| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.9984, | |
| "eval_steps": 100, | |
| "global_step": 468, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "completion_length": 633.2446681976319, | |
| "epoch": 0.010666666666666666, | |
| "grad_norm": 2.2443933486938477, | |
| "kl": 0.00011417865753173828, | |
| "learning_rate": 3.1914893617021275e-07, | |
| "loss": 0.0, | |
| "reward": 1.138736367225647, | |
| "reward_std": 0.8278621450066567, | |
| "rewards/accuracy_reward": 0.5946428831666708, | |
| "rewards/cosine_scaled_reward": 0.2899268216686323, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.25416668243706225, | |
| "step": 5 | |
| }, | |
| { | |
| "completion_length": 600.8857383728027, | |
| "epoch": 0.021333333333333333, | |
| "grad_norm": 5.001251220703125, | |
| "kl": 0.00020779371261596679, | |
| "learning_rate": 6.382978723404255e-07, | |
| "loss": 0.0, | |
| "reward": 1.2528822764754295, | |
| "reward_std": 0.8592379853129387, | |
| "rewards/accuracy_reward": 0.6553571775555611, | |
| "rewards/cosine_scaled_reward": 0.34097747248015364, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.25654763616621495, | |
| "step": 10 | |
| }, | |
| { | |
| "completion_length": 601.8518112182617, | |
| "epoch": 0.032, | |
| "grad_norm": 3.453845500946045, | |
| "kl": 0.00034580230712890627, | |
| "learning_rate": 9.574468085106384e-07, | |
| "loss": 0.0, | |
| "reward": 1.2825960636138916, | |
| "reward_std": 0.7762525148689747, | |
| "rewards/accuracy_reward": 0.6642857484519482, | |
| "rewards/cosine_scaled_reward": 0.3486674582702108, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.26964287189766767, | |
| "step": 15 | |
| }, | |
| { | |
| "completion_length": 620.7839553833007, | |
| "epoch": 0.042666666666666665, | |
| "grad_norm": 63.01131057739258, | |
| "kl": 0.001246500015258789, | |
| "learning_rate": 1.276595744680851e-06, | |
| "loss": 0.0001, | |
| "reward": 1.2914750523865224, | |
| "reward_std": 0.7945833645761013, | |
| "rewards/accuracy_reward": 0.6571428865194321, | |
| "rewards/cosine_scaled_reward": 0.3593321413063677, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.2750000203028321, | |
| "step": 20 | |
| }, | |
| { | |
| "completion_length": 639.3946762084961, | |
| "epoch": 0.05333333333333334, | |
| "grad_norm": 1.1951252222061157, | |
| "kl": 0.001938199996948242, | |
| "learning_rate": 1.5957446808510639e-06, | |
| "loss": 0.0001, | |
| "reward": 1.2197763450443744, | |
| "reward_std": 0.7964548453688621, | |
| "rewards/accuracy_reward": 0.6285714630037547, | |
| "rewards/cosine_scaled_reward": 0.323942980915308, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.2672619212418795, | |
| "step": 25 | |
| }, | |
| { | |
| "completion_length": 645.9482414245606, | |
| "epoch": 0.064, | |
| "grad_norm": 0.5322187542915344, | |
| "kl": 0.0028698921203613283, | |
| "learning_rate": 1.9148936170212767e-06, | |
| "loss": 0.0001, | |
| "reward": 1.34233574308455, | |
| "reward_std": 0.7051636058837175, | |
| "rewards/accuracy_reward": 0.6821428902447224, | |
| "rewards/cosine_scaled_reward": 0.38400235488079487, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.2761904950253665, | |
| "step": 30 | |
| }, | |
| { | |
| "completion_length": 630.1071678161621, | |
| "epoch": 0.07466666666666667, | |
| "grad_norm": 0.686019241809845, | |
| "kl": 0.00424489974975586, | |
| "learning_rate": 2.2340425531914894e-06, | |
| "loss": 0.0002, | |
| "reward": 1.2706220560474322, | |
| "reward_std": 0.7081292014569044, | |
| "rewards/accuracy_reward": 0.6839286010712385, | |
| "rewards/cosine_scaled_reward": 0.34145535016432405, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.2452381114475429, | |
| "step": 35 | |
| }, | |
| { | |
| "completion_length": 663.8464553833007, | |
| "epoch": 0.08533333333333333, | |
| "grad_norm": 10619385856.0, | |
| "kl": 11324620.806011772, | |
| "learning_rate": 2.553191489361702e-06, | |
| "loss": 453134.65, | |
| "reward": 1.4818414891138674, | |
| "reward_std": 0.724718413501978, | |
| "rewards/accuracy_reward": 0.7196428954601288, | |
| "rewards/cosine_scaled_reward": 0.43124618427827954, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.3309524044394493, | |
| "step": 40 | |
| }, | |
| { | |
| "completion_length": 636.5178840637207, | |
| "epoch": 0.096, | |
| "grad_norm": 0.4083445370197296, | |
| "kl": 0.1388763427734375, | |
| "learning_rate": 2.872340425531915e-06, | |
| "loss": 0.0055, | |
| "reward": 1.5206772923469543, | |
| "reward_std": 0.6890950493514538, | |
| "rewards/accuracy_reward": 0.7428571715950966, | |
| "rewards/cosine_scaled_reward": 0.4444867596961558, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.3333333550952375, | |
| "step": 45 | |
| }, | |
| { | |
| "completion_length": 624.0178833007812, | |
| "epoch": 0.10666666666666667, | |
| "grad_norm": 0.6491600275039673, | |
| "kl": 0.014713478088378907, | |
| "learning_rate": 2.9996241442585123e-06, | |
| "loss": 0.0006, | |
| "reward": 1.5073627218604089, | |
| "reward_std": 0.7132997542619706, | |
| "rewards/accuracy_reward": 0.712500025331974, | |
| "rewards/cosine_scaled_reward": 0.41093407664448023, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.38392860516905786, | |
| "step": 50 | |
| }, | |
| { | |
| "completion_length": 631.5339569091797, | |
| "epoch": 0.11733333333333333, | |
| "grad_norm": 0.7147920727729797, | |
| "kl": 0.007195663452148437, | |
| "learning_rate": 2.9973279301399446e-06, | |
| "loss": 0.0003, | |
| "reward": 1.5377919152379036, | |
| "reward_std": 0.76092077344656, | |
| "rewards/accuracy_reward": 0.7232143200933934, | |
| "rewards/cosine_scaled_reward": 0.4282680474221706, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.386309552192688, | |
| "step": 55 | |
| }, | |
| { | |
| "completion_length": 627.9214561462402, | |
| "epoch": 0.128, | |
| "grad_norm": 0.8942143321037292, | |
| "kl": 0.008642578125, | |
| "learning_rate": 2.992947502998804e-06, | |
| "loss": 0.0003, | |
| "reward": 1.6543699458241463, | |
| "reward_std": 0.7264986954629421, | |
| "rewards/accuracy_reward": 0.7214285999536514, | |
| "rewards/cosine_scaled_reward": 0.40972703909501434, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.5232143249362707, | |
| "step": 60 | |
| }, | |
| { | |
| "completion_length": 633.0232421875, | |
| "epoch": 0.13866666666666666, | |
| "grad_norm": 6.921348571777344, | |
| "kl": 0.01439208984375, | |
| "learning_rate": 2.9864889601923268e-06, | |
| "loss": 0.0006, | |
| "reward": 1.7206872910261155, | |
| "reward_std": 0.7344334974884987, | |
| "rewards/accuracy_reward": 0.725000036507845, | |
| "rewards/cosine_scaled_reward": 0.43497296012938025, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.5607143200933933, | |
| "step": 65 | |
| }, | |
| { | |
| "completion_length": 656.7178894042969, | |
| "epoch": 0.14933333333333335, | |
| "grad_norm": 0.6442045569419861, | |
| "kl": 0.01673736572265625, | |
| "learning_rate": 2.977961291721137e-06, | |
| "loss": 0.0007, | |
| "reward": 1.8801582887768746, | |
| "reward_std": 0.7263622097671032, | |
| "rewards/accuracy_reward": 0.7571428894996644, | |
| "rewards/cosine_scaled_reward": 0.47301534870639445, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.6500000521540642, | |
| "step": 70 | |
| }, | |
| { | |
| "completion_length": 619.4536033630371, | |
| "epoch": 0.16, | |
| "grad_norm": 1.7239394187927246, | |
| "kl": 0.026496124267578126, | |
| "learning_rate": 2.9673763677155655e-06, | |
| "loss": 0.0011, | |
| "reward": 1.8051109313964844, | |
| "reward_std": 0.7346500240266323, | |
| "rewards/accuracy_reward": 0.7160714596509934, | |
| "rewards/cosine_scaled_reward": 0.39439656864851713, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.6946429140865803, | |
| "step": 75 | |
| }, | |
| { | |
| "completion_length": 623.1785926818848, | |
| "epoch": 0.17066666666666666, | |
| "grad_norm": 0.6716666221618652, | |
| "kl": 0.018997955322265624, | |
| "learning_rate": 2.9547489219129666e-06, | |
| "loss": 0.0008, | |
| "reward": 1.9212585434317588, | |
| "reward_std": 0.634969700500369, | |
| "rewards/accuracy_reward": 0.7785714574158191, | |
| "rewards/cosine_scaled_reward": 0.4653060721466318, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.6773809991776943, | |
| "step": 80 | |
| }, | |
| { | |
| "completion_length": 690.1518196105957, | |
| "epoch": 0.18133333333333335, | |
| "grad_norm": 1.1456305980682373, | |
| "kl": 0.02204437255859375, | |
| "learning_rate": 2.9400965311490175e-06, | |
| "loss": 0.0009, | |
| "reward": 1.9084690719842912, | |
| "reward_std": 0.7263222638517618, | |
| "rewards/accuracy_reward": 0.7303571783006191, | |
| "rewards/cosine_scaled_reward": 0.4507309086387977, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.7273810178041458, | |
| "step": 85 | |
| }, | |
| { | |
| "completion_length": 650.4768188476562, | |
| "epoch": 0.192, | |
| "grad_norm": 29.814361572265625, | |
| "kl": 0.078216552734375, | |
| "learning_rate": 2.9234395908915565e-06, | |
| "loss": 0.0031, | |
| "reward": 1.8972563683986663, | |
| "reward_std": 0.7165740359574556, | |
| "rewards/accuracy_reward": 0.6875000324100256, | |
| "rewards/cosine_scaled_reward": 0.4055896209087223, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8041667267680168, | |
| "step": 90 | |
| }, | |
| { | |
| "completion_length": 668.3339584350585, | |
| "epoch": 0.20266666666666666, | |
| "grad_norm": 0.48750847578048706, | |
| "kl": 0.02767181396484375, | |
| "learning_rate": 2.904801286851009e-06, | |
| "loss": 0.0011, | |
| "reward": 1.9524270623922348, | |
| "reward_std": 0.6363851364701987, | |
| "rewards/accuracy_reward": 0.7035714564844966, | |
| "rewards/cosine_scaled_reward": 0.42206980669870975, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.826785783469677, | |
| "step": 95 | |
| }, | |
| { | |
| "completion_length": 645.9428840637207, | |
| "epoch": 0.21333333333333335, | |
| "grad_norm": 0.8315287232398987, | |
| "kl": 0.02986602783203125, | |
| "learning_rate": 2.884207562706925e-06, | |
| "loss": 0.0012, | |
| "reward": 2.0384097367525102, | |
| "reward_std": 0.6786769151687622, | |
| "rewards/accuracy_reward": 0.7517857387661934, | |
| "rewards/cosine_scaled_reward": 0.4657905898289755, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.820833396166563, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.21333333333333335, | |
| "eval_completion_length": 688.0076597412109, | |
| "eval_kl": 0.0332870361328125, | |
| "eval_loss": 0.0013802805915474892, | |
| "eval_reward": 1.86520801551342, | |
| "eval_reward_std": 0.7114028903335333, | |
| "eval_rewards/accuracy_reward": 0.650542886838317, | |
| "eval_rewards/cosine_scaled_reward": 0.3737031816519331, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_rewards/reasoning_steps_reward": 0.8409619681358338, | |
| "eval_runtime": 32350.4437, | |
| "eval_samples_per_second": 0.155, | |
| "eval_steps_per_second": 0.011, | |
| "step": 100 | |
| }, | |
| { | |
| "completion_length": 717.150033569336, | |
| "epoch": 0.224, | |
| "grad_norm": 1.5486549139022827, | |
| "kl": 0.03196563720703125, | |
| "learning_rate": 2.8616870839955444e-06, | |
| "loss": 0.0013, | |
| "reward": 2.0346583992242815, | |
| "reward_std": 0.7014419212937355, | |
| "rewards/accuracy_reward": 0.7232143215835094, | |
| "rewards/cosine_scaled_reward": 0.457277343980968, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8541667237877846, | |
| "step": 105 | |
| }, | |
| { | |
| "completion_length": 708.8571708679199, | |
| "epoch": 0.23466666666666666, | |
| "grad_norm": 0.5981384515762329, | |
| "kl": 0.02979583740234375, | |
| "learning_rate": 2.837271198208662e-06, | |
| "loss": 0.0012, | |
| "reward": 2.0179374665021896, | |
| "reward_std": 0.6652137346565723, | |
| "rewards/accuracy_reward": 0.7250000320374965, | |
| "rewards/cosine_scaled_reward": 0.47091358043253423, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8220238700509072, | |
| "step": 110 | |
| }, | |
| { | |
| "completion_length": 632.7732406616211, | |
| "epoch": 0.24533333333333332, | |
| "grad_norm": 0.7111315131187439, | |
| "kl": 0.02539825439453125, | |
| "learning_rate": 2.8109938911593322e-06, | |
| "loss": 0.001, | |
| "reward": 2.0148118153214454, | |
| "reward_std": 0.6429756574332715, | |
| "rewards/accuracy_reward": 0.728571455553174, | |
| "rewards/cosine_scaled_reward": 0.44754982106387614, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8386905357241631, | |
| "step": 115 | |
| }, | |
| { | |
| "completion_length": 655.8321723937988, | |
| "epoch": 0.256, | |
| "grad_norm": 0.5316483974456787, | |
| "kl": 0.02179107666015625, | |
| "learning_rate": 2.7828917396751474e-06, | |
| "loss": 0.0009, | |
| "reward": 1.9900789648294448, | |
| "reward_std": 0.6477071691304446, | |
| "rewards/accuracy_reward": 0.7160714656114578, | |
| "rewards/cosine_scaled_reward": 0.43412648113444446, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8398810118436814, | |
| "step": 120 | |
| }, | |
| { | |
| "completion_length": 644.7321693420411, | |
| "epoch": 0.26666666666666666, | |
| "grad_norm": 0.4458823800086975, | |
| "kl": 0.025299072265625, | |
| "learning_rate": 2.753003860684943e-06, | |
| "loss": 0.001, | |
| "reward": 2.1427780210971834, | |
| "reward_std": 0.6711063630878925, | |
| "rewards/accuracy_reward": 0.7750000268220901, | |
| "rewards/cosine_scaled_reward": 0.5183731818571686, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8494048312306404, | |
| "step": 125 | |
| }, | |
| { | |
| "completion_length": 684.2911033630371, | |
| "epoch": 0.2773333333333333, | |
| "grad_norm": 0.7146270871162415, | |
| "kl": 0.034222412109375, | |
| "learning_rate": 2.721371856769793e-06, | |
| "loss": 0.0014, | |
| "reward": 1.9814838409423827, | |
| "reward_std": 0.7353869907557964, | |
| "rewards/accuracy_reward": 0.6625000331550837, | |
| "rewards/cosine_scaled_reward": 0.3981504186260281, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9208333924412727, | |
| "step": 130 | |
| }, | |
| { | |
| "completion_length": 650.483960723877, | |
| "epoch": 0.288, | |
| "grad_norm": 0.8331003189086914, | |
| "kl": 0.046978759765625, | |
| "learning_rate": 2.688039758254093e-06, | |
| "loss": 0.0019, | |
| "reward": 2.223627084493637, | |
| "reward_std": 0.6465678755193949, | |
| "rewards/accuracy_reward": 0.7732143219560385, | |
| "rewards/cosine_scaled_reward": 0.506960358901415, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.94345243871212, | |
| "step": 135 | |
| }, | |
| { | |
| "completion_length": 702.9536026000976, | |
| "epoch": 0.2986666666666667, | |
| "grad_norm": 1.9107334613800049, | |
| "kl": 0.0536590576171875, | |
| "learning_rate": 2.65305396191733e-06, | |
| "loss": 0.0021, | |
| "reward": 2.1239778250455856, | |
| "reward_std": 0.6765143848955631, | |
| "rewards/accuracy_reward": 0.7071428891271353, | |
| "rewards/cosine_scaled_reward": 0.4555253505706787, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9613095715641975, | |
| "step": 140 | |
| }, | |
| { | |
| "completion_length": 733.6089630126953, | |
| "epoch": 0.30933333333333335, | |
| "grad_norm": 0.5300867557525635, | |
| "kl": 0.05316162109375, | |
| "learning_rate": 2.61646316641186e-06, | |
| "loss": 0.0021, | |
| "reward": 2.1554796636104583, | |
| "reward_std": 0.6578622825443745, | |
| "rewards/accuracy_reward": 0.7303571704775095, | |
| "rewards/cosine_scaled_reward": 0.47036054339259864, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9547619551420212, | |
| "step": 145 | |
| }, | |
| { | |
| "completion_length": 713.221459197998, | |
| "epoch": 0.32, | |
| "grad_norm": 0.6026062369346619, | |
| "kl": 0.0533843994140625, | |
| "learning_rate": 2.5783183044765715e-06, | |
| "loss": 0.0021, | |
| "reward": 2.1126459658145906, | |
| "reward_std": 0.5920085646212101, | |
| "rewards/accuracy_reward": 0.7089285995811224, | |
| "rewards/cosine_scaled_reward": 0.4566935421898961, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9470238655805587, | |
| "step": 150 | |
| }, | |
| { | |
| "completion_length": 678.6428886413574, | |
| "epoch": 0.33066666666666666, | |
| "grad_norm": 0.6598377227783203, | |
| "kl": 0.049908447265625, | |
| "learning_rate": 2.5386724720408135e-06, | |
| "loss": 0.002, | |
| "reward": 2.243595580756664, | |
| "reward_std": 0.6088640403002501, | |
| "rewards/accuracy_reward": 0.7767857441678643, | |
| "rewards/cosine_scaled_reward": 0.5435954930260778, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9232143476605416, | |
| "step": 155 | |
| }, | |
| { | |
| "completion_length": 683.9268142700196, | |
| "epoch": 0.3413333333333333, | |
| "grad_norm": 0.6654959321022034, | |
| "kl": 0.0447540283203125, | |
| "learning_rate": 2.49758085431725e-06, | |
| "loss": 0.0018, | |
| "reward": 2.0952899247407912, | |
| "reward_std": 0.6968366518616677, | |
| "rewards/accuracy_reward": 0.7232143208384514, | |
| "rewards/cosine_scaled_reward": 0.4637422326952219, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9083333939313889, | |
| "step": 160 | |
| }, | |
| { | |
| "completion_length": 691.3464614868165, | |
| "epoch": 0.352, | |
| "grad_norm": 0.689552903175354, | |
| "kl": 0.0448211669921875, | |
| "learning_rate": 2.455100648986533e-06, | |
| "loss": 0.0018, | |
| "reward": 2.0519487097859384, | |
| "reward_std": 0.7221721112728119, | |
| "rewards/accuracy_reward": 0.6964286031201482, | |
| "rewards/cosine_scaled_reward": 0.4602819522842765, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8952381581068038, | |
| "step": 165 | |
| }, | |
| { | |
| "completion_length": 696.5268180847168, | |
| "epoch": 0.3626666666666667, | |
| "grad_norm": 1.0024878978729248, | |
| "kl": 0.065167236328125, | |
| "learning_rate": 2.4112909865807053e-06, | |
| "loss": 0.0026, | |
| "reward": 1.7887505039572715, | |
| "reward_std": 0.7482936225831509, | |
| "rewards/accuracy_reward": 0.571428600884974, | |
| "rewards/cosine_scaled_reward": 0.3333932981360704, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8839286401867866, | |
| "step": 170 | |
| }, | |
| { | |
| "completion_length": 703.2714614868164, | |
| "epoch": 0.37333333333333335, | |
| "grad_norm": 0.5711168050765991, | |
| "kl": 0.093731689453125, | |
| "learning_rate": 2.366212848176164e-06, | |
| "loss": 0.0037, | |
| "reward": 1.9069189459085465, | |
| "reward_std": 0.8069212771952152, | |
| "rewards/accuracy_reward": 0.6500000327825546, | |
| "rewards/cosine_scaled_reward": 0.42358550764620306, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8333333879709244, | |
| "step": 175 | |
| }, | |
| { | |
| "completion_length": 714.2536003112793, | |
| "epoch": 0.384, | |
| "grad_norm": 3.1069464683532715, | |
| "kl": 0.1747802734375, | |
| "learning_rate": 2.319928980510752e-06, | |
| "loss": 0.007, | |
| "reward": 1.6917703241109847, | |
| "reward_std": 0.8836216881871224, | |
| "rewards/accuracy_reward": 0.6089285977184773, | |
| "rewards/cosine_scaled_reward": 0.35307975246978457, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.7297619581222534, | |
| "step": 180 | |
| }, | |
| { | |
| "completion_length": 727.7018188476562, | |
| "epoch": 0.39466666666666667, | |
| "grad_norm": 1.1932159662246704, | |
| "kl": 0.193988037109375, | |
| "learning_rate": 2.272503808643123e-06, | |
| "loss": 0.0078, | |
| "reward": 1.7027929693460464, | |
| "reward_std": 0.7921728197485208, | |
| "rewards/accuracy_reward": 0.6267857421189547, | |
| "rewards/cosine_scaled_reward": 0.3605310095474124, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.7154762461781502, | |
| "step": 185 | |
| }, | |
| { | |
| "completion_length": 677.6518127441407, | |
| "epoch": 0.4053333333333333, | |
| "grad_norm": 0.6525413393974304, | |
| "kl": 0.1227813720703125, | |
| "learning_rate": 2.2240033462759628e-06, | |
| "loss": 0.0049, | |
| "reward": 2.055608908832073, | |
| "reward_std": 0.6409808352589608, | |
| "rewards/accuracy_reward": 0.7428571667522192, | |
| "rewards/cosine_scaled_reward": 0.4907278836122714, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8220238700509072, | |
| "step": 190 | |
| }, | |
| { | |
| "completion_length": 729.3125358581543, | |
| "epoch": 0.416, | |
| "grad_norm": 0.470821738243103, | |
| "kl": 0.1053009033203125, | |
| "learning_rate": 2.1744951038678905e-06, | |
| "loss": 0.0042, | |
| "reward": 2.1352262631058694, | |
| "reward_std": 0.6541992913931608, | |
| "rewards/accuracy_reward": 0.7446428880095481, | |
| "rewards/cosine_scaled_reward": 0.5340357202105224, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8565476804971695, | |
| "step": 195 | |
| }, | |
| { | |
| "completion_length": 736.6607482910156, | |
| "epoch": 0.4266666666666667, | |
| "grad_norm": 0.3663829267024994, | |
| "kl": 0.145220947265625, | |
| "learning_rate": 2.124047994661941e-06, | |
| "loss": 0.0058, | |
| "reward": 2.0683016672730448, | |
| "reward_std": 0.6785697277635336, | |
| "rewards/accuracy_reward": 0.7107143150642514, | |
| "rewards/cosine_scaled_reward": 0.4861587251536548, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8714286342263222, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.4266666666666667, | |
| "eval_completion_length": 743.3604330322265, | |
| "eval_kl": 0.1699279296875, | |
| "eval_loss": 0.006734147202223539, | |
| "eval_reward": 1.8947704853653908, | |
| "eval_reward_std": 0.7092250557422638, | |
| "eval_rewards/accuracy_reward": 0.6307143133163452, | |
| "eval_rewards/cosine_scaled_reward": 0.39257041423644407, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_rewards/reasoning_steps_reward": 0.871485775399208, | |
| "eval_runtime": 32670.592, | |
| "eval_samples_per_second": 0.153, | |
| "eval_steps_per_second": 0.011, | |
| "step": 200 | |
| }, | |
| { | |
| "completion_length": 752.7053955078125, | |
| "epoch": 0.43733333333333335, | |
| "grad_norm": 0.5299625396728516, | |
| "kl": 0.1930633544921875, | |
| "learning_rate": 2.072732238761434e-06, | |
| "loss": 0.0077, | |
| "reward": 1.8860187515616418, | |
| "reward_std": 0.7606242794543505, | |
| "rewards/accuracy_reward": 0.6446428863331676, | |
| "rewards/cosine_scaled_reward": 0.40447108587541153, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8369048193097115, | |
| "step": 205 | |
| }, | |
| { | |
| "completion_length": 733.603606414795, | |
| "epoch": 0.448, | |
| "grad_norm": 1.6152819395065308, | |
| "kl": 0.219268798828125, | |
| "learning_rate": 2.0206192653867536e-06, | |
| "loss": 0.0088, | |
| "reward": 1.997245892137289, | |
| "reward_std": 0.7402419943362475, | |
| "rewards/accuracy_reward": 0.7017857382073999, | |
| "rewards/cosine_scaled_reward": 0.47284105569124224, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8226191058754921, | |
| "step": 210 | |
| }, | |
| { | |
| "completion_length": 844.0661102294922, | |
| "epoch": 0.45866666666666667, | |
| "grad_norm": 7.516280651092529, | |
| "kl": 0.27982177734375, | |
| "learning_rate": 1.967781613449095e-06, | |
| "loss": 0.0112, | |
| "reward": 1.5464881896972655, | |
| "reward_std": 0.8091491930186748, | |
| "rewards/accuracy_reward": 0.49107144959270954, | |
| "rewards/cosine_scaled_reward": 0.21672622584737838, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8386905357241631, | |
| "step": 215 | |
| }, | |
| { | |
| "completion_length": 814.1696807861329, | |
| "epoch": 0.4693333333333333, | |
| "grad_norm": 0.4684678018093109, | |
| "kl": 0.194140625, | |
| "learning_rate": 1.9142928305795637e-06, | |
| "loss": 0.0078, | |
| "reward": 1.8477135568857193, | |
| "reward_std": 0.7414120733737946, | |
| "rewards/accuracy_reward": 0.6178571652621031, | |
| "rewards/cosine_scaled_reward": 0.3584277655696496, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8714286401867867, | |
| "step": 220 | |
| }, | |
| { | |
| "completion_length": 754.1857452392578, | |
| "epoch": 0.48, | |
| "grad_norm": 0.4328997731208801, | |
| "kl": 0.12838134765625, | |
| "learning_rate": 1.8602273707541886e-06, | |
| "loss": 0.0051, | |
| "reward": 2.1135876968503, | |
| "reward_std": 0.6965163860470056, | |
| "rewards/accuracy_reward": 0.742857176810503, | |
| "rewards/cosine_scaled_reward": 0.5159685641527176, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8547619715332985, | |
| "step": 225 | |
| }, | |
| { | |
| "completion_length": 742.7750381469726, | |
| "epoch": 0.49066666666666664, | |
| "grad_norm": 0.4649052619934082, | |
| "kl": 0.1558837890625, | |
| "learning_rate": 1.8056604906573418e-06, | |
| "loss": 0.0062, | |
| "reward": 2.0384344711899756, | |
| "reward_std": 0.6620127268135547, | |
| "rewards/accuracy_reward": 0.7035714626312256, | |
| "rewards/cosine_scaled_reward": 0.483077246020548, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8517857760190963, | |
| "step": 230 | |
| }, | |
| { | |
| "completion_length": 739.6268203735351, | |
| "epoch": 0.5013333333333333, | |
| "grad_norm": 1.5264660120010376, | |
| "kl": 0.145806884765625, | |
| "learning_rate": 1.7506681449278226e-06, | |
| "loss": 0.0058, | |
| "reward": 1.999456986784935, | |
| "reward_std": 0.7032103724777699, | |
| "rewards/accuracy_reward": 0.6785714574158191, | |
| "rewards/cosine_scaled_reward": 0.45302835907787087, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8678572103381157, | |
| "step": 235 | |
| }, | |
| { | |
| "completion_length": 725.905387878418, | |
| "epoch": 0.512, | |
| "grad_norm": 13.703657150268555, | |
| "kl": 0.354132080078125, | |
| "learning_rate": 1.6953268804334257e-06, | |
| "loss": 0.0142, | |
| "reward": 2.012031316757202, | |
| "reward_std": 0.6349152896553278, | |
| "rewards/accuracy_reward": 0.6660714553669095, | |
| "rewards/cosine_scaled_reward": 0.46024551438167693, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8857143551111222, | |
| "step": 240 | |
| }, | |
| { | |
| "completion_length": 711.9410980224609, | |
| "epoch": 0.5226666666666666, | |
| "grad_norm": 42.922752380371094, | |
| "kl": 0.81356201171875, | |
| "learning_rate": 1.6397137297211436e-06, | |
| "loss": 0.0325, | |
| "reward": 2.129089578986168, | |
| "reward_std": 0.699107101932168, | |
| "rewards/accuracy_reward": 0.7160714577883482, | |
| "rewards/cosine_scaled_reward": 0.5064704709046055, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9065476730465889, | |
| "step": 245 | |
| }, | |
| { | |
| "completion_length": 738.9821746826171, | |
| "epoch": 0.5333333333333333, | |
| "grad_norm": 212.6622314453125, | |
| "kl": 1.157550048828125, | |
| "learning_rate": 1.5839061037913395e-06, | |
| "loss": 0.0463, | |
| "reward": 2.1009622782468798, | |
| "reward_std": 0.7158728931099176, | |
| "rewards/accuracy_reward": 0.7000000283122063, | |
| "rewards/cosine_scaled_reward": 0.5027479250915349, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8982143506407738, | |
| "step": 250 | |
| }, | |
| { | |
| "completion_length": 760.2428916931152, | |
| "epoch": 0.544, | |
| "grad_norm": 10.118670463562012, | |
| "kl": 0.637158203125, | |
| "learning_rate": 1.527981684345115e-06, | |
| "loss": 0.0255, | |
| "reward": 1.9621681660413741, | |
| "reward_std": 0.67494813259691, | |
| "rewards/accuracy_reward": 0.639285740070045, | |
| "rewards/cosine_scaled_reward": 0.4276442806003615, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8952381491661072, | |
| "step": 255 | |
| }, | |
| { | |
| "completion_length": 754.6803894042969, | |
| "epoch": 0.5546666666666666, | |
| "grad_norm": 7.878048419952393, | |
| "kl": 0.972845458984375, | |
| "learning_rate": 1.4720183156548855e-06, | |
| "loss": 0.0389, | |
| "reward": 1.9780788227915764, | |
| "reward_std": 0.6262619759887457, | |
| "rewards/accuracy_reward": 0.6339285982772708, | |
| "rewards/cosine_scaled_reward": 0.4304597085807472, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9136905416846275, | |
| "step": 260 | |
| }, | |
| { | |
| "completion_length": 751.5857498168946, | |
| "epoch": 0.5653333333333334, | |
| "grad_norm": 12.42583179473877, | |
| "kl": 3.09744873046875, | |
| "learning_rate": 1.4160938962086612e-06, | |
| "loss": 0.1241, | |
| "reward": 2.0433208346366882, | |
| "reward_std": 0.661328698694706, | |
| "rewards/accuracy_reward": 0.676785740442574, | |
| "rewards/cosine_scaled_reward": 0.44689220561413096, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9196429163217544, | |
| "step": 265 | |
| }, | |
| { | |
| "completion_length": 729.028604888916, | |
| "epoch": 0.576, | |
| "grad_norm": 7.453009605407715, | |
| "kl": 2.2955322265625, | |
| "learning_rate": 1.3602862702788567e-06, | |
| "loss": 0.0917, | |
| "reward": 2.094664843380451, | |
| "reward_std": 0.6356621380895376, | |
| "rewards/accuracy_reward": 0.7000000346451998, | |
| "rewards/cosine_scaled_reward": 0.46371242445893585, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9309524431824684, | |
| "step": 270 | |
| }, | |
| { | |
| "completion_length": 730.825032043457, | |
| "epoch": 0.5866666666666667, | |
| "grad_norm": 7.0367817878723145, | |
| "kl": 0.6509521484375, | |
| "learning_rate": 1.3046731195665748e-06, | |
| "loss": 0.0261, | |
| "reward": 2.083331751823425, | |
| "reward_std": 0.6676435235887765, | |
| "rewards/accuracy_reward": 0.6821428818628192, | |
| "rewards/cosine_scaled_reward": 0.45714118536561726, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.944047674536705, | |
| "step": 275 | |
| }, | |
| { | |
| "completion_length": 742.180387878418, | |
| "epoch": 0.5973333333333334, | |
| "grad_norm": 1.3236949443817139, | |
| "kl": 4.09298095703125, | |
| "learning_rate": 1.2493318550721775e-06, | |
| "loss": 0.1637, | |
| "reward": 2.075996032357216, | |
| "reward_std": 0.6379393456503749, | |
| "rewards/accuracy_reward": 0.6857143174856901, | |
| "rewards/cosine_scaled_reward": 0.4563530746847391, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9339286297559738, | |
| "step": 280 | |
| }, | |
| { | |
| "completion_length": 708.1018157958985, | |
| "epoch": 0.608, | |
| "grad_norm": 5.264936447143555, | |
| "kl": 0.21192626953125, | |
| "learning_rate": 1.1943395093426585e-06, | |
| "loss": 0.0085, | |
| "reward": 2.1390477627515794, | |
| "reward_std": 0.600306774303317, | |
| "rewards/accuracy_reward": 0.7196428820490837, | |
| "rewards/cosine_scaled_reward": 0.49619057439267633, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9232143506407737, | |
| "step": 285 | |
| }, | |
| { | |
| "completion_length": 715.4125289916992, | |
| "epoch": 0.6186666666666667, | |
| "grad_norm": 2.6887574195861816, | |
| "kl": 2.8669677734375, | |
| "learning_rate": 1.1397726292458115e-06, | |
| "loss": 0.1151, | |
| "reward": 2.1179503470659258, | |
| "reward_std": 0.5490788316354156, | |
| "rewards/accuracy_reward": 0.7053571708500386, | |
| "rewards/cosine_scaled_reward": 0.4905693273060024, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9220238789916039, | |
| "step": 290 | |
| }, | |
| { | |
| "completion_length": 742.6803916931152, | |
| "epoch": 0.6293333333333333, | |
| "grad_norm": 6.9418721199035645, | |
| "kl": 0.39151611328125, | |
| "learning_rate": 1.085707169420437e-06, | |
| "loss": 0.0157, | |
| "reward": 1.8962592497467994, | |
| "reward_std": 0.6060247957706452, | |
| "rewards/accuracy_reward": 0.5964285938069225, | |
| "rewards/cosine_scaled_reward": 0.3754258565604687, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.924404813349247, | |
| "step": 295 | |
| }, | |
| { | |
| "completion_length": 716.3464584350586, | |
| "epoch": 0.64, | |
| "grad_norm": 4.2906060218811035, | |
| "kl": 0.57667236328125, | |
| "learning_rate": 1.0322183865509054e-06, | |
| "loss": 0.0231, | |
