{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 0.5681818181818182, "eval_steps": 500, "global_step": 50, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "clip_ratio": 2.6910373094324314e-06, "completions/clipped_ratio": 0.36328125, "completions/max_length": 4096.0, "completions/max_terminated_length": 4087.0, "completions/mean_length": 2821.140625, "completions/mean_terminated_length": 2098.60107421875, "completions/min_length": 363.0, "completions/min_terminated_length": 363.0, "epoch": 0.011363636363636364, "frac_reward_zero_std": 0.125, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 1549064.0, "register/okay": 0.0, "register/okay_any": 0.9375, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.5703125, "reward_std": 0.4955156147480011, "rewards/correct/mean": 0.5703125, "rewards/correct/std": 0.4955156147480011, "rollout/chars": 7862.0, "rollout/commit": 0.84375, "rollout/reward": 0.6875, "rollout/trunc": 0.28125, "sampling/importance_sampling_ratio/max": 2.773446559906006, "sampling/importance_sampling_ratio/mean": 0.9391552209854126, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 4.223902702331543, "sampling/sampling_logp_difference/mean": 0.004874078091233969, "step": 1, "step_time": 262.9772285820072 }, { "epoch": 0.011363636363636364, "grad_norm": 0.0036677950993180275, "learning_rate": 0.0, "loss": 0.07016487419605255, "step": 1 }, { "clip_ratio": 2.384185791015625e-06, "completions/clipped_ratio": 0.353515625, "completions/max_length": 4096.0, "completions/max_terminated_length": 4077.0, "completions/mean_length": 2826.212890625, "completions/mean_terminated_length": 2131.85791015625, "completions/min_length": 440.0, "completions/min_terminated_length": 440.0, "epoch": 0.022727272727272728, "frac_reward_zero_std": 0.09375, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 3072933.0, "register/okay": 0.0, "register/okay_any": 1.0, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.583984375, "reward_std": 0.493378221988678, "rewards/correct/mean": 0.583984375, "rewards/correct/std": 0.493378221988678, "rollout/chars": 6436.875, "rollout/commit": 0.875, "rollout/reward": 0.78125, "rollout/trunc": 0.1875, "sampling/importance_sampling_ratio/max": 2.993257999420166, "sampling/importance_sampling_ratio/mean": 0.9715690016746521, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.5188601016998291, "sampling/sampling_logp_difference/mean": 0.00453161308541894, "step": 2, "step_time": 234.78294498598916 }, { "epoch": 0.022727272727272728, "grad_norm": 0.0031095428857952356, "learning_rate": 2.0000000000000003e-06, "loss": 0.03657037392258644, "step": 2 }, { "clip_ratio": 1.0142219366571226e-06, "completions/clipped_ratio": 0.3125, "completions/max_length": 4096.0, "completions/max_terminated_length": 4082.0, "completions/mean_length": 2454.72265625, "completions/mean_terminated_length": 1708.6875, "completions/min_length": 262.0, "completions/min_terminated_length": 262.0, "epoch": 0.03409090909090909, "frac_reward_zero_std": 0.125, "groups/all_right": 0.5, "groups/all_wrong": 0.0, "groups/live_frac": 0.5, "groups/n": 2, "num_tokens": 4395431.0, "register/okay": 0.03125, "register/okay_any": 0.96875, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.630859375, "reward_std": 0.4830440282821655, "rewards/correct/mean": 0.630859375, "rewards/correct/std": 0.4830440282821655, "rollout/chars": 6361.1875, "rollout/commit": 0.875, "rollout/reward": 0.875, "rollout/trunc": 0.125, "sampling/importance_sampling_ratio/max": 2.924746513366699, "sampling/importance_sampling_ratio/mean": 0.9988688826560974, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 1.3344054222106934, "sampling/sampling_logp_difference/mean": 0.004491552244871855, "step": 3, "step_time": 239.0937386670048 }, { "epoch": 0.03409090909090909, "grad_norm": 0.003723646979779005, "learning_rate": 4.000000000000001e-06, "loss": 0.034827012568712234, "step": 3 }, { "clip_ratio": 1.9186707049811957e-06, "completions/clipped_ratio": 0.373046875, "completions/max_length": 4096.0, "completions/max_terminated_length": 4087.0, "completions/mean_length": 2846.294921875, "completions/mean_terminated_length": 2102.700927734375, "completions/min_length": 370.0, "completions/min_terminated_length": 370.0, "epoch": 0.045454545454545456, "frac_reward_zero_std": 0.09375, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 5927854.0, "register/okay": 0.0, "register/okay_any": 0.9375, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.5625, "reward_std": 0.49656352400779724, "rewards/correct/mean": 0.5625, "rewards/correct/std": 0.49656352400779724, "rollout/chars": 9865.4375, "rollout/commit": 0.75, "rollout/reward": 0.65625, "rollout/trunc": 0.3125, "sampling/importance_sampling_ratio/max": 2.9617812633514404, "sampling/importance_sampling_ratio/mean": 0.9208488464355469, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 1.0877385139465332, "sampling/sampling_logp_difference/mean": 0.004612394608557224, "step": 4, "step_time": 227.59205132999705 }, { "epoch": 0.045454545454545456, "grad_norm": 0.0031638338696211576, "learning_rate": 6e-06, "loss": 0.052659276872873306, "step": 4 }, { "clip_ratio": 1.481114054513455e-06, "completions/clipped_ratio": 0.298828125, "completions/max_length": 4096.0, "completions/max_terminated_length": 4078.0, "completions/mean_length": 2491.001953125, "completions/mean_terminated_length": 1806.9776611328125, "completions/min_length": 240.0, "completions/min_terminated_length": 240.0, "epoch": 0.056818181818181816, "frac_reward_zero_std": 0.125, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 7270063.0, "register/okay": 0.0625, "register/okay_any": 1.0, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.646484375, "reward_std": 0.47852855920791626, "rewards/correct/mean": 0.646484375, "rewards/correct/std": 0.47852855920791626, "rollout/chars": 9426.375, "rollout/commit": 0.59375, "rollout/reward": 0.59375, "rollout/trunc": 0.40625, "sampling/importance_sampling_ratio/max": 2.7546863555908203, "sampling/importance_sampling_ratio/mean": 0.9406723380088806, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 1.0478997230529785, "sampling/sampling_logp_difference/mean": 0.0044736433774232864, "step": 5, "step_time": 209.38289929599705 }, { "epoch": 0.056818181818181816, "grad_norm": 0.003155597485601902, "learning_rate": 8.000000000000001e-06, "loss": 0.036301881074905396, "step": 5 }, { "clip_ratio": 2.491965801709739e-06, "completions/clipped_ratio": 0.443359375, "completions/max_length": 4096.0, "completions/max_terminated_length": 4055.0, "completions/mean_length": 3193.939453125, "completions/mean_terminated_length": 2475.456298828125, "completions/min_length": 572.0, "completions/min_terminated_length": 572.0, "epoch": 0.06818181818181818, "frac_reward_zero_std": 0.15625, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 8985328.0, "register/okay": 0.0, "register/okay_any": 0.8125, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.509765625, "reward_std": 0.5003935098648071, "rewards/correct/mean": 0.509765625, "rewards/correct/std": 0.5003935098648071, "rollout/chars": 12321.75, "rollout/commit": 0.34375, "rollout/reward": 0.15625, "rollout/trunc": 0.8125, "sampling/importance_sampling_ratio/max": 2.806901693344116, "sampling/importance_sampling_ratio/mean": 0.9583785533905029, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 1.0830349922180176, "sampling/sampling_logp_difference/mean": 0.004532902501523495, "step": 6, "step_time": 232.05504834799285 }, { "epoch": 0.06818181818181818, "grad_norm": 0.0032948916777968407, "learning_rate": 1e-05, "loss": 0.022378744557499886, "step": 6 }, { "clip_ratio": 1.430511474609375e-06, "completions/clipped_ratio": 0.30859375, "completions/max_length": 4096.0, "completions/max_terminated_length": 4077.0, "completions/mean_length": 2478.0546875, "completions/mean_terminated_length": 1755.9208984375, "completions/min_length": 206.0, "completions/min_terminated_length": 206.0, "epoch": 0.07954545454545454, "frac_reward_zero_std": 0.15625, "groups/all_right": 0.5, "groups/all_wrong": 0.0, "groups/live_frac": 0.5, "groups/n": 2, "num_tokens": 10316972.0, "register/okay": 0.0, "register/okay_any": 1.0, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.6328125, "reward_std": 0.48250964283943176, "rewards/correct/mean": 0.6328125, "rewards/correct/std": 0.48250964283943176, "rollout/chars": 4690.90625, "rollout/commit": 0.96875, "rollout/reward": 0.96875, "rollout/trunc": 0.03125, "sampling/importance_sampling_ratio/max": 2.98211932182312, "sampling/importance_sampling_ratio/mean": 0.9605857133865356, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 1.7990617752075195, "sampling/sampling_logp_difference/mean": 0.00447505246847868, "step": 7, "step_time": 225.70014225300474 }, { "epoch": 0.07954545454545454, "grad_norm": 0.003252506721764803, "learning_rate": 1e-05, "loss": 0.03394366055727005, "step": 7 }, { "clip_ratio": 4.0595539871901565e-06, "completions/clipped_ratio": 0.287109375, "completions/max_length": 4096.0, "completions/max_terminated_length": 4086.0, "completions/mean_length": 2682.54296875, "completions/mean_terminated_length": 2113.287841796875, "completions/min_length": 361.0, "completions/min_terminated_length": 361.0, "epoch": 0.09090909090909091, "frac_reward_zero_std": 0.0625, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 11759986.0, "register/okay": 0.0, "register/okay_any": 0.96875, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.6484375, "reward_std": 0.4779251217842102, "rewards/correct/mean": 0.6484375, "rewards/correct/std": 0.4779251217842102, "rollout/chars": 7723.875, "rollout/commit": 0.84375, "rollout/reward": 0.75, "rollout/trunc": 0.21875, "sampling/importance_sampling_ratio/max": 2.72625732421875, "sampling/importance_sampling_ratio/mean": 0.9662811160087585, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.5347617864608765, "sampling/sampling_logp_difference/mean": 0.004520789720118046, "step": 8, "step_time": 228.13665069599665 }, { "epoch": 0.09090909090909091, "grad_norm": 0.0035843115765601397, "learning_rate": 1e-05, "loss": 0.06731684505939484, "step": 8 }, { "clip_ratio": 1.1685516028592247e-05, "completions/clipped_ratio": 0.302734375, "completions/max_length": 4096.0, "completions/max_terminated_length": 4077.0, "completions/mean_length": 2592.853515625, "completions/mean_terminated_length": 1940.2269287109375, "completions/min_length": 365.0, "completions/min_terminated_length": 365.0, "epoch": 0.10227272727272728, "frac_reward_zero_std": 0.09375, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 13154615.0, "register/okay": 0.0625, "register/okay_any": 0.8125, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.623046875, "reward_std": 0.4850969910621643, "rewards/correct/mean": 0.623046875, "rewards/correct/std": 0.4850969910621643, "rollout/chars": 6270.625, "rollout/commit": 0.96875, "rollout/reward": 0.84375, "rollout/trunc": 0.09375, "sampling/importance_sampling_ratio/max": 2.6416101455688477, "sampling/importance_sampling_ratio/mean": 0.9010199308395386, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.5122027397155762, "sampling/sampling_logp_difference/mean": 0.00482128094881773, "step": 9, "step_time": 230.5110018680025 }, { "epoch": 0.10227272727272728, "grad_norm": 0.003008148865774274, "learning_rate": 1e-05, "loss": 0.020274115726351738, "step": 9 }, { "clip_ratio": 1.5000192377101484e-05, "completions/clipped_ratio": 0.294921875, "completions/max_length": 4096.0, "completions/max_terminated_length": 4018.0, "completions/mean_length": 2688.658203125, "completions/mean_terminated_length": 2100.030517578125, "completions/min_length": 428.0, "completions/min_terminated_length": 428.0, "epoch": 0.11363636363636363, "frac_reward_zero_std": 0.0625, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 14602376.0, "register/okay": 0.0, "register/okay_any": 1.0, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.642578125, "reward_std": 0.4797092080116272, "rewards/correct/mean": 0.642578125, "rewards/correct/std": 0.4797092080116272, "rollout/chars": 5943.53125, "rollout/commit": 0.8125, "rollout/reward": 0.71875, "rollout/trunc": 0.1875, "sampling/importance_sampling_ratio/max": 2.8662335872650146, "sampling/importance_sampling_ratio/mean": 0.9329501986503601, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.4294443130493164, "sampling/sampling_logp_difference/mean": 0.005068408325314522, "step": 10, "step_time": 221.11443298699305 }, { "epoch": 0.11363636363636363, "grad_norm": 0.004120167810469866, "learning_rate": 1e-05, "loss": 0.08815088868141174, "step": 10 }, { "clip_ratio": 1.3279116956255166e-05, "completions/clipped_ratio": 0.32421875, "completions/max_length": 4096.0, "completions/max_terminated_length": 4055.0, "completions/mean_length": 2560.62890625, "completions/mean_terminated_length": 1824.0057373046875, "completions/min_length": 248.0, "completions/min_terminated_length": 248.0, "epoch": 0.125, "frac_reward_zero_std": 0.125, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 15977178.0, "register/okay": 0.0, "register/okay_any": 0.90625, "register/premium": -0.16129032258064513, "register/verify_any": 0.96875, "reward": 0.625, "reward_std": 0.4845963716506958, "rewards/correct/mean": 0.625, "rewards/correct/std": 0.4845963716506958, "rollout/chars": 6113.28125, "rollout/commit": 0.84375, "rollout/reward": 0.84375, "rollout/trunc": 0.125, "sampling/importance_sampling_ratio/max": 2.96163010597229, "sampling/importance_sampling_ratio/mean": 0.9653053283691406, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 1.036353349685669, "sampling/sampling_logp_difference/mean": 0.004582596477121115, "step": 11, "step_time": 238.2820661850019 }, { "epoch": 0.125, "grad_norm": 0.003486036090180278, "learning_rate": 1e-05, "loss": 0.06840801239013672, "step": 11 }, { "clip_ratio": 2.348333373447531e-05, "completions/clipped_ratio": 0.3515625, "completions/max_length": 4096.0, "completions/max_terminated_length": 4083.0, "completions/mean_length": 2810.513671875, "completions/mean_terminated_length": 2113.563232421875, "completions/min_length": 239.0, "completions/min_terminated_length": 239.0, "epoch": 0.13636363636363635, "frac_reward_zero_std": 0.1875, "groups/all_right": 0.5, "groups/all_wrong": 0.0, "groups/live_frac": 0.5, "groups/n": 2, "num_tokens": 17479505.0, "register/okay": 0.03125, "register/okay_any": 0.96875, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.623046875, "reward_std": 0.4850969910621643, "rewards/correct/mean": 0.623046875, "rewards/correct/std": 0.4850969910621643, "rollout/chars": 7909.25, "rollout/commit": 0.65625, "rollout/reward": 0.65625, "rollout/trunc": 0.34375, "sampling/importance_sampling_ratio/max": 2.941793918609619, "sampling/importance_sampling_ratio/mean": 0.9080376625061035, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 1.1040374040603638, "sampling/sampling_logp_difference/mean": 0.004863664973527193, "step": 12, "step_time": 223.75293871599933 }, { "epoch": 0.13636363636363635, "grad_norm": 0.003054369240999222, "learning_rate": 1e-05, "loss": 0.04905109480023384, "step": 12 }, { "clip_ratio": 3.613838339333597e-05, "completions/clipped_ratio": 0.251953125, "completions/max_length": 4096.0, "completions/max_terminated_length": 4096.0, "completions/mean_length": 2390.146484375, "completions/mean_terminated_length": 1815.590087890625, "completions/min_length": 278.0, "completions/min_terminated_length": 278.0, "epoch": 0.14772727272727273, "frac_reward_zero_std": 0.15625, "groups/all_right": 0.0, "groups/all_wrong": 0.5, "groups/live_frac": 0.5, "groups/n": 2, "num_tokens": 18765740.0, "register/okay": 0.0, "register/okay_any": 0.96875, "register/premium": -0.5483870967741935, "register/verify_any": 0.96875, "reward": 0.662109375, "reward_std": 0.4734536409378052, "rewards/correct/mean": 0.662109375, "rewards/correct/std": 0.4734536409378052, "rollout/chars": 6902.8125, "rollout/commit": 0.625, "rollout/reward": 0.46875, "rollout/trunc": 0.40625, "sampling/importance_sampling_ratio/max": 2.6642820835113525, "sampling/importance_sampling_ratio/mean": 0.9322664141654968, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.39106273651123047, "sampling/sampling_logp_difference/mean": 0.0050008250400424, "step": 13, "step_time": 204.52620178501456 }, { "epoch": 0.14772727272727273, "grad_norm": 0.0034245674032717943, "learning_rate": 1e-05, "loss": 0.03518375754356384, "step": 13 }, { "clip_ratio": 3.3756232539872144e-05, "completions/clipped_ratio": 0.384765625, "completions/max_length": 4096.0, "completions/max_terminated_length": 4080.0, "completions/mean_length": 2794.69921875, "completions/mean_terminated_length": 1980.869873046875, "completions/min_length": 274.0, "completions/min_terminated_length": 274.0, "epoch": 0.1590909090909091, "frac_reward_zero_std": 0.09375, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 20267250.0, "register/okay": 0.0, "register/okay_any": 0.84375, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.533203125, "reward_std": 0.4993842542171478, "rewards/correct/mean": 0.533203125, "rewards/correct/std": 0.4993842542171478, "rollout/chars": 6717.75, "rollout/commit": 0.5625, "rollout/reward": 0.53125, "rollout/trunc": 0.40625, "sampling/importance_sampling_ratio/max": 2.9956328868865967, "sampling/importance_sampling_ratio/mean": 0.9270258545875549, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.949026346206665, "sampling/sampling_logp_difference/mean": 0.0053371889516711235, "step": 14, "step_time": 241.89416026500476 }, { "epoch": 0.1590909090909091, "grad_norm": 0.0032463697716593742, "learning_rate": 1e-05, "loss": 0.04247649013996124, "step": 14 }, { "clip_ratio": 4.435342742681314e-05, "completions/clipped_ratio": 0.197265625, "completions/max_length": 4096.0, "completions/max_terminated_length": 4089.0, "completions/mean_length": 2334.177734375, "completions/mean_terminated_length": 1901.223876953125, "completions/min_length": 328.0, "completions/min_terminated_length": 328.0, "epoch": 0.17045454545454544, "frac_reward_zero_std": 0.1875, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 21532909.0, "register/okay": 0.0, "register/okay_any": 1.0, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.76171875, "reward_std": 0.42644867300987244, "rewards/correct/mean": 0.76171875, "rewards/correct/std": 0.42644867300987244, "rollout/chars": 8942.15625, "rollout/commit": 0.8125, "rollout/reward": 0.6875, "rollout/trunc": 0.3125, "sampling/importance_sampling_ratio/max": 2.972348928451538, "sampling/importance_sampling_ratio/mean": 0.9189268350601196, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.5567886829376221, "sampling/sampling_logp_difference/mean": 0.004913019947707653, "step": 15, "step_time": 223.23767583300287 }, { "epoch": 0.17045454545454544, "grad_norm": 0.0037231335882097483, "learning_rate": 1e-05, "loss": 0.05135633423924446, "step": 15 }, { "clip_ratio": 4.653011211530611e-05, "completions/clipped_ratio": 0.279296875, "completions/max_length": 4096.0, "completions/max_terminated_length": 4088.0, "completions/mean_length": 2597.017578125, "completions/mean_terminated_length": 2016.1112060546875, "completions/min_length": 344.0, "completions/min_terminated_length": 344.0, "epoch": 0.18181818181818182, "frac_reward_zero_std": 0.125, "groups/all_right": 0.5, "groups/all_wrong": 0.0, "groups/live_frac": 0.5, "groups/n": 2, "num_tokens": 22936550.0, "register/okay": 0.0, "register/okay_any": 1.0, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.65234375, "reward_std": 0.47669193148612976, "rewards/correct/mean": 0.65234375, "rewards/correct/std": 0.47669193148612976, "rollout/chars": 5921.5, "rollout/commit": 0.96875, "rollout/reward": 0.96875, "rollout/trunc": 0.03125, "sampling/importance_sampling_ratio/max": 2.970635175704956, "sampling/importance_sampling_ratio/mean": 0.9122588038444519, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.617943286895752, "sampling/sampling_logp_difference/mean": 0.004951866343617439, "step": 16, "step_time": 213.3137985569956 }, { "epoch": 0.18181818181818182, "grad_norm": 0.002896440913900733, "learning_rate": 1e-05, "loss": 0.046279728412628174, "step": 16 }, { "clip_ratio": 4.9786167437559925e-05, "completions/clipped_ratio": 0.130859375, "completions/max_length": 4096.0, "completions/max_terminated_length": 4090.0, "completions/mean_length": 1984.505859375, "completions/mean_terminated_length": 1666.5955810546875, "completions/min_length": 202.0, "completions/min_terminated_length": 202.0, "epoch": 0.19318181818181818, "frac_reward_zero_std": 0.0625, "groups/all_right": 0.5, "groups/all_wrong": 0.0, "groups/live_frac": 0.5, "groups/n": 2, "num_tokens": 24023145.0, "register/okay": 0.0, "register/okay_any": 0.84375, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.765625, "reward_std": 0.42402184009552, "rewards/correct/mean": 0.765625, "rewards/correct/std": 0.42402184009552, "rollout/chars": 6103.28125, "rollout/commit": 0.9375, "rollout/reward": 0.84375, "rollout/trunc": 0.125, "sampling/importance_sampling_ratio/max": 2.654937267303467, "sampling/importance_sampling_ratio/mean": 0.9073137640953064, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.5429407358169556, "sampling/sampling_logp_difference/mean": 0.0048791998997330666, "step": 17, "step_time": 198.31722998900295 }, { "epoch": 0.19318181818181818, "grad_norm": 0.004091218579560518, "learning_rate": 1e-05, "loss": 0.05541734769940376, "step": 17 }, { "clip_ratio": 5.2998024330008775e-05, "completions/clipped_ratio": 0.36328125, "completions/max_length": 4096.0, "completions/max_terminated_length": 4088.0, "completions/mean_length": 2802.869140625, "completions/mean_terminated_length": 2065.070556640625, "completions/min_length": 337.0, "completions/min_terminated_length": 337.0, "epoch": 0.20454545454545456, "frac_reward_zero_std": 0.1875, "groups/all_right": 0.5, "groups/all_wrong": 0.0, "groups/live_frac": 0.5, "groups/n": 2, "num_tokens": 25519670.0, "register/okay": 0.0, "register/okay_any": 0.90625, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.578125, "reward_std": 0.49434176087379456, "rewards/correct/mean": 0.578125, "rewards/correct/std": 0.49434176087379456, "rollout/chars": 3893.9375, "rollout/commit": 0.90625, "rollout/reward": 0.90625, "rollout/trunc": 0.03125, "sampling/importance_sampling_ratio/max": 2.7872073650360107, "sampling/importance_sampling_ratio/mean": 0.9027870893478394, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.47285139560699463, "sampling/sampling_logp_difference/mean": 0.004941096995025873, "step": 18, "step_time": 244.71225519200016 }, { "epoch": 0.20454545454545456, "grad_norm": 0.0028813090175390244, "learning_rate": 1e-05, "loss": 0.029713578522205353, "step": 18 }, { "clip_ratio": 4.2906002590825665e-05, "completions/clipped_ratio": 0.291015625, "completions/max_length": 4096.0, "completions/max_terminated_length": 4077.0, "completions/mean_length": 2633.5078125, "completions/mean_terminated_length": 2033.201171875, "completions/min_length": 286.0, "completions/min_terminated_length": 286.0, "epoch": 0.2159090909090909, "frac_reward_zero_std": 0.125, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 26949290.0, "register/okay": 0.0, "register/okay_any": 0.96875, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.66796875, "reward_std": 0.47140273451805115, "rewards/correct/mean": 0.66796875, "rewards/correct/std": 0.47140273451805115, "rollout/chars": 9273.59375, "rollout/commit": 0.5, "rollout/reward": 0.46875, "rollout/trunc": 0.5, "sampling/importance_sampling_ratio/max": 2.972933530807495, "sampling/importance_sampling_ratio/mean": 0.8908332586288452, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 1.4295125007629395, "sampling/sampling_logp_difference/mean": 0.004852948244661093, "step": 19, "step_time": 210.78038853899852 }, { "epoch": 0.2159090909090909, "grad_norm": 0.003327382495626807, "learning_rate": 1e-05, "loss": 0.06632369756698608, "step": 19 }, { "clip_ratio": 5.486932650455856e-05, "completions/clipped_ratio": 0.427734375, "completions/max_length": 4096.0, "completions/max_terminated_length": 4071.0, "completions/mean_length": 3086.205078125, "completions/mean_terminated_length": 2331.443603515625, "completions/min_length": 704.0, "completions/min_terminated_length": 704.0, "epoch": 0.22727272727272727, "frac_reward_zero_std": 0.0625, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 28610611.0, "register/okay": 0.0, "register/okay_any": 1.0, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.5234375, "reward_std": 0.49993884563446045, "rewards/correct/mean": 0.5234375, "rewards/correct/std": 0.49993884563446045, "rollout/chars": 9274.625, "rollout/commit": 0.53125, "rollout/reward": 0.375, "rollout/trunc": 0.5, "sampling/importance_sampling_ratio/max": 2.9439356327056885, "sampling/importance_sampling_ratio/mean": 0.9422547221183777, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.7304327487945557, "sampling/sampling_logp_difference/mean": 0.004762661177664995, "step": 20, "step_time": 241.5823824050094 }, { "epoch": 0.22727272727272727, "grad_norm": 0.003230234608054161, "learning_rate": 1e-05, "loss": 0.0657850056886673, "step": 20 }, { "clip_ratio": 5.2212915420568606e-05, "completions/clipped_ratio": 0.328125, "completions/max_length": 4096.0, "completions/max_terminated_length": 4057.0, "completions/mean_length": 2670.73046875, "completions/mean_terminated_length": 1974.6685791015625, "completions/min_length": 406.0, "completions/min_terminated_length": 406.0, "epoch": 0.23863636363636365, "frac_reward_zero_std": 0.125, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 30047289.0, "register/okay": 0.03125, "register/okay_any": 1.0, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.6171875, "reward_std": 0.486548513174057, "rewards/correct/mean": 0.6171875, "rewards/correct/std": 0.486548513174057, "rollout/chars": 8537.1875, "rollout/commit": 0.625, "rollout/reward": 0.53125, "rollout/trunc": 0.4375, "sampling/importance_sampling_ratio/max": 2.9177842140197754, "sampling/importance_sampling_ratio/mean": 0.9250489473342896, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.4100780487060547, "sampling/sampling_logp_difference/mean": 0.005068977363407612, "step": 21, "step_time": 230.3553039509934 }, { "epoch": 0.23863636363636365, "grad_norm": 0.003675194690003991, "learning_rate": 1e-05, "loss": 0.059521663933992386, "step": 21 }, { "clip_ratio": 6.675696363345196e-05, "completions/clipped_ratio": 0.255859375, "completions/max_length": 4096.0, "completions/max_terminated_length": 4076.0, "completions/mean_length": 2579.5390625, "completions/mean_terminated_length": 2058.13134765625, "completions/min_length": 272.0, "completions/min_terminated_length": 272.0, "epoch": 0.25, "frac_reward_zero_std": 0.1875, "groups/all_right": 0.5, "groups/all_wrong": 0.0, "groups/live_frac": 0.5, "groups/n": 2, "num_tokens": 31434285.0, "register/okay": 0.0, "register/okay_any": 1.0, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.6796875, "reward_std": 0.4670529365539551, "rewards/correct/mean": 0.6796875, "rewards/correct/std": 0.4670529365539551, "rollout/chars": 6334.8125, "rollout/commit": 0.65625, "rollout/reward": 0.59375, "rollout/trunc": 0.40625, "sampling/importance_sampling_ratio/max": 2.882204294204712, "sampling/importance_sampling_ratio/mean": 0.936687707901001, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 1.2923583984375, "sampling/sampling_logp_difference/mean": 0.004951627925038338, "step": 22, "step_time": 196.23515353100447 }, { "epoch": 0.25, "grad_norm": 0.003399793989956379, "learning_rate": 1e-05, "loss": 0.036319706588983536, "step": 22 }, { "clip_ratio": 5.6475377618880884e-05, "completions/clipped_ratio": 0.26171875, "completions/max_length": 4096.0, "completions/max_terminated_length": 4096.0, "completions/mean_length": 2351.69140625, "completions/mean_terminated_length": 1733.338623046875, "completions/min_length": 290.0, "completions/min_terminated_length": 290.0, "epoch": 0.26136363636363635, "frac_reward_zero_std": 0.15625, "groups/all_right": 0.0, "groups/all_wrong": 0.5, "groups/live_frac": 0.5, "groups/n": 2, "num_tokens": 32720095.0, "register/okay": 0.0, "register/okay_any": 1.0, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.6484375, "reward_std": 0.4779251217842102, "rewards/correct/mean": 0.6484375, "rewards/correct/std": 0.4779251217842102, "rollout/chars": 11748.21875, "rollout/commit": 0.15625, "rollout/reward": 0.0625, "rollout/trunc": 0.875, "sampling/importance_sampling_ratio/max": 2.905789852142334, "sampling/importance_sampling_ratio/mean": 0.9464510083198547, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 1.0502471923828125, "sampling/sampling_logp_difference/mean": 0.005386264994740486, "step": 23, "step_time": 234.36768228199435 }, { "epoch": 0.26136363636363635, "grad_norm": 0.004066686611622572, "learning_rate": 1e-05, "loss": 0.07233693450689316, "step": 23 }, { "clip_ratio": 5.326732753019314e-05, "completions/clipped_ratio": 0.421875, "completions/max_length": 4096.0, "completions/max_terminated_length": 4090.0, "completions/mean_length": 3116.578125, "completions/mean_terminated_length": 2401.9560546875, "completions/min_length": 509.0, "completions/min_terminated_length": 509.0, "epoch": 0.2727272727272727, "frac_reward_zero_std": 0.0625, "groups/all_right": 0.5, "groups/all_wrong": 0.0, "groups/live_frac": 0.5, "groups/n": 2, "num_tokens": 34398855.0, "register/okay": 0.0, "register/okay_any": 0.90625, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.52734375, "reward_std": 0.49974003434181213, "rewards/correct/mean": 0.52734375, "rewards/correct/std": 0.49974003434181213, "rollout/chars": 7776.96875, "rollout/commit": 0.71875, "rollout/reward": 0.71875, "rollout/trunc": 0.28125, "sampling/importance_sampling_ratio/max": 2.990600109100342, "sampling/importance_sampling_ratio/mean": 0.8955240249633789, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.8885411024093628, "sampling/sampling_logp_difference/mean": 0.004893308971077204, "step": 24, "step_time": 248.16172832200755 }, { "epoch": 0.2727272727272727, "grad_norm": 0.003279327414929867, "learning_rate": 1e-05, "loss": 0.03035794571042061, "step": 24 }, { "clip_ratio": 4.4846863659131486e-05, "completions/clipped_ratio": 0.32421875, "completions/max_length": 4096.0, "completions/max_terminated_length": 4039.0, "completions/mean_length": 2545.95703125, "completions/mean_terminated_length": 1802.2947998046875, "completions/min_length": 200.0, "completions/min_terminated_length": 200.0, "epoch": 0.2840909090909091, "frac_reward_zero_std": 0.125, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 35778657.0, "register/okay": 0.0, "register/okay_any": 0.96875, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.619140625, "reward_std": 0.48607301712036133, "rewards/correct/mean": 0.619140625, "rewards/correct/std": 0.48607301712036133, "rollout/chars": 7533.4375, "rollout/commit": 0.59375, "rollout/reward": 0.46875, "rollout/trunc": 0.28125, "sampling/importance_sampling_ratio/max": 2.9910829067230225, "sampling/importance_sampling_ratio/mean": 0.9145698547363281, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.5044717788696289, "sampling/sampling_logp_difference/mean": 0.005069740582257509, "step": 25, "step_time": 230.56244620599318 }, { "epoch": 0.2840909090909091, "grad_norm": 0.0035241309087723494, "learning_rate": 1e-05, "loss": 0.04239988327026367, "step": 25 }, { "clip_ratio": 6.0805309203715296e-05, "completions/clipped_ratio": 0.310546875, "completions/max_length": 4096.0, "completions/max_terminated_length": 4089.0, "completions/mean_length": 2703.7734375, "completions/mean_terminated_length": 2076.679931640625, "completions/min_length": 289.0, "completions/min_terminated_length": 289.0, "epoch": 0.29545454545454547, "frac_reward_zero_std": 0.1875, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 37239005.0, "register/okay": 0.0, "register/okay_any": 0.84375, "register/premium": -0.06451612903225812, "register/verify_any": 0.96875, "reward": 0.658203125, "reward_std": 0.4747757613658905, "rewards/correct/mean": 0.658203125, "rewards/correct/std": 0.4747757613658905, "rollout/chars": 3381.46875, "rollout/commit": 1.0, "rollout/reward": 0.9375, "rollout/trunc": 0.0, "sampling/importance_sampling_ratio/max": 2.837442636489868, "sampling/importance_sampling_ratio/mean": 0.8787897229194641, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.8435883522033691, "sampling/sampling_logp_difference/mean": 0.0050262752920389175, "step": 26, "step_time": 233.77161700500073 }, { "epoch": 0.29545454545454547, "grad_norm": 0.0030401181429624557, "learning_rate": 1e-05, "loss": 0.019256848841905594, "step": 26 }, { "clip_ratio": 4.174492096353788e-05, "completions/clipped_ratio": 0.27734375, "completions/max_length": 4096.0, "completions/max_terminated_length": 4070.0, "completions/mean_length": 2551.8125, "completions/mean_terminated_length": 1959.178466796875, "completions/min_length": 338.0, "completions/min_terminated_length": 338.0, "epoch": 0.3068181818181818, "frac_reward_zero_std": 0.125, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 38608157.0, "register/okay": 0.03125, "register/okay_any": 1.0, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.650390625, "reward_std": 0.47731292247772217, "rewards/correct/mean": 0.650390625, "rewards/correct/std": 0.47731292247772217, "rollout/chars": 10290.8125, "rollout/commit": 0.21875, "rollout/reward": 0.15625, "rollout/trunc": 0.84375, "sampling/importance_sampling_ratio/max": 2.992258071899414, "sampling/importance_sampling_ratio/mean": 0.9573348164558411, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.7723093032836914, "sampling/sampling_logp_difference/mean": 0.004969725385308266, "step": 27, "step_time": 224.3992258420003 }, { "epoch": 0.3068181818181818, "grad_norm": 0.0036496196407824755, "learning_rate": 1e-05, "loss": 0.04576047509908676, "step": 27 }, { "clip_ratio": 4.57169376204547e-05, "completions/clipped_ratio": 0.279296875, "completions/max_length": 4096.0, "completions/max_terminated_length": 4050.0, "completions/mean_length": 2679.435546875, "completions/mean_terminated_length": 2130.468994140625, "completions/min_length": 420.0, "completions/min_terminated_length": 420.0, "epoch": 0.3181818181818182, "frac_reward_zero_std": 0.0625, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 40047020.0, "register/okay": 0.0, "register/okay_any": 0.84375, "register/premium": 0.8387096774193549, "register/verify_any": 0.96875, "reward": 0.646484375, "reward_std": 0.47852855920791626, "rewards/correct/mean": 0.646484375, "rewards/correct/std": 0.47852855920791626, "rollout/chars": 6973.1875, "rollout/commit": 0.9375, "rollout/reward": 0.8125, "rollout/trunc": 0.0625, "sampling/importance_sampling_ratio/max": 2.9695754051208496, "sampling/importance_sampling_ratio/mean": 0.8856563568115234, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 2.3932266235351562, "sampling/sampling_logp_difference/mean": 0.004716905765235424, "step": 28, "step_time": 233.92598559500766 }, { "epoch": 0.3181818181818182, "grad_norm": 0.003118283813819289, "learning_rate": 1e-05, "loss": 0.03405182063579559, "step": 28 }, { "clip_ratio": 6.522296916955383e-05, "completions/clipped_ratio": 0.28125, "completions/max_length": 4096.0, "completions/max_terminated_length": 4092.0, "completions/mean_length": 2800.392578125, "completions/mean_terminated_length": 2294.535400390625, "completions/min_length": 163.0, "completions/min_terminated_length": 163.0, "epoch": 0.32954545454545453, "frac_reward_zero_std": 0.0625, "groups/all_right": 0.5, "groups/all_wrong": 0.0, "groups/live_frac": 0.5, "groups/n": 2, "num_tokens": 41565365.0, "register/okay": 0.0, "register/okay_any": 0.90625, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.673828125, "reward_std": 0.4692695140838623, "rewards/correct/mean": 0.673828125, "rewards/correct/std": 0.4692695140838623, "rollout/chars": 6729.40625, "rollout/commit": 0.90625, "rollout/reward": 0.875, "rollout/trunc": 0.09375, "sampling/importance_sampling_ratio/max": 2.92781662940979, "sampling/importance_sampling_ratio/mean": 0.8985303640365601, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.8867032527923584, "sampling/sampling_logp_difference/mean": 0.00486964825540781, "step": 29, "step_time": 238.03783438399478 }, { "epoch": 0.32954545454545453, "grad_norm": 0.0032935026101768017, "learning_rate": 1e-05, "loss": 0.052372947335243225, "step": 29 }, { "clip_ratio": 8.097707109300245e-05, "completions/clipped_ratio": 0.25390625, "completions/max_length": 4096.0, "completions/max_terminated_length": 4058.0, "completions/mean_length": 2527.5546875, "completions/mean_terminated_length": 1993.7906494140625, "completions/min_length": 289.0, "completions/min_terminated_length": 289.0, "epoch": 0.3409090909090909, "frac_reward_zero_std": 0.03125, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 42921617.0, "register/okay": 0.0, "register/okay_any": 1.0, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.619140625, "reward_std": 0.48607301712036133, "rewards/correct/mean": 0.619140625, "rewards/correct/std": 0.48607301712036133, "rollout/chars": 6342.75, "rollout/commit": 0.8125, "rollout/reward": 0.5625, "rollout/trunc": 0.28125, "sampling/importance_sampling_ratio/max": 2.988611936569214, "sampling/importance_sampling_ratio/mean": 0.8912668228149414, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.7639431953430176, "sampling/sampling_logp_difference/mean": 0.0050128307193517685, "step": 30, "step_time": 224.68840066700432 }, { "epoch": 0.3409090909090909, "grad_norm": 0.003287514206022024, "learning_rate": 1e-05, "loss": 0.04755740240216255, "step": 30 }, { "clip_ratio": 7.465987437171862e-05, "completions/clipped_ratio": 0.3046875, "completions/max_length": 4096.0, "completions/max_terminated_length": 4068.0, "completions/mean_length": 2656.521484375, "completions/mean_terminated_length": 2025.73876953125, "completions/min_length": 239.0, "completions/min_terminated_length": 239.0, "epoch": 0.3522727272727273, "frac_reward_zero_std": 0.0625, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 44355548.0, "register/okay": 0.0, "register/okay_any": 0.9375, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.55859375, "reward_std": 0.4970405399799347, "rewards/correct/mean": 0.55859375, "rewards/correct/std": 0.4970405399799347, "rollout/chars": 4952.1875, "rollout/commit": 0.90625, "rollout/reward": 0.5, "rollout/trunc": 0.03125, "sampling/importance_sampling_ratio/max": 2.961225748062134, "sampling/importance_sampling_ratio/mean": 0.8841115236282349, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.722623348236084, "sampling/sampling_logp_difference/mean": 0.005319459363818169, "step": 31, "step_time": 213.5908911309998 }, { "epoch": 0.3522727272727273, "grad_norm": 0.0033462513238191605, "learning_rate": 1e-05, "loss": 0.04553345590829849, "step": 31 }, { "clip_ratio": 6.642126277256466e-05, "completions/clipped_ratio": 0.333984375, "completions/max_length": 4096.0, "completions/max_terminated_length": 4091.0, "completions/mean_length": 2756.259765625, "completions/mean_terminated_length": 2084.42529296875, "completions/min_length": 345.0, "completions/min_terminated_length": 345.0, "epoch": 0.36363636363636365, "frac_reward_zero_std": 0.0625, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 45837921.0, "register/okay": 0.0, "register/okay_any": 0.875, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.572265625, "reward_std": 0.4952339828014374, "rewards/correct/mean": 0.572265625, "rewards/correct/std": 0.4952339828014374, "rollout/chars": 7447.9375, "rollout/commit": 0.875, "rollout/reward": 0.71875, "rollout/trunc": 0.21875, "sampling/importance_sampling_ratio/max": 2.9116132259368896, "sampling/importance_sampling_ratio/mean": 0.9092390537261963, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 1.0500521659851074, "sampling/sampling_logp_difference/mean": 0.004691188223659992, "step": 32, "step_time": 235.9118720449951 }, { "epoch": 0.36363636363636365, "grad_norm": 0.003006890183314681, "learning_rate": 1e-05, "loss": 0.03242364525794983, "step": 32 }, { "clip_ratio": 5.475359353113163e-05, "completions/clipped_ratio": 0.29296875, "completions/max_length": 4096.0, "completions/max_terminated_length": 4088.0, "completions/mean_length": 2574.208984375, "completions/mean_terminated_length": 1943.6326904296875, "completions/min_length": 135.0, "completions/min_terminated_length": 135.0, "epoch": 0.375, "frac_reward_zero_std": 0.0625, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 47223516.0, "register/okay": 0.0, "register/okay_any": 0.9375, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.640625, "reward_std": 0.48028653860092163, "rewards/correct/mean": 0.640625, "rewards/correct/std": 0.48028653860092163, "rollout/chars": 6007.3125, "rollout/commit": 0.625, "rollout/reward": 0.53125, "rollout/trunc": 0.4375, "sampling/importance_sampling_ratio/max": 2.9517362117767334, "sampling/importance_sampling_ratio/mean": 0.9390583634376526, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.7950553894042969, "sampling/sampling_logp_difference/mean": 0.004755455069243908, "step": 33, "step_time": 206.17870632700215 }, { "epoch": 0.375, "grad_norm": 0.003763378830626607, "learning_rate": 1e-05, "loss": 0.06001519411802292, "step": 33 }, { "clip_ratio": 4.351094924004428e-05, "completions/clipped_ratio": 0.298828125, "completions/max_length": 4096.0, "completions/max_terminated_length": 4083.0, "completions/mean_length": 2655.25390625, "completions/mean_terminated_length": 2041.231201171875, "completions/min_length": 299.0, "completions/min_terminated_length": 299.0, "epoch": 0.38636363636363635, "frac_reward_zero_std": 0.125, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 48646558.0, "register/okay": 0.0, "register/okay_any": 1.0, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.6328125, "reward_std": 0.48250964283943176, "rewards/correct/mean": 0.6328125, "rewards/correct/std": 0.48250964283943176, "rollout/chars": 9312.15625, "rollout/commit": 0.75, "rollout/reward": 0.6875, "rollout/trunc": 0.3125, "sampling/importance_sampling_ratio/max": 2.745772123336792, "sampling/importance_sampling_ratio/mean": 0.8967394828796387, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.7830262780189514, "sampling/sampling_logp_difference/mean": 0.004709095694124699, "step": 34, "step_time": 209.81075684600728 }, { "epoch": 0.38636363636363635, "grad_norm": 0.0032068060245364904, "learning_rate": 1e-05, "loss": 0.05563589930534363, "step": 34 }, { "clip_ratio": 7.661448302087592e-05, "completions/clipped_ratio": 0.1953125, "completions/max_length": 4096.0, "completions/max_terminated_length": 4045.0, "completions/mean_length": 2417.951171875, "completions/mean_terminated_length": 2010.6578369140625, "completions/min_length": 242.0, "completions/min_terminated_length": 242.0, "epoch": 0.3977272727272727, "frac_reward_zero_std": 0.09375, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 49943429.0, "register/okay": 0.03125, "register/okay_any": 1.0, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.7109375, "reward_std": 0.45377036929130554, "rewards/correct/mean": 0.7109375, "rewards/correct/std": 0.45377036929130554, "rollout/chars": 5155.21875, "rollout/commit": 0.96875, "rollout/reward": 0.84375, "rollout/trunc": 0.03125, "sampling/importance_sampling_ratio/max": 2.74753737449646, "sampling/importance_sampling_ratio/mean": 0.9276185035705566, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 1.055890440940857, "sampling/sampling_logp_difference/mean": 0.004778267350047827, "step": 35, "step_time": 206.7543564430016 }, { "epoch": 0.3977272727272727, "grad_norm": 0.003544215578585863, "learning_rate": 1e-05, "loss": 0.04388181120157242, "step": 35 }, { "clip_ratio": 5.353016700837543e-05, "completions/clipped_ratio": 0.2890625, "completions/max_length": 4096.0, "completions/max_terminated_length": 4087.0, "completions/mean_length": 2484.22265625, "completions/mean_terminated_length": 1828.8846435546875, "completions/min_length": 268.0, "completions/min_terminated_length": 268.0, "epoch": 0.4090909090909091, "frac_reward_zero_std": 0.125, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 51278839.0, "register/okay": 0.0, "register/okay_any": 0.96875, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.59375, "reward_std": 0.49161264300346375, "rewards/correct/mean": 0.59375, "rewards/correct/std": 0.49161264300346375, "rollout/chars": 6862.0625, "rollout/commit": 0.59375, "rollout/reward": 0.53125, "rollout/trunc": 0.375, "sampling/importance_sampling_ratio/max": 2.991304874420166, "sampling/importance_sampling_ratio/mean": 0.9369009733200073, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.5460691452026367, "sampling/sampling_logp_difference/mean": 0.005069076083600521, "step": 36, "step_time": 206.1250407339976 }, { "epoch": 0.4090909090909091, "grad_norm": 0.003753333119675517, "learning_rate": 1e-05, "loss": 0.03270333260297775, "step": 36 }, { "clip_ratio": 4.700995953044185e-05, "completions/clipped_ratio": 0.294921875, "completions/max_length": 4096.0, "completions/max_terminated_length": 4056.0, "completions/mean_length": 2447.91796875, "completions/mean_terminated_length": 1758.553955078125, "completions/min_length": 242.0, "completions/min_terminated_length": 242.0, "epoch": 0.42045454545454547, "frac_reward_zero_std": 0.0625, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 52603213.0, "register/okay": 0.0, "register/okay_any": 1.0, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.611328125, "reward_std": 0.4879252314567566, "rewards/correct/mean": 0.611328125, "rewards/correct/std": 0.4879252314567566, "rollout/chars": 7591.4375, "rollout/commit": 0.65625, "rollout/reward": 0.5625, "rollout/trunc": 0.4375, "sampling/importance_sampling_ratio/max": 2.9310998916625977, "sampling/importance_sampling_ratio/mean": 0.9377585649490356, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.7711648941040039, "sampling/sampling_logp_difference/mean": 0.005007673520594835, "step": 37, "step_time": 229.42526798800827 }, { "epoch": 0.42045454545454547, "grad_norm": 0.0034484269563108683, "learning_rate": 1e-05, "loss": 0.06802354753017426, "step": 37 }, { "clip_ratio": 3.721059533745574e-05, "completions/clipped_ratio": 0.193359375, "completions/max_length": 4096.0, "completions/max_terminated_length": 4057.0, "completions/mean_length": 2208.328125, "completions/mean_terminated_length": 1755.83544921875, "completions/min_length": 284.0, "completions/min_terminated_length": 284.0, "epoch": 0.4318181818181818, "frac_reward_zero_std": 0.15625, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 53789989.0, "register/okay": 0.0, "register/okay_any": 0.90625, "register/premium": -0.19354838709677424, "register/verify_any": 0.96875, "reward": 0.75390625, "reward_std": 0.4311550557613373, "rewards/correct/mean": 0.75390625, "rewards/correct/std": 0.4311550557613373, "rollout/chars": 4564.59375, "rollout/commit": 0.84375, "rollout/reward": 0.8125, "rollout/trunc": 0.0625, "sampling/importance_sampling_ratio/max": 2.7628633975982666, "sampling/importance_sampling_ratio/mean": 0.9271668195724487, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.45870041847229004, "sampling/sampling_logp_difference/mean": 0.004600748419761658, "step": 38, "step_time": 186.41656695598795 }, { "epoch": 0.4318181818181818, "grad_norm": 0.003216351615265012, "learning_rate": 1e-05, "loss": 0.022610703483223915, "step": 38 }, { "clip_ratio": 6.092197099860641e-05, "completions/clipped_ratio": 0.2578125, "completions/max_length": 4096.0, "completions/max_terminated_length": 4096.0, "completions/mean_length": 2408.974609375, "completions/mean_terminated_length": 1822.955322265625, "completions/min_length": 265.0, "completions/min_terminated_length": 265.0, "epoch": 0.4431818181818182, "frac_reward_zero_std": 0.1875, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 55082712.0, "register/okay": 0.0, "register/okay_any": 1.0, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.662109375, "reward_std": 0.4734536409378052, "rewards/correct/mean": 0.662109375, "rewards/correct/std": 0.4734536409378052, "rollout/chars": 7649.53125, "rollout/commit": 0.875, "rollout/reward": 0.75, "rollout/trunc": 0.1875, "sampling/importance_sampling_ratio/max": 2.948693037033081, "sampling/importance_sampling_ratio/mean": 0.9154539108276367, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 3.4746341705322266, "sampling/sampling_logp_difference/mean": 0.005190804600715637, "step": 39, "step_time": 208.00693321699873 }, { "epoch": 0.4431818181818182, "grad_norm": 0.0036431732587516308, "learning_rate": 1e-05, "loss": 0.03268931061029434, "step": 39 }, { "clip_ratio": 4.239076474732428e-05, "completions/clipped_ratio": 0.19140625, "completions/max_length": 4096.0, "completions/max_terminated_length": 4095.0, "completions/mean_length": 2321.060546875, "completions/mean_terminated_length": 1900.90576171875, "completions/min_length": 158.0, "completions/min_terminated_length": 158.0, "epoch": 0.45454545454545453, "frac_reward_zero_std": 0.09375, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 56331847.0, "register/okay": 0.0, "register/okay_any": 0.9375, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.7265625, "reward_std": 0.4461594223976135, "rewards/correct/mean": 0.7265625, "rewards/correct/std": 0.4461594223976135, "rollout/chars": 6050.65625, "rollout/commit": 0.90625, "rollout/reward": 0.875, "rollout/trunc": 0.09375, "sampling/importance_sampling_ratio/max": 2.9441235065460205, "sampling/importance_sampling_ratio/mean": 0.9550347328186035, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.5566263198852539, "sampling/sampling_logp_difference/mean": 0.004640746861696243, "step": 40, "step_time": 232.20252943600644 }, { "epoch": 0.45454545454545453, "grad_norm": 0.003447507508099079, "learning_rate": 1e-05, "loss": 0.040430452674627304, "step": 40 }, { "clip_ratio": 4.128463280039796e-05, "completions/clipped_ratio": 0.23046875, "completions/max_length": 4096.0, "completions/max_terminated_length": 4061.0, "completions/mean_length": 2350.501953125, "completions/mean_terminated_length": 1827.738525390625, "completions/min_length": 233.0, "completions/min_terminated_length": 233.0, "epoch": 0.4659090909090909, "frac_reward_zero_std": 0.09375, "groups/all_right": 0.5, "groups/all_wrong": 0.0, "groups/live_frac": 0.5, "groups/n": 2, "num_tokens": 57599928.0, "register/okay": 0.0, "register/okay_any": 0.6875, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.658203125, "reward_std": 0.4747757613658905, "rewards/correct/mean": 0.658203125, "rewards/correct/std": 0.4747757613658905, "rollout/chars": 6086.21875, "rollout/commit": 0.90625, "rollout/reward": 0.84375, "rollout/trunc": 0.09375, "sampling/importance_sampling_ratio/max": 2.7746007442474365, "sampling/importance_sampling_ratio/mean": 0.9619885683059692, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 1.888315200805664, "sampling/sampling_logp_difference/mean": 0.004755809437483549, "step": 41, "step_time": 234.87280439599635 }, { "epoch": 0.4659090909090909, "grad_norm": 0.003600901225581765, "learning_rate": 1e-05, "loss": 0.03945820778608322, "step": 41 }, { "clip_ratio": 4.081156771462702e-05, "completions/clipped_ratio": 0.240234375, "completions/max_length": 4096.0, "completions/max_terminated_length": 4091.0, "completions/mean_length": 2359.01171875, "completions/mean_terminated_length": 1809.7840576171875, "completions/min_length": 261.0, "completions/min_terminated_length": 261.0, "epoch": 0.4772727272727273, "frac_reward_zero_std": 0.125, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 58867902.0, "register/okay": 0.0, "register/okay_any": 0.96875, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.677734375, "reward_std": 0.46780112385749817, "rewards/correct/mean": 0.677734375, "rewards/correct/std": 0.46780112385749817, "rollout/chars": 9607.8125, "rollout/commit": 0.71875, "rollout/reward": 0.40625, "rollout/trunc": 0.375, "sampling/importance_sampling_ratio/max": 2.874417304992676, "sampling/importance_sampling_ratio/mean": 0.9376507997512817, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.5530819892883301, "sampling/sampling_logp_difference/mean": 0.004874600097537041, "step": 42, "step_time": 207.32737054100653 }, { "epoch": 0.4772727272727273, "grad_norm": 0.003691115416586399, "learning_rate": 1e-05, "loss": 0.035165682435035706, "step": 42 }, { "clip_ratio": 4.7922681119416666e-05, "completions/clipped_ratio": 0.2109375, "completions/max_length": 4096.0, "completions/max_terminated_length": 4095.0, "completions/mean_length": 2121.927734375, "completions/mean_terminated_length": 1594.2054443359375, "completions/min_length": 258.0, "completions/min_terminated_length": 258.0, "epoch": 0.48863636363636365, "frac_reward_zero_std": 0.3125, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 60021737.0, "register/okay": 0.0, "register/okay_any": 0.96875, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.751953125, "reward_std": 0.4323015511035919, "rewards/correct/mean": 0.751953125, "rewards/correct/std": 0.4323015511035919, "rollout/chars": 7986.0, "rollout/commit": 0.625, "rollout/reward": 0.59375, "rollout/trunc": 0.375, "sampling/importance_sampling_ratio/max": 2.9013590812683105, "sampling/importance_sampling_ratio/mean": 0.9236944317817688, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.5296273231506348, "sampling/sampling_logp_difference/mean": 0.004975184798240662, "step": 43, "step_time": 222.71566601800077 }, { "epoch": 0.48863636363636365, "grad_norm": 0.0032141930423676968, "learning_rate": 1e-05, "loss": 0.031207803636789322, "step": 43 }, { "clip_ratio": 4.502689955643291e-05, "completions/clipped_ratio": 0.2734375, "completions/max_length": 4096.0, "completions/max_terminated_length": 4081.0, "completions/mean_length": 2710.71484375, "completions/mean_terminated_length": 2189.370849609375, "completions/min_length": 399.0, "completions/min_terminated_length": 399.0, "epoch": 0.5, "frac_reward_zero_std": 0.09375, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 61482727.0, "register/okay": 0.0, "register/okay_any": 1.0, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.65234375, "reward_std": 0.47669193148612976, "rewards/correct/mean": 0.65234375, "rewards/correct/std": 0.47669193148612976, "rollout/chars": 13552.09375, "rollout/commit": 0.40625, "rollout/reward": 0.25, "rollout/trunc": 0.625, "sampling/importance_sampling_ratio/max": 2.9828054904937744, "sampling/importance_sampling_ratio/mean": 0.9278571009635925, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.7164409160614014, "sampling/sampling_logp_difference/mean": 0.005057974718511105, "step": 44, "step_time": 205.39486570699955 }, { "epoch": 0.5, "grad_norm": 0.0035538969095796347, "learning_rate": 1e-05, "loss": 0.04245685040950775, "step": 44 }, { "clip_ratio": 3.6159511012101575e-05, "completions/clipped_ratio": 0.208984375, "completions/max_length": 4096.0, "completions/max_terminated_length": 4075.0, "completions/mean_length": 2432.134765625, "completions/mean_terminated_length": 1992.5457763671875, "completions/min_length": 224.0, "completions/min_terminated_length": 224.0, "epoch": 0.5113636363636364, "frac_reward_zero_std": 0.15625, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 62793372.0, "register/okay": 0.0, "register/okay_any": 0.9375, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.7421875, "reward_std": 0.43785804510116577, "rewards/correct/mean": 0.7421875, "rewards/correct/std": 0.43785804510116577, "rollout/chars": 6969.3125, "rollout/commit": 0.8125, "rollout/reward": 0.8125, "rollout/trunc": 0.1875, "sampling/importance_sampling_ratio/max": 2.979311943054199, "sampling/importance_sampling_ratio/mean": 0.9486267566680908, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 1.080000877380371, "sampling/sampling_logp_difference/mean": 0.004703765735030174, "step": 45, "step_time": 213.5226564270124 }, { "epoch": 0.5113636363636364, "grad_norm": 0.004064072854816914, "learning_rate": 1e-05, "loss": 0.06307149678468704, "step": 45 }, { "clip_ratio": 6.182664350262712e-05, "completions/clipped_ratio": 0.251953125, "completions/max_length": 4096.0, "completions/max_terminated_length": 4064.0, "completions/mean_length": 2435.552734375, "completions/mean_terminated_length": 1876.289794921875, "completions/min_length": 353.0, "completions/min_terminated_length": 353.0, "epoch": 0.5227272727272727, "frac_reward_zero_std": 0.0625, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 64113431.0, "register/okay": 0.09375, "register/okay_any": 0.90625, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.6328125, "reward_std": 0.48250964283943176, "rewards/correct/mean": 0.6328125, "rewards/correct/std": 0.48250964283943176, "rollout/chars": 7549.25, "rollout/commit": 0.90625, "rollout/reward": 0.8125, "rollout/trunc": 0.15625, "sampling/importance_sampling_ratio/max": 2.989046573638916, "sampling/importance_sampling_ratio/mean": 0.8794365525245667, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 2.0649478435516357, "sampling/sampling_logp_difference/mean": 0.005066515412181616, "step": 46, "step_time": 238.6869710050014 }, { "epoch": 0.5227272727272727, "grad_norm": 0.0037526381202042103, "learning_rate": 1e-05, "loss": 0.05647933483123779, "step": 46 }, { "clip_ratio": 4.784553817671622e-05, "completions/clipped_ratio": 0.17578125, "completions/max_length": 4096.0, "completions/max_terminated_length": 4086.0, "completions/mean_length": 2084.8671875, "completions/mean_terminated_length": 1655.95263671875, "completions/min_length": 180.0, "completions/min_terminated_length": 180.0, "epoch": 0.5340909090909091, "frac_reward_zero_std": 0.21875, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 65256483.0, "register/okay": 0.0, "register/okay_any": 0.96875, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.76171875, "reward_std": 0.42644867300987244, "rewards/correct/mean": 0.76171875, "rewards/correct/std": 0.42644867300987244, "rollout/chars": 3682.75, "rollout/commit": 0.90625, "rollout/reward": 0.8125, "rollout/trunc": 0.0625, "sampling/importance_sampling_ratio/max": 2.84405779838562, "sampling/importance_sampling_ratio/mean": 0.9413059949874878, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.5454630255699158, "sampling/sampling_logp_difference/mean": 0.005082877352833748, "step": 47, "step_time": 215.25980535299823 }, { "epoch": 0.5340909090909091, "grad_norm": 0.003414996201172471, "learning_rate": 1e-05, "loss": 0.04025837779045105, "step": 47 }, { "clip_ratio": 4.7273616303300514e-05, "completions/clipped_ratio": 0.365234375, "completions/max_length": 4096.0, "completions/max_terminated_length": 4090.0, "completions/mean_length": 2880.806640625, "completions/mean_terminated_length": 2181.60302734375, "completions/min_length": 430.0, "completions/min_terminated_length": 430.0, "epoch": 0.5454545454545454, "frac_reward_zero_std": 0.15625, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 66795920.0, "register/okay": 0.0, "register/okay_any": 1.0, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.56640625, "reward_std": 0.4960552453994751, "rewards/correct/mean": 0.56640625, "rewards/correct/std": 0.4960552453994751, "rollout/chars": 7730.71875, "rollout/commit": 0.71875, "rollout/reward": 0.65625, "rollout/trunc": 0.28125, "sampling/importance_sampling_ratio/max": 2.967653751373291, "sampling/importance_sampling_ratio/mean": 0.8913018703460693, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.6522915363311768, "sampling/sampling_logp_difference/mean": 0.004959162324666977, "step": 48, "step_time": 232.4684189019972 }, { "epoch": 0.5454545454545454, "grad_norm": 0.0030148394871503115, "learning_rate": 1e-05, "loss": 0.03712388128042221, "step": 48 }, { "clip_ratio": 4.2036647414533945e-05, "completions/clipped_ratio": 0.275390625, "completions/max_length": 4096.0, "completions/max_terminated_length": 4095.0, "completions/mean_length": 2501.763671875, "completions/mean_terminated_length": 1895.867919921875, "completions/min_length": 327.0, "completions/min_terminated_length": 327.0, "epoch": 0.5568181818181818, "frac_reward_zero_std": 0.1875, "groups/all_right": 0.5, "groups/all_wrong": 0.0, "groups/live_frac": 0.5, "groups/n": 2, "num_tokens": 68155639.0, "register/okay": 0.0625, "register/okay_any": 0.9375, "register/premium": -0.06451612903225812, "register/verify_any": 0.96875, "reward": 0.65234375, "reward_std": 0.47669193148612976, "rewards/correct/mean": 0.65234375, "rewards/correct/std": 0.47669193148612976, "rollout/chars": 4332.3125, "rollout/commit": 1.0, "rollout/reward": 0.9375, "rollout/trunc": 0.0, "sampling/importance_sampling_ratio/max": 2.96474289894104, "sampling/importance_sampling_ratio/mean": 0.9402225017547607, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 1.1063487529754639, "sampling/sampling_logp_difference/mean": 0.004922760184854269, "step": 49, "step_time": 242.5428965280007 }, { "epoch": 0.5568181818181818, "grad_norm": 0.003150244476273656, "learning_rate": 1e-05, "loss": 0.02683938853442669, "step": 49 }, { "clip_ratio": 5.262581521492393e-05, "completions/clipped_ratio": 0.181640625, "completions/max_length": 4096.0, "completions/max_terminated_length": 4083.0, "completions/mean_length": 2219.458984375, "completions/mean_terminated_length": 1802.9571533203125, "completions/min_length": 202.0, "completions/min_terminated_length": 202.0, "epoch": 0.5681818181818182, "frac_reward_zero_std": 0.125, "groups/all_right": 0.0, "groups/all_wrong": 0.0, "groups/live_frac": 1.0, "groups/n": 2, "num_tokens": 69365058.0, "register/okay": 0.0, "register/okay_any": 0.96875, "register/premium": NaN, "register/verify_any": 1.0, "reward": 0.716796875, "reward_std": 0.4509948492050171, "rewards/correct/mean": 0.716796875, "rewards/correct/std": 0.4509948492050171, "rollout/chars": 7746.8125, "rollout/commit": 0.65625, "rollout/reward": 0.46875, "rollout/trunc": 0.375, "sampling/importance_sampling_ratio/max": 2.8404057025909424, "sampling/importance_sampling_ratio/mean": 0.9204064607620239, "sampling/importance_sampling_ratio/min": 0.0, "sampling/sampling_logp_difference/max": 0.4059598445892334, "sampling/sampling_logp_difference/mean": 0.004588037263602018, "step": 50, "step_time": 194.41973535799843 }, { "epoch": 0.5681818181818182, "grad_norm": 0.003299945965409279, "learning_rate": 1e-05, "loss": 0.02735048718750477, "step": 50 } ], "logging_steps": 1, "max_steps": 300, "num_input_tokens_seen": 69365058, "num_train_epochs": 4, "save_steps": 25, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": true, "should_training_stop": false }, "attributes": {} } }, "total_flos": 0.0, "train_batch_size": 2, "trial_name": null, "trial_params": null }