| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.9984, | |
| "eval_steps": 400, | |
| "global_step": 234, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 997.4687957763672, | |
| "epoch": 0.004266666666666667, | |
| "grad_norm": 0.4476054012775421, | |
| "kl": 0.0, | |
| "learning_rate": 1.25e-07, | |
| "loss": 0.2117, | |
| "num_tokens": 254317.0, | |
| "reward": 0.5580357387661934, | |
| "reward_std": 0.39642567932605743, | |
| "rewards/accuracy_reward": 0.558035708963871, | |
| "rewards/format_reward": 0.0, | |
| "step": 1 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 833.5591859817505, | |
| "epoch": 0.021333333333333333, | |
| "grad_norm": 0.8373767733573914, | |
| "kl": 0.00017589330673217773, | |
| "learning_rate": 6.25e-07, | |
| "loss": 0.1622, | |
| "num_tokens": 1126402.0, | |
| "reward": 0.5602678819559515, | |
| "reward_std": 0.39950425596907735, | |
| "rewards/accuracy_reward": 0.5602678558789194, | |
| "rewards/format_reward": 0.0, | |
| "step": 5 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 842.7527206420898, | |
| "epoch": 0.042666666666666665, | |
| "grad_norm": 0.7621635794639587, | |
| "kl": 0.0002469301223754883, | |
| "learning_rate": 1.25e-06, | |
| "loss": 0.1515, | |
| "num_tokens": 2246671.0, | |
| "reward": 0.5714285939931869, | |
| "reward_std": 0.3970031665638089, | |
| "rewards/accuracy_reward": 0.5714285712689161, | |
| "rewards/format_reward": 0.0, | |
| "step": 10 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 904.8804016113281, | |
| "epoch": 0.064, | |
| "grad_norm": 1.0540382862091064, | |
| "kl": 0.0004149436950683594, | |
| "learning_rate": 1.875e-06, | |
| "loss": 0.1689, | |
| "num_tokens": 3425519.0, | |
| "reward": 0.598214315623045, | |
| "reward_std": 0.37588157430291175, | |
| "rewards/accuracy_reward": 0.5982142850756645, | |
| "rewards/format_reward": 0.0, | |
| "step": 15 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 803.2098541259766, | |
| "epoch": 0.08533333333333333, | |
| "grad_norm": 0.24481388926506042, | |
| "kl": 0.0009911060333251953, | |
| "learning_rate": 2.5e-06, | |
| "loss": 0.1647, | |
| "num_tokens": 4485645.0, | |
| "reward": 0.6892857506871224, | |
| "reward_std": 0.305409730784595, | |
| "rewards/accuracy_reward": 0.6892857134342194, | |
| "rewards/format_reward": 0.0, | |
| "step": 20 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 742.3911071777344, | |
| "epoch": 0.10666666666666667, | |
| "grad_norm": 0.3323942720890045, | |
| "kl": 0.001956653594970703, | |
| "learning_rate": 2.99983215271541e-06, | |
| "loss": 0.1555, | |
| "num_tokens": 5470801.0, | |
| "reward": 0.7464286014437675, | |
| "reward_std": 0.23826294410973786, | |
| "rewards/accuracy_reward": 0.7464285716414452, | |
| "rewards/format_reward": 0.0, | |
| "step": 25 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 698.3303901672364, | |
| "epoch": 0.128, | |
| "grad_norm": 0.2109682857990265, | |
| "kl": 0.0024506568908691405, | |
| "learning_rate": 2.993961440992859e-06, | |
| "loss": 0.1116, | |
| "num_tokens": 6403683.0, | |
| "reward": 0.8080357491970063, | |
| "reward_std": 0.19967383295297622, | |
| "rewards/accuracy_reward": 0.8080357126891613, | |
| "rewards/format_reward": 0.0, | |
| "step": 30 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 654.9661010742187, | |
| "epoch": 0.14933333333333335, | |
| "grad_norm": 0.1639912873506546, | |
| "kl": 0.0032644271850585938, | |
| "learning_rate": 2.979735890885282e-06, | |
| "loss": 0.0765, | |
| "num_tokens": 7290748.0, | |
| "reward": 0.8232143156230449, | |
| "reward_std": 0.18937840089201927, | |
| "rewards/accuracy_reward": 0.8232142880558968, | |
| "rewards/format_reward": 0.0, | |
| "step": 35 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 707.6750289916993, | |