| "reward": 2.1815308302640917, | |
| "reward_std": 0.6235232371836901, | |
| "rewards/accuracy_reward": 0.7428571732714773, | |
| "rewards/cosine_scaled_reward": 0.5255783690838143, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.913095298409462, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "eval_completion_length": 728.9849459716797, | |
| "eval_kl": 22.31169453125, | |
| "eval_loss": 0.8926114439964294, | |
| "eval_reward": 1.9843467233777046, | |
| "eval_reward_std": 0.6538388645738363, | |
| "eval_rewards/accuracy_reward": 0.6382285982251167, | |
| "eval_rewards/cosine_scaled_reward": 0.41530855364510644, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_rewards/reasoning_steps_reward": 0.9308095807313919, | |
| "eval_runtime": 32207.7986, | |
| "eval_samples_per_second": 0.155, | |
| "eval_steps_per_second": 0.011, | |
| "step": 300 | |
| }, | |
| { | |
| "completion_length": 723.2625328063965, | |
| "epoch": 0.6506666666666666, | |
| "grad_norm": 79.97950744628906, | |
| "kl": 487.1179443359375, | |
| "learning_rate": 9.793807346132464e-07, | |
| "loss": 19.4474, | |
| "reward": 2.162437987327576, | |
| "reward_std": 0.6324797321110964, | |
| "rewards/accuracy_reward": 0.7267857410013676, | |
| "rewards/cosine_scaled_reward": 0.5112474345514784, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9244048178195954, | |
| "step": 305 | |
| }, | |
| { | |
| "completion_length": 739.6375335693359, | |
| "epoch": 0.6613333333333333, | |
| "grad_norm": 9.395992279052734, | |
| "kl": 0.60579833984375, | |
| "learning_rate": 9.272677612385667e-07, | |
| "loss": 0.0242, | |
| "reward": 2.004467612504959, | |
| "reward_std": 0.6282935816794634, | |
| "rewards/accuracy_reward": 0.6607143184170127, | |
| "rewards/cosine_scaled_reward": 0.42589613443706187, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9178571999073029, | |
| "step": 310 | |
| }, | |
| { | |
| "completion_length": 735.6286071777344, | |
| "epoch": 0.672, | |
| "grad_norm": 12.830111503601074, | |
| "kl": 0.9565673828125, | |
| "learning_rate": 8.759520053380591e-07, | |
| "loss": 0.0383, | |
| "reward": 1.9197196617722512, | |
| "reward_std": 0.6299623921513557, | |
| "rewards/accuracy_reward": 0.6035714576020836, | |
| "rewards/cosine_scaled_reward": 0.39055290608666837, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9255953043699264, | |
| "step": 315 | |
| }, | |
| { | |
| "completion_length": 718.0571731567383, | |
| "epoch": 0.6826666666666666, | |
| "grad_norm": 176.6972198486328, | |
| "kl": 1.54287109375, | |
| "learning_rate": 8.255048961321088e-07, | |
| "loss": 0.0618, | |
| "reward": 2.1281729131937026, | |
| "reward_std": 0.6808584026992321, | |
| "rewards/accuracy_reward": 0.714285746216774, | |
| "rewards/cosine_scaled_reward": 0.4888871216215193, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9250000536441803, | |
| "step": 320 | |
| }, | |
| { | |
| "completion_length": 721.4732475280762, | |
| "epoch": 0.6933333333333334, | |
| "grad_norm": 6.025720119476318, | |
| "kl": 0.98104248046875, | |
| "learning_rate": 7.759966537240373e-07, | |
| "loss": 0.0392, | |
| "reward": 2.054315000772476, | |
| "reward_std": 0.6834255807101727, | |
| "rewards/accuracy_reward": 0.6714285992085933, | |
| "rewards/cosine_scaled_reward": 0.45312447142787277, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9297619640827179, | |
| "step": 325 | |
| }, | |
| { | |
| "completion_length": 729.3982498168946, | |
| "epoch": 0.704, | |
| "grad_norm": 6.682721138000488, | |
| "kl": 2.40982666015625, | |
| "learning_rate": 7.274961913568773e-07, | |
| "loss": 0.0964, | |
| "reward": 2.0376005843281746, | |
| "reward_std": 0.7055317234247923, | |
| "rewards/accuracy_reward": 0.6660714562982321, | |
| "rewards/cosine_scaled_reward": 0.4655766852200031, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9059524461627007, | |
| "step": 330 | |
| }, | |
| { | |
| "completion_length": 737.005387878418, | |
| "epoch": 0.7146666666666667, | |
| "grad_norm": 21.818754196166992, | |
| "kl": 0.653094482421875, | |
| "learning_rate": 6.800710194892484e-07, | |
| "loss": 0.0261, | |
| "reward": 2.056803268194199, | |
| "reward_std": 0.7108213260769844, | |
| "rewards/accuracy_reward": 0.6660714574158192, | |
| "rewards/cosine_scaled_reward": 0.45680318772792816, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9339286327362061, | |
| "step": 335 | |
| }, | |
| { | |
| "completion_length": 729.6393203735352, | |
| "epoch": 0.7253333333333334, | |
| "grad_norm": 4.025352954864502, | |
| "kl": 0.63848876953125, | |
| "learning_rate": 6.33787151823836e-07, | |
| "loss": 0.0256, | |
| "reward": 1.9720933943986894, | |
| "reward_std": 0.6898978160694241, | |
| "rewards/accuracy_reward": 0.6250000264495611, | |
| "rewards/cosine_scaled_reward": 0.42685523356776683, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9202381581068039, | |
| "step": 340 | |
| }, | |
| { | |
| "completion_length": 699.1571701049804, | |
| "epoch": 0.736, | |
| "grad_norm": 5.142830848693848, | |
| "kl": 0.65721435546875, | |
| "learning_rate": 5.887090134192947e-07, | |
| "loss": 0.0263, | |
| "reward": 2.100009024143219, | |
| "reward_std": 0.6496724892407656, | |
| "rewards/accuracy_reward": 0.6910714615136385, | |
| "rewards/cosine_scaled_reward": 0.4851280112750828, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9238095805048943, | |
| "step": 345 | |
| }, | |
| { | |
| "completion_length": 723.3910995483399, | |
| "epoch": 0.7466666666666667, | |
| "grad_norm": 4.602946758270264, | |
| "kl": 0.394140625, | |
| "learning_rate": 5.448993510134669e-07, | |
| "loss": 0.0158, | |
| "reward": 2.0926264360547067, | |
| "reward_std": 0.6916316740214825, | |
| "rewards/accuracy_reward": 0.6857143180444837, | |
| "rewards/cosine_scaled_reward": 0.4831025514518842, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9238095790147781, | |
| "step": 350 | |
| }, | |
| { | |
| "completion_length": 722.5375305175781, | |
| "epoch": 0.7573333333333333, | |
| "grad_norm": 6.0756731033325195, | |
| "kl": 1.08592529296875, | |
| "learning_rate": 5.024191456827498e-07, | |
| "loss": 0.0435, | |
| "reward": 2.0994770556688307, | |
| "reward_std": 0.666194306127727, | |
| "rewards/accuracy_reward": 0.6982143167406321, | |
| "rewards/cosine_scaled_reward": 0.4917388891801238, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9095238700509072, | |
| "step": 355 | |
| }, | |
| { | |
| "completion_length": 713.5250350952149, | |
| "epoch": 0.768, | |
| "grad_norm": 7.16264533996582, | |
| "kl": 26.27894287109375, | |
| "learning_rate": 4.6132752795918667e-07, | |
| "loss": 1.0497, | |
| "reward": 2.055359125137329, | |
| "reward_std": 0.7066416556015611, | |
| "rewards/accuracy_reward": 0.6678571753203869, | |
| "rewards/cosine_scaled_reward": 0.4732161985710263, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9142857760190963, | |
| "step": 360 | |
| }, | |
| { | |
| "completion_length": 751.5964584350586, | |
| "epoch": 0.7786666666666666, | |
| "grad_norm": 3.023808002471924, | |
| "kl": 1.154327392578125, | |
| "learning_rate": 4.2168169552342905e-07, | |
| "loss": 0.0462, | |
| "reward": 1.9766315311193465, | |
| "reward_std": 0.7433438140898943, | |
| "rewards/accuracy_reward": 0.6339286021888256, | |
| "rewards/cosine_scaled_reward": 0.42544099894585086, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9172619596123696, | |
| "step": 365 | |
| }, | |
| { | |
| "completion_length": 704.278596496582, | |
| "epoch": 0.7893333333333333, | |
| "grad_norm": 1.0741926431655884, | |
| "kl": 0.53934326171875, | |
| "learning_rate": 3.8353683358814046e-07, | |
| "loss": 0.0216, | |
| "reward": 2.0491741001605988, | |
| "reward_std": 0.587555892020464, | |
| "rewards/accuracy_reward": 0.6678571693599225, | |
| "rewards/cosine_scaled_reward": 0.46226926781237127, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9190476790070534, | |
| "step": 370 | |
| }, | |
| { | |
| "completion_length": 738.875033569336, | |
| "epoch": 0.8, | |
| "grad_norm": 41.52888870239258, | |
| "kl": 0.6643310546875, | |
| "learning_rate": 3.469460380826697e-07, | |
| "loss": 0.0265, | |
| "reward": 2.0449665546417237, | |
| "reward_std": 0.6989724855870009, | |
| "rewards/accuracy_reward": 0.6625000312924385, | |
| "rewards/cosine_scaled_reward": 0.4574664521496743, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9250000640749931, | |
| "step": 375 | |
| }, | |
| { | |
| "completion_length": 724.0678855895997, | |
| "epoch": 0.8106666666666666, | |
| "grad_norm": 4.322193145751953, | |
| "kl": 0.7086669921875, | |
| "learning_rate": 3.119602417459075e-07, | |
| "loss": 0.0284, | |
| "reward": 2.055614770948887, | |
| "reward_std": 0.6039443843066692, | |
| "rewards/accuracy_reward": 0.667857171408832, | |
| "rewards/cosine_scaled_reward": 0.46275755076203495, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9250000655651093, | |
| "step": 380 | |
| }, | |
| { | |
| "completion_length": 739.5125350952148, | |
| "epoch": 0.8213333333333334, | |
| "grad_norm": 4.056361198425293, | |
| "kl": 0.7447509765625, | |
| "learning_rate": 2.786281432302071e-07, | |
| "loss": 0.0298, | |
| "reward": 2.0523035705089567, | |
| "reward_std": 0.6267267379909753, | |
| "rewards/accuracy_reward": 0.6750000279396773, | |
| "rewards/cosine_scaled_reward": 0.4463511134439614, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9309524476528168, | |
| "step": 385 | |
| }, | |
| { | |
| "completion_length": 722.548243713379, | |
| "epoch": 0.832, | |
| "grad_norm": 1.378568410873413, | |
| "kl": 0.501007080078125, | |
| "learning_rate": 2.46996139315057e-07, | |
| "loss": 0.02, | |
| "reward": 2.0793206453323365, | |
| "reward_std": 0.6533296214416623, | |
| "rewards/accuracy_reward": 0.6875000290572644, | |
| "rewards/cosine_scaled_reward": 0.4781300783797633, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9136905401945115, | |
| "step": 390 | |
| }, | |
| { | |
| "completion_length": 732.4714630126953, | |
| "epoch": 0.8426666666666667, | |
| "grad_norm": 2.4824626445770264, | |
| "kl": 0.71015625, | |
| "learning_rate": 2.1710826032485286e-07, | |
| "loss": 0.0284, | |
| "reward": 2.1464335188269614, | |
| "reward_std": 0.6267410140484572, | |
| "rewards/accuracy_reward": 0.7071428874507546, | |
| "rewards/cosine_scaled_reward": 0.5136953465640545, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9255953013896943, | |
| "step": 395 | |
| }, | |
| { | |
| "completion_length": 769.8607528686523, | |
| "epoch": 0.8533333333333334, | |
| "grad_norm": 5.1279401779174805, | |
| "kl": 0.787158203125, | |
| "learning_rate": 1.8900610884066817e-07, | |
| "loss": 0.0315, | |
| "reward": 1.9811220198869706, | |
| "reward_std": 0.6900037627667188, | |
| "rewards/accuracy_reward": 0.6357143126428128, | |
| "rewards/cosine_scaled_reward": 0.4329076783033088, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.912500062584877, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.8533333333333334, | |
| "eval_completion_length": 738.6478045166016, | |
| "eval_kl": 0.67065634765625, | |
| "eval_loss": 0.026821324601769447, | |
| "eval_reward": 1.9358687758922577, | |
| "eval_reward_std": 0.681571420711279, | |
| "eval_rewards/accuracy_reward": 0.6160857413113118, | |
| "eval_rewards/cosine_scaled_reward": 0.4032782297934056, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_rewards/reasoning_steps_reward": 0.9165048221349716, | |
| "eval_runtime": 32285.4404, | |
| "eval_samples_per_second": 0.155, | |
| "eval_steps_per_second": 0.011, | |
| "step": 400 | |
| }, | |
| { | |
| "completion_length": 763.4339599609375, | |
| "epoch": 0.864, | |
| "grad_norm": 4.143102169036865, | |
| "kl": 0.609136962890625, | |
| "learning_rate": 1.627288017913383e-07, | |
| "loss": 0.0244, | |
| "reward": 1.9788720414042473, | |
| "reward_std": 0.6925495602190495, | |
| "rewards/accuracy_reward": 0.6375000275671482, | |
| "rewards/cosine_scaled_reward": 0.42411007191985844, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9172619670629502, | |
| "step": 405 | |
| }, | |
| { | |
| "completion_length": 754.2500312805175, | |
| "epoch": 0.8746666666666667, | |
| "grad_norm": 4.33268928527832, | |
| "kl": 0.9586181640625, | |
| "learning_rate": 1.3831291600445573e-07, | |
| "loss": 0.0383, | |
| "reward": 1.9650759071111679, | |
| "reward_std": 0.6423604141920805, | |
| "rewards/accuracy_reward": 0.6303571704775095, | |
| "rewards/cosine_scaled_reward": 0.4222186904400587, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.912500062584877, | |
| "step": 410 | |
| }, | |
| { | |
| "completion_length": 751.8071762084961, | |
| "epoch": 0.8853333333333333, | |
| "grad_norm": 7.097233295440674, | |
| "kl": 0.8556884765625, | |
| "learning_rate": 1.1579243729307487e-07, | |
| "loss": 0.0342, | |
| "reward": 1.9338065341114998, | |
| "reward_std": 0.7414230849593878, | |
| "rewards/accuracy_reward": 0.6321428898721934, | |
| "rewards/cosine_scaled_reward": 0.41178265907801687, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8898810192942619, | |
| "step": 415 | |
| }, | |
| { | |
| "completion_length": 752.8571723937988, | |
| "epoch": 0.896, | |
| "grad_norm": 3.0274124145507812, | |
| "kl": 0.67294921875, | |
| "learning_rate": 9.519871314899092e-08, | |
| "loss": 0.0269, | |
| "reward": 1.9913182631134987, | |
| "reward_std": 0.7086525153368711, | |
| "rewards/accuracy_reward": 0.6571428876370191, | |
| "rewards/cosine_scaled_reward": 0.4359610580140725, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8982143491506577, | |
| "step": 420 | |
| }, | |
| { | |
| "completion_length": 751.7411056518555, | |
| "epoch": 0.9066666666666666, | |
| "grad_norm": 1.3194289207458496, | |
| "kl": 0.722802734375, | |
| "learning_rate": 7.656040910844358e-08, | |
| "loss": 0.0289, | |
| "reward": 2.0188252568244933, | |
| "reward_std": 0.7707155652344226, | |
| "rewards/accuracy_reward": 0.644642885029316, | |
| "rewards/cosine_scaled_reward": 0.44144419142976404, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9327381521463394, | |
| "step": 425 | |
| }, | |
| { | |
| "completion_length": 755.0464630126953, | |
| "epoch": 0.9173333333333333, | |
| "grad_norm": 4.276956081390381, | |
| "kl": 0.9569580078125, | |
| "learning_rate": 5.990346885098235e-08, | |
| "loss": 0.0383, | |
| "reward": 2.000167742371559, | |
| "reward_std": 0.7376608021557332, | |
| "rewards/accuracy_reward": 0.6589285988360644, | |
| "rewards/cosine_scaled_reward": 0.45314384531229734, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8880952954292297, | |
| "step": 430 | |
| }, | |
| { | |
| "completion_length": 727.2536087036133, | |
| "epoch": 0.928, | |
| "grad_norm": 19.139204025268555, | |
| "kl": 1.32947998046875, | |
| "learning_rate": 4.5251078087033493e-08, | |
| "loss": 0.0532, | |
| "reward": 2.039694218337536, | |
| "reward_std": 0.6533694989979267, | |
| "rewards/accuracy_reward": 0.6732143165543676, | |
| "rewards/cosine_scaled_reward": 0.4462417368311435, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9202381521463394, | |
| "step": 435 | |
| }, | |
| { | |
| "completion_length": 734.5536064147949, | |
| "epoch": 0.9386666666666666, | |
| "grad_norm": 9.922527313232422, | |
| "kl": 1.4177001953125, | |
| "learning_rate": 3.262363228443427e-08, | |
| "loss": 0.0567, | |
| "reward": 1.9774114236235618, | |
| "reward_std": 0.7198221303522587, | |
| "rewards/accuracy_reward": 0.6571428865194321, | |
| "rewards/cosine_scaled_reward": 0.4309827778954059, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.8892857789993286, | |
| "step": 440 | |
| }, | |
| { | |
| "completion_length": 755.5053962707519, | |
| "epoch": 0.9493333333333334, | |
| "grad_norm": 3.058717727661133, | |
| "kl": 1.02747802734375, | |
| "learning_rate": 2.2038708278862952e-08, | |
| "loss": 0.0411, | |
| "reward": 1.9413904681801797, | |
| "reward_std": 0.6192027345299721, | |
| "rewards/accuracy_reward": 0.6214285951107741, | |
| "rewards/cosine_scaled_reward": 0.41579519272781906, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9041667267680168, | |
| "step": 445 | |
| }, | |
| { | |
| "completion_length": 723.6143173217773, | |
| "epoch": 0.96, | |
| "grad_norm": 2.64345383644104, | |
| "kl": 0.74544677734375, | |
| "learning_rate": 1.3511039807673209e-08, | |
| "loss": 0.0298, | |
| "reward": 2.1570381984114646, | |
| "reward_std": 0.6153812855482101, | |
| "rewards/accuracy_reward": 0.7089286003261804, | |
| "rewards/cosine_scaled_reward": 0.5165619559586048, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9315476790070534, | |
| "step": 450 | |
| }, | |
| { | |
| "completion_length": 728.894679260254, | |
| "epoch": 0.9706666666666667, | |
| "grad_norm": 2.217505693435669, | |
| "kl": 0.678607177734375, | |
| "learning_rate": 7.0524970011963675e-09, | |
| "loss": 0.0272, | |
| "reward": 2.2157696574926375, | |
| "reward_std": 0.6317826233804226, | |
| "rewards/accuracy_reward": 0.7500000305473804, | |
| "rewards/cosine_scaled_reward": 0.5425553207285703, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9232143491506577, | |
| "step": 455 | |
| }, | |
| { | |
| "completion_length": 722.3839637756348, | |
| "epoch": 0.9813333333333333, | |
| "grad_norm": 3.196773052215576, | |
| "kl": 0.709228515625, | |
| "learning_rate": 2.6720698600553595e-09, | |
| "loss": 0.0284, | |
| "reward": 2.122882993519306, | |
| "reward_std": 0.599827627837658, | |
| "rewards/accuracy_reward": 0.7017857432365417, | |
| "rewards/cosine_scaled_reward": 0.5175257750786841, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9035714983940124, | |
| "step": 460 | |
| }, | |
| { | |
| "completion_length": 754.775032043457, | |
| "epoch": 0.992, | |
| "grad_norm": 8.455827713012695, | |
| "kl": 0.835205078125, | |
| "learning_rate": 3.7585574148779613e-10, | |
| "loss": 0.0334, | |
| "reward": 1.9985675051808358, | |
| "reward_std": 0.7642196819186211, | |
| "rewards/accuracy_reward": 0.6500000316649676, | |
| "rewards/cosine_scaled_reward": 0.4402340850589098, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.9083333894610405, | |
| "step": 465 | |
| }, | |
| { | |
| "completion_length": 746.1607462565104, | |
| "epoch": 0.9984, | |
| "kl": 0.8069661458333334, | |
| "reward": 2.0161508160332837, | |
| "reward_std": 0.7148686709503332, | |
| "rewards/accuracy_reward": 0.6636905111372471, | |
| "rewards/cosine_scaled_reward": 0.4536507367156446, | |
| "rewards/format_reward": 0.0, | |
| "rewards/reasoning_steps_reward": 0.898809589445591, | |
| "step": 468, | |
| "total_flos": 0.0, | |
| "train_loss": 4841.422249500714, | |
| "train_runtime": 180396.3107, | |
| "train_samples_per_second": 0.042, | |
| "train_steps_per_second": 0.003 | |
| } | |
| ], | |
| "logging_steps": 5, | |
| "max_steps": 468, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 1, | |
| "save_steps": 200, | |
| "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 | |
| } | |