diff --git "a/checkpoint-200/trainer_state.json" "b/checkpoint-200/trainer_state.json" new file mode 100644--- /dev/null +++ "b/checkpoint-200/trainer_state.json" @@ -0,0 +1,3634 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "episode": 3200, + "epoch": 0.019172927825909816, + "eval_steps": 500, + "global_step": 200, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "episode": 16, + "epoch": 9.586463912954908e-05, + "loss/policy_avg": 0.0339290015399456, + "lr": 1e-05, + "objective/entropy": 78.48619842529297, + "objective/kl": 5.6675214767456055, + "objective/non_score_reward": -0.2833760380744934, + "objective/rlhf_reward": 3.2664958328008655, + "objective/scores": 1.1, + "policy/approxkl_avg": 56.270538330078125, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.5703125, + "step": 0, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 2.000828266143799 + }, + { + "episode": 32, + "epoch": 0.00019172927825909816, + "loss/policy_avg": 0.032509539276361465, + "lr": 9.999360940695298e-06, + "objective/entropy": 39.34157943725586, + "objective/kl": 8.134885787963867, + "objective/non_score_reward": -0.40674424171447754, + "objective/rlhf_reward": -1.6269769463688135, + "objective/scores": 0.0, + "policy/approxkl_avg": 125.53129577636719, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.46875, + "step": 1, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9994192123413086 + }, + { + "episode": 48, + "epoch": 0.00028759391738864725, + "loss/policy_avg": 0.2574540972709656, + "lr": 9.998721881390595e-06, + "objective/entropy": 35.90438461303711, + "objective/kl": 10.056818008422852, + "objective/non_score_reward": -0.5028409957885742, + "objective/rlhf_reward": -4.011363983154297, + "objective/scores": -0.5, + "policy/approxkl_avg": 197.81790161132812, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.73828125, + "step": 2, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.99604332447052 + }, + { + "episode": 64, + "epoch": 0.0003834585565181963, + "loss/policy_avg": 0.1315518617630005, + "lr": 9.99808282208589e-06, + "objective/entropy": 163.52642822265625, + "objective/kl": 12.497467041015625, + "objective/non_score_reward": -0.6248733997344971, + "objective/rlhf_reward": -2.499493680894375, + "objective/scores": 0.0, + "policy/approxkl_avg": 280.7725830078125, + "policy/clipfrac_avg": 1.5, + "policy/entropy_avg": 0.716796875, + "step": 3, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.999171257019043 + }, + { + "episode": 80, + "epoch": 0.0004793231956477454, + "loss/policy_avg": 0.024046147242188454, + "lr": 9.997443762781187e-06, + "objective/entropy": 118.5094223022461, + "objective/kl": 5.982309818267822, + "objective/non_score_reward": -0.29911553859710693, + "objective/rlhf_reward": -1.196462158113718, + "objective/scores": 0.0, + "policy/approxkl_avg": 52.543487548828125, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.552734375, + "step": 4, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 2.001847743988037 + }, + { + "episode": 96, + "epoch": 0.0005751878347772945, + "loss/policy_avg": 0.10632362961769104, + "lr": 9.996804703476484e-06, + "objective/entropy": 152.1885986328125, + "objective/kl": 7.815367698669434, + "objective/non_score_reward": -0.3907684087753296, + "objective/rlhf_reward": 0.0987858943933384, + "objective/scores": 0.41546487678572874, + "policy/approxkl_avg": 88.51527404785156, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.5859375, + "step": 5, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9992973804473877 + }, + { + "episode": 112, + "epoch": 0.0006710524739068436, + "loss/policy_avg": 0.08422186970710754, + "lr": 9.99616564417178e-06, + "objective/entropy": 75.0154037475586, + "objective/kl": 17.52770233154297, + "objective/non_score_reward": -0.8763852119445801, + "objective/rlhf_reward": -3.5055407360196114, + "objective/scores": 0.0, + "policy/approxkl_avg": 443.602294921875, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.505859375, + "step": 6, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9972598552703857 + }, + { + "episode": 128, + "epoch": 0.0007669171130363926, + "loss/policy_avg": 0.007405903190374374, + "lr": 9.995526584867077e-06, + "objective/entropy": 51.67172622680664, + "objective/kl": 11.563663482666016, + "objective/non_score_reward": -0.5781831741333008, + "objective/rlhf_reward": -4.312732696533203, + "objective/scores": -0.5, + "policy/approxkl_avg": 126.90079498291016, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.546875, + "step": 7, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 2.0016322135925293 + }, + { + "episode": 144, + "epoch": 0.0008627817521659417, + "loss/policy_avg": 0.13771404325962067, + "lr": 9.994887525562374e-06, + "objective/entropy": 240.35464477539062, + "objective/kl": 18.096904754638672, + "objective/non_score_reward": -0.9048453569412231, + "objective/rlhf_reward": -1.6719702733325317, + "objective/scores": 0.4868528072345416, + "policy/approxkl_avg": 460.8926696777344, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.90625, + "step": 8, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9999654293060303 + }, + { + "episode": 160, + "epoch": 0.0009586463912954908, + "loss/policy_avg": 0.41069674491882324, + "lr": 9.99424846625767e-06, + "objective/entropy": 224.78262329101562, + "objective/kl": 11.231921195983887, + "objective/non_score_reward": -0.5615960955619812, + "objective/rlhf_reward": -0.8225522383051791, + "objective/scores": 0.35595802480981553, + "policy/approxkl_avg": 167.4181671142578, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.7578125, + "step": 9, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9952213764190674 + }, + { + "episode": 176, + "epoch": 0.0010545110304250398, + "loss/policy_avg": 0.2340843677520752, + "lr": 9.993609406952966e-06, + "objective/entropy": 77.48204040527344, + "objective/kl": 13.726895332336426, + "objective/non_score_reward": -0.6863448619842529, + "objective/rlhf_reward": -0.34537934362888345, + "objective/scores": 0.6, + "policy/approxkl_avg": 270.8516845703125, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.876953125, + "step": 10, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 6, + "val/ratio": 1.9955778121948242 + }, + { + "episode": 192, + "epoch": 0.001150375669554589, + "loss/policy_avg": 0.1845349222421646, + "lr": 9.992970347648263e-06, + "objective/entropy": -45.138362884521484, + "objective/kl": 14.76271915435791, + "objective/non_score_reward": -0.7381359338760376, + "objective/rlhf_reward": -1.2192103425661722, + "objective/scores": 0.43333333333333335, + "policy/approxkl_avg": 207.85874938964844, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.703125, + "step": 11, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 2.000040292739868 + }, + { + "episode": 208, + "epoch": 0.001246240308684138, + "loss/policy_avg": 0.6059431433677673, + "lr": 9.992331288343558e-06, + "objective/entropy": 40.190372467041016, + "objective/kl": 19.720378875732422, + "objective/non_score_reward": -0.9860190749168396, + "objective/rlhf_reward": -1.821370030120883, + "objective/scores": 0.5306765580733931, + "policy/approxkl_avg": 268.6492919921875, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.654296875, + "step": 12, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 1.9966726303100586 + }, + { + "episode": 224, + "epoch": 0.0013421049478136871, + "loss/policy_avg": -0.0064672790467739105, + "lr": 9.991692229038855e-06, + "objective/entropy": 108.48332214355469, + "objective/kl": 5.689068794250488, + "objective/non_score_reward": -0.28445348143577576, + "objective/rlhf_reward": -1.1378139406442642, + "objective/scores": 0.0, + "policy/approxkl_avg": 15.834894180297852, + "policy/clipfrac_avg": 1.75, + "policy/entropy_avg": 0.4716796875, + "step": 13, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 2.0026183128356934 + }, + { + "episode": 240, + "epoch": 0.001437969586943236, + "loss/policy_avg": 0.6670212745666504, + "lr": 9.991053169734152e-06, + "objective/entropy": 0.18174362182617188, + "objective/kl": 12.982845306396484, + "objective/non_score_reward": -0.6491422653198242, + "objective/rlhf_reward": -2.596569076180458, + "objective/scores": 0.0, + "policy/approxkl_avg": 330.118896484375, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.759765625, + "step": 14, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9997687339782715 + }, + { + "episode": 256, + "epoch": 0.0015338342260727853, + "loss/policy_avg": 0.21263472735881805, + "lr": 9.990414110429449e-06, + "objective/entropy": 249.88232421875, + "objective/kl": 9.040252685546875, + "objective/non_score_reward": -0.45201271772384644, + "objective/rlhf_reward": -1.808050960302353, + "objective/scores": 0.0, + "policy/approxkl_avg": 102.57914733886719, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.744140625, + "step": 15, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 0, + "val/ratio": 2.000203847885132 + }, + { + "episode": 272, + "epoch": 0.0016296988652023342, + "loss/policy_avg": 0.01660698838531971, + "lr": 9.989775051124744e-06, + "objective/entropy": 106.64703369140625, + "objective/kl": 11.038640022277832, + "objective/non_score_reward": -0.5519319772720337, + "objective/rlhf_reward": -2.2077280431985855, + "objective/scores": 0.0, + "policy/approxkl_avg": 164.40890502929688, + "policy/clipfrac_avg": 1.5, + "policy/entropy_avg": 0.654296875, + "step": 16, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 2.000194549560547 + }, + { + "episode": 288, + "epoch": 0.0017255635043318834, + "loss/policy_avg": 0.17964985966682434, + "lr": 9.989135991820041e-06, + "objective/entropy": 29.59412956237793, + "objective/kl": 11.429637908935547, + "objective/non_score_reward": -0.5714819431304932, + "objective/rlhf_reward": -2.2859277576208115, + "objective/scores": 0.0, + "policy/approxkl_avg": 113.22151184082031, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.642578125, + "step": 17, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9986356496810913 + }, + { + "episode": 304, + "epoch": 0.0018214281434614326, + "loss/policy_avg": 0.1845007836818695, + "lr": 9.988496932515338e-06, + "objective/entropy": -2.3180160522460938, + "objective/kl": 15.66268539428711, + "objective/non_score_reward": -0.7831343412399292, + "objective/rlhf_reward": -1.773287498687191, + "objective/scores": 0.33981246656813147, + "policy/approxkl_avg": 162.00823974609375, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.8046875, + "step": 18, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 2.0002946853637695 + }, + { + "episode": 320, + "epoch": 0.0019172927825909815, + "loss/policy_avg": 0.14623276889324188, + "lr": 9.987857873210635e-06, + "objective/entropy": 26.79373550415039, + "objective/kl": 16.199674606323242, + "objective/non_score_reward": -0.8099837303161621, + "objective/rlhf_reward": -3.2399348318576813, + "objective/scores": 0.0, + "policy/approxkl_avg": 345.71685791015625, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.7109375, + "step": 19, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9992010593414307 + }, + { + "episode": 336, + "epoch": 0.0020131574217205307, + "loss/policy_avg": 0.061316944658756256, + "lr": 9.987218813905932e-06, + "objective/entropy": 30.27604866027832, + "objective/kl": 13.349930763244629, + "objective/non_score_reward": -0.6674965620040894, + "objective/rlhf_reward": -4.669986248016357, + "objective/scores": -0.5, + "policy/approxkl_avg": 182.816650390625, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.56640625, + "step": 20, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.996535301208496 + }, + { + "episode": 352, + "epoch": 0.0021090220608500796, + "loss/policy_avg": -0.08272892981767654, + "lr": 9.986579754601228e-06, + "objective/entropy": 198.63003540039062, + "objective/kl": 11.5382719039917, + "objective/non_score_reward": -0.5769136548042297, + "objective/rlhf_reward": -0.9076545149087907, + "objective/scores": 0.35, + "policy/approxkl_avg": 159.97686767578125, + "policy/clipfrac_avg": 1.5, + "policy/entropy_avg": 0.974609375, + "step": 21, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.998477578163147 + }, + { + "episode": 368, + "epoch": 0.0022048866999796286, + "loss/policy_avg": 0.23813551664352417, + "lr": 9.985940695296524e-06, + "objective/entropy": 181.51829528808594, + "objective/kl": 13.08276653289795, + "objective/non_score_reward": -0.6541383266448975, + "objective/rlhf_reward": -4.61655330657959, + "objective/scores": -0.5, + "policy/approxkl_avg": 108.47281646728516, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.6484375, + "step": 22, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9991295337677002 + }, + { + "episode": 384, + "epoch": 0.002300751339109178, + "loss/policy_avg": 0.36420387029647827, + "lr": 9.98530163599182e-06, + "objective/entropy": 257.93609619140625, + "objective/kl": 14.696407318115234, + "objective/non_score_reward": -0.7348203063011169, + "objective/rlhf_reward": -1.3351611531415755, + "objective/scores": 0.40102999566398123, + "policy/approxkl_avg": 150.4597625732422, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.849609375, + "step": 23, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9980016946792603 + }, + { + "episode": 400, + "epoch": 0.002396615978238727, + "loss/policy_avg": 0.20780539512634277, + "lr": 9.984662576687117e-06, + "objective/entropy": -139.27951049804688, + "objective/kl": 15.462644577026367, + "objective/non_score_reward": -0.77313232421875, + "objective/rlhf_reward": -5.092529296875, + "objective/scores": -0.5, + "policy/approxkl_avg": 237.78317260742188, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.6484375, + "step": 24, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 7, + "val/ratio": 1.999894618988037 + }, + { + "episode": 416, + "epoch": 0.002492480617368276, + "loss/policy_avg": 0.2547074556350708, + "lr": 9.984023517382414e-06, + "objective/entropy": 103.24639892578125, + "objective/kl": 17.307334899902344, + "objective/non_score_reward": -0.8653668165206909, + "objective/rlhf_reward": -5.461467266082764, + "objective/scores": -0.5, + "policy/approxkl_avg": 167.2418212890625, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.65625, + "step": 25, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 2.000932216644287 + }, + { + "episode": 432, + "epoch": 0.002588345256497825, + "loss/policy_avg": 0.3379603624343872, + "lr": 9.983384458077711e-06, + "objective/entropy": 120.86388397216797, + "objective/kl": 14.275808334350586, + "objective/non_score_reward": -0.7137903571128845, + "objective/rlhf_reward": -2.8551614582538605, + "objective/scores": 0.0, + "policy/approxkl_avg": 99.34181213378906, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.73828125, + "step": 26, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9994611740112305 + }, + { + "episode": 448, + "epoch": 0.0026842098956273742, + "loss/policy_avg": 0.1328231394290924, + "lr": 9.982745398773006e-06, + "objective/entropy": 154.86619567871094, + "objective/kl": 14.35202693939209, + "objective/non_score_reward": -0.7176014184951782, + "objective/rlhf_reward": -4.870405673980713, + "objective/scores": -0.5, + "policy/approxkl_avg": 91.33482360839844, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.724609375, + "step": 27, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9957685470581055 + }, + { + "episode": 464, + "epoch": 0.002780074534756923, + "loss/policy_avg": 0.2314174473285675, + "lr": 9.982106339468303e-06, + "objective/entropy": 60.19127655029297, + "objective/kl": 12.188166618347168, + "objective/non_score_reward": -0.6094082593917847, + "objective/rlhf_reward": -2.4376331865787506, + "objective/scores": 0.0, + "policy/approxkl_avg": 205.1094970703125, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.810546875, + "step": 28, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.994727611541748 + }, + { + "episode": 480, + "epoch": 0.002875939173886472, + "loss/policy_avg": 0.1261996328830719, + "lr": 9.9814672801636e-06, + "objective/entropy": 67.08200073242188, + "objective/kl": 16.607372283935547, + "objective/non_score_reward": -0.8303685784339905, + "objective/rlhf_reward": -3.3214742839336395, + "objective/scores": 0.0, + "policy/approxkl_avg": 226.6929168701172, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.53125, + "step": 29, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9976527690887451 + }, + { + "episode": 496, + "epoch": 0.0029718038130160216, + "loss/policy_avg": 0.35239556431770325, + "lr": 9.980828220858897e-06, + "objective/entropy": 204.22116088867188, + "objective/kl": 14.144770622253418, + "objective/non_score_reward": -0.7072385549545288, + "objective/rlhf_reward": 1.5710457801818851, + "objective/scores": 1.1, + "policy/approxkl_avg": 124.6588363647461, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.7578125, + "step": 30, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9977011680603027 + }, + { + "episode": 512, + "epoch": 0.0030676684521455705, + "loss/policy_avg": 0.26766547560691833, + "lr": 9.980189161554194e-06, + "objective/entropy": 107.69725036621094, + "objective/kl": 12.877479553222656, + "objective/non_score_reward": -0.643873929977417, + "objective/rlhf_reward": -2.5754958018660545, + "objective/scores": 0.0, + "policy/approxkl_avg": 186.40504455566406, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.701171875, + "step": 31, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9972684383392334 + }, + { + "episode": 528, + "epoch": 0.0031635330912751195, + "loss/policy_avg": 0.017455143854022026, + "lr": 9.97955010224949e-06, + "objective/entropy": 69.81261444091797, + "objective/kl": 15.58060073852539, + "objective/non_score_reward": -0.7790300250053406, + "objective/rlhf_reward": -0.19240116024133824, + "objective/scores": 0.7309297535714575, + "policy/approxkl_avg": 326.45733642578125, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.671875, + "step": 32, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9994511604309082 + }, + { + "episode": 544, + "epoch": 0.0032593977304046684, + "loss/policy_avg": 0.1626880019903183, + "lr": 9.978911042944786e-06, + "objective/entropy": 49.244285583496094, + "objective/kl": 11.25068473815918, + "objective/non_score_reward": -0.5625342130661011, + "objective/rlhf_reward": -0.871534817901951, + "objective/scores": 0.34465054211822604, + "policy/approxkl_avg": 50.66204071044922, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.7109375, + "step": 33, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9966607093811035 + }, + { + "episode": 560, + "epoch": 0.003355262369534218, + "loss/policy_avg": 0.18031546473503113, + "lr": 9.978271983640083e-06, + "objective/entropy": 147.95474243164062, + "objective/kl": 15.950370788574219, + "objective/non_score_reward": -0.7975186109542847, + "objective/rlhf_reward": -3.190074533224106, + "objective/scores": 0.0, + "policy/approxkl_avg": 201.51080322265625, + "policy/clipfrac_avg": 1.5, + "policy/entropy_avg": 0.515625, + "step": 34, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 0, + "val/ratio": 2.0006275177001953 + }, + { + "episode": 576, + "epoch": 0.0034511270086637668, + "loss/policy_avg": -0.05689749866724014, + "lr": 9.977632924335378e-06, + "objective/entropy": 4.240108489990234, + "objective/kl": 13.38272762298584, + "objective/non_score_reward": -0.6691364049911499, + "objective/rlhf_reward": -4.6765456199646, + "objective/scores": -0.5, + "policy/approxkl_avg": 247.30409240722656, + "policy/clipfrac_avg": 1.5, + "policy/entropy_avg": 0.796875, + "step": 35, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9973469972610474 + }, + { + "episode": 592, + "epoch": 0.0035469916477933157, + "loss/policy_avg": 0.16461774706840515, + "lr": 9.976993865030675e-06, + "objective/entropy": 123.00151824951172, + "objective/kl": 11.21810531616211, + "objective/non_score_reward": -0.5609052181243896, + "objective/rlhf_reward": -4.243620872497559, + "objective/scores": -0.5, + "policy/approxkl_avg": 64.79019927978516, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.712890625, + "step": 36, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 2.0026960372924805 + }, + { + "episode": 608, + "epoch": 0.003642856286922865, + "loss/policy_avg": 0.15750062465667725, + "lr": 9.976354805725972e-06, + "objective/entropy": 30.60162925720215, + "objective/kl": 17.013538360595703, + "objective/non_score_reward": -0.850676953792572, + "objective/rlhf_reward": -3.4027078449726105, + "objective/scores": 0.0, + "policy/approxkl_avg": 234.95870971679688, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.7421875, + "step": 37, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 1.998824119567871 + }, + { + "episode": 624, + "epoch": 0.003738720926052414, + "loss/policy_avg": 0.03158241882920265, + "lr": 9.975715746421269e-06, + "objective/entropy": 115.71566772460938, + "objective/kl": 15.653677940368652, + "objective/non_score_reward": -0.7826838493347168, + "objective/rlhf_reward": -3.1307354420423508, + "objective/scores": 0.0, + "policy/approxkl_avg": 233.744873046875, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.765625, + "step": 38, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.997577428817749 + }, + { + "episode": 640, + "epoch": 0.003834585565181963, + "loss/policy_avg": -0.031586866825819016, + "lr": 9.975076687116566e-06, + "objective/entropy": 70.69473266601562, + "objective/kl": 13.526529312133789, + "objective/non_score_reward": -0.6763265132904053, + "objective/rlhf_reward": 1.6946939915418628, + "objective/scores": 1.1, + "policy/approxkl_avg": 101.47872924804688, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.677734375, + "step": 39, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 2.0043745040893555 + }, + { + "episode": 656, + "epoch": 0.003930450204311512, + "loss/policy_avg": 0.12032957375049591, + "lr": 9.97443762781186e-06, + "objective/entropy": 172.61669921875, + "objective/kl": 16.604652404785156, + "objective/non_score_reward": -0.8302326798439026, + "objective/rlhf_reward": -5.320930480957031, + "objective/scores": -0.5, + "policy/approxkl_avg": 220.98178100585938, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.642578125, + "step": 40, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9957454204559326 + }, + { + "episode": 672, + "epoch": 0.004026314843441061, + "loss/policy_avg": 0.0876859575510025, + "lr": 9.973798568507158e-06, + "objective/entropy": 12.816411972045898, + "objective/kl": 13.908916473388672, + "objective/non_score_reward": -0.6954457759857178, + "objective/rlhf_reward": -4.781783103942871, + "objective/scores": -0.5, + "policy/approxkl_avg": 74.01300048828125, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.640625, + "step": 41, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 2.0018563270568848 + }, + { + "episode": 688, + "epoch": 0.00412217948257061, + "loss/policy_avg": 0.1008758619427681, + "lr": 9.973159509202454e-06, + "objective/entropy": 257.1292724609375, + "objective/kl": 11.528783798217773, + "objective/non_score_reward": -0.5764391422271729, + "objective/rlhf_reward": -2.3057566583156586, + "objective/scores": 0.0, + "policy/approxkl_avg": 84.59580993652344, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.75390625, + "step": 42, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 0, + "val/ratio": 1.9929132461547852 + }, + { + "episode": 704, + "epoch": 0.004218044121700159, + "loss/policy_avg": 0.44639891386032104, + "lr": 9.972520449897751e-06, + "objective/entropy": 62.445350646972656, + "objective/kl": 13.397602081298828, + "objective/non_score_reward": -0.6698801517486572, + "objective/rlhf_reward": -4.679520606994629, + "objective/scores": -0.5, + "policy/approxkl_avg": 185.67079162597656, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.533203125, + "step": 43, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9992148876190186 + }, + { + "episode": 720, + "epoch": 0.004313908760829708, + "loss/policy_avg": 0.09568839520215988, + "lr": 9.971881390593048e-06, + "objective/entropy": 129.84619140625, + "objective/kl": 14.350381851196289, + "objective/non_score_reward": -0.7175191640853882, + "objective/rlhf_reward": -0.9226653081940968, + "objective/scores": 0.4868528072345416, + "policy/approxkl_avg": 253.10037231445312, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.7421875, + "step": 44, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9957365989685059 + }, + { + "episode": 736, + "epoch": 0.004409773399959257, + "loss/policy_avg": 0.22084593772888184, + "lr": 9.971242331288345e-06, + "objective/entropy": -42.89992904663086, + "objective/kl": 15.893115997314453, + "objective/non_score_reward": -0.7946557998657227, + "objective/rlhf_reward": -1.44528977672259, + "objective/scores": 0.43333333333333335, + "policy/approxkl_avg": 121.7098388671875, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.751953125, + "step": 45, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 2.000582218170166 + }, + { + "episode": 752, + "epoch": 0.004505638039088807, + "loss/policy_avg": 0.22875869274139404, + "lr": 9.97060327198364e-06, + "objective/entropy": 9.025165557861328, + "objective/kl": 22.01996612548828, + "objective/non_score_reward": -1.1009982824325562, + "objective/rlhf_reward": -1.48027405583975, + "objective/scores": 0.7309297535714575, + "policy/approxkl_avg": 321.845703125, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.5859375, + "step": 46, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9953069686889648 + }, + { + "episode": 768, + "epoch": 0.004601502678218356, + "loss/policy_avg": 0.07661572843790054, + "lr": 9.969964212678937e-06, + "objective/entropy": 10.382087707519531, + "objective/kl": 13.358439445495605, + "objective/non_score_reward": -0.6679220199584961, + "objective/rlhf_reward": 1.7283119499683384, + "objective/scores": 1.1, + "policy/approxkl_avg": 29.50304412841797, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.638671875, + "step": 47, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 1.999579668045044 + }, + { + "episode": 784, + "epoch": 0.004697367317347905, + "loss/policy_avg": 0.19636262953281403, + "lr": 9.969325153374234e-06, + "objective/entropy": 0.6832618713378906, + "objective/kl": 8.836541175842285, + "objective/non_score_reward": -0.4418269991874695, + "objective/rlhf_reward": -1.767308071255684, + "objective/scores": 0.0, + "policy/approxkl_avg": 94.3209228515625, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.712890625, + "step": 48, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.998420000076294 + }, + { + "episode": 800, + "epoch": 0.004793231956477454, + "loss/policy_avg": -0.02652953751385212, + "lr": 9.968686094069531e-06, + "objective/entropy": 125.6042709350586, + "objective/kl": 15.016199111938477, + "objective/non_score_reward": -0.7508100271224976, + "objective/rlhf_reward": -5.00324010848999, + "objective/scores": -0.5, + "policy/approxkl_avg": 207.12213134765625, + "policy/clipfrac_avg": 1.5, + "policy/entropy_avg": 0.7421875, + "step": 49, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 2.002742052078247 + }, + { + "episode": 816, + "epoch": 0.004889096595607003, + "loss/policy_avg": 0.10039197653532028, + "lr": 9.968047034764828e-06, + "objective/entropy": -24.506595611572266, + "objective/kl": 18.726213455200195, + "objective/non_score_reward": -0.9363107085227966, + "objective/rlhf_reward": -5.745243072509766, + "objective/scores": -0.5, + "policy/approxkl_avg": 300.677490234375, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.6953125, + "step": 50, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 5, + "val/ratio": 1.9997317790985107 + }, + { + "episode": 832, + "epoch": 0.004984961234736552, + "loss/policy_avg": 0.18666991591453552, + "lr": 9.967407975460123e-06, + "objective/entropy": -47.62429428100586, + "objective/kl": 13.258740425109863, + "objective/non_score_reward": -0.6629370450973511, + "objective/rlhf_reward": -1.0954889049201753, + "objective/scores": 0.38906482631788786, + "policy/approxkl_avg": 110.42059326171875, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.685546875, + "step": 51, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 5, + "val/ratio": 1.9965816736221313 + }, + { + "episode": 848, + "epoch": 0.005080825873866101, + "loss/policy_avg": 0.195734903216362, + "lr": 9.96676891615542e-06, + "objective/entropy": 57.396114349365234, + "objective/kl": 15.980720520019531, + "objective/non_score_reward": -0.7990360856056213, + "objective/rlhf_reward": -5.196144104003906, + "objective/scores": -0.5, + "policy/approxkl_avg": 125.00595092773438, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.775390625, + "step": 52, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 1.995786428451538 + }, + { + "episode": 864, + "epoch": 0.00517669051299565, + "loss/policy_avg": 0.24296848475933075, + "lr": 9.966129856850717e-06, + "objective/entropy": 99.57502746582031, + "objective/kl": 17.62392807006836, + "objective/non_score_reward": -0.8811964988708496, + "objective/rlhf_reward": -2.0090143916928134, + "objective/scores": 0.37894294565112985, + "policy/approxkl_avg": 99.32807922363281, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.662109375, + "step": 53, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 2.0016443729400635 + }, + { + "episode": 880, + "epoch": 0.0052725551521251995, + "loss/policy_avg": 0.5130509734153748, + "lr": 9.965490797546014e-06, + "objective/entropy": 34.0892219543457, + "objective/kl": 14.999124526977539, + "objective/non_score_reward": -0.7499562501907349, + "objective/rlhf_reward": -1.4840530840479695, + "objective/scores": 0.37894294565112985, + "policy/approxkl_avg": 130.593017578125, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.75390625, + "step": 54, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 2.000711679458618 + }, + { + "episode": 896, + "epoch": 0.0053684197912547485, + "loss/policy_avg": 0.3232521116733551, + "lr": 9.96485173824131e-06, + "objective/entropy": 69.26298522949219, + "objective/kl": 21.724315643310547, + "objective/non_score_reward": -1.086215853691101, + "objective/rlhf_reward": -2.788603871074274, + "objective/scores": 0.38906482631788786, + "policy/approxkl_avg": 234.98104858398438, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.68359375, + "step": 55, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9952917098999023 + }, + { + "episode": 912, + "epoch": 0.0054642844303842975, + "loss/policy_avg": 0.10791392624378204, + "lr": 9.964212678936606e-06, + "objective/entropy": 32.22584533691406, + "objective/kl": 15.846414566040039, + "objective/non_score_reward": -0.7923207879066467, + "objective/rlhf_reward": 1.230716893076897, + "objective/scores": 1.1, + "policy/approxkl_avg": 256.9724426269531, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.4736328125, + "step": 56, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9984806776046753 + }, + { + "episode": 928, + "epoch": 0.005560149069513846, + "loss/policy_avg": 0.37206730246543884, + "lr": 9.963573619631903e-06, + "objective/entropy": -6.195688247680664, + "objective/kl": 12.801559448242188, + "objective/non_score_reward": -0.6400780081748962, + "objective/rlhf_reward": -2.5603120028972626, + "objective/scores": 0.0, + "policy/approxkl_avg": 40.83631896972656, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.744140625, + "step": 57, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 2.0007452964782715 + }, + { + "episode": 944, + "epoch": 0.005656013708643395, + "loss/policy_avg": 0.05091024935245514, + "lr": 9.9629345603272e-06, + "objective/entropy": -18.476280212402344, + "objective/kl": 18.95052719116211, + "objective/non_score_reward": -0.9475262761116028, + "objective/rlhf_reward": -3.790105164051056, + "objective/scores": 0.0, + "policy/approxkl_avg": 144.4001007080078, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.580078125, + "step": 58, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9988869428634644 + }, + { + "episode": 960, + "epoch": 0.005751878347772944, + "loss/policy_avg": -0.008312445133924484, + "lr": 9.962295501022495e-06, + "objective/entropy": 217.08169555664062, + "objective/kl": 14.908738136291504, + "objective/non_score_reward": -0.7454369068145752, + "objective/rlhf_reward": -2.9817477762699127, + "objective/scores": 0.0, + "policy/approxkl_avg": 93.395751953125, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.77734375, + "step": 59, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9973504543304443 + }, + { + "episode": 976, + "epoch": 0.005847742986902493, + "loss/policy_avg": 0.03407389298081398, + "lr": 9.961656441717792e-06, + "objective/entropy": 140.58189392089844, + "objective/kl": 20.377920150756836, + "objective/non_score_reward": -1.0188961029052734, + "objective/rlhf_reward": -6.075584411621094, + "objective/scores": -0.5, + "policy/approxkl_avg": 66.34793090820312, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.673828125, + "step": 60, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9996695518493652 + }, + { + "episode": 992, + "epoch": 0.005943607626032043, + "loss/policy_avg": 0.08645053207874298, + "lr": 9.961017382413088e-06, + "objective/entropy": 19.108230590820312, + "objective/kl": 13.644828796386719, + "objective/non_score_reward": -0.6822414994239807, + "objective/rlhf_reward": -2.7289658784866333, + "objective/scores": 0.0, + "policy/approxkl_avg": 35.71690368652344, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.564453125, + "step": 61, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9962527751922607 + }, + { + "episode": 1008, + "epoch": 0.006039472265161592, + "loss/policy_avg": -0.17965860664844513, + "lr": 9.960378323108385e-06, + "objective/entropy": 168.96075439453125, + "objective/kl": 11.691057205200195, + "objective/non_score_reward": -0.5845528841018677, + "objective/rlhf_reward": -4.338212013244629, + "objective/scores": -0.5, + "policy/approxkl_avg": 65.8020248413086, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.775390625, + "step": 62, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 2.0171313285827637 + }, + { + "episode": 1024, + "epoch": 0.006135336904291141, + "loss/policy_avg": 0.061459362506866455, + "lr": 9.959739263803682e-06, + "objective/entropy": 117.6607437133789, + "objective/kl": 15.35727310180664, + "objective/non_score_reward": -0.7678636312484741, + "objective/rlhf_reward": -3.071454644203186, + "objective/scores": 0.0, + "policy/approxkl_avg": 187.7376708984375, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.9375, + "step": 63, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9990172386169434 + }, + { + "episode": 1040, + "epoch": 0.00623120154342069, + "loss/policy_avg": 0.07200516015291214, + "lr": 9.959100204498979e-06, + "objective/entropy": 20.476089477539062, + "objective/kl": 13.475000381469727, + "objective/non_score_reward": -0.6737500429153442, + "objective/rlhf_reward": 1.7049996197223667, + "objective/scores": 1.1, + "policy/approxkl_avg": 68.92333984375, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.6796875, + "step": 64, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 6, + "val/ratio": 1.9985809326171875 + }, + { + "episode": 1056, + "epoch": 0.006327066182550239, + "loss/policy_avg": 0.08365275710821152, + "lr": 9.958461145194274e-06, + "objective/entropy": -127.62371826171875, + "objective/kl": 22.050678253173828, + "objective/non_score_reward": -1.1025339365005493, + "objective/rlhf_reward": -4.410135626792908, + "objective/scores": 0.0, + "policy/approxkl_avg": 301.7841491699219, + "policy/clipfrac_avg": 0.5, + "policy/entropy_avg": 0.720703125, + "step": 65, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 5, + "val/ratio": 1.9990687370300293 + }, + { + "episode": 1072, + "epoch": 0.006422930821679788, + "loss/policy_avg": 0.040758199989795685, + "lr": 9.957822085889571e-06, + "objective/entropy": 70.77458190917969, + "objective/kl": 18.2130069732666, + "objective/non_score_reward": -0.9106502532958984, + "objective/rlhf_reward": -1.9807416550522907, + "objective/scores": 0.41546487678572874, + "policy/approxkl_avg": 190.14797973632812, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.439453125, + "step": 66, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9984623193740845 + }, + { + "episode": 1088, + "epoch": 0.006518795460809337, + "loss/policy_avg": 0.05295582860708237, + "lr": 9.957183026584868e-06, + "objective/entropy": 97.35667419433594, + "objective/kl": 24.64842987060547, + "objective/non_score_reward": -1.2324215173721313, + "objective/rlhf_reward": -4.929685860872269, + "objective/scores": 0.0, + "policy/approxkl_avg": 394.2406921386719, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.67578125, + "step": 67, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 2.0007405281066895 + }, + { + "episode": 1104, + "epoch": 0.006614660099938887, + "loss/policy_avg": 0.14266067743301392, + "lr": 9.956543967280165e-06, + "objective/entropy": 85.57185363769531, + "objective/kl": 14.76464557647705, + "objective/non_score_reward": -0.7382322549819946, + "objective/rlhf_reward": -0.029210095049115647, + "objective/scores": 0.7309297535714575, + "policy/approxkl_avg": 171.19406127929688, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.6015625, + "step": 68, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 1.9964725971221924 + }, + { + "episode": 1120, + "epoch": 0.006710524739068436, + "loss/policy_avg": 0.11469551920890808, + "lr": 9.955904907975462e-06, + "objective/entropy": 21.974023818969727, + "objective/kl": 16.630640029907227, + "objective/non_score_reward": -0.8315319418907166, + "objective/rlhf_reward": -3.326127827167511, + "objective/scores": 0.0, + "policy/approxkl_avg": 63.985679626464844, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.72265625, + "step": 69, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 1.999800682067871 + }, + { + "episode": 1136, + "epoch": 0.006806389378197985, + "loss/policy_avg": 0.10287429392337799, + "lr": 9.955265848670757e-06, + "objective/entropy": 43.38239288330078, + "objective/kl": 22.02418327331543, + "objective/non_score_reward": -1.101209282875061, + "objective/rlhf_reward": -4.404837071895599, + "objective/scores": 0.0, + "policy/approxkl_avg": 135.89984130859375, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.69140625, + "step": 70, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9973747730255127 + }, + { + "episode": 1152, + "epoch": 0.0069022540173275335, + "loss/policy_avg": 2.0731570720672607, + "lr": 9.954626789366054e-06, + "objective/entropy": -7.300925254821777, + "objective/kl": 17.635089874267578, + "objective/non_score_reward": -0.8817545175552368, + "objective/rlhf_reward": -3.5270181000232697, + "objective/scores": 0.0, + "policy/approxkl_avg": 183.6417236328125, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.65625, + "step": 71, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 5, + "val/ratio": 1.9976426362991333 + }, + { + "episode": 1168, + "epoch": 0.0069981186564570825, + "loss/policy_avg": 0.24466943740844727, + "lr": 9.95398773006135e-06, + "objective/entropy": -26.054595947265625, + "objective/kl": 15.905699729919434, + "objective/non_score_reward": -0.7952849864959717, + "objective/rlhf_reward": -5.181139945983887, + "objective/scores": -0.5, + "policy/approxkl_avg": 46.05584716796875, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.625, + "step": 72, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9955050945281982 + }, + { + "episode": 1184, + "epoch": 0.0070939832955866314, + "loss/policy_avg": 0.4031391739845276, + "lr": 9.953348670756648e-06, + "objective/entropy": -107.55976867675781, + "objective/kl": 19.68102264404297, + "objective/non_score_reward": -0.9840512275695801, + "objective/rlhf_reward": -3.936204746365547, + "objective/scores": 0.0, + "policy/approxkl_avg": 232.6634521484375, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.751953125, + "step": 73, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 6, + "val/ratio": 2.00066876411438 + }, + { + "episode": 1200, + "epoch": 0.00718984793471618, + "loss/policy_avg": 0.1890699565410614, + "lr": 9.952709611451944e-06, + "objective/entropy": 118.900146484375, + "objective/kl": 21.680133819580078, + "objective/non_score_reward": -1.084006667137146, + "objective/rlhf_reward": -6.336027145385742, + "objective/scores": -0.5, + "policy/approxkl_avg": 266.20465087890625, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.720703125, + "step": 74, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.997557282447815 + }, + { + "episode": 1216, + "epoch": 0.00728571257384573, + "loss/policy_avg": 0.383888304233551, + "lr": 9.952070552147241e-06, + "objective/entropy": 124.33120727539062, + "objective/kl": 21.27002716064453, + "objective/non_score_reward": -1.0635013580322266, + "objective/rlhf_reward": -4.254005193710327, + "objective/scores": 0.0, + "policy/approxkl_avg": 86.44483184814453, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.611328125, + "step": 75, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9962941408157349 + }, + { + "episode": 1232, + "epoch": 0.007381577212975279, + "loss/policy_avg": 0.23960661888122559, + "lr": 9.951431492842536e-06, + "objective/entropy": 40.334468841552734, + "objective/kl": 17.827497482299805, + "objective/non_score_reward": -0.891374945640564, + "objective/rlhf_reward": -3.5654996633529663, + "objective/scores": 0.0, + "policy/approxkl_avg": 94.15713500976562, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.548828125, + "step": 76, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9984164237976074 + }, + { + "episode": 1248, + "epoch": 0.007477441852104828, + "loss/policy_avg": 0.4706483781337738, + "lr": 9.950792433537833e-06, + "objective/entropy": 106.07322692871094, + "objective/kl": 19.12630844116211, + "objective/non_score_reward": -0.9563154578208923, + "objective/rlhf_reward": -3.8252618312835693, + "objective/scores": 0.0, + "policy/approxkl_avg": 185.7378387451172, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.87890625, + "step": 77, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 5, + "val/ratio": 1.9966304302215576 + }, + { + "episode": 1264, + "epoch": 0.007573306491234377, + "loss/policy_avg": 0.0665474385023117, + "lr": 9.950153374233129e-06, + "objective/entropy": 34.984527587890625, + "objective/kl": 23.865880966186523, + "objective/non_score_reward": -1.1932940483093262, + "objective/rlhf_reward": -3.373176074028015, + "objective/scores": 0.35, + "policy/approxkl_avg": 336.36712646484375, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.53515625, + "step": 78, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.997634768486023 + }, + { + "episode": 1280, + "epoch": 0.007669171130363926, + "loss/policy_avg": 0.3842596113681793, + "lr": 9.949514314928425e-06, + "objective/entropy": 229.88047790527344, + "objective/kl": 27.72378921508789, + "objective/non_score_reward": -1.386189579963684, + "objective/rlhf_reward": -1.144758558273315, + "objective/scores": 1.1, + "policy/approxkl_avg": 524.0328369140625, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.720703125, + "step": 79, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 0, + "val/ratio": 1.997296690940857 + }, + { + "episode": 1296, + "epoch": 0.007765035769493475, + "loss/policy_avg": 0.4347228705883026, + "lr": 9.948875255623722e-06, + "objective/entropy": -43.351566314697266, + "objective/kl": 18.37939453125, + "objective/non_score_reward": -0.9189697504043579, + "objective/rlhf_reward": -3.6758789718151093, + "objective/scores": 0.0, + "policy/approxkl_avg": 87.53759002685547, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.521484375, + "step": 80, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9968631267547607 + }, + { + "episode": 1312, + "epoch": 0.007860900408623025, + "loss/policy_avg": 0.5703809261322021, + "lr": 9.94823619631902e-06, + "objective/entropy": 182.94879150390625, + "objective/kl": 24.6871337890625, + "objective/non_score_reward": -1.2343567609786987, + "objective/rlhf_reward": -6.937427043914795, + "objective/scores": -0.5, + "policy/approxkl_avg": 274.44744873046875, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.71875, + "step": 81, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9995486736297607 + }, + { + "episode": 1328, + "epoch": 0.007956765047752574, + "loss/policy_avg": 0.10641711950302124, + "lr": 9.947597137014316e-06, + "objective/entropy": 73.29893493652344, + "objective/kl": 17.603548049926758, + "objective/non_score_reward": -0.88017737865448, + "objective/rlhf_reward": -3.52070951461792, + "objective/scores": 0.0, + "policy/approxkl_avg": 123.0771255493164, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.505859375, + "step": 82, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.998863935470581 + }, + { + "episode": 1344, + "epoch": 0.008052629686882123, + "loss/policy_avg": 0.12928956747055054, + "lr": 9.946958077709611e-06, + "objective/entropy": 108.6548080444336, + "objective/kl": 18.59684944152832, + "objective/non_score_reward": -0.929842472076416, + "objective/rlhf_reward": -2.3407678390420497, + "objective/scores": 0.34465054211822604, + "policy/approxkl_avg": 120.68421936035156, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.57421875, + "step": 83, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.998254656791687 + }, + { + "episode": 1360, + "epoch": 0.008148494326011672, + "loss/policy_avg": 0.14865761995315552, + "lr": 9.946319018404908e-06, + "objective/entropy": 102.67412567138672, + "objective/kl": 23.651020050048828, + "objective/non_score_reward": -1.1825510263442993, + "objective/rlhf_reward": -4.7302040457725525, + "objective/scores": 0.0, + "policy/approxkl_avg": 73.66981506347656, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.4033203125, + "step": 84, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 2.000584602355957 + }, + { + "episode": 1376, + "epoch": 0.00824435896514122, + "loss/policy_avg": 0.07400541007518768, + "lr": 9.945679959100205e-06, + "objective/entropy": 133.18292236328125, + "objective/kl": 11.912694931030273, + "objective/non_score_reward": -0.5956346988677979, + "objective/rlhf_reward": -0.4351277453469593, + "objective/scores": 0.4868528072345416, + "policy/approxkl_avg": 49.34624099731445, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.59375, + "step": 85, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 2.000187397003174 + }, + { + "episode": 1392, + "epoch": 0.00834022360427077, + "loss/policy_avg": 0.13357847929000854, + "lr": 9.945040899795502e-06, + "objective/entropy": 112.34770202636719, + "objective/kl": 20.725894927978516, + "objective/non_score_reward": -1.0362948179244995, + "objective/rlhf_reward": -2.7665772224343836, + "objective/scores": 0.34465054211822604, + "policy/approxkl_avg": 157.26473999023438, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.56640625, + "step": 86, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9963226318359375 + }, + { + "episode": 1408, + "epoch": 0.008436088243400319, + "loss/policy_avg": 0.32753437757492065, + "lr": 9.944401840490799e-06, + "objective/entropy": 43.2598762512207, + "objective/kl": 19.98666000366211, + "objective/non_score_reward": -0.9993331432342529, + "objective/rlhf_reward": -5.997332572937012, + "objective/scores": -0.5, + "policy/approxkl_avg": 257.4547424316406, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.599609375, + "step": 87, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9966068267822266 + }, + { + "episode": 1424, + "epoch": 0.008531952882529868, + "loss/policy_avg": 0.09795168787240982, + "lr": 9.943762781186096e-06, + "objective/entropy": -59.364646911621094, + "objective/kl": 14.953709602355957, + "objective/non_score_reward": -0.7476855516433716, + "objective/rlhf_reward": -4.990742206573486, + "objective/scores": -0.5, + "policy/approxkl_avg": 55.110633850097656, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.640625, + "step": 88, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 1.9969828128814697 + }, + { + "episode": 1440, + "epoch": 0.008627817521659416, + "loss/policy_avg": 0.06303240358829498, + "lr": 9.94312372188139e-06, + "objective/entropy": 50.4556770324707, + "objective/kl": 19.505146026611328, + "objective/non_score_reward": -0.9752573370933533, + "objective/rlhf_reward": -3.9010292887687683, + "objective/scores": 0.0, + "policy/approxkl_avg": 34.922752380371094, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.607421875, + "step": 89, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 1.9964232444763184 + }, + { + "episode": 1456, + "epoch": 0.008723682160788965, + "loss/policy_avg": 0.2796894907951355, + "lr": 9.942484662576688e-06, + "objective/entropy": 135.44993591308594, + "objective/kl": 22.230022430419922, + "objective/non_score_reward": -1.1115009784698486, + "objective/rlhf_reward": -4.446004092693329, + "objective/scores": 0.0, + "policy/approxkl_avg": 48.1524658203125, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.56640625, + "step": 90, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.998495101928711 + }, + { + "episode": 1472, + "epoch": 0.008819546799918514, + "loss/policy_avg": 0.12483496963977814, + "lr": 9.941845603271985e-06, + "objective/entropy": 148.17709350585938, + "objective/kl": 17.60011100769043, + "objective/non_score_reward": -0.8800055384635925, + "objective/rlhf_reward": -5.520022392272949, + "objective/scores": -0.5, + "policy/approxkl_avg": 142.25204467773438, + "policy/clipfrac_avg": 1.5, + "policy/entropy_avg": 0.6171875, + "step": 91, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.998748779296875 + }, + { + "episode": 1488, + "epoch": 0.008915411439048063, + "loss/policy_avg": 0.0684453696012497, + "lr": 9.941206543967281e-06, + "objective/entropy": 25.60771942138672, + "objective/kl": 17.530319213867188, + "objective/non_score_reward": -0.8765159845352173, + "objective/rlhf_reward": -5.506063938140869, + "objective/scores": -0.5, + "policy/approxkl_avg": 70.80863189697266, + "policy/clipfrac_avg": 1.5, + "policy/entropy_avg": 0.625, + "step": 92, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9999198913574219 + }, + { + "episode": 1504, + "epoch": 0.009011276078177614, + "loss/policy_avg": 0.13488999009132385, + "lr": 9.940567484662578e-06, + "objective/entropy": -75.2538070678711, + "objective/kl": 19.696504592895508, + "objective/non_score_reward": -0.9848252534866333, + "objective/rlhf_reward": -3.939300984144211, + "objective/scores": 0.0, + "policy/approxkl_avg": 214.8182373046875, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.515625, + "step": 93, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9989702701568604 + }, + { + "episode": 1520, + "epoch": 0.009107140717307163, + "loss/policy_avg": 0.02409663423895836, + "lr": 9.939928425357874e-06, + "objective/entropy": 8.831840515136719, + "objective/kl": 25.456069946289062, + "objective/non_score_reward": -1.272803544998169, + "objective/rlhf_reward": -3.5754421589695777, + "objective/scores": 0.37894294565112985, + "policy/approxkl_avg": 176.86953735351562, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.4873046875, + "step": 94, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9964553117752075 + }, + { + "episode": 1536, + "epoch": 0.009203005356436712, + "loss/policy_avg": 0.0426328219473362, + "lr": 9.93928936605317e-06, + "objective/entropy": 185.92372131347656, + "objective/kl": 19.176239013671875, + "objective/non_score_reward": -0.95881187915802, + "objective/rlhf_reward": -3.8352474570274353, + "objective/scores": 0.0, + "policy/approxkl_avg": 269.95849609375, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.70703125, + "step": 95, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9930847883224487 + }, + { + "episode": 1552, + "epoch": 0.009298869995566261, + "loss/policy_avg": 0.3135666251182556, + "lr": 9.938650306748467e-06, + "objective/entropy": -119.88722229003906, + "objective/kl": 18.911632537841797, + "objective/non_score_reward": -0.9455816745758057, + "objective/rlhf_reward": -3.782326579093933, + "objective/scores": 0.0, + "policy/approxkl_avg": 136.56689453125, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.630859375, + "step": 96, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 2.0003180503845215 + }, + { + "episode": 1568, + "epoch": 0.00939473463469581, + "loss/policy_avg": 0.1893162876367569, + "lr": 9.938011247443764e-06, + "objective/entropy": 179.721435546875, + "objective/kl": 21.26153564453125, + "objective/non_score_reward": -1.0630767345428467, + "objective/rlhf_reward": -1.3285883411180701, + "objective/scores": 0.7309297535714575, + "policy/approxkl_avg": 203.61773681640625, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.70703125, + "step": 97, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.994171142578125 + }, + { + "episode": 1584, + "epoch": 0.009490599273825359, + "loss/policy_avg": 0.5632504224777222, + "lr": 9.937372188139061e-06, + "objective/entropy": 3.3514366149902344, + "objective/kl": 19.21142578125, + "objective/non_score_reward": -0.9605712890625, + "objective/rlhf_reward": -2.483035289977474, + "objective/scores": 0.33981246656813147, + "policy/approxkl_avg": 74.39619445800781, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.6953125, + "step": 98, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.99905526638031 + }, + { + "episode": 1600, + "epoch": 0.009586463912954908, + "loss/policy_avg": 0.20837292075157166, + "lr": 9.936733128834358e-06, + "objective/entropy": 121.03665161132812, + "objective/kl": 13.999438285827637, + "objective/non_score_reward": -0.6999719142913818, + "objective/rlhf_reward": -4.799887657165527, + "objective/scores": -0.5, + "policy/approxkl_avg": 32.508689880371094, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.375, + "step": 99, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 0, + "val/ratio": 1.9994678497314453 + }, + { + "episode": 1616, + "epoch": 0.009682328552084457, + "loss/policy_avg": 0.2726283669471741, + "lr": 9.936094069529653e-06, + "objective/entropy": 110.09475708007812, + "objective/kl": 15.960447311401367, + "objective/non_score_reward": -0.798022449016571, + "objective/rlhf_reward": -3.192089796066284, + "objective/scores": 0.0, + "policy/approxkl_avg": 61.767425537109375, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.55859375, + "step": 100, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 2.0027127265930176 + }, + { + "episode": 1632, + "epoch": 0.009778193191214006, + "loss/policy_avg": 0.2845292091369629, + "lr": 9.93545501022495e-06, + "objective/entropy": -153.4110107421875, + "objective/kl": 16.05643081665039, + "objective/non_score_reward": -0.8028215765953064, + "objective/rlhf_reward": 1.1887137234210972, + "objective/scores": 1.1, + "policy/approxkl_avg": 95.32630157470703, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.669921875, + "step": 101, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 1.997380256652832 + }, + { + "episode": 1648, + "epoch": 0.009874057830343555, + "loss/policy_avg": 0.18602727353572845, + "lr": 9.934815950920245e-06, + "objective/entropy": -13.683324813842773, + "objective/kl": 23.494054794311523, + "objective/non_score_reward": -1.174702763557434, + "objective/rlhf_reward": -6.698811054229736, + "objective/scores": -0.5, + "policy/approxkl_avg": 145.7095947265625, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.4912109375, + "step": 102, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9994981288909912 + }, + { + "episode": 1664, + "epoch": 0.009969922469473104, + "loss/policy_avg": 0.2709546983242035, + "lr": 9.934176891615542e-06, + "objective/entropy": -101.46907043457031, + "objective/kl": 22.274028778076172, + "objective/non_score_reward": -1.113701581954956, + "objective/rlhf_reward": -2.507394979672368, + "objective/scores": 0.4868528072345416, + "policy/approxkl_avg": 308.6561584472656, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.4912109375, + "step": 103, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 1.9981049299240112 + }, + { + "episode": 1680, + "epoch": 0.010065787108602653, + "loss/policy_avg": 0.0334465391933918, + "lr": 9.933537832310839e-06, + "objective/entropy": 130.1453857421875, + "objective/kl": 29.715213775634766, + "objective/non_score_reward": -1.4857605695724487, + "objective/rlhf_reward": -4.601406863241821, + "objective/scores": 0.33540891336663825, + "policy/approxkl_avg": 351.42138671875, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.5234375, + "step": 104, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.998427152633667 + }, + { + "episode": 1696, + "epoch": 0.010161651747732202, + "loss/policy_avg": 0.24828088283538818, + "lr": 9.932898773006136e-06, + "objective/entropy": 125.98516082763672, + "objective/kl": 15.857706069946289, + "objective/non_score_reward": -0.7928853034973145, + "objective/rlhf_reward": -1.720943163247451, + "objective/scores": 0.36264953503719355, + "policy/approxkl_avg": 74.20083618164062, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.578125, + "step": 105, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9993948936462402 + }, + { + "episode": 1712, + "epoch": 0.01025751638686175, + "loss/policy_avg": 0.2954871356487274, + "lr": 9.932259713701433e-06, + "objective/entropy": 97.68868255615234, + "objective/kl": 12.135580062866211, + "objective/non_score_reward": -0.6067790389060974, + "objective/rlhf_reward": -2.4271161258220673, + "objective/scores": 0.0, + "policy/approxkl_avg": 67.66595458984375, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.59375, + "step": 106, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9992032051086426 + }, + { + "episode": 1728, + "epoch": 0.0103533810259913, + "loss/policy_avg": 0.10418711602687836, + "lr": 9.931620654396728e-06, + "objective/entropy": -43.816890716552734, + "objective/kl": 19.110689163208008, + "objective/non_score_reward": -0.9555345773696899, + "objective/rlhf_reward": -5.82213830947876, + "objective/scores": -0.5, + "policy/approxkl_avg": 160.15283203125, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.6796875, + "step": 107, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 5, + "val/ratio": 1.9975645542144775 + }, + { + "episode": 1744, + "epoch": 0.01044924566512085, + "loss/policy_avg": 0.23229390382766724, + "lr": 9.930981595092025e-06, + "objective/entropy": 91.57461547851562, + "objective/kl": 18.9378662109375, + "objective/non_score_reward": -0.9468932747840881, + "objective/rlhf_reward": -3.787573218345642, + "objective/scores": 0.0, + "policy/approxkl_avg": 155.35989379882812, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.66796875, + "step": 108, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9963436126708984 + }, + { + "episode": 1760, + "epoch": 0.010545110304250399, + "loss/policy_avg": 0.3382238447666168, + "lr": 9.930342535787322e-06, + "objective/entropy": -49.52970886230469, + "objective/kl": 17.919204711914062, + "objective/non_score_reward": -0.89596027135849, + "objective/rlhf_reward": -5.583841323852539, + "objective/scores": -0.5, + "policy/approxkl_avg": 151.97140502929688, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.73828125, + "step": 109, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 1.9960722923278809 + }, + { + "episode": 1776, + "epoch": 0.010640974943379948, + "loss/policy_avg": 0.16102033853530884, + "lr": 9.929703476482619e-06, + "objective/entropy": -40.16828155517578, + "objective/kl": 15.826179504394531, + "objective/non_score_reward": -0.7913089990615845, + "objective/rlhf_reward": -3.1652360260486603, + "objective/scores": 0.0, + "policy/approxkl_avg": 32.21597671508789, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.767578125, + "step": 110, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 1.9961647987365723 + }, + { + "episode": 1792, + "epoch": 0.010736839582509497, + "loss/policy_avg": 0.08855805546045303, + "lr": 9.929064417177915e-06, + "objective/entropy": 187.74282836914062, + "objective/kl": 22.12034797668457, + "objective/non_score_reward": -1.1060173511505127, + "objective/rlhf_reward": -6.424069404602051, + "objective/scores": -0.5, + "policy/approxkl_avg": 89.00162506103516, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.80859375, + "step": 111, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9971915483474731 + }, + { + "episode": 1808, + "epoch": 0.010832704221639046, + "loss/policy_avg": 0.3315132260322571, + "lr": 9.928425357873212e-06, + "objective/entropy": -130.41551208496094, + "objective/kl": 20.600021362304688, + "objective/non_score_reward": -1.030001163482666, + "objective/rlhf_reward": -2.7414021278298915, + "objective/scores": 0.34465054211822604, + "policy/approxkl_avg": 231.54774475097656, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.671875, + "step": 112, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 2.0009350776672363 + }, + { + "episode": 1824, + "epoch": 0.010928568860768595, + "loss/policy_avg": 0.5379164814949036, + "lr": 9.927786298568507e-06, + "objective/entropy": 122.76021575927734, + "objective/kl": 24.667219161987305, + "objective/non_score_reward": -1.2333608865737915, + "objective/rlhf_reward": -6.933443546295166, + "objective/scores": -0.5, + "policy/approxkl_avg": 214.82647705078125, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.4638671875, + "step": 113, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9977924823760986 + }, + { + "episode": 1840, + "epoch": 0.011024433499898144, + "loss/policy_avg": 0.19039300084114075, + "lr": 9.927147239263804e-06, + "objective/entropy": -26.283668518066406, + "objective/kl": 21.043611526489258, + "objective/non_score_reward": -1.0521806478500366, + "objective/rlhf_reward": -2.6929507491909828, + "objective/scores": 0.37894294565112985, + "policy/approxkl_avg": 337.57025146484375, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.62890625, + "step": 114, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 1.9954912662506104 + }, + { + "episode": 1856, + "epoch": 0.011120298139027693, + "loss/policy_avg": 0.030586296692490578, + "lr": 9.926508179959101e-06, + "objective/entropy": 189.2314910888672, + "objective/kl": 18.47957992553711, + "objective/non_score_reward": -0.9239791035652161, + "objective/rlhf_reward": -3.6959164142608643, + "objective/scores": 0.0, + "policy/approxkl_avg": 158.6993865966797, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.8671875, + "step": 115, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9987783432006836 + }, + { + "episode": 1872, + "epoch": 0.011216162778157242, + "loss/policy_avg": 0.23665881156921387, + "lr": 9.925869120654398e-06, + "objective/entropy": 73.73204803466797, + "objective/kl": 19.052127838134766, + "objective/non_score_reward": -0.9526063203811646, + "objective/rlhf_reward": -5.810425281524658, + "objective/scores": -0.5, + "policy/approxkl_avg": 45.58797073364258, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.55859375, + "step": 116, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9978525638580322 + }, + { + "episode": 1888, + "epoch": 0.01131202741728679, + "loss/policy_avg": 0.11659398674964905, + "lr": 9.925230061349695e-06, + "objective/entropy": 128.39474487304688, + "objective/kl": 25.3045597076416, + "objective/non_score_reward": -1.265228033065796, + "objective/rlhf_reward": -2.6609121322631832, + "objective/scores": 0.6, + "policy/approxkl_avg": 76.14613342285156, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.64453125, + "step": 117, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9966604709625244 + }, + { + "episode": 1904, + "epoch": 0.01140789205641634, + "loss/policy_avg": 0.19203245639801025, + "lr": 9.92459100204499e-06, + "objective/entropy": 57.626686096191406, + "objective/kl": 30.407909393310547, + "objective/non_score_reward": -1.5203955173492432, + "objective/rlhf_reward": -1.6815817117691036, + "objective/scores": 1.1, + "policy/approxkl_avg": 324.47161865234375, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.544921875, + "step": 118, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 2.0000627040863037 + }, + { + "episode": 1920, + "epoch": 0.011503756695545889, + "loss/policy_avg": -0.02956828847527504, + "lr": 9.923951942740287e-06, + "objective/entropy": 160.20449829101562, + "objective/kl": 13.33430290222168, + "objective/non_score_reward": -0.666715145111084, + "objective/rlhf_reward": -2.666860580444336, + "objective/scores": 0.0, + "policy/approxkl_avg": 62.17939758300781, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.587890625, + "step": 119, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 2.003843307495117 + }, + { + "episode": 1936, + "epoch": 0.011599621334675438, + "loss/policy_avg": 0.3666956424713135, + "lr": 9.923312883435584e-06, + "objective/entropy": 173.73385620117188, + "objective/kl": 25.82461929321289, + "objective/non_score_reward": -1.2912311553955078, + "objective/rlhf_reward": -7.164924621582031, + "objective/scores": -0.5, + "policy/approxkl_avg": 248.4417724609375, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.71484375, + "step": 120, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 2.0001635551452637 + }, + { + "episode": 1952, + "epoch": 0.011695485973804987, + "loss/policy_avg": 0.07095953077077866, + "lr": 9.92267382413088e-06, + "objective/entropy": 60.89289855957031, + "objective/kl": 21.512653350830078, + "objective/non_score_reward": -1.0756325721740723, + "objective/rlhf_reward": -2.851932506175384, + "objective/scores": 0.36264953503719355, + "policy/approxkl_avg": 187.26104736328125, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.654296875, + "step": 121, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9986932277679443 + }, + { + "episode": 1968, + "epoch": 0.011791350612934537, + "loss/policy_avg": 0.11872611939907074, + "lr": 9.922034764826178e-06, + "objective/entropy": -24.511760711669922, + "objective/kl": 22.253305435180664, + "objective/non_score_reward": -1.1126651763916016, + "objective/rlhf_reward": -4.450661063194275, + "objective/scores": 0.0, + "policy/approxkl_avg": 199.84397888183594, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.6484375, + "step": 122, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 5, + "val/ratio": 1.9950015544891357 + }, + { + "episode": 1984, + "epoch": 0.011887215252064086, + "loss/policy_avg": 0.5726426839828491, + "lr": 9.921395705521473e-06, + "objective/entropy": 102.35612487792969, + "objective/kl": 32.768287658691406, + "objective/non_score_reward": -1.6384142637252808, + "objective/rlhf_reward": -8.553656578063965, + "objective/scores": -0.5, + "policy/approxkl_avg": 327.3544921875, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.533203125, + "step": 123, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.995248556137085 + }, + { + "episode": 2000, + "epoch": 0.011983079891193635, + "loss/policy_avg": 0.19069992005825043, + "lr": 9.92075664621677e-06, + "objective/entropy": 7.145952224731445, + "objective/kl": 17.727392196655273, + "objective/non_score_reward": -0.8863697052001953, + "objective/rlhf_reward": -3.545478705316782, + "objective/scores": 0.0, + "policy/approxkl_avg": 26.600868225097656, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.57421875, + "step": 124, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 2.0011134147644043 + }, + { + "episode": 2016, + "epoch": 0.012078944530323184, + "loss/policy_avg": 0.2572447657585144, + "lr": 9.920117586912067e-06, + "objective/entropy": 109.04229736328125, + "objective/kl": 17.79098129272461, + "objective/non_score_reward": -0.8895490765571594, + "objective/rlhf_reward": -5.558196067810059, + "objective/scores": -0.5, + "policy/approxkl_avg": 111.66732788085938, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.763671875, + "step": 125, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9968197345733643 + }, + { + "episode": 2032, + "epoch": 0.012174809169452733, + "loss/policy_avg": 0.043444547802209854, + "lr": 9.919478527607362e-06, + "objective/entropy": 75.83810424804688, + "objective/kl": 27.20602798461914, + "objective/non_score_reward": -1.3603014945983887, + "objective/rlhf_reward": -2.51748690450308, + "objective/scores": 0.7309297535714575, + "policy/approxkl_avg": 274.793701171875, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.521484375, + "step": 126, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9991501569747925 + }, + { + "episode": 2048, + "epoch": 0.012270673808582282, + "loss/policy_avg": 0.2138219177722931, + "lr": 9.918839468302659e-06, + "objective/entropy": 21.247840881347656, + "objective/kl": 14.299978256225586, + "objective/non_score_reward": -0.7149989008903503, + "objective/rlhf_reward": -2.8599955439567566, + "objective/scores": 0.0, + "policy/approxkl_avg": 21.416780471801758, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.80078125, + "step": 127, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 5, + "val/ratio": 1.998870849609375 + }, + { + "episode": 2064, + "epoch": 0.012366538447711831, + "loss/policy_avg": 0.23010344803333282, + "lr": 9.918200408997956e-06, + "objective/entropy": -76.91316223144531, + "objective/kl": 13.382017135620117, + "objective/non_score_reward": -0.6691007614135742, + "objective/rlhf_reward": -2.676403224468231, + "objective/scores": 0.0, + "policy/approxkl_avg": 37.060523986816406, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.677734375, + "step": 128, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 1.998915195465088 + }, + { + "episode": 2080, + "epoch": 0.01246240308684138, + "loss/policy_avg": 0.4017820954322815, + "lr": 9.917561349693252e-06, + "objective/entropy": 198.82456970214844, + "objective/kl": 22.337753295898438, + "objective/non_score_reward": -1.1168878078460693, + "objective/rlhf_reward": -2.9112917771011144, + "objective/scores": 0.38906482631788786, + "policy/approxkl_avg": 41.49570846557617, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.59765625, + "step": 129, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 0, + "val/ratio": 2.001622200012207 + }, + { + "episode": 2096, + "epoch": 0.012558267725970929, + "loss/policy_avg": 0.45664405822753906, + "lr": 9.91692229038855e-06, + "objective/entropy": 96.67610168457031, + "objective/kl": 13.830822944641113, + "objective/non_score_reward": -0.6915411353111267, + "objective/rlhf_reward": -2.7661644518375397, + "objective/scores": 0.0, + "policy/approxkl_avg": 49.5977783203125, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.87890625, + "step": 130, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 2.0018720626831055 + }, + { + "episode": 2112, + "epoch": 0.012654132365100478, + "loss/policy_avg": 0.18199189007282257, + "lr": 9.916283231083844e-06, + "objective/entropy": 12.164558410644531, + "objective/kl": 17.693878173828125, + "objective/non_score_reward": -0.8846939206123352, + "objective/rlhf_reward": -3.538775682449341, + "objective/scores": 0.0, + "policy/approxkl_avg": 3.7435173988342285, + "policy/clipfrac_avg": 1.5, + "policy/entropy_avg": 0.62890625, + "step": 131, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 2.0015268325805664 + }, + { + "episode": 2128, + "epoch": 0.012749997004230027, + "loss/policy_avg": 0.21469825506210327, + "lr": 9.915644171779141e-06, + "objective/entropy": -11.569038391113281, + "objective/kl": 14.204147338867188, + "objective/non_score_reward": -0.7102073431015015, + "objective/rlhf_reward": -1.481579744552059, + "objective/scores": 0.33981246656813147, + "policy/approxkl_avg": 91.71839141845703, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.658203125, + "step": 132, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 2.0028629302978516 + }, + { + "episode": 2144, + "epoch": 0.012845861643359576, + "loss/policy_avg": 0.27063143253326416, + "lr": 9.915005112474438e-06, + "objective/entropy": 180.4578857421875, + "objective/kl": 24.935741424560547, + "objective/non_score_reward": -1.2467870712280273, + "objective/rlhf_reward": -6.987148284912109, + "objective/scores": -0.5, + "policy/approxkl_avg": 138.678955078125, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.529296875, + "step": 133, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 0, + "val/ratio": 1.9967637062072754 + }, + { + "episode": 2160, + "epoch": 0.012941726282489125, + "loss/policy_avg": 0.1394023448228836, + "lr": 9.914366053169735e-06, + "objective/entropy": -29.98552703857422, + "objective/kl": 13.385698318481445, + "objective/non_score_reward": -0.6692849397659302, + "objective/rlhf_reward": -2.6771397292613983, + "objective/scores": 0.0, + "policy/approxkl_avg": 73.47354125976562, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.6171875, + "step": 134, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 2.000580310821533 + }, + { + "episode": 2176, + "epoch": 0.013037590921618674, + "loss/policy_avg": 0.0048561920411884785, + "lr": 9.913726993865032e-06, + "objective/entropy": 88.89292907714844, + "objective/kl": 28.03160858154297, + "objective/non_score_reward": -1.4015804529190063, + "objective/rlhf_reward": -4.227719643203121, + "objective/scores": 0.34465054211822604, + "policy/approxkl_avg": 370.16766357421875, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.666015625, + "step": 135, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.995047926902771 + }, + { + "episode": 2192, + "epoch": 0.013133455560748224, + "loss/policy_avg": 0.1565648913383484, + "lr": 9.913087934560329e-06, + "objective/entropy": 48.28108596801758, + "objective/kl": 22.514755249023438, + "objective/non_score_reward": -1.1257379055023193, + "objective/rlhf_reward": -0.10295168161392176, + "objective/scores": 1.1, + "policy/approxkl_avg": 272.63470458984375, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.58203125, + "step": 136, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 5, + "val/ratio": 2.0003504753112793 + }, + { + "episode": 2208, + "epoch": 0.013229320199877773, + "loss/policy_avg": 0.1350177526473999, + "lr": 9.912448875255624e-06, + "objective/entropy": 184.51797485351562, + "objective/kl": 30.795909881591797, + "objective/non_score_reward": -1.5397955179214478, + "objective/rlhf_reward": -8.159181594848633, + "objective/scores": -0.5, + "policy/approxkl_avg": 407.7762145996094, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.578125, + "step": 137, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 0, + "val/ratio": 1.997571349143982 + }, + { + "episode": 2224, + "epoch": 0.013325184839007322, + "loss/policy_avg": 0.2587956190109253, + "lr": 9.911809815950921e-06, + "objective/entropy": 14.785064697265625, + "objective/kl": 23.858671188354492, + "objective/non_score_reward": -1.1929335594177246, + "objective/rlhf_reward": -4.77173438668251, + "objective/scores": 0.0, + "policy/approxkl_avg": 258.4976501464844, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.58984375, + "step": 138, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.996340036392212 + }, + { + "episode": 2240, + "epoch": 0.013421049478136871, + "loss/policy_avg": 0.03932709991931915, + "lr": 9.911170756646218e-06, + "objective/entropy": -88.26953887939453, + "objective/kl": 11.428003311157227, + "objective/non_score_reward": -0.5714001655578613, + "objective/rlhf_reward": -4.285600662231445, + "objective/scores": -0.5, + "policy/approxkl_avg": 41.424224853515625, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.775390625, + "step": 139, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 2.0010673999786377 + }, + { + "episode": 2256, + "epoch": 0.01351691411726642, + "loss/policy_avg": 0.20215287804603577, + "lr": 9.910531697341515e-06, + "objective/entropy": 66.65933227539062, + "objective/kl": 20.929710388183594, + "objective/non_score_reward": -1.046485424041748, + "objective/rlhf_reward": 0.2140582442283634, + "objective/scores": 1.1, + "policy/approxkl_avg": 124.57344055175781, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.7421875, + "step": 140, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9977262020111084 + }, + { + "episode": 2272, + "epoch": 0.01361277875639597, + "loss/policy_avg": 0.5699018836021423, + "lr": 9.909892638036812e-06, + "objective/entropy": 10.020034790039062, + "objective/kl": 16.74536895751953, + "objective/non_score_reward": -0.8372684717178345, + "objective/rlhf_reward": 1.0509260237216953, + "objective/scores": 1.1, + "policy/approxkl_avg": 193.53329467773438, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.640625, + "step": 141, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9986698627471924 + }, + { + "episode": 2288, + "epoch": 0.013708643395525518, + "loss/policy_avg": 0.15781471133232117, + "lr": 9.909253578732107e-06, + "objective/entropy": -44.31187438964844, + "objective/kl": 30.742799758911133, + "objective/non_score_reward": -1.537139892578125, + "objective/rlhf_reward": -8.1485595703125, + "objective/scores": -0.5, + "policy/approxkl_avg": 158.5760498046875, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.73828125, + "step": 142, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 1.9987871646881104 + }, + { + "episode": 2304, + "epoch": 0.013804508034655067, + "loss/policy_avg": 0.09526471048593521, + "lr": 9.908614519427404e-06, + "objective/entropy": 52.471221923828125, + "objective/kl": 19.550655364990234, + "objective/non_score_reward": -0.9775327444076538, + "objective/rlhf_reward": -2.4595327778771967, + "objective/scores": 0.36264953503719355, + "policy/approxkl_avg": 176.07566833496094, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.91796875, + "step": 143, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9970924854278564 + }, + { + "episode": 2320, + "epoch": 0.013900372673784616, + "loss/policy_avg": 0.03243420645594597, + "lr": 9.9079754601227e-06, + "objective/entropy": 128.66928100585938, + "objective/kl": 21.24932861328125, + "objective/non_score_reward": -1.0624663829803467, + "objective/rlhf_reward": -1.849865472316742, + "objective/scores": 0.6, + "policy/approxkl_avg": 188.10623168945312, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.470703125, + "step": 144, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9996135234832764 + }, + { + "episode": 2336, + "epoch": 0.013996237312914165, + "loss/policy_avg": 0.25250673294067383, + "lr": 9.907336400817996e-06, + "objective/entropy": 198.611083984375, + "objective/kl": 21.650169372558594, + "objective/non_score_reward": -1.0825085639953613, + "objective/rlhf_reward": -6.330034255981445, + "objective/scores": -0.5, + "policy/approxkl_avg": 130.0052490234375, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.892578125, + "step": 145, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9994676113128662 + }, + { + "episode": 2352, + "epoch": 0.014092101952043714, + "loss/policy_avg": -0.034668684005737305, + "lr": 9.906697341513293e-06, + "objective/entropy": -27.681907653808594, + "objective/kl": 26.334529876708984, + "objective/non_score_reward": -1.316726565361023, + "objective/rlhf_reward": -3.907656395171566, + "objective/scores": 0.33981246656813147, + "policy/approxkl_avg": 210.81857299804688, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.552734375, + "step": 146, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 1.9972717761993408 + }, + { + "episode": 2368, + "epoch": 0.014187966591173263, + "loss/policy_avg": 0.12787118554115295, + "lr": 9.90605828220859e-06, + "objective/entropy": -43.486568450927734, + "objective/kl": 20.714540481567383, + "objective/non_score_reward": -1.0357270240783691, + "objective/rlhf_reward": -4.142907917499542, + "objective/scores": 0.0, + "policy/approxkl_avg": 86.59358978271484, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.673828125, + "step": 147, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 1.999436378479004 + }, + { + "episode": 2384, + "epoch": 0.014283831230302812, + "loss/policy_avg": 0.03302329033613205, + "lr": 9.905419222903886e-06, + "objective/entropy": 173.50836181640625, + "objective/kl": 23.32859230041504, + "objective/non_score_reward": -1.1664297580718994, + "objective/rlhf_reward": -3.3402058220206925, + "objective/scores": 0.3313782131597591, + "policy/approxkl_avg": 224.7312774658203, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.759765625, + "step": 148, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9986599683761597 + }, + { + "episode": 2400, + "epoch": 0.01437969586943236, + "loss/policy_avg": 0.04894339293241501, + "lr": 9.904780163599183e-06, + "objective/entropy": 48.957122802734375, + "objective/kl": 25.618064880371094, + "objective/non_score_reward": -1.2809032201766968, + "objective/rlhf_reward": -3.699780781467525, + "objective/scores": 0.35595802480981553, + "policy/approxkl_avg": 230.8973388671875, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.576171875, + "step": 149, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 5, + "val/ratio": 1.9963643550872803 + }, + { + "episode": 2416, + "epoch": 0.01447556050856191, + "loss/policy_avg": 0.36329030990600586, + "lr": 9.904141104294478e-06, + "objective/entropy": 157.50445556640625, + "objective/kl": 23.93838882446289, + "objective/non_score_reward": -1.1969194412231445, + "objective/rlhf_reward": -4.787678003311157, + "objective/scores": 0.0, + "policy/approxkl_avg": 200.30621337890625, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.6484375, + "step": 150, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9953033924102783 + }, + { + "episode": 2432, + "epoch": 0.01457142514769146, + "loss/policy_avg": 0.3649589419364929, + "lr": 9.903502044989775e-06, + "objective/entropy": 213.43943786621094, + "objective/kl": 19.777463912963867, + "objective/non_score_reward": -0.9888731837272644, + "objective/rlhf_reward": -2.6138571410471494, + "objective/scores": 0.33540891336663825, + "policy/approxkl_avg": 81.07418060302734, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.677734375, + "step": 151, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 0, + "val/ratio": 2.0006325244903564 + }, + { + "episode": 2448, + "epoch": 0.01466728978682101, + "loss/policy_avg": 0.35868164896965027, + "lr": 9.902862985685072e-06, + "objective/entropy": 2.8281936645507812, + "objective/kl": 23.836688995361328, + "objective/non_score_reward": -1.191834568977356, + "objective/rlhf_reward": -6.767337799072266, + "objective/scores": -0.5, + "policy/approxkl_avg": 91.45345306396484, + "policy/clipfrac_avg": 0.5, + "policy/entropy_avg": 0.5703125, + "step": 152, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 1.9970684051513672 + }, + { + "episode": 2464, + "epoch": 0.014763154425950558, + "loss/policy_avg": 0.07516692578792572, + "lr": 9.902223926380369e-06, + "objective/entropy": 171.71034240722656, + "objective/kl": 21.95725440979004, + "objective/non_score_reward": -1.097862720489502, + "objective/rlhf_reward": -4.3914510905742645, + "objective/scores": 0.0, + "policy/approxkl_avg": 221.61212158203125, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.53515625, + "step": 153, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 0, + "val/ratio": 1.9973832368850708 + }, + { + "episode": 2480, + "epoch": 0.014859019065080107, + "loss/policy_avg": 0.01844581961631775, + "lr": 9.901584867075666e-06, + "objective/entropy": 64.6440200805664, + "objective/kl": 17.436233520507812, + "objective/non_score_reward": -0.8718117475509644, + "objective/rlhf_reward": -5.487246990203857, + "objective/scores": -0.5, + "policy/approxkl_avg": 70.52445983886719, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.755859375, + "step": 154, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 2.000408172607422 + }, + { + "episode": 2496, + "epoch": 0.014954883704209656, + "loss/policy_avg": 0.05899505689740181, + "lr": 9.900945807770961e-06, + "objective/entropy": 123.7980728149414, + "objective/kl": 25.07213592529297, + "objective/non_score_reward": -1.253606915473938, + "objective/rlhf_reward": -7.014427661895752, + "objective/scores": -0.5, + "policy/approxkl_avg": 88.28120422363281, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.51171875, + "step": 155, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 2.000736951828003 + }, + { + "episode": 2512, + "epoch": 0.015050748343339205, + "loss/policy_avg": 0.4545804560184479, + "lr": 9.900306748466258e-06, + "objective/entropy": 10.871131896972656, + "objective/kl": 26.028505325317383, + "objective/non_score_reward": -1.3014252185821533, + "objective/rlhf_reward": -5.205701023340225, + "objective/scores": 0.0, + "policy/approxkl_avg": 217.84939575195312, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.5859375, + "step": 156, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9983347654342651 + }, + { + "episode": 2528, + "epoch": 0.015146612982468754, + "loss/policy_avg": 0.4191577136516571, + "lr": 9.899667689161555e-06, + "objective/entropy": 109.52301025390625, + "objective/kl": 27.344154357910156, + "objective/non_score_reward": -1.3672077655792236, + "objective/rlhf_reward": -7.4688310623168945, + "objective/scores": -0.5, + "policy/approxkl_avg": 89.15927124023438, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.62109375, + "step": 157, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.999230146408081 + }, + { + "episode": 2544, + "epoch": 0.015242477621598303, + "loss/policy_avg": 0.1439390629529953, + "lr": 9.899028629856852e-06, + "objective/entropy": 220.28952026367188, + "objective/kl": 19.178768157958984, + "objective/non_score_reward": -0.9589384198188782, + "objective/rlhf_reward": -3.8357537388801575, + "objective/scores": 0.0, + "policy/approxkl_avg": 71.46617126464844, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.8828125, + "step": 158, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.997718334197998 + }, + { + "episode": 2560, + "epoch": 0.015338342260727852, + "loss/policy_avg": 0.30983591079711914, + "lr": 9.898389570552149e-06, + "objective/entropy": 104.64752197265625, + "objective/kl": 27.657455444335938, + "objective/non_score_reward": -1.3828728199005127, + "objective/rlhf_reward": -3.7981575886408487, + "objective/scores": 0.43333333333333335, + "policy/approxkl_avg": 138.6593017578125, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.560546875, + "step": 159, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9977858066558838 + }, + { + "episode": 2576, + "epoch": 0.015434206899857401, + "loss/policy_avg": 0.07094208896160126, + "lr": 9.897750511247446e-06, + "objective/entropy": 90.89056396484375, + "objective/kl": 21.30394744873047, + "objective/non_score_reward": -1.065197467803955, + "objective/rlhf_reward": -4.260790050029755, + "objective/scores": 0.0, + "policy/approxkl_avg": 297.8140563964844, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.80859375, + "step": 160, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 2.0023467540740967 + }, + { + "episode": 2592, + "epoch": 0.01553007153898695, + "loss/policy_avg": 0.16327880322933197, + "lr": 9.89711145194274e-06, + "objective/entropy": 100.65301513671875, + "objective/kl": 24.78557586669922, + "objective/non_score_reward": -1.239278793334961, + "objective/rlhf_reward": -3.1322867229309788, + "objective/scores": 0.4562071871080222, + "policy/approxkl_avg": 208.9399871826172, + "policy/clipfrac_avg": 1.5, + "policy/entropy_avg": 0.625, + "step": 161, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 2.0041985511779785 + }, + { + "episode": 2608, + "epoch": 0.0156259361781165, + "loss/policy_avg": 0.216099351644516, + "lr": 9.896472392638038e-06, + "objective/entropy": 57.33685302734375, + "objective/kl": 29.304649353027344, + "objective/non_score_reward": -1.4652326107025146, + "objective/rlhf_reward": -7.8609299659729, + "objective/scores": -0.5, + "policy/approxkl_avg": 207.6790313720703, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.833984375, + "step": 162, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9980382919311523 + }, + { + "episode": 2624, + "epoch": 0.01572180081724605, + "loss/policy_avg": 0.0953613817691803, + "lr": 9.895833333333334e-06, + "objective/entropy": -133.8697967529297, + "objective/kl": 16.73604965209961, + "objective/non_score_reward": -0.83680260181427, + "objective/rlhf_reward": -3.347210466861725, + "objective/scores": 0.0, + "policy/approxkl_avg": 145.06759643554688, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.802734375, + "step": 163, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 6, + "val/ratio": 1.9988582134246826 + }, + { + "episode": 2640, + "epoch": 0.0158176654563756, + "loss/policy_avg": 0.42890581488609314, + "lr": 9.895194274028631e-06, + "objective/entropy": 180.38433837890625, + "objective/kl": 25.974315643310547, + "objective/non_score_reward": -1.2987157106399536, + "objective/rlhf_reward": -7.1948628425598145, + "objective/scores": -0.5, + "policy/approxkl_avg": 126.24812316894531, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.73046875, + "step": 164, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9987969398498535 + }, + { + "episode": 2656, + "epoch": 0.015913530095505148, + "loss/policy_avg": 0.1958284080028534, + "lr": 9.894555214723928e-06, + "objective/entropy": 174.94032287597656, + "objective/kl": 25.64311981201172, + "objective/non_score_reward": -1.282155990600586, + "objective/rlhf_reward": -3.786988249331146, + "objective/scores": 0.33540891336663825, + "policy/approxkl_avg": 86.50934600830078, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.705078125, + "step": 165, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9944665431976318 + }, + { + "episode": 2672, + "epoch": 0.016009394734634697, + "loss/policy_avg": 0.3368389904499054, + "lr": 9.893916155419225e-06, + "objective/entropy": 10.10284423828125, + "objective/kl": 24.560279846191406, + "objective/non_score_reward": -1.2280139923095703, + "objective/rlhf_reward": -6.912055969238281, + "objective/scores": -0.5, + "policy/approxkl_avg": 80.86394500732422, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.720703125, + "step": 166, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9986299276351929 + }, + { + "episode": 2688, + "epoch": 0.016105259373764245, + "loss/policy_avg": 0.11198948323726654, + "lr": 9.89327709611452e-06, + "objective/entropy": 161.62661743164062, + "objective/kl": 39.22645568847656, + "objective/non_score_reward": -1.9613227844238281, + "objective/rlhf_reward": -7.845290899276733, + "objective/scores": 0.0, + "policy/approxkl_avg": 164.2472381591797, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.841796875, + "step": 167, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 2.000530958175659 + }, + { + "episode": 2704, + "epoch": 0.016201124012893794, + "loss/policy_avg": 0.19894596934318542, + "lr": 9.892638036809815e-06, + "objective/entropy": 137.1651153564453, + "objective/kl": 24.861934661865234, + "objective/non_score_reward": -1.2430968284606934, + "objective/rlhf_reward": -3.147558684619974, + "objective/scores": 0.4562071871080222, + "policy/approxkl_avg": 133.68283081054688, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.66796875, + "step": 168, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.997298240661621 + }, + { + "episode": 2720, + "epoch": 0.016296988652023343, + "loss/policy_avg": 0.12182526290416718, + "lr": 9.891998977505112e-06, + "objective/entropy": 18.006725311279297, + "objective/kl": 17.92361068725586, + "objective/non_score_reward": -0.8961805701255798, + "objective/rlhf_reward": -2.134124155254707, + "objective/scores": 0.36264953503719355, + "policy/approxkl_avg": 192.45278930664062, + "policy/clipfrac_avg": 0.5, + "policy/entropy_avg": 0.583984375, + "step": 169, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 2.0006425380706787 + }, + { + "episode": 2736, + "epoch": 0.016392853291152892, + "loss/policy_avg": -0.012196972966194153, + "lr": 9.89135991820041e-06, + "objective/entropy": 115.16173553466797, + "objective/kl": 21.65146827697754, + "objective/non_score_reward": -1.082573413848877, + "objective/rlhf_reward": -2.5054651006785145, + "objective/scores": 0.4562071871080222, + "policy/approxkl_avg": 145.09487915039062, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.71484375, + "step": 170, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9983795881271362 + }, + { + "episode": 2752, + "epoch": 0.01648871793028244, + "loss/policy_avg": 0.5326859951019287, + "lr": 9.890720858895706e-06, + "objective/entropy": 95.13655853271484, + "objective/kl": 24.00056266784668, + "objective/non_score_reward": -1.2000280618667603, + "objective/rlhf_reward": -4.800112426280975, + "objective/scores": 0.0, + "policy/approxkl_avg": 119.50138854980469, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.634765625, + "step": 171, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 2.0006933212280273 + }, + { + "episode": 2768, + "epoch": 0.01658458256941199, + "loss/policy_avg": 0.3668867349624634, + "lr": 9.890081799591003e-06, + "objective/entropy": 132.36126708984375, + "objective/kl": 21.386262893676758, + "objective/non_score_reward": -1.0693132877349854, + "objective/rlhf_reward": -1.8772529125213624, + "objective/scores": 0.6, + "policy/approxkl_avg": 48.470794677734375, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.666015625, + "step": 172, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9946337938308716 + }, + { + "episode": 2784, + "epoch": 0.01668044720854154, + "loss/policy_avg": 0.460104763507843, + "lr": 9.8894427402863e-06, + "objective/entropy": 129.8038330078125, + "objective/kl": 25.860858917236328, + "objective/non_score_reward": -1.2930430173873901, + "objective/rlhf_reward": -3.656400167735752, + "objective/scores": 0.37894294565112985, + "policy/approxkl_avg": 146.58050537109375, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.5625, + "step": 173, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9949872493743896 + }, + { + "episode": 2800, + "epoch": 0.016776311847671088, + "loss/policy_avg": 0.11980315297842026, + "lr": 9.888803680981595e-06, + "objective/entropy": 173.85202026367188, + "objective/kl": 23.159679412841797, + "objective/non_score_reward": -1.1579840183258057, + "objective/rlhf_reward": -6.631936073303223, + "objective/scores": -0.5, + "policy/approxkl_avg": 84.56037902832031, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.7734375, + "step": 174, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9972279071807861 + }, + { + "episode": 2816, + "epoch": 0.016872176486800637, + "loss/policy_avg": 0.060305699706077576, + "lr": 9.888164621676892e-06, + "objective/entropy": 90.63494110107422, + "objective/kl": 23.04631805419922, + "objective/non_score_reward": -1.1523159742355347, + "objective/rlhf_reward": -4.609263688325882, + "objective/scores": 0.0, + "policy/approxkl_avg": 204.8768310546875, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.564453125, + "step": 175, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 1.9992055892944336 + }, + { + "episode": 2832, + "epoch": 0.016968041125930186, + "loss/policy_avg": 0.5130124092102051, + "lr": 9.887525562372189e-06, + "objective/entropy": 66.25984191894531, + "objective/kl": 32.70683288574219, + "objective/non_score_reward": -1.635341763496399, + "objective/rlhf_reward": -6.541367173194885, + "objective/scores": 0.0, + "policy/approxkl_avg": 265.0827941894531, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.609375, + "step": 176, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.996222734451294 + }, + { + "episode": 2848, + "epoch": 0.017063905765059735, + "loss/policy_avg": 0.008577877655625343, + "lr": 9.886886503067486e-06, + "objective/entropy": -118.17359924316406, + "objective/kl": 21.02519989013672, + "objective/non_score_reward": -1.051259994506836, + "objective/rlhf_reward": -4.205039799213409, + "objective/scores": 0.0, + "policy/approxkl_avg": 260.4126892089844, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.66015625, + "step": 177, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 6, + "val/ratio": 2.000894069671631 + }, + { + "episode": 2864, + "epoch": 0.017159770404189284, + "loss/policy_avg": 0.18238189816474915, + "lr": 9.886247443762783e-06, + "objective/entropy": 130.0546875, + "objective/kl": 32.371009826660156, + "objective/non_score_reward": -1.6185506582260132, + "objective/rlhf_reward": -8.474203109741211, + "objective/scores": -0.5, + "policy/approxkl_avg": 228.2266387939453, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.7890625, + "step": 178, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9942381381988525 + }, + { + "episode": 2880, + "epoch": 0.017255635043318833, + "loss/policy_avg": 0.18286140263080597, + "lr": 9.88560838445808e-06, + "objective/entropy": -10.639881134033203, + "objective/kl": 29.253890991210938, + "objective/non_score_reward": -1.462694525718689, + "objective/rlhf_reward": -7.850778102874756, + "objective/scores": -0.5, + "policy/approxkl_avg": 148.62832641601562, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.798828125, + "step": 179, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 1.9988160133361816 + }, + { + "episode": 2896, + "epoch": 0.017351499682448382, + "loss/policy_avg": 0.07891340553760529, + "lr": 9.884969325153375e-06, + "objective/entropy": -120.97007751464844, + "objective/kl": 21.97601890563965, + "objective/non_score_reward": -1.0988008975982666, + "objective/rlhf_reward": -2.2724974177041393, + "objective/scores": 0.5306765580733931, + "policy/approxkl_avg": 200.60455322265625, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.533203125, + "step": 180, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 5, + "val/ratio": 2.000365972518921 + }, + { + "episode": 2912, + "epoch": 0.01744736432157793, + "loss/policy_avg": 0.06744587421417236, + "lr": 9.884330265848671e-06, + "objective/entropy": 73.97216796875, + "objective/kl": 19.66523551940918, + "objective/non_score_reward": -0.9832619428634644, + "objective/rlhf_reward": -5.933047771453857, + "objective/scores": -0.5, + "policy/approxkl_avg": 168.08172607421875, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.603515625, + "step": 181, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.997870683670044 + }, + { + "episode": 2928, + "epoch": 0.01754322896070748, + "loss/policy_avg": 0.12424597889184952, + "lr": 9.883691206543968e-06, + "objective/entropy": 77.82262420654297, + "objective/kl": 21.0150146484375, + "objective/non_score_reward": -1.050750732421875, + "objective/rlhf_reward": 0.1969969511032108, + "objective/scores": 1.1, + "policy/approxkl_avg": 109.60333251953125, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.693359375, + "step": 182, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.996559500694275 + }, + { + "episode": 2944, + "epoch": 0.01763909359983703, + "loss/policy_avg": 0.2606327533721924, + "lr": 9.883052147239265e-06, + "objective/entropy": 172.60293579101562, + "objective/kl": 29.473426818847656, + "objective/non_score_reward": -1.4736714363098145, + "objective/rlhf_reward": -5.894685626029968, + "objective/scores": 0.0, + "policy/approxkl_avg": 194.51976013183594, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.818359375, + "step": 183, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9962568283081055 + }, + { + "episode": 2960, + "epoch": 0.017734958238966578, + "loss/policy_avg": 0.10910254716873169, + "lr": 9.882413087934562e-06, + "objective/entropy": 240.20162963867188, + "objective/kl": 15.176373481750488, + "objective/non_score_reward": -0.7588187456130981, + "objective/rlhf_reward": 1.3647250771522526, + "objective/scores": 1.1, + "policy/approxkl_avg": 43.88645935058594, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.8203125, + "step": 184, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 2.0006511211395264 + }, + { + "episode": 2976, + "epoch": 0.017830822878096127, + "loss/policy_avg": 0.5547807812690735, + "lr": 9.881774028629857e-06, + "objective/entropy": 85.18072509765625, + "objective/kl": 21.537092208862305, + "objective/non_score_reward": -1.0768545866012573, + "objective/rlhf_reward": -6.307418346405029, + "objective/scores": -0.5, + "policy/approxkl_avg": 161.64654541015625, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.75390625, + "step": 185, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9996278285980225 + }, + { + "episode": 2992, + "epoch": 0.017926687517225676, + "loss/policy_avg": 0.3181283175945282, + "lr": 9.881134969325154e-06, + "objective/entropy": 129.28257751464844, + "objective/kl": 28.59075927734375, + "objective/non_score_reward": -1.4295378923416138, + "objective/rlhf_reward": -5.718151569366455, + "objective/scores": 0.0, + "policy/approxkl_avg": 53.536468505859375, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.705078125, + "step": 186, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.998319149017334 + }, + { + "episode": 3008, + "epoch": 0.018022552156355228, + "loss/policy_avg": 0.37268152832984924, + "lr": 9.880495910020451e-06, + "objective/entropy": 183.25308227539062, + "objective/kl": 29.692989349365234, + "objective/non_score_reward": -1.484649419784546, + "objective/rlhf_reward": -7.938598155975342, + "objective/scores": -0.5, + "policy/approxkl_avg": 91.56600952148438, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.599609375, + "step": 187, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9983760118484497 + }, + { + "episode": 3024, + "epoch": 0.018118416795484777, + "loss/policy_avg": 0.7035294771194458, + "lr": 9.879856850715748e-06, + "objective/entropy": -141.20687866210938, + "objective/kl": 16.28227996826172, + "objective/non_score_reward": -0.8141138553619385, + "objective/rlhf_reward": -5.256455421447754, + "objective/scores": -0.5, + "policy/approxkl_avg": 39.17454528808594, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.912109375, + "step": 188, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 5, + "val/ratio": 2.000697612762451 + }, + { + "episode": 3040, + "epoch": 0.018214281434614326, + "loss/policy_avg": 0.23234151303768158, + "lr": 9.879217791411043e-06, + "objective/entropy": 41.34138107299805, + "objective/kl": 27.16008758544922, + "objective/non_score_reward": -1.3580043315887451, + "objective/rlhf_reward": -4.05341557511459, + "objective/scores": 0.34465054211822604, + "policy/approxkl_avg": 271.9233093261719, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.51171875, + "step": 189, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.999565839767456 + }, + { + "episode": 3056, + "epoch": 0.018310146073743875, + "loss/policy_avg": 0.07211380451917648, + "lr": 9.87857873210634e-06, + "objective/entropy": 20.789365768432617, + "objective/kl": 19.183855056762695, + "objective/non_score_reward": -0.9591927528381348, + "objective/rlhf_reward": -5.836771011352539, + "objective/scores": -0.5, + "policy/approxkl_avg": 108.85220336914062, + "policy/clipfrac_avg": 0.5, + "policy/entropy_avg": 0.6328125, + "step": 190, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 2.0007762908935547 + }, + { + "episode": 3072, + "epoch": 0.018406010712873424, + "loss/policy_avg": 0.7360332012176514, + "lr": 9.877939672801637e-06, + "objective/entropy": 219.01002502441406, + "objective/kl": 30.353984832763672, + "objective/non_score_reward": -1.5176992416381836, + "objective/rlhf_reward": -4.692195155707699, + "objective/scores": 0.34465054211822604, + "policy/approxkl_avg": 60.275230407714844, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.73046875, + "step": 191, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 1.9968408346176147 + }, + { + "episode": 3088, + "epoch": 0.018501875352002973, + "loss/policy_avg": 0.7660672664642334, + "lr": 9.877300613496934e-06, + "objective/entropy": 192.5721435546875, + "objective/kl": 18.974138259887695, + "objective/non_score_reward": -0.948706865310669, + "objective/rlhf_reward": -5.794827461242676, + "objective/scores": -0.5, + "policy/approxkl_avg": 208.0977783203125, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.568359375, + "step": 192, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 0, + "val/ratio": 1.996788740158081 + }, + { + "episode": 3104, + "epoch": 0.018597739991132522, + "loss/policy_avg": 0.4530583620071411, + "lr": 9.876661554192229e-06, + "objective/entropy": 185.2235107421875, + "objective/kl": 24.102296829223633, + "objective/non_score_reward": -1.2051149606704712, + "objective/rlhf_reward": -0.4204598426818844, + "objective/scores": 1.1, + "policy/approxkl_avg": 34.94757080078125, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.728515625, + "step": 193, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9967865943908691 + }, + { + "episode": 3120, + "epoch": 0.01869360463026207, + "loss/policy_avg": 0.03487266227602959, + "lr": 9.876022494887526e-06, + "objective/entropy": 183.6469268798828, + "objective/kl": 20.133817672729492, + "objective/non_score_reward": -1.0066908597946167, + "objective/rlhf_reward": -4.026763558387756, + "objective/scores": 0.0, + "policy/approxkl_avg": 168.301025390625, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.7578125, + "step": 194, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 1, + "val/ratio": 2.0027122497558594 + }, + { + "episode": 3136, + "epoch": 0.01878946926939162, + "loss/policy_avg": -0.029073666781187057, + "lr": 9.875383435582823e-06, + "objective/entropy": 138.25656127929688, + "objective/kl": 18.322010040283203, + "objective/non_score_reward": -0.9161005020141602, + "objective/rlhf_reward": -5.664402008056641, + "objective/scores": -0.5, + "policy/approxkl_avg": 60.4761848449707, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.673828125, + "step": 195, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 2, + "val/ratio": 1.9991329908370972 + }, + { + "episode": 3152, + "epoch": 0.01888533390852117, + "loss/policy_avg": 0.14693962037563324, + "lr": 9.87474437627812e-06, + "objective/entropy": 71.7930908203125, + "objective/kl": 19.49433135986328, + "objective/non_score_reward": -0.9747166633605957, + "objective/rlhf_reward": -3.8988667130470276, + "objective/scores": 0.0, + "policy/approxkl_avg": 126.81082153320312, + "policy/clipfrac_avg": 1.25, + "policy/entropy_avg": 0.767578125, + "step": 196, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.9977989196777344 + }, + { + "episode": 3168, + "epoch": 0.018981198547650718, + "loss/policy_avg": 0.6557031869888306, + "lr": 9.874105316973416e-06, + "objective/entropy": -9.503684997558594, + "objective/kl": 21.540775299072266, + "objective/non_score_reward": -1.0770388841629028, + "objective/rlhf_reward": -2.9295533085740626, + "objective/scores": 0.34465054211822604, + "policy/approxkl_avg": 100.91127014160156, + "policy/clipfrac_avg": 0.75, + "policy/entropy_avg": 0.619140625, + "step": 197, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 4, + "val/ratio": 1.9948209524154663 + }, + { + "episode": 3184, + "epoch": 0.019077063186780267, + "loss/policy_avg": 0.23461255431175232, + "lr": 9.873466257668712e-06, + "objective/entropy": -49.81024932861328, + "objective/kl": 20.112146377563477, + "objective/non_score_reward": -1.0056073665618896, + "objective/rlhf_reward": -6.0224289894104, + "objective/scores": -0.5, + "policy/approxkl_avg": 135.8631134033203, + "policy/clipfrac_avg": 1.0, + "policy/entropy_avg": 0.58203125, + "step": 198, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 1.999420404434204 + }, + { + "episode": 3200, + "epoch": 0.019172927825909816, + "loss/policy_avg": -0.5007312297821045, + "lr": 9.872827198364009e-06, + "objective/entropy": -25.718414306640625, + "objective/kl": 15.317103385925293, + "objective/non_score_reward": -0.7658551931381226, + "objective/rlhf_reward": -5.06342077255249, + "objective/scores": -0.5, + "policy/approxkl_avg": 154.1348876953125, + "policy/clipfrac_avg": 1.75, + "policy/entropy_avg": 0.796875, + "step": 199, + "val/clipfrac_avg": 0.0, + "val/num_eos_tokens": 3, + "val/ratio": 2.01686954498291 + } + ], + "logging_steps": 500, + "max_steps": 7824, + "num_input_tokens_seen": 0, + "num_train_epochs": 3.0, + "save_steps": 200, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": false + }, + "attributes": {} + } + }, + "total_flos": 0, + "train_batch_size": null, + "trial_name": null, + "trial_params": null +}