| "epoch": 0.17066666666666666, | |
| "grad_norm": 0.1635686606168747, | |
| "kl": 0.002868843078613281, | |
| "learning_rate": 2.957235057439301e-06, | |
| "loss": 0.0723, | |
| "num_tokens": 8244687.0, | |
| "reward": 0.7946428894996643, | |
| "reward_std": 0.18750199358910322, | |
| "rewards/accuracy_reward": 0.7946428544819355, | |
| "rewards/format_reward": 0.0, | |
| "step": 40 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 676.180387878418, | |
| "epoch": 0.192, | |
| "grad_norm": 0.14903444051742554, | |
| "kl": 0.003135490417480469, | |
| "learning_rate": 2.9265847744427307e-06, | |
| "loss": 0.0618, | |
| "num_tokens": 9153419.0, | |
| "reward": 0.8321428894996643, | |
| "reward_std": 0.1699943056330085, | |
| "rewards/accuracy_reward": 0.8321428552269936, | |
| "rewards/format_reward": 0.0, | |
| "step": 45 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 623.0259208679199, | |
| "epoch": 0.21333333333333335, | |
| "grad_norm": 0.2258339673280716, | |
| "kl": 0.0038270950317382812, | |
| "learning_rate": 2.887956450710995e-06, | |
| "loss": 0.0603, | |
| "num_tokens": 10003724.0, | |
| "reward": 0.8401786059141159, | |
| "reward_std": 0.16125054527074098, | |
| "rewards/accuracy_reward": 0.8401785671710968, | |
| "rewards/format_reward": 0.0, | |
| "step": 50 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 659.4062774658203, | |
| "epoch": 0.23466666666666666, | |
| "grad_norm": 0.30922192335128784, | |
| "kl": 0.003952980041503906, | |
| "learning_rate": 2.8415661114995055e-06, | |
| "loss": 0.0489, | |
| "num_tokens": 10890953.0, | |
| "reward": 0.808035746216774, | |
| "reward_std": 0.154349534958601, | |
| "rewards/accuracy_reward": 0.8080357141792774, | |
| "rewards/format_reward": 0.0, | |
| "step": 55 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 639.1562812805175, | |
| "epoch": 0.256, | |
| "grad_norm": 0.1367289125919342, | |
| "kl": 0.004214859008789063, | |
| "learning_rate": 2.7876731904027993e-06, | |
| "loss": 0.0428, | |
| "num_tokens": 11784818.0, | |
| "reward": 0.811607176065445, | |
| "reward_std": 0.15857027508318425, | |
| "rewards/accuracy_reward": 0.8116071425378323, | |
| "rewards/format_reward": 0.0, | |
| "step": 60 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 642.5893127441407, | |
| "epoch": 0.2773333333333333, | |
| "grad_norm": 0.3067258298397064, | |
| "kl": 0.004591751098632813, | |
| "learning_rate": 2.726579078496647e-06, | |
| "loss": 0.0456, | |
| "num_tokens": 12662200.0, | |
| "reward": 0.8410714611411094, | |
| "reward_std": 0.15819188673049212, | |
| "rewards/accuracy_reward": 0.8410714238882064, | |
| "rewards/format_reward": 0.0, | |
| "step": 65 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 682.150032043457, | |
| "epoch": 0.2986666666666667, | |
| "grad_norm": 1.2549998760223389, | |
| "kl": 0.005356979370117187, | |
| "learning_rate": 2.6586254388368995e-06, | |
| "loss": 0.0582, | |
| "num_tokens": 13583589.0, | |
| "reward": 0.7830357477068901, | |
| "reward_std": 0.18484641518443823, | |
| "rewards/accuracy_reward": 0.7830357119441033, | |
| "rewards/format_reward": 0.0, | |
| "step": 70 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 705.7393188476562, | |
| "epoch": 0.32, | |
| "grad_norm": 0.2942885756492615, | |
| "kl": 0.00424346923828125, | |
| "learning_rate": 2.584192295741087e-06, | |
| "loss": 0.0507, | |
| "num_tokens": 14528325.0, | |
| "reward": 0.797321455180645, | |
| "reward_std": 0.17518679071217774, | |
| "rewards/accuracy_reward": 0.7973214276134968, | |
| "rewards/format_reward": 0.0, | |
| "step": 75 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 619.9482444763183, | |
| "epoch": 0.3413333333333333, | |
| "grad_norm": 0.17190858721733093, | |
