GavinChan1105/X-R1-3B-cn-math

1{2  "best_metric": null,3  "best_model_checkpoint": null,4  "epoch": 1.0,5  "eval_steps": 10,6  "global_step": 750,7  "is_hyper_param_search": false,8  "is_local_process_zero": true,9  "is_world_process_zero": true,10  "log_history": [11    {12      "clip_ratio": 0.0,13      "completion_length": 298.3437587738037,14      "epoch": 0.013333333333333334,15      "grad_norm": 0.16560879349708557,16      "kl": 0.00018159300088882446,17      "learning_rate": 4e-07,18      "loss": 0.0049,19      "reward": 0.15208333674818278,20      "reward_std": 0.2552751675248146,21      "rewards/accuracy_reward": 0.06666666846722365,22      "rewards/format_reward": 0.08541666883975267,23      "step": 1024    },25    {26      "clip_ratio": 0.0,27      "completion_length": 314.44792423248293,28      "epoch": 0.02666666666666667,29      "grad_norm": 0.2575075328350067,30      "kl": 0.0009123936295509338,31      "learning_rate": 8e-07,32      "loss": 0.0058,33      "reward": 0.20625000558793544,34      "reward_std": 0.2727560464292765,35      "rewards/accuracy_reward": 0.06458333488553762,36      "rewards/format_reward": 0.14166667070239783,37      "step": 2038    },39    {40      "clip_ratio": 0.0,41      "completion_length": 252.6500078201294,42      "epoch": 0.04,43      "grad_norm": 0.46988776326179504,44      "kl": 0.03377872705459595,45      "learning_rate": 1.2000000000000002e-06,46      "loss": 0.0205,47      "reward": 0.5270833471789956,48      "reward_std": 0.4601708807051182,49      "rewards/accuracy_reward": 0.07916666902601718,50      "rewards/format_reward": 0.44791667852550743,51      "step": 3052    },53    {54      "clip_ratio": 0.0,55      "completion_length": 242.21250705718995,56      "epoch": 0.05333333333333334,57      "grad_norm": 0.19061416387557983,58      "kl": 0.06111354827880859,59      "learning_rate": 1.6e-06,60      "loss": 0.048,61      "reward": 0.8041666891425848,62      "reward_std": 0.4267156172543764,63      "rewards/accuracy_reward": 0.06250000167638063,64      "rewards/format_reward": 0.7416666835546494,65      "step": 4066    },67    {68      "clip_ratio": 0.0,69      "completion_length": 197.73125562667846,70      "epoch": 0.06666666666666667,71      "grad_norm": 0.2476302981376648,72      "kl": 0.0408905029296875,73      "learning_rate": 2e-06,74      "loss": 0.021,75      "reward": 0.8895833583548665,76      "reward_std": 0.4103152878582478,77      "rewards/accuracy_reward": 0.09375000242143869,78      "rewards/format_reward": 0.7958333509042859,79      "step": 5080    },81    {82      "clip_ratio": 0.0,83      "completion_length": 198.86250495910645,84      "epoch": 0.08,85      "grad_norm": 0.13683611154556274,86      "kl": 0.0466217041015625,87      "learning_rate": 2.4000000000000003e-06,88      "loss": 0.0316,89      "reward": 0.9916666936129331,90      "reward_std": 0.36507078595459463,91      "rewards/accuracy_reward": 0.11666666958481073,92      "rewards/format_reward": 0.8750000182539225,93      "step": 6094    },95    {96      "clip_ratio": 0.0,97      "completion_length": 225.4541726589203,98      "epoch": 0.09333333333333334,99      "grad_norm": 0.17669692635536194,100      "kl": 0.06876106262207031,101      "learning_rate": 2.8000000000000003e-06,102      "loss": 0.0388,103      "reward": 1.0520833637565374,104      "reward_std": 0.4207858666777611,105      "rewards/accuracy_reward": 0.19375000540167092,106      "rewards/format_reward": 0.8583333497866988,107      "step": 70108    },109    {110      "clip_ratio": 0.0,111      "completion_length": 194.30625553131102,112      "epoch": 0.10666666666666667,113      "grad_norm": 0.15089309215545654,114      "kl": 0.08316802978515625,115      "learning_rate": 2.9995938617691924e-06,116      "loss": 0.03,117      "reward": 1.1625000346451997,118      "reward_std": 0.44051293805241587,119      "rewards/accuracy_reward": 0.2520833399146795,120      "rewards/format_reward": 0.91041667945683,121      "step": 80122    },123    {124      "clip_ratio": 0.0,125      "completion_length": 202.31667232513428,126      "epoch": 0.12,127      "grad_norm": 0.2549769878387451,128      "kl": 0.06892852783203125,129      "learning_rate": 2.9963460753897363e-06,130      "loss": 0.0229,131      "reward": 1.1770833685994149,132      "reward_std": 0.4563848368823528,133      "rewards/accuracy_reward": 0.2687500063329935,134      "rewards/format_reward": 0.9083333484828472,135      "step": 90136    },137    {138      "clip_ratio": 0.0,139      "completion_length": 215.91458954811097,140      "epoch": 0.13333333333333333,141      "grad_norm": 3.687699556350708,142      "kl": 0.1230712890625,143      "learning_rate": 2.989857536612915e-06,144      "loss": 0.0166,145      "reward": 1.235416704416275,146      "reward_std": 0.4576607421040535,147      "rewards/accuracy_reward": 0.32083334121853113,148      "rewards/format_reward": 0.914583346247673,149      "step": 100150    },151    {152      "clip_ratio": 0.0,153      "completion_length": 231.20208988189697,154      "epoch": 0.14666666666666667,155      "grad_norm": 0.19754290580749512,156      "kl": 0.09471588134765625,157      "learning_rate": 2.980142298168869e-06,158      "loss": 0.0774,159      "reward": 1.204166702181101,160      "reward_std": 0.5094456784427166,161      "rewards/accuracy_reward": 0.329166673310101,162      "rewards/format_reward": 0.8750000186264515,163      "step": 110164    },165    {166      "clip_ratio": 0.0,167      "completion_length": 210.604172039032,168      "epoch": 0.16,169      "grad_norm": 0.9343002438545227,170      "kl": 0.10347900390625,171      "learning_rate": 2.9672214011007086e-06,172      "loss": 0.0425,173      "reward": 1.2375000335276127,174      "reward_std": 0.5164696607738734,175      "rewards/accuracy_reward": 0.35208334103226663,176      "rewards/format_reward": 0.8854166835546493,177      "step": 120178    },179    {180      "clip_ratio": 0.0,181      "completion_length": 254.06667470932007,182      "epoch": 0.17333333333333334,183      "grad_norm": 0.25169649720191956,184      "kl": 0.148199462890625,185      "learning_rate": 2.951122829194296e-06,186      "loss": 0.0796,187      "reward": 1.1187500342726708,188      "reward_std": 0.5015288021415472,189      "rewards/accuracy_reward": 0.26250000670552254,190      "rewards/format_reward": 0.8562500193715096,191      "step": 130192    },193    {194      "clip_ratio": 0.0,195      "completion_length": 233.81875615119935,196      "epoch": 0.18666666666666668,197      "grad_norm": 0.7912704348564148,198      "kl": 2.357049560546875,199      "learning_rate": 2.9318814483715983e-06,200      "loss": 0.1295,201      "reward": 0.9895833631977439,202      "reward_std": 0.5210714556276799,203      "rewards/accuracy_reward": 0.26250000577419996,204      "rewards/format_reward": 0.7270833453163504,205      "step": 140206    },207    {208      "clip_ratio": 0.0,209      "completion_length": 227.87708988189698,210      "epoch": 0.2,211      "grad_norm": 1.25613272190094,212      "kl": 0.46229248046875,213      "learning_rate": 2.9095389311788626e-06,214      "loss": 0.0646,215      "reward": 1.129166703671217,216      "reward_std": 0.49879880994558334,217      "rewards/accuracy_reward": 0.2687500067055225,218      "rewards/format_reward": 0.8604166861623526,219      "step": 150220    },221    {222      "clip_ratio": 0.0,223      "completion_length": 194.6770890235901,224      "epoch": 0.21333333333333335,225      "grad_norm": 0.4988195300102234,226      "kl": 0.613861083984375,227      "learning_rate": 2.8841436665331635e-06,228      "loss": 0.0366,229      "reward": 1.2395833693444729,230      "reward_std": 0.48180873580276967,231      "rewards/accuracy_reward": 0.31666667591780423,232      "rewards/format_reward": 0.922916679084301,233      "step": 160234    },235    {236      "clip_ratio": 0.0,237      "completion_length": 168.8062547683716,238      "epoch": 0.22666666666666666,239      "grad_norm": 0.9171387553215027,240      "kl": 1.17796630859375,241      "learning_rate": 2.855750654922781e-06,242      "loss": 0.0156,243      "reward": 1.1541667036712169,244      "reward_std": 0.5033410575240851,245      "rewards/accuracy_reward": 0.25416667219251393,246      "rewards/format_reward": 0.9000000156462192,247      "step": 170248    },249    {250      "clip_ratio": 0.0,251      "completion_length": 241.8708402633667,252      "epoch": 0.24,253      "grad_norm": 25.64914894104004,254      "kl": 114.61384887695313,255      "learning_rate": 2.8244213892883906e-06,256      "loss": 2.7595,257      "reward": 1.181250037252903,258      "reward_std": 0.51867739520967,259      "rewards/accuracy_reward": 0.31041667331010103,260      "rewards/format_reward": 0.8708333529531955,261      "step": 180262    },263    {264      "clip_ratio": 0.0,265      "completion_length": 314.0875092506409,266      "epoch": 0.25333333333333335,267      "grad_norm": 0.8180872797966003,268      "kl": 2.062310791015625,269      "learning_rate": 2.7902237218430485e-06,270      "loss": 0.1151,271      "reward": 1.1354166999459268,272      "reward_std": 0.5186546068638563,273      "rewards/accuracy_reward": 0.26458334010094403,274      "rewards/format_reward": 0.8708333499729634,275      "step": 190276    },277    {278      "clip_ratio": 0.0,279      "completion_length": 271.76875801086425,280      "epoch": 0.26666666666666666,281      "grad_norm": 2.5058634281158447,282      "kl": 1.689434814453125,283      "learning_rate": 2.753231717119405e-06,284      "loss": 0.0826,285      "reward": 1.1770833715796472,286      "reward_std": 0.5468059632927179,287      "rewards/accuracy_reward": 0.3104166744276881,288      "rewards/format_reward": 0.8666666835546494,289      "step": 200290    },291    {292      "clip_ratio": 0.0,293      "completion_length": 189.7958378791809,294      "epoch": 0.28,295      "grad_norm": 0.8333961367607117,296      "kl": 2.188385009765625,297      "learning_rate": 2.713525491562421e-06,298      "loss": 0.159,299      "reward": 1.1291666947305203,300      "reward_std": 0.48742703087627887,301      "rewards/accuracy_reward": 0.2375000048428774,302      "rewards/format_reward": 0.8916666850447654,303      "step": 210304    },305    {306      "clip_ratio": 0.0,307      "completion_length": 169.8958384513855,308      "epoch": 0.29333333333333333,309      "grad_norm": 0.48866531252861023,310      "kl": 1.365594482421875,311      "learning_rate": 2.671191040014989e-06,312      "loss": 0.1571,313      "reward": 1.214583370089531,314      "reward_std": 0.45735101476311685,315      "rewards/accuracy_reward": 0.2812500067055225,316      "rewards/format_reward": 0.9333333462476731,317      "step": 220318    },319    {320      "clip_ratio": 0.0,321      "completion_length": 210.16042280197144,322      "epoch": 0.30666666666666664,323      "grad_norm": 0.4208928346633911,324      "kl": 1.861163330078125,325      "learning_rate": 2.626320049472249e-06,326      "loss": 0.2113,327      "reward": 1.1687500353902578,328      "reward_std": 0.48322329856455326,329      "rewards/accuracy_reward": 0.3020833406597376,330      "rewards/format_reward": 0.866666679084301,331      "step": 230332    },333    {334      "clip_ratio": 0.0,335      "completion_length": 213.62292337417603,336      "epoch": 0.32,337      "grad_norm": 0.9921014308929443,338      "kl": 5.52261962890625,339      "learning_rate": 2.5790097005079765e-06,340      "loss": 0.4186,341      "reward": 1.1104166995733977,342      "reward_std": 0.5192686680704355,343      "rewards/accuracy_reward": 0.26250000651925803,344      "rewards/format_reward": 0.8479166831821203,345      "step": 240346    },347    {348      "clip_ratio": 0.0,349      "completion_length": 173.20625500679017,350      "epoch": 0.3333333333333333,351      "grad_norm": 0.5788355469703674,352      "kl": 0.586700439453125,353      "learning_rate": 2.529362456803101e-06,354      "loss": 0.0873,355      "reward": 1.2354166999459266,356      "reward_std": 0.5144322618842125,357      "rewards/accuracy_reward": 0.32708334140479567,358      "rewards/format_reward": 0.9083333492279053,359      "step": 250360    },361    {362      "clip_ratio": 0.0,363      "completion_length": 170.66458702087402,364      "epoch": 0.3466666666666667,365      "grad_norm": 1.29031503200531,366      "kl": 3.8255126953125,367      "learning_rate": 2.477485843232183e-06,368      "loss": 0.2837,369      "reward": 1.2604167006909848,370      "reward_std": 0.5114245742559433,371      "rewards/accuracy_reward": 0.3416666738688946,372      "rewards/format_reward": 0.918750012665987,373      "step": 260374    },375    {376      "clip_ratio": 0.0,377      "completion_length": 167.50208835601808,378      "epoch": 0.36,379      "grad_norm": 0.6127444505691528,380      "kl": 0.5264404296875,381      "learning_rate": 2.4234922129884873e-06,382      "loss": 0.0213,383      "reward": 1.3562500387430192,384      "reward_std": 0.42282451651990416,385      "rewards/accuracy_reward": 0.38958334233611824,386      "rewards/format_reward": 0.9666666738688946,387      "step": 270388    },389    {390      "clip_ratio": 0.0,391      "completion_length": 177.09583835601808,392      "epoch": 0.37333333333333335,393      "grad_norm": 0.6640939712524414,394      "kl": 1.479010009765625,395      "learning_rate": 2.36749850425198e-06,396      "loss": 0.1688,397      "reward": 1.250000035762787,398      "reward_std": 0.4655508127063513,399      "rewards/accuracy_reward": 0.3479166755452752,400      "rewards/format_reward": 0.9020833499729634,401      "step": 280402    },403    {404      "clip_ratio": 0.0,405      "completion_length": 198.85000591278077,406      "epoch": 0.38666666666666666,407      "grad_norm": 0.14313329756259918,408      "kl": 1.668316650390625,409      "learning_rate": 2.3096259869272697e-06,410      "loss": 0.1292,411      "reward": 1.289583370834589,412      "reward_std": 0.47122009098529816,413      "rewards/accuracy_reward": 0.3645833419635892,414      "rewards/format_reward": 0.9250000134110451,415      "step": 290416    },417    {418      "clip_ratio": 0.0,419      "completion_length": 230.99375743865966,420      "epoch": 0.4,421      "grad_norm": 0.10921728610992432,422      "kl": 0.85283203125,423      "learning_rate": 2.25e-06,424      "loss": 0.0644,425      "reward": 1.3041667066514493,426      "reward_std": 0.4090763859450817,427      "rewards/accuracy_reward": 0.37500000894069674,428      "rewards/format_reward": 0.9291666768491268,429      "step": 300430    },431    {432      "clip_ratio": 0.0,433      "completion_length": 270.9208402633667,434      "epoch": 0.41333333333333333,435      "grad_norm": 1.005725383758545,436      "kl": 0.2567535400390625,437      "learning_rate": 2.1887496800805174e-06,438      "loss": 0.035,439      "reward": 1.333333370089531,440      "reward_std": 0.46985283866524696,441      "rewards/accuracy_reward": 0.3937500087544322,442      "rewards/format_reward": 0.939583346992731,443      "step": 310444    },445    {446      "clip_ratio": 0.0,447      "completion_length": 238.92500734329224,448      "epoch": 0.4266666666666667,449      "grad_norm": 1.076439380645752,450      "kl": 1.3035552978515625,451      "learning_rate": 2.126007681722727e-06,452      "loss": 0.1024,453      "reward": 1.2500000342726707,454      "reward_std": 0.48043784201145173,455      "rewards/accuracy_reward": 0.35625000931322576,456      "rewards/format_reward": 0.8937500141561031,457      "step": 320458    },459    {460      "clip_ratio": 0.0,461      "completion_length": 414.82501125335693,462      "epoch": 0.44,463      "grad_norm": 0.6154415011405945,464      "kl": 4.590509033203125,465      "learning_rate": 2.061909890123868e-06,466      "loss": 0.2873,467      "reward": 0.777083358168602,468      "reward_std": 0.4854964125901461,469      "rewards/accuracy_reward": 0.22083333935588598,470      "rewards/format_reward": 0.5562500126659871,471      "step": 330472    },473    {474      "clip_ratio": 0.0,475      "completion_length": 237.14375581741334,476      "epoch": 0.4533333333333333,477      "grad_norm": 0.4070974886417389,478      "kl": 0.4917999267578125,479      "learning_rate": 1.9965951268274372e-06,480      "loss": 0.0908,481      "reward": 1.347916703671217,482      "reward_std": 0.5186698414385319,483      "rewards/accuracy_reward": 0.4395833432674408,484      "rewards/format_reward": 0.9083333507180213,485      "step": 340486    },487    {488      "clip_ratio": 0.0,489      "completion_length": 211.02292304039003,490      "epoch": 0.4666666666666667,491      "grad_norm": 0.7467241883277893,492      "kl": 1.289691162109375,493      "learning_rate": 1.9302048490666355e-06,494      "loss": 0.1216,495      "reward": 1.3833333637565375,496      "reward_std": 0.4452923540025949,497      "rewards/accuracy_reward": 0.4604166755452752,498      "rewards/format_reward": 0.9229166809469461,499      "step": 350500    },501    {502      "clip_ratio": 0.0,503      "completion_length": 207.78333921432494,504      "epoch": 0.48,505      "grad_norm": 0.5213920474052429,506      "kl": 0.3299652099609375,507      "learning_rate": 1.8628828433995015e-06,508      "loss": 0.0802,509      "reward": 1.354166705906391,510      "reward_std": 0.43137194626033304,511      "rewards/accuracy_reward": 0.4145833427086473,512      "rewards/format_reward": 0.9395833432674408,513      "step": 360514    },515    {516      "clip_ratio": 0.0,517      "completion_length": 214.07083945274354,518      "epoch": 0.49333333333333335,519      "grad_norm": 11.964993476867676,520      "kl": 1.5061492919921875,521      "learning_rate": 1.7947749142992453e-06,522      "loss": 0.1615,523      "reward": 1.2437500320374966,524      "reward_std": 0.4532905198633671,525      "rewards/accuracy_reward": 0.3145833395421505,526      "rewards/format_reward": 0.9291666775941849,527      "step": 370528    },529    {530      "clip_ratio": 0.0,531      "completion_length": 231.10625724792482,532      "epoch": 0.5066666666666667,533      "grad_norm": 0.7543414235115051,534      "kl": 0.93128662109375,535      "learning_rate": 1.7260285683742248e-06,536      "loss": 0.1419,537      "reward": 1.3000000320374965,538      "reward_std": 0.5549088928848505,539      "rewards/accuracy_reward": 0.4020833412185311,540      "rewards/format_reward": 0.8979166842997074,541      "step": 380542    },543    {544      "clip_ratio": 0.0,545      "completion_length": 234.94167280197144,546      "epoch": 0.52,547      "grad_norm": 1.8742121458053589,548      "kl": 37.191220092773435,549      "learning_rate": 1.6567926949014804e-06,550      "loss": 0.5318,551      "reward": 1.3187500394880771,552      "reward_std": 0.5167587421834469,553      "rewards/accuracy_reward": 0.40833334214985373,554      "rewards/format_reward": 0.9104166820645332,555      "step": 390556    },557    {558      "clip_ratio": 0.0,559      "completion_length": 188.943754863739,560      "epoch": 0.5333333333333333,561      "grad_norm": 0.14478746056556702,562      "kl": 0.573187255859375,563      "learning_rate": 1.5872172433657137e-06,564      "loss": 0.0723,565      "reward": 1.3020833745598792,566      "reward_std": 0.4792703174054623,567      "rewards/accuracy_reward": 0.3645833421498537,568      "rewards/format_reward": 0.9375000111758709,569      "step": 400570    },571    {572      "clip_ratio": 0.0,573      "completion_length": 178.75833768844603,574      "epoch": 0.5466666666666666,575      "grad_norm": 1.8087667226791382,576      "kl": 0.8369659423828125,577      "learning_rate": 1.5174528987020958e-06,578      "loss": 0.0923,579      "reward": 1.3125000409781933,580      "reward_std": 0.48596611246466637,581      "rewards/accuracy_reward": 0.39166667591780424,582      "rewards/format_reward": 0.9208333469927311,583      "step": 410584    },585    {586      "clip_ratio": 0.0,587      "completion_length": 256.539590215683,588      "epoch": 0.56,589      "grad_norm": 0.5763727426528931,590      "kl": 2.3948883056640624,591      "learning_rate": 1.4476507549462489e-06,592      "loss": 0.2606,593      "reward": 1.1541667029261589,594      "reward_std": 0.6029179282486439,595      "rewards/accuracy_reward": 0.35000000819563865,596      "rewards/format_reward": 0.8041666833683848,597      "step": 420598    },599    {600      "clip_ratio": 0.0,601      "completion_length": 163.55833940505983,602      "epoch": 0.5733333333333334,603      "grad_norm": 0.23491673171520233,604      "kl": 0.215203857421875,605      "learning_rate": 1.3779619879982127e-06,606      "loss": 0.023,607      "reward": 1.420833373069763,608      "reward_std": 0.41848115585744383,609      "rewards/accuracy_reward": 0.4520833443850279,610      "rewards/format_reward": 0.9687500074505806,611      "step": 430612    },613    {614      "clip_ratio": 0.0,615      "completion_length": 183.7666721343994,616      "epoch": 0.5866666666666667,617      "grad_norm": 0.1774040013551712,618      "kl": 0.300738525390625,619      "learning_rate": 1.308537528209108e-06,620      "loss": 0.0643,621      "reward": 1.4270833656191826,622      "reward_std": 0.4158630233258009,623      "rewards/accuracy_reward": 0.46666667573153975,624      "rewards/format_reward": 0.9604166753590107,625      "step": 440626    },627    {628      "clip_ratio": 0.0,629      "completion_length": 212.7937566280365,630      "epoch": 0.6,631      "grad_norm": 0.20609110593795776,632      "kl": 0.5560638427734375,633      "learning_rate": 1.2395277334996047e-06,634      "loss": 0.0963,635      "reward": 1.3229166999459268,636      "reward_std": 0.4623309187591076,637      "rewards/accuracy_reward": 0.3937500096857548,638      "rewards/format_reward": 0.9291666783392429,639      "step": 450640    },641    {642      "clip_ratio": 0.0,643      "completion_length": 189.80208778381348,644      "epoch": 0.6133333333333333,645      "grad_norm": 0.49082836508750916,646      "kl": 0.7411895751953125,647      "learning_rate": 1.1710820637181448e-06,648      "loss": 0.0805,649      "reward": 1.320833370834589,650      "reward_std": 0.4459747776389122,651      "rewards/accuracy_reward": 0.37916667480021715,652      "rewards/format_reward": 0.9416666768491269,653      "step": 460654    },655    {656      "clip_ratio": 0.0,657      "completion_length": 194.38542304039,658      "epoch": 0.6266666666666667,659      "grad_norm": 0.48382291197776794,660      "kl": 0.4151397705078125,661      "learning_rate": 1.103348756944197e-06,662      "loss": 0.0534,663      "reward": 1.3812500394880771,664      "reward_std": 0.480151966586709,665      "rewards/accuracy_reward": 0.4250000108033419,666      "rewards/format_reward": 0.9562500081956387,667      "step": 470668    },669    {670      "clip_ratio": 0.0,671      "completion_length": 220.32292351722717,672      "epoch": 0.64,673      "grad_norm": 0.2328379899263382,674      "kl": 0.377593994140625,675      "learning_rate": 1.036474508437579e-06,676      "loss": 0.0734,677      "reward": 1.318750038743019,678      "reward_std": 0.47938378117978575,679      "rewards/accuracy_reward": 0.37708334289491174,680      "rewards/format_reward": 0.9416666783392429,681      "step": 480682    },683    {684      "clip_ratio": 0.0,685      "completion_length": 213.46458911895752,686      "epoch": 0.6533333333333333,687      "grad_norm": 0.6423675417900085,688      "kl": 0.6955718994140625,689      "learning_rate": 9.70604152929197e-07,690      "loss": 0.1096,691      "reward": 1.3416667044162751,692      "reward_std": 0.46137820817530156,693      "rewards/accuracy_reward": 0.4145833419635892,694      "rewards/format_reward": 0.9270833462476731,695      "step": 490696    },697    {698      "clip_ratio": 0.0,699      "completion_length": 214.08125677108765,700      "epoch": 0.6666666666666666,701      "grad_norm": 0.14841440320014954,702      "kl": 0.7181640625,703      "learning_rate": 9.058803509412648e-07,704      "loss": 0.0987,705      "reward": 1.397916704416275,706      "reward_std": 0.44944200329482553,707      "rewards/accuracy_reward": 0.46041667945683,708      "rewards/format_reward": 0.9375000111758709,709      "step": 500710    },711    {712      "clip_ratio": 0.0,713      "completion_length": 226.89792413711547,714      "epoch": 0.68,715      "grad_norm": 0.20204541087150574,716      "kl": 0.37396240234375,717      "learning_rate": 8.424432798163837e-07,718      "loss": 0.075,719      "reward": 1.4020833641290664,720      "reward_std": 0.3638565935194492,721      "rewards/accuracy_reward": 0.45000000968575476,722      "rewards/format_reward": 0.9520833425223827,723      "step": 510724    },725    {726      "clip_ratio": 0.0,727      "completion_length": 227.1395909309387,728      "epoch": 0.6933333333333334,729      "grad_norm": 0.170625239610672,730      "kl": 0.7568450927734375,731      "learning_rate": 7.804303301246311e-07,732      "loss": 0.0984,733      "reward": 1.331250037252903,734      "reward_std": 0.48546464554965496,735      "rewards/accuracy_reward": 0.40000001061707735,736      "rewards/format_reward": 0.9312500089406968,737      "step": 520738    },739    {740      "clip_ratio": 0.0,741      "completion_length": 222.33542160987855,742      "epoch": 0.7066666666666667,743      "grad_norm": 0.35064828395843506,744      "kl": 0.613885498046875,745      "learning_rate": 7.19975808106177e-07,746      "loss": 0.0488,747      "reward": 1.3625000312924385,748      "reward_std": 0.44792898930609226,749      "rewards/accuracy_reward": 0.4104166755452752,750      "rewards/format_reward": 0.9520833417773247,751      "step": 530752    },753    {754      "clip_ratio": 0.0,755      "completion_length": 237.87292375564576,756      "epoch": 0.72,757      "grad_norm": 0.3742374777793884,758      "kl": 0.584375,759      "learning_rate": 6.6121064479388e-07,760      "loss": 0.0612,761      "reward": 1.4020833656191827,762      "reward_std": 0.4303410712629557,763      "rewards/accuracy_reward": 0.46041667759418486,764      "rewards/format_reward": 0.9416666768491269,765      "step": 540766    },767    {768      "clip_ratio": 0.0,769      "completion_length": 244.73542499542236,770      "epoch": 0.7333333333333333,771      "grad_norm": 0.346811980009079,772      "kl": 0.83583984375,773      "learning_rate": 6.04262112445821e-07,774      "loss": 0.1029,775      "reward": 1.293750037997961,776      "reward_std": 0.48419367931783197,777      "rewards/accuracy_reward": 0.3791666770353913,778      "rewards/format_reward": 0.9145833477377892,779      "step": 550780    },781    {782      "clip_ratio": 0.0,783      "completion_length": 268.9645917892456,784      "epoch": 0.7466666666666667,785      "grad_norm": 0.14408645033836365,786      "kl": 0.513336181640625,787      "learning_rate": 5.492535489019345e-07,788      "loss": 0.084,789      "reward": 1.2875000312924385,790      "reward_std": 0.48581976890563966,791      "rewards/accuracy_reward": 0.3770833395421505,792      "rewards/format_reward": 0.9104166757315397,793      "step": 560794    },795    {796      "clip_ratio": 0.0,797      "completion_length": 241.53333921432494,798      "epoch": 0.76,799      "grad_norm": 1.40211021900177,800      "kl": 0.8751678466796875,801      "learning_rate": 4.963040904617131e-07,802      "loss": 0.1203,803      "reward": 1.3145833693444728,804      "reward_std": 0.49484665393829347,805      "rewards/accuracy_reward": 0.4145833423361182,806      "rewards/format_reward": 0.9000000134110451,807      "step": 570808    },809    {810      "clip_ratio": 0.0,811      "completion_length": 252.40000743865966,812      "epoch": 0.7733333333333333,813      "grad_norm": 0.4801377058029175,814      "kl": 1.2982696533203124,815      "learning_rate": 4.4552841386150737e-07,816      "loss": 0.1253,817      "reward": 1.2708333659917117,818      "reward_std": 0.5930769924074412,819      "rewards/accuracy_reward": 0.43541667610406876,820      "rewards/format_reward": 0.8354166846722364,821      "step": 580822    },823    {824      "clip_ratio": 0.0,825      "completion_length": 259.3583409309387,826      "epoch": 0.7866666666666666,827      "grad_norm": 1.0408235788345337,828      "kl": 1.4053619384765625,829      "learning_rate": 3.9703648791025716e-07,830      "loss": 0.183,831      "reward": 1.1916666943579912,832      "reward_std": 0.6191505286842585,833      "rewards/accuracy_reward": 0.3854166746139526,834      "rewards/format_reward": 0.8062500186264515,835      "step": 590836    },837    {838      "clip_ratio": 0.0,839      "completion_length": 243.28750715255737,840      "epoch": 0.8,841      "grad_norm": 0.2709506154060364,842      "kl": 0.96761474609375,843      "learning_rate": 3.5093333532153313e-07,844      "loss": 0.1034,845      "reward": 1.316666693240404,846      "reward_std": 0.46122562885284424,847      "rewards/accuracy_reward": 0.4354166748002172,848      "rewards/format_reward": 0.8812500096857547,849      "step": 600850    },851    {852      "clip_ratio": 0.0,853      "completion_length": 218.83542261123657,854      "epoch": 0.8133333333333334,855      "grad_norm": 0.19969280064105988,856      "kl": 0.4908477783203125,857      "learning_rate": 3.073188052577282e-07,858      "loss": 0.0814,859      "reward": 1.3458333656191825,860      "reward_std": 0.4190520565956831,861      "rewards/accuracy_reward": 0.4041666744276881,862      "rewards/format_reward": 0.9416666775941849,863      "step": 610864    },865    {866      "clip_ratio": 0.0,867      "completion_length": 216.3875057220459,868      "epoch": 0.8266666666666667,869      "grad_norm": 0.23758280277252197,870      "kl": 0.64254150390625,871      "learning_rate": 2.6628735707900655e-07,872      "loss": 0.0776,873      "reward": 1.3395833685994147,874      "reward_std": 0.4737738098949194,875      "rewards/accuracy_reward": 0.4250000070780516,876      "rewards/format_reward": 0.9145833484828472,877      "step": 620878    },879    {880      "clip_ratio": 0.0,881      "completion_length": 245.11459074020385,882      "epoch": 0.84,883      "grad_norm": 0.4293162226676941,884      "kl": 1.1108795166015626,885      "learning_rate": 2.2792785576536108e-07,886      "loss": 0.1537,887      "reward": 1.24375003837049,888      "reward_std": 0.520675316080451,889      "rewards/accuracy_reward": 0.3729166738688946,890      "rewards/format_reward": 0.8708333484828472,891      "step": 630892    },893    {894      "clip_ratio": 0.0,895      "completion_length": 249.4729232788086,896      "epoch": 0.8533333333333334,897      "grad_norm": 0.17477993667125702,898      "kl": 1.014031982421875,899      "learning_rate": 1.9232337945485655e-07,900      "loss": 0.1481,901      "reward": 1.293750035017729,902      "reward_std": 0.5492795780301094,903      "rewards/accuracy_reward": 0.4104166742414236,904      "rewards/format_reward": 0.8833333525806666,905      "step": 640906    },907    {908      "clip_ratio": 0.0,909      "completion_length": 271.8937582015991,910      "epoch": 0.8666666666666667,911      "grad_norm": 0.683167040348053,912      "kl": 1.2961456298828125,913      "learning_rate": 1.5955103951488177e-07,914      "loss": 0.1557,915      "reward": 1.1854166965931654,916      "reward_std": 0.5265001837164164,917      "rewards/accuracy_reward": 0.33541667349636556,918      "rewards/format_reward": 0.8500000169500709,919      "step": 650920    },921    {922      "clip_ratio": 0.0,923      "completion_length": 238.02084150314332,924      "epoch": 0.88,925      "grad_norm": 0.47596192359924316,926      "kl": 0.965118408203125,927      "learning_rate": 1.2968181353609853e-07,928      "loss": 0.1338,929      "reward": 1.3020833618938923,930      "reward_std": 0.49198094978928564,931      "rewards/accuracy_reward": 0.40833334252238274,932      "rewards/format_reward": 0.8937500178813934,933      "step": 660934    },935    {936      "clip_ratio": 0.0,937      "completion_length": 221.03750729560852,938      "epoch": 0.8933333333333333,939      "grad_norm": 0.4685460031032562,940      "kl": 0.8098968505859375,941      "learning_rate": 1.0278039161078634e-07,942      "loss": 0.1018,943      "reward": 1.3375000290572643,944      "reward_std": 0.4632135137915611,945      "rewards/accuracy_reward": 0.43750000912696124,946      "rewards/format_reward": 0.9000000119209289,947      "step": 670948    },949    {950      "clip_ratio": 0.0,951      "completion_length": 253.50209112167357,952      "epoch": 0.9066666666666666,953      "grad_norm": 0.2506866753101349,954      "kl": 0.906805419921875,955      "learning_rate": 7.89050362285062e-08,956      "loss": 0.1365,957      "reward": 1.2375000305473804,958      "reward_std": 0.517885773256421,959      "rewards/accuracy_reward": 0.3645833428949118,960      "rewards/format_reward": 0.8729166854172945,961      "step": 680962    },963    {964      "clip_ratio": 0.0,965      "completion_length": 239.60625705718994,966      "epoch": 0.92,967      "grad_norm": 0.5413881540298462,968      "kl": 0.835858154296875,969      "learning_rate": 5.810745609252166e-08,970      "loss": 0.1281,971      "reward": 1.2187500409781933,972      "reward_std": 0.5353617053478956,973      "rewards/accuracy_reward": 0.3395833415910602,974      "rewards/format_reward": 0.8791666828095913,975      "step": 690976    },977    {978      "clip_ratio": 0.0,979      "completion_length": 232.96667356491088,980      "epoch": 0.9333333333333333,981      "grad_norm": 0.41816839575767517,982      "kl": 1.4766021728515626,983      "learning_rate": 4.0432694130264294e-08,984      "loss": 0.1412,985      "reward": 1.2375000312924385,986      "reward_std": 0.5488180216401816,987      "rewards/accuracy_reward": 0.37708334121853115,988      "rewards/format_reward": 0.8604166824370623,989      "step": 700990    },991    {992      "clip_ratio": 0.0,993      "completion_length": 278.4333411693573,994      "epoch": 0.9466666666666667,995      "grad_norm": 0.5456490516662598,996      "kl": 1.7320938110351562,997      "learning_rate": 2.5919029940380145e-08,998      "loss": 0.2249,999      "reward": 1.2062500290572644,1000      "reward_std": 0.6055900201201438,1001      "rewards/accuracy_reward": 0.3833333427086473,1002      "rewards/format_reward": 0.8229166869074106,1003      "step": 7101004    },1005    {1006      "clip_ratio": 0.0,1007      "completion_length": 239.62500748634338,1008      "epoch": 0.96,1009      "grad_norm": 0.32890596985816956,1010      "kl": 1.458221435546875,1011      "learning_rate": 1.4597896887644457e-08,1012      "loss": 0.1827,1013      "reward": 1.264583370089531,1014      "reward_std": 0.5755864661186934,1015      "rewards/accuracy_reward": 0.4187500111758709,1016      "rewards/format_reward": 0.8458333492279053,1017      "step": 7201018    },1019    {1020      "clip_ratio": 0.0,1021      "completion_length": 274.1791749954224,1022      "epoch": 0.9733333333333334,1023      "grad_norm": 0.3116457760334015,1024      "kl": 1.3982940673828126,1025      "learning_rate": 6.493814025293476e-09,1026      "loss": 0.1776,1027      "reward": 1.231250035017729,1028      "reward_std": 0.5566708967089653,1029      "rewards/accuracy_reward": 0.40000000949949027,1030      "rewards/format_reward": 0.8312500137835741,1031      "step": 7301032    },1033    {1034      "clip_ratio": 0.0,1035      "completion_length": 255.23750734329224,1036      "epoch": 0.9866666666666667,1037      "grad_norm": 0.4142087697982788,1038      "kl": 1.2847900390625,1039      "learning_rate": 1.624332992213151e-09,1040      "loss": 0.1855,1041      "reward": 1.2229166984558106,1042      "reward_std": 0.5469876442104578,1043      "rewards/accuracy_reward": 0.38333334159106014,1044      "rewards/format_reward": 0.8395833514630795,1045      "step": 7401046    },1047    {1048      "clip_ratio": 0.0,1049      "completion_length": 279.8500079154968,1050      "epoch": 1.0,1051      "grad_norm": 0.3987623155117035,1052      "kl": 1.2904449462890626,1053      "learning_rate": 0.0,1054      "loss": 0.2137,1055      "reward": 1.2583333723247052,1056      "reward_std": 0.6197128046303988,1057      "rewards/accuracy_reward": 0.4229166766628623,1058      "rewards/format_reward": 0.8354166867211461,1059      "step": 7501060    },1061    {1062      "epoch": 1.0,1063      "step": 750,1064      "total_flos": 0.0,1065      "train_loss": 0.1488674604743719,1066      "train_runtime": 66920.1081,1067      "train_samples_per_second": 0.09,1068      "train_steps_per_second": 0.0111069    }1070  ],1071  "logging_steps": 10,1072  "max_steps": 750,1073  "num_input_tokens_seen": 0,1074  "num_train_epochs": 1,1075  "save_steps": 500,1076  "stateful_callbacks": {1077    "TrainerControl": {1078      "args": {1079        "should_epoch_stop": false,1080        "should_evaluate": false,1081        "should_log": false,1082        "should_save": true,1083        "should_training_stop": true1084      },1085      "attributes": {}1086    }1087  },1088  "total_flos": 0.0,1089  "train_batch_size": 2,1090  "trial_name": null,1091  "trial_params": null1092}1093