dbenv-week2-HW2-ppo / checkpoint-100 /trainer_state.json
ntraore's picture
Upload folder using huggingface_hub
aaa86ac verified
{
"best_global_step": null,
"best_metric": null,
"best_model_checkpoint": null,
"episode": 800,
"epoch": 1.0,
"eval_steps": 500,
"global_step": 100,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"episode": 8,
"epoch": 0.01,
"eps": 0,
"loss/policy_avg": 0.09475220739841461,
"loss/value_avg": 37.87782287597656,
"lr": 5e-05,
"objective/entropy": -46.64240264892578,
"objective/kl": 0.24876117706298828,
"objective/non_score_reward": -0.012438058853149414,
"objective/rlhf_reward": -7.19224739074707,
"objective/scores": -7.1798095703125,
"policy/approxkl_avg": 0.23542489111423492,
"policy/clipfrac_avg": 0.442307710647583,
"policy/entropy_avg": 1.8903515338897705,
"step": 1,
"val/clipfrac_avg": 0.75,
"val/num_eos_tokens": 9,
"val/ratio": 1.0843846797943115,
"val/ratio_var": 0.015354042872786522
},
{
"episode": 16,
"epoch": 0.02,
"eps": 0,
"loss/policy_avg": -0.09378913044929504,
"loss/value_avg": 17.912017822265625,
"lr": 4.9500000000000004e-05,
"objective/entropy": -43.38051986694336,
"objective/kl": -0.39787864685058594,
"objective/non_score_reward": 0.019893934950232506,
"objective/rlhf_reward": -7.087190628051758,
"objective/scores": -7.10708475112915,
"policy/approxkl_avg": 0.031495485454797745,
"policy/clipfrac_avg": 0.510869562625885,
"policy/entropy_avg": 1.8803576231002808,
"step": 2,
"val/clipfrac_avg": 0.75,
"val/num_eos_tokens": 9,
"val/ratio": 1.0658495426177979,
"val/ratio_var": 0.0052874707616865635
},
{
"episode": 24,
"epoch": 0.03,
"eps": 0,
"loss/policy_avg": -0.048949725925922394,
"loss/value_avg": 14.420023918151855,
"lr": 4.9e-05,
"objective/entropy": -19.564510345458984,
"objective/kl": 8.628725051879883,
"objective/non_score_reward": -0.4314362406730652,
"objective/rlhf_reward": -7.2510809898376465,
"objective/scores": -6.819644927978516,
"policy/approxkl_avg": 0.02875935286283493,
"policy/clipfrac_avg": 0.42465752363204956,
"policy/entropy_avg": 2.1179404258728027,
"step": 3,
"val/clipfrac_avg": 0.7191358208656311,
"val/num_eos_tokens": 8,
"val/ratio": 1.0251271724700928,
"val/ratio_var": 0.0014984746230766177
},
{
"episode": 32,
"epoch": 0.04,
"eps": 0,
"loss/policy_avg": -0.058641545474529266,
"loss/value_avg": 6.136292934417725,
"lr": 4.85e-05,
"objective/entropy": 2.016387462615967,
"objective/kl": 12.600632667541504,
"objective/non_score_reward": -0.6300315856933594,
"objective/rlhf_reward": -7.23648738861084,
"objective/scores": -6.6064558029174805,
"policy/approxkl_avg": 0.026198051869869232,
"policy/clipfrac_avg": 0.279629647731781,
"policy/entropy_avg": 2.0662665367126465,
"step": 4,
"val/clipfrac_avg": 0.6328670978546143,
"val/num_eos_tokens": 9,
"val/ratio": 1.0075256824493408,
"val/ratio_var": 3.9297356124734506e-05
},
{
"episode": 40,
"epoch": 0.05,
"eps": 0,
"loss/policy_avg": -0.06558828055858612,
"loss/value_avg": 1.9106276035308838,
"lr": 4.8e-05,
"objective/entropy": 50.39515686035156,
"objective/kl": 18.737285614013672,
"objective/non_score_reward": -0.9368643164634705,
"objective/rlhf_reward": -8.13361644744873,
"objective/scores": -7.196752548217773,
"policy/approxkl_avg": 0.017262864857912064,
"policy/clipfrac_avg": 0.22424891591072083,
"policy/entropy_avg": 2.027545928955078,
"step": 5,
"val/clipfrac_avg": 0.5177083015441895,
"val/num_eos_tokens": 7,
"val/ratio": 1.0054020881652832,
"val/ratio_var": 0.00012101430911570787
},
{
"episode": 48,
"epoch": 0.06,
"eps": 0,
"loss/policy_avg": -0.029659481719136238,
"loss/value_avg": 0.9584429264068604,
"lr": 4.75e-05,
"objective/entropy": 31.915897369384766,
"objective/kl": 17.292184829711914,
"objective/non_score_reward": -0.8646091818809509,
"objective/rlhf_reward": -7.128448009490967,
"objective/scores": -6.263838768005371,
"policy/approxkl_avg": 0.00980815663933754,
"policy/clipfrac_avg": 0.1230158805847168,
"policy/entropy_avg": 1.5745285749435425,
"step": 6,
"val/clipfrac_avg": 0.4331395626068115,
"val/num_eos_tokens": 7,
"val/ratio": 0.9889805316925049,
"val/ratio_var": 0.00010295213724020869
},
{
"episode": 56,
"epoch": 0.07,
"eps": 0,
"loss/policy_avg": -0.05306820571422577,
"loss/value_avg": 0.9622513055801392,
"lr": 4.7e-05,
"objective/entropy": 12.873809814453125,
"objective/kl": 17.415592193603516,
"objective/non_score_reward": -0.8707795739173889,
"objective/rlhf_reward": -7.566481590270996,
"objective/scores": -6.695702075958252,
"policy/approxkl_avg": 0.016156045719981194,
"policy/clipfrac_avg": 0.2081081122159958,
"policy/entropy_avg": 1.9123384952545166,
"step": 7,
"val/clipfrac_avg": 0.5481771230697632,
"val/num_eos_tokens": 8,
"val/ratio": 0.9937976598739624,
"val/ratio_var": 5.228411828284152e-05
},
{
"episode": 64,
"epoch": 0.08,
"eps": 0,
"loss/policy_avg": -0.022447112947702408,
"loss/value_avg": 0.4383925497531891,
"lr": 4.6500000000000005e-05,
"objective/entropy": 6.965538024902344,
"objective/kl": 21.360532760620117,
"objective/non_score_reward": -1.0680266618728638,
"objective/rlhf_reward": -6.4422149658203125,
"objective/scores": -5.374188423156738,
"policy/approxkl_avg": 0.0097531508654356,
"policy/clipfrac_avg": 0.0854545384645462,
"policy/entropy_avg": 1.2789623737335205,
"step": 8,
"val/clipfrac_avg": 0.1944444477558136,
"val/num_eos_tokens": 6,
"val/ratio": 1.0197606086730957,
"val/ratio_var": 0.00018990044191014022
},
{
"episode": 72,
"epoch": 0.09,
"eps": 0,
"loss/policy_avg": -0.03375006467103958,
"loss/value_avg": 0.20129632949829102,
"lr": 4.600000000000001e-05,
"objective/entropy": 25.035720825195312,
"objective/kl": 21.478357315063477,
"objective/non_score_reward": -1.0739178657531738,
"objective/rlhf_reward": -6.791517734527588,
"objective/scores": -5.717599868774414,
"policy/approxkl_avg": 0.03056078404188156,
"policy/clipfrac_avg": 0.12059858441352844,
"policy/entropy_avg": 1.3224170207977295,
"step": 9,
"val/clipfrac_avg": 0.1833910048007965,
"val/num_eos_tokens": 5,
"val/ratio": 1.0177733898162842,
"val/ratio_var": 0.000167887206771411
},
{
"episode": 80,
"epoch": 0.1,
"eps": 0,
"loss/policy_avg": -0.018979491665959358,
"loss/value_avg": 0.5691134333610535,
"lr": 4.55e-05,
"objective/entropy": 34.5059814453125,
"objective/kl": 27.583213806152344,
"objective/non_score_reward": -1.3791606426239014,
"objective/rlhf_reward": -8.178009033203125,
"objective/scores": -6.7988481521606445,
"policy/approxkl_avg": 0.01339983381330967,
"policy/clipfrac_avg": 0.08974358439445496,
"policy/entropy_avg": 1.328680157661438,
"step": 10,
"val/clipfrac_avg": 0.45820188522338867,
"val/num_eos_tokens": 6,
"val/ratio": 1.0151302814483643,
"val/ratio_var": 0.00030882572173140943
},
{
"episode": 88,
"epoch": 0.11,
"eps": 0,
"loss/policy_avg": -0.01908348686993122,
"loss/value_avg": 0.7630579471588135,
"lr": 4.5e-05,
"objective/entropy": 25.90598487854004,
"objective/kl": 21.881332397460938,
"objective/non_score_reward": -1.0940666198730469,
"objective/rlhf_reward": -7.77611780166626,
"objective/scores": -6.682051181793213,
"policy/approxkl_avg": 0.009299904108047485,
"policy/clipfrac_avg": 0.08960843086242676,
"policy/entropy_avg": 1.1157985925674438,
"step": 11,
"val/clipfrac_avg": 0.1889880895614624,
"val/num_eos_tokens": 5,
"val/ratio": 1.0160670280456543,
"val/ratio_var": 0.00019202969269827008
},
{
"episode": 96,
"epoch": 0.12,
"eps": 0,
"loss/policy_avg": -0.007455375976860523,
"loss/value_avg": 0.11711454391479492,
"lr": 4.4500000000000004e-05,
"objective/entropy": 4.318185806274414,
"objective/kl": 25.807804107666016,
"objective/non_score_reward": -1.2903902530670166,
"objective/rlhf_reward": -5.789934158325195,
"objective/scores": -4.4995436668396,
"policy/approxkl_avg": 0.0064476775005459785,
"policy/clipfrac_avg": 0.0356217622756958,
"policy/entropy_avg": 0.33628159761428833,
"step": 12,
"val/clipfrac_avg": 0.27583980560302734,
"val/num_eos_tokens": 1,
"val/ratio": 1.0104122161865234,
"val/ratio_var": 1.4566742720489856e-05
},
{
"episode": 104,
"epoch": 0.13,
"eps": 0,
"loss/policy_avg": -0.0054306937381625175,
"loss/value_avg": 0.6040916442871094,
"lr": 4.4000000000000006e-05,
"objective/entropy": 4.192093372344971,
"objective/kl": 24.524070739746094,
"objective/non_score_reward": -1.226203441619873,
"objective/rlhf_reward": -7.248112201690674,
"objective/scores": -6.021908760070801,
"policy/approxkl_avg": 0.013138508424162865,
"policy/clipfrac_avg": 0.03693181648850441,
"policy/entropy_avg": 0.5951633453369141,
"step": 13,
"val/clipfrac_avg": 0.2570621371269226,
"val/num_eos_tokens": 2,
"val/ratio": 0.9956222772598267,
"val/ratio_var": 1.213013911183225e-05
},
{
"episode": 112,
"epoch": 0.14,
"eps": 0,
"loss/policy_avg": -0.016900666058063507,
"loss/value_avg": 0.7295329570770264,
"lr": 4.35e-05,
"objective/entropy": 13.571832656860352,
"objective/kl": 25.062355041503906,
"objective/non_score_reward": -1.2531176805496216,
"objective/rlhf_reward": -5.660390377044678,
"objective/scores": -4.407272815704346,
"policy/approxkl_avg": 0.010249488987028599,
"policy/clipfrac_avg": 0.037146225571632385,
"policy/entropy_avg": 0.2530284523963928,
"step": 14,
"val/clipfrac_avg": 0.26238206028938293,
"val/num_eos_tokens": 0,
"val/ratio": 1.0186505317687988,
"val/ratio_var": 0.00026969113969244063
},
{
"episode": 120,
"epoch": 0.15,
"eps": 0,
"loss/policy_avg": -0.0026563892606645823,
"loss/value_avg": 0.13339246809482574,
"lr": 4.3e-05,
"objective/entropy": 2.422701358795166,
"objective/kl": 19.802749633789062,
"objective/non_score_reward": -0.99013751745224,
"objective/rlhf_reward": -6.004833698272705,
"objective/scores": -5.01469612121582,
"policy/approxkl_avg": 0.003422472393140197,
"policy/clipfrac_avg": 0.009043928235769272,
"policy/entropy_avg": 0.2299969494342804,
"step": 15,
"val/clipfrac_avg": 0.10180411487817764,
"val/num_eos_tokens": 1,
"val/ratio": 1.0032926797866821,
"val/ratio_var": 1.6620078895357437e-05
},
{
"episode": 128,
"epoch": 0.16,
"eps": 0,
"loss/policy_avg": -0.04060552269220352,
"loss/value_avg": 0.30010226368904114,
"lr": 4.25e-05,
"objective/entropy": 41.43886184692383,
"objective/kl": 33.01524353027344,
"objective/non_score_reward": -1.6507622003555298,
"objective/rlhf_reward": -6.505797386169434,
"objective/scores": -4.855035305023193,
"policy/approxkl_avg": 0.04338652640581131,
"policy/clipfrac_avg": 0.09436275064945221,
"policy/entropy_avg": 0.8455747961997986,
"step": 16,
"val/clipfrac_avg": 0.07823961228132248,
"val/num_eos_tokens": 1,
"val/ratio": 0.9889528751373291,
"val/ratio_var": 2.141590630344581e-05
},
{
"episode": 136,
"epoch": 0.17,
"eps": 0,
"loss/policy_avg": -0.000526640797033906,
"loss/value_avg": 0.16104120016098022,
"lr": 4.2e-05,
"objective/entropy": 0.7853485345840454,
"objective/kl": 22.7672176361084,
"objective/non_score_reward": -1.138360857963562,
"objective/rlhf_reward": -5.805202484130859,
"objective/scores": -4.666841506958008,
"policy/approxkl_avg": 0.00014227806241251528,
"policy/clipfrac_avg": 0.001179245300590992,
"policy/entropy_avg": 0.011943262070417404,
"step": 17,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 1.0015124082565308,
"val/ratio_var": 6.323739398794714e-07
},
{
"episode": 144,
"epoch": 0.18,
"eps": 0,
"loss/policy_avg": -0.0006135053699836135,
"loss/value_avg": 0.06096681207418442,
"lr": 4.15e-05,
"objective/entropy": 0.09207843244075775,
"objective/kl": 22.14600944519043,
"objective/non_score_reward": -1.1073005199432373,
"objective/rlhf_reward": -5.784175872802734,
"objective/scores": -4.676875591278076,
"policy/approxkl_avg": 1.3936476534581743e-05,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.00949782133102417,
"step": 18,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 1.000463843345642,
"val/ratio_var": 2.575906987711818e-10
},
{
"episode": 152,
"epoch": 0.19,
"eps": 0,
"loss/policy_avg": -0.0019343766616657376,
"loss/value_avg": 0.1923561841249466,
"lr": 4.1e-05,
"objective/entropy": 0.16122427582740784,
"objective/kl": 22.323423385620117,
"objective/non_score_reward": -1.1161713600158691,
"objective/rlhf_reward": -6.342621326446533,
"objective/scores": -5.226449966430664,
"policy/approxkl_avg": 0.00015607287059538066,
"policy/clipfrac_avg": 0.004716981202363968,
"policy/entropy_avg": 0.010668131522834301,
"step": 19,
"val/clipfrac_avg": 0.2299528270959854,
"val/num_eos_tokens": 0,
"val/ratio": 1.001634120941162,
"val/ratio_var": 1.4890058075422985e-08
},
{
"episode": 160,
"epoch": 0.2,
"eps": 0,
"loss/policy_avg": -0.004725292790681124,
"loss/value_avg": 0.16123607754707336,
"lr": 4.05e-05,
"objective/entropy": 8.322664260864258,
"objective/kl": 24.683185577392578,
"objective/non_score_reward": -1.2341593503952026,
"objective/rlhf_reward": -5.897225379943848,
"objective/scores": -4.6630659103393555,
"policy/approxkl_avg": 0.009079387411475182,
"policy/clipfrac_avg": 0.010613207705318928,
"policy/entropy_avg": 0.029525920748710632,
"step": 20,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 1.020014762878418,
"val/ratio_var": 0.0007286081090569496
},
{
"episode": 168,
"epoch": 0.21,
"eps": 0,
"loss/policy_avg": 1.005048216029536e-05,
"loss/value_avg": 0.10769544541835785,
"lr": 4e-05,
"objective/entropy": 0.6765559911727905,
"objective/kl": 21.308088302612305,
"objective/non_score_reward": -1.0654044151306152,
"objective/rlhf_reward": -5.437287330627441,
"objective/scores": -4.371882915496826,
"policy/approxkl_avg": 4.5833239710191265e-05,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.004317480139434338,
"step": 21,
"val/clipfrac_avg": 0.08490566164255142,
"val/num_eos_tokens": 0,
"val/ratio": 0.9998916387557983,
"val/ratio_var": 4.1950055873485326e-08
},
{
"episode": 176,
"epoch": 0.22,
"eps": 0,
"loss/policy_avg": -0.002242459449917078,
"loss/value_avg": 0.08367361128330231,
"lr": 3.9500000000000005e-05,
"objective/entropy": 11.297447204589844,
"objective/kl": 20.96246337890625,
"objective/non_score_reward": -1.0481232404708862,
"objective/rlhf_reward": -5.453763484954834,
"objective/scores": -4.405640125274658,
"policy/approxkl_avg": 0.008936449885368347,
"policy/clipfrac_avg": 0.009124087169766426,
"policy/entropy_avg": 0.09280973672866821,
"step": 22,
"val/clipfrac_avg": 0.0024271844886243343,
"val/num_eos_tokens": 1,
"val/ratio": 0.9998486638069153,
"val/ratio_var": 1.0913846608673339e-06
},
{
"episode": 184,
"epoch": 0.23,
"eps": 0,
"loss/policy_avg": -2.541125832067337e-05,
"loss/value_avg": 0.04419855773448944,
"lr": 3.9000000000000006e-05,
"objective/entropy": 0.021149635314941406,
"objective/kl": 21.688682556152344,
"objective/non_score_reward": -1.0844340324401855,
"objective/rlhf_reward": -5.5749287605285645,
"objective/scores": -4.490494728088379,
"policy/approxkl_avg": 3.073716925428016e-07,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.0026977781672030687,
"step": 23,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 1.0000680685043335,
"val/ratio_var": 6.224354365258478e-12
},
{
"episode": 192,
"epoch": 0.24,
"eps": 0,
"loss/policy_avg": -5.5702646932331845e-05,
"loss/value_avg": 0.12123075127601624,
"lr": 3.85e-05,
"objective/entropy": 0.02785348892211914,
"objective/kl": 21.36163330078125,
"objective/non_score_reward": -1.0680818557739258,
"objective/rlhf_reward": -5.926137924194336,
"objective/scores": -4.85805606842041,
"policy/approxkl_avg": 1.9325676703374484e-07,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.003609274746850133,
"step": 24,
"val/clipfrac_avg": 0.11674527823925018,
"val/num_eos_tokens": 0,
"val/ratio": 1.0000431537628174,
"val/ratio_var": 7.615597041876754e-11
},
{
"episode": 200,
"epoch": 0.25,
"eps": 0,
"loss/policy_avg": 1.5368334061349742e-05,
"loss/value_avg": 0.10563156753778458,
"lr": 3.8e-05,
"objective/entropy": 0.03280210494995117,
"objective/kl": 21.531850814819336,
"objective/non_score_reward": -1.0765926837921143,
"objective/rlhf_reward": -5.382946968078613,
"objective/scores": -4.306354522705078,
"policy/approxkl_avg": 4.6442153234238504e-07,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.004041329957544804,
"step": 25,
"val/clipfrac_avg": 0.08726415038108826,
"val/num_eos_tokens": 0,
"val/ratio": 1.0000438690185547,
"val/ratio_var": 7.789253686496522e-10
},
{
"episode": 208,
"epoch": 0.26,
"eps": 0,
"loss/policy_avg": -0.0069130496121943,
"loss/value_avg": 0.049083612859249115,
"lr": 3.7500000000000003e-05,
"objective/entropy": -0.11079001426696777,
"objective/kl": 18.664615631103516,
"objective/non_score_reward": -0.9332309365272522,
"objective/rlhf_reward": -5.723866939544678,
"objective/scores": -4.79063606262207,
"policy/approxkl_avg": 0.008058368228375912,
"policy/clipfrac_avg": 0.009842519648373127,
"policy/entropy_avg": 0.20962515473365784,
"step": 26,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 1,
"val/ratio": 0.9940447807312012,
"val/ratio_var": 1.0320742148905993e-05
},
{
"episode": 216,
"epoch": 0.27,
"eps": 0,
"loss/policy_avg": -0.019339777529239655,
"loss/value_avg": 0.20152565836906433,
"lr": 3.7e-05,
"objective/entropy": 19.612171173095703,
"objective/kl": 19.863204956054688,
"objective/non_score_reward": -0.9931602478027344,
"objective/rlhf_reward": -5.119001388549805,
"objective/scores": -4.12584114074707,
"policy/approxkl_avg": 0.0681137889623642,
"policy/clipfrac_avg": 0.028301885351538658,
"policy/entropy_avg": 0.12574249505996704,
"step": 27,
"val/clipfrac_avg": 0.19929245114326477,
"val/num_eos_tokens": 0,
"val/ratio": 1.4276061058044434,
"val/ratio_var": 0.5271331667900085
},
{
"episode": 224,
"epoch": 0.28,
"eps": 0,
"loss/policy_avg": -0.003817332675680518,
"loss/value_avg": 0.1855505108833313,
"lr": 3.65e-05,
"objective/entropy": 17.2297306060791,
"objective/kl": 24.35909652709961,
"objective/non_score_reward": -1.2179547548294067,
"objective/rlhf_reward": -6.254249572753906,
"objective/scores": -5.036294937133789,
"policy/approxkl_avg": 0.07349927723407745,
"policy/clipfrac_avg": 0.011848341673612595,
"policy/entropy_avg": 0.03583555296063423,
"step": 28,
"val/clipfrac_avg": 0.18853428959846497,
"val/num_eos_tokens": 1,
"val/ratio": 0.9938024878501892,
"val/ratio_var": 1.8159549654228613e-05
},
{
"episode": 232,
"epoch": 0.29,
"eps": 0,
"loss/policy_avg": -0.0002830463636200875,
"loss/value_avg": 0.29814815521240234,
"lr": 3.6e-05,
"objective/entropy": 1.003382682800293,
"objective/kl": 22.782087326049805,
"objective/non_score_reward": -1.1391044855117798,
"objective/rlhf_reward": -6.658185958862305,
"objective/scores": -5.5190815925598145,
"policy/approxkl_avg": 5.4168755013961345e-05,
"policy/clipfrac_avg": 0.000589622650295496,
"policy/entropy_avg": 0.00883533339947462,
"step": 29,
"val/clipfrac_avg": 0.14799527823925018,
"val/num_eos_tokens": 0,
"val/ratio": 1.000295639038086,
"val/ratio_var": 4.9585271000296416e-08
},
{
"episode": 240,
"epoch": 0.3,
"eps": 0,
"loss/policy_avg": 4.81777751701884e-05,
"loss/value_avg": 0.04622498154640198,
"lr": 3.55e-05,
"objective/entropy": 0.02696990966796875,
"objective/kl": 20.711894989013672,
"objective/non_score_reward": -1.0355949401855469,
"objective/rlhf_reward": -5.1848955154418945,
"objective/scores": -4.149300575256348,
"policy/approxkl_avg": 4.969482461092412e-07,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.002766390796750784,
"step": 30,
"val/clipfrac_avg": 0.1320754736661911,
"val/num_eos_tokens": 0,
"val/ratio": 1.0001327991485596,
"val/ratio_var": 7.744499264106253e-09
},
{
"episode": 248,
"epoch": 0.31,
"eps": 0,
"loss/policy_avg": -9.446142030355986e-06,
"loss/value_avg": 0.03906877338886261,
"lr": 3.5e-05,
"objective/entropy": 0.018702030181884766,
"objective/kl": 20.70791244506836,
"objective/non_score_reward": -1.0353957414627075,
"objective/rlhf_reward": -5.228009223937988,
"objective/scores": -4.19261360168457,
"policy/approxkl_avg": 7.734050200269849e-08,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.0023581646382808685,
"step": 31,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 1.0000567436218262,
"val/ratio_var": 8.75753414231184e-10
},
{
"episode": 256,
"epoch": 0.32,
"eps": 0,
"loss/policy_avg": -0.0007870738045312464,
"loss/value_avg": 0.04817962646484375,
"lr": 3.45e-05,
"objective/entropy": 7.295212745666504,
"objective/kl": 20.889162063598633,
"objective/non_score_reward": -1.0444581508636475,
"objective/rlhf_reward": -5.3556060791015625,
"objective/scores": -4.311148166656494,
"policy/approxkl_avg": 0.03791780769824982,
"policy/clipfrac_avg": 0.001768867950886488,
"policy/entropy_avg": 0.0018975285347551107,
"step": 32,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 5.218703269958496,
"val/ratio_var": 57.09228515625
},
{
"episode": 264,
"epoch": 0.33,
"eps": 0,
"loss/policy_avg": -1.0527459380682558e-05,
"loss/value_avg": 0.04428659379482269,
"lr": 3.4000000000000007e-05,
"objective/entropy": 0.011212348937988281,
"objective/kl": 21.30687141418457,
"objective/non_score_reward": -1.0653434991836548,
"objective/rlhf_reward": -5.890254497528076,
"objective/scores": -4.824911117553711,
"policy/approxkl_avg": 2.762976292203234e-09,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.0016911793500185013,
"step": 33,
"val/clipfrac_avg": 0.12028302252292633,
"val/num_eos_tokens": 0,
"val/ratio": 1.0000114440917969,
"val/ratio_var": 3.550345087366402e-11
},
{
"episode": 272,
"epoch": 0.34,
"eps": 0,
"loss/policy_avg": -0.0006948995869606733,
"loss/value_avg": 0.32256680727005005,
"lr": 3.35e-05,
"objective/entropy": 16.441970825195312,
"objective/kl": 19.257633209228516,
"objective/non_score_reward": -0.9628816843032837,
"objective/rlhf_reward": -5.559625148773193,
"objective/scores": -4.596743583679199,
"policy/approxkl_avg": 0.010671013966202736,
"policy/clipfrac_avg": 0.004716981202363968,
"policy/entropy_avg": 0.0016686677699908614,
"step": 34,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 1.067878246307373,
"val/ratio_var": 0.008366185240447521
},
{
"episode": 280,
"epoch": 0.35,
"eps": 0,
"loss/policy_avg": -9.87138719210634e-06,
"loss/value_avg": 0.09626642614603043,
"lr": 3.3e-05,
"objective/entropy": 0.011440753936767578,
"objective/kl": 22.810344696044922,
"objective/non_score_reward": -1.1405171155929565,
"objective/rlhf_reward": -6.108673095703125,
"objective/scores": -4.968155860900879,
"policy/approxkl_avg": 2.907630047843668e-08,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.0016481179045513272,
"step": 35,
"val/clipfrac_avg": 0.13148584961891174,
"val/num_eos_tokens": 0,
"val/ratio": 1.000030517578125,
"val/ratio_var": 7.212008767965017e-11
},
{
"episode": 288,
"epoch": 0.36,
"eps": 0,
"loss/policy_avg": -0.00041965150739997625,
"loss/value_avg": 0.34581074118614197,
"lr": 3.2500000000000004e-05,
"objective/entropy": 5.809754848480225,
"objective/kl": 19.50436782836914,
"objective/non_score_reward": -0.9752184152603149,
"objective/rlhf_reward": -6.9395976066589355,
"objective/scores": -5.96437931060791,
"policy/approxkl_avg": 0.0018271624576300383,
"policy/clipfrac_avg": 0.001768867950886488,
"policy/entropy_avg": 0.00134533760137856,
"step": 36,
"val/clipfrac_avg": 0.21049527823925018,
"val/num_eos_tokens": 0,
"val/ratio": 0.9987656474113464,
"val/ratio_var": 7.995101896085544e-07
},
{
"episode": 296,
"epoch": 0.37,
"eps": 0,
"loss/policy_avg": -0.0005644646007567644,
"loss/value_avg": 0.03859845548868179,
"lr": 3.2000000000000005e-05,
"objective/entropy": 10.471796035766602,
"objective/kl": 24.136667251586914,
"objective/non_score_reward": -1.2068334817886353,
"objective/rlhf_reward": -6.471621990203857,
"objective/scores": -5.264788627624512,
"policy/approxkl_avg": 0.003387878183275461,
"policy/clipfrac_avg": 0.002358490601181984,
"policy/entropy_avg": 0.001884155673906207,
"step": 37,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 1.011574387550354,
"val/ratio_var": 0.0001694993843557313
},
{
"episode": 304,
"epoch": 0.38,
"eps": 0,
"loss/policy_avg": 1.0252632591800648e-06,
"loss/value_avg": 0.12985283136367798,
"lr": 3.15e-05,
"objective/entropy": 0.009371757507324219,
"objective/kl": 22.376888275146484,
"objective/non_score_reward": -1.1188445091247559,
"objective/rlhf_reward": -5.919074058532715,
"objective/scores": -4.800229549407959,
"policy/approxkl_avg": 3.2507914138335536e-09,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.0015507801435887814,
"step": 38,
"val/clipfrac_avg": 0.09669811278581619,
"val/num_eos_tokens": 0,
"val/ratio": 0.9999986886978149,
"val/ratio_var": 4.625277938430372e-11
},
{
"episode": 312,
"epoch": 0.39,
"eps": 0,
"loss/policy_avg": -0.0002961930294986814,
"loss/value_avg": 0.22089466452598572,
"lr": 3.1e-05,
"objective/entropy": 7.122612476348877,
"objective/kl": 16.583417892456055,
"objective/non_score_reward": -0.8291708827018738,
"objective/rlhf_reward": -5.8148112297058105,
"objective/scores": -4.985640525817871,
"policy/approxkl_avg": 0.001451514894142747,
"policy/clipfrac_avg": 0.001768867950886488,
"policy/entropy_avg": 0.001556121977046132,
"step": 39,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 1.0047032833099365,
"val/ratio_var": 2.341072104172781e-05
},
{
"episode": 320,
"epoch": 0.4,
"eps": 0,
"loss/policy_avg": -0.0014894630294293165,
"loss/value_avg": 0.12523594498634338,
"lr": 3.05e-05,
"objective/entropy": 7.267007827758789,
"objective/kl": 18.883316040039062,
"objective/non_score_reward": -0.944165825843811,
"objective/rlhf_reward": -5.483607769012451,
"objective/scores": -4.53944206237793,
"policy/approxkl_avg": 0.026282615959644318,
"policy/clipfrac_avg": 0.00294811325147748,
"policy/entropy_avg": 0.0025057026650756598,
"step": 40,
"val/clipfrac_avg": 0.2045990526676178,
"val/num_eos_tokens": 0,
"val/ratio": 2.0079188346862793,
"val/ratio_var": 2.947521686553955
},
{
"episode": 328,
"epoch": 0.41,
"eps": 0,
"loss/policy_avg": 1.742775202728808e-05,
"loss/value_avg": 0.08753944933414459,
"lr": 3e-05,
"objective/entropy": 0.014935731887817383,
"objective/kl": 21.758373260498047,
"objective/non_score_reward": -1.0879186391830444,
"objective/rlhf_reward": -5.788181781768799,
"objective/scores": -4.700263023376465,
"policy/approxkl_avg": 8.582184563010742e-08,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.002552807331085205,
"step": 41,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 0.999954879283905,
"val/ratio_var": 1.7294539134127263e-09
},
{
"episode": 336,
"epoch": 0.42,
"eps": 0,
"loss/policy_avg": 3.205050006727106e-06,
"loss/value_avg": 0.1433621644973755,
"lr": 2.95e-05,
"objective/entropy": 0.0236055850982666,
"objective/kl": 22.17743492126465,
"objective/non_score_reward": -1.1088719367980957,
"objective/rlhf_reward": -5.329220294952393,
"objective/scores": -4.220348358154297,
"policy/approxkl_avg": 6.920163997392592e-08,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.0035035195760428905,
"step": 42,
"val/clipfrac_avg": 0.26768869161605835,
"val/num_eos_tokens": 0,
"val/ratio": 0.9999628067016602,
"val/ratio_var": 1.2399977711297083e-09
},
{
"episode": 344,
"epoch": 0.43,
"eps": 0,
"loss/policy_avg": -0.0001469778799219057,
"loss/value_avg": 0.06910190731287003,
"lr": 2.9e-05,
"objective/entropy": 4.376821517944336,
"objective/kl": 23.600093841552734,
"objective/non_score_reward": -1.1800047159194946,
"objective/rlhf_reward": -5.795012950897217,
"objective/scores": -4.615008354187012,
"policy/approxkl_avg": 0.0004178693052381277,
"policy/clipfrac_avg": 0.001179245300590992,
"policy/entropy_avg": 0.004530872218310833,
"step": 43,
"val/clipfrac_avg": 0.05896226316690445,
"val/num_eos_tokens": 0,
"val/ratio": 0.9992430806159973,
"val/ratio_var": 4.377442337499815e-07
},
{
"episode": 352,
"epoch": 0.44,
"eps": 0,
"loss/policy_avg": -2.1136806026333943e-05,
"loss/value_avg": 0.17057329416275024,
"lr": 2.8499999999999998e-05,
"objective/entropy": 0.029170513153076172,
"objective/kl": 23.519248962402344,
"objective/non_score_reward": -1.1759625673294067,
"objective/rlhf_reward": -6.7914838790893555,
"objective/scores": -5.615521430969238,
"policy/approxkl_avg": 3.24759383829587e-08,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.003596365451812744,
"step": 44,
"val/clipfrac_avg": 0.18278300762176514,
"val/num_eos_tokens": 0,
"val/ratio": 1.0000362396240234,
"val/ratio_var": 6.971466182115194e-10
},
{
"episode": 360,
"epoch": 0.45,
"eps": 0,
"loss/policy_avg": 9.945049896487035e-06,
"loss/value_avg": 0.07444664090871811,
"lr": 2.8000000000000003e-05,
"objective/entropy": 0.026226326823234558,
"objective/kl": 20.804168701171875,
"objective/non_score_reward": -1.0402084589004517,
"objective/rlhf_reward": -5.101940631866455,
"objective/scores": -4.061732292175293,
"policy/approxkl_avg": 1.7325481493912775e-08,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.0033669061958789825,
"step": 45,
"val/clipfrac_avg": 0.09021226316690445,
"val/num_eos_tokens": 0,
"val/ratio": 1.0000312328338623,
"val/ratio_var": 1.461017973269918e-10
},
{
"episode": 368,
"epoch": 0.46,
"eps": 0,
"loss/policy_avg": -2.804990072036162e-05,
"loss/value_avg": 0.1255272775888443,
"lr": 2.7500000000000004e-05,
"objective/entropy": 0.615976870059967,
"objective/kl": 21.274887084960938,
"objective/non_score_reward": -1.063744306564331,
"objective/rlhf_reward": -5.944283485412598,
"objective/scores": -4.8805389404296875,
"policy/approxkl_avg": 7.317021663766354e-05,
"policy/clipfrac_avg": 0.001179245300590992,
"policy/entropy_avg": 0.005477185361087322,
"step": 46,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 1.0005582571029663,
"val/ratio_var": 2.227838393764614e-07
},
{
"episode": 376,
"epoch": 0.47,
"eps": 0,
"loss/policy_avg": -0.0003541487967595458,
"loss/value_avg": 0.10688911378383636,
"lr": 2.7000000000000002e-05,
"objective/entropy": 5.619821548461914,
"objective/kl": 21.989046096801758,
"objective/non_score_reward": -1.0994524955749512,
"objective/rlhf_reward": -5.335377216339111,
"objective/scores": -4.23592472076416,
"policy/approxkl_avg": 0.0005093185463920236,
"policy/clipfrac_avg": 0.00294811325147748,
"policy/entropy_avg": 0.004899248480796814,
"step": 47,
"val/clipfrac_avg": 0.002358490601181984,
"val/num_eos_tokens": 0,
"val/ratio": 0.9987145662307739,
"val/ratio_var": 1.1634953125394532e-06
},
{
"episode": 384,
"epoch": 0.48,
"eps": 0,
"loss/policy_avg": -0.00010589355952106416,
"loss/value_avg": 0.178936168551445,
"lr": 2.6500000000000004e-05,
"objective/entropy": 6.092685699462891,
"objective/kl": 23.867069244384766,
"objective/non_score_reward": -1.1933534145355225,
"objective/rlhf_reward": -6.364969253540039,
"objective/scores": -5.1716156005859375,
"policy/approxkl_avg": 0.0002781845105346292,
"policy/clipfrac_avg": 0.001179245300590992,
"policy/entropy_avg": 0.004049578681588173,
"step": 48,
"val/clipfrac_avg": 0.036556605249643326,
"val/num_eos_tokens": 0,
"val/ratio": 1.0012478828430176,
"val/ratio_var": 1.7213611727129319e-06
},
{
"episode": 392,
"epoch": 0.49,
"eps": 0,
"loss/policy_avg": -0.0019225918222218752,
"loss/value_avg": 0.2773953974246979,
"lr": 2.6000000000000002e-05,
"objective/entropy": 12.097511291503906,
"objective/kl": 24.415565490722656,
"objective/non_score_reward": -1.220778465270996,
"objective/rlhf_reward": -7.17116641998291,
"objective/scores": -5.950387954711914,
"policy/approxkl_avg": 0.012523818761110306,
"policy/clipfrac_avg": 0.003537735901772976,
"policy/entropy_avg": 0.009605629369616508,
"step": 49,
"val/clipfrac_avg": 0.24174529314041138,
"val/num_eos_tokens": 0,
"val/ratio": 0.999590277671814,
"val/ratio_var": 6.160067300697847e-07
},
{
"episode": 400,
"epoch": 0.5,
"eps": 0,
"loss/policy_avg": -8.95777702680789e-06,
"loss/value_avg": 0.09243077039718628,
"lr": 2.5500000000000003e-05,
"objective/entropy": 0.029079481959342957,
"objective/kl": 21.414337158203125,
"objective/non_score_reward": -1.0707169771194458,
"objective/rlhf_reward": -5.165630340576172,
"objective/scores": -4.094913482666016,
"policy/approxkl_avg": 3.28439213603815e-08,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.00365700526162982,
"step": 50,
"val/clipfrac_avg": 0.17216980457305908,
"val/num_eos_tokens": 0,
"val/ratio": 0.9999852776527405,
"val/ratio_var": 6.366533966684074e-10
},
{
"episode": 408,
"epoch": 0.51,
"eps": 0,
"loss/policy_avg": -0.0005268001696094871,
"loss/value_avg": 0.3501094579696655,
"lr": 2.5e-05,
"objective/entropy": 7.087305545806885,
"objective/kl": 19.910663604736328,
"objective/non_score_reward": -0.9955330491065979,
"objective/rlhf_reward": -6.61014461517334,
"objective/scores": -5.614611625671387,
"policy/approxkl_avg": 0.0009080818854272366,
"policy/clipfrac_avg": 0.005306603852659464,
"policy/entropy_avg": 0.0055799842812120914,
"step": 51,
"val/clipfrac_avg": 0.1350235939025879,
"val/num_eos_tokens": 0,
"val/ratio": 1.0038042068481445,
"val/ratio_var": 1.224496190843638e-05
},
{
"episode": 416,
"epoch": 0.52,
"eps": 0,
"loss/policy_avg": -2.6310954126529396e-05,
"loss/value_avg": 0.15892276167869568,
"lr": 2.45e-05,
"objective/entropy": 0.056726157665252686,
"objective/kl": 22.944862365722656,
"objective/non_score_reward": -1.1472431421279907,
"objective/rlhf_reward": -5.8535990715026855,
"objective/scores": -4.706356048583984,
"policy/approxkl_avg": 1.523222181276651e-06,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.006282071582973003,
"step": 52,
"val/clipfrac_avg": 0.18278300762176514,
"val/num_eos_tokens": 0,
"val/ratio": 0.999813437461853,
"val/ratio_var": 2.821833788857475e-08
},
{
"episode": 424,
"epoch": 0.53,
"eps": 0,
"loss/policy_avg": 6.494516128441319e-06,
"loss/value_avg": 0.058799147605895996,
"lr": 2.4e-05,
"objective/entropy": 0.09326425194740295,
"objective/kl": 21.923828125,
"objective/non_score_reward": -1.0961912870407104,
"objective/rlhf_reward": -5.75281286239624,
"objective/scores": -4.65662145614624,
"policy/approxkl_avg": 2.059743565041572e-06,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.00874295923858881,
"step": 53,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 0.9997859597206116,
"val/ratio_var": 3.57754323943027e-08
},
{
"episode": 432,
"epoch": 0.54,
"eps": 0,
"loss/policy_avg": -2.164826037187595e-05,
"loss/value_avg": 0.04323159158229828,
"lr": 2.35e-05,
"objective/entropy": 0.11925557255744934,
"objective/kl": 21.405750274658203,
"objective/non_score_reward": -1.0702874660491943,
"objective/rlhf_reward": -5.812557220458984,
"objective/scores": -4.742269515991211,
"policy/approxkl_avg": 1.437954608718428e-07,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.009416119195520878,
"step": 54,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 1.0000391006469727,
"val/ratio_var": 5.505272238082171e-09
},
{
"episode": 440,
"epoch": 0.55,
"eps": 0,
"loss/policy_avg": -0.0003278666699770838,
"loss/value_avg": 0.0747484341263771,
"lr": 2.3000000000000003e-05,
"objective/entropy": 5.897934913635254,
"objective/kl": 22.183162689208984,
"objective/non_score_reward": -1.1091580390930176,
"objective/rlhf_reward": -5.195685863494873,
"objective/scores": -4.0865278244018555,
"policy/approxkl_avg": 0.0002924558939412236,
"policy/clipfrac_avg": 0.001768867950886488,
"policy/entropy_avg": 0.011092341504991055,
"step": 55,
"val/clipfrac_avg": 0.08195754885673523,
"val/num_eos_tokens": 0,
"val/ratio": 1.0012880563735962,
"val/ratio_var": 1.8590358195069712e-06
},
{
"episode": 448,
"epoch": 0.56,
"eps": 0,
"loss/policy_avg": -2.5927633032551967e-05,
"loss/value_avg": 0.1024797111749649,
"lr": 2.25e-05,
"objective/entropy": 0.43828973174095154,
"objective/kl": 21.43529510498047,
"objective/non_score_reward": -1.0717647075653076,
"objective/rlhf_reward": -5.657342910766602,
"objective/scores": -4.585577964782715,
"policy/approxkl_avg": 0.0002615810662973672,
"policy/clipfrac_avg": 0.001768867950886488,
"policy/entropy_avg": 0.015617892146110535,
"step": 56,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 0.9997985363006592,
"val/ratio_var": 2.572847712656312e-08
},
{
"episode": 456,
"epoch": 0.57,
"eps": 0,
"loss/policy_avg": -0.0005400018999353051,
"loss/value_avg": 0.052224770188331604,
"lr": 2.2000000000000003e-05,
"objective/entropy": 0.4916686415672302,
"objective/kl": 21.881145477294922,
"objective/non_score_reward": -1.094057321548462,
"objective/rlhf_reward": -5.429174423217773,
"objective/scores": -4.335117340087891,
"policy/approxkl_avg": 0.0003721084212884307,
"policy/clipfrac_avg": 0.004127358552068472,
"policy/entropy_avg": 0.018060289323329926,
"step": 57,
"val/clipfrac_avg": 0.06191037595272064,
"val/num_eos_tokens": 0,
"val/ratio": 0.9988338947296143,
"val/ratio_var": 1.326149572378199e-06
},
{
"episode": 464,
"epoch": 0.58,
"eps": 0,
"loss/policy_avg": -0.0007685062009841204,
"loss/value_avg": 0.46044227480888367,
"lr": 2.15e-05,
"objective/entropy": 5.862210750579834,
"objective/kl": 24.761629104614258,
"objective/non_score_reward": -1.238081455230713,
"objective/rlhf_reward": -5.525996685028076,
"objective/scores": -4.287915229797363,
"policy/approxkl_avg": 0.0017531004268676043,
"policy/clipfrac_avg": 0.008254717104136944,
"policy/entropy_avg": 0.01944148726761341,
"step": 58,
"val/clipfrac_avg": 0.033608488738536835,
"val/num_eos_tokens": 0,
"val/ratio": 0.9990991353988647,
"val/ratio_var": 4.7575713324476965e-07
},
{
"episode": 472,
"epoch": 0.59,
"eps": 0,
"loss/policy_avg": -0.00046633119927719235,
"loss/value_avg": 0.18473884463310242,
"lr": 2.1e-05,
"objective/entropy": 1.0018765926361084,
"objective/kl": 23.490116119384766,
"objective/non_score_reward": -1.1745058298110962,
"objective/rlhf_reward": -4.755178451538086,
"objective/scores": -3.5806727409362793,
"policy/approxkl_avg": 0.00026144221192225814,
"policy/clipfrac_avg": 0.001768867950886488,
"policy/entropy_avg": 0.012994904071092606,
"step": 59,
"val/clipfrac_avg": 0.23231130838394165,
"val/num_eos_tokens": 0,
"val/ratio": 1.0006680488586426,
"val/ratio_var": 2.8671664153989695e-07
},
{
"episode": 480,
"epoch": 0.6,
"eps": 0,
"loss/policy_avg": -0.00043274153722450137,
"loss/value_avg": 0.08987745642662048,
"lr": 2.05e-05,
"objective/entropy": 0.7680976390838623,
"objective/kl": 24.150394439697266,
"objective/non_score_reward": -1.2075196504592896,
"objective/rlhf_reward": -4.8026533126831055,
"objective/scores": -3.5951337814331055,
"policy/approxkl_avg": 0.0007885855156928301,
"policy/clipfrac_avg": 0.001768867950886488,
"policy/entropy_avg": 0.007194040343165398,
"step": 60,
"val/clipfrac_avg": 0.13030660152435303,
"val/num_eos_tokens": 0,
"val/ratio": 0.9989591836929321,
"val/ratio_var": 6.779022214686847e-07
},
{
"episode": 488,
"epoch": 0.61,
"eps": 0,
"loss/policy_avg": -0.00015176059969235212,
"loss/value_avg": 0.05639723315834999,
"lr": 2e-05,
"objective/entropy": 0.042793840169906616,
"objective/kl": 24.719331741333008,
"objective/non_score_reward": -1.235966682434082,
"objective/rlhf_reward": -4.184081077575684,
"objective/scores": -2.9481143951416016,
"policy/approxkl_avg": 1.515730446044472e-06,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.0038767289370298386,
"step": 61,
"val/clipfrac_avg": 0.2895047068595886,
"val/num_eos_tokens": 0,
"val/ratio": 1.0002093315124512,
"val/ratio_var": 2.6440673650540703e-08
},
{
"episode": 496,
"epoch": 0.62,
"eps": 0,
"loss/policy_avg": -0.00039324647514149547,
"loss/value_avg": 0.11202812939882278,
"lr": 1.9500000000000003e-05,
"objective/entropy": 0.8435674905776978,
"objective/kl": 23.85810089111328,
"objective/non_score_reward": -1.192905068397522,
"objective/rlhf_reward": -4.872404098510742,
"objective/scores": -3.6794989109039307,
"policy/approxkl_avg": 9.81152625172399e-05,
"policy/clipfrac_avg": 0.001179245300590992,
"policy/entropy_avg": 0.004706778563559055,
"step": 62,
"val/clipfrac_avg": 0.08136792480945587,
"val/num_eos_tokens": 0,
"val/ratio": 1.0006957054138184,
"val/ratio_var": 5.388393446992268e-07
},
{
"episode": 504,
"epoch": 0.63,
"eps": 0,
"loss/policy_avg": -0.00048710883129388094,
"loss/value_avg": 0.3882809281349182,
"lr": 1.9e-05,
"objective/entropy": 6.33875036239624,
"objective/kl": 23.255828857421875,
"objective/non_score_reward": -1.1627914905548096,
"objective/rlhf_reward": -5.508538246154785,
"objective/scores": -4.345746994018555,
"policy/approxkl_avg": 0.0013963093515485525,
"policy/clipfrac_avg": 0.00294811325147748,
"policy/entropy_avg": 0.00336352176964283,
"step": 63,
"val/clipfrac_avg": 0.1362028419971466,
"val/num_eos_tokens": 0,
"val/ratio": 0.9983474612236023,
"val/ratio_var": 1.6093348449430778e-06
},
{
"episode": 512,
"epoch": 0.64,
"eps": 0,
"loss/policy_avg": -0.0009707484859973192,
"loss/value_avg": 0.07156310975551605,
"lr": 1.85e-05,
"objective/entropy": 5.413853168487549,
"objective/kl": 25.648712158203125,
"objective/non_score_reward": -1.2824357748031616,
"objective/rlhf_reward": -4.381594181060791,
"objective/scores": -3.099158525466919,
"policy/approxkl_avg": 0.00428994745016098,
"policy/clipfrac_avg": 0.001768867950886488,
"policy/entropy_avg": 0.0022988244891166687,
"step": 64,
"val/clipfrac_avg": 0.10908018797636032,
"val/num_eos_tokens": 0,
"val/ratio": 0.9985601902008057,
"val/ratio_var": 1.0099387282025418e-06
},
{
"episode": 520,
"epoch": 0.65,
"eps": 0,
"loss/policy_avg": -0.0018268069252371788,
"loss/value_avg": 0.037753816694021225,
"lr": 1.8e-05,
"objective/entropy": 5.359340190887451,
"objective/kl": 23.45077133178711,
"objective/non_score_reward": -1.17253839969635,
"objective/rlhf_reward": -4.046453475952148,
"objective/scores": -2.873914957046509,
"policy/approxkl_avg": 0.030117690563201904,
"policy/clipfrac_avg": 0.001768867950886488,
"policy/entropy_avg": 0.0016830174718052149,
"step": 65,
"val/clipfrac_avg": 0.0383254699409008,
"val/num_eos_tokens": 0,
"val/ratio": 0.9982761144638062,
"val/ratio_var": 1.3298521253091167e-06
},
{
"episode": 528,
"epoch": 0.66,
"eps": 0,
"loss/policy_avg": -0.0002859890228137374,
"loss/value_avg": 0.021327916532754898,
"lr": 1.75e-05,
"objective/entropy": 6.843641757965088,
"objective/kl": 24.481531143188477,
"objective/non_score_reward": -1.224076509475708,
"objective/rlhf_reward": -4.482547283172607,
"objective/scores": -3.2584707736968994,
"policy/approxkl_avg": 0.008686334826052189,
"policy/clipfrac_avg": 0.001768867950886488,
"policy/entropy_avg": 0.0017893059412017465,
"step": 66,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 0.9984126091003418,
"val/ratio_var": 1.1801236041719676e-06
},
{
"episode": 536,
"epoch": 0.67,
"eps": 0,
"loss/policy_avg": -0.0005837440257892013,
"loss/value_avg": 0.053136035799980164,
"lr": 1.7000000000000003e-05,
"objective/entropy": 6.8793745040893555,
"objective/kl": 22.919111251831055,
"objective/non_score_reward": -1.1459555625915527,
"objective/rlhf_reward": -4.624814033508301,
"objective/scores": -3.478858232498169,
"policy/approxkl_avg": 0.0112453643232584,
"policy/clipfrac_avg": 0.001768867950886488,
"policy/entropy_avg": 0.001637791283428669,
"step": 67,
"val/clipfrac_avg": 0.10613207519054413,
"val/num_eos_tokens": 0,
"val/ratio": 0.9983804821968079,
"val/ratio_var": 1.2074211781509803e-06
},
{
"episode": 544,
"epoch": 0.68,
"eps": 0,
"loss/policy_avg": 0.00033806508872658014,
"loss/value_avg": 0.048039525747299194,
"lr": 1.65e-05,
"objective/entropy": 14.75048542022705,
"objective/kl": 23.478042602539062,
"objective/non_score_reward": -1.1739022731781006,
"objective/rlhf_reward": -4.242119789123535,
"objective/scores": -3.0682172775268555,
"policy/approxkl_avg": 0.027489835396409035,
"policy/clipfrac_avg": 0.001768867950886488,
"policy/entropy_avg": 0.001942639471963048,
"step": 68,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 0.9966699481010437,
"val/ratio_var": 5.0052558435709216e-06
},
{
"episode": 552,
"epoch": 0.69,
"eps": 0,
"loss/policy_avg": -0.00016136097838170826,
"loss/value_avg": 0.26613786816596985,
"lr": 1.6000000000000003e-05,
"objective/entropy": 5.09047269821167,
"objective/kl": 23.750701904296875,
"objective/non_score_reward": -1.1875351667404175,
"objective/rlhf_reward": -5.098639011383057,
"objective/scores": -3.9111037254333496,
"policy/approxkl_avg": 0.0023809224367141724,
"policy/clipfrac_avg": 0.001768867950886488,
"policy/entropy_avg": 0.0014511016197502613,
"step": 69,
"val/clipfrac_avg": 0.08254716545343399,
"val/num_eos_tokens": 0,
"val/ratio": 0.9986908435821533,
"val/ratio_var": 8.683288683641877e-07
},
{
"episode": 560,
"epoch": 0.7,
"eps": 0,
"loss/policy_avg": 5.313775091053685e-06,
"loss/value_avg": 0.04069013148546219,
"lr": 1.55e-05,
"objective/entropy": 0.004417300224304199,
"objective/kl": 22.818960189819336,
"objective/non_score_reward": -1.140947937965393,
"objective/rlhf_reward": -4.2426886558532715,
"objective/scores": -3.101740598678589,
"policy/approxkl_avg": 1.6673491476382196e-09,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.000669731292873621,
"step": 70,
"val/clipfrac_avg": 0.033018868416547775,
"val/num_eos_tokens": 0,
"val/ratio": 1.0000087022781372,
"val/ratio_var": 8.521776166670048e-12
},
{
"episode": 568,
"epoch": 0.71,
"eps": 0,
"loss/policy_avg": 3.097684384556487e-06,
"loss/value_avg": 0.07564935088157654,
"lr": 1.5e-05,
"objective/entropy": 0.0037894248962402344,
"objective/kl": 24.71213722229004,
"objective/non_score_reward": -1.2356069087982178,
"objective/rlhf_reward": -4.127857208251953,
"objective/scores": -2.8922500610351562,
"policy/approxkl_avg": 1.0003454731233319e-09,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.0005924965953454375,
"step": 71,
"val/clipfrac_avg": 0.1179245263338089,
"val/num_eos_tokens": 0,
"val/ratio": 1.000006079673767,
"val/ratio_var": 2.1884716261411086e-12
},
{
"episode": 576,
"epoch": 0.72,
"eps": 0,
"loss/policy_avg": -3.769466502490104e-06,
"loss/value_avg": 0.018075991421937943,
"lr": 1.45e-05,
"objective/entropy": 0.004113674163818359,
"objective/kl": 24.736343383789062,
"objective/non_score_reward": -1.2368173599243164,
"objective/rlhf_reward": -4.213069438934326,
"objective/scores": -2.9762520790100098,
"policy/approxkl_avg": 7.042492145004076e-10,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.0006491912645287812,
"step": 72,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 1.0000057220458984,
"val/ratio_var": 9.189685687763238e-13
},
{
"episode": 584,
"epoch": 0.73,
"eps": 0,
"loss/policy_avg": -8.439961675321683e-06,
"loss/value_avg": 0.02476009726524353,
"lr": 1.4000000000000001e-05,
"objective/entropy": 0.004063129425048828,
"objective/kl": 24.381113052368164,
"objective/non_score_reward": -1.2190558910369873,
"objective/rlhf_reward": -4.345273971557617,
"objective/scores": -3.12621808052063,
"policy/approxkl_avg": 1.6590684381867504e-09,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.0006231879233382642,
"step": 73,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 1.0000081062316895,
"val/ratio_var": 8.526512829121202e-13
},
{
"episode": 592,
"epoch": 0.74,
"eps": 0,
"loss/policy_avg": -0.0015291509917005897,
"loss/value_avg": 0.0400041788816452,
"lr": 1.3500000000000001e-05,
"objective/entropy": 5.482907295227051,
"objective/kl": 19.337940216064453,
"objective/non_score_reward": -0.9668970704078674,
"objective/rlhf_reward": -4.515492916107178,
"objective/scores": -3.548595666885376,
"policy/approxkl_avg": 0.04833262413740158,
"policy/clipfrac_avg": 0.0018703242531046271,
"policy/entropy_avg": 0.11862978339195251,
"step": 74,
"val/clipfrac_avg": 0.007462686393409967,
"val/num_eos_tokens": 1,
"val/ratio": 0.9982566833496094,
"val/ratio_var": 1.3634206652568537e-06
},
{
"episode": 600,
"epoch": 0.75,
"eps": 0,
"loss/policy_avg": 4.78032063710998e-07,
"loss/value_avg": 0.02071548067033291,
"lr": 1.3000000000000001e-05,
"objective/entropy": 0.003952503204345703,
"objective/kl": 24.335468292236328,
"objective/non_score_reward": -1.216773509979248,
"objective/rlhf_reward": -4.120009899139404,
"objective/scores": -2.9032363891601562,
"policy/approxkl_avg": 4.740483561249675e-10,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.0005925038713030517,
"step": 75,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 1.0000042915344238,
"val/ratio_var": 3.884300379172062e-13
},
{
"episode": 608,
"epoch": 0.76,
"eps": 0,
"loss/policy_avg": -1.0813885182869853e-06,
"loss/value_avg": 0.016514349728822708,
"lr": 1.25e-05,
"objective/entropy": 0.0032129287719726562,
"objective/kl": 24.535846710205078,
"objective/non_score_reward": -1.2267922163009644,
"objective/rlhf_reward": -4.2841620445251465,
"objective/scores": -3.0573699474334717,
"policy/approxkl_avg": 6.217855119672322e-10,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.0005110952188260853,
"step": 76,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 1.000004529953003,
"val/ratio_var": 2.03688913067053e-13
},
{
"episode": 616,
"epoch": 0.77,
"eps": 0,
"loss/policy_avg": -7.110946626198711e-06,
"loss/value_avg": 0.047897208482027054,
"lr": 1.2e-05,
"objective/entropy": 0.003764629364013672,
"objective/kl": 25.578466415405273,
"objective/non_score_reward": -1.2789231538772583,
"objective/rlhf_reward": -4.72192907333374,
"objective/scores": -3.4430060386657715,
"policy/approxkl_avg": 5.927374147063347e-09,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.0005579590797424316,
"step": 77,
"val/clipfrac_avg": 0.004127358552068472,
"val/num_eos_tokens": 0,
"val/ratio": 1.0000112056732178,
"val/ratio_var": 2.03688913067053e-13
},
{
"episode": 624,
"epoch": 0.78,
"eps": 0,
"loss/policy_avg": -3.5153384487784933e-06,
"loss/value_avg": 0.014893004670739174,
"lr": 1.1500000000000002e-05,
"objective/entropy": 0.003036022186279297,
"objective/kl": 24.764558792114258,
"objective/non_score_reward": -1.2382278442382812,
"objective/rlhf_reward": -4.514825820922852,
"objective/scores": -3.2765979766845703,
"policy/approxkl_avg": 1.0030707375818793e-09,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.0004918260383419693,
"step": 78,
"val/clipfrac_avg": 0.010613207705318928,
"val/num_eos_tokens": 0,
"val/ratio": 1.0000040531158447,
"val/ratio_var": 1.0421293683690255e-13
},
{
"episode": 632,
"epoch": 0.79,
"eps": 0,
"loss/policy_avg": -0.0017197050619870424,
"loss/value_avg": 0.059836748987436295,
"lr": 1.1000000000000001e-05,
"objective/entropy": 10.20138168334961,
"objective/kl": 18.15127182006836,
"objective/non_score_reward": -0.9075635075569153,
"objective/rlhf_reward": -4.070981502532959,
"objective/scores": -3.1634178161621094,
"policy/approxkl_avg": 0.672481894493103,
"policy/clipfrac_avg": 0.001768867950886488,
"policy/entropy_avg": 0.0007890131091699004,
"step": 79,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 118013739008.0,
"val/ratio_var": 5.060441798977168e+22
},
{
"episode": 640,
"epoch": 0.8,
"eps": 0,
"loss/policy_avg": -0.0004759056319016963,
"loss/value_avg": 0.014939786866307259,
"lr": 1.05e-05,
"objective/entropy": 7.734498977661133,
"objective/kl": 23.559078216552734,
"objective/non_score_reward": -1.1779539585113525,
"objective/rlhf_reward": -4.140482425689697,
"objective/scores": -2.9625284671783447,
"policy/approxkl_avg": 0.10851067304611206,
"policy/clipfrac_avg": 0.001768867950886488,
"policy/entropy_avg": 0.00210343929938972,
"step": 80,
"val/clipfrac_avg": 0.01945754699409008,
"val/num_eos_tokens": 0,
"val/ratio": 381.6966552734375,
"val/ratio_var": 399719.8125
},
{
"episode": 648,
"epoch": 0.81,
"eps": 0,
"loss/policy_avg": -0.0009191618300974369,
"loss/value_avg": 0.45779138803482056,
"lr": 1e-05,
"objective/entropy": 10.361689567565918,
"objective/kl": 25.474790573120117,
"objective/non_score_reward": -1.2737394571304321,
"objective/rlhf_reward": -5.300909042358398,
"objective/scores": -4.027169704437256,
"policy/approxkl_avg": 0.001493292860686779,
"policy/clipfrac_avg": 0.001768867950886488,
"policy/entropy_avg": 0.004847034811973572,
"step": 81,
"val/clipfrac_avg": 0.057193394750356674,
"val/num_eos_tokens": 0,
"val/ratio": 1.0019978284835815,
"val/ratio_var": 2.976292307721451e-06
},
{
"episode": 656,
"epoch": 0.82,
"eps": 0,
"loss/policy_avg": -0.0005446273135021329,
"loss/value_avg": 0.2114354372024536,
"lr": 9.5e-06,
"objective/entropy": 0.7502050399780273,
"objective/kl": 24.138385772705078,
"objective/non_score_reward": -1.2069193124771118,
"objective/rlhf_reward": -4.927611351013184,
"objective/scores": -3.7206919193267822,
"policy/approxkl_avg": 0.00017879472579807043,
"policy/clipfrac_avg": 0.001179245300590992,
"policy/entropy_avg": 0.013909805566072464,
"step": 82,
"val/clipfrac_avg": 0.04245283082127571,
"val/num_eos_tokens": 0,
"val/ratio": 0.999313235282898,
"val/ratio_var": 3.8654417267025565e-07
},
{
"episode": 664,
"epoch": 0.83,
"eps": 0,
"loss/policy_avg": -0.0008736238232813776,
"loss/value_avg": 0.019180217757821083,
"lr": 9e-06,
"objective/entropy": 0.9455151557922363,
"objective/kl": 24.899396896362305,
"objective/non_score_reward": -1.2449698448181152,
"objective/rlhf_reward": -4.538572311401367,
"objective/scores": -3.293602705001831,
"policy/approxkl_avg": 0.00010491647117305547,
"policy/clipfrac_avg": 0.001179245300590992,
"policy/entropy_avg": 0.006164656486362219,
"step": 83,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 0.999660849571228,
"val/ratio_var": 1.1305993297128225e-07
},
{
"episode": 672,
"epoch": 0.84,
"eps": 0,
"loss/policy_avg": 1.0482122888788581e-05,
"loss/value_avg": 0.058847103267908096,
"lr": 8.500000000000002e-06,
"objective/entropy": 0.026853561401367188,
"objective/kl": 27.00433349609375,
"objective/non_score_reward": -1.3502166271209717,
"objective/rlhf_reward": -4.081714153289795,
"objective/scores": -2.7314975261688232,
"policy/approxkl_avg": 6.986047296209108e-09,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.004489392973482609,
"step": 84,
"val/clipfrac_avg": 0.03537735715508461,
"val/num_eos_tokens": 0,
"val/ratio": 1.0000537633895874,
"val/ratio_var": 1.511182290414581e-09
},
{
"episode": 680,
"epoch": 0.85,
"eps": 0,
"loss/policy_avg": -2.936266810138477e-06,
"loss/value_avg": 0.016916655004024506,
"lr": 8.000000000000001e-06,
"objective/entropy": 0.020194053649902344,
"objective/kl": 24.650644302368164,
"objective/non_score_reward": -1.2325321435928345,
"objective/rlhf_reward": -4.201452255249023,
"objective/scores": -2.9689202308654785,
"policy/approxkl_avg": 1.4028140871147343e-09,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.0036443518474698067,
"step": 85,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 1.0000245571136475,
"val/ratio_var": 3.302128848137187e-10
},
{
"episode": 688,
"epoch": 0.86,
"eps": 0,
"loss/policy_avg": -0.00116700679063797,
"loss/value_avg": 0.03304029628634453,
"lr": 7.5e-06,
"objective/entropy": 10.308148384094238,
"objective/kl": 26.107011795043945,
"objective/non_score_reward": -1.3053505420684814,
"objective/rlhf_reward": -4.5653228759765625,
"objective/scores": -3.259972095489502,
"policy/approxkl_avg": 0.0007528926944360137,
"policy/clipfrac_avg": 0.003537735901772976,
"policy/entropy_avg": 0.003782036015763879,
"step": 86,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 0.9984903931617737,
"val/ratio_var": 1.3541301768782432e-06
},
{
"episode": 696,
"epoch": 0.87,
"eps": 0,
"loss/policy_avg": -0.0012483038008213043,
"loss/value_avg": 0.020644810050725937,
"lr": 7.000000000000001e-06,
"objective/entropy": 4.934552192687988,
"objective/kl": 25.656095504760742,
"objective/non_score_reward": -1.2828046083450317,
"objective/rlhf_reward": -4.459787845611572,
"objective/scores": -3.17698335647583,
"policy/approxkl_avg": 0.0011864990228787065,
"policy/clipfrac_avg": 0.001768867950886488,
"policy/entropy_avg": 0.0023125973530113697,
"step": 87,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 0.9989151358604431,
"val/ratio_var": 6.361988198477775e-07
},
{
"episode": 704,
"epoch": 0.88,
"eps": 0,
"loss/policy_avg": -0.0006414442323148251,
"loss/value_avg": 0.36193448305130005,
"lr": 6.5000000000000004e-06,
"objective/entropy": 4.682817459106445,
"objective/kl": 25.37774658203125,
"objective/non_score_reward": -1.2688874006271362,
"objective/rlhf_reward": -4.902969837188721,
"objective/scores": -3.634082555770874,
"policy/approxkl_avg": 0.00025906978407874703,
"policy/clipfrac_avg": 0.001179245300590992,
"policy/entropy_avg": 0.0020385892130434513,
"step": 88,
"val/clipfrac_avg": 0.03773584961891174,
"val/num_eos_tokens": 0,
"val/ratio": 0.9993649125099182,
"val/ratio_var": 2.928149456238316e-07
},
{
"episode": 712,
"epoch": 0.89,
"eps": 0,
"loss/policy_avg": -0.001403487753123045,
"loss/value_avg": 0.045823924243450165,
"lr": 6e-06,
"objective/entropy": 7.662001132965088,
"objective/kl": 26.13595962524414,
"objective/non_score_reward": -1.306797981262207,
"objective/rlhf_reward": -4.496212005615234,
"objective/scores": -3.1894140243530273,
"policy/approxkl_avg": 0.001071479171514511,
"policy/clipfrac_avg": 0.00294811325147748,
"policy/entropy_avg": 0.0017833821475505829,
"step": 89,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 0.9985510110855103,
"val/ratio_var": 1.2452730970835546e-06
},
{
"episode": 720,
"epoch": 0.9,
"eps": 0,
"loss/policy_avg": 2.8316421776253264e-06,
"loss/value_avg": 0.02468748763203621,
"lr": 5.500000000000001e-06,
"objective/entropy": 0.006358623504638672,
"objective/kl": 25.177474975585938,
"objective/non_score_reward": -1.258873701095581,
"objective/rlhf_reward": -4.340834617614746,
"objective/scores": -3.081960678100586,
"policy/approxkl_avg": 3.0429461728154195e-10,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.0012125474167987704,
"step": 90,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 1.000009298324585,
"val/ratio_var": 3.510081114654895e-11
},
{
"episode": 728,
"epoch": 0.91,
"eps": 0,
"loss/policy_avg": -0.003975835628807545,
"loss/value_avg": 0.0309587549418211,
"lr": 5e-06,
"objective/entropy": 9.262920379638672,
"objective/kl": 27.77571678161621,
"objective/non_score_reward": -1.3887858390808105,
"objective/rlhf_reward": -4.151398658752441,
"objective/scores": -2.762612819671631,
"policy/approxkl_avg": 0.0016623970586806536,
"policy/clipfrac_avg": 0.004716981202363968,
"policy/entropy_avg": 0.025041375309228897,
"step": 91,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 0.9995808601379395,
"val/ratio_var": 2.4279825083795004e-07
},
{
"episode": 736,
"epoch": 0.92,
"eps": 0,
"loss/policy_avg": -3.4936194879264804e-06,
"loss/value_avg": 0.021127665415406227,
"lr": 4.5e-06,
"objective/entropy": 0.003822803497314453,
"objective/kl": 24.14543342590332,
"objective/non_score_reward": -1.207271695137024,
"objective/rlhf_reward": -4.3759307861328125,
"objective/scores": -3.168658971786499,
"policy/approxkl_avg": 2.551990563315343e-10,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.000724360637832433,
"step": 92,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 1.000006914138794,
"val/ratio_var": 1.4746129664566787e-11
},
{
"episode": 744,
"epoch": 0.93,
"eps": 0,
"loss/policy_avg": -0.0001179491518996656,
"loss/value_avg": 0.0367477647960186,
"lr": 4.000000000000001e-06,
"objective/entropy": 5.264230251312256,
"objective/kl": 24.125144958496094,
"objective/non_score_reward": -1.2062572240829468,
"objective/rlhf_reward": -4.48881721496582,
"objective/scores": -3.282560110092163,
"policy/approxkl_avg": 3.359060792718083e-05,
"policy/clipfrac_avg": 0.000589622650295496,
"policy/entropy_avg": 0.0007435465813614428,
"step": 93,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 0.9997267127037048,
"val/ratio_var": 6.120335171999614e-08
},
{
"episode": 752,
"epoch": 0.94,
"eps": 0,
"loss/policy_avg": -0.0003386925673112273,
"loss/value_avg": 0.010906368494033813,
"lr": 3.5000000000000004e-06,
"objective/entropy": 5.7897210121154785,
"objective/kl": 23.875186920166016,
"objective/non_score_reward": -1.1937594413757324,
"objective/rlhf_reward": -4.192901611328125,
"objective/scores": -2.9991419315338135,
"policy/approxkl_avg": 0.00011795548925874755,
"policy/clipfrac_avg": 0.001179245300590992,
"policy/entropy_avg": 0.0006920391460880637,
"step": 94,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 0.9995193481445312,
"val/ratio_var": 1.679601666637609e-07
},
{
"episode": 760,
"epoch": 0.95,
"eps": 0,
"loss/policy_avg": -7.089016889949562e-06,
"loss/value_avg": 0.023602645844221115,
"lr": 3e-06,
"objective/entropy": 0.0036764144897460938,
"objective/kl": 24.60375213623047,
"objective/non_score_reward": -1.2301876544952393,
"objective/rlhf_reward": -4.6299896240234375,
"objective/scores": -3.399801731109619,
"policy/approxkl_avg": 8.320251154714242e-09,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.0006472456734627485,
"step": 95,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 1.0000110864639282,
"val/ratio_var": 1.4495071809506044e-12
},
{
"episode": 768,
"epoch": 0.96,
"eps": 0,
"loss/policy_avg": 4.672603211020032e-07,
"loss/value_avg": 0.0325840599834919,
"lr": 2.5e-06,
"objective/entropy": 0.0027709007263183594,
"objective/kl": 24.18463897705078,
"objective/non_score_reward": -1.2092320919036865,
"objective/rlhf_reward": -4.501224517822266,
"objective/scores": -3.291992664337158,
"policy/approxkl_avg": 1.2074271582562801e-09,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.0005334683228284121,
"step": 96,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 1.0000051259994507,
"val/ratio_var": 3.268496584496461e-13
},
{
"episode": 776,
"epoch": 0.97,
"eps": 0,
"loss/policy_avg": -0.0007012633141130209,
"loss/value_avg": 0.025306837633252144,
"lr": 2.0000000000000003e-06,
"objective/entropy": 6.557465076446533,
"objective/kl": 24.751359939575195,
"objective/non_score_reward": -1.2375680208206177,
"objective/rlhf_reward": -4.222304821014404,
"objective/scores": -2.984736919403076,
"policy/approxkl_avg": 0.00011602477025007829,
"policy/clipfrac_avg": 0.001179245300590992,
"policy/entropy_avg": 0.0004479637718759477,
"step": 97,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 0.9995274543762207,
"val/ratio_var": 1.737082584440941e-07
},
{
"episode": 784,
"epoch": 0.98,
"eps": 0,
"loss/policy_avg": -3.1152703741099685e-05,
"loss/value_avg": 0.01748664304614067,
"lr": 1.5e-06,
"objective/entropy": 6.415027141571045,
"objective/kl": 23.67858123779297,
"objective/non_score_reward": -1.1839290857315063,
"objective/rlhf_reward": -4.280977249145508,
"objective/scores": -3.097048282623291,
"policy/approxkl_avg": 2.8390975785441697e-05,
"policy/clipfrac_avg": 0.000589622650295496,
"policy/entropy_avg": 0.00046760181430727243,
"step": 98,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 0.9997364282608032,
"val/ratio_var": 4.89392171232339e-08
},
{
"episode": 792,
"epoch": 0.99,
"eps": 0,
"loss/policy_avg": -0.00032194965751841664,
"loss/value_avg": 0.17389172315597534,
"lr": 1.0000000000000002e-06,
"objective/entropy": 1.3875689506530762,
"objective/kl": 23.84311866760254,
"objective/non_score_reward": -1.1921560764312744,
"objective/rlhf_reward": -5.213768005371094,
"objective/scores": -4.02161169052124,
"policy/approxkl_avg": 7.786412425048184e-06,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.0005042271804995835,
"step": 99,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 0.9998690485954285,
"val/ratio_var": 1.8143422764183015e-08
},
{
"episode": 800,
"epoch": 1.0,
"eps": 0,
"loss/policy_avg": -0.0009410554193891585,
"loss/value_avg": 0.45231789350509644,
"lr": 5.000000000000001e-07,
"objective/entropy": 8.080781936645508,
"objective/kl": 22.65479278564453,
"objective/non_score_reward": -1.1327396631240845,
"objective/rlhf_reward": -5.200520992279053,
"objective/scores": -4.067781448364258,
"policy/approxkl_avg": 2.2665039068670012e-05,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.00041846060776151717,
"step": 100,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 0,
"val/ratio": 0.9996796250343323,
"val/ratio_var": 1.0175851627991506e-07
}
],
"logging_steps": 10,
"max_steps": 100,
"num_input_tokens_seen": 0,
"num_train_epochs": 1.0,
"save_steps": 500,
"stateful_callbacks": {
"TrainerControl": {
"args": {
"should_epoch_stop": false,
"should_evaluate": false,
"should_log": true,
"should_save": true,
"should_training_stop": true
},
"attributes": {}
}
},
"total_flos": 0,
"train_batch_size": null,
"trial_name": null,
"trial_params": null
}