| "kl": 0.00408172607421875, | |
| "learning_rate": 2.5036959095382875e-06, | |
| "loss": 0.0474, | |
| "num_tokens": 15383695.0, | |
| "reward": 0.8223214603960514, | |
| "reward_std": 0.13084097821265459, | |
| "rewards/accuracy_reward": 0.8223214283585548, | |
| "rewards/format_reward": 0.0, | |
| "step": 80 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 651.5598556518555, | |
| "epoch": 0.3626666666666667, | |
| "grad_norm": 0.46220502257347107, | |
| "kl": 0.0038448333740234374, | |
| "learning_rate": 2.4175864486725093e-06, | |
| "loss": 0.0395, | |
| "num_tokens": 16274673.0, | |
| "reward": 0.8464286021888257, | |
| "reward_std": 0.14037740416824818, | |
| "rewards/accuracy_reward": 0.846428568661213, | |
| "rewards/format_reward": 0.0, | |
| "step": 85 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 669.5187744140625, | |
| "epoch": 0.384, | |
| "grad_norm": 0.2906588315963745, | |
| "kl": 0.004624176025390625, | |
| "learning_rate": 2.3263454721781537e-06, | |
| "loss": 0.045, | |
| "num_tokens": 17179556.0, | |
| "reward": 0.8419643200933933, | |
| "reward_std": 0.16921061147004365, | |
| "rewards/accuracy_reward": 0.8419642858207226, | |
| "rewards/format_reward": 0.0, | |
| "step": 90 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 671.370565032959, | |
| "epoch": 0.4053333333333333, | |
| "grad_norm": 0.25627151131629944, | |
| "kl": 0.00406341552734375, | |
| "learning_rate": 2.230483236606551e-06, | |
| "loss": 0.0304, | |
| "num_tokens": 18091140.0, | |
| "reward": 0.7982143193483353, | |
| "reward_std": 0.1869086405262351, | |
| "rewards/accuracy_reward": 0.7982142873108387, | |
| "rewards/format_reward": 0.0, | |
| "step": 95 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 673.0446693420411, | |
| "epoch": 0.4266666666666667, | |
| "grad_norm": 0.8661001920700073, | |
| "kl": 0.004367828369140625, | |
| "learning_rate": 2.1305358424643485e-06, | |
| "loss": 0.0367, | |
| "num_tokens": 18998523.0, | |
| "reward": 0.8035714536905288, | |
| "reward_std": 0.17903811391443014, | |
| "rewards/accuracy_reward": 0.8035714350640774, | |
| "rewards/format_reward": 0.0, | |
| "step": 100 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 663.0018180847168, | |
| "epoch": 0.448, | |
| "grad_norm": 0.28504350781440735, | |
| "kl": 0.004234695434570312, | |
| "learning_rate": 2.027062236122014e-06, | |
| "loss": 0.023, | |
| "num_tokens": 19901861.0, | |
| "reward": 0.8026786029338837, | |
| "reward_std": 0.19563668854534627, | |
| "rewards/accuracy_reward": 0.802678569406271, | |
| "rewards/format_reward": 0.0, | |
| "step": 105 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 668.9250289916993, | |
| "epoch": 0.4693333333333333, | |
| "grad_norm": 0.23065027594566345, | |
| "kl": 0.005011749267578125, | |
| "learning_rate": 1.9206410839590043e-06, | |
| "loss": 0.0265, | |
| "num_tokens": 20801382.0, | |
| "reward": 0.8017857454717159, | |
| "reward_std": 0.1697748264297843, | |
| "rewards/accuracy_reward": 0.8017857126891613, | |
| "rewards/format_reward": 0.0, | |
| "step": 110 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 669.5696678161621, | |
| "epoch": 0.49066666666666664, | |
| "grad_norm": 0.25070255994796753, | |
| "kl": 0.004204177856445312, | |
| "learning_rate": 1.8118675362266389e-06, | |
| "loss": 0.04, | |
| "num_tokens": 21713070.0, | |
| "reward": 0.8071428939700127, | |
| "reward_std": 0.15041764192283152, | |
| "rewards/accuracy_reward": 0.8071428596973419, | |
| "rewards/format_reward": 0.0, | |
| "step": 115 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 745.8312896728515, | |
| "epoch": 0.512, | |
| "grad_norm": 0.2231619954109192, | |
| "kl": 0.004381752014160157, | |
| "learning_rate": 1.7013498987264833e-06, | |
| "loss": 0.0291, | |
| "num_tokens": 22706608.0, | |
| "reward": 0.7928571775555611, | |
| "reward_std": 0.15372475292533636, | |
| "rewards/accuracy_reward": 0.7928571425378322, | |
| "rewards/format_reward": 0.0, | |
| "step": 120 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 683.2152099609375, | |
| "epoch": 0.5333333333333333, | |
| "grad_norm": 0.22776393592357635, | |
| "kl": 0.004504776000976563, | |
| "learning_rate": 1.5897062309175513e-06, | |
| "loss": 0.0457, | |
| "num_tokens": 23648209.0, | |
| "reward": 0.7892857491970062, | |
| "reward_std": 0.17449938021600248, | |
| "rewards/accuracy_reward": 0.7892857134342194, | |
| "rewards/format_reward": 0.0, | |
| "step": 125 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 657.4553871154785, | |
| "epoch": 0.5546666666666666, | |
| "grad_norm": 0.16024665534496307, | |
| "kl": 0.00522613525390625, | |
| "learning_rate": 1.4775608894771048e-06, | |
| "loss": 0.0366, | |
| "num_tokens": 24549815.0, | |
| "reward": 0.8125000387430191, | |
| "reward_std": 0.1693068951368332, | |
| "rewards/accuracy_reward": 0.8124999962747097, | |
| "rewards/format_reward": 0.0, | |
| "step": 130 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 642.0161018371582, | |
| "epoch": 0.576, | |
| "grad_norm": 0.15500028431415558, | |
| "kl": 0.004698562622070313, | |
| "learning_rate": 1.3655410366448499e-06, | |
| "loss": 0.0268, | |
| "num_tokens": 25438763.0, | |
| "reward": 0.8062500298023224, | |
| "reward_std": 0.1475358337163925, | |
| "rewards/accuracy_reward": 0.806250000745058, | |
| "rewards/format_reward": 0.0, | |
| "step": 135 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 729.4768188476562, | |
| "epoch": 0.5973333333333334, | |
| "grad_norm": 0.19548794627189636, | |
| "kl": 0.004253196716308594, | |
| "learning_rate": 1.2542731328772936e-06, | |
| "loss": 0.0278, | |
| "num_tokens": 26409399.0, | |
| "reward": 0.8008928880095482, | |
| "reward_std": 0.14697162248194218, | |
| "rewards/accuracy_reward": 0.8008928537368775, | |
| "rewards/format_reward": 0.0, | |
| "step": 140 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 665.8884208679199, | |
| "epoch": 0.6186666666666667, | |
| "grad_norm": 0.23221784830093384, | |
| "kl": 0.005530166625976563, | |
| "learning_rate": 1.1443794334267539e-06, | |
| "loss": 0.0404, | |
| "num_tokens": 27322886.0, | |
| "reward": 0.8187500342726708, | |
| "reward_std": 0.15602440275251866, | |
| "rewards/accuracy_reward": 0.8187499970197678, | |
| "rewards/format_reward": 0.0, | |
| "step": 145 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 676.9500312805176, | |
| "epoch": 0.64, | |
| "grad_norm": 0.18050934374332428, | |
| "kl": 0.0050506591796875, | |
| "learning_rate": 1.036474508437579e-06, | |
| "loss": 0.0525, | |
| "num_tokens": 28241951.0, | |
| "reward": 0.8491071745753288, | |
| "reward_std": 0.16172744631767272, | |
| "rewards/accuracy_reward": 0.8491071447730064, | |
| "rewards/format_reward": 0.0, | |
| "step": 150 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 642.0652069091797, | |
| "epoch": 0.6613333333333333, | |
| "grad_norm": 0.12173999845981598, | |
| "kl": 0.004921722412109375, | |
| "learning_rate": 9.311618060206075e-07, | |
| "loss": 0.0422, | |
| "num_tokens": 29117563.0, | |
| "reward": 0.805357176065445, | |
| "reward_std": 0.17488674409687519, | |
| "rewards/accuracy_reward": 0.8053571425378323, | |
| "rewards/format_reward": 0.0, | |
| "step": 155 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 716.8884292602539, | |
| "epoch": 0.6826666666666666, | |
| "grad_norm": 0.2920779585838318, | |
| "kl": 0.004403495788574218, | |
| "learning_rate": 8.290302775265509e-07, | |
| "loss": 0.0223, | |
| "num_tokens": 30082031.0, | |
| "reward": 0.7910714626312256, | |
| "reward_std": 0.1914942855015397, | |
| "rewards/accuracy_reward": 0.7910714291036129, | |
| "rewards/format_reward": 0.0, | |
| "step": 160 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 701.8036018371582, | |
| "epoch": 0.704, | |
| "grad_norm": 0.9731239080429077, | |
| "kl": 0.00501861572265625, | |
| "learning_rate": 7.30651083891141e-07, | |
| "loss": 0.0393, | |
| "num_tokens": 31018320.0, | |
| "reward": 0.8223214641213417, | |
| "reward_std": 0.19152794536203147, | |
| "rewards/accuracy_reward": 0.8223214253783226, | |
| "rewards/format_reward": 0.0, | |
| "step": 165 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 686.908959197998, | |
| "epoch": 0.7253333333333334, | |
| "grad_norm": 0.21161052584648132, | |
| "kl": 0.004227638244628906, | |
| "learning_rate": 6.3657440147149e-07, | |
| "loss": 0.0509, | |
| "num_tokens": 31945256.0, | |
| "reward": 0.8053571827709675, | |
| "reward_std": 0.1896941650658846, | |
| "rewards/accuracy_reward": 0.8110118992626667, | |
| "rewards/format_reward": 0.0, | |
| "step": 170 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 729.9232475280762, | |
| "epoch": 0.7466666666666667, | |
| "grad_norm": 0.15954989194869995, | |
| "kl": 0.004137992858886719, | |
| "learning_rate": 5.473263452367318e-07, | |
| "loss": 0.0326, | |
| "num_tokens": 32932758.0, | |
| "reward": 0.7705357417464256, | |
| "reward_std": 0.18078458309173584, | |
| "rewards/accuracy_reward": 0.7705357111990452, | |
| "rewards/format_reward": 0.0, | |
| "step": 175 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 649.5544960021973, | |
| "epoch": 0.768, | |
| "grad_norm": 0.13724558055400848, | |
| "kl": 0.004343414306640625, | |
| "learning_rate": 4.63406026519703e-07, | |
| "loss": 0.027, | |
| "num_tokens": 33826439.0, | |
| "reward": 0.8089286006987095, | |
| "reward_std": 0.15814926102757454, | |
| "rewards/accuracy_reward": 0.8089285723865032, | |
| "rewards/format_reward": 0.0, | |
| "step": 180 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 718.4714622497559, | |
| "epoch": 0.7893333333333333, | |
| "grad_norm": 0.14736232161521912, | |
| "kl": 0.004001426696777344, | |
| "learning_rate": 3.852827617839085e-07, | |
| "loss": 0.0637, | |
| "num_tokens": 34789978.0, | |
| "reward": 0.8017857491970062, | |
| "reward_std": 0.1630306661128998, | |
| "rewards/accuracy_reward": 0.8017857111990452, | |
| "rewards/format_reward": 0.0, | |
| "step": 185 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 673.543782043457, | |
| "epoch": 0.8106666666666666, | |
| "grad_norm": 0.3745248019695282, | |
| "kl": 0.004261016845703125, | |
| "learning_rate": 3.133934480154885e-07, | |
| "loss": 0.032, | |
| "num_tokens": 35712907.0, | |
| "reward": 0.8008928954601288, | |
| "reward_std": 0.1807129830121994, | |
| "rewards/accuracy_reward": 0.8008928552269936, | |
| "rewards/format_reward": 0.0, | |
| "step": 190 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 664.7464561462402, | |
| "epoch": 0.832, | |
| "grad_norm": 0.2295093536376953, | |
| "kl": 0.004295730590820312, | |
| "learning_rate": 2.48140119418046e-07, | |
| "loss": 0.0484, | |
| "num_tokens": 36621153.0, | |
| "reward": 0.8062500327825546, | |
| "reward_std": 0.16623926647007464, | |
| "rewards/accuracy_reward": 0.8122023768723011, | |
| "rewards/format_reward": 0.0, | |
| "step": 195 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 727.7053886413574, | |
| "epoch": 0.8533333333333334, | |
| "grad_norm": 0.31535524129867554, | |
| "kl": 0.003837394714355469, | |
| "learning_rate": 1.8988769907430552e-07, | |
| "loss": 0.0474, | |
| "num_tokens": 37591786.0, | |
| "reward": 0.7875000357627868, | |
| "reward_std": 0.16348065678030252, | |
| "rewards/accuracy_reward": 0.7875000007450581, | |
| "rewards/format_reward": 0.0, | |
| "step": 200 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 673.6312782287598, | |
| "epoch": 0.8746666666666667, | |
| "grad_norm": 0.2970116138458252, | |
| "kl": 0.0037149429321289063, | |
| "learning_rate": 1.3896195814820269e-07, | |
| "loss": 0.0408, | |
| "num_tokens": 38502885.0, | |
| "reward": 0.842857176065445, | |
| "reward_std": 0.15663124285638333, | |
| "rewards/accuracy_reward": 0.8428571417927742, | |
| "rewards/format_reward": 0.0, | |
| "step": 205 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 658.6187789916992, | |
| "epoch": 0.896, | |
| "grad_norm": 0.3235113322734833, | |
| "kl": 0.005364990234375, | |
| "learning_rate": 9.564769404039419e-08, | |
| "loss": 0.0322, | |
| "num_tokens": 39400271.0, | |
| "reward": 0.7964285999536515, | |
| "reward_std": 0.19183696284890175, | |
| "rewards/accuracy_reward": 0.7964285723865032, | |
| "rewards/format_reward": 0.0, | |
| "step": 210 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 658.0768119812012, | |
| "epoch": 0.9173333333333333, | |
| "grad_norm": 0.19358721375465393, | |
| "kl": 0.0041599273681640625, | |
| "learning_rate": 6.018713768566658e-08, | |
| "loss": 0.024, | |
| "num_tokens": 40305926.0, | |
| "reward": 0.8125000312924385, | |
| "reward_std": 0.16783579289913178, | |
| "rewards/accuracy_reward": 0.8125, | |
| "rewards/format_reward": 0.0, | |
| "step": 215 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 672.805387878418, | |
| "epoch": 0.9386666666666666, | |
| "grad_norm": 0.2503073513507843, | |
| "kl": 0.004539871215820312, | |
| "learning_rate": 3.277859889929147e-08, | |
| "loss": 0.0397, | |
| "num_tokens": 41216982.0, | |
| "reward": 0.8482143253087997, | |
| "reward_std": 0.20670701321214438, | |
| "rewards/accuracy_reward": 0.8482142858207226, | |
| "rewards/format_reward": 0.0, | |
| "step": 220 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 654.6687858581543, | |
| "epoch": 0.96, | |
| "grad_norm": 0.17779560387134552, | |
| "kl": 0.0040149688720703125, | |
| "learning_rate": 1.357535734809795e-08, | |
| "loss": 0.0372, | |
| "num_tokens": 42114081.0, | |
| "reward": 0.8000000335276127, | |
| "reward_std": 0.16292541287839413, | |
| "rewards/accuracy_reward": 0.7999999985098839, | |
| "rewards/format_reward": 0.0, | |
| "step": 225 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 646.3321731567382, | |
| "epoch": 0.9813333333333333, | |
| "grad_norm": 0.2772572338581085, | |
| "kl": 0.004202651977539063, | |
| "learning_rate": 2.684805348397268e-09, | |
| "loss": 0.0411, | |
| "num_tokens": 43004377.0, | |
| "reward": 0.805357177555561, | |
| "reward_std": 0.17196922078728677, | |
| "rewards/accuracy_reward": 0.8053571373224259, | |
| "rewards/format_reward": 0.0, | |
| "step": 230 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 680.2171030044556, | |
| "epoch": 0.9984, | |
| "kl": 0.003808736801147461, | |
| "num_tokens": 43742868.0, | |
| "reward": 0.8024553880095482, | |
| "reward_std": 0.17644294258207083, | |
| "rewards/accuracy_reward": 0.802455360069871, | |
| "rewards/format_reward": 0.0, | |
| "step": 234, | |
| "total_flos": 0.0, | |
| "train_loss": 0.055906044398872264, | |
| "train_runtime": 36259.8049, | |
| "train_samples_per_second": 0.207, | |
| "train_steps_per_second": 0.006 | |
| } | |
| ], | |
| "logging_steps": 5, | |
| "max_steps": 234, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 1, | |
| "save_steps": 100, | |
| "stateful_callbacks": { | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": true | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 0.0, | |
| "train_batch_size": 4, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |