| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.9998757609640949, | |
| "eval_steps": 100, | |
| "global_step": 1509, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "completion_length": 544.525013923645, | |
| "epoch": 0.0033130409574688366, | |
| "grad_norm": 0.2902330414514582, | |
| "kl": 0.00014667510986328126, | |
| "learning_rate": 6.622516556291392e-07, | |
| "loss": 0.0, | |
| "reward": 0.20833333805203438, | |
| "reward_std": 0.1532064698636532, | |
| "rewards/accuracy_reward": 0.20833333805203438, | |
| "rewards/format_reward": 0.0, | |
| "step": 5 | |
| }, | |
| { | |
| "completion_length": 539.743765258789, | |
| "epoch": 0.006626081914937673, | |
| "grad_norm": 0.4239089441988848, | |
| "kl": 0.0002197861671447754, | |
| "learning_rate": 1.3245033112582784e-06, | |
| "loss": 0.0, | |
| "reward": 0.20000000428408385, | |
| "reward_std": 0.14731391314417125, | |
| "rewards/accuracy_reward": 0.20000000428408385, | |
| "rewards/format_reward": 0.0, | |
| "step": 10 | |
| }, | |
| { | |
| "completion_length": 555.6875169754028, | |
| "epoch": 0.00993912287240651, | |
| "grad_norm": 0.5514350439282207, | |
| "kl": 0.0004414021968841553, | |
| "learning_rate": 1.9867549668874175e-06, | |
| "loss": 0.0, | |
| "reward": 0.19375000502914191, | |
| "reward_std": 0.13258252199739218, | |
| "rewards/accuracy_reward": 0.1916666716337204, | |
| "rewards/format_reward": 0.002083333395421505, | |
| "step": 15 | |
| }, | |
| { | |
| "completion_length": 569.2666831970215, | |
| "epoch": 0.013252163829875346, | |
| "grad_norm": 0.3228506299764182, | |
| "kl": 0.0017175555229187011, | |
| "learning_rate": 2.6490066225165567e-06, | |
| "loss": 0.0001, | |
| "reward": 0.24375000558793544, | |
| "reward_std": 0.15026019159704446, | |
| "rewards/accuracy_reward": 0.24375000558793544, | |
| "rewards/format_reward": 0.0, | |
| "step": 20 | |
| }, | |
| { | |
| "completion_length": 547.4958526611329, | |
| "epoch": 0.016565204787344183, | |
| "grad_norm": 0.4765963281561274, | |
| "kl": 0.0036331653594970704, | |
| "learning_rate": 3.311258278145696e-06, | |
| "loss": 0.0001, | |
| "reward": 0.24166667219251395, | |
| "reward_std": 0.13552880007773638, | |
| "rewards/accuracy_reward": 0.24166667219251395, | |
| "rewards/format_reward": 0.0, | |
| "step": 25 | |
| }, | |
| { | |
| "completion_length": 544.0520998001099, | |
| "epoch": 0.01987824574481302, | |
| "grad_norm": 0.21762305936684645, | |
| "kl": 0.00529632568359375, | |
| "learning_rate": 3.973509933774835e-06, | |
| "loss": 0.0002, | |
| "reward": 0.260416672565043, | |
| "reward_std": 0.10311973933130503, | |
| "rewards/accuracy_reward": 0.260416672565043, | |
| "rewards/format_reward": 0.0, | |
| "step": 30 | |
| }, | |
| { | |
| "completion_length": 561.004182434082, | |
| "epoch": 0.023191286702281856, | |
| "grad_norm": 0.42754577646802083, | |
| "kl": 0.0049117088317871095, | |
| "learning_rate": 4.635761589403974e-06, | |
| "loss": 0.0002, | |
| "reward": 0.23541667256504298, | |
| "reward_std": 0.15026019141077995, | |
| "rewards/accuracy_reward": 0.23541667256504298, | |
| "rewards/format_reward": 0.0, | |
| "step": 35 | |
| }, | |
| { | |
| "completion_length": 558.4416833877564, | |
| "epoch": 0.026504327659750693, | |
| "grad_norm": 0.38474962439429805, | |
| "kl": 0.00634613037109375, | |
| "learning_rate": 5.2980132450331135e-06, | |
| "loss": 0.0003, | |
| "reward": 0.23958333898335696, | |
| "reward_std": 0.13258252181112767, | |
| "rewards/accuracy_reward": 0.23958333898335696, | |
| "rewards/format_reward": 0.0, | |
| "step": 40 | |
| }, | |
| { | |
| "completion_length": 587.2520998001098, | |
| "epoch": 0.02981736861721953, | |
| "grad_norm": 0.38278137652247257, | |
| "kl": 0.008282661437988281, | |
| "learning_rate": 5.960264900662252e-06, | |
| "loss": 0.0003, | |
| "reward": 0.26666667237877845, | |
| "reward_std": 0.12963624354451894, | |
| "rewards/accuracy_reward": 0.26666667237877845, | |
| "rewards/format_reward": 0.0, | |
| "step": 45 | |
| }, | |
| { | |
| "completion_length": 585.193766784668, | |
| "epoch": 0.033130409574688366, | |
| "grad_norm": 0.46609514480011244, | |
| "kl": 0.011644363403320312, | |
| "learning_rate": 6.622516556291392e-06, | |
| "loss": 0.0005, | |
| "reward": 0.28750000689178706, | |
| "reward_std": 0.1590990262106061, | |
| "rewards/accuracy_reward": 0.28750000689178706, | |
| "rewards/format_reward": 0.0, | |
| "step": 50 | |
| }, | |
| { | |
| "completion_length": 604.7958530426025, | |
| "epoch": 0.0364434505321572, | |
| "grad_norm": 0.39658289344370634, | |
| "kl": 0.014037322998046876, | |
| "learning_rate": 7.28476821192053e-06, | |
| "loss": 0.0006, | |
| "reward": 0.2145833380520344, | |
| "reward_std": 0.1207974087446928, | |
| "rewards/accuracy_reward": 0.2145833380520344, | |
| "rewards/format_reward": 0.0, | |
| "step": 55 | |
| }, | |
| { | |
| "completion_length": 573.2937698364258, | |
| "epoch": 0.03975649148962604, | |
| "grad_norm": 0.3562470913913186, | |
| "kl": 0.015195465087890625, | |
| "learning_rate": 7.94701986754967e-06, | |
| "loss": 0.0006, | |
| "reward": 0.2666666736826301, | |
| "reward_std": 0.16499158274382353, | |
| "rewards/accuracy_reward": 0.2666666736826301, | |
| "rewards/format_reward": 0.0, | |
| "step": 60 | |
| }, | |
| { | |
| "completion_length": 523.7500129699707, | |
| "epoch": 0.043069532447094876, | |
| "grad_norm": 0.44620950890040406, | |
| "kl": 0.021143341064453126, | |
| "learning_rate": 8.609271523178809e-06, | |
| "loss": 0.0008, | |
| "reward": 0.2583333391696215, | |
| "reward_std": 0.1590990262106061, | |
| "rewards/accuracy_reward": 0.2583333391696215, | |
| "rewards/format_reward": 0.0, | |
| "step": 65 | |
| }, | |
| { | |
| "completion_length": 570.8146015167237, | |
| "epoch": 0.04638257340456371, | |
| "grad_norm": 0.3096717335604667, | |
| "kl": 0.02764739990234375, | |
| "learning_rate": 9.271523178807948e-06, | |
| "loss": 0.0011, | |
| "reward": 0.28541667349636557, | |
| "reward_std": 0.1797229742631316, | |
| "rewards/accuracy_reward": 0.28541667349636557, | |
| "rewards/format_reward": 0.0, | |
| "step": 70 | |
| }, | |
| { | |
| "completion_length": 560.4291816711426, | |
| "epoch": 0.04969561436203255, | |
| "grad_norm": 0.41540849647955475, | |
| "kl": 0.03727569580078125, | |
| "learning_rate": 9.933774834437086e-06, | |
| "loss": 0.0015, | |
| "reward": 0.3145833402872086, | |
| "reward_std": 0.1738304177299142, | |
| "rewards/accuracy_reward": 0.3145833402872086, | |
| "rewards/format_reward": 0.0, | |
| "step": 75 | |
| }, | |
| { | |
| "completion_length": 588.3562690734864, | |
| "epoch": 0.053008655319501385, | |
| "grad_norm": 0.3491763818431896, | |
| "kl": 0.0417724609375, | |
| "learning_rate": 1.0596026490066227e-05, | |
| "loss": 0.0017, | |
| "reward": 0.2979166731238365, | |
| "reward_std": 0.16793786101043223, | |
| "rewards/accuracy_reward": 0.2979166731238365, | |
| "rewards/format_reward": 0.0, | |
| "step": 80 | |
| }, | |
| { | |
| "completion_length": 591.312519454956, | |
| "epoch": 0.05632169627697022, | |
| "grad_norm": 0.2088425343382679, | |
| "kl": 0.05334014892578125, | |
| "learning_rate": 1.1258278145695364e-05, | |
| "loss": 0.0021, | |
| "reward": 0.21458333786576986, | |
| "reward_std": 0.11490485221147537, | |
| "rewards/accuracy_reward": 0.21458333786576986, | |
| "rewards/format_reward": 0.0, | |
| "step": 85 | |
| }, | |
| { | |
| "completion_length": 559.370846748352, | |
| "epoch": 0.05963473723443906, | |
| "grad_norm": 0.2070181739527896, | |
| "kl": 0.05329437255859375, | |
| "learning_rate": 1.1920529801324505e-05, | |
| "loss": 0.0021, | |
| "reward": 0.2020833384245634, | |
| "reward_std": 0.1561527479439974, | |
| "rewards/accuracy_reward": 0.2020833384245634, | |
| "rewards/format_reward": 0.0, | |
| "step": 90 | |
| }, | |
| { | |
| "completion_length": 510.6500156402588, | |
| "epoch": 0.0629477781919079, | |
| "grad_norm": 0.37541642613803017, | |
| "kl": 0.0571197509765625, | |
| "learning_rate": 1.2582781456953644e-05, | |
| "loss": 0.0023, | |
| "reward": 0.2145833382382989, | |
| "reward_std": 0.12079740893095732, | |
| "rewards/accuracy_reward": 0.2145833382382989, | |
| "rewards/format_reward": 0.0, | |
| "step": 95 | |
| }, | |
| { | |
| "completion_length": 525.6958492279052, | |
| "epoch": 0.06626081914937673, | |
| "grad_norm": 0.42588005101800375, | |
| "kl": 0.0678314208984375, | |
| "learning_rate": 1.3245033112582784e-05, | |
| "loss": 0.0027, | |
| "reward": 0.19166667126119136, | |
| "reward_std": 0.12963624373078347, | |
| "rewards/accuracy_reward": 0.19166667126119136, | |
| "rewards/format_reward": 0.0, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.06626081914937673, | |
| "eval_completion_length": 585.5294297162225, | |
| "eval_kl": 0.0954733455882353, | |
| "eval_loss": 0.0038791955448687077, | |
| "eval_reward": 0.1617647111415863, | |
| "eval_reward_std": 0.09012145242270302, | |
| "eval_rewards/accuracy_reward": 0.1617647111415863, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_runtime": 83.6505, | |
| "eval_samples_per_second": 1.183, | |
| "eval_steps_per_second": 0.108, | |
| "step": 100 | |
| }, | |
| { | |
| "completion_length": 605.8437679290771, | |
| "epoch": 0.06957386010684558, | |
| "grad_norm": 0.38186731473973345, | |
| "kl": 0.087103271484375, | |
| "learning_rate": 1.3907284768211921e-05, | |
| "loss": 0.0035, | |
| "reward": 0.17083333767950534, | |
| "reward_std": 0.12374368719756604, | |
| "rewards/accuracy_reward": 0.17083333767950534, | |
| "rewards/format_reward": 0.0, | |
| "step": 105 | |
| }, | |
| { | |
| "completion_length": 612.3166820526124, | |
| "epoch": 0.0728869010643144, | |
| "grad_norm": 0.31009682442787734, | |
| "kl": 0.0819427490234375, | |
| "learning_rate": 1.456953642384106e-05, | |
| "loss": 0.0033, | |
| "reward": 0.17916667107492684, | |
| "reward_std": 0.12963624373078347, | |
| "rewards/accuracy_reward": 0.17916667107492684, | |
| "rewards/format_reward": 0.0, | |
| "step": 110 | |
| }, | |
| { | |
| "completion_length": 588.9187660217285, | |
| "epoch": 0.07619994202178325, | |
| "grad_norm": 0.42260706748908966, | |
| "kl": 0.080755615234375, | |
| "learning_rate": 1.52317880794702e-05, | |
| "loss": 0.0032, | |
| "reward": 0.1541666703298688, | |
| "reward_std": 0.10017346106469631, | |
| "rewards/accuracy_reward": 0.1541666703298688, | |
| "rewards/format_reward": 0.0, | |
| "step": 115 | |
| }, | |
| { | |
| "completion_length": 499.27501373291017, | |
| "epoch": 0.07951298297925208, | |
| "grad_norm": 0.40235696345652067, | |
| "kl": 0.096820068359375, | |
| "learning_rate": 1.589403973509934e-05, | |
| "loss": 0.0039, | |
| "reward": 0.15000000447034836, | |
| "reward_std": 0.08838834799826145, | |
| "rewards/accuracy_reward": 0.15000000447034836, | |
| "rewards/format_reward": 0.0, | |
| "step": 120 | |
| }, | |
| { | |
| "completion_length": 500.2541803359985, | |
| "epoch": 0.08282602393672092, | |
| "grad_norm": 0.3313046600076135, | |
| "kl": 0.1076812744140625, | |
| "learning_rate": 1.6556291390728477e-05, | |
| "loss": 0.0043, | |
| "reward": 0.14166667088866233, | |
| "reward_std": 0.10606601759791374, | |
| "rewards/accuracy_reward": 0.14166667088866233, | |
| "rewards/format_reward": 0.0, | |
| "step": 125 | |
| }, | |
| { | |
| "completion_length": 549.8687664031983, | |
| "epoch": 0.08613906489418975, | |
| "grad_norm": 0.3284282629828314, | |
| "kl": 0.1057586669921875, | |
| "learning_rate": 1.7218543046357617e-05, | |
| "loss": 0.0042, | |
| "reward": 0.14375000391155482, | |
| "reward_std": 0.12668996546417474, | |
| "rewards/accuracy_reward": 0.14375000391155482, | |
| "rewards/format_reward": 0.0, | |
| "step": 130 | |
| }, | |
| { | |
| "completion_length": 554.0104347229004, | |
| "epoch": 0.0894521058516586, | |
| "grad_norm": 0.3273867814905218, | |
| "kl": 0.1071990966796875, | |
| "learning_rate": 1.7880794701986758e-05, | |
| "loss": 0.0043, | |
| "reward": 0.11041666958481074, | |
| "reward_std": 0.09722718279808759, | |
| "rewards/accuracy_reward": 0.11041666958481074, | |
| "rewards/format_reward": 0.0, | |
| "step": 135 | |
| }, | |
| { | |
| "completion_length": 580.7979354858398, | |
| "epoch": 0.09276514680912742, | |
| "grad_norm": 0.3375694940246247, | |
| "kl": 0.1095489501953125, | |
| "learning_rate": 1.8543046357615895e-05, | |
| "loss": 0.0044, | |
| "reward": 0.13125000316649676, | |
| "reward_std": 0.10311973933130503, | |
| "rewards/accuracy_reward": 0.13125000316649676, | |
| "rewards/format_reward": 0.0, | |
| "step": 140 | |
| }, | |
| { | |
| "completion_length": 530.2166820526123, | |
| "epoch": 0.09607818776659627, | |
| "grad_norm": 0.4769828906690383, | |
| "kl": 0.12447509765625, | |
| "learning_rate": 1.9205298013245036e-05, | |
| "loss": 0.005, | |
| "reward": 0.14375000298023224, | |
| "reward_std": 0.12079740893095732, | |
| "rewards/accuracy_reward": 0.14375000298023224, | |
| "rewards/format_reward": 0.0, | |
| "step": 145 | |
| }, | |
| { | |
| "completion_length": 517.9875141143799, | |
| "epoch": 0.0993912287240651, | |
| "grad_norm": 0.3406117455598423, | |
| "kl": 0.1377685546875, | |
| "learning_rate": 1.9867549668874173e-05, | |
| "loss": 0.0055, | |
| "reward": 0.11250000260770321, | |
| "reward_std": 0.10017346087843179, | |
| "rewards/accuracy_reward": 0.11250000260770321, | |
| "rewards/format_reward": 0.0, | |
| "step": 150 | |
| }, | |
| { | |
| "completion_length": 529.8146003723144, | |
| "epoch": 0.10270426968153394, | |
| "grad_norm": 0.4711687133275435, | |
| "kl": 0.138232421875, | |
| "learning_rate": 1.999957185872951e-05, | |
| "loss": 0.0055, | |
| "reward": 0.14791667014360427, | |
| "reward_std": 0.1207974087446928, | |
| "rewards/accuracy_reward": 0.14791667014360427, | |
| "rewards/format_reward": 0.0, | |
| "step": 155 | |
| }, | |
| { | |
| "completion_length": 570.939599609375, | |
| "epoch": 0.10601731063900277, | |
| "grad_norm": 0.25580347999222874, | |
| "kl": 0.1270263671875, | |
| "learning_rate": 1.999783259765003e-05, | |
| "loss": 0.0051, | |
| "reward": 0.12500000298023223, | |
| "reward_std": 0.11195857413113117, | |
| "rewards/accuracy_reward": 0.12500000298023223, | |
| "rewards/format_reward": 0.0, | |
| "step": 160 | |
| }, | |
| { | |
| "completion_length": 494.0750160217285, | |
| "epoch": 0.10933035159647161, | |
| "grad_norm": 0.4334715879020743, | |
| "kl": 0.19808349609375, | |
| "learning_rate": 1.9994755690455154e-05, | |
| "loss": 0.0079, | |
| "reward": 0.12708333618938922, | |
| "reward_std": 0.12079740893095732, | |
| "rewards/accuracy_reward": 0.12708333618938922, | |
| "rewards/format_reward": 0.0, | |
| "step": 165 | |
| }, | |
| { | |
| "completion_length": 465.81459465026853, | |
| "epoch": 0.11264339255394044, | |
| "grad_norm": 0.19728139906086956, | |
| "kl": 0.20181884765625, | |
| "learning_rate": 1.99903415488154e-05, | |
| "loss": 0.0081, | |
| "reward": 0.08541666902601719, | |
| "reward_std": 0.06187184359878302, | |
| "rewards/accuracy_reward": 0.08541666902601719, | |
| "rewards/format_reward": 0.0, | |
| "step": 170 | |
| }, | |
| { | |
| "completion_length": 511.3020980834961, | |
| "epoch": 0.11595643351140929, | |
| "grad_norm": 0.34016272031635464, | |
| "kl": 0.22132568359375, | |
| "learning_rate": 1.9984590763314722e-05, | |
| "loss": 0.0089, | |
| "reward": 0.07083333544433117, | |
| "reward_std": 0.0589255653321743, | |
| "rewards/accuracy_reward": 0.07083333544433117, | |
| "rewards/format_reward": 0.0, | |
| "step": 175 | |
| }, | |
| { | |
| "completion_length": 591.8833505630494, | |
| "epoch": 0.11926947446887812, | |
| "grad_norm": 0.23300547193780718, | |
| "kl": 0.193109130859375, | |
| "learning_rate": 1.997750410337147e-05, | |
| "loss": 0.0077, | |
| "reward": 0.0416666679084301, | |
| "reward_std": 0.047140452265739444, | |
| "rewards/accuracy_reward": 0.0416666679084301, | |
| "rewards/format_reward": 0.0, | |
| "step": 180 | |
| }, | |
| { | |
| "completion_length": 607.9291858673096, | |
| "epoch": 0.12258251542634696, | |
| "grad_norm": 0.23907198670526106, | |
| "kl": 0.180279541015625, | |
| "learning_rate": 1.9969082517135463e-05, | |
| "loss": 0.0072, | |
| "reward": 0.07916666846722364, | |
| "reward_std": 0.06481812186539174, | |
| "rewards/accuracy_reward": 0.07916666846722364, | |
| "rewards/format_reward": 0.0, | |
| "step": 185 | |
| }, | |
| { | |
| "completion_length": 533.2396003723145, | |
| "epoch": 0.1258955563838158, | |
| "grad_norm": 0.32468582498361115, | |
| "kl": 0.1746826171875, | |
| "learning_rate": 1.995932713136112e-05, | |
| "loss": 0.007, | |
| "reward": 0.08125000223517417, | |
| "reward_std": 0.07954951319843531, | |
| "rewards/accuracy_reward": 0.08125000223517417, | |
| "rewards/format_reward": 0.0, | |
| "step": 190 | |
| }, | |
| { | |
| "completion_length": 578.1500179290772, | |
| "epoch": 0.12920859734128462, | |
| "grad_norm": 0.6318480566887794, | |
| "kl": 0.17467041015625, | |
| "learning_rate": 1.994823925125672e-05, | |
| "loss": 0.007, | |
| "reward": 0.10833333637565375, | |
| "reward_std": 0.10606601759791374, | |
| "rewards/accuracy_reward": 0.10833333637565375, | |
| "rewards/format_reward": 0.0, | |
| "step": 195 | |
| }, | |
| { | |
| "completion_length": 532.8479331970215, | |
| "epoch": 0.13252163829875346, | |
| "grad_norm": 0.31162375914822543, | |
| "kl": 0.209014892578125, | |
| "learning_rate": 1.993582036030978e-05, | |
| "loss": 0.0084, | |
| "reward": 0.11041666977107525, | |
| "reward_std": 0.09722718279808759, | |
| "rewards/accuracy_reward": 0.11041666977107525, | |
| "rewards/format_reward": 0.0, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.13252163829875346, | |
| "eval_completion_length": 461.04903097713697, | |
| "eval_kl": 0.22242647058823528, | |
| "eval_loss": 0.009009506553411484, | |
| "eval_reward": 0.1274509846287615, | |
| "eval_reward_std": 0.08318903253358953, | |
| "eval_rewards/accuracy_reward": 0.1274509846287615, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_runtime": 64.4615, | |
| "eval_samples_per_second": 1.536, | |
| "eval_steps_per_second": 0.14, | |
| "step": 200 | |
| }, | |
| { | |
| "completion_length": 461.4708471298218, | |
| "epoch": 0.1358346792562223, | |
| "grad_norm": 0.941704833763444, | |
| "kl": 0.1978759765625, | |
| "learning_rate": 1.9922072120088537e-05, | |
| "loss": 0.0079, | |
| "reward": 0.08125000242143869, | |
| "reward_std": 0.07954951319843531, | |
| "rewards/accuracy_reward": 0.08125000242143869, | |
| "rewards/format_reward": 0.0, | |
| "step": 205 | |
| }, | |
| { | |
| "completion_length": 525.7250125885009, | |
| "epoch": 0.13914772021369115, | |
| "grad_norm": 0.429496265642979, | |
| "kl": 0.1792236328125, | |
| "learning_rate": 1.9906996370019692e-05, | |
| "loss": 0.0072, | |
| "reward": 0.0916666692122817, | |
| "reward_std": 0.09428090453147889, | |
| "rewards/accuracy_reward": 0.0916666692122817, | |
| "rewards/format_reward": 0.0, | |
| "step": 210 | |
| }, | |
| { | |
| "completion_length": 566.6875190734863, | |
| "epoch": 0.14246076117115997, | |
| "grad_norm": 1.1240583407373201, | |
| "kl": 15.97672119140625, | |
| "learning_rate": 1.989059512714227e-05, | |
| "loss": 0.6397, | |
| "reward": 0.0916666690260172, | |
| "reward_std": 0.08838834781199693, | |
| "rewards/accuracy_reward": 0.0916666690260172, | |
| "rewards/format_reward": 0.0, | |
| "step": 215 | |
| }, | |
| { | |
| "completion_length": 542.343769454956, | |
| "epoch": 0.1457738021286288, | |
| "grad_norm": 0.3232677681544197, | |
| "kl": 0.189068603515625, | |
| "learning_rate": 1.9872870585837757e-05, | |
| "loss": 0.0076, | |
| "reward": 0.12500000298023223, | |
| "reward_std": 0.10606601759791374, | |
| "rewards/accuracy_reward": 0.12500000298023223, | |
| "rewards/format_reward": 0.0, | |
| "step": 220 | |
| }, | |
| { | |
| "completion_length": 546.4791807174682, | |
| "epoch": 0.14908684308609765, | |
| "grad_norm": 0.8875397199463153, | |
| "kl": 0.188201904296875, | |
| "learning_rate": 1.9853825117536522e-05, | |
| "loss": 0.0075, | |
| "reward": 0.13333333637565375, | |
| "reward_std": 0.0766032349318266, | |
| "rewards/accuracy_reward": 0.13333333637565375, | |
| "rewards/format_reward": 0.0, | |
| "step": 225 | |
| }, | |
| { | |
| "completion_length": 540.5750137329102, | |
| "epoch": 0.1523998840435665, | |
| "grad_norm": 0.42221258320463867, | |
| "kl": 0.19306640625, | |
| "learning_rate": 1.983346127040053e-05, | |
| "loss": 0.0077, | |
| "reward": 0.12083333656191826, | |
| "reward_std": 0.12374368719756604, | |
| "rewards/accuracy_reward": 0.12083333656191826, | |
| "rewards/format_reward": 0.0, | |
| "step": 230 | |
| }, | |
| { | |
| "completion_length": 448.9229295730591, | |
| "epoch": 0.1557129250010353, | |
| "grad_norm": 0.3606831391556023, | |
| "kl": 0.24759521484375, | |
| "learning_rate": 1.9811781768982392e-05, | |
| "loss": 0.0099, | |
| "reward": 0.08125000223517417, | |
| "reward_std": 0.10311973914504051, | |
| "rewards/accuracy_reward": 0.08125000223517417, | |
| "rewards/format_reward": 0.0, | |
| "step": 235 | |
| }, | |
| { | |
| "completion_length": 479.3083478927612, | |
| "epoch": 0.15902596595850416, | |
| "grad_norm": 0.519351205279088, | |
| "kl": 0.19686279296875, | |
| "learning_rate": 1.9788789513860875e-05, | |
| "loss": 0.0079, | |
| "reward": 0.10000000279396773, | |
| "reward_std": 0.09428090453147889, | |
| "rewards/accuracy_reward": 0.10000000279396773, | |
| "rewards/format_reward": 0.0, | |
| "step": 240 | |
| }, | |
| { | |
| "completion_length": 543.2770980834961, | |
| "epoch": 0.162339006915973, | |
| "grad_norm": 0.2739435166284146, | |
| "kl": 0.192205810546875, | |
| "learning_rate": 1.9764487581252787e-05, | |
| "loss": 0.0077, | |
| "reward": 0.10416666883975267, | |
| "reward_std": 0.10017346087843179, | |
| "rewards/accuracy_reward": 0.10416666883975267, | |
| "rewards/format_reward": 0.0, | |
| "step": 245 | |
| }, | |
| { | |
| "completion_length": 604.7562683105468, | |
| "epoch": 0.16565204787344184, | |
| "grad_norm": 0.22639850925377564, | |
| "kl": 0.18382568359375, | |
| "learning_rate": 1.9738879222601425e-05, | |
| "loss": 0.0074, | |
| "reward": 0.08750000204890966, | |
| "reward_std": 0.08249579146504402, | |
| "rewards/accuracy_reward": 0.08750000204890966, | |
| "rewards/format_reward": 0.0, | |
| "step": 250 | |
| }, | |
| { | |
| "completion_length": 508.62084617614744, | |
| "epoch": 0.16896508883091066, | |
| "grad_norm": 0.34102274429245655, | |
| "kl": 26.4016845703125, | |
| "learning_rate": 1.9711967864141542e-05, | |
| "loss": 1.0641, | |
| "reward": 0.12083333600312471, | |
| "reward_std": 0.11195857394486666, | |
| "rewards/accuracy_reward": 0.12083333600312471, | |
| "rewards/format_reward": 0.0, | |
| "step": 255 | |
| }, | |
| { | |
| "completion_length": 479.06876335144045, | |
| "epoch": 0.1722781297883795, | |
| "grad_norm": 0.3926595209195284, | |
| "kl": 0.193994140625, | |
| "learning_rate": 1.968375710644093e-05, | |
| "loss": 0.0078, | |
| "reward": 0.11875000279396772, | |
| "reward_std": 0.1207974087446928, | |
| "rewards/accuracy_reward": 0.11875000279396772, | |
| "rewards/format_reward": 0.0, | |
| "step": 260 | |
| }, | |
| { | |
| "completion_length": 587.6750179290772, | |
| "epoch": 0.17559117074584835, | |
| "grad_norm": 0.24108678658537444, | |
| "kl": 0.18203125, | |
| "learning_rate": 1.9654250723918706e-05, | |
| "loss": 0.0073, | |
| "reward": 0.1104166692122817, | |
| "reward_std": 0.08544206954538822, | |
| "rewards/accuracy_reward": 0.1104166692122817, | |
| "rewards/format_reward": 0.0, | |
| "step": 265 | |
| }, | |
| { | |
| "completion_length": 677.4437683105468, | |
| "epoch": 0.1789042117033172, | |
| "grad_norm": 0.18106832646177998, | |
| "kl": 0.18656005859375, | |
| "learning_rate": 1.9623452664340305e-05, | |
| "loss": 0.0075, | |
| "reward": 0.07708333525806665, | |
| "reward_std": 0.05597928706556558, | |
| "rewards/accuracy_reward": 0.07708333525806665, | |
| "rewards/format_reward": 0.0, | |
| "step": 270 | |
| }, | |
| { | |
| "completion_length": 580.7875162124634, | |
| "epoch": 0.182217252660786, | |
| "grad_norm": 0.3207457026346929, | |
| "kl": 0.19422607421875, | |
| "learning_rate": 1.9591367048289297e-05, | |
| "loss": 0.0078, | |
| "reward": 0.07916666883975268, | |
| "reward_std": 0.07071067839860916, | |
| "rewards/accuracy_reward": 0.07916666883975268, | |
| "rewards/format_reward": 0.0, | |
| "step": 275 | |
| }, | |
| { | |
| "completion_length": 529.9458477020264, | |
| "epoch": 0.18553029361825485, | |
| "grad_norm": 0.3268544056694053, | |
| "kl": 0.183233642578125, | |
| "learning_rate": 1.9557998168616087e-05, | |
| "loss": 0.0073, | |
| "reward": 0.11875000316649675, | |
| "reward_std": 0.09133462626487017, | |
| "rewards/accuracy_reward": 0.11875000316649675, | |
| "rewards/format_reward": 0.0, | |
| "step": 280 | |
| }, | |
| { | |
| "completion_length": 465.0104305267334, | |
| "epoch": 0.1888433345757237, | |
| "grad_norm": 0.48827428252889743, | |
| "kl": 0.17584228515625, | |
| "learning_rate": 1.9523350489863545e-05, | |
| "loss": 0.007, | |
| "reward": 0.1437500037252903, | |
| "reward_std": 0.10901229586452246, | |
| "rewards/accuracy_reward": 0.1437500037252903, | |
| "rewards/format_reward": 0.0, | |
| "step": 285 | |
| }, | |
| { | |
| "completion_length": 444.8500112533569, | |
| "epoch": 0.19215637553319254, | |
| "grad_norm": 0.4683984486186781, | |
| "kl": 0.18759765625, | |
| "learning_rate": 1.9487428647669688e-05, | |
| "loss": 0.0075, | |
| "reward": 0.11458333637565374, | |
| "reward_std": 0.09722718279808759, | |
| "rewards/accuracy_reward": 0.11458333637565374, | |
| "rewards/format_reward": 0.0, | |
| "step": 290 | |
| }, | |
| { | |
| "completion_length": 500.3625148773193, | |
| "epoch": 0.19546941649066138, | |
| "grad_norm": 0.1814212886292743, | |
| "kl": 0.18865966796875, | |
| "learning_rate": 1.9450237448147463e-05, | |
| "loss": 0.0076, | |
| "reward": 0.0958333358168602, | |
| "reward_std": 0.08249579146504402, | |
| "rewards/accuracy_reward": 0.0958333358168602, | |
| "rewards/format_reward": 0.0, | |
| "step": 295 | |
| }, | |
| { | |
| "completion_length": 566.7708499908447, | |
| "epoch": 0.1987824574481302, | |
| "grad_norm": 0.24448480046541934, | |
| "kl": 0.201904296875, | |
| "learning_rate": 1.9411781867241718e-05, | |
| "loss": 0.0081, | |
| "reward": 0.0916666692122817, | |
| "reward_std": 0.09428090453147889, | |
| "rewards/accuracy_reward": 0.0916666692122817, | |
| "rewards/format_reward": 0.0, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.1987824574481302, | |
| "eval_completion_length": 547.3529555376838, | |
| "eval_kl": 0.18393841911764705, | |
| "eval_loss": 0.007359905168414116, | |
| "eval_reward": 0.112745100961012, | |
| "eval_reward_std": 0.10398629176266053, | |
| "eval_rewards/accuracy_reward": 0.112745100961012, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_runtime": 77.7803, | |
| "eval_samples_per_second": 1.273, | |
| "eval_steps_per_second": 0.116, | |
| "step": 300 | |
| }, | |
| { | |
| "completion_length": 530.7312660217285, | |
| "epoch": 0.20209549840559904, | |
| "grad_norm": 0.20881065313746844, | |
| "kl": 0.196630859375, | |
| "learning_rate": 1.937206705006344e-05, | |
| "loss": 0.0079, | |
| "reward": 0.12708333600312471, | |
| "reward_std": 0.07954951319843531, | |
| "rewards/accuracy_reward": 0.12708333600312471, | |
| "rewards/format_reward": 0.0, | |
| "step": 305 | |
| }, | |
| { | |
| "completion_length": 467.48334827423093, | |
| "epoch": 0.20540853936306788, | |
| "grad_norm": 0.2581604337535273, | |
| "kl": 0.188153076171875, | |
| "learning_rate": 1.9331098310201392e-05, | |
| "loss": 0.0075, | |
| "reward": 0.10416666921228171, | |
| "reward_std": 0.10017346087843179, | |
| "rewards/accuracy_reward": 0.10416666921228171, | |
| "rewards/format_reward": 0.0, | |
| "step": 310 | |
| }, | |
| { | |
| "completion_length": 550.3041828155517, | |
| "epoch": 0.20872158032053673, | |
| "grad_norm": 0.3018033078636091, | |
| "kl": 0.183837890625, | |
| "learning_rate": 1.9288881129011177e-05, | |
| "loss": 0.0074, | |
| "reward": 0.12083333656191826, | |
| "reward_std": 0.12374368719756604, | |
| "rewards/accuracy_reward": 0.12083333656191826, | |
| "rewards/format_reward": 0.0, | |
| "step": 315 | |
| }, | |
| { | |
| "completion_length": 554.0333499908447, | |
| "epoch": 0.21203462127800554, | |
| "grad_norm": 0.37948238789004246, | |
| "kl": 0.2573486328125, | |
| "learning_rate": 1.9245421154881873e-05, | |
| "loss": 0.0103, | |
| "reward": 0.10416666958481073, | |
| "reward_std": 0.10606601759791374, | |
| "rewards/accuracy_reward": 0.10416666958481073, | |
| "rewards/format_reward": 0.0, | |
| "step": 320 | |
| }, | |
| { | |
| "completion_length": 495.84793186187744, | |
| "epoch": 0.21534766223547439, | |
| "grad_norm": 0.3507554355169013, | |
| "kl": 0.20789794921875, | |
| "learning_rate": 1.9200724202480305e-05, | |
| "loss": 0.0083, | |
| "reward": 0.08125000223517417, | |
| "reward_std": 0.08544206973165273, | |
| "rewards/accuracy_reward": 0.08125000223517417, | |
| "rewards/format_reward": 0.0, | |
| "step": 325 | |
| }, | |
| { | |
| "completion_length": 439.99584617614744, | |
| "epoch": 0.21866070319294323, | |
| "grad_norm": 0.33415650171071337, | |
| "kl": 0.20986328125, | |
| "learning_rate": 1.9154796251973092e-05, | |
| "loss": 0.0084, | |
| "reward": 0.11250000316649675, | |
| "reward_std": 0.08249579146504402, | |
| "rewards/accuracy_reward": 0.11250000316649675, | |
| "rewards/format_reward": 0.0, | |
| "step": 330 | |
| }, | |
| { | |
| "completion_length": 502.4770965576172, | |
| "epoch": 0.22197374415041207, | |
| "grad_norm": 0.29077054623985626, | |
| "kl": 0.2102783203125, | |
| "learning_rate": 1.9107643448226536e-05, | |
| "loss": 0.0084, | |
| "reward": 0.12291667032986879, | |
| "reward_std": 0.09722718279808759, | |
| "rewards/accuracy_reward": 0.12291667032986879, | |
| "rewards/format_reward": 0.0, | |
| "step": 335 | |
| }, | |
| { | |
| "completion_length": 492.98126487731935, | |
| "epoch": 0.2252867851078809, | |
| "grad_norm": 0.32276994073724186, | |
| "kl": 0.2305908203125, | |
| "learning_rate": 1.905927209998447e-05, | |
| "loss": 0.0092, | |
| "reward": 0.13750000335276127, | |
| "reward_std": 0.10606601759791374, | |
| "rewards/accuracy_reward": 0.13750000335276127, | |
| "rewards/format_reward": 0.0, | |
| "step": 340 | |
| }, | |
| { | |
| "completion_length": 496.520845413208, | |
| "epoch": 0.22859982606534973, | |
| "grad_norm": 0.3397890414322317, | |
| "kl": 0.20244140625, | |
| "learning_rate": 1.900968867902419e-05, | |
| "loss": 0.0081, | |
| "reward": 0.1000000024214387, | |
| "reward_std": 0.10017346106469631, | |
| "rewards/accuracy_reward": 0.1000000024214387, | |
| "rewards/format_reward": 0.0, | |
| "step": 345 | |
| }, | |
| { | |
| "completion_length": 539.0625156402588, | |
| "epoch": 0.23191286702281858, | |
| "grad_norm": 0.34646557191148036, | |
| "kl": 0.18717041015625, | |
| "learning_rate": 1.8958899819290592e-05, | |
| "loss": 0.0075, | |
| "reward": 0.1020833358168602, | |
| "reward_std": 0.08544206973165273, | |
| "rewards/accuracy_reward": 0.1020833358168602, | |
| "rewards/format_reward": 0.0, | |
| "step": 350 | |
| }, | |
| { | |
| "completion_length": 534.3270980834961, | |
| "epoch": 0.23522590798028742, | |
| "grad_norm": 0.306741354263696, | |
| "kl": 0.2046142578125, | |
| "learning_rate": 1.890691231600856e-05, | |
| "loss": 0.0082, | |
| "reward": 0.0916666692122817, | |
| "reward_std": 0.09428090434521437, | |
| "rewards/accuracy_reward": 0.0916666692122817, | |
| "rewards/format_reward": 0.0, | |
| "step": 355 | |
| }, | |
| { | |
| "completion_length": 551.9895969390869, | |
| "epoch": 0.23853894893775623, | |
| "grad_norm": 0.293113566609471, | |
| "kl": 0.2113525390625, | |
| "learning_rate": 1.8853733124773837e-05, | |
| "loss": 0.0085, | |
| "reward": 0.12083333656191826, | |
| "reward_std": 0.10017346106469631, | |
| "rewards/accuracy_reward": 0.12083333656191826, | |
| "rewards/format_reward": 0.0, | |
| "step": 360 | |
| }, | |
| { | |
| "completion_length": 487.33751411437987, | |
| "epoch": 0.24185198989522508, | |
| "grad_norm": 0.3636802453570497, | |
| "kl": 0.20135498046875, | |
| "learning_rate": 1.8799369360622394e-05, | |
| "loss": 0.0081, | |
| "reward": 0.13750000316649674, | |
| "reward_std": 0.10606601722538471, | |
| "rewards/accuracy_reward": 0.13750000316649674, | |
| "rewards/format_reward": 0.0, | |
| "step": 365 | |
| }, | |
| { | |
| "completion_length": 432.9791788101196, | |
| "epoch": 0.24516503085269392, | |
| "grad_norm": 0.7560117165844005, | |
| "kl": 0.20098876953125, | |
| "learning_rate": 1.8743828297078485e-05, | |
| "loss": 0.008, | |
| "reward": 0.1062500026077032, | |
| "reward_std": 0.09133462626487017, | |
| "rewards/accuracy_reward": 0.1062500026077032, | |
| "rewards/format_reward": 0.0, | |
| "step": 370 | |
| }, | |
| { | |
| "completion_length": 461.3520961761475, | |
| "epoch": 0.24847807181016277, | |
| "grad_norm": 0.24674434059917383, | |
| "kl": 0.2183349609375, | |
| "learning_rate": 1.8687117365181514e-05, | |
| "loss": 0.0087, | |
| "reward": 0.12916666995733977, | |
| "reward_std": 0.10606601759791374, | |
| "rewards/accuracy_reward": 0.12916666995733977, | |
| "rewards/format_reward": 0.0, | |
| "step": 375 | |
| }, | |
| { | |
| "completion_length": 571.5291835784913, | |
| "epoch": 0.2517911127676316, | |
| "grad_norm": 0.25861125382341793, | |
| "kl": 0.2226318359375, | |
| "learning_rate": 1.8629244152491773e-05, | |
| "loss": 0.0089, | |
| "reward": 0.07291666846722364, | |
| "reward_std": 0.06187184359878302, | |
| "rewards/accuracy_reward": 0.07291666846722364, | |
| "rewards/format_reward": 0.0, | |
| "step": 380 | |
| }, | |
| { | |
| "completion_length": 584.9791839599609, | |
| "epoch": 0.2551041537251004, | |
| "grad_norm": 0.37996774263610955, | |
| "kl": 0.20263671875, | |
| "learning_rate": 1.8570216402075326e-05, | |
| "loss": 0.0081, | |
| "reward": 0.10416666958481073, | |
| "reward_std": 0.10606601759791374, | |
| "rewards/accuracy_reward": 0.10416666958481073, | |
| "rewards/format_reward": 0.0, | |
| "step": 385 | |
| }, | |
| { | |
| "completion_length": 480.7541809082031, | |
| "epoch": 0.25841719468256924, | |
| "grad_norm": 0.4807126789722509, | |
| "kl": 0.2365234375, | |
| "learning_rate": 1.8510042011467978e-05, | |
| "loss": 0.0095, | |
| "reward": 0.0979166692122817, | |
| "reward_std": 0.09722718279808759, | |
| "rewards/accuracy_reward": 0.0979166692122817, | |
| "rewards/format_reward": 0.0, | |
| "step": 390 | |
| }, | |
| { | |
| "completion_length": 464.12292823791506, | |
| "epoch": 0.2617302356400381, | |
| "grad_norm": 0.2806539928587211, | |
| "kl": 0.22197265625, | |
| "learning_rate": 1.8448729031618687e-05, | |
| "loss": 0.0089, | |
| "reward": 0.07083333525806665, | |
| "reward_std": 0.07071067839860916, | |
| "rewards/accuracy_reward": 0.07083333525806665, | |
| "rewards/format_reward": 0.0, | |
| "step": 395 | |
| }, | |
| { | |
| "completion_length": 521.7875158309937, | |
| "epoch": 0.2650432765975069, | |
| "grad_norm": 0.11840433673859241, | |
| "kl": 0.1973876953125, | |
| "learning_rate": 1.838628566581236e-05, | |
| "loss": 0.0079, | |
| "reward": 0.08333333544433116, | |
| "reward_std": 0.0766032349318266, | |
| "rewards/accuracy_reward": 0.08333333544433116, | |
| "rewards/format_reward": 0.0, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.2650432765975069, | |
| "eval_completion_length": 560.7108028636259, | |
| "eval_kl": 0.17957261029411764, | |
| "eval_loss": 0.007205521687865257, | |
| "eval_reward": 0.07352941307951422, | |
| "eval_reward_std": 0.06239177505759632, | |
| "eval_rewards/accuracy_reward": 0.07352941307951422, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_runtime": 83.1386, | |
| "eval_samples_per_second": 1.191, | |
| "eval_steps_per_second": 0.108, | |
| "step": 400 | |
| }, | |
| { | |
| "completion_length": 595.2062665939332, | |
| "epoch": 0.2683563175549758, | |
| "grad_norm": 0.33533763786101467, | |
| "kl": 0.1843994140625, | |
| "learning_rate": 1.8322720268572333e-05, | |
| "loss": 0.0074, | |
| "reward": 0.09375000242143869, | |
| "reward_std": 0.07365695666521788, | |
| "rewards/accuracy_reward": 0.09375000242143869, | |
| "rewards/format_reward": 0.0, | |
| "step": 405 | |
| }, | |
| { | |
| "completion_length": 549.1041831970215, | |
| "epoch": 0.2716693585124446, | |
| "grad_norm": 0.503465226947629, | |
| "kl": 0.19075927734375, | |
| "learning_rate": 1.8258041344542567e-05, | |
| "loss": 0.0076, | |
| "reward": 0.0895833358168602, | |
| "reward_std": 0.07954951319843531, | |
| "rewards/accuracy_reward": 0.0895833358168602, | |
| "rewards/format_reward": 0.0, | |
| "step": 410 | |
| }, | |
| { | |
| "completion_length": 539.4666854858399, | |
| "epoch": 0.27498239946991343, | |
| "grad_norm": 0.2542908863882778, | |
| "kl": 0.200390625, | |
| "learning_rate": 1.8192257547349805e-05, | |
| "loss": 0.008, | |
| "reward": 0.08541666883975267, | |
| "reward_std": 0.07365695666521788, | |
| "rewards/accuracy_reward": 0.08541666883975267, | |
| "rewards/format_reward": 0.0, | |
| "step": 415 | |
| }, | |
| { | |
| "completion_length": 490.5791805267334, | |
| "epoch": 0.2782954404273823, | |
| "grad_norm": 0.3338735472126353, | |
| "kl": 0.19786376953125, | |
| "learning_rate": 1.8125377678445755e-05, | |
| "loss": 0.0079, | |
| "reward": 0.08958333563059569, | |
| "reward_std": 0.07954951319843531, | |
| "rewards/accuracy_reward": 0.08958333563059569, | |
| "rewards/format_reward": 0.0, | |
| "step": 420 | |
| }, | |
| { | |
| "completion_length": 442.22501316070554, | |
| "epoch": 0.2816084813848511, | |
| "grad_norm": 0.31309794250892437, | |
| "kl": 0.1982177734375, | |
| "learning_rate": 1.8057410685929505e-05, | |
| "loss": 0.0079, | |
| "reward": 0.09375000223517418, | |
| "reward_std": 0.08544206954538822, | |
| "rewards/accuracy_reward": 0.09375000223517418, | |
| "rewards/format_reward": 0.0, | |
| "step": 425 | |
| }, | |
| { | |
| "completion_length": 610.3291862487793, | |
| "epoch": 0.28492152234231993, | |
| "grad_norm": 0.18677755115855318, | |
| "kl": 0.18875732421875, | |
| "learning_rate": 1.7988365663350352e-05, | |
| "loss": 0.0076, | |
| "reward": 0.09375000279396772, | |
| "reward_std": 0.07954951319843531, | |
| "rewards/accuracy_reward": 0.09375000279396772, | |
| "rewards/format_reward": 0.0, | |
| "step": 430 | |
| }, | |
| { | |
| "completion_length": 618.6645980834961, | |
| "epoch": 0.2882345632997888, | |
| "grad_norm": 0.4382724584840247, | |
| "kl": 0.201416015625, | |
| "learning_rate": 1.7918251848491118e-05, | |
| "loss": 0.0081, | |
| "reward": 0.09375000279396772, | |
| "reward_std": 0.07954951319843531, | |
| "rewards/accuracy_reward": 0.09375000279396772, | |
| "rewards/format_reward": 0.0, | |
| "step": 435 | |
| }, | |
| { | |
| "completion_length": 509.11668395996094, | |
| "epoch": 0.2915476042572576, | |
| "grad_norm": 0.741790732534879, | |
| "kl": 0.22264404296875, | |
| "learning_rate": 1.7847078622132202e-05, | |
| "loss": 0.0089, | |
| "reward": 0.11666667014360428, | |
| "reward_std": 0.1178511306643486, | |
| "rewards/accuracy_reward": 0.11666667014360428, | |
| "rewards/format_reward": 0.0, | |
| "step": 440 | |
| }, | |
| { | |
| "completion_length": 455.8645965576172, | |
| "epoch": 0.2948606452147265, | |
| "grad_norm": 0.22712749456065462, | |
| "kl": 0.2342041015625, | |
| "learning_rate": 1.7774855506796497e-05, | |
| "loss": 0.0094, | |
| "reward": 0.08750000223517418, | |
| "reward_std": 0.07071067839860916, | |
| "rewards/accuracy_reward": 0.08750000223517418, | |
| "rewards/format_reward": 0.0, | |
| "step": 445 | |
| }, | |
| { | |
| "completion_length": 508.2083498001099, | |
| "epoch": 0.2981736861721953, | |
| "grad_norm": 0.2862921365130213, | |
| "kl": 0.19876708984375, | |
| "learning_rate": 1.770159216547532e-05, | |
| "loss": 0.008, | |
| "reward": 0.10833333637565375, | |
| "reward_std": 0.10017346106469631, | |
| "rewards/accuracy_reward": 0.10833333637565375, | |
| "rewards/format_reward": 0.0, | |
| "step": 450 | |
| }, | |
| { | |
| "completion_length": 610.0416835784912, | |
| "epoch": 0.3014867271296641, | |
| "grad_norm": 0.20694738981947886, | |
| "kl": 0.18927001953125, | |
| "learning_rate": 1.76272984003356e-05, | |
| "loss": 0.0076, | |
| "reward": 0.09583333600312471, | |
| "reward_std": 0.08249579146504402, | |
| "rewards/accuracy_reward": 0.09583333600312471, | |
| "rewards/format_reward": 0.0, | |
| "step": 455 | |
| }, | |
| { | |
| "completion_length": 579.829182434082, | |
| "epoch": 0.304799768087133, | |
| "grad_norm": 0.27667252528639497, | |
| "kl": 0.203564453125, | |
| "learning_rate": 1.7551984151408363e-05, | |
| "loss": 0.0081, | |
| "reward": 0.13125000335276127, | |
| "reward_std": 0.1207974087446928, | |
| "rewards/accuracy_reward": 0.13125000335276127, | |
| "rewards/format_reward": 0.0, | |
| "step": 460 | |
| }, | |
| { | |
| "completion_length": 590.5333518981934, | |
| "epoch": 0.3081128090446018, | |
| "grad_norm": 0.2881508370061625, | |
| "kl": 0.22376708984375, | |
| "learning_rate": 1.7475659495258864e-05, | |
| "loss": 0.0089, | |
| "reward": 0.10416666939854621, | |
| "reward_std": 0.08838834799826145, | |
| "rewards/accuracy_reward": 0.10416666939854621, | |
| "rewards/format_reward": 0.0, | |
| "step": 465 | |
| }, | |
| { | |
| "completion_length": 533.4229331970215, | |
| "epoch": 0.3114258500020706, | |
| "grad_norm": 0.2567790764625792, | |
| "kl": 0.211279296875, | |
| "learning_rate": 1.739833464363838e-05, | |
| "loss": 0.0085, | |
| "reward": 0.11875000298023224, | |
| "reward_std": 0.10311973933130503, | |
| "rewards/accuracy_reward": 0.11875000298023224, | |
| "rewards/format_reward": 0.0, | |
| "step": 470 | |
| }, | |
| { | |
| "completion_length": 534.5958490371704, | |
| "epoch": 0.3147388909595395, | |
| "grad_norm": 0.28822211835842837, | |
| "kl": 0.21998291015625, | |
| "learning_rate": 1.7320019942117954e-05, | |
| "loss": 0.0088, | |
| "reward": 0.11666666995733976, | |
| "reward_std": 0.09428090453147889, | |
| "rewards/accuracy_reward": 0.11666666995733976, | |
| "rewards/format_reward": 0.0, | |
| "step": 475 | |
| }, | |
| { | |
| "completion_length": 502.02084770202634, | |
| "epoch": 0.3180519319170083, | |
| "grad_norm": 0.33502620138635836, | |
| "kl": 0.2275390625, | |
| "learning_rate": 1.7240725868704218e-05, | |
| "loss": 0.0091, | |
| "reward": 0.08541666902601719, | |
| "reward_std": 0.07365695666521788, | |
| "rewards/accuracy_reward": 0.08541666902601719, | |
| "rewards/format_reward": 0.0, | |
| "step": 480 | |
| }, | |
| { | |
| "completion_length": 524.7958484649658, | |
| "epoch": 0.3213649728744772, | |
| "grad_norm": 0.3472442436711827, | |
| "kl": 0.223516845703125, | |
| "learning_rate": 1.71604630324375e-05, | |
| "loss": 0.0089, | |
| "reward": 0.08958333563059569, | |
| "reward_std": 0.08544206973165273, | |
| "rewards/accuracy_reward": 0.08958333563059569, | |
| "rewards/format_reward": 0.0, | |
| "step": 485 | |
| }, | |
| { | |
| "completion_length": 524.3166828155518, | |
| "epoch": 0.324678013831946, | |
| "grad_norm": 0.3263087112920367, | |
| "kl": 0.22652587890625, | |
| "learning_rate": 1.7079242171972417e-05, | |
| "loss": 0.0091, | |
| "reward": 0.10416666921228171, | |
| "reward_std": 0.09428090453147889, | |
| "rewards/accuracy_reward": 0.10416666921228171, | |
| "rewards/format_reward": 0.0, | |
| "step": 490 | |
| }, | |
| { | |
| "completion_length": 571.7645992279053, | |
| "epoch": 0.3279910547894148, | |
| "grad_norm": 0.28845204101541877, | |
| "kl": 0.20472412109375, | |
| "learning_rate": 1.6997074154141097e-05, | |
| "loss": 0.0082, | |
| "reward": 0.10833333600312471, | |
| "reward_std": 0.08838834799826145, | |
| "rewards/accuracy_reward": 0.10833333600312471, | |
| "rewards/format_reward": 0.0, | |
| "step": 495 | |
| }, | |
| { | |
| "completion_length": 551.5854331970215, | |
| "epoch": 0.3313040957468837, | |
| "grad_norm": 0.2761223611585464, | |
| "kl": 0.1948486328125, | |
| "learning_rate": 1.6913969972499272e-05, | |
| "loss": 0.0078, | |
| "reward": 0.12291666977107525, | |
| "reward_std": 0.10311973914504051, | |
| "rewards/accuracy_reward": 0.12291666977107525, | |
| "rewards/format_reward": 0.0, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.3313040957468837, | |
| "eval_completion_length": 505.10295553768384, | |
| "eval_kl": 0.19996553308823528, | |
| "eval_loss": 0.008075407706201077, | |
| "eval_reward": 0.13725490517476024, | |
| "eval_reward_std": 0.1386483881403418, | |
| "eval_rewards/accuracy_reward": 0.13725490517476024, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_runtime": 76.805, | |
| "eval_samples_per_second": 1.289, | |
| "eval_steps_per_second": 0.117, | |
| "step": 500 | |
| }, | |
| { | |
| "completion_length": 499.3750162124634, | |
| "epoch": 0.3346171367043525, | |
| "grad_norm": 0.3713745412028642, | |
| "kl": 0.19395751953125, | |
| "learning_rate": 1.682994074585541e-05, | |
| "loss": 0.0078, | |
| "reward": 0.13750000298023224, | |
| "reward_std": 0.10606601741164923, | |
| "rewards/accuracy_reward": 0.13750000298023224, | |
| "rewards/format_reward": 0.0, | |
| "step": 505 | |
| }, | |
| { | |
| "completion_length": 504.960432434082, | |
| "epoch": 0.3379301776618213, | |
| "grad_norm": 0.2985183472381298, | |
| "kl": 0.208966064453125, | |
| "learning_rate": 1.674499771678309e-05, | |
| "loss": 0.0084, | |
| "reward": 0.11041666958481074, | |
| "reward_std": 0.06776440013200044, | |
| "rewards/accuracy_reward": 0.11041666958481074, | |
| "rewards/format_reward": 0.0, | |
| "step": 510 | |
| }, | |
| { | |
| "completion_length": 534.1958522796631, | |
| "epoch": 0.3412432186192902, | |
| "grad_norm": 2600.854085321423, | |
| "kl": 52.59376220703125, | |
| "learning_rate": 1.665915225011681e-05, | |
| "loss": 2.0976, | |
| "reward": 0.11458333600312472, | |
| "reward_std": 0.10901229586452246, | |
| "rewards/accuracy_reward": 0.11458333600312472, | |
| "rewards/format_reward": 0.0, | |
| "step": 515 | |
| }, | |
| { | |
| "completion_length": 599.9729377746582, | |
| "epoch": 0.344556259576759, | |
| "grad_norm": 0.2386666186123784, | |
| "kl": 0.892706298828125, | |
| "learning_rate": 1.6572415831431466e-05, | |
| "loss": 0.0358, | |
| "reward": 0.1062500026077032, | |
| "reward_std": 0.10311973933130503, | |
| "rewards/accuracy_reward": 0.1062500026077032, | |
| "rewards/format_reward": 0.0, | |
| "step": 520 | |
| }, | |
| { | |
| "completion_length": 546.0208488464356, | |
| "epoch": 0.3478693005342279, | |
| "grad_norm": 0.34730751193742443, | |
| "kl": 0.313232421875, | |
| "learning_rate": 1.6484800065505627e-05, | |
| "loss": 0.0125, | |
| "reward": 0.07708333563059569, | |
| "reward_std": 0.07954951319843531, | |
| "rewards/accuracy_reward": 0.07708333563059569, | |
| "rewards/format_reward": 0.0, | |
| "step": 525 | |
| }, | |
| { | |
| "completion_length": 464.66667861938475, | |
| "epoch": 0.3511823414916967, | |
| "grad_norm": 0.39701952165391247, | |
| "kl": 0.2570068359375, | |
| "learning_rate": 1.6396316674768914e-05, | |
| "loss": 0.0103, | |
| "reward": 0.10208333600312472, | |
| "reward_std": 0.10311973933130503, | |
| "rewards/accuracy_reward": 0.10208333600312472, | |
| "rewards/format_reward": 0.0, | |
| "step": 530 | |
| }, | |
| { | |
| "completion_length": 522.2875156402588, | |
| "epoch": 0.3544953824491655, | |
| "grad_norm": 0.28714956052647667, | |
| "kl": 0.222015380859375, | |
| "learning_rate": 1.630697749773359e-05, | |
| "loss": 0.0089, | |
| "reward": 0.12291666958481073, | |
| "reward_std": 0.12079740893095732, | |
| "rewards/accuracy_reward": 0.12291666958481073, | |
| "rewards/format_reward": 0.0, | |
| "step": 535 | |
| }, | |
| { | |
| "completion_length": 589.6312671661377, | |
| "epoch": 0.3578084234066344, | |
| "grad_norm": 0.30471851904044167, | |
| "kl": 0.219708251953125, | |
| "learning_rate": 1.621679448741067e-05, | |
| "loss": 0.0088, | |
| "reward": 0.0937500026077032, | |
| "reward_std": 0.07365695666521788, | |
| "rewards/accuracy_reward": 0.0937500026077032, | |
| "rewards/format_reward": 0.0, | |
| "step": 540 | |
| }, | |
| { | |
| "completion_length": 574.7833515167237, | |
| "epoch": 0.3611214643641032, | |
| "grad_norm": 0.2748841871191923, | |
| "kl": 0.220672607421875, | |
| "learning_rate": 1.6125779709710668e-05, | |
| "loss": 0.0088, | |
| "reward": 0.1104166692122817, | |
| "reward_std": 0.07365695666521788, | |
| "rewards/accuracy_reward": 0.1104166692122817, | |
| "rewards/format_reward": 0.0, | |
| "step": 545 | |
| }, | |
| { | |
| "completion_length": 610.558349609375, | |
| "epoch": 0.364434505321572, | |
| "grad_norm": 0.17413907087385358, | |
| "kl": 0.19656982421875, | |
| "learning_rate": 1.603394534182925e-05, | |
| "loss": 0.0079, | |
| "reward": 0.12291666939854622, | |
| "reward_std": 0.08544206973165273, | |
| "rewards/accuracy_reward": 0.12291666939854622, | |
| "rewards/format_reward": 0.0, | |
| "step": 550 | |
| }, | |
| { | |
| "completion_length": 632.8937694549561, | |
| "epoch": 0.3677475462790409, | |
| "grad_norm": 0.28335081427323106, | |
| "kl": 0.18609619140625, | |
| "learning_rate": 1.5941303670618018e-05, | |
| "loss": 0.0074, | |
| "reward": 0.11041666902601718, | |
| "reward_std": 0.10901229586452246, | |
| "rewards/accuracy_reward": 0.11041666902601718, | |
| "rewards/format_reward": 0.0, | |
| "step": 555 | |
| }, | |
| { | |
| "completion_length": 593.0187686920166, | |
| "epoch": 0.3710605872365097, | |
| "grad_norm": 0.34731414645367464, | |
| "kl": 0.191357421875, | |
| "learning_rate": 1.5847867090940602e-05, | |
| "loss": 0.0077, | |
| "reward": 0.12083333656191826, | |
| "reward_std": 0.1178511306643486, | |
| "rewards/accuracy_reward": 0.12083333656191826, | |
| "rewards/format_reward": 0.0, | |
| "step": 560 | |
| }, | |
| { | |
| "completion_length": 551.2416843414306, | |
| "epoch": 0.37437362819397857, | |
| "grad_norm": 0.3247113074997217, | |
| "kl": 0.196209716796875, | |
| "learning_rate": 1.57536481040143e-05, | |
| "loss": 0.0078, | |
| "reward": 0.11041666958481074, | |
| "reward_std": 0.08544206973165273, | |
| "rewards/accuracy_reward": 0.11041666958481074, | |
| "rewards/format_reward": 0.0, | |
| "step": 565 | |
| }, | |
| { | |
| "completion_length": 515.1666835784912, | |
| "epoch": 0.3776866691514474, | |
| "grad_norm": 0.2586024101464412, | |
| "kl": 0.2087890625, | |
| "learning_rate": 1.5658659315737505e-05, | |
| "loss": 0.0083, | |
| "reward": 0.11250000279396773, | |
| "reward_std": 0.10606601759791374, | |
| "rewards/accuracy_reward": 0.11250000279396773, | |
| "rewards/format_reward": 0.0, | |
| "step": 570 | |
| }, | |
| { | |
| "completion_length": 510.70001316070557, | |
| "epoch": 0.3809997101089162, | |
| "grad_norm": 0.29072360950689996, | |
| "kl": 0.20093994140625, | |
| "learning_rate": 1.5562913435003113e-05, | |
| "loss": 0.008, | |
| "reward": 0.12083333674818278, | |
| "reward_std": 0.09428090453147889, | |
| "rewards/accuracy_reward": 0.12083333674818278, | |
| "rewards/format_reward": 0.0, | |
| "step": 575 | |
| }, | |
| { | |
| "completion_length": 537.1416828155518, | |
| "epoch": 0.38431275106638507, | |
| "grad_norm": 0.4567470923989011, | |
| "kl": 0.19588623046875, | |
| "learning_rate": 1.5466423271998144e-05, | |
| "loss": 0.0078, | |
| "reward": 0.11666666939854622, | |
| "reward_std": 0.0766032349318266, | |
| "rewards/accuracy_reward": 0.11666666939854622, | |
| "rewards/format_reward": 0.0, | |
| "step": 580 | |
| }, | |
| { | |
| "completion_length": 545.7729301452637, | |
| "epoch": 0.3876257920238539, | |
| "grad_norm": 0.22942721449166853, | |
| "kl": 0.210009765625, | |
| "learning_rate": 1.536920173648984e-05, | |
| "loss": 0.0084, | |
| "reward": 0.10416666939854621, | |
| "reward_std": 0.08838834781199693, | |
| "rewards/accuracy_reward": 0.10416666939854621, | |
| "rewards/format_reward": 0.0, | |
| "step": 585 | |
| }, | |
| { | |
| "completion_length": 562.1437660217285, | |
| "epoch": 0.39093883298132276, | |
| "grad_norm": 0.16727493145277006, | |
| "kl": 0.21328125, | |
| "learning_rate": 1.5271261836098403e-05, | |
| "loss": 0.0085, | |
| "reward": 0.07708333544433117, | |
| "reward_std": 0.06776440013200044, | |
| "rewards/accuracy_reward": 0.07708333544433117, | |
| "rewards/format_reward": 0.0, | |
| "step": 590 | |
| }, | |
| { | |
| "completion_length": 559.0500179290772, | |
| "epoch": 0.3942518739387916, | |
| "grad_norm": 0.26024076771764676, | |
| "kl": 0.210040283203125, | |
| "learning_rate": 1.5172616674556673e-05, | |
| "loss": 0.0084, | |
| "reward": 0.0979166692122817, | |
| "reward_std": 0.09133462626487017, | |
| "rewards/accuracy_reward": 0.0979166692122817, | |
| "rewards/format_reward": 0.0, | |
| "step": 595 | |
| }, | |
| { | |
| "completion_length": 555.5812656402588, | |
| "epoch": 0.3975649148962604, | |
| "grad_norm": 0.2500300158618388, | |
| "kl": 0.17904052734375, | |
| "learning_rate": 1.5073279449956916e-05, | |
| "loss": 0.0072, | |
| "reward": 0.14375000298023224, | |
| "reward_std": 0.10311973914504051, | |
| "rewards/accuracy_reward": 0.14375000298023224, | |
| "rewards/format_reward": 0.0, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.3975649148962604, | |
| "eval_completion_length": 543.9804059196921, | |
| "eval_kl": 0.17497702205882354, | |
| "eval_loss": 0.006994770839810371, | |
| "eval_reward": 0.15196078928077922, | |
| "eval_reward_std": 0.09012145242270302, | |
| "eval_rewards/accuracy_reward": 0.15196078928077922, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_runtime": 78.4023, | |
| "eval_samples_per_second": 1.263, | |
| "eval_steps_per_second": 0.115, | |
| "step": 600 | |
| }, | |
| { | |
| "completion_length": 567.908349609375, | |
| "epoch": 0.40087795585372926, | |
| "grad_norm": 0.30328216491481214, | |
| "kl": 0.16656494140625, | |
| "learning_rate": 1.4973263452985023e-05, | |
| "loss": 0.0067, | |
| "reward": 0.15208333749324082, | |
| "reward_std": 0.12079740893095732, | |
| "rewards/accuracy_reward": 0.15208333749324082, | |
| "rewards/format_reward": 0.0, | |
| "step": 605 | |
| }, | |
| { | |
| "completion_length": 587.0479328155518, | |
| "epoch": 0.4041909968111981, | |
| "grad_norm": 0.28902742765864486, | |
| "kl": 0.166668701171875, | |
| "learning_rate": 1.4872582065142285e-05, | |
| "loss": 0.0067, | |
| "reward": 0.12291666958481073, | |
| "reward_std": 0.10311973914504051, | |
| "rewards/accuracy_reward": 0.12291666958481073, | |
| "rewards/format_reward": 0.0, | |
| "step": 610 | |
| }, | |
| { | |
| "completion_length": 580.5958499908447, | |
| "epoch": 0.4075040377686669, | |
| "grad_norm": 0.26061142390184844, | |
| "kl": 0.179718017578125, | |
| "learning_rate": 1.4771248756955042e-05, | |
| "loss": 0.0072, | |
| "reward": 0.15416667014360427, | |
| "reward_std": 0.11785113047808408, | |
| "rewards/accuracy_reward": 0.15416667014360427, | |
| "rewards/format_reward": 0.0, | |
| "step": 615 | |
| }, | |
| { | |
| "completion_length": 621.1146015167236, | |
| "epoch": 0.41081707872613576, | |
| "grad_norm": 0.3726707315873621, | |
| "kl": 0.20567626953125, | |
| "learning_rate": 1.4669277086172406e-05, | |
| "loss": 0.0082, | |
| "reward": 0.10416666958481073, | |
| "reward_std": 0.10017346106469631, | |
| "rewards/accuracy_reward": 0.10416666958481073, | |
| "rewards/format_reward": 0.0, | |
| "step": 620 | |
| }, | |
| { | |
| "completion_length": 566.5250190734863, | |
| "epoch": 0.4141301196836046, | |
| "grad_norm": 0.28592154315160423, | |
| "kl": 0.21568603515625, | |
| "learning_rate": 1.4566680695952333e-05, | |
| "loss": 0.0086, | |
| "reward": 0.10625000279396772, | |
| "reward_std": 0.09722718279808759, | |
| "rewards/accuracy_reward": 0.10625000279396772, | |
| "rewards/format_reward": 0.0, | |
| "step": 625 | |
| }, | |
| { | |
| "completion_length": 470.1937656402588, | |
| "epoch": 0.41744316064107345, | |
| "grad_norm": 0.3545452795184588, | |
| "kl": 0.217047119140625, | |
| "learning_rate": 1.4463473313036241e-05, | |
| "loss": 0.0087, | |
| "reward": 0.10625000298023224, | |
| "reward_std": 0.09722718279808759, | |
| "rewards/accuracy_reward": 0.10625000298023224, | |
| "rewards/format_reward": 0.0, | |
| "step": 630 | |
| }, | |
| { | |
| "completion_length": 489.2375156402588, | |
| "epoch": 0.42075620159854227, | |
| "grad_norm": 0.281093922778775, | |
| "kl": 0.183544921875, | |
| "learning_rate": 1.4359668745912472e-05, | |
| "loss": 0.0073, | |
| "reward": 0.15416667014360427, | |
| "reward_std": 0.10017346106469631, | |
| "rewards/accuracy_reward": 0.15416667014360427, | |
| "rewards/format_reward": 0.0, | |
| "step": 635 | |
| }, | |
| { | |
| "completion_length": 509.7854309082031, | |
| "epoch": 0.4240692425560111, | |
| "grad_norm": 0.3083814307787709, | |
| "kl": 0.183203125, | |
| "learning_rate": 1.4255280882968787e-05, | |
| "loss": 0.0073, | |
| "reward": 0.1416666701436043, | |
| "reward_std": 0.11195857413113117, | |
| "rewards/accuracy_reward": 0.1416666701436043, | |
| "rewards/format_reward": 0.0, | |
| "step": 640 | |
| }, | |
| { | |
| "completion_length": 531.0479331970215, | |
| "epoch": 0.42738228351347995, | |
| "grad_norm": 0.4133177147977361, | |
| "kl": 0.219439697265625, | |
| "learning_rate": 1.415032369063422e-05, | |
| "loss": 0.0088, | |
| "reward": 0.1458333371207118, | |
| "reward_std": 0.15320646949112415, | |
| "rewards/accuracy_reward": 0.1458333371207118, | |
| "rewards/format_reward": 0.0, | |
| "step": 645 | |
| }, | |
| { | |
| "completion_length": 534.3729320526123, | |
| "epoch": 0.43069532447094877, | |
| "grad_norm": 0.24836791560619703, | |
| "kl": 0.194482421875, | |
| "learning_rate": 1.4044811211510419e-05, | |
| "loss": 0.0078, | |
| "reward": 0.11666666977107525, | |
| "reward_std": 0.10017346106469631, | |
| "rewards/accuracy_reward": 0.11666666977107525, | |
| "rewards/format_reward": 0.0, | |
| "step": 650 | |
| }, | |
| { | |
| "completion_length": 495.2333480834961, | |
| "epoch": 0.4340083654284176, | |
| "grad_norm": 0.26953166407001816, | |
| "kl": 0.20745849609375, | |
| "learning_rate": 1.3938757562492873e-05, | |
| "loss": 0.0083, | |
| "reward": 0.10416666977107525, | |
| "reward_std": 0.09428090453147889, | |
| "rewards/accuracy_reward": 0.10416666977107525, | |
| "rewards/format_reward": 0.0, | |
| "step": 655 | |
| }, | |
| { | |
| "completion_length": 448.07918033599856, | |
| "epoch": 0.43732140638588646, | |
| "grad_norm": 0.2813525701476749, | |
| "kl": 0.20928955078125, | |
| "learning_rate": 1.3832176932882136e-05, | |
| "loss": 0.0084, | |
| "reward": 0.1250000024214387, | |
| "reward_std": 0.09428090453147889, | |
| "rewards/accuracy_reward": 0.1250000024214387, | |
| "rewards/format_reward": 0.0, | |
| "step": 660 | |
| }, | |
| { | |
| "completion_length": 482.58334884643557, | |
| "epoch": 0.4406344473433553, | |
| "grad_norm": 0.23983882743615745, | |
| "kl": 0.20230712890625, | |
| "learning_rate": 1.3725083582485397e-05, | |
| "loss": 0.0081, | |
| "reward": 0.12083333618938923, | |
| "reward_std": 0.10017346106469631, | |
| "rewards/accuracy_reward": 0.12083333618938923, | |
| "rewards/format_reward": 0.0, | |
| "step": 665 | |
| }, | |
| { | |
| "completion_length": 531.0250171661377, | |
| "epoch": 0.44394748830082414, | |
| "grad_norm": 0.23057904466712248, | |
| "kl": 0.18568115234375, | |
| "learning_rate": 1.3617491839708614e-05, | |
| "loss": 0.0074, | |
| "reward": 0.11875000316649675, | |
| "reward_std": 0.08544206973165273, | |
| "rewards/accuracy_reward": 0.11875000316649675, | |
| "rewards/format_reward": 0.0, | |
| "step": 670 | |
| }, | |
| { | |
| "completion_length": 546.8812679290771, | |
| "epoch": 0.44726052925829296, | |
| "grad_norm": 0.2918960689101493, | |
| "kl": 0.17314453125, | |
| "learning_rate": 1.3509416099639456e-05, | |
| "loss": 0.0069, | |
| "reward": 0.14375000316649675, | |
| "reward_std": 0.13847507834434508, | |
| "rewards/accuracy_reward": 0.14375000316649675, | |
| "rewards/format_reward": 0.0, | |
| "step": 675 | |
| }, | |
| { | |
| "completion_length": 525.9000158309937, | |
| "epoch": 0.4505735702157618, | |
| "grad_norm": 0.33464307998263254, | |
| "kl": 0.18521728515625, | |
| "learning_rate": 1.3400870822121348e-05, | |
| "loss": 0.0074, | |
| "reward": 0.11250000298023224, | |
| "reward_std": 0.10017346106469631, | |
| "rewards/accuracy_reward": 0.11250000298023224, | |
| "rewards/format_reward": 0.0, | |
| "step": 680 | |
| }, | |
| { | |
| "completion_length": 496.16668090820315, | |
| "epoch": 0.45388661117323065, | |
| "grad_norm": 0.24577312673005983, | |
| "kl": 0.2126708984375, | |
| "learning_rate": 1.3291870529818809e-05, | |
| "loss": 0.0085, | |
| "reward": 0.08333333544433116, | |
| "reward_std": 0.07071067839860916, | |
| "rewards/accuracy_reward": 0.08333333544433116, | |
| "rewards/format_reward": 0.0, | |
| "step": 685 | |
| }, | |
| { | |
| "completion_length": 516.3520980834961, | |
| "epoch": 0.45719965213069946, | |
| "grad_norm": 0.28279663282344375, | |
| "kl": 0.241748046875, | |
| "learning_rate": 1.3182429806274442e-05, | |
| "loss": 0.0097, | |
| "reward": 0.1145833358168602, | |
| "reward_std": 0.10311973914504051, | |
| "rewards/accuracy_reward": 0.1145833358168602, | |
| "rewards/format_reward": 0.0, | |
| "step": 690 | |
| }, | |
| { | |
| "completion_length": 489.8125160217285, | |
| "epoch": 0.4605126930881683, | |
| "grad_norm": 0.24489140060416123, | |
| "kl": 0.17784423828125, | |
| "learning_rate": 1.3072563293957725e-05, | |
| "loss": 0.0071, | |
| "reward": 0.10625000279396772, | |
| "reward_std": 0.09722718279808759, | |
| "rewards/accuracy_reward": 0.10625000279396772, | |
| "rewards/format_reward": 0.0, | |
| "step": 695 | |
| }, | |
| { | |
| "completion_length": 515.912516784668, | |
| "epoch": 0.46382573404563715, | |
| "grad_norm": 0.25229498956295204, | |
| "kl": 0.17552490234375, | |
| "learning_rate": 1.2962285692305964e-05, | |
| "loss": 0.007, | |
| "reward": 0.12916666958481074, | |
| "reward_std": 0.12374368701130152, | |
| "rewards/accuracy_reward": 0.12916666958481074, | |
| "rewards/format_reward": 0.0, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.46382573404563715, | |
| "eval_completion_length": 477.3774611529182, | |
| "eval_kl": 0.1923828125, | |
| "eval_loss": 0.007697770372033119, | |
| "eval_reward": 0.15196078752770142, | |
| "eval_reward_std": 0.10398629176266053, | |
| "eval_rewards/accuracy_reward": 0.15196078752770142, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_runtime": 73.7384, | |
| "eval_samples_per_second": 1.343, | |
| "eval_steps_per_second": 0.122, | |
| "step": 700 | |
| }, | |
| { | |
| "completion_length": 521.3458486557007, | |
| "epoch": 0.46713877500310597, | |
| "grad_norm": 0.3114651168291986, | |
| "kl": 0.19656982421875, | |
| "learning_rate": 1.2851611755757587e-05, | |
| "loss": 0.0079, | |
| "reward": 0.08750000186264514, | |
| "reward_std": 0.0766032349318266, | |
| "rewards/accuracy_reward": 0.08750000186264514, | |
| "rewards/format_reward": 0.0, | |
| "step": 705 | |
| }, | |
| { | |
| "completion_length": 505.2958469390869, | |
| "epoch": 0.47045181596057484, | |
| "grad_norm": 0.2616041148953221, | |
| "kl": 0.18330078125, | |
| "learning_rate": 1.2740556291778096e-05, | |
| "loss": 0.0073, | |
| "reward": 0.1479166703298688, | |
| "reward_std": 0.10311973914504051, | |
| "rewards/accuracy_reward": 0.1479166703298688, | |
| "rewards/format_reward": 0.0, | |
| "step": 710 | |
| }, | |
| { | |
| "completion_length": 501.8895986557007, | |
| "epoch": 0.47376485691804365, | |
| "grad_norm": 0.23670699846471074, | |
| "kl": 0.20030517578125, | |
| "learning_rate": 1.2629134158878919e-05, | |
| "loss": 0.008, | |
| "reward": 0.11666666939854622, | |
| "reward_std": 0.09428090453147889, | |
| "rewards/accuracy_reward": 0.11666666939854622, | |
| "rewards/format_reward": 0.0, | |
| "step": 715 | |
| }, | |
| { | |
| "completion_length": 569.3500164031982, | |
| "epoch": 0.47707789787551247, | |
| "grad_norm": 0.1767297956772261, | |
| "kl": 0.184912109375, | |
| "learning_rate": 1.2517360264629463e-05, | |
| "loss": 0.0074, | |
| "reward": 0.07083333525806665, | |
| "reward_std": 0.05303300879895687, | |
| "rewards/accuracy_reward": 0.07083333525806665, | |
| "rewards/format_reward": 0.0, | |
| "step": 720 | |
| }, | |
| { | |
| "completion_length": 549.1854320526123, | |
| "epoch": 0.48039093883298134, | |
| "grad_norm": 0.2515191189191733, | |
| "kl": 0.198480224609375, | |
| "learning_rate": 1.2405249563662539e-05, | |
| "loss": 0.0079, | |
| "reward": 0.07500000204890966, | |
| "reward_std": 0.08838834799826145, | |
| "rewards/accuracy_reward": 0.07500000204890966, | |
| "rewards/format_reward": 0.0, | |
| "step": 725 | |
| }, | |
| { | |
| "completion_length": 502.370849609375, | |
| "epoch": 0.48370397979045016, | |
| "grad_norm": 0.32374848605233897, | |
| "kl": 0.21995849609375, | |
| "learning_rate": 1.2292817055673543e-05, | |
| "loss": 0.0088, | |
| "reward": 0.08125000223517417, | |
| "reward_std": 0.07954951319843531, | |
| "rewards/accuracy_reward": 0.08125000223517417, | |
| "rewards/format_reward": 0.0, | |
| "step": 730 | |
| }, | |
| { | |
| "completion_length": 496.44584503173826, | |
| "epoch": 0.48701702074791897, | |
| "grad_norm": 0.3225185847351344, | |
| "kl": 0.195855712890625, | |
| "learning_rate": 1.2180077783413601e-05, | |
| "loss": 0.0078, | |
| "reward": 0.08125000223517417, | |
| "reward_std": 0.06776440013200044, | |
| "rewards/accuracy_reward": 0.08125000223517417, | |
| "rewards/format_reward": 0.0, | |
| "step": 735 | |
| }, | |
| { | |
| "completion_length": 478.5145969390869, | |
| "epoch": 0.49033006170538784, | |
| "grad_norm": 0.3624783956571456, | |
| "kl": 0.192364501953125, | |
| "learning_rate": 1.2067046830676947e-05, | |
| "loss": 0.0077, | |
| "reward": 0.1479166705161333, | |
| "reward_std": 0.13258252181112767, | |
| "rewards/accuracy_reward": 0.1479166705161333, | |
| "rewards/format_reward": 0.0, | |
| "step": 740 | |
| }, | |
| { | |
| "completion_length": 498.3958461761475, | |
| "epoch": 0.49364310266285666, | |
| "grad_norm": 0.29628298606274384, | |
| "kl": 0.195697021484375, | |
| "learning_rate": 1.1953739320282778e-05, | |
| "loss": 0.0078, | |
| "reward": 0.12708333674818278, | |
| "reward_std": 0.09722718279808759, | |
| "rewards/accuracy_reward": 0.12708333674818278, | |
| "rewards/format_reward": 0.0, | |
| "step": 745 | |
| }, | |
| { | |
| "completion_length": 522.5895965576171, | |
| "epoch": 0.49695614362032553, | |
| "grad_norm": 0.273049970135215, | |
| "kl": 0.18468017578125, | |
| "learning_rate": 1.1840170412051957e-05, | |
| "loss": 0.0074, | |
| "reward": 0.10833333544433117, | |
| "reward_std": 0.08838834799826145, | |
| "rewards/accuracy_reward": 0.10833333544433117, | |
| "rewards/format_reward": 0.0, | |
| "step": 750 | |
| }, | |
| { | |
| "completion_length": 523.5166801452636, | |
| "epoch": 0.5002691845777943, | |
| "grad_norm": 0.4360470753971284, | |
| "kl": 0.20830078125, | |
| "learning_rate": 1.1726355300778693e-05, | |
| "loss": 0.0083, | |
| "reward": 0.10416666939854621, | |
| "reward_std": 0.12374368719756604, | |
| "rewards/accuracy_reward": 0.10416666939854621, | |
| "rewards/format_reward": 0.0, | |
| "step": 755 | |
| }, | |
| { | |
| "completion_length": 585.5500165939332, | |
| "epoch": 0.5035822255352632, | |
| "grad_norm": 0.18267493572339938, | |
| "kl": 0.21558837890625, | |
| "learning_rate": 1.1612309214197599e-05, | |
| "loss": 0.0086, | |
| "reward": 0.10000000260770321, | |
| "reward_std": 0.0766032349318266, | |
| "rewards/accuracy_reward": 0.10000000260770321, | |
| "rewards/format_reward": 0.0, | |
| "step": 760 | |
| }, | |
| { | |
| "completion_length": 567.3645992279053, | |
| "epoch": 0.506895266492732, | |
| "grad_norm": 0.24654038356285773, | |
| "kl": 0.23746337890625, | |
| "learning_rate": 1.1498047410946307e-05, | |
| "loss": 0.0095, | |
| "reward": 0.0875000024214387, | |
| "reward_std": 0.10017346087843179, | |
| "rewards/accuracy_reward": 0.0875000024214387, | |
| "rewards/format_reward": 0.0, | |
| "step": 765 | |
| }, | |
| { | |
| "completion_length": 611.8812675476074, | |
| "epoch": 0.5102083074502008, | |
| "grad_norm": 0.17786553342995753, | |
| "kl": 0.222509765625, | |
| "learning_rate": 1.1383585178523955e-05, | |
| "loss": 0.0089, | |
| "reward": 0.09583333600312471, | |
| "reward_std": 0.07071067839860916, | |
| "rewards/accuracy_reward": 0.09583333600312471, | |
| "rewards/format_reward": 0.0, | |
| "step": 770 | |
| }, | |
| { | |
| "completion_length": 556.170849609375, | |
| "epoch": 0.5135213484076697, | |
| "grad_norm": 0.32776722432627464, | |
| "kl": 0.197369384765625, | |
| "learning_rate": 1.126893783124583e-05, | |
| "loss": 0.0079, | |
| "reward": 0.11875000335276127, | |
| "reward_std": 0.11490485239773988, | |
| "rewards/accuracy_reward": 0.11875000335276127, | |
| "rewards/format_reward": 0.0, | |
| "step": 775 | |
| }, | |
| { | |
| "completion_length": 489.2875148773193, | |
| "epoch": 0.5168343893651385, | |
| "grad_norm": 0.22661050449666004, | |
| "kl": 0.199169921875, | |
| "learning_rate": 1.1154120708194398e-05, | |
| "loss": 0.008, | |
| "reward": 0.08333333563059568, | |
| "reward_std": 0.07071067839860916, | |
| "rewards/accuracy_reward": 0.08333333563059568, | |
| "rewards/format_reward": 0.0, | |
| "step": 780 | |
| }, | |
| { | |
| "completion_length": 483.7916805267334, | |
| "epoch": 0.5201474303226074, | |
| "grad_norm": 0.3631761942043462, | |
| "kl": 0.200616455078125, | |
| "learning_rate": 1.1039149171167046e-05, | |
| "loss": 0.008, | |
| "reward": 0.12291666977107525, | |
| "reward_std": 0.10901229586452246, | |
| "rewards/accuracy_reward": 0.12291666977107525, | |
| "rewards/format_reward": 0.0, | |
| "step": 785 | |
| }, | |
| { | |
| "completion_length": 473.125016784668, | |
| "epoch": 0.5234604712800762, | |
| "grad_norm": 0.26699712541162623, | |
| "kl": 0.192742919921875, | |
| "learning_rate": 1.0924038602620757e-05, | |
| "loss": 0.0077, | |
| "reward": 0.11041666958481074, | |
| "reward_std": 0.09133462626487017, | |
| "rewards/accuracy_reward": 0.11041666958481074, | |
| "rewards/format_reward": 0.0, | |
| "step": 790 | |
| }, | |
| { | |
| "completion_length": 500.1333484649658, | |
| "epoch": 0.526773512237545, | |
| "grad_norm": 0.41463867334904875, | |
| "kl": 0.18306884765625, | |
| "learning_rate": 1.0808804403614044e-05, | |
| "loss": 0.0073, | |
| "reward": 0.12291666939854622, | |
| "reward_std": 0.10901229567825794, | |
| "rewards/accuracy_reward": 0.12291666939854622, | |
| "rewards/format_reward": 0.0, | |
| "step": 795 | |
| }, | |
| { | |
| "completion_length": 500.9916820526123, | |
| "epoch": 0.5300865531950139, | |
| "grad_norm": 0.35472513450823207, | |
| "kl": 0.18797607421875, | |
| "learning_rate": 1.0693461991746389e-05, | |
| "loss": 0.0075, | |
| "reward": 0.1416666703298688, | |
| "reward_std": 0.10017346106469631, | |
| "rewards/accuracy_reward": 0.1416666703298688, | |
| "rewards/format_reward": 0.0, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.5300865531950139, | |
| "eval_completion_length": 496.34805118336396, | |
| "eval_kl": 0.20209099264705882, | |
| "eval_loss": 0.007962403818964958, | |
| "eval_reward": 0.1421568661051638, | |
| "eval_reward_std": 0.1178511306643486, | |
| "eval_rewards/accuracy_reward": 0.1421568661051638, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_runtime": 74.1818, | |
| "eval_samples_per_second": 1.335, | |
| "eval_steps_per_second": 0.121, | |
| "step": 800 | |
| }, | |
| { | |
| "completion_length": 525.1229301452637, | |
| "epoch": 0.5333995941524827, | |
| "grad_norm": 0.34019503442109345, | |
| "kl": 0.232171630859375, | |
| "learning_rate": 1.0578026799095464e-05, | |
| "loss": 0.0093, | |
| "reward": 0.14583333693444728, | |
| "reward_std": 0.12374368701130152, | |
| "rewards/accuracy_reward": 0.14583333693444728, | |
| "rewards/format_reward": 0.0, | |
| "step": 805 | |
| }, | |
| { | |
| "completion_length": 615.4854366302491, | |
| "epoch": 0.5367126351099516, | |
| "grad_norm": 0.28699140537614237, | |
| "kl": 0.2682373046875, | |
| "learning_rate": 1.046251427015241e-05, | |
| "loss": 0.0107, | |
| "reward": 0.1416666705161333, | |
| "reward_std": 0.10017346106469631, | |
| "rewards/accuracy_reward": 0.1416666705161333, | |
| "rewards/format_reward": 0.0, | |
| "step": 810 | |
| }, | |
| { | |
| "completion_length": 609.1979358673095, | |
| "epoch": 0.5400256760674204, | |
| "grad_norm": 0.29816057971503035, | |
| "kl": 0.22767333984375, | |
| "learning_rate": 1.0346939859755481e-05, | |
| "loss": 0.0091, | |
| "reward": 0.10416666921228171, | |
| "reward_std": 0.11195857413113117, | |
| "rewards/accuracy_reward": 0.10416666921228171, | |
| "rewards/format_reward": 0.0, | |
| "step": 815 | |
| }, | |
| { | |
| "completion_length": 649.9166870117188, | |
| "epoch": 0.5433387170248892, | |
| "grad_norm": 1.1960413989590513, | |
| "kl": 0.33731689453125, | |
| "learning_rate": 1.023131903102226e-05, | |
| "loss": 0.0135, | |
| "reward": 0.09791666902601719, | |
| "reward_std": 0.10311973933130503, | |
| "rewards/accuracy_reward": 0.09791666902601719, | |
| "rewards/format_reward": 0.0, | |
| "step": 820 | |
| }, | |
| { | |
| "completion_length": 554.097931098938, | |
| "epoch": 0.5466517579823581, | |
| "grad_norm": 0.30872099739539816, | |
| "kl": 0.247076416015625, | |
| "learning_rate": 1.0115667253280817e-05, | |
| "loss": 0.0099, | |
| "reward": 0.11041666958481074, | |
| "reward_std": 0.09722718279808759, | |
| "rewards/accuracy_reward": 0.11041666958481074, | |
| "rewards/format_reward": 0.0, | |
| "step": 825 | |
| }, | |
| { | |
| "completion_length": 552.6354309082031, | |
| "epoch": 0.5499647989398269, | |
| "grad_norm": 0.20054998735675889, | |
| "kl": 0.206219482421875, | |
| "learning_rate": 1e-05, | |
| "loss": 0.0082, | |
| "reward": 0.12500000298023223, | |
| "reward_std": 0.09428090434521437, | |
| "rewards/accuracy_reward": 0.12500000298023223, | |
| "rewards/format_reward": 0.0, | |
| "step": 830 | |
| }, | |
| { | |
| "completion_length": 546.9625179290772, | |
| "epoch": 0.5532778398972957, | |
| "grad_norm": 0.5641613086876952, | |
| "kl": 0.255999755859375, | |
| "learning_rate": 9.884332746719186e-06, | |
| "loss": 0.0102, | |
| "reward": 0.12708333674818278, | |
| "reward_std": 0.1207974087446928, | |
| "rewards/accuracy_reward": 0.12708333674818278, | |
| "rewards/format_reward": 0.0, | |
| "step": 835 | |
| }, | |
| { | |
| "completion_length": 554.7666828155518, | |
| "epoch": 0.5565908808547646, | |
| "grad_norm": 0.3738960340913415, | |
| "kl": 0.1666259765625, | |
| "learning_rate": 9.768680968977743e-06, | |
| "loss": 0.0067, | |
| "reward": 0.12083333674818278, | |
| "reward_std": 0.1178511306643486, | |
| "rewards/accuracy_reward": 0.12083333674818278, | |
| "rewards/format_reward": 0.0, | |
| "step": 840 | |
| }, | |
| { | |
| "completion_length": 566.0395988464355, | |
| "epoch": 0.5599039218122334, | |
| "grad_norm": 0.22732871482096867, | |
| "kl": 0.190960693359375, | |
| "learning_rate": 9.653060140244524e-06, | |
| "loss": 0.0076, | |
| "reward": 0.10000000223517418, | |
| "reward_std": 0.08838834799826145, | |
| "rewards/accuracy_reward": 0.10000000223517418, | |
| "rewards/format_reward": 0.0, | |
| "step": 845 | |
| }, | |
| { | |
| "completion_length": 618.0916870117187, | |
| "epoch": 0.5632169627697022, | |
| "grad_norm": 0.40863278743755727, | |
| "kl": 0.255059814453125, | |
| "learning_rate": 9.537485729847594e-06, | |
| "loss": 0.0102, | |
| "reward": 0.12500000298023223, | |
| "reward_std": 0.11195857413113117, | |
| "rewards/accuracy_reward": 0.12500000298023223, | |
| "rewards/format_reward": 0.0, | |
| "step": 850 | |
| }, | |
| { | |
| "completion_length": 615.8875164031982, | |
| "epoch": 0.5665300037271711, | |
| "grad_norm": 0.29927787657044663, | |
| "kl": 0.259979248046875, | |
| "learning_rate": 9.421973200904538e-06, | |
| "loss": 0.0104, | |
| "reward": 0.12291667051613331, | |
| "reward_std": 0.12668996546417474, | |
| "rewards/accuracy_reward": 0.12291667051613331, | |
| "rewards/format_reward": 0.0, | |
| "step": 855 | |
| }, | |
| { | |
| "completion_length": 564.025016784668, | |
| "epoch": 0.5698430446846399, | |
| "grad_norm": 0.5404660152394563, | |
| "kl": 0.21617431640625, | |
| "learning_rate": 9.306538008253611e-06, | |
| "loss": 0.0086, | |
| "reward": 0.13541667014360428, | |
| "reward_std": 0.1266899650916457, | |
| "rewards/accuracy_reward": 0.13541667014360428, | |
| "rewards/format_reward": 0.0, | |
| "step": 860 | |
| }, | |
| { | |
| "completion_length": 544.654182434082, | |
| "epoch": 0.5731560856421087, | |
| "grad_norm": 0.3233447008213159, | |
| "kl": 0.181341552734375, | |
| "learning_rate": 9.19119559638596e-06, | |
| "loss": 0.0073, | |
| "reward": 0.14583333749324084, | |
| "reward_std": 0.10017346106469631, | |
| "rewards/accuracy_reward": 0.14583333749324084, | |
| "rewards/format_reward": 0.0, | |
| "step": 865 | |
| }, | |
| { | |
| "completion_length": 572.0687698364258, | |
| "epoch": 0.5764691265995776, | |
| "grad_norm": 0.3691524796796649, | |
| "kl": 0.217767333984375, | |
| "learning_rate": 9.075961397379247e-06, | |
| "loss": 0.0087, | |
| "reward": 0.13333333656191826, | |
| "reward_std": 0.10017346106469631, | |
| "rewards/accuracy_reward": 0.13333333656191826, | |
| "rewards/format_reward": 0.0, | |
| "step": 870 | |
| }, | |
| { | |
| "completion_length": 615.9208526611328, | |
| "epoch": 0.5797821675570465, | |
| "grad_norm": 0.2985439339832434, | |
| "kl": 0.21961669921875, | |
| "learning_rate": 8.960850828832958e-06, | |
| "loss": 0.0088, | |
| "reward": 0.12708333618938922, | |
| "reward_std": 0.09133462607860565, | |
| "rewards/accuracy_reward": 0.12708333618938922, | |
| "rewards/format_reward": 0.0, | |
| "step": 875 | |
| }, | |
| { | |
| "completion_length": 651.279183959961, | |
| "epoch": 0.5830952085145152, | |
| "grad_norm": 0.2039953541987364, | |
| "kl": 0.229315185546875, | |
| "learning_rate": 8.845879291805605e-06, | |
| "loss": 0.0092, | |
| "reward": 0.11250000298023224, | |
| "reward_std": 0.0766032349318266, | |
| "rewards/accuracy_reward": 0.11250000298023224, | |
| "rewards/format_reward": 0.0, | |
| "step": 880 | |
| }, | |
| { | |
| "completion_length": 635.308352279663, | |
| "epoch": 0.5864082494719841, | |
| "grad_norm": 0.33333680886000144, | |
| "kl": 0.180218505859375, | |
| "learning_rate": 8.731062168754174e-06, | |
| "loss": 0.0072, | |
| "reward": 0.13125000335276127, | |
| "reward_std": 0.12079740893095732, | |
| "rewards/accuracy_reward": 0.13125000335276127, | |
| "rewards/format_reward": 0.0, | |
| "step": 885 | |
| }, | |
| { | |
| "completion_length": 575.0104343414307, | |
| "epoch": 0.589721290429453, | |
| "grad_norm": 8.973633703652538, | |
| "kl": 0.467108154296875, | |
| "learning_rate": 8.616414821476048e-06, | |
| "loss": 0.0188, | |
| "reward": 0.10416666921228171, | |
| "reward_std": 0.0589255653321743, | |
| "rewards/accuracy_reward": 0.10416666921228171, | |
| "rewards/format_reward": 0.0, | |
| "step": 890 | |
| }, | |
| { | |
| "completion_length": 574.4812694549561, | |
| "epoch": 0.5930343313869217, | |
| "grad_norm": 0.26182341985351215, | |
| "kl": 0.16265869140625, | |
| "learning_rate": 8.501952589053694e-06, | |
| "loss": 0.0065, | |
| "reward": 0.13541667014360428, | |
| "reward_std": 0.10901229586452246, | |
| "rewards/accuracy_reward": 0.13541667014360428, | |
| "rewards/format_reward": 0.0, | |
| "step": 895 | |
| }, | |
| { | |
| "completion_length": 579.7250156402588, | |
| "epoch": 0.5963473723443906, | |
| "grad_norm": 0.3188530106914609, | |
| "kl": 0.18770751953125, | |
| "learning_rate": 8.387690785802403e-06, | |
| "loss": 0.0075, | |
| "reward": 0.1500000037252903, | |
| "reward_std": 0.12963624373078347, | |
| "rewards/accuracy_reward": 0.1500000037252903, | |
| "rewards/format_reward": 0.0, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.5963473723443906, | |
| "eval_completion_length": 560.4804077148438, | |
| "eval_kl": 0.21737132352941177, | |
| "eval_loss": 0.008583012968301773, | |
| "eval_reward": 0.17156863168758504, | |
| "eval_reward_std": 0.1455808080294553, | |
| "eval_rewards/accuracy_reward": 0.17156863168758504, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_runtime": 80.338, | |
| "eval_samples_per_second": 1.232, | |
| "eval_steps_per_second": 0.112, | |
| "step": 900 | |
| }, | |
| { | |
| "completion_length": 593.8625198364258, | |
| "epoch": 0.5996604133018595, | |
| "grad_norm": 0.2613535270080971, | |
| "kl": 0.201318359375, | |
| "learning_rate": 8.273644699221309e-06, | |
| "loss": 0.0081, | |
| "reward": 0.08958333600312471, | |
| "reward_std": 0.08544206973165273, | |
| "rewards/accuracy_reward": 0.08958333600312471, | |
| "rewards/format_reward": 0.0, | |
| "step": 905 | |
| }, | |
| { | |
| "completion_length": 599.837515258789, | |
| "epoch": 0.6029734542593282, | |
| "grad_norm": 0.19506591405221774, | |
| "kl": 0.188922119140625, | |
| "learning_rate": 8.159829587948048e-06, | |
| "loss": 0.0076, | |
| "reward": 0.12291666977107525, | |
| "reward_std": 0.12079740893095732, | |
| "rewards/accuracy_reward": 0.12291666977107525, | |
| "rewards/format_reward": 0.0, | |
| "step": 910 | |
| }, | |
| { | |
| "completion_length": 650.3021049499512, | |
| "epoch": 0.6062864952167971, | |
| "grad_norm": 0.27346090434561215, | |
| "kl": 0.176104736328125, | |
| "learning_rate": 8.046260679717225e-06, | |
| "loss": 0.007, | |
| "reward": 0.1395833369344473, | |
| "reward_std": 0.10901229567825794, | |
| "rewards/accuracy_reward": 0.1395833369344473, | |
| "rewards/format_reward": 0.0, | |
| "step": 915 | |
| }, | |
| { | |
| "completion_length": 615.7187690734863, | |
| "epoch": 0.609599536174266, | |
| "grad_norm": 0.34804926597561486, | |
| "kl": 0.18919677734375, | |
| "learning_rate": 7.932953169323057e-06, | |
| "loss": 0.0076, | |
| "reward": 0.14375000353902578, | |
| "reward_std": 0.1207974087446928, | |
| "rewards/accuracy_reward": 0.14375000353902578, | |
| "rewards/format_reward": 0.0, | |
| "step": 920 | |
| }, | |
| { | |
| "completion_length": 565.602099609375, | |
| "epoch": 0.6129125771317347, | |
| "grad_norm": 0.23449254684560134, | |
| "kl": 0.1919677734375, | |
| "learning_rate": 7.8199222165864e-06, | |
| "loss": 0.0077, | |
| "reward": 0.12291666995733977, | |
| "reward_std": 0.10311973933130503, | |
| "rewards/accuracy_reward": 0.12291666995733977, | |
| "rewards/format_reward": 0.0, | |
| "step": 925 | |
| }, | |
| { | |
| "completion_length": 589.8625164031982, | |
| "epoch": 0.6162256180892036, | |
| "grad_norm": 0.2764473971277821, | |
| "kl": 0.21224365234375, | |
| "learning_rate": 7.70718294432646e-06, | |
| "loss": 0.0085, | |
| "reward": 0.11875000279396772, | |
| "reward_std": 0.10311973933130503, | |
| "rewards/accuracy_reward": 0.11875000279396772, | |
| "rewards/format_reward": 0.0, | |
| "step": 930 | |
| }, | |
| { | |
| "completion_length": 549.8187686920166, | |
| "epoch": 0.6195386590466725, | |
| "grad_norm": 1.0463426634032484, | |
| "kl": 0.55484619140625, | |
| "learning_rate": 7.594750436337467e-06, | |
| "loss": 0.0222, | |
| "reward": 0.15000000335276126, | |
| "reward_std": 0.13552880026400088, | |
| "rewards/accuracy_reward": 0.15000000335276126, | |
| "rewards/format_reward": 0.0, | |
| "step": 935 | |
| }, | |
| { | |
| "completion_length": 557.4937660217286, | |
| "epoch": 0.6228517000041413, | |
| "grad_norm": 0.18999745413152938, | |
| "kl": 0.24376220703125, | |
| "learning_rate": 7.482639735370536e-06, | |
| "loss": 0.0098, | |
| "reward": 0.15833333767950536, | |
| "reward_std": 0.1178511306643486, | |
| "rewards/accuracy_reward": 0.15833333767950536, | |
| "rewards/format_reward": 0.0, | |
| "step": 940 | |
| }, | |
| { | |
| "completion_length": 565.875016784668, | |
| "epoch": 0.6261647409616101, | |
| "grad_norm": 0.2942645244909035, | |
| "kl": 0.235906982421875, | |
| "learning_rate": 7.37086584112108e-06, | |
| "loss": 0.0094, | |
| "reward": 0.1416666705161333, | |
| "reward_std": 0.15320646967738866, | |
| "rewards/accuracy_reward": 0.1416666705161333, | |
| "rewards/format_reward": 0.0, | |
| "step": 945 | |
| }, | |
| { | |
| "completion_length": 515.4104309082031, | |
| "epoch": 0.629477781919079, | |
| "grad_norm": 0.284453920534699, | |
| "kl": 0.207122802734375, | |
| "learning_rate": 7.2594437082219074e-06, | |
| "loss": 0.0083, | |
| "reward": 0.13333333656191826, | |
| "reward_std": 0.10606601741164923, | |
| "rewards/accuracy_reward": 0.13333333656191826, | |
| "rewards/format_reward": 0.0, | |
| "step": 950 | |
| }, | |
| { | |
| "completion_length": 519.2750148773193, | |
| "epoch": 0.6327908228765479, | |
| "grad_norm": 0.2551459658462794, | |
| "kl": 0.213092041015625, | |
| "learning_rate": 7.148388244242414e-06, | |
| "loss": 0.0085, | |
| "reward": 0.1354166703298688, | |
| "reward_std": 0.11490485239773988, | |
| "rewards/accuracy_reward": 0.1354166703298688, | |
| "rewards/format_reward": 0.0, | |
| "step": 955 | |
| }, | |
| { | |
| "completion_length": 551.9812677383422, | |
| "epoch": 0.6361038638340166, | |
| "grad_norm": 0.23334554711541383, | |
| "kl": 0.188629150390625, | |
| "learning_rate": 7.037714307694038e-06, | |
| "loss": 0.0075, | |
| "reward": 0.11250000335276127, | |
| "reward_std": 0.08838834799826145, | |
| "rewards/accuracy_reward": 0.11250000335276127, | |
| "rewards/format_reward": 0.0, | |
| "step": 960 | |
| }, | |
| { | |
| "completion_length": 575.2812675476074, | |
| "epoch": 0.6394169047914855, | |
| "grad_norm": 0.21745310342706664, | |
| "kl": 0.201898193359375, | |
| "learning_rate": 6.927436706042276e-06, | |
| "loss": 0.0081, | |
| "reward": 0.12708333618938922, | |
| "reward_std": 0.10311973933130503, | |
| "rewards/accuracy_reward": 0.12708333618938922, | |
| "rewards/format_reward": 0.0, | |
| "step": 965 | |
| }, | |
| { | |
| "completion_length": 544.3521011352539, | |
| "epoch": 0.6427299457489544, | |
| "grad_norm": 0.21392354942012623, | |
| "kl": 0.193035888671875, | |
| "learning_rate": 6.8175701937255645e-06, | |
| "loss": 0.0077, | |
| "reward": 0.1395833369344473, | |
| "reward_std": 0.10901229586452246, | |
| "rewards/accuracy_reward": 0.1395833369344473, | |
| "rewards/format_reward": 0.0, | |
| "step": 970 | |
| }, | |
| { | |
| "completion_length": 594.6437686920166, | |
| "epoch": 0.6460429867064231, | |
| "grad_norm": 0.3297497458421446, | |
| "kl": 0.20833740234375, | |
| "learning_rate": 6.708129470181197e-06, | |
| "loss": 0.0083, | |
| "reward": 0.1541666707023978, | |
| "reward_std": 0.14731391314417125, | |
| "rewards/accuracy_reward": 0.1541666707023978, | |
| "rewards/format_reward": 0.0, | |
| "step": 975 | |
| }, | |
| { | |
| "completion_length": 612.2333488464355, | |
| "epoch": 0.649356027663892, | |
| "grad_norm": 0.25579146126693203, | |
| "kl": 0.24373779296875, | |
| "learning_rate": 6.5991291778786556e-06, | |
| "loss": 0.0097, | |
| "reward": 0.12291666977107525, | |
| "reward_std": 0.07954951319843531, | |
| "rewards/accuracy_reward": 0.12291666977107525, | |
| "rewards/format_reward": 0.0, | |
| "step": 980 | |
| }, | |
| { | |
| "completion_length": 593.9500152587891, | |
| "epoch": 0.6526690686213609, | |
| "grad_norm": 0.308305547519083, | |
| "kl": 0.262152099609375, | |
| "learning_rate": 6.490583900360543e-06, | |
| "loss": 0.0105, | |
| "reward": 0.11875000260770321, | |
| "reward_std": 0.1207974087446928, | |
| "rewards/accuracy_reward": 0.11875000260770321, | |
| "rewards/format_reward": 0.0, | |
| "step": 985 | |
| }, | |
| { | |
| "completion_length": 569.4208499908448, | |
| "epoch": 0.6559821095788296, | |
| "grad_norm": 0.3215020178086238, | |
| "kl": 0.2141845703125, | |
| "learning_rate": 6.38250816029139e-06, | |
| "loss": 0.0086, | |
| "reward": 0.11458333637565374, | |
| "reward_std": 0.10901229586452246, | |
| "rewards/accuracy_reward": 0.11458333637565374, | |
| "rewards/format_reward": 0.0, | |
| "step": 990 | |
| }, | |
| { | |
| "completion_length": 569.345849609375, | |
| "epoch": 0.6592951505362985, | |
| "grad_norm": 0.3323215086416121, | |
| "kl": 0.219927978515625, | |
| "learning_rate": 6.274916417514605e-06, | |
| "loss": 0.0088, | |
| "reward": 0.13125000353902577, | |
| "reward_std": 0.1384750785306096, | |
| "rewards/accuracy_reward": 0.13125000353902577, | |
| "rewards/format_reward": 0.0, | |
| "step": 995 | |
| }, | |
| { | |
| "completion_length": 611.7208526611328, | |
| "epoch": 0.6626081914937674, | |
| "grad_norm": 0.3161012671184911, | |
| "kl": 0.3405517578125, | |
| "learning_rate": 6.167823067117868e-06, | |
| "loss": 0.0136, | |
| "reward": 0.08958333600312471, | |
| "reward_std": 0.09722718279808759, | |
| "rewards/accuracy_reward": 0.08958333600312471, | |
| "rewards/format_reward": 0.0, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.6626081914937674, | |
| "eval_completion_length": 612.1666834214154, | |
| "eval_kl": 0.27849264705882354, | |
| "eval_loss": 0.01135108433663845, | |
| "eval_reward": 0.11764706145314609, | |
| "eval_reward_std": 0.11091871077523512, | |
| "eval_rewards/accuracy_reward": 0.11764706145314609, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_runtime": 87.2908, | |
| "eval_samples_per_second": 1.134, | |
| "eval_steps_per_second": 0.103, | |
| "step": 1000 | |
| }, | |
| { | |
| "completion_length": 591.775019454956, | |
| "epoch": 0.6659212324512361, | |
| "grad_norm": 0.32740344668707216, | |
| "kl": 0.218963623046875, | |
| "learning_rate": 6.061242437507131e-06, | |
| "loss": 0.0088, | |
| "reward": 0.1104166692122817, | |
| "reward_std": 0.10311973933130503, | |
| "rewards/accuracy_reward": 0.1104166692122817, | |
| "rewards/format_reward": 0.0, | |
| "step": 1005 | |
| }, | |
| { | |
| "completion_length": 578.4812664031982, | |
| "epoch": 0.669234273408705, | |
| "grad_norm": 0.2198210321894444, | |
| "kl": 0.179779052734375, | |
| "learning_rate": 5.955188788489583e-06, | |
| "loss": 0.0072, | |
| "reward": 0.1729166705161333, | |
| "reward_std": 0.12668996527791024, | |
| "rewards/accuracy_reward": 0.1729166705161333, | |
| "rewards/format_reward": 0.0, | |
| "step": 1010 | |
| }, | |
| { | |
| "completion_length": 633.9875148773193, | |
| "epoch": 0.6725473143661739, | |
| "grad_norm": 0.24817253204515607, | |
| "kl": 0.180908203125, | |
| "learning_rate": 5.849676309365786e-06, | |
| "loss": 0.0072, | |
| "reward": 0.1395833369344473, | |
| "reward_std": 0.09722718279808759, | |
| "rewards/accuracy_reward": 0.1395833369344473, | |
| "rewards/format_reward": 0.0, | |
| "step": 1015 | |
| }, | |
| { | |
| "completion_length": 625.5687683105468, | |
| "epoch": 0.6758603553236426, | |
| "grad_norm": 0.26537873271258117, | |
| "kl": 0.197027587890625, | |
| "learning_rate": 5.744719117031217e-06, | |
| "loss": 0.0079, | |
| "reward": 0.15833333767950536, | |
| "reward_std": 0.1178511306643486, | |
| "rewards/accuracy_reward": 0.15833333767950536, | |
| "rewards/format_reward": 0.0, | |
| "step": 1020 | |
| }, | |
| { | |
| "completion_length": 642.2458499908447, | |
| "epoch": 0.6791733962811115, | |
| "grad_norm": 0.2811132286389355, | |
| "kl": 0.227191162109375, | |
| "learning_rate": 5.6403312540875325e-06, | |
| "loss": 0.0091, | |
| "reward": 0.10416666939854621, | |
| "reward_std": 0.09428090453147889, | |
| "rewards/accuracy_reward": 0.10416666939854621, | |
| "rewards/format_reward": 0.0, | |
| "step": 1025 | |
| }, | |
| { | |
| "completion_length": 605.6250198364257, | |
| "epoch": 0.6824864372385804, | |
| "grad_norm": 0.29826246716685967, | |
| "kl": 0.1943115234375, | |
| "learning_rate": 5.536526686963762e-06, | |
| "loss": 0.0078, | |
| "reward": 0.14375000391155482, | |
| "reward_std": 0.12079740893095732, | |
| "rewards/accuracy_reward": 0.14375000391155482, | |
| "rewards/format_reward": 0.0, | |
| "step": 1030 | |
| }, | |
| { | |
| "completion_length": 572.9312675476074, | |
| "epoch": 0.6857994781960493, | |
| "grad_norm": 0.19368256939132425, | |
| "kl": 0.156805419921875, | |
| "learning_rate": 5.433319304047666e-06, | |
| "loss": 0.0063, | |
| "reward": 0.15208333730697632, | |
| "reward_std": 0.09722718261182309, | |
| "rewards/accuracy_reward": 0.15208333730697632, | |
| "rewards/format_reward": 0.0, | |
| "step": 1035 | |
| }, | |
| { | |
| "completion_length": 556.6062641143799, | |
| "epoch": 0.689112519153518, | |
| "grad_norm": 0.3992963343728779, | |
| "kl": 0.2367431640625, | |
| "learning_rate": 5.330722913827594e-06, | |
| "loss": 0.0095, | |
| "reward": 0.17291667088866233, | |
| "reward_std": 0.10311973914504051, | |
| "rewards/accuracy_reward": 0.17291667088866233, | |
| "rewards/format_reward": 0.0, | |
| "step": 1040 | |
| }, | |
| { | |
| "completion_length": 560.204182434082, | |
| "epoch": 0.6924255601109869, | |
| "grad_norm": 0.3434721576041372, | |
| "kl": 0.209393310546875, | |
| "learning_rate": 5.228751243044961e-06, | |
| "loss": 0.0084, | |
| "reward": 0.16041667051613331, | |
| "reward_std": 0.12079740893095732, | |
| "rewards/accuracy_reward": 0.16041667051613331, | |
| "rewards/format_reward": 0.0, | |
| "step": 1045 | |
| }, | |
| { | |
| "completion_length": 605.5666820526124, | |
| "epoch": 0.6957386010684558, | |
| "grad_norm": 0.2306193550339744, | |
| "kl": 0.222296142578125, | |
| "learning_rate": 5.127417934857718e-06, | |
| "loss": 0.0089, | |
| "reward": 0.13333333600312472, | |
| "reward_std": 0.123743686825037, | |
| "rewards/accuracy_reward": 0.13333333600312472, | |
| "rewards/format_reward": 0.0, | |
| "step": 1050 | |
| }, | |
| { | |
| "completion_length": 526.712515258789, | |
| "epoch": 0.6990516420259245, | |
| "grad_norm": 0.6065517464028967, | |
| "kl": 0.213092041015625, | |
| "learning_rate": 5.026736547014981e-06, | |
| "loss": 0.0085, | |
| "reward": 0.13958333674818277, | |
| "reward_std": 0.12079740893095732, | |
| "rewards/accuracy_reward": 0.13958333674818277, | |
| "rewards/format_reward": 0.0, | |
| "step": 1055 | |
| }, | |
| { | |
| "completion_length": 547.4145992279052, | |
| "epoch": 0.7023646829833934, | |
| "grad_norm": 0.28705779880249954, | |
| "kl": 0.23438720703125, | |
| "learning_rate": 4.926720550043089e-06, | |
| "loss": 0.0094, | |
| "reward": 0.1166666692122817, | |
| "reward_std": 0.08249579127877951, | |
| "rewards/accuracy_reward": 0.1166666692122817, | |
| "rewards/format_reward": 0.0, | |
| "step": 1060 | |
| }, | |
| { | |
| "completion_length": 534.8395999908447, | |
| "epoch": 0.7056777239408623, | |
| "grad_norm": 0.2921526699948002, | |
| "kl": 0.238226318359375, | |
| "learning_rate": 4.827383325443331e-06, | |
| "loss": 0.0095, | |
| "reward": 0.1479166703298688, | |
| "reward_std": 0.12079740893095732, | |
| "rewards/accuracy_reward": 0.1479166703298688, | |
| "rewards/format_reward": 0.0, | |
| "step": 1065 | |
| }, | |
| { | |
| "completion_length": 556.1479335784912, | |
| "epoch": 0.708990764898331, | |
| "grad_norm": 0.3462960302006907, | |
| "kl": 0.23875732421875, | |
| "learning_rate": 4.728738163901597e-06, | |
| "loss": 0.0095, | |
| "reward": 0.13541667014360428, | |
| "reward_std": 0.10311973933130503, | |
| "rewards/accuracy_reward": 0.13541667014360428, | |
| "rewards/format_reward": 0.0, | |
| "step": 1070 | |
| }, | |
| { | |
| "completion_length": 561.583348083496, | |
| "epoch": 0.7123038058557999, | |
| "grad_norm": 0.3220348313143619, | |
| "kl": 0.186767578125, | |
| "learning_rate": 4.630798263510162e-06, | |
| "loss": 0.0075, | |
| "reward": 0.14375000353902578, | |
| "reward_std": 0.12079740893095732, | |
| "rewards/accuracy_reward": 0.14375000353902578, | |
| "rewards/format_reward": 0.0, | |
| "step": 1075 | |
| }, | |
| { | |
| "completion_length": 574.2791835784913, | |
| "epoch": 0.7156168468132688, | |
| "grad_norm": 0.361410974713343, | |
| "kl": 0.313189697265625, | |
| "learning_rate": 4.533576728001858e-06, | |
| "loss": 0.0125, | |
| "reward": 0.1458333371207118, | |
| "reward_std": 0.10606601759791374, | |
| "rewards/accuracy_reward": 0.1458333371207118, | |
| "rewards/format_reward": 0.0, | |
| "step": 1080 | |
| }, | |
| { | |
| "completion_length": 584.5437686920166, | |
| "epoch": 0.7189298877707375, | |
| "grad_norm": 0.3433650794021713, | |
| "kl": 0.16435546875, | |
| "learning_rate": 4.437086564996891e-06, | |
| "loss": 0.0066, | |
| "reward": 0.1479166705161333, | |
| "reward_std": 0.12079740893095732, | |
| "rewards/accuracy_reward": 0.1479166705161333, | |
| "rewards/format_reward": 0.0, | |
| "step": 1085 | |
| }, | |
| { | |
| "completion_length": 589.4520965576172, | |
| "epoch": 0.7222429287282064, | |
| "grad_norm": 0.28664499371278185, | |
| "kl": 0.362109375, | |
| "learning_rate": 4.341340684262498e-06, | |
| "loss": 0.0145, | |
| "reward": 0.16666667070239782, | |
| "reward_std": 0.12963624354451894, | |
| "rewards/accuracy_reward": 0.16666667070239782, | |
| "rewards/format_reward": 0.0, | |
| "step": 1090 | |
| }, | |
| { | |
| "completion_length": 589.1021018981934, | |
| "epoch": 0.7255559696856753, | |
| "grad_norm": 0.6726902176410475, | |
| "kl": 0.17657470703125, | |
| "learning_rate": 4.246351895985702e-06, | |
| "loss": 0.0071, | |
| "reward": 0.1687500037252903, | |
| "reward_std": 0.14436763487756252, | |
| "rewards/accuracy_reward": 0.1687500037252903, | |
| "rewards/format_reward": 0.0, | |
| "step": 1095 | |
| }, | |
| { | |
| "completion_length": 608.1708499908448, | |
| "epoch": 0.728869010643144, | |
| "grad_norm": 0.19762245990532015, | |
| "kl": 0.151568603515625, | |
| "learning_rate": 4.152132909059402e-06, | |
| "loss": 0.0061, | |
| "reward": 0.15000000316649675, | |
| "reward_std": 0.10017346087843179, | |
| "rewards/accuracy_reward": 0.15000000316649675, | |
| "rewards/format_reward": 0.0, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.728869010643144, | |
| "eval_completion_length": 567.0490381577436, | |
| "eval_kl": 0.16104664522058823, | |
| "eval_loss": 0.006509152241051197, | |
| "eval_reward": 0.18627451579360402, | |
| "eval_reward_std": 0.1386483881403418, | |
| "eval_rewards/accuracy_reward": 0.18627451579360402, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_runtime": 85.6746, | |
| "eval_samples_per_second": 1.156, | |
| "eval_steps_per_second": 0.105, | |
| "step": 1100 | |
| }, | |
| { | |
| "completion_length": 586.3041831970215, | |
| "epoch": 0.7321820516006129, | |
| "grad_norm": 0.26497157645231056, | |
| "kl": 0.153692626953125, | |
| "learning_rate": 4.058696329381987e-06, | |
| "loss": 0.0061, | |
| "reward": 0.22500000540167092, | |
| "reward_std": 0.11785113029181957, | |
| "rewards/accuracy_reward": 0.22500000540167092, | |
| "rewards/format_reward": 0.0, | |
| "step": 1105 | |
| }, | |
| { | |
| "completion_length": 640.6333503723145, | |
| "epoch": 0.7354950925580818, | |
| "grad_norm": 0.175788774023308, | |
| "kl": 0.155810546875, | |
| "learning_rate": 3.966054658170754e-06, | |
| "loss": 0.0062, | |
| "reward": 0.12916667014360428, | |
| "reward_std": 0.11195857413113117, | |
| "rewards/accuracy_reward": 0.12916667014360428, | |
| "rewards/format_reward": 0.0, | |
| "step": 1110 | |
| }, | |
| { | |
| "completion_length": 614.8416843414307, | |
| "epoch": 0.7388081335155506, | |
| "grad_norm": 0.39182645951451167, | |
| "kl": 0.19365234375, | |
| "learning_rate": 3.874220290289337e-06, | |
| "loss": 0.0077, | |
| "reward": 0.17500000391155482, | |
| "reward_std": 0.11785113047808408, | |
| "rewards/accuracy_reward": 0.17500000391155482, | |
| "rewards/format_reward": 0.0, | |
| "step": 1115 | |
| }, | |
| { | |
| "completion_length": 597.1479331970215, | |
| "epoch": 0.7421211744730194, | |
| "grad_norm": 0.3847191431848802, | |
| "kl": 0.201531982421875, | |
| "learning_rate": 3.7832055125893318e-06, | |
| "loss": 0.0081, | |
| "reward": 0.13750000316649674, | |
| "reward_std": 0.1178511306643486, | |
| "rewards/accuracy_reward": 0.13750000316649674, | |
| "rewards/format_reward": 0.0, | |
| "step": 1120 | |
| }, | |
| { | |
| "completion_length": 584.2437702178955, | |
| "epoch": 0.7454342154304883, | |
| "grad_norm": 0.42174504116546924, | |
| "kl": 0.158538818359375, | |
| "learning_rate": 3.6930225022664136e-06, | |
| "loss": 0.0063, | |
| "reward": 0.17500000409781932, | |
| "reward_std": 0.12374368719756604, | |
| "rewards/accuracy_reward": 0.17500000409781932, | |
| "rewards/format_reward": 0.0, | |
| "step": 1125 | |
| }, | |
| { | |
| "completion_length": 580.464596939087, | |
| "epoch": 0.7487472563879571, | |
| "grad_norm": 0.2445425181891289, | |
| "kl": 0.18485107421875, | |
| "learning_rate": 3.6036833252310887e-06, | |
| "loss": 0.0074, | |
| "reward": 0.18750000447034837, | |
| "reward_std": 0.12374368719756604, | |
| "rewards/accuracy_reward": 0.18750000447034837, | |
| "rewards/format_reward": 0.0, | |
| "step": 1130 | |
| }, | |
| { | |
| "completion_length": 580.1958534240723, | |
| "epoch": 0.7520602973454259, | |
| "grad_norm": 0.36465004329450484, | |
| "kl": 0.198504638671875, | |
| "learning_rate": 3.515199934494373e-06, | |
| "loss": 0.0079, | |
| "reward": 0.1625000037252903, | |
| "reward_std": 0.14142135679721832, | |
| "rewards/accuracy_reward": 0.1625000037252903, | |
| "rewards/format_reward": 0.0, | |
| "step": 1135 | |
| }, | |
| { | |
| "completion_length": 601.0125175476074, | |
| "epoch": 0.7553733383028948, | |
| "grad_norm": 0.2483965145552254, | |
| "kl": 0.170068359375, | |
| "learning_rate": 3.427584168568535e-06, | |
| "loss": 0.0068, | |
| "reward": 0.1687500037252903, | |
| "reward_std": 0.12079740893095732, | |
| "rewards/accuracy_reward": 0.1687500037252903, | |
| "rewards/format_reward": 0.0, | |
| "step": 1140 | |
| }, | |
| { | |
| "completion_length": 599.8041851043702, | |
| "epoch": 0.7586863792603636, | |
| "grad_norm": 0.21066990688347378, | |
| "kl": 0.18428955078125, | |
| "learning_rate": 3.3408477498831917e-06, | |
| "loss": 0.0074, | |
| "reward": 0.1333333369344473, | |
| "reward_std": 0.10606601759791374, | |
| "rewards/accuracy_reward": 0.1333333369344473, | |
| "rewards/format_reward": 0.0, | |
| "step": 1145 | |
| }, | |
| { | |
| "completion_length": 586.0396009445191, | |
| "epoch": 0.7619994202178324, | |
| "grad_norm": 0.28233525146520194, | |
| "kl": 0.170452880859375, | |
| "learning_rate": 3.2550022832169125e-06, | |
| "loss": 0.0068, | |
| "reward": 0.18125000465661287, | |
| "reward_std": 0.10311973933130503, | |
| "rewards/accuracy_reward": 0.18125000465661287, | |
| "rewards/format_reward": 0.0, | |
| "step": 1150 | |
| }, | |
| { | |
| "completion_length": 590.0333492279053, | |
| "epoch": 0.7653124611753013, | |
| "grad_norm": 0.3931119421182273, | |
| "kl": 0.184490966796875, | |
| "learning_rate": 3.170059254144593e-06, | |
| "loss": 0.0074, | |
| "reward": 0.17291667070239783, | |
| "reward_std": 0.1266899650916457, | |
| "rewards/accuracy_reward": 0.17291667070239783, | |
| "rewards/format_reward": 0.0, | |
| "step": 1155 | |
| }, | |
| { | |
| "completion_length": 609.7041851043701, | |
| "epoch": 0.7686255021327701, | |
| "grad_norm": 0.25191777564937673, | |
| "kl": 0.2096435546875, | |
| "learning_rate": 3.086030027500728e-06, | |
| "loss": 0.0084, | |
| "reward": 0.11666666995733976, | |
| "reward_std": 0.0766032349318266, | |
| "rewards/accuracy_reward": 0.11666666995733976, | |
| "rewards/format_reward": 0.0, | |
| "step": 1160 | |
| }, | |
| { | |
| "completion_length": 585.3791835784912, | |
| "epoch": 0.7719385430902389, | |
| "grad_norm": 0.48874888580232095, | |
| "kl": 0.167364501953125, | |
| "learning_rate": 3.002925845858905e-06, | |
| "loss": 0.0067, | |
| "reward": 0.19166667181998492, | |
| "reward_std": 0.14142135679721832, | |
| "rewards/accuracy_reward": 0.19166667181998492, | |
| "rewards/format_reward": 0.0, | |
| "step": 1165 | |
| }, | |
| { | |
| "completion_length": 603.406270980835, | |
| "epoch": 0.7752515840477078, | |
| "grad_norm": 0.31203818869127303, | |
| "kl": 0.171881103515625, | |
| "learning_rate": 2.920757828027586e-06, | |
| "loss": 0.0069, | |
| "reward": 0.16250000298023223, | |
| "reward_std": 0.11195857394486666, | |
| "rewards/accuracy_reward": 0.16250000298023223, | |
| "rewards/format_reward": 0.0, | |
| "step": 1170 | |
| }, | |
| { | |
| "completion_length": 597.1666839599609, | |
| "epoch": 0.7785646250051766, | |
| "grad_norm": 0.22045928574301496, | |
| "kl": 0.1704345703125, | |
| "learning_rate": 2.839536967562504e-06, | |
| "loss": 0.0068, | |
| "reward": 0.16250000428408384, | |
| "reward_std": 0.10606601759791374, | |
| "rewards/accuracy_reward": 0.16250000428408384, | |
| "rewards/format_reward": 0.0, | |
| "step": 1175 | |
| }, | |
| { | |
| "completion_length": 582.6396030426025, | |
| "epoch": 0.7818776659626455, | |
| "grad_norm": 0.31490753838871305, | |
| "kl": 0.1723388671875, | |
| "learning_rate": 2.759274131295787e-06, | |
| "loss": 0.0069, | |
| "reward": 0.17708333749324084, | |
| "reward_std": 0.14436763506382705, | |
| "rewards/accuracy_reward": 0.17708333749324084, | |
| "rewards/format_reward": 0.0, | |
| "step": 1180 | |
| }, | |
| { | |
| "completion_length": 607.9500186920166, | |
| "epoch": 0.7851907069201143, | |
| "grad_norm": 0.2500499737554532, | |
| "kl": 0.198590087890625, | |
| "learning_rate": 2.679980057882049e-06, | |
| "loss": 0.0079, | |
| "reward": 0.14791667107492684, | |
| "reward_std": 0.09722718279808759, | |
| "rewards/accuracy_reward": 0.14791667107492684, | |
| "rewards/format_reward": 0.0, | |
| "step": 1185 | |
| }, | |
| { | |
| "completion_length": 590.0312683105469, | |
| "epoch": 0.7885037478775831, | |
| "grad_norm": 0.3890063180094862, | |
| "kl": 0.21483154296875, | |
| "learning_rate": 2.60166535636162e-06, | |
| "loss": 0.0086, | |
| "reward": 0.12916667032986878, | |
| "reward_std": 0.12963624373078347, | |
| "rewards/accuracy_reward": 0.12916667032986878, | |
| "rewards/format_reward": 0.0, | |
| "step": 1190 | |
| }, | |
| { | |
| "completion_length": 592.3979331970215, | |
| "epoch": 0.791816788835052, | |
| "grad_norm": 0.34955868788134764, | |
| "kl": 0.20556640625, | |
| "learning_rate": 2.5243405047411353e-06, | |
| "loss": 0.0082, | |
| "reward": 0.15208333730697632, | |
| "reward_std": 0.13258252199739218, | |
| "rewards/accuracy_reward": 0.15208333730697632, | |
| "rewards/format_reward": 0.0, | |
| "step": 1195 | |
| }, | |
| { | |
| "completion_length": 602.2000190734864, | |
| "epoch": 0.7951298297925208, | |
| "grad_norm": 0.320862144828303, | |
| "kl": 0.1858642578125, | |
| "learning_rate": 2.448015848591638e-06, | |
| "loss": 0.0074, | |
| "reward": 0.16458333786576987, | |
| "reward_std": 0.16204530466347933, | |
| "rewards/accuracy_reward": 0.16458333786576987, | |
| "rewards/format_reward": 0.0, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.7951298297925208, | |
| "eval_completion_length": 595.4313946892233, | |
| "eval_kl": 0.16259765625, | |
| "eval_loss": 0.006604722701013088, | |
| "eval_reward": 0.20588235907694874, | |
| "eval_reward_std": 0.12478355011519264, | |
| "eval_rewards/accuracy_reward": 0.20588235907694874, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_runtime": 82.8847, | |
| "eval_samples_per_second": 1.194, | |
| "eval_steps_per_second": 0.109, | |
| "step": 1200 | |
| }, | |
| { | |
| "completion_length": 613.1854331970214, | |
| "epoch": 0.7984428707499897, | |
| "grad_norm": 0.2661889633504761, | |
| "kl": 0.18961181640625, | |
| "learning_rate": 2.3727015996644043e-06, | |
| "loss": 0.0076, | |
| "reward": 0.12708333656191825, | |
| "reward_std": 0.10311973933130503, | |
| "rewards/accuracy_reward": 0.12708333656191825, | |
| "rewards/format_reward": 0.0, | |
| "step": 1205 | |
| }, | |
| { | |
| "completion_length": 594.962519454956, | |
| "epoch": 0.8017559117074585, | |
| "grad_norm": 0.33195445099683457, | |
| "kl": 0.17481689453125, | |
| "learning_rate": 2.298407834524682e-06, | |
| "loss": 0.007, | |
| "reward": 0.16041667051613331, | |
| "reward_std": 0.12079740893095732, | |
| "rewards/accuracy_reward": 0.16041667051613331, | |
| "rewards/format_reward": 0.0, | |
| "step": 1210 | |
| }, | |
| { | |
| "completion_length": 562.3083518981933, | |
| "epoch": 0.8050689526649273, | |
| "grad_norm": 0.36379765983184786, | |
| "kl": 1.331353759765625, | |
| "learning_rate": 2.2251444932035094e-06, | |
| "loss": 0.0532, | |
| "reward": 0.1875000050291419, | |
| "reward_std": 0.13552880007773638, | |
| "rewards/accuracy_reward": 0.1875000050291419, | |
| "rewards/format_reward": 0.0, | |
| "step": 1215 | |
| }, | |
| { | |
| "completion_length": 535.5166809082032, | |
| "epoch": 0.8083819936223962, | |
| "grad_norm": 0.22606794243792797, | |
| "kl": 0.16956787109375, | |
| "learning_rate": 2.1529213778677993e-06, | |
| "loss": 0.0068, | |
| "reward": 0.1666666703298688, | |
| "reward_std": 0.10606601759791374, | |
| "rewards/accuracy_reward": 0.1666666703298688, | |
| "rewards/format_reward": 0.0, | |
| "step": 1220 | |
| }, | |
| { | |
| "completion_length": 559.0458503723145, | |
| "epoch": 0.811695034579865, | |
| "grad_norm": 0.3145793732636785, | |
| "kl": 0.15970458984375, | |
| "learning_rate": 2.081748151508883e-06, | |
| "loss": 0.0064, | |
| "reward": 0.18750000335276126, | |
| "reward_std": 0.10606601741164923, | |
| "rewards/accuracy_reward": 0.18750000335276126, | |
| "rewards/format_reward": 0.0, | |
| "step": 1225 | |
| }, | |
| { | |
| "completion_length": 546.4958501815796, | |
| "epoch": 0.8150080755373338, | |
| "grad_norm": 0.1573772953559393, | |
| "kl": 0.17611083984375, | |
| "learning_rate": 2.0116343366496493e-06, | |
| "loss": 0.007, | |
| "reward": 0.14791667070239783, | |
| "reward_std": 0.08544206973165273, | |
| "rewards/accuracy_reward": 0.14791667070239783, | |
| "rewards/format_reward": 0.0, | |
| "step": 1230 | |
| }, | |
| { | |
| "completion_length": 549.3166814804077, | |
| "epoch": 0.8183211164948027, | |
| "grad_norm": 0.45017301026338924, | |
| "kl": 0.1666015625, | |
| "learning_rate": 1.942589314070494e-06, | |
| "loss": 0.0067, | |
| "reward": 0.18541667126119138, | |
| "reward_std": 0.12668996546417474, | |
| "rewards/accuracy_reward": 0.18541667126119138, | |
| "rewards/format_reward": 0.0, | |
| "step": 1235 | |
| }, | |
| { | |
| "completion_length": 595.6729320526123, | |
| "epoch": 0.8216341574522715, | |
| "grad_norm": 0.2792656285293251, | |
| "kl": 0.1830078125, | |
| "learning_rate": 1.8746223215542482e-06, | |
| "loss": 0.0073, | |
| "reward": 0.16041666977107524, | |
| "reward_std": 0.10901229567825794, | |
| "rewards/accuracy_reward": 0.16041666977107524, | |
| "rewards/format_reward": 0.0, | |
| "step": 1240 | |
| }, | |
| { | |
| "completion_length": 589.0687671661377, | |
| "epoch": 0.8249471984097403, | |
| "grad_norm": 0.34000796378112624, | |
| "kl": 0.1674072265625, | |
| "learning_rate": 1.8077424526501964e-06, | |
| "loss": 0.0067, | |
| "reward": 0.16875000428408385, | |
| "reward_std": 0.15026019159704446, | |
| "rewards/accuracy_reward": 0.16875000428408385, | |
| "rewards/format_reward": 0.0, | |
| "step": 1245 | |
| }, | |
| { | |
| "completion_length": 590.1375190734864, | |
| "epoch": 0.8282602393672092, | |
| "grad_norm": 0.1827720037966532, | |
| "kl": 0.1564697265625, | |
| "learning_rate": 1.7419586554574364e-06, | |
| "loss": 0.0063, | |
| "reward": 0.15833333674818278, | |
| "reward_std": 0.11195857413113117, | |
| "rewards/accuracy_reward": 0.15833333674818278, | |
| "rewards/format_reward": 0.0, | |
| "step": 1250 | |
| }, | |
| { | |
| "completion_length": 547.691682434082, | |
| "epoch": 0.831573280324678, | |
| "grad_norm": 0.2333616789960661, | |
| "kl": 0.148370361328125, | |
| "learning_rate": 1.6772797314276712e-06, | |
| "loss": 0.0059, | |
| "reward": 0.18750000428408384, | |
| "reward_std": 0.12963624373078347, | |
| "rewards/accuracy_reward": 0.18750000428408384, | |
| "rewards/format_reward": 0.0, | |
| "step": 1255 | |
| }, | |
| { | |
| "completion_length": 590.2541835784912, | |
| "epoch": 0.8348863212821469, | |
| "grad_norm": 0.4458824925413339, | |
| "kl": 0.154559326171875, | |
| "learning_rate": 1.6137143341876439e-06, | |
| "loss": 0.0062, | |
| "reward": 0.19166667051613331, | |
| "reward_std": 0.12374368701130152, | |
| "rewards/accuracy_reward": 0.19166667051613331, | |
| "rewards/format_reward": 0.0, | |
| "step": 1260 | |
| }, | |
| { | |
| "completion_length": 578.4666858673096, | |
| "epoch": 0.8381993622396157, | |
| "grad_norm": 0.2507774814936557, | |
| "kl": 0.16309814453125, | |
| "learning_rate": 1.5512709683813165e-06, | |
| "loss": 0.0065, | |
| "reward": 0.17083333767950534, | |
| "reward_std": 0.11195857413113117, | |
| "rewards/accuracy_reward": 0.17083333767950534, | |
| "rewards/format_reward": 0.0, | |
| "step": 1265 | |
| }, | |
| { | |
| "completion_length": 578.7125144958496, | |
| "epoch": 0.8415124031970845, | |
| "grad_norm": 0.29941656086318214, | |
| "kl": 0.2466796875, | |
| "learning_rate": 1.4899579885320237e-06, | |
| "loss": 0.0099, | |
| "reward": 0.1958333371207118, | |
| "reward_std": 0.12374368719756604, | |
| "rewards/accuracy_reward": 0.1958333371207118, | |
| "rewards/format_reward": 0.0, | |
| "step": 1270 | |
| }, | |
| { | |
| "completion_length": 611.9833530426025, | |
| "epoch": 0.8448254441545534, | |
| "grad_norm": 0.5120540878644168, | |
| "kl": 0.161627197265625, | |
| "learning_rate": 1.4297835979246777e-06, | |
| "loss": 0.0065, | |
| "reward": 0.11875000316649675, | |
| "reward_std": 0.09722718279808759, | |
| "rewards/accuracy_reward": 0.11875000316649675, | |
| "rewards/format_reward": 0.0, | |
| "step": 1275 | |
| }, | |
| { | |
| "completion_length": 554.3375129699707, | |
| "epoch": 0.8481384851120222, | |
| "grad_norm": 0.24272356502293516, | |
| "kl": 0.166864013671875, | |
| "learning_rate": 1.370755847508226e-06, | |
| "loss": 0.0067, | |
| "reward": 0.1750000048428774, | |
| "reward_std": 0.12374368719756604, | |
| "rewards/accuracy_reward": 0.1750000048428774, | |
| "rewards/format_reward": 0.0, | |
| "step": 1280 | |
| }, | |
| { | |
| "completion_length": 566.3208484649658, | |
| "epoch": 0.851451526069491, | |
| "grad_norm": 0.33649577414263443, | |
| "kl": 0.176239013671875, | |
| "learning_rate": 1.3128826348184886e-06, | |
| "loss": 0.007, | |
| "reward": 0.21458333749324082, | |
| "reward_std": 0.15615274775773286, | |
| "rewards/accuracy_reward": 0.21458333749324082, | |
| "rewards/format_reward": 0.0, | |
| "step": 1285 | |
| }, | |
| { | |
| "completion_length": 570.1416812896729, | |
| "epoch": 0.8547645670269599, | |
| "grad_norm": 0.2521899833365297, | |
| "kl": 0.175726318359375, | |
| "learning_rate": 1.256171702921516e-06, | |
| "loss": 0.007, | |
| "reward": 0.1979166716337204, | |
| "reward_std": 0.13847507834434508, | |
| "rewards/accuracy_reward": 0.1979166716337204, | |
| "rewards/format_reward": 0.0, | |
| "step": 1290 | |
| }, | |
| { | |
| "completion_length": 585.5687664031982, | |
| "epoch": 0.8580776079844287, | |
| "grad_norm": 0.26502078855241107, | |
| "kl": 0.19686279296875, | |
| "learning_rate": 1.200630639377609e-06, | |
| "loss": 0.0079, | |
| "reward": 0.15625000316649676, | |
| "reward_std": 0.10901229567825794, | |
| "rewards/accuracy_reward": 0.15625000316649676, | |
| "rewards/format_reward": 0.0, | |
| "step": 1295 | |
| }, | |
| { | |
| "completion_length": 595.7041839599609, | |
| "epoch": 0.8613906489418975, | |
| "grad_norm": 0.24839009821426125, | |
| "kl": 0.20594482421875, | |
| "learning_rate": 1.1462668752261652e-06, | |
| "loss": 0.0082, | |
| "reward": 0.17083333767950534, | |
| "reward_std": 0.13552880007773638, | |
| "rewards/accuracy_reward": 0.17083333767950534, | |
| "rewards/format_reward": 0.0, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.8613906489418975, | |
| "eval_completion_length": 596.6029609231388, | |
| "eval_kl": 0.24149816176470587, | |
| "eval_loss": 0.009774969890713692, | |
| "eval_reward": 0.13235294380608728, | |
| "eval_reward_std": 0.09012145286097246, | |
| "eval_rewards/accuracy_reward": 0.13235294380608728, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_runtime": 84.0185, | |
| "eval_samples_per_second": 1.178, | |
| "eval_steps_per_second": 0.107, | |
| "step": 1300 | |
| }, | |
| { | |
| "completion_length": 579.6208499908447, | |
| "epoch": 0.8647036898993664, | |
| "grad_norm": 0.23156140004960932, | |
| "kl": 0.208966064453125, | |
| "learning_rate": 1.0930876839914418e-06, | |
| "loss": 0.0084, | |
| "reward": 0.16458333730697633, | |
| "reward_std": 0.12668996527791024, | |
| "rewards/accuracy_reward": 0.16458333730697633, | |
| "rewards/format_reward": 0.0, | |
| "step": 1305 | |
| }, | |
| { | |
| "completion_length": 581.4416831970215, | |
| "epoch": 0.8680167308568352, | |
| "grad_norm": 0.28452284185332194, | |
| "kl": 0.2054443359375, | |
| "learning_rate": 1.04110018070941e-06, | |
| "loss": 0.0082, | |
| "reward": 0.16666667088866233, | |
| "reward_std": 0.12374368719756604, | |
| "rewards/accuracy_reward": 0.16666667088866233, | |
| "rewards/format_reward": 0.0, | |
| "step": 1310 | |
| }, | |
| { | |
| "completion_length": 597.6521030426026, | |
| "epoch": 0.871329771814304, | |
| "grad_norm": 0.2454825397955897, | |
| "kl": 0.21685791015625, | |
| "learning_rate": 9.903113209758098e-07, | |
| "loss": 0.0087, | |
| "reward": 0.11458333618938923, | |
| "reward_std": 0.09133462607860565, | |
| "rewards/accuracy_reward": 0.11458333618938923, | |
| "rewards/format_reward": 0.0, | |
| "step": 1315 | |
| }, | |
| { | |
| "completion_length": 571.9479333877564, | |
| "epoch": 0.8746428127717729, | |
| "grad_norm": 0.3731994534971755, | |
| "kl": 0.22552490234375, | |
| "learning_rate": 9.407279000155311e-07, | |
| "loss": 0.009, | |
| "reward": 0.15000000335276126, | |
| "reward_std": 0.09428090434521437, | |
| "rewards/accuracy_reward": 0.15000000335276126, | |
| "rewards/format_reward": 0.0, | |
| "step": 1320 | |
| }, | |
| { | |
| "completion_length": 599.3375160217286, | |
| "epoch": 0.8779558537292417, | |
| "grad_norm": 0.312791764382651, | |
| "kl": 0.19879150390625, | |
| "learning_rate": 8.923565517734633e-07, | |
| "loss": 0.0079, | |
| "reward": 0.10416666958481073, | |
| "reward_std": 0.07071067839860916, | |
| "rewards/accuracy_reward": 0.10416666958481073, | |
| "rewards/format_reward": 0.0, | |
| "step": 1325 | |
| }, | |
| { | |
| "completion_length": 568.7083507537842, | |
| "epoch": 0.8812688946867105, | |
| "grad_norm": 0.25551838738681937, | |
| "kl": 0.21905517578125, | |
| "learning_rate": 8.452037480269082e-07, | |
| "loss": 0.0088, | |
| "reward": 0.20000000447034835, | |
| "reward_std": 0.11195857413113117, | |
| "rewards/accuracy_reward": 0.20000000447034835, | |
| "rewards/format_reward": 0.0, | |
| "step": 1330 | |
| }, | |
| { | |
| "completion_length": 544.368769454956, | |
| "epoch": 0.8845819356441794, | |
| "grad_norm": 0.26604711389347285, | |
| "kl": 0.16533203125, | |
| "learning_rate": 7.992757975196974e-07, | |
| "loss": 0.0066, | |
| "reward": 0.1708333371207118, | |
| "reward_std": 0.10606601759791374, | |
| "rewards/accuracy_reward": 0.1708333371207118, | |
| "rewards/format_reward": 0.0, | |
| "step": 1335 | |
| }, | |
| { | |
| "completion_length": 561.4187679290771, | |
| "epoch": 0.8878949766016483, | |
| "grad_norm": 0.3433532227955832, | |
| "kl": 0.174151611328125, | |
| "learning_rate": 7.545788451181313e-07, | |
| "loss": 0.007, | |
| "reward": 0.17916667070239783, | |
| "reward_std": 0.12963624373078347, | |
| "rewards/accuracy_reward": 0.17916667070239783, | |
| "rewards/format_reward": 0.0, | |
| "step": 1340 | |
| }, | |
| { | |
| "completion_length": 574.2041835784912, | |
| "epoch": 0.891208017559117, | |
| "grad_norm": 0.271754666321704, | |
| "kl": 0.160504150390625, | |
| "learning_rate": 7.11118870988825e-07, | |
| "loss": 0.0064, | |
| "reward": 0.13750000298023224, | |
| "reward_std": 0.10017346106469631, | |
| "rewards/accuracy_reward": 0.13750000298023224, | |
| "rewards/format_reward": 0.0, | |
| "step": 1345 | |
| }, | |
| { | |
| "completion_length": 567.6791843414306, | |
| "epoch": 0.8945210585165859, | |
| "grad_norm": 0.3130447466581642, | |
| "kl": 0.164306640625, | |
| "learning_rate": 6.689016897986123e-07, | |
| "loss": 0.0066, | |
| "reward": 0.14166666977107525, | |
| "reward_std": 0.1178511306643486, | |
| "rewards/accuracy_reward": 0.14166666977107525, | |
| "rewards/format_reward": 0.0, | |
| "step": 1350 | |
| }, | |
| { | |
| "completion_length": 585.5437667846679, | |
| "epoch": 0.8978340994740548, | |
| "grad_norm": 0.19957089977636847, | |
| "kl": 0.1621734619140625, | |
| "learning_rate": 6.279329499365649e-07, | |
| "loss": 0.0065, | |
| "reward": 0.17500000447034836, | |
| "reward_std": 0.10017346106469631, | |
| "rewards/accuracy_reward": 0.17500000447034836, | |
| "rewards/format_reward": 0.0, | |
| "step": 1355 | |
| }, | |
| { | |
| "completion_length": 568.7625167846679, | |
| "epoch": 0.9011471404315236, | |
| "grad_norm": 0.24101026537901626, | |
| "kl": 0.144744873046875, | |
| "learning_rate": 5.88218132758287e-07, | |
| "loss": 0.0058, | |
| "reward": 0.15833333730697632, | |
| "reward_std": 0.12963624335825444, | |
| "rewards/accuracy_reward": 0.15833333730697632, | |
| "rewards/format_reward": 0.0, | |
| "step": 1360 | |
| }, | |
| { | |
| "completion_length": 553.6021018981934, | |
| "epoch": 0.9044601813889924, | |
| "grad_norm": 0.31615520494124105, | |
| "kl": 0.160650634765625, | |
| "learning_rate": 5.497625518525374e-07, | |
| "loss": 0.0064, | |
| "reward": 0.12291666995733977, | |
| "reward_std": 0.10311973933130503, | |
| "rewards/accuracy_reward": 0.12291666995733977, | |
| "rewards/format_reward": 0.0, | |
| "step": 1365 | |
| }, | |
| { | |
| "completion_length": 563.3395992279053, | |
| "epoch": 0.9077732223464613, | |
| "grad_norm": 0.26385870637392084, | |
| "kl": 0.143408203125, | |
| "learning_rate": 5.125713523303133e-07, | |
| "loss": 0.0057, | |
| "reward": 0.19583333767950534, | |
| "reward_std": 0.1178511306643486, | |
| "rewards/accuracy_reward": 0.19583333767950534, | |
| "rewards/format_reward": 0.0, | |
| "step": 1370 | |
| }, | |
| { | |
| "completion_length": 583.402102279663, | |
| "epoch": 0.91108626330393, | |
| "grad_norm": 0.20653175003039265, | |
| "kl": 0.169732666015625, | |
| "learning_rate": 4.7664951013645875e-07, | |
| "loss": 0.0068, | |
| "reward": 0.15625000409781933, | |
| "reward_std": 0.10311973914504051, | |
| "rewards/accuracy_reward": 0.15625000409781933, | |
| "rewards/format_reward": 0.0, | |
| "step": 1375 | |
| }, | |
| { | |
| "completion_length": 576.2833488464355, | |
| "epoch": 0.9143993042613989, | |
| "grad_norm": 0.33444825058309613, | |
| "kl": 0.15838623046875, | |
| "learning_rate": 4.420018313839147e-07, | |
| "loss": 0.0063, | |
| "reward": 0.17708333767950535, | |
| "reward_std": 0.1384750785306096, | |
| "rewards/accuracy_reward": 0.17708333767950535, | |
| "rewards/format_reward": 0.0, | |
| "step": 1380 | |
| }, | |
| { | |
| "completion_length": 564.4771015167237, | |
| "epoch": 0.9177123452188678, | |
| "grad_norm": 0.38336632998772435, | |
| "kl": 0.175140380859375, | |
| "learning_rate": 4.086329517107046e-07, | |
| "loss": 0.007, | |
| "reward": 0.19583333786576987, | |
| "reward_std": 0.14731391333043575, | |
| "rewards/accuracy_reward": 0.19583333786576987, | |
| "rewards/format_reward": 0.0, | |
| "step": 1385 | |
| }, | |
| { | |
| "completion_length": 579.0646015167237, | |
| "epoch": 0.9210253861763366, | |
| "grad_norm": 0.2664468063435788, | |
| "kl": 0.169805908203125, | |
| "learning_rate": 3.7654733565969826e-07, | |
| "loss": 0.0068, | |
| "reward": 0.16041666977107524, | |
| "reward_std": 0.13258252162486314, | |
| "rewards/accuracy_reward": 0.16041666977107524, | |
| "rewards/format_reward": 0.0, | |
| "step": 1390 | |
| }, | |
| { | |
| "completion_length": 572.5937644958497, | |
| "epoch": 0.9243384271338054, | |
| "grad_norm": 0.29069246784999064, | |
| "kl": 0.172479248046875, | |
| "learning_rate": 3.457492760812975e-07, | |
| "loss": 0.0069, | |
| "reward": 0.20833333749324084, | |
| "reward_std": 0.10606601759791374, | |
| "rewards/accuracy_reward": 0.20833333749324084, | |
| "rewards/format_reward": 0.0, | |
| "step": 1395 | |
| }, | |
| { | |
| "completion_length": 582.8104347229004, | |
| "epoch": 0.9276514680912743, | |
| "grad_norm": 0.3338040728206735, | |
| "kl": 0.18824462890625, | |
| "learning_rate": 3.1624289355907334e-07, | |
| "loss": 0.0075, | |
| "reward": 0.18125000428408383, | |
| "reward_std": 0.12668996546417474, | |
| "rewards/accuracy_reward": 0.18125000428408383, | |
| "rewards/format_reward": 0.0, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.9276514680912743, | |
| "eval_completion_length": 557.0784409466912, | |
| "eval_kl": 0.15639361213235295, | |
| "eval_loss": 0.006327385548502207, | |
| "eval_reward": 0.1960784367778722, | |
| "eval_reward_std": 0.13864838857861125, | |
| "eval_rewards/accuracy_reward": 0.1960784367778722, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_runtime": 86.7655, | |
| "eval_samples_per_second": 1.141, | |
| "eval_steps_per_second": 0.104, | |
| "step": 1400 | |
| }, | |
| { | |
| "completion_length": 589.4166835784912, | |
| "epoch": 0.9309645090487431, | |
| "grad_norm": 0.353780019491935, | |
| "kl": 0.178558349609375, | |
| "learning_rate": 2.8803213585846036e-07, | |
| "loss": 0.0071, | |
| "reward": 0.16458333767950534, | |
| "reward_std": 0.11490485221147537, | |
| "rewards/accuracy_reward": 0.16458333767950534, | |
| "rewards/format_reward": 0.0, | |
| "step": 1405 | |
| }, | |
| { | |
| "completion_length": 579.3000144958496, | |
| "epoch": 0.9342775500062119, | |
| "grad_norm": 0.34672099101912396, | |
| "kl": 0.158709716796875, | |
| "learning_rate": 2.6112077739857465e-07, | |
| "loss": 0.0063, | |
| "reward": 0.15833333674818278, | |
| "reward_std": 0.12374368701130152, | |
| "rewards/accuracy_reward": 0.15833333674818278, | |
| "rewards/format_reward": 0.0, | |
| "step": 1410 | |
| }, | |
| { | |
| "completion_length": 578.6437679290772, | |
| "epoch": 0.9375905909636808, | |
| "grad_norm": 0.2557223102198384, | |
| "kl": 0.175360107421875, | |
| "learning_rate": 2.3551241874721353e-07, | |
| "loss": 0.007, | |
| "reward": 0.11666666977107525, | |
| "reward_std": 0.0766032349318266, | |
| "rewards/accuracy_reward": 0.11666666977107525, | |
| "rewards/format_reward": 0.0, | |
| "step": 1415 | |
| }, | |
| { | |
| "completion_length": 564.343769454956, | |
| "epoch": 0.9409036319211497, | |
| "grad_norm": 0.31991282109346414, | |
| "kl": 0.162542724609375, | |
| "learning_rate": 2.1121048613912843e-07, | |
| "loss": 0.0065, | |
| "reward": 0.19583333749324083, | |
| "reward_std": 0.14142135679721832, | |
| "rewards/accuracy_reward": 0.19583333749324083, | |
| "rewards/format_reward": 0.0, | |
| "step": 1420 | |
| }, | |
| { | |
| "completion_length": 574.3541858673095, | |
| "epoch": 0.9442166728786184, | |
| "grad_norm": 0.28826580975238403, | |
| "kl": 0.19622802734375, | |
| "learning_rate": 1.8821823101760949e-07, | |
| "loss": 0.0079, | |
| "reward": 0.1812500037252903, | |
| "reward_std": 0.12079740893095732, | |
| "rewards/accuracy_reward": 0.1812500037252903, | |
| "rewards/format_reward": 0.0, | |
| "step": 1425 | |
| }, | |
| { | |
| "completion_length": 562.9854351043701, | |
| "epoch": 0.9475297138360873, | |
| "grad_norm": 0.34971148204114416, | |
| "kl": 0.149298095703125, | |
| "learning_rate": 1.665387295994747e-07, | |
| "loss": 0.006, | |
| "reward": 0.1541666705161333, | |
| "reward_std": 0.1178511306643486, | |
| "rewards/accuracy_reward": 0.1541666705161333, | |
| "rewards/format_reward": 0.0, | |
| "step": 1430 | |
| }, | |
| { | |
| "completion_length": 562.6541831970214, | |
| "epoch": 0.9508427547935562, | |
| "grad_norm": 0.36546930503986563, | |
| "kl": 0.158831787109375, | |
| "learning_rate": 1.4617488246348012e-07, | |
| "loss": 0.0064, | |
| "reward": 0.20208333767950534, | |
| "reward_std": 0.13258252199739218, | |
| "rewards/accuracy_reward": 0.20208333767950534, | |
| "rewards/format_reward": 0.0, | |
| "step": 1435 | |
| }, | |
| { | |
| "completion_length": 590.1208526611329, | |
| "epoch": 0.9541557957510249, | |
| "grad_norm": 0.208162230921283, | |
| "kl": 0.180303955078125, | |
| "learning_rate": 1.271294141622459e-07, | |
| "loss": 0.0072, | |
| "reward": 0.1395833369344473, | |
| "reward_std": 0.08544206973165273, | |
| "rewards/accuracy_reward": 0.1395833369344473, | |
| "rewards/format_reward": 0.0, | |
| "step": 1440 | |
| }, | |
| { | |
| "completion_length": 587.8396018981933, | |
| "epoch": 0.9574688367084938, | |
| "grad_norm": 0.2594006165397625, | |
| "kl": 0.173162841796875, | |
| "learning_rate": 1.094048728577346e-07, | |
| "loss": 0.0069, | |
| "reward": 0.14583333674818277, | |
| "reward_std": 0.10017346106469631, | |
| "rewards/accuracy_reward": 0.14583333674818277, | |
| "rewards/format_reward": 0.0, | |
| "step": 1445 | |
| }, | |
| { | |
| "completion_length": 567.8000190734863, | |
| "epoch": 0.9607818776659627, | |
| "grad_norm": 0.32429036574157794, | |
| "kl": 0.1535888671875, | |
| "learning_rate": 9.300362998030832e-08, | |
| "loss": 0.0061, | |
| "reward": 0.17291667070239783, | |
| "reward_std": 0.12668996546417474, | |
| "rewards/accuracy_reward": 0.17291667070239783, | |
| "rewards/format_reward": 0.0, | |
| "step": 1450 | |
| }, | |
| { | |
| "completion_length": 565.8521003723145, | |
| "epoch": 0.9640949186234314, | |
| "grad_norm": 0.2978650282959838, | |
| "kl": 0.150830078125, | |
| "learning_rate": 7.792787991146356e-08, | |
| "loss": 0.006, | |
| "reward": 0.1958333382382989, | |
| "reward_std": 0.11195857394486666, | |
| "rewards/accuracy_reward": 0.1958333382382989, | |
| "rewards/format_reward": 0.0, | |
| "step": 1455 | |
| }, | |
| { | |
| "completion_length": 561.8375183105469, | |
| "epoch": 0.9674079595809003, | |
| "grad_norm": 0.2897071588264315, | |
| "kl": 0.1702880859375, | |
| "learning_rate": 6.417963969022389e-08, | |
| "loss": 0.0068, | |
| "reward": 0.21250000540167094, | |
| "reward_std": 0.13552880026400088, | |
| "rewards/accuracy_reward": 0.21250000540167094, | |
| "rewards/format_reward": 0.0, | |
| "step": 1460 | |
| }, | |
| { | |
| "completion_length": 586.7416843414306, | |
| "epoch": 0.9707210005383692, | |
| "grad_norm": 0.38925090389086237, | |
| "kl": 0.15640869140625, | |
| "learning_rate": 5.176074874327919e-08, | |
| "loss": 0.0063, | |
| "reward": 0.18541667088866234, | |
| "reward_std": 0.1384750785306096, | |
| "rewards/accuracy_reward": 0.18541667088866234, | |
| "rewards/format_reward": 0.0, | |
| "step": 1465 | |
| }, | |
| { | |
| "completion_length": 561.2791835784913, | |
| "epoch": 0.9740340414958379, | |
| "grad_norm": 0.34721664371729727, | |
| "kl": 0.1898681640625, | |
| "learning_rate": 4.067286863888131e-08, | |
| "loss": 0.0076, | |
| "reward": 0.1895833384245634, | |
| "reward_std": 0.15026019141077995, | |
| "rewards/accuracy_reward": 0.1895833384245634, | |
| "rewards/format_reward": 0.0, | |
| "step": 1470 | |
| }, | |
| { | |
| "completion_length": 590.7000167846679, | |
| "epoch": 0.9773470824533068, | |
| "grad_norm": 0.2964515178540727, | |
| "kl": 0.168414306640625, | |
| "learning_rate": 3.091748286453866e-08, | |
| "loss": 0.0067, | |
| "reward": 0.1541666707023978, | |
| "reward_std": 0.13552880026400088, | |
| "rewards/accuracy_reward": 0.1541666707023978, | |
| "rewards/format_reward": 0.0, | |
| "step": 1475 | |
| }, | |
| { | |
| "completion_length": 567.4812641143799, | |
| "epoch": 0.9806601234107757, | |
| "grad_norm": 0.22436357518659494, | |
| "kl": 0.182574462890625, | |
| "learning_rate": 2.2495896628529355e-08, | |
| "loss": 0.0073, | |
| "reward": 0.1833333382382989, | |
| "reward_std": 0.12374368719756604, | |
| "rewards/accuracy_reward": 0.1833333382382989, | |
| "rewards/format_reward": 0.0, | |
| "step": 1480 | |
| }, | |
| { | |
| "completion_length": 576.3062686920166, | |
| "epoch": 0.9839731643682446, | |
| "grad_norm": 0.2682867728590303, | |
| "kl": 0.160321044921875, | |
| "learning_rate": 1.5409236685277608e-08, | |
| "loss": 0.0064, | |
| "reward": 0.15625000335276126, | |
| "reward_std": 0.13258252199739218, | |
| "rewards/accuracy_reward": 0.15625000335276126, | |
| "rewards/format_reward": 0.0, | |
| "step": 1485 | |
| }, | |
| { | |
| "completion_length": 578.2854351043701, | |
| "epoch": 0.9872862053257133, | |
| "grad_norm": 0.3291475555349833, | |
| "kl": 0.1758056640625, | |
| "learning_rate": 9.658451184600959e-09, | |
| "loss": 0.007, | |
| "reward": 0.16041667032986878, | |
| "reward_std": 0.13258252181112767, | |
| "rewards/accuracy_reward": 0.16041667032986878, | |
| "rewards/format_reward": 0.0, | |
| "step": 1490 | |
| }, | |
| { | |
| "completion_length": 577.6250164031983, | |
| "epoch": 0.9905992462831822, | |
| "grad_norm": 0.28807275157813206, | |
| "kl": 0.1435791015625, | |
| "learning_rate": 5.2443095448506674e-09, | |
| "loss": 0.0057, | |
| "reward": 0.16875000428408385, | |
| "reward_std": 0.12079740893095732, | |
| "rewards/accuracy_reward": 0.16875000428408385, | |
| "rewards/format_reward": 0.0, | |
| "step": 1495 | |
| }, | |
| { | |
| "completion_length": 565.6062662124634, | |
| "epoch": 0.9939122872406511, | |
| "grad_norm": 0.34303096364094354, | |
| "kl": 0.1605682373046875, | |
| "learning_rate": 2.167402349972925e-09, | |
| "loss": 0.0064, | |
| "reward": 0.19791667070239782, | |
| "reward_std": 0.1679378604516387, | |
| "rewards/accuracy_reward": 0.19791667070239782, | |
| "rewards/format_reward": 0.0, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.9939122872406511, | |
| "eval_completion_length": 554.8088396857767, | |
| "eval_kl": 0.18235868566176472, | |
| "eval_loss": 0.007235472556203604, | |
| "eval_reward": 0.21078431869254394, | |
| "eval_reward_std": 0.13171596868949778, | |
| "eval_rewards/accuracy_reward": 0.21078431869254394, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_runtime": 83.8442, | |
| "eval_samples_per_second": 1.181, | |
| "eval_steps_per_second": 0.107, | |
| "step": 1500 | |
| }, | |
| { | |
| "completion_length": 581.8271011352539, | |
| "epoch": 0.9972253281981198, | |
| "grad_norm": 0.21304650345487455, | |
| "kl": 0.1817626953125, | |
| "learning_rate": 4.2814127048873553e-10, | |
| "loss": 0.0073, | |
| "reward": 0.2020833371207118, | |
| "reward_std": 0.15026019141077995, | |
| "rewards/accuracy_reward": 0.2020833371207118, | |
| "rewards/format_reward": 0.0, | |
| "step": 1505 | |
| }, | |
| { | |
| "completion_length": 576.2317910194397, | |
| "epoch": 0.9998757609640949, | |
| "kl": 0.15965652465820312, | |
| "reward": 0.21093750488944352, | |
| "reward_std": 0.1215339784976095, | |
| "rewards/accuracy_reward": 0.21093750488944352, | |
| "rewards/format_reward": 0.0, | |
| "step": 1509, | |
| "total_flos": 0.0, | |
| "train_loss": 0.020143496153197993, | |
| "train_runtime": 113002.1292, | |
| "train_samples_per_second": 0.641, | |
| "train_steps_per_second": 0.013 | |
| } | |
| ], | |
| "logging_steps": 5, | |
| "max_steps": 1509, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 1, | |
| "save_steps": 500, | |
| "stateful_callbacks": { | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": false, | |
| "should_training_stop": false | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 0.0, | |
| "train_batch_size": 2, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |