| { | |
| "best_global_step": null, | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.9954198473282443, | |
| "eval_steps": 2000000, | |
| "global_step": 163, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 702.7132034301758, | |
| "epoch": 0.0061068702290076335, | |
| "grad_norm": 0.13332052615614565, | |
| "kl": 0.0, | |
| "learning_rate": 5.88235294117647e-08, | |
| "loss": 0.0341, | |
| "num_tokens": 880751.0, | |
| "reward": 0.770089328289032, | |
| "reward_std": 0.13822399266064167, | |
| "rewards/accuracy_reward": 0.27678571827709675, | |
| "rewards/format_reward": 0.9866071343421936, | |
| "step": 1 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 739.6551666259766, | |
| "epoch": 0.030534351145038167, | |
| "grad_norm": 0.2044588463598013, | |
| "kl": 5.408376455307007e-05, | |
| "learning_rate": 2.941176470588235e-07, | |
| "loss": 0.0268, | |
| "num_tokens": 4507619.0, | |
| "reward": 0.7359096258878708, | |
| "reward_std": 0.1397800410632044, | |
| "rewards/accuracy_reward": 0.24414062616415322, | |
| "rewards/format_reward": 0.9835379421710968, | |
| "step": 5 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 739.3268173217773, | |
| "epoch": 0.061068702290076333, | |
| "grad_norm": 0.2606158264843265, | |
| "kl": 7.957220077514648e-05, | |
| "learning_rate": 5.88235294117647e-07, | |
| "loss": 0.0332, | |
| "num_tokens": 9079091.0, | |
| "reward": 0.7665178939700127, | |
| "reward_std": 0.14487640419974923, | |
| "rewards/accuracy_reward": 0.2741071430966258, | |
| "rewards/format_reward": 0.9848214223980903, | |
| "step": 10 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 733.3120864868164, | |
| "epoch": 0.0916030534351145, | |
| "grad_norm": 0.12473067773197893, | |
| "kl": 0.0001289844512939453, | |
| "learning_rate": 8.823529411764705e-07, | |
| "loss": 0.0328, | |
| "num_tokens": 13590105.0, | |
| "reward": 0.7473214596509934, | |
| "reward_std": 0.1489832980558276, | |
| "rewards/accuracy_reward": 0.2566964281722903, | |
| "rewards/format_reward": 0.9812499925494194, | |
| "step": 15 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 713.0060607910157, | |
| "epoch": 0.12213740458015267, | |
| "grad_norm": 0.10652200791094474, | |
| "kl": 0.0002826213836669922, | |
| "learning_rate": 9.989585804326962e-07, | |
| "loss": 0.0185, | |
| "num_tokens": 18037460.0, | |
| "reward": 0.7446428909897804, | |
| "reward_std": 0.13395796837285162, | |
| "rewards/accuracy_reward": 0.2495535710826516, | |
| "rewards/format_reward": 0.9901785641908646, | |
| "step": 20 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 721.1808349609375, | |
| "epoch": 0.15267175572519084, | |
| "grad_norm": 0.20214832387838885, | |
| "kl": 0.0041507244110107425, | |
| "learning_rate": 9.926100533780304e-07, | |
| "loss": 0.0254, | |
| "num_tokens": 22546158.0, | |
| "reward": 0.7579241439700126, | |
| "reward_std": 0.14215883370488883, | |
| "rewards/accuracy_reward": 0.26495535727590325, | |
| "rewards/format_reward": 0.9859374895691871, | |
| "step": 25 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 742.2864181518555, | |
| "epoch": 0.183206106870229, | |
| "grad_norm": 0.1594816777583089, | |
| "kl": 0.0016126632690429688, | |
| "learning_rate": 9.805648919361503e-07, | |
| "loss": 0.0357, | |
| "num_tokens": 27073905.0, | |
| "reward": 0.7584821790456772, | |
| "reward_std": 0.12880926295183598, | |
| "rewards/accuracy_reward": 0.2645089291036129, | |
| "rewards/format_reward": 0.9879464194178581, | |
| "step": 30 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 725.3341888427734, | |
| "epoch": 0.21374045801526717, | |
| "grad_norm": 0.09889141258603189, | |
| "kl": 0.001922607421875, | |
| "learning_rate": 9.62962388596925e-07, | |
| "loss": 0.0234, | |
| "num_tokens": 31605946.0, | |
| "reward": 0.7930803909897804, | |
| "reward_std": 0.12912328215315938, | |
| "rewards/accuracy_reward": 0.2970982141792774, | |
| "rewards/format_reward": 0.9919642791152, | |
| "step": 35 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 710.5219085693359, | |
| "epoch": 0.24427480916030533, | |
| "grad_norm": 0.12468270475594573, | |
| "kl": 0.0033203125, | |
| "learning_rate": 9.400061019867678e-07, | |
| "loss": 0.0202, | |
| "num_tokens": 36024036.0, | |
| "reward": 0.7656250327825547, | |
| "reward_std": 0.12553255576640368, | |
| "rewards/accuracy_reward": 0.26986607145518066, | |
| "rewards/format_reward": 0.9915178537368774, | |
| "step": 40 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 691.2645370483399, | |
| "epoch": 0.2748091603053435, | |
| "grad_norm": 0.10338268423256407, | |
| "kl": 0.004708480834960937, | |
| "learning_rate": 9.11961502878777e-07, | |
| "loss": 0.025, | |
| "num_tokens": 40342221.0, | |
| "reward": 0.789732177555561, | |
| "reward_std": 0.13373914500698447, | |
| "rewards/accuracy_reward": 0.29419642593711615, | |
| "rewards/format_reward": 0.9910714223980903, | |
| "step": 45 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 689.1975814819336, | |
| "epoch": 0.3053435114503817, | |
| "grad_norm": 0.11801883674056522, | |
| "kl": 0.006087493896484375, | |
| "learning_rate": 8.791529042392812e-07, | |
| "loss": 0.0221, | |
| "num_tokens": 44648378.0, | |
| "reward": 0.7669643178582192, | |
| "reward_std": 0.12678753938525916, | |
| "rewards/accuracy_reward": 0.2709821423981339, | |
| "rewards/format_reward": 0.9919642791152, | |
| "step": 50 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 681.3627517700195, | |
| "epoch": 0.33587786259541985, | |
| "grad_norm": 0.13594270458309027, | |
| "kl": 0.007537078857421875, | |
| "learning_rate": 8.419597108123053e-07, | |
| "loss": 0.0159, | |
| "num_tokens": 48948507.0, | |
| "reward": 0.7919643267989158, | |
| "reward_std": 0.1272535071708262, | |
| "rewards/accuracy_reward": 0.2941964280791581, | |
| "rewards/format_reward": 0.9955357074737549, | |
| "step": 55 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 667.7931106567382, | |
| "epoch": 0.366412213740458, | |
| "grad_norm": 0.12780945861669307, | |
| "kl": 0.00973663330078125, | |
| "learning_rate": 8.008120316124611e-07, | |
| "loss": 0.0151, | |
| "num_tokens": 53166244.0, | |
| "reward": 0.7524553954601287, | |
| "reward_std": 0.12692366167902946, | |
| "rewards/accuracy_reward": 0.25491071604192256, | |
| "rewards/format_reward": 0.9950892791152001, | |
| "step": 60 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 696.0245849609375, | |
| "epoch": 0.3969465648854962, | |
| "grad_norm": 0.12051116100632073, | |
| "kl": 0.01003570556640625, | |
| "learning_rate": 7.561857060642119e-07, | |
| "loss": 0.0107, | |
| "num_tokens": 57558962.0, | |
| "reward": 0.7633928894996643, | |
| "reward_std": 0.13064181264489888, | |
| "rewards/accuracy_reward": 0.26562499944120643, | |
| "rewards/format_reward": 0.9955357104539871, | |
| "step": 65 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 662.5156509399415, | |
| "epoch": 0.42748091603053434, | |
| "grad_norm": 0.11229836339791986, | |
| "kl": 0.01019439697265625, | |
| "learning_rate": 7.085968013061584e-07, | |
| "loss": 0.0178, | |
| "num_tokens": 61733288.0, | |
| "reward": 0.7448661029338837, | |
| "reward_std": 0.11164179369807244, | |
| "rewards/accuracy_reward": 0.2475446429103613, | |
| "rewards/format_reward": 0.9946428507566452, | |
| "step": 70 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 682.3205673217774, | |
| "epoch": 0.4580152671755725, | |
| "grad_norm": 0.12731781754060964, | |
| "kl": 0.009069061279296875, | |
| "learning_rate": 6.585956442945531e-07, | |
| "loss": 0.0177, | |
| "num_tokens": 66027892.0, | |
| "reward": 0.7617187812924385, | |
| "reward_std": 0.11499134246259927, | |
| "rewards/accuracy_reward": 0.26450892849825325, | |
| "rewards/format_reward": 0.9944196343421936, | |
| "step": 75 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 700.5924423217773, | |
| "epoch": 0.48854961832061067, | |
| "grad_norm": 0.11204035822940968, | |
| "kl": 0.009134674072265625, | |
| "learning_rate": 6.06760457719898e-07, | |
| "loss": 0.0133, | |
| "num_tokens": 70433842.0, | |
| "reward": 0.8136161074042321, | |
| "reward_std": 0.12341635385528207, | |
| "rewards/accuracy_reward": 0.3149553577415645, | |
| "rewards/format_reward": 0.9973214238882064, | |
| "step": 80 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 667.9774856567383, | |
| "epoch": 0.5190839694656488, | |
| "grad_norm": 0.08504179774118455, | |
| "kl": 0.010693359375, | |
| "learning_rate": 5.536906733320815e-07, | |
| "loss": 0.0095, | |
| "num_tokens": 74684629.0, | |
| "reward": 0.7792411088943482, | |
| "reward_std": 0.10279256403446198, | |
| "rewards/accuracy_reward": 0.28125000055879357, | |
| "rewards/format_reward": 0.9959821373224258, | |
| "step": 85 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 672.8207870483399, | |
| "epoch": 0.549618320610687, | |
| "grad_norm": 0.09677151139504983, | |
| "kl": 0.0111236572265625, | |
| "learning_rate": 5e-07, | |
| "loss": 0.0082, | |
| "num_tokens": 78941554.0, | |
| "reward": 0.7789062857627869, | |
| "reward_std": 0.10907009486109018, | |
| "rewards/accuracy_reward": 0.2801339304074645, | |
| "rewards/format_reward": 0.997544638812542, | |
| "step": 90 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 676.2569496154786, | |
| "epoch": 0.5801526717557252, | |
| "grad_norm": 0.11724442552756359, | |
| "kl": 0.0115997314453125, | |
| "learning_rate": 4.463093266679185e-07, | |
| "loss": 0.0106, | |
| "num_tokens": 83189777.0, | |
| "reward": 0.7709821790456772, | |
| "reward_std": 0.11432018820196391, | |
| "rewards/accuracy_reward": 0.27232142791617664, | |
| "rewards/format_reward": 0.9973214238882064, | |
| "step": 95 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 675.6011459350586, | |
| "epoch": 0.6106870229007634, | |
| "grad_norm": 0.10650310445039746, | |
| "kl": 0.01142425537109375, | |
| "learning_rate": 3.932395422801019e-07, | |
| "loss": 0.0117, | |
| "num_tokens": 87446302.0, | |
| "reward": 0.796875037252903, | |
| "reward_std": 0.11783077660948038, | |
| "rewards/accuracy_reward": 0.2979910722468048, | |
| "rewards/format_reward": 0.9977678522467613, | |
| "step": 100 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 666.8801651000977, | |
| "epoch": 0.6412213740458015, | |
| "grad_norm": 0.10284875734947345, | |
| "kl": 0.01230621337890625, | |
| "learning_rate": 3.41404355705447e-07, | |
| "loss": 0.0094, | |
| "num_tokens": 91633661.0, | |
| "reward": 0.8254464611411094, | |
| "reward_std": 0.11524229180067777, | |
| "rewards/accuracy_reward": 0.3265624988824129, | |
| "rewards/format_reward": 0.9977678552269935, | |
| "step": 105 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 708.1185592651367, | |
| "epoch": 0.6717557251908397, | |
| "grad_norm": 0.09648724954516992, | |
| "kl": 0.0115264892578125, | |
| "learning_rate": 2.914031986938417e-07, | |
| "loss": 0.0124, | |
| "num_tokens": 96055112.0, | |
| "reward": 0.7448661044239998, | |
| "reward_std": 0.12669871849939227, | |
| "rewards/accuracy_reward": 0.24642857336439192, | |
| "rewards/format_reward": 0.9968749955296516, | |
| "step": 110 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 677.4897605895997, | |
| "epoch": 0.7022900763358778, | |
| "grad_norm": 0.1011113055763225, | |
| "kl": 0.0118255615234375, | |
| "learning_rate": 2.4381429393578815e-07, | |
| "loss": 0.009, | |
| "num_tokens": 100317890.0, | |
| "reward": 0.7813616394996643, | |
| "reward_std": 0.1330147437751293, | |
| "rewards/accuracy_reward": 0.2825892847031355, | |
| "rewards/format_reward": 0.997544638812542, | |
| "step": 115 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 678.3256973266601, | |
| "epoch": 0.732824427480916, | |
| "grad_norm": 0.15245123106678932, | |
| "kl": 0.01221771240234375, | |
| "learning_rate": 1.991879683875386e-07, | |
| "loss": 0.0104, | |
| "num_tokens": 104592629.0, | |
| "reward": 0.7625000357627869, | |
| "reward_std": 0.11710045160725713, | |
| "rewards/accuracy_reward": 0.2633928569033742, | |
| "rewards/format_reward": 0.9982142806053161, | |
| "step": 120 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 670.6040435791016, | |
| "epoch": 0.7633587786259542, | |
| "grad_norm": 0.11742459382336433, | |
| "kl": 0.01245880126953125, | |
| "learning_rate": 1.5804028918769485e-07, | |
| "loss": 0.0114, | |
| "num_tokens": 108835759.0, | |
| "reward": 0.7925223559141159, | |
| "reward_std": 0.1248929288238287, | |
| "rewards/accuracy_reward": 0.29352678582072256, | |
| "rewards/format_reward": 0.9979910671710968, | |
| "step": 125 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 659.5288223266601, | |
| "epoch": 0.7938931297709924, | |
| "grad_norm": 0.09688477664215772, | |
| "kl": 0.012872314453125, | |
| "learning_rate": 1.2084709576071883e-07, | |
| "loss": 0.0099, | |
| "num_tokens": 112970592.0, | |
| "reward": 0.7906250342726707, | |
| "reward_std": 0.12112973481416703, | |
| "rewards/accuracy_reward": 0.29174106996506455, | |
| "rewards/format_reward": 0.9977678537368775, | |
| "step": 130 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 698.9047195434571, | |
| "epoch": 0.8244274809160306, | |
| "grad_norm": 0.10609676202620595, | |
| "kl": 0.0119049072265625, | |
| "learning_rate": 8.803849712122291e-08, | |
| "loss": 0.0084, | |
| "num_tokens": 117351077.0, | |
| "reward": 0.7845982551574707, | |
| "reward_std": 0.11648994972929358, | |
