Muqeeth's picture
Training in progress, step 34, checkpoint
7ff5d57 verified
{
"best_global_step": null,
"best_metric": null,
"best_model_checkpoint": null,
"episode": 17408,
"epoch": 0.07457162440027416,
"eval_steps": 500,
"global_step": 34,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"episode": 512,
"epoch": 0.0021932830705962986,
"eps": 6,
"loss/policy_avg": 0.18060211837291718,
"lr": 3e-06,
"objective/entropy": -44.39540100097656,
"objective/kl": 0.23749515414237976,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 4.621584892272949,
"objective/scores": 4.621584892272949,
"policy/approxkl_avg": 0.09178640693426132,
"policy/clipfrac_avg": 0.271484375,
"policy/entropy_avg": 0.7151485681533813,
"step": 0,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 5024,
"val/ratio": 1.0001094341278076,
"val/ratio_var": 1.1996758075838443e-06
},
{
"episode": 1024,
"epoch": 0.004386566141192597,
"eps": 6,
"loss/policy_avg": 0.10823587328195572,
"lr": 2.9882812500000002e-06,
"objective/entropy": -40.33261489868164,
"objective/kl": 0.8562099933624268,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 4.949611186981201,
"objective/scores": 4.949611186981201,
"policy/approxkl_avg": 0.09872549772262573,
"policy/clipfrac_avg": 0.29296875,
"policy/entropy_avg": 0.6909540891647339,
"step": 1,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 3988,
"val/ratio": 1.0000126361846924,
"val/ratio_var": 1.5850138197492925e-06
},
{
"episode": 1536,
"epoch": 0.006579849211788897,
"eps": 6,
"loss/policy_avg": 0.24130496382713318,
"lr": 2.9765625e-06,
"objective/entropy": -37.140586853027344,
"objective/kl": 2.1461260318756104,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 5.433126449584961,
"objective/scores": 5.433126449584961,
"policy/approxkl_avg": 0.1151532381772995,
"policy/clipfrac_avg": 0.30078125,
"policy/entropy_avg": 0.6479897499084473,
"step": 2,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 2924,
"val/ratio": 0.999896764755249,
"val/ratio_var": 1.4944761232982273e-06
},
{
"episode": 2048,
"epoch": 0.008773132282385195,
"eps": 6,
"loss/policy_avg": 0.11201904714107513,
"lr": 2.96484375e-06,
"objective/entropy": -23.162431716918945,
"objective/kl": 4.706851005554199,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 5.616954803466797,
"objective/scores": 5.616954803466797,
"policy/approxkl_avg": 0.13956432044506073,
"policy/clipfrac_avg": 0.365234375,
"policy/entropy_avg": 0.6664378643035889,
"step": 3,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 2146,
"val/ratio": 0.9996535778045654,
"val/ratio_var": 2.1856692455912707e-06
},
{
"episode": 2560,
"epoch": 0.010966415352981495,
"eps": 6,
"loss/policy_avg": 0.21055178344249725,
"lr": 2.953125e-06,
"objective/entropy": -17.498517990112305,
"objective/kl": 6.893625259399414,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 5.975703716278076,
"objective/scores": 5.975703716278076,
"policy/approxkl_avg": 0.14695778489112854,
"policy/clipfrac_avg": 0.33203125,
"policy/entropy_avg": 0.6391204595565796,
"step": 4,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 1715,
"val/ratio": 0.9999969005584717,
"val/ratio_var": 2.4044477413553977e-06
},
{
"episode": 3072,
"epoch": 0.013159698423577794,
"eps": 6,
"loss/policy_avg": 0.28335219621658325,
"lr": 2.94140625e-06,
"objective/entropy": -16.186603546142578,
"objective/kl": 8.215164184570312,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 6.314523696899414,
"objective/scores": 6.314523696899414,
"policy/approxkl_avg": 0.15406616032123566,
"policy/clipfrac_avg": 0.34375,
"policy/entropy_avg": 0.6234536170959473,
"step": 5,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 1743,
"val/ratio": 1.0000813007354736,
"val/ratio_var": 2.2491783511213725e-06
},
{
"episode": 3584,
"epoch": 0.015352981494174092,
"eps": 6,
"loss/policy_avg": 0.2581750154495239,
"lr": 2.9296875e-06,
"objective/entropy": -16.82525634765625,
"objective/kl": 10.012128829956055,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 6.68141508102417,
"objective/scores": 6.68141508102417,
"policy/approxkl_avg": 0.15727216005325317,
"policy/clipfrac_avg": 0.306640625,
"policy/entropy_avg": 0.6279575824737549,
"step": 6,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 2024,
"val/ratio": 1.0000149011611938,
"val/ratio_var": 3.910876330337487e-06
},
{
"episode": 4096,
"epoch": 0.01754626456477039,
"eps": 6,
"loss/policy_avg": 0.4508611857891083,
"lr": 2.91796875e-06,
"objective/entropy": -13.531968116760254,
"objective/kl": 11.674339294433594,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 6.797825813293457,
"objective/scores": 6.797825813293457,
"policy/approxkl_avg": 0.17787906527519226,
"policy/clipfrac_avg": 0.333984375,
"policy/entropy_avg": 0.6140714287757874,
"step": 7,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 2248,
"val/ratio": 0.9998712539672852,
"val/ratio_var": 2.3673501345911063e-06
},
{
"episode": 4608,
"epoch": 0.01973954763536669,
"eps": 6,
"loss/policy_avg": 0.25109419226646423,
"lr": 2.90625e-06,
"objective/entropy": -15.348489761352539,
"objective/kl": 12.88166332244873,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 7.097723484039307,
"objective/scores": 7.097723484039307,
"policy/approxkl_avg": 0.18050235509872437,
"policy/clipfrac_avg": 0.302734375,
"policy/entropy_avg": 0.6190727949142456,
"step": 8,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 2698,
"val/ratio": 1.000298023223877,
"val/ratio_var": 2.175341933252639e-06
},
{
"episode": 5120,
"epoch": 0.02193283070596299,
"eps": 6,
"loss/policy_avg": 0.2478555142879486,
"lr": 2.89453125e-06,
"objective/entropy": -20.807220458984375,
"objective/kl": 12.126293182373047,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 7.294116020202637,
"objective/scores": 7.294116020202637,
"policy/approxkl_avg": 0.16978076100349426,
"policy/clipfrac_avg": 0.31640625,
"policy/entropy_avg": 0.5916793346405029,
"step": 9,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 3166,
"val/ratio": 0.999906063079834,
"val/ratio_var": 2.3764121124258963e-06
},
{
"episode": 5632,
"epoch": 0.024126113776559287,
"eps": 6,
"loss/policy_avg": 0.2194754183292389,
"lr": 2.8828125e-06,
"objective/entropy": -23.919815063476562,
"objective/kl": 13.048431396484375,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 7.783618927001953,
"objective/scores": 7.783618927001953,
"policy/approxkl_avg": 0.1719370186328888,
"policy/clipfrac_avg": 0.3359375,
"policy/entropy_avg": 0.6023513078689575,
"step": 10,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 3395,
"val/ratio": 0.9998449683189392,
"val/ratio_var": 1.784112214409106e-06
},
{
"episode": 6144,
"epoch": 0.026319396847155587,
"eps": 6,
"loss/policy_avg": 0.13436169922351837,
"lr": 2.87109375e-06,
"objective/entropy": -21.395191192626953,
"objective/kl": 15.173453330993652,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 7.682435512542725,
"objective/scores": 7.682435512542725,
"policy/approxkl_avg": 0.19314789772033691,
"policy/clipfrac_avg": 0.369140625,
"policy/entropy_avg": 0.6016405820846558,
"step": 11,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 3629,
"val/ratio": 0.9996920824050903,
"val/ratio_var": 1.8341333998250775e-06
},
{
"episode": 6656,
"epoch": 0.028512679917751884,
"eps": 6,
"loss/policy_avg": 0.27384719252586365,
"lr": 2.859375e-06,
"objective/entropy": -23.660961151123047,
"objective/kl": 15.459722518920898,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 8.303886413574219,
"objective/scores": 8.303886413574219,
"policy/approxkl_avg": 0.1933055967092514,
"policy/clipfrac_avg": 0.359375,
"policy/entropy_avg": 0.5889922976493835,
"step": 12,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 3434,
"val/ratio": 1.0000567436218262,
"val/ratio_var": 2.589923496998381e-06
},
{
"episode": 7168,
"epoch": 0.030705962988348184,
"eps": 6,
"loss/policy_avg": 0.20046143233776093,
"lr": 2.84765625e-06,
"objective/entropy": -24.642715454101562,
"objective/kl": 17.541698455810547,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 8.798042297363281,
"objective/scores": 8.798042297363281,
"policy/approxkl_avg": 0.1510610282421112,
"policy/clipfrac_avg": 0.337890625,
"policy/entropy_avg": 0.552618145942688,
"step": 13,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 3045,
"val/ratio": 1.0000343322753906,
"val/ratio_var": 2.4570517780375667e-06
},
{
"episode": 7680,
"epoch": 0.03289924605894448,
"eps": 6,
"loss/policy_avg": 0.22966830432415009,
"lr": 2.8359375e-06,
"objective/entropy": -19.635862350463867,
"objective/kl": 18.844629287719727,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 8.394830703735352,
"objective/scores": 8.394830703735352,
"policy/approxkl_avg": 0.1538720726966858,
"policy/clipfrac_avg": 0.3359375,
"policy/entropy_avg": 0.5296885371208191,
"step": 14,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 2402,
"val/ratio": 1.0000078678131104,
"val/ratio_var": 2.192526608268963e-06
},
{
"episode": 8192,
"epoch": 0.03509252912954078,
"eps": 6,
"loss/policy_avg": 0.2523574233055115,
"lr": 2.82421875e-06,
"objective/entropy": -19.664382934570312,
"objective/kl": 19.9348201751709,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 8.683082580566406,
"objective/scores": 8.683082580566406,
"policy/approxkl_avg": 0.216833233833313,
"policy/clipfrac_avg": 0.34375,
"policy/entropy_avg": 0.5138433575630188,
"step": 15,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 2232,
"val/ratio": 0.9998021721839905,
"val/ratio_var": 1.7273125649808208e-06
},
{
"episode": 8704,
"epoch": 0.03728581220013708,
"eps": 6,
"loss/policy_avg": 0.1616867184638977,
"lr": 2.8125e-06,
"objective/entropy": -19.297401428222656,
"objective/kl": 20.52892303466797,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 8.915764808654785,
"objective/scores": 8.915764808654785,
"policy/approxkl_avg": 0.1600520759820938,
"policy/clipfrac_avg": 0.318359375,
"policy/entropy_avg": 0.47873735427856445,
"step": 16,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 1931,
"val/ratio": 0.9997690916061401,
"val/ratio_var": 2.2947442630538717e-06
},
{
"episode": 9216,
"epoch": 0.03947909527073338,
"eps": 6,
"loss/policy_avg": 0.163500115275383,
"lr": 2.80078125e-06,
"objective/entropy": -16.193389892578125,
"objective/kl": 22.477802276611328,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 8.87047290802002,
"objective/scores": 8.87047290802002,
"policy/approxkl_avg": 0.17684966325759888,
"policy/clipfrac_avg": 0.33984375,
"policy/entropy_avg": 0.4807279706001282,
"step": 17,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 1798,
"val/ratio": 0.9998293519020081,
"val/ratio_var": 2.4208854938478908e-06
},
{
"episode": 9728,
"epoch": 0.041672378341329676,
"eps": 6,
"loss/policy_avg": 0.19272606074810028,
"lr": 2.7890625e-06,
"objective/entropy": -12.483476638793945,
"objective/kl": 24.110105514526367,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 8.77183723449707,
"objective/scores": 8.77183723449707,
"policy/approxkl_avg": 0.1895884871482849,
"policy/clipfrac_avg": 0.36328125,
"policy/entropy_avg": 0.4732131361961365,
"step": 18,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 1580,
"val/ratio": 1.0000896453857422,
"val/ratio_var": 2.6988111585524166e-06
},
{
"episode": 10240,
"epoch": 0.04386566141192598,
"eps": 6,
"loss/policy_avg": 0.31339162588119507,
"lr": 2.77734375e-06,
"objective/entropy": -14.506279945373535,
"objective/kl": 23.571977615356445,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 8.617586135864258,
"objective/scores": 8.617586135864258,
"policy/approxkl_avg": 0.18291257321834564,
"policy/clipfrac_avg": 0.349609375,
"policy/entropy_avg": 0.4734336733818054,
"step": 19,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 1565,
"val/ratio": 0.9999023675918579,
"val/ratio_var": 2.692267571546836e-06
},
{
"episode": 10752,
"epoch": 0.04605894448252228,
"eps": 6,
"loss/policy_avg": 0.4666588306427002,
"lr": 2.765625e-06,
"objective/entropy": -12.953289985656738,
"objective/kl": 24.32935333251953,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 8.778571128845215,
"objective/scores": 8.778571128845215,
"policy/approxkl_avg": 0.20027068257331848,
"policy/clipfrac_avg": 0.30859375,
"policy/entropy_avg": 0.4659819006919861,
"step": 20,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 1506,
"val/ratio": 1.0001583099365234,
"val/ratio_var": 3.233054258089396e-06
},
{
"episode": 11264,
"epoch": 0.048252227553118573,
"eps": 6,
"loss/policy_avg": 0.27946847677230835,
"lr": 2.75390625e-06,
"objective/entropy": -11.48613166809082,
"objective/kl": 24.395483016967773,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 9.02904224395752,
"objective/scores": 9.02904224395752,
"policy/approxkl_avg": 0.18909631669521332,
"policy/clipfrac_avg": 0.333984375,
"policy/entropy_avg": 0.48570871353149414,
"step": 21,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 1560,
"val/ratio": 1.0002957582473755,
"val/ratio_var": 2.959979838124127e-06
},
{
"episode": 11776,
"epoch": 0.05044551062371487,
"eps": 6,
"loss/policy_avg": 0.2717439830303192,
"lr": 2.7421875e-06,
"objective/entropy": -9.29305362701416,
"objective/kl": 26.03940200805664,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 9.028656005859375,
"objective/scores": 9.028656005859375,
"policy/approxkl_avg": 0.1726618856191635,
"policy/clipfrac_avg": 0.34765625,
"policy/entropy_avg": 0.5084494352340698,
"step": 22,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 1547,
"val/ratio": 1.0000401735305786,
"val/ratio_var": 3.0132259780657478e-06
},
{
"episode": 12288,
"epoch": 0.052638793694311174,
"eps": 6,
"loss/policy_avg": 0.19815856218338013,
"lr": 2.73046875e-06,
"objective/entropy": -13.52733039855957,
"objective/kl": 25.79766273498535,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 9.250341415405273,
"objective/scores": 9.250341415405273,
"policy/approxkl_avg": 0.1738530993461609,
"policy/clipfrac_avg": 0.3125,
"policy/entropy_avg": 0.4977012574672699,
"step": 23,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 1648,
"val/ratio": 0.9999476671218872,
"val/ratio_var": 2.497435616533039e-06
},
{
"episode": 12800,
"epoch": 0.05483207676490747,
"eps": 6,
"loss/policy_avg": 0.2345161885023117,
"lr": 2.71875e-06,
"objective/entropy": -9.971322059631348,
"objective/kl": 26.595800399780273,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 9.097506523132324,
"objective/scores": 9.097506523132324,
"policy/approxkl_avg": 0.18735162913799286,
"policy/clipfrac_avg": 0.33203125,
"policy/entropy_avg": 0.4894530177116394,
"step": 24,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 1580,
"val/ratio": 1.000110149383545,
"val/ratio_var": 2.1261726033117156e-06
},
{
"episode": 13312,
"epoch": 0.05702535983550377,
"eps": 6,
"loss/policy_avg": 0.2765505909919739,
"lr": 2.70703125e-06,
"objective/entropy": -8.664510726928711,
"objective/kl": 27.207202911376953,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 9.248998641967773,
"objective/scores": 9.248998641967773,
"policy/approxkl_avg": 0.21922221779823303,
"policy/clipfrac_avg": 0.3671875,
"policy/entropy_avg": 0.4936632513999939,
"step": 25,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 1689,
"val/ratio": 0.9997528195381165,
"val/ratio_var": 2.2092231120041106e-06
},
{
"episode": 13824,
"epoch": 0.05921864290610007,
"eps": 6,
"loss/policy_avg": 0.20894497632980347,
"lr": 2.6953125e-06,
"objective/entropy": -13.794546127319336,
"objective/kl": 27.585330963134766,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 9.134411811828613,
"objective/scores": 9.134411811828613,
"policy/approxkl_avg": 0.19758901000022888,
"policy/clipfrac_avg": 0.337890625,
"policy/entropy_avg": 0.4570607841014862,
"step": 26,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 1664,
"val/ratio": 0.999937117099762,
"val/ratio_var": 3.2459099656989565e-06
},
{
"episode": 14336,
"epoch": 0.06141192597669637,
"eps": 6,
"loss/policy_avg": 0.187041237950325,
"lr": 2.68359375e-06,
"objective/entropy": -18.817930221557617,
"objective/kl": 26.306095123291016,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 9.336585998535156,
"objective/scores": 9.336585998535156,
"policy/approxkl_avg": 0.20577886700630188,
"policy/clipfrac_avg": 0.279296875,
"policy/entropy_avg": 0.4518057107925415,
"step": 27,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 1731,
"val/ratio": 1.0001513957977295,
"val/ratio_var": 2.708766032810672e-06
},
{
"episode": 14848,
"epoch": 0.06360520904729267,
"eps": 6,
"loss/policy_avg": 0.15964296460151672,
"lr": 2.671875e-06,
"objective/entropy": -19.800785064697266,
"objective/kl": 27.51376724243164,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 9.47825813293457,
"objective/scores": 9.47825813293457,
"policy/approxkl_avg": 0.19460758566856384,
"policy/clipfrac_avg": 0.34765625,
"policy/entropy_avg": 0.4188844859600067,
"step": 28,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 1687,
"val/ratio": 0.9997619390487671,
"val/ratio_var": 2.6442203306942247e-06
},
{
"episode": 15360,
"epoch": 0.06579849211788896,
"eps": 6,
"loss/policy_avg": 0.15885861217975616,
"lr": 2.66015625e-06,
"objective/entropy": -17.79833221435547,
"objective/kl": 27.949443817138672,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 9.496625900268555,
"objective/scores": 9.496625900268555,
"policy/approxkl_avg": 0.16567930579185486,
"policy/clipfrac_avg": 0.34765625,
"policy/entropy_avg": 0.4135560691356659,
"step": 29,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 1689,
"val/ratio": 1.0000743865966797,
"val/ratio_var": 3.2084180929814465e-06
},
{
"episode": 15872,
"epoch": 0.06799177518848526,
"eps": 6,
"loss/policy_avg": 0.16605864465236664,
"lr": 2.6484375e-06,
"objective/entropy": -18.510845184326172,
"objective/kl": 29.722782135009766,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 9.869793891906738,
"objective/scores": 9.869793891906738,
"policy/approxkl_avg": 0.1928042322397232,
"policy/clipfrac_avg": 0.421875,
"policy/entropy_avg": 0.3933394253253937,
"step": 30,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 1600,
"val/ratio": 1.0000554323196411,
"val/ratio_var": 2.9572597668447997e-06
},
{
"episode": 16384,
"epoch": 0.07018505825908156,
"eps": 6,
"loss/policy_avg": 0.19288870692253113,
"lr": 2.63671875e-06,
"objective/entropy": -17.214672088623047,
"objective/kl": 30.913219451904297,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 9.56223201751709,
"objective/scores": 9.56223201751709,
"policy/approxkl_avg": 0.1885090172290802,
"policy/clipfrac_avg": 0.345703125,
"policy/entropy_avg": 0.3990550935268402,
"step": 31,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 1621,
"val/ratio": 0.9998235106468201,
"val/ratio_var": 2.50669563683914e-06
},
{
"episode": 16896,
"epoch": 0.07237834132967787,
"eps": 6,
"loss/policy_avg": 0.23322975635528564,
"lr": 2.6250000000000003e-06,
"objective/entropy": -18.789386749267578,
"objective/kl": 30.845539093017578,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 9.760330200195312,
"objective/scores": 9.760330200195312,
"policy/approxkl_avg": 0.17573946714401245,
"policy/clipfrac_avg": 0.30859375,
"policy/entropy_avg": 0.38172662258148193,
"step": 32,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 1705,
"val/ratio": 1.0002737045288086,
"val/ratio_var": 2.0741492789966287e-06
},
{
"episode": 17408,
"epoch": 0.07457162440027416,
"eps": 6,
"loss/policy_avg": 0.21965008974075317,
"lr": 2.61328125e-06,
"objective/entropy": -14.802291870117188,
"objective/kl": 32.42792510986328,
"objective/non_score_reward": 0.0,
"objective/rlhf_reward": 9.586631774902344,
"objective/scores": 9.586631774902344,
"policy/approxkl_avg": 0.22661904990673065,
"policy/clipfrac_avg": 0.36328125,
"policy/entropy_avg": 0.40180861949920654,
"step": 33,
"val/clipfrac_avg": 0.0,
"val/num_eos_tokens": 1797,
"val/ratio": 1.0000510215759277,
"val/ratio_var": 2.326285539311357e-06
}
],
"logging_steps": 1,
"max_steps": 128,
"num_input_tokens_seen": 0,
"num_train_epochs": 1.122960932145305,
"save_steps": 2,
"stateful_callbacks": {
"TrainerControl": {
"args": {
"should_epoch_stop": false,
"should_evaluate": false,
"should_log": true,
"should_save": true,
"should_training_stop": false
},
"attributes": {}
}
},
"total_flos": 0,
"train_batch_size": null,
"trial_name": null,
"trial_params": null
}