660246849,
"rewards-max/forensics_tempsamp_iou_reward": 0.975481390953064,
"rewards/forensics_tempsamp_iou_reward": 0.5578437149524689,
"step": 919
},
{
"completion_length": 24.953125,
"epoch": 3.8493723849372383,
"grad_norm": 4.198416233062744,
"kl": 0.33203125,
"learning_rate": 3.8702928870292885e-08,
"loss": 0.0133,
"reward": 0.578032523393631,
"reward_std": 0.12932458892464638,
"reward_std_all": 0.165772944688797,
"rewards-max/forensics_tempsamp_iou_reward": 0.9754117727279663,
"rewards/forensics_tempsamp_iou_reward": 0.5403764843940735,
"step": 920
},
{
"completion_length": 22.53125,
"epoch": 3.8535564853556483,
"grad_norm": 3.611314296722412,
"kl": 0.2841796875,
"learning_rate": 3.765690376569038e-08,
"loss": 0.0114,
"reward": 0.6531776189804077,
"reward_std": 0.06621057167649269,
"reward_std_all": 0.09814953431487083,
"rewards-max/forensics_tempsamp_iou_reward": 0.9589038193225861,
"rewards/forensics_tempsamp_iou_reward": 0.6304429471492767,
"step": 921
},
{
"completion_length": 23.65625,
"epoch": 3.8577405857740583,
"grad_norm": 3.1453349590301514,
"kl": 0.3408203125,
"learning_rate": 3.6610878661087864e-08,
"loss": 0.0136,
"reward": 0.612339586019516,
"reward_std": 0.1331213228404522,
"reward_std_all": 0.15164828300476074,
"rewards-max/forensics_tempsamp_iou_reward": 0.8986673653125763,
"rewards/forensics_tempsamp_iou_reward": 0.5591473281383514,
"step": 922
},
{
"completion_length": 23.375,
"epoch": 3.8619246861924688,
"grad_norm": 3.3675780296325684,
"kl": 0.2978515625,
"learning_rate": 3.556485355648536e-08,
"loss": 0.0119,
"reward": 0.6146429777145386,
"reward_std": 0.1413417086005211,
"reward_std_all": 0.15720140933990479,
"rewards-max/forensics_tempsamp_iou_reward": 0.9878578782081604,
"rewards/forensics_tempsamp_iou_reward": 0.587394654750824,
"step": 923
},
{
"completion_length": 23.359375,
"epoch": 3.8661087866108788,
"grad_norm": 3.649900436401367,
"kl": 0.298828125,
"learning_rate": 3.4518828451882844e-08,
"loss": 0.0119,
"reward": 0.6848908066749573,
"reward_std": 0.07526717334985733,
"reward_std_all": 0.08670604601502419,
"rewards-max/forensics_tempsamp_iou_reward": 0.9577077925205231,
"rewards/forensics_tempsamp_iou_reward": 0.6764519214630127,
"step": 924
},
{
"completion_length": 21.5,
"epoch": 3.8702928870292888,
"grad_norm": 5.696201324462891,
"kl": 0.4326171875,
"learning_rate": 3.347280334728033e-08,
"loss": 0.0173,
"reward": 0.7163218855857849,
"reward_std": 0.08484910801053047,
"reward_std_all": 0.08410285413265228,
"rewards-max/forensics_tempsamp_iou_reward": 0.9796935021877289,
"rewards/forensics_tempsamp_iou_reward": 0.7156825959682465,
"step": 925
},
{
"completion_length": 21.796875,
"epoch": 3.8744769874476988,
"grad_norm": 4.2482733726501465,
"kl": 0.341796875,
"learning_rate": 3.242677824267782e-08,
"loss": 0.0136,
"reward": 0.6680137813091278,
"reward_std": 0.07828610762953758,
"reward_std_all": 0.09661782532930374,
"rewards-max/forensics_tempsamp_iou_reward": 0.9428873360157013,
"rewards/forensics_tempsamp_iou_reward": 0.6637776792049408,
"step": 926
},
{
"completion_length": 20.640625,
"epoch": 3.8786610878661087,
"grad_norm": 4.880695343017578,
"kl": 0.306640625,
"learning_rate": 3.138075313807531e-08,
"loss": 0.0123,
"reward": 0.5892651975154877,
"reward_std": 0.07100336253643036,
"reward_std_all": 0.13644718751311302,
"rewards-max/forensics_tempsamp_iou_reward": 0.9687834680080414,
"rewards/forensics_tempsamp_iou_reward": 0.5680693089962006,
"step": 927
},
{
"completion_length": 21.125,
"epoch": 3.8828451882845187,
"grad_norm": 3.957054615020752,
"kl": 0.287109375,
"learning_rate": 3.03347280334728e-08,
"loss": 0.0115,
"reward": 0.6973958313465118,
"reward_std": 0.09553158655762672,
"reward_std_all": 0.10078664124011993,
"rewards-max/forensics_tempsamp_iou_reward": 0.9921888411045074,
"rewards/forensics_tempsamp_iou_reward": 0.7156790196895599,
"step": 928
},
{
"completion_length": 21.375,
"epoch": 3.8870292887029287,
"grad_norm": 4.0022406578063965,
"kl": 0.291015625,
"learning_rate": 2.9288702928870292e-08,
"loss": 0.0116,
"reward": 0.6788120567798615,
"reward_std": 0.03657440282404423,
"reward_std_all": 0.0726332776248455,
"rewards-max/forensics_tempsamp_iou_reward": 0.991425633430481,
"rewards/forensics_tempsamp_iou_reward": 0.7121764421463013,
"step": 929
},
{
"completion_length": 21.4375,
"epoch": 3.891213389121339,
"grad_norm": 3.970301628112793,
"kl": 0.3076171875,
"learning_rate": 2.8242677824267782e-08,
"loss": 0.0123,
"reward": 0.6575943231582642,
"reward_std": 0.050754569470882416,
"reward_std_all": 0.09140124917030334,
"rewards-max/forensics_tempsamp_iou_reward": 0.9812597930431366,
"rewards/forensics_tempsamp_iou_reward": 0.6476146578788757,
"step": 930
},
{
"completion_length": 22.5,
"epoch": 3.895397489539749,
"grad_norm": 4.57827615737915,
"kl": 0.3486328125,
"learning_rate": 2.719665271966527e-08,
"loss": 0.0139,
"reward": 0.5588297545909882,
"reward_std": 0.09875202551484108,
"reward_std_all": 0.16188780590891838,
"rewards-max/forensics_tempsamp_iou_reward": 0.931830644607544,
"rewards/forensics_tempsamp_iou_reward": 0.5228331685066223,
"step": 931
},
{
"completion_length": 24.203125,
"epoch": 3.899581589958159,
"grad_norm": 3.895090103149414,
"kl": 0.2978515625,
"learning_rate": 2.6150627615062758e-08,
"loss": 0.0119,
"reward": 0.6733531951904297,
"reward_std": 0.05230311118066311,
"reward_std_all": 0.08610033243894577,
"rewards-max/forensics_tempsamp_iou_reward": 0.9852098226547241,
"rewards/forensics_tempsamp_iou_reward": 0.6775158047676086,
"step": 932
},
{
"completion_length": 23.90625,
"epoch": 3.903765690376569,
"grad_norm": 3.4022293090820312,
"kl": 0.2646484375,
"learning_rate": 2.510460251046025e-08,
"loss": 0.0106,
"reward": 0.6822870671749115,
"reward_std": 0.0841967836022377,
"reward_std_all": 0.09335251525044441,
"rewards-max/forensics_tempsamp_iou_reward": 0.9679335951805115,
"rewards/forensics_tempsamp_iou_reward": 0.6719159781932831,
"step": 933
},
{
"completion_length": 22.125,
"epoch": 3.907949790794979,
"grad_norm": 3.9217264652252197,
"kl": 0.22802734375,
"learning_rate": 2.405857740585774e-08,
"loss": 0.0091,
"reward": 0.5825020968914032,
"reward_std": 0.09399706497788429,
"reward_std_all": 0.14431558549404144,
"rewards-max/forensics_tempsamp_iou_reward": 0.9819277226924896,
"rewards/forensics_tempsamp_iou_reward": 0.5595098733901978,
"step": 934
},
{
"completion_length": 22.015625,
"epoch": 3.912133891213389,
"grad_norm": 4.365074634552002,
"kl": 0.3828125,
"learning_rate": 2.301255230125523e-08,
"loss": 0.0153,
"reward": 0.6624384224414825,
"reward_std": 0.08193648420274258,
"reward_std_all": 0.10500656813383102,
"rewards-max/forensics_tempsamp_iou_reward": 0.9284738004207611,
"rewards/forensics_tempsamp_iou_reward": 0.6390271484851837,
"step": 935
},
{
"completion_length": 20.875,
"epoch": 3.916317991631799,
"grad_norm": 4.2274322509765625,
"kl": 0.4130859375,
"learning_rate": 2.1966527196652717e-08,
"loss": 0.0165,
"reward": 0.6611848175525665,
"reward_std": 0.07888626866042614,
"reward_std_all": 0.09993824735283852,
"rewards-max/forensics_tempsamp_iou_reward": 0.9921667277812958,
"rewards/forensics_tempsamp_iou_reward": 0.6490730345249176,
"step": 936
},
{
"completion_length": 25.890625,
"epoch": 3.920502092050209,
"grad_norm": 3.7898221015930176,
"kl": 0.279296875,
"learning_rate": 2.0920502092050206e-08,
"loss": 0.0111,
"reward": 0.6077411472797394,
"reward_std": 0.08884499222040176,
"reward_std_all": 0.1312670148909092,
"rewards-max/forensics_tempsamp_iou_reward": 0.9812169373035431,
"rewards/forensics_tempsamp_iou_reward": 0.605519026517868,
"step": 937
},
{
"completion_length": 26.15625,
"epoch": 3.924686192468619,
"grad_norm": 3.4783947467803955,
"kl": 0.359375,
"learning_rate": 1.9874476987447696e-08,
"loss": 0.0144,
"reward": 0.5737885534763336,
"reward_std": 0.10801371932029724,
"reward_std_all": 0.15838073194026947,
"rewards-max/forensics_tempsamp_iou_reward": 0.9669712483882904,
"rewards/forensics_tempsamp_iou_reward": 0.5519327521324158,
"step": 938
},
{
"completion_length": 22.984375,
"epoch": 3.928870292887029,
"grad_norm": 4.137996196746826,
"kl": 0.4013671875,
"learning_rate": 1.882845188284519e-08,
"loss": 0.016,
"reward": 0.5872654914855957,
"reward_std": 0.12187778577208519,
"reward_std_all": 0.1577468328177929,
"rewards-max/forensics_tempsamp_iou_reward": 0.9860146343708038,
"rewards/forensics_tempsamp_iou_reward": 0.5738124549388885,
"step": 939
},
{
"completion_length": 23.84375,
"epoch": 3.933054393305439,
"grad_norm": 5.695178508758545,
"kl": 0.421875,
"learning_rate": 1.778242677824268e-08,
"loss": 0.017,
"reward": 0.6048561036586761,
"reward_std": 0.1106899194419384,
"reward_std_all": 0.14769892394542694,
"rewards-max/forensics_tempsamp_iou_reward": 0.9272052645683289,
"rewards/forensics_tempsamp_iou_reward": 0.56010502576828,
"step": 940
},
{
"completion_length": 22.890625,
"epoch": 3.937238493723849,
"grad_norm": 3.68876576423645,
"kl": 0.2998046875,
"learning_rate": 1.6736401673640165e-08,
"loss": 0.012,
"reward": 0.6464340388774872,
"reward_std": 0.052803413942456245,
"reward_std_all": 0.09727136418223381,
"rewards-max/forensics_tempsamp_iou_reward": 0.9612617790699005,
"rewards/forensics_tempsamp_iou_reward": 0.6423720121383667,
"step": 941
},
{
"completion_length": 25.296875,
"epoch": 3.941422594142259,
"grad_norm": 3.633321523666382,
"kl": 0.369140625,
"learning_rate": 1.5690376569037655e-08,
"loss": 0.0148,
"reward": 0.6459762454032898,
"reward_std": 0.11734186112880707,
"reward_std_all": 0.12514468654990196,
"rewards-max/forensics_tempsamp_iou_reward": 0.9622135162353516,
"rewards/forensics_tempsamp_iou_reward": 0.6274979710578918,
"step": 942
},
{
"completion_length": 19.625,
"epoch": 3.9456066945606696,
"grad_norm": 4.421983242034912,
"kl": 0.3134765625,
"learning_rate": 1.4644351464435146e-08,
"loss": 0.0125,
"reward": 0.6432606279850006,
"reward_std": 0.0786641277372837,
"reward_std_all": 0.11154136806726456,
"rewards-max/forensics_tempsamp_iou_reward": 0.9799275696277618,
"rewards/forensics_tempsamp_iou_reward": 0.6429593861103058,
"step": 943
},
{
"completion_length": 21.421875,
"epoch": 3.9497907949790796,
"grad_norm": 4.0187153816223145,
"kl": 0.298828125,
"learning_rate": 1.3598326359832636e-08,
"loss": 0.0119,
"reward": 0.7106819450855255,
"reward_std": 0.09048251435160637,
"reward_std_all": 0.09315777942538261,
"rewards-max/forensics_tempsamp_iou_reward": 0.9679808020591736,
"rewards/forensics_tempsamp_iou_reward": 0.7335850596427917,
"step": 944
},
{
"completion_length": 23.375,
"epoch": 3.9539748953974896,
"grad_norm": 4.419652938842773,
"kl": 0.26416015625,
"learning_rate": 1.2552301255230125e-08,
"loss": 0.0106,
"reward": 0.5755804479122162,
"reward_std": 0.11616823077201843,
"reward_std_all": 0.15799596905708313,
"rewards-max/forensics_tempsamp_iou_reward": 0.9661780297756195,
"rewards/forensics_tempsamp_iou_reward": 0.5476838052272797,
"step": 945
},
{
"completion_length": 23.9375,
"epoch": 3.9581589958158996,
"grad_norm": 3.9612879753112793,
"kl": 0.28955078125,
"learning_rate": 1.1506276150627615e-08,
"loss": 0.0116,
"reward": 0.6181075572967529,
"reward_std": 0.11996693909168243,
"reward_std_all": 0.14073332771658897,
"rewards-max/forensics_tempsamp_iou_reward": 0.9677874743938446,
"rewards/forensics_tempsamp_iou_reward": 0.5825429558753967,
"step": 946
},
{
"completion_length": 21.53125,
"epoch": 3.9623430962343096,
"grad_norm": 4.088210582733154,
"kl": 0.33984375,
"learning_rate": 1.0460251046025103e-08,
"loss": 0.0136,
"reward": 0.6096947193145752,
"reward_std": 0.08745855651795864,
"reward_std_all": 0.12916648015379906,
"rewards-max/forensics_tempsamp_iou_reward": 0.9378291070461273,
"rewards/forensics_tempsamp_iou_reward": 0.5722188055515289,
"step": 947
},
{
"completion_length": 20.859375,
"epoch": 3.9665271966527196,
"grad_norm": 4.124889373779297,
"kl": 0.29150390625,
"learning_rate": 9.414225941422594e-09,
"loss": 0.0117,
"reward": 0.6375137269496918,
"reward_std": 0.0933135375380516,
"reward_std_all": 0.11896751821041107,
"rewards-max/forensics_tempsamp_iou_reward": 0.9327073693275452,
"rewards/forensics_tempsamp_iou_reward": 0.5956218838691711,
"step": 948
},
{
"completion_length": 20.46875,
"epoch": 3.9707112970711296,
"grad_norm": 4.608794212341309,
"kl": 0.2900390625,
"learning_rate": 8.368200836820083e-09,
"loss": 0.0116,
"reward": 0.6584287285804749,
"reward_std": 0.1122908815741539,
"reward_std_all": 0.12204866111278534,
"rewards-max/forensics_tempsamp_iou_reward": 0.9188196957111359,
"rewards/forensics_tempsamp_iou_reward": 0.630940318107605,
"step": 949
},
{
"completion_length": 20.984375,
"epoch": 3.97489539748954,
"grad_norm": 4.057939529418945,
"kl": 0.3857421875,
"learning_rate": 7.322175732217573e-09,
"loss": 0.0154,
"reward": 0.6136142313480377,
"reward_std": 0.08369151130318642,
"reward_std_all": 0.12873724848031998,
"rewards-max/forensics_tempsamp_iou_reward": 0.9150943458080292,
"rewards/forensics_tempsamp_iou_reward": 0.5841339826583862,
"step": 950
},
{
"completion_length": 19.5625,
"epoch": 3.97907949790795,
"grad_norm": 4.380617141723633,
"kl": 0.28515625,
"learning_rate": 6.276150627615063e-09,
"loss": 0.0114,
"reward": 0.7000117599964142,
"reward_std": 0.06516311690211296,
"reward_std_all": 0.0776391550898552,
"rewards-max/forensics_tempsamp_iou_reward": 0.9619520902633667,
"rewards/forensics_tempsamp_iou_reward": 0.7163766026496887,
"step": 951
},
{
"completion_length": 19.8125,
"epoch": 3.98326359832636,
"grad_norm": 4.497678279876709,
"kl": 0.3154296875,
"learning_rate": 5.2301255230125516e-09,
"loss": 0.0126,
"reward": 0.6311633884906769,
"reward_std": 0.1011255644261837,
"reward_std_all": 0.12433922290802002,
"rewards-max/forensics_tempsamp_iou_reward": 0.9924454689025879,
"rewards/forensics_tempsamp_iou_reward": 0.6102850437164307,
"step": 952
},
{
"completion_length": 21.65625,
"epoch": 3.98744769874477,
"grad_norm": 3.9524948596954346,
"kl": 0.333984375,
"learning_rate": 4.184100418410041e-09,
"loss": 0.0134,
"reward": 0.6072176396846771,
"reward_std": 0.11647656001150608,
"reward_std_all": 0.15526938065886497,
"rewards-max/forensics_tempsamp_iou_reward": 0.909515380859375,
"rewards/forensics_tempsamp_iou_reward": 0.5656242966651917,
"step": 953
},
{
"completion_length": 20.875,
"epoch": 3.99163179916318,
"grad_norm": 3.17622709274292,
"kl": 0.37890625,
"learning_rate": 3.1380753138075314e-09,
"loss": 0.0152,
"reward": 0.7282925844192505,
"reward_std": 0.07994503900408745,
"reward_std_all": 0.07912097871303558,
"rewards-max/forensics_tempsamp_iou_reward": 0.9914014637470245,
"rewards/forensics_tempsamp_iou_reward": 0.7788703441619873,
"step": 954
},
{
"completion_length": 21.75,
"epoch": 3.99581589958159,
"grad_norm": 4.384639263153076,
"kl": 0.3056640625,
"learning_rate": 2.0920502092050206e-09,
"loss": 0.0122,
"reward": 0.6780132055282593,
"reward_std": 0.08848053961992264,
"reward_std_all": 0.10389265045523643,
"rewards-max/forensics_tempsamp_iou_reward": 0.9830898344516754,
"rewards/forensics_tempsamp_iou_reward": 0.6761464476585388,
"step": 955
},
{
"completion_length": 22.71428680419922,
"epoch": 4.0,
"grad_norm": 3.990787982940674,
"kl": 0.38671875,
"learning_rate": 1.0460251046025103e-09,
"loss": 0.0156,
"reward": 0.6871524751186371,
"reward_std": 0.05144812352955341,
"reward_std_all": 0.06930763646960258,
"rewards-max/forensics_tempsamp_iou_reward": 0.8512924015522003,
"rewards/forensics_tempsamp_iou_reward": 0.6577990353107452,
"step": 956
}
],
"logging_steps": 1.0,
"max_steps": 956,
"num_input_tokens_seen": 0,
"num_train_epochs": 4,
"save_steps": 30,
"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
}