| "rewards/accuracy_reward": 0.28571428507566454, | |
| "rewards/format_reward": 0.9977678522467613, | |
| "step": 135 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 688.1453430175782, | |
| "epoch": 0.8549618320610687, | |
| "grad_norm": 0.08115857291878435, | |
| "kl": 0.0126739501953125, | |
| "learning_rate": 5.999389801323218e-08, | |
| "loss": 0.0113, | |
| "num_tokens": 121661224.0, | |
| "reward": 0.8302455723285675, | |
| "reward_std": 0.12670655427500604, | |
| "rewards/accuracy_reward": 0.3314732125028968, | |
| "rewards/format_reward": 0.9975446373224258, | |
| "step": 140 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 672.133511352539, | |
| "epoch": 0.8854961832061069, | |
| "grad_norm": 0.10287154621758528, | |
| "kl": 0.01266937255859375, | |
| "learning_rate": 3.7037611403075096e-08, | |
| "loss": 0.0093, | |
| "num_tokens": 125919062.0, | |
| "reward": 0.8181920021772384, | |
| "reward_std": 0.12157530700787902, | |
| "rewards/accuracy_reward": 0.3189732149243355, | |
| "rewards/format_reward": 0.9984374955296517, | |
| "step": 145 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 659.8361923217774, | |
| "epoch": 0.916030534351145, | |
| "grad_norm": 0.11539905461898202, | |
| "kl": 0.01181793212890625, | |
| "learning_rate": 1.943510806384968e-08, | |
| "loss": 0.0061, | |
| "num_tokens": 130131040.0, | |
| "reward": 0.781919676065445, | |
| "reward_std": 0.10376817025244237, | |
| "rewards/accuracy_reward": 0.2823660712689161, | |
| "rewards/format_reward": 0.9991071403026581, | |
| "step": 150 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 663.9966735839844, | |
| "epoch": 0.9465648854961832, | |
| "grad_norm": 0.14485603251022727, | |
| "kl": 0.01304473876953125, | |
| "learning_rate": 7.389946621969678e-09, | |
| "loss": 0.0114, | |
| "num_tokens": 134339337.0, | |
| "reward": 0.759709857404232, | |
| "reward_std": 0.12261311169713736, | |
| "rewards/accuracy_reward": 0.26138392817229034, | |
| "rewards/format_reward": 0.996651777625084, | |
| "step": 155 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 664.3817291259766, | |
| "epoch": 0.9770992366412213, | |
| "grad_norm": 0.10778077233274512, | |
| "kl": 0.0122100830078125, | |
| "learning_rate": 1.0414195673039138e-09, | |
| "loss": 0.0035, | |
| "num_tokens": 138539071.0, | |
| "reward": 0.7744419991970062, | |
| "reward_std": 0.10305451611056923, | |
| "rewards/accuracy_reward": 0.27477678619325163, | |
| "rewards/format_reward": 0.9993303567171097, | |
| "step": 160 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 704.554547627767, | |
| "epoch": 0.9954198473282443, | |
| "kl": 0.012700398763020834, | |
| "num_tokens": 141142234.0, | |
| "reward": 0.788318489988645, | |
| "reward_std": 0.12267326470464468, | |
| "rewards/accuracy_reward": 0.2898065475746989, | |
| "rewards/format_reward": 0.9970238034923872, | |
| "step": 163, | |
| "total_flos": 0.0, | |
| "train_loss": 0.015952993104337183, | |
| "train_runtime": 32110.3153, | |
| "train_samples_per_second": 0.571, | |
| "train_steps_per_second": 0.005 | |
| } | |
| ], | |
| "logging_steps": 5, | |
| "max_steps": 163, | |
| "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": true, | |
| "should_training_stop": true | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 0.0, | |
| "train_batch_size": 16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |